Langmuir 1994,10, 267-275
267
Structure of Polymerizable Lipid Bilayers. 6. Bilayer Structure of Three Polymerizable Diacetylenic Glutamate Lipids David G . Rhodes,**+ David A. Frankel,$ Thauming Kuo,~and David F. O'Brient Pharmaceutics Division, College of Pharmacy, University of Texas at Austin, Austin, Texas 78712-1074, and C. S.Marvel Laboratories, Department of Chemistry, University of Arizona, Tucson, Arizona 85721 Received June 11, 1993. In Final Form: October 5, 199P The profile structurehas been determined for bilayers of three highly polymerizablediacetylenic glutamate lipids (DGL),bis(docosa-10,12-diyyl)N - [6-(triethyla"onio)hexanoyll-~-glutamate bromide (DGL-11, N- [11-(trimethyla"onino)undecanoyll-~-glutamatebromide (DGL-21, and N - [4-(trimethyla"onio)butoxybenzoyl] glutamate bromide (DGL-3). Bilayers of each lipid are similar in structure, with lamellar profile and equatorial reflections that are consistent with a tilted lamellar phase with MIIpacking. The structures are distinct from previously reported structures of diacetylenic glycerophosphocholines,since the acyl chains do not appear to be offset in the bilayer normal direction and are not as crystalline. Langmuir compression isotherms of DGL-1 and DGL-2 monolayers on water show no clear transitions, whereas those of DGL-3 have several transitions. All three lipids are readily polymerizable as Langmuir films. Model structures are proposed for the DGL bilayers which distinguish them from the previously reported diacetylenic glycerophosphocholine bilayer structures.
INTRODUCTION Diacetylenic lipids have been the focus of considerable interest due primarily to their ability to form robust supramolecular assemblies and to the morphologies of these assemblies. Recent reports1t2on the properties of a diacetylenic glutamate lipid (DGL), bis(docosa-10,12diynyl) N-[6-(triethylammonio)hexanoylanoyll glutamate bromide (DGL-11, indicated that multilayer cast films of this material photopolymerize rapidly and completely upon exposure to 254-nm light. For example, 180 s of UV exposure (254 nm) produced a film which was at least -75% polymerized and stable in the presence of chloroform, chlorobenzene, THF, DMSO, or DMF. The dried films were deep purple at 22 OC and exhibited a reversible thermochromic effect, turning red-orange at 50 "C. The presence of long wavelength absorption indicated the presence of long-chain conjugated p~lydiacetylene.~ Similar observations were reported for homologues of DGL-2 and DGL-3 by Kuo and O'Brien.4 CHfiHfiO-O-(CH2)(rC
I
EC-C EC-(CH2)&H3
R - C O - N H - C H - C ~ C H ~ ) Q ~ E=C-(CH2)84H3 C-C
R = (Cl+-C&)rN+-(CH2)5(CH3)3-N+-(CHdio-
DOL-1
DOL-2
( C H ~ ) ~ - N + - ( C H Z ) ~ O - C ~ ~ - DOL-3
This behavior was in marked contrast to that of the diacetylenic glycerophosphocholines (DAPC). These PC's usually polymerize to a lesser extent than the DGL systems
* Author to whom inquires should be addressed. t University of Texas at Austin. Diffraction work was performed when D.G.R. was at the Biomolecular Structure Analysis Center, at the University of Connecticut Health Center,Farmington, CT 060302017. University of Arizona. Abstract published in Advance ACS Abstracts, December 15, 1993. (1) Kuo, T.; OBrien, D. F. J. Am. Chem. Soc. 1988,110, 7571. (2) Kuo, T.; OBrien, D. F. Macromolecules 1989, 23, 3225. (3) Tieke,B.; Liester,G.;Wegner, G. J. Polym. Sci., Polym. Chem. Ed. 1979,17,1631. (4) Kuo, T.; OBrien, D. F. Langmuir 1991, 7, 584. @
and are not resistant to organic solvent or temperature. Most pure DAPC's (e.g. DC8,gPC) turn red upon exposure to 254-nm ultraviolet light, although some DAPC's with short proximal acyl chains, D C ~ J O P Cand , ~ mixtures of long chain DAPC and saturated short chain glycerophosphocholine (PC), e.g. DCs,gPC/DNPC,Gturn blue. Those which do turn blue turn irreversibly red upon heating to temperatures higher than the main chain melting temperature, Tm,or upon exposure to organic solvent. Solvent exposure, e.g. CHCl3,usually disrupts solid films of DAPC, regardless of the initial color, although some of the polymeric species are only marginally soluble. The DGL's examined here all have a common acyl chain structure, comparable to the extensively studied DCs,gPC, but with an odd number of methylene groups in the proximal acyl chain. The glutamate backbone structure provides a chiral center, as does a glycerol backbone, but imparts more flexibility than does the glycerol backbone. The amide portion of the headgroup has the potential to form hydrogen bonds with adjacent lipid molecules. The cationic quaternary ammonium headgroup termini might be expected to interact strongly with a facing cationic surface. Finally, the conformation of the hydrophobic or,butoxyphenyl) between the amine "connector" (C5,C ~ O and the amide is expected to adapt to the spacing imposed by the acyl chains and might even provide an additional hydrophobic domain. The possibility of 7~ orbital interaction between phenyl groups in DGL-3 must also be considered. Bilayers of each of these DGL's are easily polymerizable at room temperature and form blue polymer, which indicates that the assembly is a highly ordered structure that favors the formation of very long polymer chains.4 A feature which distinguishes polymerized DGL films from those made of other diacetylenic lipids is their resistance to organic solvent. This suggests that poly-DGL films are covalently cross-linked, but no mechanism or structural data to explain that cross-linking has been previously (5) Rhodes, D. G.; Xu, 2.;Bittman, R. Biochim. Biophys. Acta 1992, 1128,93. (6) Rhodes, D. G.; Singh, A. Chem. Phys. Lipids 1991,59, 215.
0743-746319412410-0267$04.50/0 0 1994 American Chemical Society
268 Langmuir, Vol. 10, No. 1, 1994 proposed. The polymerizability of diacetylenes provides limits to the possible conformations. Diacetylene polymerization requires that the molecules be aligned in a highly ordered solid-state lattice for successful polymerization to begin but must have sufficient flexibility to allow some conformational shift to occur. Suitable geometries have been established with model diacetylenic compounds, for which crystal structures have been determined in the monomeric and polymeric states. For polymer to form, the monomer alignment must be such that the inter-diacetylene distance, d l , is 4.4-5.4 A and the angle of the linear diacetylene 45-55' with respect to the polymerization axis.7 The limits to dl arise from limits to the displacement upon polymerization. The measured dl in polydiacetylenes is 4.91 A. This report describes diffraction studies to characterize the bilayer structure of the poly-DGL lipid assemblies, in order to understand the basis of their robust behavior. The results of these diffraction experiments will be compared to those of previous studies of DAPC's. MATERIALS AND METHODS Materials. All solvents were HPLC grade (Aldrich) and were used without further purification. Water was purified with a deionization/ultrafiltrationsystem (Vanguard RGW-5). Lipids, synthesized as described previ0usly,~~4 were stored in chloroform or dichloromethane at -20 "C.Liposome preparations were made by shell drying the lipid from solvent with a stream of nitrogen and then subjecting the sample to mechanical vacuum overnight. The lipid film was rehydrated by adding water to produce a final concentration of 4 mg/mL at 50 "C, allowing the sample to hydrate, and vortexing gently. Multibilayers. Multibilayers were formed as described previou~ly,~~~~8~9 using a variation of the method of Clark et al.1° . Aliquots of concentrated liposome suspension containing 200 pg of lipid were added to sedimentation cells designed to fit in a swinging bucket of a Beckman SW-28 rotor. The caps of the buckets have been modified to include a 100-pm pinhole. With the rotor spinning at 1000 rpm, the centrifuge vacuum was applied, removing the water through the pinhole. Within 1 h, most of the water had evaporated, leaving a partially hydrated multibilayer stack on a thin aluminum foil. Multibilayers on foil were polymerized in the cold at - 4 "C, -95 % relative humidity. Unpolymerized multibilayer samples were placed in a shallow plastic container, the edges of which supported the UV light at - 5 mm from the samples. Nitrogen was flowed through the container. The samples were then exposed to 254-nm light (UVG-11, rated at 1.8 X 10lSphotons/(s/cm2)at 76 mm) for 30 s to effect complete polymerization. Samples were then mounted on curved glass and placed in sealed brass canisters with a plastic cup of saturated salt solution to regulate relative humidity (98% to 66 % 1. Most diffraction data reported here were obtained at 15 "C, so canisters were placed at this temperature and allowed to equilibrate for at least 18 h prior to data collection. Following any change of conditions (relative humidity, temperature), samples were reequilibrated for 18h. Mounted samples were not removed from (7) Baughman, R. H.; Chance, R. R. Ann. N.Y. Acad. Sci. 1978,313,
Rhodes et al.
the canisters during the course of the experiment in order to improve mechanical reproducibility. Alternatively, lipid in solvent could be slowly dripped onto a substrate, allowing the solvent to evaporate between drops. The resulting film was then polymerized as before and equilibrated for diffraction. This procedure could be used to deposit the sample onto a foil substrate or directly onto the curved glass. Diffraction. Diffraction data were collected as described previo~sly8*~ using a GX-18 rotating anode microfocus generator (Marconi Avionics) and a fixed geometry beamline with a Franks mirror providing line focus at the detection plane. A Ni filter (0.025 mm) was used with a Cu target to provide Ka X-rays (A = 1.54 A). A one-dimensional quartz wire electronic detector (Braun, Innovative Technology, Inc.) was used for collecting lamellar scattering data and Kodak DEF-5 film was used with a (mirror/mirror) point focus beamline to obtain equatorial data. Diffraction data were analyzed as described previously586p819by correcting the raw lamellar intensity data using a two exponential background and then integrating the observed reflections. Unresolved peaks were integrated by fitting the profile with a sum of Gaussian profiles representing the individual peaks. Film recording of data in two dimensions data was used to identify nonlamellar peaks. These data were not integrated but were used to determine the spacing in the in-plane direction ( x - y ) . Spacings from films were measured using an Enraf Nonius film reader. Phase determination used the swelling method,ll in which the lamellar repeat was varied by equilibrating samples to a series of different relative humidities. Lamellar intensity data from these samples were Lorentz corrected and plotted as structure factor (F(z*)vs h/d (= 2 sin e/A) for a series of data sets for equivalent samples obtained at different relative humidities. No other correction factors were applied to the intensity data, but the accuracy of these data was adequate for the analysis performed here. The continuous structure factor function was then estimated by interpolation, and relative phases . ~ ~electron ~~~~~ were determined as described p r e v i ~ u s l y The density profilewas calculated using the intensity and phase data to calculate a one-dimensional Fourier transform. The phase assignments were verified by a Fourier reconstruction, in which the calculated electron density profiles were subjected to a second Fourier transform, resulting in aseries of&*). If these functions were similar, the phases were judged to be correct. Modeling. Model structures were designed using CHEMX software (developed and distributed by Chemical Design Ltd., Oxford, U.K.) which fit the electron density profile, positioning moieties of high electron density (Br, 0,N, C=C-C=C) at peaks and moieties of low electron density (chain termini) at minima. The models were additionally constrained by requiring that the diacetylene moieties in the model fit a diacetylenic polymer backbone constructed using published crystal structures.12J3 Finally, the model was required to include only energetically allowableconformations throughout and to be compatible with three-dimensional packing to propagate a bilayer structure without prohibitive steric interactions. An electron summing method9J4 was used to calculate an
706.
(8)Rhodes, D. G.; Blechner, S.; Yager, P.; Schoen, P. Chem. Phys. Lipids 1988,49, 39. (9) Blechner, S . L.; Morris, W.; Schoen, P. E.; Yager, P.; Singh, A,; Rhodes, D.G. Chem. Phys. Lipids 1991,58, 41. (10) Clark,N.; Rothschild, K.; Luippold, D.; Simon, B. Biophys. J. 1980, 20, 246.
(11)Franks, N.J.Mol. Biol. 1976, 100, 345. (12) Enkelman, V. In Polydiocetylenes; Cantow, H . , Ed.; Springer-Verlag: Berlin, 1984; p 91, (13) Gross,H.;Sixl,H.;Kr6hnke,C.;Enkelmann,V. Chem. Phys. 1980, 45, 15. (14) Blechner, S. L. Thesis, University of Connecticut 1990.
Structure of Polymeriza bZe Lipid Bilayers
Langmuir, VoZ. 10, No. 1, 1994 269
electron density profile for this model for comparison with eperimental data. Thin Films. Compression isotherms were obtained using a KSV 5000 system in a class 100 clean room at 20 “C. The trough surface was cleaned by repeated “blank” compression and aspiration until a 10:1 compression produced 50 “C, but this thermochromic transition was fully reversible, as reported by Kuo and O’Brien.4 No attempt was made to quantify the spectral characteristics. The sample color was sensitive to exposure to certain organic solvents (e.g. chloroform). The solvent-induced transition was fully reversible; blue color returned upon evaporation of the solvent. There was no apparent loss of material from the multibilayers upon extended exposure to chloroform, so the polymerization under these conditions is likely to be as complete as reported previously for somewhat different polymerization condition^.^^^ The intensity and definition of diffraction from multibilayers of DGL’s were not as strong as might be expected based on the ordering apparent from the polymerization. In particular, it was surprising to observe the moderately large mosaic spread in DGL lipid samples. As demonstrated in Figure 1,diffraction patterns were rich in detail, including not only the lamellar reflections which dominate most multibilayer patterns but also equatorial reflections at two different spacings and some off-axis reflections. Qualitatively,these patterns were similar to those obtained from DAPC m~ltibilayers.6~8~~ Experiments with DGL-1were used to look for structural changes induced by perturbants that would be disruptive to multibilayers of conventional lipids (Figure 2). The DGL 1sample diffracted quite well, even followingrinsing with chloroform (Figure ab). Moreover, as shown in Figure 2c, when this solvent-rinsed sample was heated to 57 “C, at least 8 orders are visible. Because the thermal phase transition temperature is 18.5 “C in the fully hydrated state: and multibilayer samples were red at 57 “C it is likely that the bilayers are in a liquid crystalline configuration at this temperature and humidity. However, DSC data have not been obtained for partially dehydrated samples, so the phase of multibilayers at 57 “C is not absolutely certain. Data from one dimensional wire detectors were used to determine the lamellar repeat distance. Figure 3 shows typical detector patterns for DGL-1, DGL-2, and DGL-3. Film data (Figure 1)were used to verify these results and to determine repeat spacings for equatorial and off-axis
-
Figure 1. ,X-ray diffraction patterns from multibilayers of (a, top) DGL-1 (98% relative humidity), (b, middle) DGL-2 (72% relative humidity), and (c, bottom) DGL-3 (84% relative humidity). All were obtained using a point focus beamline with samples maintained at 15 “C.The speckled patterns are scatter from the aluminum substrate and cannister windows.
reflections. The lists of lamellar and equatorial repeat distancesfor DGL-1, DGL-2, and DGL-3are given in Table 1. The lamellar repeat distances are relatively large, but not as large as one would expect for all-trans chains normal to the bilayer plane. In all cases, it is necessaryto postulate some mechanism (tilting or disordering) for “shortening” the acyl chain length. Possible model structures are described below. In some data sets the scattered reflections along the lamellar axis do not all index as a single simple lattice. Although there is a lattice repeat, there are other reflections present that initially made interpretation and integration of the lamellar data difficult. Film exposures revealed that these reflections arose from a ring at -9.7 A-l, reflections at -6 A-l at 45” from the lamellar axis, and equatorial reflections at -4.8 and 4.3 A-l, some of which cross the lamellar axis due to mosaic disorder in the sample. In some cases, film data indicated that off-axis reflections were at the same s as the seventh or ninth order lamellar reflection. For this reason, many data sets could not be analyzed in detail for h 1 7. The intensity for each reflection was determined by integration, and phases were determined by a swelling analysis. Background intensity resulting from substrate
Rhodes et al.
270 Langmuir, Vol. 10,No. 1, 1994 ”- wt%7w5-m
I1
-4
I
I
a 9
5 6 7 8
2 3
C
0
250
750 28 (channels)
lo00 ___
Figure 3. - Detector patterns for (a) DGL-1, (b) DGL-2, and (c) DGL-3. The data are presented as relative intensity in terms of X-ray counts, as a function of position along the detector wire in channels. The calibration from channels to millimeters is determined by exposuresto an attenuated beam while displacing the detector by fixed increments. Conversion to 2 8 follows from knowing the sample-to-detector distance. Table 1. Repeat Spacings for Diacetylenic Glutamate Lipids. lamellar equatorial lipid headgroup (Br) spacing (A) spacing(s)(A) DGL-1 EtN+(CH2)&ONHDGL-2 MesN+(CH2)&ONHDGL-3 MesN+(CH2)4OCeH4CONH-
Figure 2. The DGL multibilayers retained significant order even under conditions that would denature conventional bilayer systems. (a, top) This diffraction pattern, from a DGL-1 sample at 15 “C, 98% relative humidity, is comparable to Figure la. (b, middle) This pattern is from the same sample following CHCl3 rinsing (15 “C, 84% relative humidity). Washing the specimen with CHCl3had almost no effect on the DGL diffraction patterns. The apparent increase in order may be due to lower humidity. (c, bottom) Heating the solvent-washedDGL-1 sample to 57 “C (98% relative humidity) turned it red, indicating disorder, but many lamellar reflections, as well as the equatorial reflections, are still visible.
scatter, beam spreading,and incoherent sample scattering was subtracted by fitting a multiple exponential background. For well-separated peaks, areas were determined by simple addition of all counts between baseline points. In some cases, where nonlamellar peaks overlapped lamellar peaks, detector data were empirically fitted as a sum of Gaussian profiles. Intensity data (I) for all peaks were plotted as normalized V2against correspondingh/d. This process results in plots of the structure factor (F) as a function of s (= 2 sin e/X = h / d ) , in which discrete data points are used to infer a continuous function. Figure 4 shows structure factor plots for DGL-1, DGL-2, and DGL3.
64 68 80
9.6,4.a, 4 . 4 , ~ 11.9,8.5,4.4,4.1,3.8 9.8,4.6,4.4*
a Data for lamellar spacings are typical spacings at high humidity and were best obtained from electronic detector data. Equatorial spacingswere relatively independent of humidity and were obtained from film data.
Although the relative phases are clear in most places, ambiguity remains at some nodes. Based on previous results with DAPC’s or PC’s and the shape of resulting real-space transforms, structure factor functions comparable to a sinc function are more likely than those which “arbitrarily” oscillate above and below the abscissca. Nevertheless, the DGL lipids are not typical. Although other reports have described empirical fitting (essentially based on the expected form of the electron density profile), the present systems are not phospholipids, and the “intuitive” approaches are not valid. It is expected that the Br counterion will contribute a relatively strong peak to the profile, but the extent to which the Br is constrained in a region of z is not clear. Thus, ambiguities of phasing had to be resolved without prejudice. The details of this procedure follow for each case, and the final assignments were verified by a Fourier reconstruction p r o c e d ~ r e . ~ ~ ~ Figure 4a is a reduced data set from a swelling series (relativehumidity = 98-13 % ) of two multibilayer samples
Structure of Polymerizable Lipid Bilayers
Langmuir, Vol. 10, No. 1, 1994 271
1
U
0.1
0
s (k’)
0.2
i i
1
3 i
11
-l
r
I
a
profiles from all data sets were evaluated for consistency, i.e. profile structures at 91% and at 72% relative humidity were required to be similar. Spacingsof significantfeatures of electron density data seta were examined to determine the region of the profile corresponding to the water space. One would expect that the profile should remain invariant with humidity except in this region, and that the extent of this region should decrease with dehydration. Gradually, then, higher orders were included, using h I9 and finally complete data sets (see below). Similar criteria were used, with the additional constraint that the proposed electron density profile be consistent with previous profiles for the same data set; i.e. was the profile calculated from low orders a low resolution equivalent of the profile calculated with all orders? Based on these criteria, the function traced in Figure 4a was selected. Because of interference with the lamellar pattern by nonlamellar reflections, some data at seventh or ninth order could not be used. Films obtained on the line-focus beamline following detector runs were used to determine whether any intensity detected along the lamellar axis resulted from nonlamellar scattering. Most data sets revealed potential ambiguities with ninth order data, so no attempt was made to obtain a tight fit for these data points (Figure 4a). To estimate complete fits, empirical estimates were made to approximate the contribution of the ninth order to the overall electron density profile. Although small details of the profile changed (ripples or exaggerated peaks),the overallform of the electron density profile was preserved over a wide range (0.8X - 0.3X) of scale factors. Figure 4b shows a complete reduced DGL-2 data set (including both positive and negative structure factor possibilities) for two samples from a single preparation, including data obtained from samples at relative humidities from 98 % to 13 %. The lamellar repeat varied from 87 to BOA, based on at least five and as many as nine lamellar reflections . Nodes were observed at the fourth order (s 0.05 A-l-1) and sixth order (s c 0.07 A-1). Like phases were apparent for first and second orders and for nonzero sixth and seventh orders. In addition to the phase combination illustrated (-, -, +,0, -O/+, +,-, +),alternative possibilities involving nodes at s = 0.028 (-, -, -, 0, +, ...) and/or without nodes at s > 0.08 (..., O/+, +, +, + and ..., 01-, -, -, -) or s > 0.10 (..., O/+, +, -, -) were considered. Data were evaluated in the same manner as used for DGL1, first using h I5, then h I7, and finally h I9. This analysis indicated that the function shown in Figure 4b is correct. The data for DGL-3 (Figure 4c) represent samples at relative humidities ranging from 98% to 13%. The lamellar repeat varied from 70.4 to 60.2 A, and h,, varied from 6 to 13, so a greater range of reciprocalspace is covered by the sampled data than in the case of DGL-2. True nodes appear at s 0.085,0.115, and 0.13 A-l and nodes can easily be inferred at s = 0.035,0.05 A-l. Like signs can be assumed for orders 1 and 2 and for orders 4 and 5. Ambiguity arises at h > 6 and was approached in a manner analogous to that described above. Profile structures for h I6 were calculated initially. Comparison of these profiles to those calculated to higher resolution allowed decisions to be made as to whether nodes at s c 0.0115 and s e 0.013 resulted from the structure factor function approaching or crossing the abcissa. The fit illustrated results in an electron density profile that is relatively invariant for the bilayer interior and which accounts for changes in the lamellar repeat by changes in the hydrated region of the unit cell.
I 0
0.1
s (A1)
o.2
Figure 4. Structurefactor of (a, top) DGL-1, (b, middle) DGL2, and (c, bottom) DGL-3as a function of s. Discrete data points (+/-) are plotted and the interpolated fit to these data is shown for one choice of sign. The relative phases are determined from these plots, and the overall choice of sign is determined by inspection of the resulting real-space transform.
of DGL-1. Because phase information is lost in the diffraction process, both possibilities (+ or -, assuming a centrosymmetric structure) are plotted. Typically, 12 or more lamellar reflections were observed, with well-defined nodes for s e 0.045A-1(h = 3), s e 0.075-0.095 A-l (h = 5, 6), and s = 0.155-0.177 A-l (h = 10, 11). Because the extent of swelling was limited, the data did not sample a large region of s. This made phasing more difficult, but no orders changed sign during swelling, and the same phase combination could be used for all data sets. The phase combination illustrated (-, -, +IO, -, +IO, O/+, -, +IO, +, 0,O/+, +, +),was determined by first examining data for h I4, for which the phases are obvious, to determine a “low resolution” estimate of the profile structure. Data for h 5 7 were included and resulting electron density
Rhodes et al.
272 Langmuir, Vol. 10,No.1, 1994 n [mN/ml 1 1
40
DGL-1
(C4)
DGL-2
(C10)
-
-M
Figure 5. Electron density profiles of (a) DGL-1, (b) DGL-2, and (c) DGL-3 at 15 "C.These electron density scale of these profiles is relative and independent of the others. Because of uncertainties in the intensity of higher orders due to interference from broadly arced equatorialand off-axisreflections,the profiles are calculated with only eight orders.
li
I/
-
\
Calculated electron density profiles based on these phases are shown in Figure 5. Profiles of DGL-1 and DGL-2 show electron density minima at the bilayer center and maxima near the unit cell edge, presumably due to the electron-dense Br counterion. In addition, peaks were observed at the interior and at f25 A for both systems. These maxima appear to correspond to the diacetylene and backbone regions, respectively (see below). The headgroup peaks in the electron density profile of DGL-3 bilayers were shifted. Peaks at -f30 A,near the unit cell edge, were displaced - 5 A from the corresponding peaks in profiles of DGL-1 or DGL-3. The presumed backbone peak appeared at f22 A and the diacetylene peak at f12
A.
Equatorial reflections at -4-5 A from lipid systems arise from chain-packing repeat distances. As these represent interchain spacing, a parameter that changes as the lipid phase changes, they are often used as indicators of phase state and morphology of liquid crystallinephases. In film patterns collected from DGL systems, reflections nearly always appear at -4.4 A (Table 1). Other reflections have been observed at 4.8 and 3.6 A (DGL-11, 4.1 and 3.8 A (DGL-P), or 4.6 A (DGL-3). Larger equatorial spacings are less common. Reports from this lab0ratory5~6~~~~ and others15 have described reflections at a -10 A repeat in DAPC systems. These appear to correspond to interheadgroup spacings. Similar reflections at 10-11 A appear in the present system (Table 1). Langmuir compression isotherms can disclose features of intermolecular packing in the bilayer (or monolayer) plane. Compression isotherms of DGL-1 and DGL-3 (Figure 6a,c) were quite similar in overall sha e, with a sudden onset of surface pressure at 150-170 and an apparent collapse at 40 mN/m (-50 A2). Isotherms from DGL-2 (Figure 6b) were more complex, with transitions at 27, 37,and 45 mN/m, before collapse at 50 mN/m at slightly smaller molecular area. The molecular areas of each system in the condensed phase are close to those expected from the equatorial diffraction data and a packing model to be discussed below. Alignment and crystallization of the extended acyl chain in the DGL-2 headgroup may be responsible for transitions in the DGL-2 isotherms
l2
(15) Caffrey, M.; Hogan, J.; Rudolph, A. Biochemistry 1991,30,2134.
DGL-3 n ImN/ml
(C4-Phenoxy)
-
Figure 6. Compression isothermsof (a,top) DGL-1, (b,middle) DGL-2, and (c, bottom) DGL-3. All were obtained at 20 O C on HzO, as described in the methods. and the elevated T (compared to that of DGL-1 or DGL-3) in the expanded phase (see Discussion). DGL-1, DGL-2, or DGL-3 monolayers could be polymerized on the surface by exposure to W light. A UVG-11 lamp mounted at a distance of 1cm above the film was used to illuminate (215 min) a film compressed to 35 dyn/ cm. The surface pressure decreased during illumination. A dark haze appeared on the surface and a dark blue line appeared on the Wilhelmy plate, suggesting a polymerization-associated color change. Blodgett deposition of these polymerized films onto clean glass resulted in visible
-
Structure of Polymerizable Lipid Bilayers
purple/magenta tinting. The formation of purple polydiacetylene, rather than blue, suggests that the polymerization may not have been as complete or as well ordered as in the solvent-deposited thick films or centrifugedeposited multibilayers. Hysteresis curves, not shown, revealvery little difference between the compression and decompression traces. The reversible interaction between adjacent lipids in the monolayer doesnot preclude specificfavorableinteractions in the lateral association but does suggest low energies of association. That is, the lack of hysteresis suggests that the lipids on the surface do not spontaneously aggregate or crystallize as one observes for very long chain saturated PC's. The hysteresis curves also indicate that the plateau at 40 mN/m is a collapse rather than a transition.
Langmuir, Vol. 10, No. 1, 1994 273
U .30
-10
-20
0
20
10
30
DISCUSSION The DGL lipids are particularly interesting because of their facile, complete polymerization and the robust nature of the resulting polymerized films. Polymerization behavior has been studied for many DAPC's and most of these systemsexhibit only partial polymerization, usually to the less well ordered red form. Other diacetylenic lipid systems (e.g. DC2,#C6 or DC&'C/DNPC mixtures6)have been shown to rapidly react under UV light to yield the blue polymerized form. In both of these cases, in contrast to the DGL lipids, the color turns irreversibly from blue to red upon heating to T > Tmor brief exposure to solvent. It has been suggested616 that packing constraints in the highly ordered bilayer structure of DCmpPC's restrict the ability of these molecules to accommodate the conformational changes necessary for polydiacetyleneformation. The data presented here are consistent with that hypothesisand, furthermore, suggest that the solvent and melting resistance of the polymerized DGL lipid bilayers is due to a unique packing motif. In order to avoid uncertainty due to off-lamellar reflections (see above) the electron density profile for DGL-1 illustrated here is for h I7 only. Electron density maxima in the bilayer interior ( i 1 5 A) and near the edge (f25 A) probably correspond to diacetylene and the densely packed backbone region (which also contains 0 and N atoms), respectively. The maxima at the unit cell edge are probably due to the Br counterion. The model structure suggested in Figure 7a, while not ideal, is a reasonable beginning. In particular, it is not obvious why the electron density minimum in the bilayer center is so broad. Model fitting to account for the diacetylene and backbone peaks does not easily acount for the breadth of this minimum. The profile structure of DGL-2 could not be obtained at high resolution, due in part to the extraordinarily large lamellar repeat spacing. With h I7, the profiles at all relative humidity show electron density peaks at f12 and f 2 5 A, corresponding to the diacetylene and backbone regions. In addition,an electron density maximum appears at the edge of the unit cell. In some cases (especially at high humidity) this appears to be a broad, sometimes doubled maximum, but most often is a single, well-defined peak. A good model structure proposed from these data involves parallel, tilted ( 3 0 O ) acyl chains. The fit for the diacetylene moiety, to the electron density maximum at f15 A, and the tightly packed backbone region, to the maximum at f26 A, is straightforward (Figure 7b). Although an exact conformation cannot be deduced from the one-dimensional data obtained here, this tilt would
40
-30
0
-20
-20
-10
0
20
10
40
20
30
Figure 7. Approximate model structures of (a, top) DGL-1, (b, middle) DGL-2, and (c, bottom) DGL-3 compared to the corresponding electron density profiles at 15 "C.Although the exact conformations cannot be determined from data at this resolution, these profiles are reasonable approximations, consistent with all available data. The inset in (c) shows a possible interheadgroup conformation (see Discussion).
result in the tilt for the diacetylene which is required for polymer formation. The headgroup conformation cannot be determined either. Figure 7b illustrates two possible conformations, either of which would be consistent with the profile structure data. A disordered headgroupmight help to resolve the question of how the molecules may pack tightly with two chains in the bilayer interior but only one outside of the backbone. A disordered (CH2)lo chain would have a large overall cross section and might account for the relative disorder in the lamellar stacking of this derivative. The profile structure of DGL-3 did not include electron density maxima at fdI2. This suggests that the Br may be deeper in the headgroup structure than in the case of DGL-1 or DGL-2 or, as indicated in the proposed model structure, that the headgroups may interpenetrate, with phenyl groups associated with Br/NMe3+ from opposing layers. This would help to explain the small degree of swelling observed in this system. (Note the limited s-spread in the data of Figure 4c.)
Rhodes et al.
274 Langmuir, Vol. 10,No. 1, 1994
Figure 8. Assignment of the nonlamellar reflections. The spacings (see Table 1)are derived from film measurements. The tilt angle of the diacetylene is based on known crystal structures for other, smaller, diacetylenes which polymerize well and form three-dimensional crystals. The inset illustrates that the 3.8 8, lattice will typically consist of about five repeats and will exist throughout the multibilayer.
a
b
Figure 9. A proposed packing model for DGL lipids (a) and DAPC lipids (b). The view is along the bilayer normal. Acyl chains are parallel and indicated by circles. Individual lipid molecules are represented by rectangles. The dashed lines represent polymer chains.
The bilayer structures of these three DGL’s are quite similar. The electron density profiles appear to be somewhat unusual, as most published data is from phosphate-containing lipids, for which there is a clearly defined electron density maximum associated with the P and 0 atoms of the phosphocholine. In the case of the DGL lipids, the profile structure is less intuitive. One expects an electron density maximum associated with the Br counterion, but the degree of disorder in this region apparently dissipates the peak. Other maxima are expected to arise from the backbone, as this portion of the molecule contains four 0 atoms and this region of the bilayer tends to be tightly packed. Model studies (rep6and unpublished observations) suggest that even for bilayers of phosphate-containing lipids, a significant contribution to the so-called Itheadgroup peak” of electron density comes from the backbone region of the molecule. Gaussian fitting of DPPC electron density (16) Weiner, M.;
Suter, R. M.; Nagle, J. F. Biophys. J. 1989,55,315.
profiles16showed that the best fit for models in which the headgroup was modeled by two Gaussians resulted from a large Gaussian at smaller z (corresponding to the backbone) and a smaller Gaussian superimposed at slightly larger z (corresponding to the phosphate group). Thus, the extent to which these profiles resemble phospholipid bilayer profiles should not be surprising. Maxima from the diacetylene moiety should have somewhat higher electron density than CHz and should be tightly packed. The maxima ascribed to the diacetylene in the profile are similar in relative magnitude to that reported for the diacetylene in DAPC’S.~,~ The Langmuir compression isotherms of DGL-1 and DGL-3 are relatively featureless, and suggest low energy differences between available conformations. In compression isotherms of simple molecules like fatty acids, conformational variations are limited to little more than trans/gauche isomerization, and the isotherms appear quite similar in form to those of DGL-1 and DGL-3. In more complex structures, features appear in the compression which have been attributed to higher order structural events such as headgroup ordering.17 The relatively short, bulky headgroups of DGL-1 and DGL-3 are unlikely to have much conformationalfreedom, which is consistent with a simple compression isotherm. However, the headgroup of DGL-2 is long and can assume several conformations which could correspond to the observed transitions. No attempt will be made at this time to assign specific structures to these phases. In the case of all three lipids, the collapse area is close to that expected from the equatorial reflections in the diffractionexperiments (seebelow). The fact that all three compounds may be polymerized as Langmuir monolayers suggests a packing arrangement similar to that in the multibilayer. This is supported by the observation that the blue line deposited on the Wilhelmy plate during surface polymerization was extremely difficult to remove, even with solvents such as methanol, ethanol, chloroform, or acetone. Based on the reflections observed in the 3-5 A range, the most likely chain packing lattice is Milor Figure 8 shows a possible packing motif for the diacetylenemodified acyl chains which accounts for the 4.4-4.5 A reflections (polymerizationdirection) and for the 10-11 A reflections (in a roughly orthogonal direction) as primary in-plane lattice repeats. The smaller (3.8-4.0A repeat could corrrespond to an interchain packing distance in the chain-normaldirection. With chains eight to nine CH2 units long at this tilt angle, -five repeats at this distance would appear throughout the bilayer (see inset, Figure 7). By the same argument, a very small spacing corresponding to an inter-diacetylene spacing, would have very few repeats and would be difficult to observe. van der Waals volumes of these lipids compared to the unit cell volume per lipid, allowing 10 X 4.4 X d / 2 A3per lipid molecule, fill approximately 50 % of the total volume. This compares to 64 % for DMPC in the dihydrate crystal as determined by Pearson and Pascherl9 or 60 % for DC8,9PC, a DAPC which is thought to form highly crystalline bilayers. These data suggest that DGL’s do not form crystals as tightly packed as the long chain DAPC’s such as DC8,9PC or DC8,13PC. Diacetylene-containing molecules must be in a solidstate, well ordered aggregate with precise geometric
-
(17) Lando, J. B.; Sudiwala, R. V. Chem. Mater. 1990,2,594. (18) Hernqvist, L. In Crystallization and Polymorphism of Fats and Fatty Acids; Garti, N., Sato, K., Eds.; Dekker: New York, 1988; p 97. (19) Pearson,; Pascher, I. Nature 1979,281, 870.
Structure of Polymerizable Lipid Bilayers constraints for successful polymerization12 and must be able to accommodate the conformational change that accompanies polymer formation. It has been suggested that the restricted environment of the DAPC‘s is responsible for the relatively limited extent of polymerization.6J3 This is supported by structural studies of mixtures of longchain DAPC’s and short-chain saturated PC’s,S as well as by studies with DAPC’s with various chain lengths.6~8t9 The present data suggest that (1)the DGL’s exist in an ordered, but not crystalline, bilayer in which some packing flexibility may arise from the additional CH2 in the backbone and (2) the DGL acyl chains of equivalent geometry are offset in the plane of the bilayer but not in the bilayer normal direction. A packing/polymerization model is proposed in Figure 9, which takes into account the psuedoequivalence of the acyl chains. This model would allow for formation of long, well-ordered polymers, which are frequently cross-linked to adjacent polymer chains. Instead of chains of polydiacetylene linking together rows of lipids, as would be expected for DAPC’sF a mesh is created. In the presence of organicsolvents or elevated temperature, a set of parallel polydiacetylene chains may become disordered and misaligned, decreasing the effective conjugation length and shifting the absorption to shorter wavelength (blue to red, red to orange or yellow). In the case of polymer from DAPC, the polymer is unconstrained, and the structure will not return to ita original alignment when the tem-
Langmuir, Vol. 10,No. 1, 1994 275 perature is lowered (or solvent removed). When the crosslinked lattice formed by DGL melts, the polydiacetylene is temporarily disordered, and the color shifts to red. Upon cooling, all constituent molecules are in place, and the polydiacetylene can recrystallize to its extended, blue, linear form. This model explains chromatic and solubility characteristics of DGL’s and diacetylenic PC’s. Work is underway to verify the details of the packing arrangement of these and other diacetylenic lipids and to relate these properties to functional characteristics. Nomenclature DAPC = diacetylenic glycerophosphocholine DGL = diacetylenic glutamate lipid DCa,$C = 1,2-bis(tricosa-10,l2-diynoyl)-sn-glycero-3-phosphocholine DNPC = 1,2-dinonanoyl-sn-glycero-3-phosphocholine PC = phosphocholine Acknowledgment. The authors wish to thank John Czausz and Stephen Poole for technical assistance and the staff of the Biomolecular Structure Analysis Center for maintaining the Center and its equipment. Support (D.G.R.) was provided by NSF (CTS-8904938). Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for partial support of this research (D.F.O’B., PRF-19173AC72).