Alteration of Synaptic Membrane Cholesterol/Phospholipid Ratio Using a Lipid Transfer Protein EFFECT ON y-AMINOBUTYRIC ACID UPTAKE*

A procedure was developed to vary the cholesterol- to-phospholipid (Ch/PL) ratio of synaptic plasma membranes and synaptosomes using a nonspecific lipid transfer protein so that membrane lipid composition could be correlated with presynaptic function. In syn- aptic plasma membranes, Ch/PL molar ratios from 0.21 to 1.19 were produced from a normal value of 0.52 _+ 0.01 by incubation with the transfer protein and an excess of either phosphatidylcholine or cholesterol/ phosphatidylcholine liposomes for 60 min at 32 “C. In synaptosomes, Ch/PL ratios from 0.16 to 0.81 were similarly produced from a normal value of 0.38 t 0.04. Cholesterol loading or depletion of the membranes was accompanied by a decrease or increase, respectively in the phospholipid-to-protein ratio. The fluidity of the synaptic plasma membrane, as estimated by 1,6-di- phenylhexatriene anisotropy measurements, was increased by lowering the Ch/PL ratio and decreased by raising the Ch/PL ratio. Decreasing the Ch/PL ratio of synaptosomes and synaptic plasma membrane vesicles resulted in loss of sodium-dependent y-aminobutyric acid (GABA) uptake (70-100% loss at Ch/PL ratios decreased to 40% of normal) and reduction in the number of accessible GABA-binding sites. Choline uptake was not affected in these same preparations. GABA uptake was restored by reinserting cholesterol into the membrane. Synap- tosomal 40% of KH synaptosomes resuspended in KHS medium During lipid low

A procedure was developed to vary the cholesterolto-phospholipid (Ch/PL) ratio of synaptic plasma membranes and synaptosomes using a nonspecific lipid transfer protein so that membrane lipid composition could be correlated with presynaptic function. In synaptic plasma membranes, Ch/PL molar ratios from 0.21 to 1.19 were produced from a normal value of 0.52 _+ 0.01 by incubation with the transfer protein and an excess of either phosphatidylcholine or cholesterol/ phosphatidylcholine liposomes for 60 min at 32 "C. In synaptosomes, Ch/PL ratios from 0.16 to 0.81 were similarly produced from a normal value of 0.38 t 0.04. Cholesterol loading or depletion of the membranes was accompanied by a decrease or increase, respectively in the phospholipid-to-protein ratio. The fluidity of the synaptic plasma membrane, as estimated by 1,6-diphenylhexatriene anisotropy measurements, was increased by lowering the Ch/PL ratio and decreased by raising the Ch/PL ratio.
Decreasing the Ch/PL ratio of synaptosomes and synaptic plasma membrane vesicles resulted in loss of sodium-dependent y-aminobutyric acid (GABA) uptake (70-100% loss at Ch/PL ratios decreased to 40% of normal) and reduction in the number of accessible GABA-binding sites. Choline uptake was not affected in these same preparations. GABA uptake was restored by reinserting cholesterol into the membrane. Synaptosomal membrane potential and synaptic plasma membrane sodium permeability were not affected by changing the Ch/PL ratio. Increase in the Ch/PL ratio above norma1 had no effect on either choline or GABA uptake. Both the decrease in the Ch/PL ratio and the increase in the lipid-to-protein ratio increase membrane "fluidity," which may modulate the vertical displacement and motional characteristics of the GABA transporter.
The ratio of Ch/PL' in mammalian plasma membranes is a major determinant of membrane viscosity (1). The influence of the Ch/PL ratio (and therefore membrane viscosity) on ' The abbreviations used are: Ch/PL, cholesterol/phospholipid (molar ratio); CC5, 3,3'-dipentyl 2,2'-oxacarbocyanine; DPH, 1,G-di-" membrane functions has been studied by manipulation of membrane cholesterol content in specific cells or organelles, either by dietary means (2) or by incubation of membranes in uitro with PC or cholesterol/PC liposomes. The latter method relies on sterol partitioning between liposomes and membranes without significant transfer of phospholipid (3). These techniques have been used to study cholesterol effects on membrane transport (4-6), receptor mobility or signaling ( 7 , 8), and membrane-associated enzymatic activities (9)(10)(11)). In addition, malignant transformation (12) and spur cell formation in liver disease (13) have been correlated with altered membrane cholesterol content. However, these methods, used in the past for altering membrane cholesterol content, have serious disadvantages. Dietary manipulation (in vivo or in cell culture) is not effective for all cell types and results in fatty acid changes as well (14,15). The liposome sterol partitioning method has been effective for cells of the circulatory system, but changes in other types of membranes are difficult to accomplish by this method. We have found that several hours of incubation achieve only minor changes (10-20%) in the Ch/ PL ratio of synaptic plasma membranes when sticking of liposomes to the membranes is accounted for by using a nonexchangeable marker. ' Shapiro and Barchi (16) have reported larger changes in the synaptic plasma membrane Ch/ PL ratio with much longer (24-h) incubations. However, degradative changes which can occur within this time span are a serious problem. Because of these limitations, we have explored the use of a lipid transfer protein to facilitate cholesterol transfer, and have achieved large changes in membrane lipid composition within a relatively brief time span. This study is concerned with evaluating the role of cholesterol in synaptosomal function using a nonspecific lipid transfer protein to vary membrane Ch/PL ratio over a wide range.
For use in the cholesterol depletion or loading experiments, PC (egg or dioleoyl) with or without 1.7-1.9 mol of cholesterol/mol of PC was mixed in chloroform with a trace (2,000-10,OOO dpm/pg of P) of ["Hltriolein or ["Cltriolein (and sometimes 'H-labeled PC). After evaporation of the solvent under a stream of Nz, the lipids were further dried under vacuum in a desiccator preflushed with N2, and then dispersed in 0.32 M sucrose, 10 mM HEPES, pH 7.4, at PC concentrations of 5-10 mg/ml for the PC liposomes and 3-5 mg/ml for the cholesterol/PC liposomes. The PC liposomes were sonicated under Nz until translucent in the bath sonicator. Liposomes containing cholesterol were sonicated using a Branson sonicator with a %inch tipped horn under a stream of N,. These dispersions were then centrifuged at 10, OOO X g,, X 45 min at 20 "C to remove titanium fragments and undispersed lipid.

Preparation of Synaptosomes and Synaptic Plasma Membranes
Rat forebrain synaptosomes were prepared from a washed crude mitochondrial pellet essentially by the method of Gray and Whittaker (20) as modified by Hajos (21). In the final step of the procedure, the synaptosomes were centrifuged into a layer of 0.8 M sucrose, 2.5 mM HEPES, pH 7.4. This layer was collected and diluted over a 1-h period with 3 volumes of KH medium with periodic swirling. The diluted synaptosomes were recovered by centrifugation at 9900 X g,, for 20 min and resuspended in KHS medium to 10-20 mg of protein/ ml. The synaptosomes were used directly after preparation for lipid exchange.
Synaptic plasma membranes were prepared from the washed crude mitochondrial pellet in a manner similar to that of Jones and Matus (22). As described by Jones and Matus, the plasma membrane band was collected at the 28.5-348 (w/w) sucrose interface and then diluted 2-fold with cold water. The synaptic plasma membranes were recovered by centrifugation at 87,000 x g,,.
x 120 min, forming a pellet with a white outer rim and a slightly darker center. The outer rim (Type I SPM) and the center (Type I1 SPM) were separated with a spatula and suspended in SH medium at 10-30 mg of protein/ml. The purer SPM fraction (Type I) had a Ch/PL molar ratio of 0.52 -C 0.01, a phosphorus-to-protein ratio of 33-37 pg/mg, and a sodium-dependent GABA uptake activity of 25-45 pmol/min/mg of protein at 0.15 p~ GABA. This fraction was used for the lipid exchanges unless stated otherwise. The cruder SPM fraction (Type 11) had somewhat lower Ch/PL and PL/protein ratios and a GABA uptake activity of 10-15 pmol/min/mg. This fraction was also suitable for lipid exchange and gave results qualitatively similar to those obtained with the purer SPM fraction. If not, used immediately, the membranes were quickfrozen in liquid nitrogen and stored at -85 "C. Sodium-dependent GABA uptake was stable for at least 1 month of frozen storage.

Protein and Lipid Analysis
Protein was determined according to Lowry et al. (23) or, when mercaptoethanol was present, as described by Ross and Schatz (24). Phosphorus was measured by the method of Chen et al. (25) as described in Rouser and Fleischer (26). Cholesterol was determined using an assay based on enzymic oxidation of cholesterol to cholest-4en-3-one (Sigma Kit 350). Membranes (0.05-0.3 mg of protein) and cholesterol/PC liposomes were added directly to the enzyme-reagent mix containing cholate, without prior lipid extraction, and the absorbance at 500 nm was allowed to develop. Turbidity was normally responsible for 4-10% of the total absorbance for the membrane samples and was corrected for by remeasuring the absorbance after addition of ascorbic acid.
Phospholipid composition was determined by thin layer chromatography and phosphorus analysis as described by Rouser and Fleischer (26). Precoated Silica Gel H plates (Analtech) and a chloroform-methanol-acetic acid-water (50:25:6:3, by volume) solvent system were used. Liposome sticking was corrected for using the nonexchangeable marker ['Hltriolein as described in a later section.

Anisotropy Measurements
DPH was introduced into the synaptic plasma membranes essentially according to the methods of Shinitzky and Barenholz (27) and Shinitzky and Inbar (28). The membranes (2 pg of P/ml) were incubated for 60 min at 25 "C in 0.1 M NaCl, 10 mM HEPES, pH 7.4, containing 0.1 ~L M DPH, giving a phospholipid-to-probe molar ratio of about 650. A sample without DPH was prepared as a blank. Steady state fluorescence polarization of the equilibrated samples was measured using a Perkin-Elmer MPF-44B fluorescence spectrometer with polarizers placed in the excitation and observation light paths. Excitation and emission wavelengths were set at 360 and 430 nm, with respective band widths of 3 and 20 nrn. With the excitation beam vertically polarized, the fluorescence intensity was measured with the emission polarizer in the parallel position (Ill) and then perpendicular position ( I J . The steady state emission anisotropy, r, was calculated according to where G is an instrumental correction factor defined as ZJIll for DPH freely tumbling in ethanol and excited by vertically polarized light. A small correction in the apparent anisotropy was also made for the presence of contaminating liposomes (seee "Results"). Increased anisotropy represents increased restriction of the probe's motion and bas been shown in membrane systems to reflect increased "microviscosity" (or decreased "fluidity") in the core of the lipid bilayer (7). We acknowledge the limitations of applying a macroscopic term to a highly anisotropic micro-environment.

Measurements of GABA Uptake by Synaptic Plasma Membrane Vesicles
Sodium gradient-dependent uptake of [2,3-"HIGABA by synaptic plasma membrane vesicles at 25 "C was studied by the filtration method of Kanner (29). Measurements were performed directly following lipid transfer experiments. A 15-min incubation was used to load the vesicles with potassium phosphate as described by Kanner (see Fig. 4). The loading of the vesicles (normal and cholesterolaltered) with potassium phosphate by this procedure was found, by following passive "'Rh uptake, to be virtually complete after 10 min. The membranes were collected on Millipore HAMK, 0.45-p filters. These fdters quantitatively trapped liposomes present in the samples, allowing reliable subtraction of their contribution to the total '"H radioactivity. This contribution was less than 10% of the total radioactivity on the filter.

Measurement of Synaptosomal Uptake of Choline and GABA
['HIGABA and ['4C]choline uptake by synaptosomes was measured in a single assay mixture by a centrifugation method described by Simon and Kuhar (30) for ["H]choline. The incubation medium was KHS containing 1 p~ [2,3-"HIGABA and 2 p~ [n~ethyl-'~C] choline (37 "C). Sample blanks without radioactive GABA and choline were used to subtract the contribution of liposomal I4C-and/or "H-labeled triolein to the total radioactivity of the synaptosomal pellets.

Measurement of GABA Binding
Specific binding of [2,3-'H]GABA to synaptic plasma membrane vesicles was measured in 0.1 M NaCl at 0-4 "C in the absence of ionic gradients across the vesicle membrane. Specific and nonspecific binding was differentiated by observing the difference between binding in the presence and absence of a large excess of unlabeled GABA. The membrane vesicles were loaded with 0.1 M NaCl, 1 mM MgCI2, 25 mM HEPES, pH 7.4, by incubation at 37 " C for 15 min in that solution, 10.0 p~ [2,3-'H] GABA (2.5-25 Ci/mmol) f 1 mM unlabeled GABA. The equilibrated samples (20 pl in volume) were transferred directly to Amicon 0.22-pm cellulose ester filters and filtered under vacuum. The filters were washed with two 4-ml portions of cold buffer without isotope, air-dried on fiiter paper, dissolved in 1 ml of ethylene glycol monoethyl ether, and counted.

Permeability Measurements
Light Scattering-Light-scattering intensity was measured at a wavelength of 350 nm 90" to the incident beam using a Durrum D-110 thennojacketed stop-flow spectrofluorometer with a soft-stop accessory at 25 "C. The synaptic plasma membranes in SH medium were rapidly mixed with an equal volume of 0.3 M NaCl in SH medium, and the decrease in light-scattering intensity upon reswelling of the vesicles was recorded. The record was digitized (20-25 points) and computer-fitted (by a method minimizing x') to a two-exponential decay of the form Y = A I . exp (-A&) + Aa-exp (-AX) + A:,, where Az and A, are the rate constants of the fast and slow decay components, AI and AJ indicate the proportional weight of each component, and As is the base-line (31). The overall half-time for decay of lightscattering intensity was read from the fitted decay curves. modified synaptic plasma membranes was also measured by loading the membrane vesicles (1.5-3.5 mg of protein/ml) for 4 h at 0-4 "C with incubation buffer containing 0.1 M NaC1,l mM MgClz (containing 50 pCi/ml "Na), followed by dilution with mixing into 40 volumes of incubation buffer without isotope at 25 "C. The half-time for loss of 22Na from the vesicles, measured by filtration through Amicon 0.22p filters at time intervals after the dilution, was used as an index of sodium permeability under these conditions. 22 Nu Permeability-The Na' permeability of control and lipid-

Membrane Potential Measurements
The synaptosomal membrane potential was measured using ['HI TPP' and [I4C]methanol (less than 5 ppm) in KHS medium based on the principles described by Rottenberg (32). The internal vesicular volume was also determined in parallel measurements (32) from the relative distributions of [14C]dextran and 3Hz0. The synaptosomes were preincubated in KHS medium containing 2.5 pg/ml oligomycin, 1.0 pg/ml antimycin A, and either 5 or 65 mM KC1 for 10 min at 37 "C before addition of the isotopes. After equilibration with the isotopes for 10 min at 37 "C, the synaptosomes were collected by centrifugation in a Beckman Microfuge. A sample of the synaptosomes which had previously been lysed in deionized water at 37 "C was also run under similar conditions and was used to determine nonspecific binding of rJHJTPP' in the absence of a membrane potential. After correcting for nonspecific binding, the ratio of TPP' inside and outside the synaptosomes was calculated and inserted into the Nernst equation to yield the membrane potential.
The possibility that the isolated synaptic plasma membrane vesicles have a membrane potential in the presence of the ionic gradients imposed for the GABA uptake assay was investigated by the ["HI TPP' method. The synaptic plasma membranes were loaded with and then resuspended in 0.1 M potassium phosphate, 1 mM MgCL, pH 7.0, as described earlier for the uptake measurements. Aliquots were diluted IO-fold into either 0.1 M NaCl, 1 mM MgC12 or 0.1 M K phosphate, 1 mM MgC12, containing ['H]TPP+ (0.25 p~) and ["C] methanol, or [',C]dextran and 'HHzO. After 2 min, the samples were centrifuged in a Beckman Airfuge at 50,000 X g,, for 2 min. The supernatants and pellets were separated and used to calculate membrane potential essentially as described for the synaptosomes, except that the vesicles with potassium phosphate both inside and outside were used as blanks for ['HH]TPP+ binding in the absence of a membrane potential.
The synaptosomal membrane pot.entia1 was also measured qualitatively using the permeant fluorescent cation CCG by a method based on that of Sims et al. (33) as applied to synaptosomes by Blaustein and Goldring (34). Synaptosomes in KHS medium were incubated at 37 "C for 10 min, and then added (4.2 pg of total phosphorus) to 2 ml of KHS medium containing 2 p~ CCB while continuously recording the fluorescence intensity of the dye in a Perkin-Elmer MPF-44B fluorescence spectrometer at about 30 "C. Excitation and emission were at 470 and 510 nm. After stabilization of the increase in fluorescence observed upon addition of the synaptosomes, 60 p1 of 3 M KC1 were added with rapid mixing to depolarize the synaptosomes. Further addition of KC1 caused no further increase in fluorescence. As an index of the membrane potential, AFe:, K / A F s mM K was calculated, where AF:, mM K is the change in fluorescence upon addition of the synaptosomes to the dye in 5 mM K+ medium, and A F~T , ,~K is the change in fluorescence upon depolarization with 65 mM KC1.

Preparation and Assay of the Lipid Transfer Protein
The nonspecific lipid transfer protein was partially purified from beef liver by the method of Crain and Zilversmit (35), omitting the heat treatment and octylagarose column chromatography. In the final stage of purification, the lipid transfer activity was eluted from a CMcellulose column with 25 mM sodium phosphate, 45 mM NaC1, 5 mM P-mercaptoethanol, 0.02% sodium azide, pH-7.4.
The transfer protein was stored at 0-4 "C in the elution buffer with little loss of activity for 1-2 months. Prior to use, an aliquot was concentrated 5-10-fold without appreciable loss of activity in an Amicon ultrafiltration cell with a PM-10 membrane under N a pressure. When applied to synaptosomes, the transfer protein was dialyzed against elution buffer without azide directly before or after concentration. Finally, the concentrate was centrifuged to sediment any aggregated material. The PC or cholesterol/PC liposomes, for cholesterol depletion or loading, respectively, were then added to start the exchange. For cholesterol depletion, the molar ratio of liposomal phospholipid to membrane phospholipid (on the basis of phosphorus) was generally 4-10. For cholesterol loading, the liposomal cholesterol-to-membrane cholesterol molar ratio was about 10. The exchange was terminated by centrifugation of the membranes through a layer of 0.5 M sucrose at 2 O C for 40 min at 1 0 0 , O O O X gsv. The pellets were suspended in 0.32 M sucrose at 2-4 mg of protein/ml. Sodium gradient-dependent uptake of [3H]GABA and the membrane potential were assayed immediately. Several small aliquots were frozen in liquid nitrogen and stored at -85 "C for subsequent permeability, DPH fluorescence polarization, and GABA-binding measurements. The [14C]triolein content of the pellet was used to correct the phosphorus and cholesterol (and in some cases ['HIPC) content of the pellet for sticking of liposome^.^ This was done using the P/I4C, ch~lesterol/~~C, and 'H/ l4C ratios of the starting liposomes. ["H]Triolein was used in some cases as the nonexchangeable marker in the absence of other '3Hlabeled lipids.
To reverse induced alterations in the membrane Ch/PL ratio, a ' "Sticking" is defined operationally using radiolabeled triolein as a tracer in the phospholipid vesicles, based on the indication that the triolein is not transferred to the membranes by the transfer protein (35). The triolein is used to correct for sticking of the comparable amount of phospholipid and cholest,erol. Some degree of fusion cannot be precluded. If fusion of lipid vesicles with the membrane had occurred in part, the lipid composition of the membrane would have been modified, and the correction for sticking would be inappropriate. This correction was maximally 25% in our studies. If all were due to fusion rather than sticking, we would he in error, at most, by this percentage. Synaptic plasma membranes (63 pg of P, 0.4 mg of cholesterol, 1.9 mg of protein) were incubated with or without egg PC liposomes (267 pg of P) or choIesterol/egg PC liposomes (3.41 mg of cholesterol, 146 pg of P) in the presence of varying amounts of transfer protein for 60 min at 32 "C as described under "Materials and Methods" (the final volume was 1.75 ml). The membranes were separated from the liposomes and transfer protein as described under "Materials and Methods." Correction was made for liposome sticking using a nonexchangable marker, [14C]triolein. The -total sample phosphorus referable to sticking of liposomes was 18-25% in the case of cholesterol depletion and 10-1546 in the case of cholesterol loading. Nanomoles of "H-labeled PC/nmol of total membrane phospholipid.
' Nanomoles of cholestero1 or phospholipid gained or lost/mg of protein during the exchange, compared to the membranes incubated without the transfer protein or liposomes.
second exchange was performed immediately following, differing from the fvst only in the type of liposome present (PC or cholesterol/PC) and its nonexchangeable marker (E3H]triolein rather than ["CJtriolein). Controls were provided by incubating membranes in the absence of liposomes. Synaptosomes-Lipid transfer with synaptosomes was accomplished as described above except that 1) the azide was removed from the transfer protein preparation by dialysis before use, 2) 10 mlrr glucose, 15% of KH medium concentration were added to the exchange mixture to preserve synaptosomal viability, 3) the synaptosomes were collected after the exchange by centrifugation at 60,000 X g , , for 30 min at 2 "C through 0.5 M sucrose containing 40% of KH medium, and 4) the synaptosomes were resuspended in KHS medium at 5-10 mg of protein/ml. During lipid transfer, low ionic strength was maintained to minimize inhibition of the transfer protein (35). Following lipid transfer, the KH medium concentration was increased in two steps (3 and 4 above) to 100% for the functional assays, which were performed immediately.

RESULTS
The lipid transfer protein was found to be an effective means of varying the Ch/PL ratio in synaptic plasma membranes and synaptosomes. In synaptic plasma membranes, Ch/PL molar ratios were varied over a range of 0.21-1.19 from a normal value of 0.52 f 0.01 by incubation with transfer protein in the presence of an excess of either egg PC or cholesterol/egg PC liposomes for 60 min at 32 "C ( Table 1). In synaptosomes, the Ch/PL ratio could be varied from 0.16 to 0.81 compared with a normal value of 0.38 f 0.04. In the latter case, changes in the Ch/PL ratio of the synaptosomal plasma membrane were similar to the changes observed in the whole synaptosome (Table 11). The degree of cholesterol loading or depletion was controlled by varying the amount of transfer activity present and was minimal (cf Table I) in the absence of transfer protein.
Induced changes in cholesterol and phospholipid content of units/ml) at 32 O C for either 45 min or 3 h as described under "Materials and Methods" in order to alter their Ch/PL ratio. The synaptosomes were re-isolated after the exchange by centrifugation through 0.5 M sucrose and aliquoted for lipid analysis. A sample of the synaptosomes was then lysed by hypotonic shock, and the plasma membranes were isolated by density gradient centrifugation (see "Materials and Methods") and analyzed for cholesterol and phospho~ rus content. Liposome sticking was corrected for using ["HJtriolein. Partitioning of lipids between membranes after the exchange incubation (during synaptosomal subfractionation) was minimized bq keeping the samples at 0-2 "C at all times. synaptosomal membranes could be reversed by a second incubation with fresh liposomes and transfer protein (Fig. 1).

When [14C]cholesterol which had been incorporated into synaptic plasma membranes by a previous incubation was transferred back out into egg PC liposomes, the specific activity of
[14C]cholesterol in the membrane did not change. Therefore, the previously introduced cholesterol had equilibrated with the bulk pool of membrane cholesterol (Table 111). Cholesterol loading or depletion of synaptosomes and synaptic plasma membranes was accompanied by a decrease or increase, respectively, in the phospholipid-to-protein ratio  "The Ch/PL ratio of the "unloaded" membranes could not be corrected for liposome sticking (since these membranes were incubated with unlabeled liposomes), and so was not reported.  '' Synaptic plasma membranes and synaptosomes were incubated with transfer protein and liposomes as described in Figs. 3 and 8.
Nanomoles of cholesterol and phospholipid gained or lost/mg of protein during the lipid exchange, compared to the membranes incubated without the transfer protein or liposomes.
' Internal volumes were measured using ["Cldextran and ,'HaO as described under "Materials and Methods." The standard deviations are for two separate determinations on the same samples. When no standard deviation is given, only one measurement was performed.

FIG.
3. DPH fluorescence anisotropy in synaptic plasma membranes at 4,25, and 37 "C as a function of the Ch/PL ratio. Synaptic plasma membranes (223 pg of cholesterol and 35 pg of P/ ml) were incubated in the presence of varying amounts of transfer protein (2-45 units/ml) with egg PC liposomes (150 pg of P/mU or cholesterol/egg PC liposomes (2105 pg of cholesterol and 91 pg of P/ ml) for 90 min at 32 "C in order to alter the Ch/PL ratio. The membranes were re-isolated after the lipid transfer for lipid analysis and DPH anisotropy measurements as described under "Materials and Methods." Liposome sticking, which increased the phosphorus content of the samples by 8 to 2370, was corrected for using ['HI triolein. Control membranes were incubated with transfer protein only (0) or with no transfer protein and no liposomes (A). Increasing anisotropy of the probe's motion indicated increasing microviscosity of the membrane, between the theoretical limits of zero and the anisotropy of DPH in propylene glycol at -50 "C, 0.362 (28). The reported anisotropies of DPH in the synaptic plasma membranes are corrected for the presence of liposomes. Correction was made on the basis of the known relative weights of liposomal and membrane total lipid (phospholipid and cholesterol) in each sample, since in a mixed system, the observed anisotropy, r, is given by r = zLf,rt, where f, is the fraction of total fluorescence intensity emitted by component i (7, 37). In mixtures of untreated SPMs and liposomes, the values for each component, whose r, values could be determined separately, were empirically determined to be equivalent to the fractional weight of phospholipid plus cholesterol associated with that component. The corrections in anisotropy due to liposome contamination were less than 10% for cholesterol depletion and 2% for cholesterol loading. The normal Ch/PL ratio (for this preparation of membranes) is marked by the dotted line.
( Table I) (Table IV), this represents, at an external GABA concentration of 0.15 p~, a GABA concentration gradient of 85-fold across the vesicle membranes. membranes could be accounted for largely by the increase in PL/protein ratio (not shown). Therefore, exchange of membrane and liposomal phospholipid was not a major factor. Phospholipid analysis of the cholesterol-depleted membranes indicated little decrease in phospholipids other than PC (less than 17% on a per milligram of protein basis), although due to PC transfer from the liposomes, the membrane PC content increased from a normal value of 43% to as much as 76% of the total phospholipid (not shown).
Freeze-fracture electron micrographs of control and cholesterol-altered synaptic plasma membranes showed vesicles of similar morphology (Fig. 2). The internal volumes of synaptic plasma membrane vesicles and synaptosomes increased with both cholesterol loading and cholesterol depletion (up to 35

GABA Uptake and Membrane
Lipids and 70%, respectively), in accord with a net transfer of lipid from the liposomes to the membranes in both cases ( Table   IV) .
The fluorescence anisotropy of DPH was measured in synaptic plasma membranes and found to increase with increase in the Ch/PL ratio (Fig. 3). An increase in DPH anisotropy indicates a decrease in membrane fluidity. Decreasing temperature (between 4 and 37 "C), like increasing Ch/PL ratio, decreased the fluidity of synaptic plasma membranes.
Sodium gradient-dependent uptake of I3H]GABA by synaptic plasma membrane vesicles became saturated at 45-90 pmol of GABA/mg of protein, depending on the membrane preparation, at 25 "C and an external GABA concentration of 0.15 p~ (Fig. 4). Uptake was dependent on the presence of a downhill sodium concentration gradient.
The measured GABA uptake capacity was therefore limited in part by the rate at which the imposed sodium gradient dissipated. The uptake rate was temperature dependent, at 4 "C being 5% of that at 25 "C. Kanner (29) has reported a K , of 2.5 ,UM for GABA uptake in a similar preparation of synaptic plasma membranes. This value agrees with that reported for intact synaptosomes (38) and is approximately the same as the dissociation constant for GABA binding (in 0.1 M NaC1) presented later (Fig. 9).
The liposomes as described in Fig. 3. Control membranes were incubated with transfer protein only (0) or with neither transfer protein nor liposomes (A). GABA uptake was measured at 25 "C in 0.15 PM [2,3-3H]GABA, 0.16 mg of membrane protein/ml in the presence of a sodium concentration gradient (0.1 M K phosphate, 1 mM MgCL inside and 0.1 M NaCI, 1 mM MgC12 outside) immediately following re-isolation of the membranes after lipid transfer. The uptake rate was approximately linear at 1 min and was saturated at 5 min. The phosphorus/protein ratios (micrograms/mg) for the 12 samples shown, reading from lowest to highest Ch/PL ratio, were 68.6, 67.2,   60.1, 43.4, 40.3, 36.7 (a), 36.0 (A), 35.6, 36.4, 32.8, 34.4, and 32.2. The horizontal and uerticaE lines denote the Ch/PL ratio and GABA uptake rate and capacity in the control synaptic plasma membranes. synaptic plasma membranes of modified lipid composition is shown in Fig. 5. The uptake rate and capacity were progressively lowered by depletion of membrane cholesterol. This pattern was consistently observed, with a 70-100% loss of uptake rate at Ch/PL ratios near 0.2. GABA uptake by synaptic plasma membranes depleted of cholesterol could be restored to 80% of the control rate by reloading the vesicles with cholesterol (Fig. 6). Incubation with transfer protein or liposomes alone did not affect GABA uptake, nor was any change seen in the membranes excessively loaded with cholesterol (Fig. 5).
Another type of reversal experiment (Fig. 7) was conducted in order to separate the effects of changes in the lipid/protein ratio from the effects of changes in the Ch/PL ratio on GABA uptake by synaptic plasma membrane vesicles. This separation was achieved by determining whether an increasing lipid/ protein ratio correlated with loss of uptake even at Ch/PL ratios above normal. Synaptic plasma membrane vesicles previously loaded with cholesterol were depleted of cholesterol by a second incubation with fresh transfer protein and egg PC liposomes. A parallel sample, incubated the first time with transfer protein only, was similarly depleted of cholesterol. For both, increases in the lipid-to-protein ratio correlated with CONTROL DEPLETED + RELOIOED+

FIG. 6. Reversal of the changes in cholesterol-depleted synaptic plasma membranes by
reinsertion of cholesterol. Synaptic plasma membrane vesicles (Type 11), 1.0 mg of protein/ml, were submitted to two consecutive incubations. The first accomplished cholesterol depletion, and the second one reloading, each with appropriate controls. The samples shown were incubated as follows: 1, both incubations with transfer protein but no liposomes; 2, incubation f i s t with egg PC liposomes (156 Fg of P / d ) and transfer protein, then with transfer protein only; 3, 4, and 5 were portions of 2 taken after the first incubation, and then incubated with cholesterol/egg PC liposomes (4.18 mg of cholesterol and 181 wg of P/ml) only ( 3 ) or cholesterol/egg PC liposomes and transfer protein (4 and 5 ) . The transfer protein concentration in each incubation was 9 units/ml. Both incubations were at 32 "C, the first for 90 min and the second for 75 min. Liposome sticking increased the phosphorus content of the samples by less than 22% and was corrected for using ['Hltriolein.
Sodium gradient-dependent GABA uptake and DPH fluorescence anisotropy measurements were conducted at 25 "C as described under "Materials and Methods." The numbers in parentheses in A are the phosphorus/protein ratios (pg of P/mg of protein) for each sample. ]triolein (see "Materials and Methods"). Sodium gradient-dependent GABA uptake was measured for each sample after the second incubation at 25 "C in 0.15 p~ [2,3-"H]CABA as described under "Materials and Methods." The lipid/protein ratio is the total weight of membrane phospholipid and cholesterol/mg of membrane protein, after correcting for liposome sticking using "H-and I4C-labeled triolein. The numbers in parentheses are the Ch/PL ratios (mole/mole) for each point.
decreasing GABA uptake (Fig. 7). For each set, GABA uptake was higher at higher Ch/PL ratios and at lower lipid/protein ratios. These results indicate that both the lipid/protein and Ch/PL ratio influence GABA uptake.
The effect of alteration in the Ch/PL ratio on function was also studied in intact synaptosomes. Whereas the isolated synaptic plasma membrane vesicles are more readily studied in terms of membrane permeability and fluidity (DPH anisotropy), the synaptosomes retain more functions which may be studied and correlated. For instance, they have a large, measurable membrane potential which is dependent on energization and is therefore a good indicator of viability. Synaptosomal GABA uptake decreased with cholesterol depletion (Fig.  8), but was not affected by the transfer protein alone or by cholesterol loading (the latter not shown). Hence, synaptosomes and synaptic plasma membranes responded similarly with regard to GABA uptake upon alteration of the membrane Ch/PL ratio. Uptake of choline by the synaptosomes was not significantly affected by cholesterol depletion (Fig. 8) or loading (not shown). Choline uptake is apparently inoperative in isolated synaptic plasma membranes and therefore could not be studied in that system.
The loss of GABA uptake observed upon lowering of the membrane Ch/PL ratio was not due to a more rapid dissipation of the required sodium concentration gradient. Lightscattering measurements of passive NaCl permeability in control and lipid-modified synaptic plasma membrane vesicles showed that the half-times of the fast and slow components of the Na' ion movement, as well as their relative proportions, did not change significantly with cholesterol depletion (Table   V). Z'Na permeability also was unchanged in cholesterol-depleted synaptic plasma membrane vesicles assayed with equal internal and external Na' concentrations, confirming that the Less than 16% of the sample phosphorus was referable to stuck liposomes in all cases. Uptake of [''C] choline (2 p~) and rJH]GABA (1 p~) was measured at 37 "C following re-isolation of the synaptosomes after the lipid transfer as described under "Materials and Methods." The rates were measured while uptake was linear (at 2 rnin). The error bars indicate the standard error of two determinations.
loss of sodium gradient-dependent GABA uptake was not due to a change in membrane permeability to sodium.
The transmembrane potential of the synaptosomal membrane, as measured by [3H]TPP' in the presence of oligomycin and antimycin A, was not significantly affected by lowering the Ch/PL ratio to 0.15 (Table VI). Increasing the external potassium concentration from 5 to 65 mM depolarized the synaptosomes, as did 0.2 mM veratridine (not shown), confirming the measured potential as referable to the synaptosomal plasma membrane. The membrane potential of normal and lipid-altered synaptosomes was also estimated using CC5 fluorescence. There was no change in membrane potential, as measured by this method, in cholesterol-depleted (Table VI) or cholesterol-loaded (not shown) synaptosomes. Furthermore, no transmembrane potential was detectable with ['HI TPP+ in isolated synaptic plasma membrane vesicles, regardless of previous exposure to transfer protein or liposomes. This was true even within 2 min of imposition of ionic gradients for GABA uptake measurements. GABA uptake was diminishing although still active in this time frame (cfi Fig. 4 ) .
Specific binding of [2,3-:'H]GABA to synaptic plasma membranes of lowered Ch/PL ratio was decreased relative to controls and cholesterol-loaded membranes, paralleling the loss of GABA uptake. The binding was studied under conditions minimizing GABA uptake: 0-4 "C in 0.1 M NaCl, but in the absence of a sodium gradient across the vesicle membrane.
Specific binding of ['HIGABA was 1.71 * 0.01 pmol of GABA/ mg of protein at 0.15 PM GABA in normal synaptic plasma membranes (Ch/PL ratio = 0.51) which had gone through a (28 f 0.4%) (72 f 0.4%) ______ Uptake was measured at 25 "C directly after the lipid exchange as described in Fig. 5 (in the presence of a downhill Na' gradient).
The synaptic plasma membranes, 0.05-0.10 mg of protein/ml in SH medium, were diluted with an equal volume ' Escape of "Na from preloaded membrane vesicles was measured at 25 "C as described under "Materials and Methods-.."

TABLE VI
Membrane potential of control a n d cholesterol-depleted synaptosomes Synaptosomes were incubated with or without egg PC liposomes and either 20 or 40 units/ml of transfer protein as described under "Materials and Methods" and in Fig. 8. The synaptosomal membrane potential was measured directly following the exchange, 30-60 min before the uptake measurements shown in Fig. 8 (2) The synaptosomal membrane potential was measured in the presence of oligomycin and antimycin A using the permeant cation ['H]TPP+ (0.21 p~) and ['4C]methanol, in conjunction with parallel measurements using [I4C]dextran and 'Hz0 to determine the intravesicular volume (see "Materials and Methods"). Without the mitochondrial inhibitors, the apparent synaptosomal potential in 5 mM K' medium was approximately -120 mV. The inhibitors were included in order to eliminate contribution of the intrasynaptosomal mitochondrial potential to the overall potential, although they may also have lowered the potential across the synaptosomal plasma membrane indirectly by decreasing the supply of ATP.
Fluorescence intensity of the permeant cationic dye CCS was measured as described under "Materials and Methods." AFr, mhl K is the fluorescence change upon addition of the synaptosomes (2.1 pg of P/ml) to the dye (2 p~) in KHS medium containing 5 mM KC1. FIG. 9. Specific GABA binding to control and cholesteroldepleted synaptic plasma membranes. Synaptic plasma membrane vesicles were incubated with either 1) transfer protein and egg PC liposomes (0) or 2) neither transfer protein nor liposomes (O), as described in Fig. 3. The membranes were stored at -85 "C after the lipid transfer before measurement of binding. Specific GABA binding in the presence of 0.1 M NaC1, 1 mM MgC12, pH 7.4, was measured a t 0-4 "C in the absence of ionic gradients (see "Materials and Methods"). Sample blanks, without ['HIGABA, were used to subtract contaminating ['Hltriolein. At most, 34 and 3.5% of the total counts on the filter were referable to ['Hltriolein and nonspecific [''HJGABA trapping on the filter, respectively. These estimates apply to measurements at low (0.15 p~) GABA concentration for the cholesteroldepleted membranes. The Ch/PL molar ratios of the control and cholesterol-depleted membranes were 0.51 and 0.18, and their respective GABA uptake rates (see Fig. 5) were 29.9 and 5.0 pmol/min/mg. The total heights of the vertical bars are twice the standard error for three separate binding measurements at each GABA concentration. repletion to '73% was accompanied by a decrease in the number of binding sites (at 0.15 PM GABA) to 46% of control and recovery to 86% of control. Binding to control and cholesteroldepleted membranes over a range of GABA concentrations (0.15-10 p~) is depicted in a double reciprocal plot in Fig. 9. Nonspecific binding, which did not show saturation, represented 17% of the total binding at 0.15 p~ GABA and 41% a t 10 p~ GABA for control membranes. For the cholesteroldepleted membranes, these percentages were 22 and 58%. For the control, the dissociation constant (KD) and binding capacity (Bmax) for specific GABA binding were 2.56 k 0.5 p~ and 31 f 5.6 pmol of GABA/mg of protein. The cholesteroldepleted membranes had a KD of 2.00 & 0.6 PM and B,,, of 16 k 4.2 pmol of GABA/mg of protein.

DISCUSSION
A procedure is described which induces large changes in the Ch/PL ratio of synaptic plasma membranes and intact synaptosomes using a nonspecific lipid transfer protein from beef liver. These changes can be achieved within a few hours under gentle conditions. With this methodology, we were able to correlate the membrane Ch/PL ratio with several membrane functions relevant to presynaptic function.
Sodium-dependent uptake of GABA which is believed to function in the termination of synaptic transmission by GABA-containing neurons (39), was markedly sensitive to reduction in membrane Ch/PL ratio in both synaptosomes and synaptic plasma membranes. In contrast, synaptosomal high affinity choline uptake and the sodium permeability of synaptic plasma membrane vesicles were not affected by alteration of the membrane Ch/PL ratio. GABA uptake could be restored by reinserting cholesterol into the membrane, suggesting that the loss of function was referable a t least in part to the removal of cholesterol.
In isolated synaptic plasma membranes, sodium gradientdependent uptake of GABA has been shown to be enhanced by agents which would be expected to create or increase a negative interior membrane potential, such as valinomycin in the presence of higher intravesicular potassium (29). In synaptosomes, a membrane potential may also influence sodiumdependent GABA uptake (40). The relative influence of the sodium concentration gradient and transmembrane potential on GABA uptake has not been determined. The membrane potential of the intact synaptosomes was not affected by protein-mediated cholesterol depletion (Table VI), and GABA uptake was retained in synaptic plasma membrane vesicles in the apparent absence of a transmembrane potential. Hence, the membrane potential is not sensitive to changes in membrane cholesterol content, nor does it seem to be required for GABA transport. Our data do not rule out the possibility that the presence of a negative electrical potential might stimulate GABA uptake. Despite the apparent absence of a membrane potential, the synaptic plasma membrane vesicles were able to generate a GABA concentration gradient of about 85-fold, based on an internal volume of 6.7 pl/mg of protein and a capacity of 85 pmol of GABA/mg of protein (cf. Fig. 4).
There is a complication in the correlation of loss of GABA uptake with decrease in membrane cholesterol content in that the decrease in the Ch/PL ratio is attended by an increase in the lipid/protein ratio during protein-mediated lipid transfer experiments. We have been able to sort out the effects referable to changes in Ch/PL and lipid/protein ratios on GABA uptake (Fig. 7). A decrease in GABA uptake correlates with both a decrease in the Ch/PL ratio and an increase in the lipid/protein ratio. However, it is clear from Fig. 7 that, at any given lipid/protein ratio, GABA uptake is inhibited with a decrease in cholesterol content. Removal and reinsertion of cholesterol can be correlated directly with loss and recovery of GABA uptake (Fig. 6): GABA uptake by cholesterol-depleted synaptic plasma membranes was restored to 80% of the control by restoring the original Ch/PL ratio, even though very little of the phospholipid incorporated into the mem-branes during cholesterol depletion was removed during reinsertion of cholesterol.
Although the influence of the lipid/protein ratio on GABA uptake is evident, a specific effect on GABA transport by egg PC, introduced into the membrane during cholesterol depletion, appears unlikely for several reasons. First, the fatty acid composition of the egg PC incorporated into the membrane was similar to that of the synaptic plasma membrane PC: both are relatively high in palmitic acid and low in long chain polyunsaturated fatty acids (41, 42). By contrast, the fatty acid compositions of phosphatidylserine and phosphatidylethanolamine in the synaptosomal plasma membrane are much higher in long chain polyunsaturated fatty acids. Second, the loss of GABA uptake could be reversed by treatment with transfer protein and cholesterol/egg PC liposomes, a process which itself results in transfer of liposomal egg PC to the membrane (cf . Table I). Third, in other studies, we found that exchange of up to 50% of the synaptosomal PC pool with several classes of synthetic PCs, including dimyristoyl PC, dioleoyl PC, and dielaidoyl PC, resulted in no change in synaptosomal uptake of GABA. These latter experiments made use of a specific PC exchange protein from beef liver! Loss of GABA uptake also does not appear to be referable to loss of specific membrane phospholipids since little of the original membrane phospholipid was lost during cholesterol depletion experiments. The major lipid alterations were exchange of liposomal egg PC with membrane cholesterol and addition of liposomal PC to the membranes without back exchange.
The decrease in the Ch/PL ratio and increase in the lipid/ protein ratio which occur concomitantly during transfer protein-mediated cholesterol depletion can both be correlated with increased membrane fluidity. Rigidification of phospholipids in their liquid-crystalline state by the presence of cholesterol has been demonstrated in liposomes using several techniques, including fluorescence (1, 43), EPR (44), and NMR (45). The incorporation of protein into phospholipid vesicles has been found to result in a small decrease in the rate of intramolecular phospholipid motions as compared to the phospholipid vesicles alone (46). In this study, we find that loss of GABA uptake by synaptic plasma membrane vesicles correlates with reduced DPH fluorescence anisotropy, and therefore increased membrane fluidity (see Figs. 3 and 6).
Thus, it is plausible that GABA uptake is lost during proteinmediated cholesterol depletion directly in response to increased membrane fluidity, resulting from both a decrease in the Ch/PL ratio and, to a lesser degree, increase in the lipid/ protein ratio. We cannot rule out the possibility of a specific requirement for cholesterol by the GABA transporter. In tissue culture studies, Goldstein et al. (47) have shown that cholesterol is essential for cell growth. The molecular basis for this essentially has not been established.
There have been several reports of reduced surface accessibility, and decreased rotational mobility, of proteins in membranes fluidized by a variety of means, including sterol exchange into PC liposomes (48)(49)(50)(51), cholesterol extraction with polyvinyl pyrrolidine dispersions of PC or linoleic acid (52,53), phospholipid enrichment of cells by liposome fusion (54), and malignant transformation (1). The response of a particular protein receptor to membrane fluidity changes would reflect its particular characteristics and orientation in the membrane and, perhaps, the method used to alter the membrane's fluid state. Four different receptors have been shown to be reduced in surface accessibility by increased fluidity of the membrane. These include P-adrenergic receptors in Chang liver cells ( 5 4 , P. North and S. Fleischer, manuscript in preparation. serotonin (52) and opiate (53) receptors in mouse brain membranes, and the glucose carrier protein in human erythrocytes (51). Other receptors, in contrast, did not respond to changes in membrane fluidity. These include P-adrenergic receptors in turkey erythrocytes fluidized by insertion of cis-vaccenic acid (55) and a-adrenergic receptors in cholesterol-depleted human platelets (8). Dependence of carrier-mediated glucose transport on membrane fluidity has also been reported in erythrocytes (51) and adipocytes (56).
Our r3H]GABA-binding analysis correlates loss of sodium gradient-dependent GABA uptake with decreased number of available GABA-binding sites (Fig. 9). The lost sites were not expelled from the membrane, since uptake could be restored by reinserting cholesterol into the membrane. (Released sites would have been removed from the system during centrifugation of the membranes through sucrose before the cholesterol-reloading experiment.) Sodium-dependent GABA-binding sites have been shown to be referable to the sodiumdependent GABA uptake system found in synaptosomal preparations (57). A decrease in sodium-dependent GABA receptors would be expected to decrease GABA transport. High affinity sodium-independent GABA binding (KO 1 0.2 PM) to what are probably postsynaptic GABA receptors (58,59) may also be included in the data presented in Fig. 9. However, the contribution of these sites in synaptic plasma membrane preparations (about 5 pmol/mg of protein) is comparatively small (57,58). Loss of GABA uptake therefore correlates with increased membrane fluidity and a reduction in the number of available sodium-dependent GABA-binding sites.
The observed loss of GABA receptor accessibility upon membrane fluidization is consistent with the model of Shinitzky and co-workers (48)(49)(50) that the vertical displacement of membrane proteins is controlled by the lateral pressure exerted by surrounding membrane lipids. Such lateral pressure would be decreased by membrane fluidization. The vertical displacement model is based largely on studies which have correlated increased membrane fluidity with decreased accessibility of membrane proteins to impermeant reagents. According to this model, the lowering of specific GABA binding and therefore decreased GABA uptake rate which we observe with decreased Ch/PL ratio in synaptic plasma membranes would be due to an inward vertical displacement of the GABA transporter protein. That is to say, the GABA-binding sites normally exposed at the surface of the membrane become buried and inaccessible. The decrease in GABA binding is not directly proportional to the loss of uptake. For the cholesteroldepleted membranes shown in Fig. 9, the rate of GABA uptake was reduced to 16.7% of the control, whereas binding capacity was reduced to only about 50% of the control. Therefore, additional factors are apparently involved. The GABA transporter protein has been solubilized and reconstituted into lipid vesicles by Kanner (60). Study of the motional characteristics of the transporter protein in a reconstituted system of defined composition could provide more definitive interpretation of the effect of membrane lipid composition on the GABA uptake mechanism.
As yet, the application of lipid transfer proteins to modification of the lipid composition of membranes has been limited. This study illustrates the power of such application. Crain and Zilversmit (61) found using the nonspecific transfer protein from beef liver that in microsomal membranes a 27% depletion of phosphatidylethanolamine produced a 37% inhibition of glucose-6-phosphate phosphohydrolase activity. Voelker and Kennedy (62) used the PC exchange protein to provide lipid substrate to an isolated plasma membrane fraction for sphingomyelin synthesis, with the implication that transfer proteins may serve in this role physiologically.