Eur J Biochem / l / , 315-324(1980) (, by FEBS 1980

Interaction of GMl Ganglioside with Bovine Serum Albumin Formation and Isolation of Multiple Complexes Maurizio TOMASI, L. Giorgio RODA, Clara AUSIELLO, Giuliano D'AGNOLO, Bruno VENhRANDO, Riccardo GHIDONI, Sandro SONNINO, and Guido TETTAMANTI

Laboratorio di Biologia Ccllulare e Immunologia, lstituto Superiore di Saniti, Koma, Laboratorio di Farmacologia, Universiti di Ancona and Istituto di Chimica Biologica, Facolti di Medicina e Chirurgia, Universiti di Milano

(Received June 27. 1979/March 24. 1980)

The binding of ganglioside C r M l to bovine serum albumin has been studied by using absorption and fluorescence properties of the protein chromophores. Differences in the ultraviolet absorption spectrum and in fluorescence quenching, as well as a marked shift of the wavelength at the fluorescence maximum provide information about the binding of this ganglioside to albumin. Ultracentrifugal studies showed that there are two forms of the GM1-protein complexes which differ markedly in their molecular weight. These two forms have been separated on this basis, by a chromatographic sieving procedure, and designated as complexes I and IT. Both complexes are characterized by a G M ~ : protein ratio of one ganglioside micelle per albumin polypeptide chain. Complex I1 polymerizes slowly and irreversibly to a dimer, complex I. These results have been correlated with the optical studies in order to draw limited inferences as to the environment of the binding sites on the native protein. The interaction between G M M I micelles and albumin is mostly hydrophobic and the two complexes are actually mixed ganglioside-protein micelles. At submicellar concentrations of ganglioside a binding of ganglioside GMI to albumin also occurs. This process is due, however, to an aspecific, reversible adhesion of GMl inolecules on the albumin surface with no apparent perturbation of the albumin' structure.

In the past few years, i t has become increasingly evident that gangliosides, glycosphingolipids contain- ing sialic acid [l], contribute to several properties of natural cell membranes. With their hydrophilic oligo- saccharide moieties protruding out of the membrane, gangliosides are well suited to play a role in the trans- mission of biochemical sigiiak across the membrane and into the cell. The biochemical information, gene- rated by ligand-ganglioside interaction at the cell surface, forins an intrinsic part of a variety of events, such as the cellular processes activated by the bacterial toxins of Vihrio cholercte [2] and C'lostridium tctnizi [3], by lectins [4,5], by hormones [6]. by interferon (71 and by cell-cell recognition [8,9]. The carbohydrate pattern of the oligosaccharide portion of the gan- gliosidcs seems to determine the specificity of the

Ahhreviation. Ganglioside GMI, galactosyl-N-acetylgalactosa- minyl-(N-acetylneuraminosyl)-ga~actosy~g~ucosy~c~ramide. ['13)- Gu,, tritium-labelled GM,.

Enzyme. Galactose oxidase (EC 1.1.3.9); Vihrio cholcwe neuraminidase (EC 3.2.1.18).

ganglioside-protein interaction. Moreover, proteins that can cross-link gangliosides on the cell surface may cause redistribution of these gangliosides into segregated regions [lo, 111, which results in a modi- fication of membrane fluidity. Apparently a higher membrane fluidity due to ganglioside aggregation is a characteristic of the structuring and stability of functional synaptic membranes [12,13]. This diffe- rent localization of gangliosides in the lipid matrix will expose proteins inside the aggregate to a different environment than that surrounding proteins outside tlic aggregate. This, and the observation of the receptor activity of gangliosides, support the idea that inter- actions between proteins and membrane gangliosides are determinants not only of membrane structure but also function.

Recent progress in the study of biological mem- branes has been obtained by the use of model systems consisting of soluble amphiphiles and various proteins. The examination of simple models has not only pro- vided precise data on molecular interactions but also, in some cases, insight into the basic parameters of

316 Interaction of Gangliosides with Bovine Serum Albumin

the biological systems. As a prerequisite to this type of understanding it is necessary that the structures and compositions as well as the properties of the models be determined. In the case of the protein-ganglioside interactions previous evidence has shown that a simple model can be provided by the association of ganglio- sides with bovine serum albumin [14,15]. The purpose of the present communication is to describe the prep- aration and the properties of bovine serum albumin complexes with gaiiglioside G M l and to define the usefulness of such a model system in the study of protein-ganglioside interactions.

MATERIALS AND METHODS

The polyacrylamide-agarose gels, Ultrogel, were purchased from LKB (Roma, Italy). BioGel A-1.5 m was obtained from Bio-Rad (Richmond, CA, USA). Delipidated bovine serum albumin and N-acctyl- neuraminic acid were obtained from Sigma (St Louis, MO, USA). Vihrio cholrrnr neuraminidase was purch- ased from Serva (Heidelberg, FRG), while galactose oxidase was obtained from KABI (Stockholm, Swe- den). Sodium bor~[~H]hydr idc was obtained from the Radiochemical Centre (Amersham, U K). Collodion bags and a filtration apparatus were obtained from Sartorius Membranfilter (Gottingen, FRG). The pre- blended liquid scintillation solution, Instagel, was obtained from Packard Instruments (Milano, Italy). All other reagents were purchased locally and were of the highest purity available. Water for routine use was freshly redistilled in a glass apparatus and solvents wcre distilled before use.

Prepurarion of Monomeric Bovine Serum Alhunzin

About 20'); of the commercial preparation of bovine serum albumin employed was in the form of polymeric aggregates. Removal of the aggregates, mostly dimers, was achieved by a chromatographic procedure. 2 tnl of a protein solution (25 nig/nil) were applied to a column (1.6 x 200 cm) of Ultrogel ACA 44 equilibrated with 0.05 M sodium phosphate buffer pH 7.5. Fractions were assayed by polyacryl- amide disc gel electrophoresis. Fractions containing the monomeric form OF bovine serum albumin were pooled and concentrated, without stirring, in a Sar- torius Membranfilter apparatus (model S.M. 26314), and stored at 4 ' -C for no more than a week before use.

Preprution und Luhelling of Gunglioside GM 1

A sample of mixed calf brain gangliosides, isolated according to Tettamanti et al. [16], was exhaustively digested with V. clzolerue neuraminidase by the method of Ghidoni et al. [17], in order to transform poly-

sialogangliosides into the monosialoganglioside GM1.

At the end of the digestion, ganglioside was iso- lated, purified and assayed as described by Sonnino et al. [I 81. Chromatographic analysis indicated that the purity of G M 1 was over 987;. The content of sphingosine bases, determined according to the method of Carter and Gaver [19], was found to be: C18:(,, 8.1 32 .4x; Czo:o, 5.494; Cz0:], 54.1 o o . The fatty acid side chain was predominantly (98 : I , , ) octa- decanoic acid.

The ganglioside GMlr employed in our studies, will have the aggregation characteristics of gangliosidc molecules, formed from an average of the monosialo- gangliosides and the polysialogangliosides [ 12,201.The gangliosides micelle was examined by light-scattering experiments. A micellar size of M , 532000 i 30000 was found for ganglioside Ghll [21], as expected on the basis of its sphingosine-base and fatty acid corn- position, and approximately 350 monomer molecules were present in each micelle.

Tritiated GMl was obtained by oxidation of the terminal galactose by galactose oxidase followed by reduction with sodium b~ro[~H]hydride using the method of Suzuki and Suzuki [22] as modified by Ghidoni et al. [23]. Because of the use of sodium b~ro[~H]hydr ide of very high specific radioactivity (5 - 20 Ci/mmol) the procedure employed gave gan- glioside Gb,l with a specific radioactivity of 1 .S c'i mniol. The radiochemical purity of "H-labelled gan- glioside GM1 was about 99 ". 31-l-labelled ganglioside GM1 was stored at 4°C in n-propanoljwater (2 .1 , by vol.) and was found to be stable for at least six months.

~l~Jc.troi,fiorrsi,s

Polyacrylamide disc gel electrophoresis was done according to the method of Davis [24] as modified by Laemmli [25]. The gels werc stained with a solution of 0.2 ',>,; Coomassie R 250.

Cliernicul Procediires N-Acetylneuraminic acid was determined by the

method of Svennerholm [26]. Protein was determined by the method of Munkres and Richards [27] or of Lowry et al. [28]. This second method was routinely used for protein determination in the presence of ganglioside GMl . Since ganglioside GMI caused inter- ference with color development, probably preventing the dye from binding to the protein, standard curves were run with pure albumin and the amount of gan- glioside employed in each experiment.

Ultrcicen tr if ugul St Lrdies

Sedimentation velocity experiments were per- formed at 10 C and 52000 rev./min in a Beckman € analytical ultracentrifuge equipped with schlieren op-

M . Tomasi. L. G. Roda, C. Ausiello. G. D'Agnolo. B. Venerando. R . Ghidoni, S. Sonnino. and G. Tettamanti 317

tics. Sample:, at a concentration of 22 pM of bovine Isolation of GAwl-( Bovine Serum AIhumin) Conzplexe.~ serum albumin and 3 mM ot' ganglioside GMI, in micellar form, were preincubated for 24 h at 20 C in 0.05 M sodium phosphate buffer pH 7.5, prior to the analysis.

Opt ical M c tliou's

Fluorescence spectra were recorded with an Amin- co-Bowman fluorescence spectrophotometer and are reported without correction for spectral efficiency. Difference spectral measurements were made in a Varian Cary 118 recording spectrophotometer at 20'C, in 0.05 M sodium phosphate buffer pH 7.5, with Yankeelow double sector cells. The sample cell, contained bovine serum albumin (22 pM) in the first compartment and varying amounts of ganglioside GMl in the second compartment. A second cell, pre- pared in the same manner, was placed in the reference beam of the spectrophotometer. After recording a good baseline, the contents of the sample cell were mixed, by inverting it several times, and the difference spectrum was recorded. Since scattering of the mixed samples increased significantly with time, the spectra were corrected by extrapolating at wavelengths below 310 nm the differential curve recorded between 400 and 310 nm.

Binding of Gungliositle G,wl to Bovirze Serum Alhuvni~z

In order to have in solution the expected physical form of ganglioside G M ~ , the following standard con- ditions were employed in preparing GMI solutions for binding experiments. To a known amount of ganglioside G M ~ , dissolved in water, 1.5 pCi of 3H- labelled ganglioside GMI was added (when the final ganglioside concentration was 0.1 pM, 0.3 pCi of 'H-labelled ganglioside GM, were employed) and the mixture was lyophilized. The residue was dissolved in 2 ml of chloroform/methanol (2/1, by vol.). The solution, containing the ganglioside as free monomeric molecules, was evaporated to dryness; the residue was taken up in 1.0 nil of 0.05 M sodium phosphate buffer and, following the observations of Formisano ct al. [15], allowed to stand at 20' C Tor at least one day. The critical micellar concentration of ganglioside GMl employed in this study was found to be lower than 1 1tM [21]. Monomeric bovine serum albumin was dissolved in 1 ml 0.05 m sodium phosphate bufT'er pH 7.5, (concentration: 22 1M) and added to the GMl solution. The resulting mixture, containing a fixed amount of albumin (1 1 pM) and varying amounts of ganglioside were allowed to stand at 20'C for a given time, then submitted to analysis.

Ganglioside-protein complexes were isolated by passing the solutions obtained from the binding experiments through a BioGel A-1 .S m column (1 x 200 cm) equilibrated with 0.05 M sodium phoa- phate buffer pH 7.5. The temperature was kept at 20 C and the column was eluted with the equilibra- tion buffer at a flow rate of 8 ml/h. Fractions (3.5 ml) were collected and analyzed for N-acetylneuraminic acid and protein content, absorbance at 280 nm and radioactivity. Radioactivity determinations were made by adding 10 ml of Tnstagel solution to the sample in a scintillation vial and counting in a Packard Tri- Carb liquid scintillation spectrometer. Results were fairly reproducible provided that newly packed col- umns, with fresh gel, were used in each experiment to avoid contamination by ganglioside GM1 adsorbed on the column materials.

RESULTS Interuction qf Gunglioside G M 1

with Bovine Serutii Allmmiti

The amphipathic structure of ganglioside mole- cules is closely connected with a pronounced surface and interfacial activity [15, 291. The binding capa- bility of gangliosides to different materials causes difficulty in carrying out binding experiments by con- ventional methods such as equilibrium dialysis. There- fore we have studied the effects of ganglioside G M ~ on some of the optical properties of bovine serum a1 bumin.

The addition or solutions of ganglioside GM, to a solution of bovine serum albumin results in charac- teristic difference spectra (Fig. l ) , with a minimum at 287 nm, which is a contribution of aromatic sidechain blue shifts due to the unmasking to the solvent of the protein chromophores [30]. This effect can be observed only at ganglioside coaentrations exceeding 0.1 mM, where most of ganglioside is present in mi- cellar form. The effect increases a s G M ~ concentration increases up to about 4 mM, where the effect is com- plete. To determine the stoichiometry of the G M ~ - albumin complex, a constant amount of protein was gradually titrated with ganglioside GM.11 (Fig. 2A). Under our experimental conditions, the amount of GMI bound to the protein was dependent upon the concentration of GM1 and a maximum of one ganglio- side micelle bound per molecule of protein was ob- served. The value of M , 532000 for thc micellar size of ganglioside G M I , determined by laser light scatter- ing measurements [21], was used in our calculations. The kinetics of the binding of ganglioside GMl to albumin is shown in Fig.2B. The difference in absorp- tion of bovine serum albumin increased upon addition of ganglioside G M ~ . This increase of the minimum

31X Interaction of <?angliosides with Bovine Scruin Albumin

amplitude at 287 nm did not follow simple kinetics. The logarithm of absorbance changes versus time (not shown) is at least bimodal, indicating a fast pro- cess at first and then a much slower one. The rate constant of formation of the complex cannot be cal- culated from this data because, in our experimental conditions, the reactive species of the ganglioside being the micelle, the incubation mixture will contain

V , I ,

3 270 29 0 310 Wavelength (nm)

Fig. 1, Difference .spectru of' GMl-hovine .serum ulhunim complex. Bovine serum albumin at a final concentration of 11 pM was mixed with different amounts of ganglioside CHI in 0.05 M sodium phosphate buffer pH 7.5 at 20 'C and the diRerence spectrum was recorded after 24 h of incubation as described under Materials and Methods. Gwl final concentrations: (- -- ) 1.5 mM : (--) 3 mM ; (0-0) 4 mM

0 A

- 2c 0 1.0 2.0 3.0 4

[Ganglioside G,,](mM)

micelles and polypeptide chains in almost equivalent amounts even at the highest G M ~ concentration em- ployed (4mM). In fact, for a bimolecular reaction of the type X + Y+C the pseudo-first-order rate con- stant of formation of the complex can be calculated only from data obtained with the ligand X present in large excess [31]. Similar results were also obtained by lowering the protein concentration in such a way to obtain micelles in excess of 2 : l or 3 : l over the protein. Attempts to run experiments at higher GMl concentrations failed because appreciable light scatter- ing became evident shortly after mixing the ganplioside and protein solutions.

To characterize further the interaction of ganglio- side GMI and bovine serum albumin, incubations simi- lar to those described in Fig. 1 were carried out and the fluorescence emission spectrum was recorded. Addi- tion of ganglioside GM1 to the bovine serum albumin produces a decrease in the maximal fluorescence in- tensity with increasing ganglioside content. Spectra shown in Fig. 3 reflect a change in the environment of tryptophans and indicate an association between the protein and the ganglioside. A limiting effect of 1 - 1.2 micelles of ganglioside per albumin polypeptide chain on the fluorescence properties was observed and an increase of the micelle/polypeptide chain ratio up to 2 causes no further spectral changes.

Serlinzcn tu t ion Brho v iow o f G M I - (Bovine Scwni Albumin) Coniplrse.5

In an attempt to evaluate the possible effect of bound ganglioside G M l on the hydrodynamic proper- ties of bovine serum albumin, sedimentation velocity experiments were performed. These studies have shown that, after binding with ganglioside GMl in

Time (h)

Fig. 2. Ejfect of' GWl conc.enfrution on,formrition o f '~un~l io . r i~ l i~-ho~~inc~ serum ulhumin coniplex. (A) Bovine serum albumin, at a final concen- tration of 11 pM. was incubated with the indicated amounts of ganglioside GMl, in 0.05 M sodium phosphate buffer pH 7.5 at 20 C. Aftcr 24 h of incubation, the difference spectrum was recorded as described under Materials and Methods. (B) Bovine serum albumin was incubated under the bame conditions as under A with ganglioside Gu, at final concentrations of 0.75 m M (&--O), 2 mM (0- n), 4 mM (O---O) for the indicated periods of time at 20 C and the difference spectrum was recordcd. Results are expressed as difference in absorbancc of bovine serum albumin at 287 nm

M. Tomasi. L. G. Roda, C. Ausiello, Ci. D'Apnolo, B. Venerando, R . Ghidoni. S. Sonnino. and Ci. Tettamanti 31')

% 25

, "300 325 350 375 400

Wavelength (nrri)

Fig. 3 . F1uowscenc.e speclru oj' GM,-(bovit ie serum ulbumin) ronz- p / ~ . w ~ . Bovine scruin albumin at a final concentration of 11 pM was mixed with different amounts of ganglioside GMI in 0.05 M sodium phosphate buffer pH 7.5 at 20°C in a final volume of 1.0 nil and fluorescence emission was recorded, after 24 h of incubation, with the excitation wavelength of 278 nm. Spectrum of free bovine serum albumin ( - ~ ) and of complcxes formed between albumin and ganglioside GMI at a final concentration of 1 .S inM (-- ) o r 3 rnM (. - -. -)

micellar form, provided that the ganglioside concen- tration exceeds 0.1 mM, bovine serum albumin ex- hibits an unusual behaviour in the ultracentrifuge. Using 3.0 mM ganglioside GMl the sedimentation pattern is characterised by the appearance of three distinct peaks, which, at 22 pM albumin concentra- tion, have apparent sedimentation coefficients of 2.5, 7.2 and 11.2 S (Fig.4). The sedimentation velocity experiments were carried out at a protein concentra- tion in excess of 2/1 with respect to the micelle con- centration. A decrease of G M ~ concentration results in progressively larger amounts of material sediment- ing at approximately 2.5 S and a decrease in the amount of material sedimenting at 11.2 S. These re- sults suggest, that the 2.5-S peak represents the free form of bovine serum albumin while the 7.2-S and 1 1.2-S peaks may represent forins ofthe protein bound to the ganglioside. Also in this case the form of the interacting ganglioside should be assumed to be the micelle. We therefore attempted to separate the three different molecular species on the basis of differences in molecular weight using a gel filtration technique.

Isolotioiz of GMI- (Bovine Serunz Albumin) Coniplest~s

Considerable efforts were made, in preliminary experiments, to find out a simple analytical system capable of separating: (a) monomeric from micellar G M ~ ; (b) G M I micelles from bovine serum albumin; (c) GM1 micelles and bovine serum albumin from GMl- protein complexes. The column chromatographic sys- ten making use of BioGel A-1.5 m was a satisfactory method (Fig. 5 A, B).

Fig. 4. Sedimentution velocit.vputter-n o f G ~ 1 - i bovine .seruin ulbirmin) complexes. Bovine serum albumin (22 pM) after incubation in 0.05 M sodium phosphate buffer pH 7.5 for 24 h at 20'C, with ganglioside GMl ( 3 mM), was subjected to sedimentation analysis. The photographs were taken 16 min (upper pattern) and 32 min (lower pattern) after reaching a speed of 52000rev.imin. Se- dimentation proceeds from right to left

Control experiments, in which pre-incubation with the protein was omitted, showed that the chromato- graphic behaviour of ganglioside G M ~ was concen- tration-dependent. When ganglioside GMl was dis- solved at concentrations below and up to 0.1 pM, a variable amount of it, especially at the lower ganglio- side concentrations, as monitored by radioactivity measurements was eluted in the column void volume. Comparing the data of several chromatographic runs we found that recovery of ganglioside GM1 was de- pendent on the batches of the gel matrixes used. In most experiments with BioGel A-1.5 m, the amount of ganglioside recovered was approximately 40 ::';; (Fig. 5A), while the remnant ganglioside could not be eluted even by exhaustive washings of the column. The latter phenomenon is probably the result of the inter- facial activity of the amphipathic ganglioside G M ~ , which causes its adhesion to the glass and the resin beads of the chromatographic system. The molecular weight of the micelle of ganglioside GM1, employed in

320 Interaction of Gangliosides with Bovine Serum Albumin

l p I n 2.5

.- D O

Elution volume [ml)

Fig. 5. G'elfiltrutioii elution prof2'le of luhelled GMl und bovine .strum

ulhumin. Chromatography of CrMl and bovine serum albumin on Bio-Gel A-1.5 m was conducted as described under Materials and Methods. (A) 'H labelled gangliocide G,u, 0.1 pM, radioactivity expressed per 0.5 ml. (B) Chromatography of 3tI-labelled ganglio- sidc GH[, 5 pM, and bovine serum albumin, 11 pM. was performcd independently, and their profiles were superimposed in this figure. (e -0) Radioactivity, expressed pcr 0.1 ml; (-- ~ - ) absorbance at 280 nm

5.0

2.5

0

I c

our studies, was found to be 53200 [21] when measured by laser light scattering; consequently, the micellar form of the ganglioside should not be eluted at the void volume of the column, since the excluded limit of BioGel A-1.5 in is approximately 1 . 5 ~ lo6. Thus, the ganglioside peak of Fig.5A. excluded from the column, is not in micellar form. However, at present we are not able to say which physical form of the ganglioside is responsible of this behaviour currently under investigation. This experiment was repeated with the ganglioside predominantly and surely in micellar form at a sample concentration of 10 pM. In this case, most of the radioactivity was recovered as a symme- trical peak with a constant and reproducible elution profile while the excluded peak practically disappeared (Fig. 5 B). The elution volume of the ganglioside peak of Fig.5B provided an estimate of the micellar size of ganglioside GMl. Assuming the micelles to be homogeneous in size, we found an apparent molecular weight of 450000 for GMI niicelles, in good agreement, within the experimental error, with the light scattering measurements. Application of albumin to the same column, in the absence of ganglioside GM, results in elution of the protein at a higher and distinct buffer volume from that of micellar ganglioside (Fig. 5 B).

Fig.6. GelJlfratiow of houinc .serum ulhurnin c o n i p k . ~ n ~ r h Gvl ur .suhmicellur concentration. Chromatography of bovine serum al- bumin, at a final concentration l l pM, pre-incubated at 20 C with 0.1 pM labelled Gul for the indicated periods of time was con- ducted as dcscribed under Materials and Methods. (A) 1 h; (B) 6 h . (C) 24 11. (G 0) Radioactivity, expressed per 0.5 ml; ( ~- ) absorbance at 280 nm

Fig. 6 shows the elution pattern of ganghoside GM1 and bovine serum albumin, in terms of radioac- tivity in dis./min and protein content, when ganglio- side at low concentration (0.1 FM) and albumin (1 1 pM) were mixed together and incubated at 20':C for different periods of time, prior to chromatography. It can be observed that after 1 h of incubation with bovine serum albumin, ganglioside G,w was eluted mostly in the void volume. A small symmetrical peak of radioactivity superimposing that of the protein can, however, also be detected (Fig. 6A). A further increase in the incubation time results in a larger portion of ganglioside GMI eluted with the protein and in a concurrent decrease in the amount of gan- glioside eluted in the excluded volume, After 24 h more than 80 7; of the eluted radioactivity was carried on by the albumin peak. The binding was a relatively slow process but i t must be pointed out that when ganglioside G M I was dissolved directly with the bovine serum albumin solution the interaction between gan-

M. 't'omasi, L. G. Roda. C. Ausicllo. G. D'Apolo. B. Vencriindo, K. Ghidoni, S. Sonnino, and G. Tettainanti 37 I

glioside and protein was a much faster phenomenon. More than 90';~:) of the ganglioside was bound within 20 min of incubation at 20 C. In numerous experi- ments the recovery of radioactivity always rose in the presence of bovine serum albumin. Very likely there is a competition between the protein and the surfaces of the chromatographic system. In the presence of albumin, ganglioside GM,, at the concentrations of the experiments reported in Fig. 5 and 6, distributed itself more on the protein surfaces than on the surface of the column glass and gel beads.

After 24 h of incubation with bovine serum albu- min, the maximum amount of ganglioside bound to the protein was one molecule of GMl per 100 molecules of bovine serum albumin. A similar ganglioside/ protein ratio was also found with longer incubation times. Since this ratio is absolutely below the maximum binding capacity of bovine serum albumin for amphi- pathic substances, the protein-ganglioside peak of Fig. 6C was further incubated with a fresh gdnglioside solution, at a concentration of 0.1 pM, for 24 h at 20 'C. The incubated material was submitted to a new chromatographic run. A peak of protein and gan- glioside with the same GW ,'bovine serum albumin molar ratio was observed in the same region as thc albumin-ganglioside peak in the previous experiment. Under these conditions, there was no further associa- tion of ganglioside GMI to albumin. These results scem to exclude that the interaction of ganglioside GMlr in the physical form occurring at these concentrations, with bovine serum albumin has the characteristics of a saturation phenomenon. Moreover, re-chromatog- raphy on a new column of the albumin-ganglioside peak, after incubation in buffer for 24 h at 20 C, led to a substantial decrease of the amount of radio- activity carried out by the protein and to a con- current decrease in the amount of radioactivity eluted from the column. This indicates a redistribution of ganglioside GM, between the protein and the column system.

Fig. 7 illustrates the elution pattern of ganglioside GMi and bovine serum albumin, monitored as in Fig. 6, when ganglioside, from 5 pM to 2 mM concen- tration, and albumin were mixed together and allowed to stand at 20' C for 24 h prior to chromatography. In these cases no radioactivity was eluted in a reason- able amount either with the excluded volume or in the protein peak. Conversely more than 95 7" of radio- activity was eluted i n two peaks, I and IT, having faster mobility than that of free micells. The same two peaks carried part of the protein, the remainder of which came out in the peak characteristic of the free albumin. By increasing ganglioside concentration the peak of free albumin gradually decreases almost disappearing while the relative proportion between peak I and I1 progressively changes in favour of peak I. These results are in good agreement with those previously observed

l A 0.50

0.2 5

0

0.50

0.25 g 0 a, N

m +

3.25

0

3.53

3.25

3

Elut ion vnliirne (nit)

Fig. I. Gel fi'ltrution qfhovitze scrutn ulhumin comple\-es kvith tnicrllur GMI. Chromatography of bovine serum albumin, at a final concen- tration of 11 pM, pre-incubated at 20 'C for 24 h with various amounts of ganglioside GMI was conducted as described under Materials and Methods. The ganglioside final concentration was: (A) 5 pM; (B) 50 p M ; (C) 0.5 m M ; (D) 2 mM. (0- --a) Radio- activity, expressed per 0.1 ml; (--- ) absorbance at 280 nm

in the ultracentrifuge experiments. Thus, peaks I and IT represent the chromatographic evidence of the formation of two complexes between bovine serum albumin and G M ~ ganglioside, predominantly in m - cellar form at these concentrations.

Complex 1 has a molecular weight of approximately 1 x lo6; complex I1 of 0.5 x lo6. The apparent sizes of the complexes were also confirmed by chromatog- raphy on different media, such as Ultrogel ACA 34, which in subsequent experiments was shown to give better resolution of the molecular species under study.

Interestingly, at a ganglioside concentration be- tween 0.1 and 5 pM, the elution profile showed an intermediate situation between those observed in the experiments of Fig. 6 and 7. At these concentrations, after incubation for 24 h at 20 "C with 11 FM bovine

322 Interaction of Gangliosides with Bovine Serum Albuniln

scruni albumin, all the applied ganglioside was eluted in two peaks, a portion at an elution volume cor- responding to complex I1 of Fig. 7 and the remainder at an elution volume corresponding to the ganglio- side-protein aggregate of Fig. 6. At these concentra- tions two different physical forms of ganglioside G M 1 (the micelle, the monomer and/or a still unknown one) are present, both displaying different binding in measurable amounts with bovine serum albumin.

Tmnsitiorz of' Conzplrxes

I order to be sure that the separation of the two complexes was not an artifact, isolated complex I and I1 were re-chromatographed on the same column independently. Complex I and I1 were eluted as a single peak at the expected buffer volume, verifying the initial separation. However, complex I1 chroina- tographed always with varying amounts of complex I . The possibility that complex I1 was undergoing conversion to complex I was tested by chronological studies in which a fixed starting concentration (50 yM) of ganglioside GML was maintained. With time the peak representing free albuniin decreased while the complex I and 11 peaks increased. Morcover complex I increased slowly at the expence of complex 11. These results suggest, thereforc, that there is a slow irreversible transition of complex I1 material to com- plex I . This was further confirmed by incubating the isolated complex I1 for different times at 20 T. When the incubated complex I1 samples werc re-chromato- graphed on the same gel system, complex I appeared i n amounts proportional to the incubation times (Fig. 8).

In addition it must be pointed out that complex I and 11 always had the same ganglioside/protein ratio. This ratio was essentially constant across the two peaks in all the experimental procedures employed, and cqual to 6.9 (range 6.7-7.1). The complexes are constituted by one bovine serum albumin polypeptide chain and approximately 320 molecules of ganglioside GMI . This result correlatcs well, within the experimental error, with the number of 350 monomers per GMI ganglioside micelle determined by laser light scatter- ing. Thus, the maximal binding of the two complexes was of one ganglioside micelle/protein polypeptide chain. In a few instances, the material eluted between complex 11 and free albumin had, instead, a much higher ganglioside content with respect to protein, thus indicating the presence of free micelles. How- ever, unbound ganglioside micelles disappeared with increasing incubation times.

A few parameters affecting the aggregation of complex 11 to I were also studied. By raising the ionic strength of the ganglioside-protein solutions the formation of complex I is inhibited. Incubation of the mixtures with 2 M NaCl at 20°C supprcsses the

Elution volume (mi)

Fig. X. Conversion of G,fl-hoi~ine serum crlbimziti uitnp/e\- I I itito

conzpler I . Bovinc serum albumin, at a final concentration of 11 pM, after incubation for 4 h at 20 C with 50 pM ganglioside GMl, was chromatographed on Bio-Gel A-1.5 m as described in Fig.7. Fractions, isolated as complex I1 from six parallel experi- ments, were pooled, concentrated without stirring in a Sartorius Membran-filter apparatus and incubated for different times at 20 C. The incubation mixtures were then chromatographed on the same column. Incubation time: (- -) 2 h ; ( ' . .) 96 h. Protein was detected by absorbance at 280 nm

--) 48 h: (.

aggregation of complex I1 to I and the former is the only species found in solution. Opposite results were obtained with ammonium sulfate (100 saturation) treatment. Unbound ganglioside GMl is soluble, under these conditions, while a precipitate containing both protein and ganglioside with the above indicated ratio can be readily isolated. The pellet, dissolved and chro- inatographed a s described, contained mostly complex I with little or no complex 11.

DISCUSSION Experimental evidence from different laboratories

support the hypothesis [32] that gangliosides, typical components of the cell plasma membranes, function as biotransducers of membrane-mediated informa- tion. This hypothesis does implicate specific interac- tions between gangliosides and some membrane pro- teins. At present ganglioside interaction with mem- brane-bound proteins is difficult to investigate at the molecular level. Therefore we considered it useful, as a preliminary approach to the problem, to develop a simple model, making use of ganglioside GM1. the most commonly used species in the studies of func- tional biochemistry of gangliosidcs, and bovine serum albumin.

Since gangliosides are known to undergo micelli- zation in aqueous solution, over a certain concentra- tion and under given conditions [20], at least two physical forms of ganglioside GMi should be consi- dered for their capability to interact with albumin. the monomer and the micelle. GMl, of the same batch used in the present investigation, was definitely shown to be present as micelle already at a concentration

M. Tomasi. L. G. Roda, C. Ausiello, G. D’Agnolo, 8. Vcnerando, R. Ghidoni, S. Sonnino, and G. Tettainanti 323

of 1 pM [21]. When bovine serum albumin is incubated with ganglioside GMl at concentrations well above 1 pM, that is with the ganglioside predominantly in micellar form, two well defined complexes, I and I1 are formed. As shown by the GM1-protein molar ratio, which is Lhe same in the two complexes and the apparent molecular weights found in the experiments of Fig.6 and 7, complex I1 is constituted by one gan- glioside micelle and one polypeptide chain, while com- plex I is a dinier of complex 11. The complex which arises first is complex I1 and its formation is dependent upon time and ganglioside concentration. Complex I , once formed, displays strong stability, is not converted to other forms and does not release the bound gan- glioside. With regards to the mode of dimerization it is well known that bovine serum albumin undergoes spontaneous polymerization to dimers and higher aggregates upon aging [33]. However, no polymeriza- tion of albumin was found to occur under our ex- perimental conditions. Moreover, when bovine serum albumin dimers, isolated during the purification step of the protein, were incubated with GMl micelles, evidence for the formation of a ganglioside-protein complex was obtained from the molecular sieving experiments with a GMl-protein molar ratio of 3.2- 3.6, one half that recorded using albumin monomers. Thus, the single complex obtained in this case is con- stituted by one ganglioside micelle and one protein dimer molecule. On the basis of this data, the most likely explanation for the observed behaviour of the two complexes is to assume that there is a polymeriza- tion of the entire complex I1 to a dimer complex I, and that complex I is not the result of the interaction between albumin aggregates and single gdnglioside micelles.

The primary driving force for the transition of complex IT to complex I appears to be the result of weak electrostatic interactions. Neutral salts might be expected to inhibit the formation of complex I by weakening attractive charge-charge interactions between exposed regions of the albumin polypeptide chain. This assumption is corroborated by the am- monium sulfate precipitation of complex I. By lower- ing water activity, ammonium sulfate, at saturation, enhances inter-molecular and intra-molecular electro- static interactions [34], inducing the transition of complex I1 to complex I.

Once the albumin has become incorporated into the ganglioside micelle, the resulting protein-amphi- phile interactions leads to a structural alteration of albumin. This process would preferentially expose to the solvent protein regions containing ionic groups and providing the contact area for the formation of the dimer complex 1.

Binding with ganglioside micelle induced detect- able alterations in the ultraviolet absorption and fluo- rescence spectra of bovine serum albumin. The spectra

have features indicative of differences in the struc- ture of the protein in the neighbourhood of some of its aromatic amino acids [30,34]. On the other hand, these alterations of the albumin structure enable the protein to bind ganglioside GMl and are consistent with the hypothesis that the basic interaction between G M I micelles and bovine serum albumin is hydro- phobic, i.c. that the micellar GMl-protein complexes are actually mixed ganglioside-protein micelles. From this point of view the interaction between G M l micelles and bovine serum albumin resembles that described for other amphiphiles, like Triton X-100, sodium do- decylsulphate and bile salts [35,36].

The interaction between bovine serum albumin and G M l monomers can be studied provided that the ganglioside is present only in its monomeric form. This is difficult to achieve since the ganglioside con- centration and, probably, the exact environmental conditions at which ganglioside aggregation occurs are still discussed and not precisely known [15,20, 37,381. What we can say is that in our chromato- graphic system GMl micelles moved as a symmetrical peak, intermediate between those of free albumin and complex 11, and that this peak completely disappeared when the ganglioside concentration was 0.1 pM. At this same concentration, however, all the ganglioside which can be eluted from the column migrates with the excluded peak, while a significant amount of gan- glioside is absorbed on the column glass and gel beads. Assuming that at this concentration the physical form of ganglioside G M I in solution is the monomer and that the excluded peak could be the result of if the interaction of the ganglioside with some components of the chromatographic system under these conditions, GM1 binds to the protein with an apparent maximum binding ratio of 1 ganglioside molecule per 100 albumin polypeptide chains. The binding is a slow (markedly slower than that with the micelles) and quite aspecific process. In fact the features of a saturation phenomc- non could not be obtained and ganglioside easily leaves the albumin surface for the surfaces of the column components. Also this process did not appear to reflect the general ability of bovine serum albumin to bind single amphipathic molecules since, if so, we should expect higher binding velocity, greater com- plex stability and stoichiometry. The binding, or, better, adhesion, of ganglioside monomers to albumin was not able to affect the protein structure, at least as revealed by the monitoring technique employed, which, conversely gave clear signals with the albumin- GMI micelle complexes. However, it should be re- membered that the lowest ligand concentration (0.1 pM) that could be usefully employed in our ex- periments might actually be too small for eomplexing a significant fraction of protein molecules ; therefore a possible perturbation of the albumin structure in this complex could not be detected by optical tech-

3 24 M. Tomasi el a].: Interaction of Gangliosides with Bovine Serum Albumin

niques. It is evident from our results that the inter- action between bovine serum albumin and GMl mono- mers, or any other physical form of aggregation which is not true, stable micelles, cannot be suggested to play an intermediate role in the formation of the protein-GMl micelle complexes. These should be viewed as a direct interaction between albumin and the ganglioside micelle, similar to those reported for the interactions of other amphiphiles [35,36].

In conclusion the reported results allow us to suggest that the most interesting GMl-albumin inler- action is that leading to complex 11, which is the result of specific, mostly hydrophobic interaction, between GM1 micelle and albumin. In fact binding with ganglioside micellcs induced detectable altera- tions in the ultraviolet absorption and fluorescence spectra of bovine serum albumin. The spectra have features indicative of differences in the structure of the protein in the neighbourhood of some of its aromatic amino acids [30,39]. These alterations of the albumin structure enable the protein to bind ganglioside G M I and are consistent with the hypo- thesis that the basic interaction between GMt micelles and bovine serum albumin is hydrophobic, i.e. that the micellar GMl -protein complexes are actually mixed ganglioside-protein micelles. A further, peculiar, con- forinational change of the protein after binding to the amphiphile brings complex 11 to give, by dimerization, complex I. Additional work on the characteristics of ganglioside GMi (and other gangliosides) in mixed niicelles with bovine serum albumin, and on the kinetics of their formation, is necessary. This would complete the study of our model and facilitate the investigation of the very complex ganglioside-protein interactions occurring at the membrane level, after binding of proteins to ganglioside receptors, or during the membrane lateral phase separation of both protein and ganglioside components [2- 1 I].

REFERENCES

1. Klenk. E. (1942) Hoi,/,e-~Tc.).l.r'.s Z. Pii.vsiol. Chrn7. 273, 76-86. 2. Holmgren, J. , Lonnroth, I. & Svennerholm, L. (1973) Inff,c/.

3. van Heyningen, W. E. (1974) Nuture (L,ond.) 24Y, 415-417. 4. Burger, M. M. & Goldberg, A. R. (1967) Proc. Nut/ Acud. Sci.

5. Maget-Dana, R., Roche. A.-C. & Monsigny, M. (1977) FEBS

Inimun. 8, 208 - 21 4.

U S A , 57, 359- 366.

L.e// . 7Y, 305-309.

6. Aloj, S. M., Kohn, L. D., Lee, G. & Meldolesi, M. F. (1977)

7. Kohn, L. D., Friedman, R. M., Holmes, J . M. & Lee, G. (1976)

8. Hakomori, S.-I. (1975) Biochim. Biopliys. Ac,/o. 417. 55-89. 9. Keenan, T. W., Schmidt, E., Franke. W. W. & Wiegandt, H.

10. Rt-vesz, T. &Greaves, M . (1975) Nature (Lond.) 257.103- 106. 11. Surolia, A,, Bachwat, B. K. & Podder. S. K. (1975) Nutiire

12. Yohe, H. C., Roark, D. E. & Rosemberg, A. (1976) J . Biol.

13. I-Iansson, H.-A., Holmgren, J. & Svennerholm, L. (1977) Proc.

14. Hagashi, K. & Katagiri, A. (1974) Biudiinz. Biop1iy.c. Acttr

15. Formisano, S . , Johnson, M . L., Lee, G., Aloj, S. M . & Edel-

16. Tettanmiiti,G., Bonali,F., Marchesini,S. & Z a m b o ~ i , V . (1973)

17. Ghidoni, R. Sonnino, S., Tettamanti, G., Wiegandt. H. & Zam-

18. Sonnino, S., Ghidoni, R., Galli, G. & Tettamanti, G. (1978)

19. Carter, M. E. & Gaver, R. C. (1967) J . Lipid Rrs. 8. 391 -395. 20. Gammack, D. B. (1963) Biocliem. J . 88, 373-383. 21. Corti, M. , Degiorgio, V.. Ghidoni, R., Sonnino. S. & Tetts-

22. Suzuki, Y. & Suzuki, K . (1972) J . Lipid Res. 13. 687-690. 23. Ghidoni, R., Tettamanti, G. & Zambot~i . V. (1974) Iiul. J .

24. Davis, B. J . (1964) Ann. N. Y . Acud. Sci. 121. 404-427. 25. Laemmli, U . K. (1970) Nuture (Lond.) 227,680-685. 26. Svennerholm, L. (1957) Biochim. Biophys. Acta. 24. 604-61 1. 27. Munkres, K . D. & Richards, F. M. (1965) Arch. Biochem.

28. Lowry, 0. H., Rosebrough, N. I., Farr. A. L. & Randall. R . J .

29. Hill, M . W. & Lester, R. (1972) Biochim. Biophys. dcfcr. 2x2.

30. Steinhardt. J., Leydy, J. & Mooney, J . P. (1972) Biochrmi.s/r~.

31. Plowman. K . M. (1972) h z ~ ~ n z c ~ Kinetic,.s. McGraw-Hill. N e b

32. Brady, R. 0. & Fishman, P. H. (1979) Adv. Enzjmol. 50.

33. Nikkcl, H. J. & Foster, J . F. (1971) Biochenris/rv. 10, 4479- 4486.

34. von Hippel, P. H. & Schleich. T. (1969) in Strur,firrc.iii?t/.Ttcrhi/iti, cf Biologicd Macromolecules (Timasheff, S. N . & Fasman. G. D., eds) pp. 417-574, Marcel Dekker, New York.

Biochim. Riophys. Res. Commnn. 74, 1053 - 1059.

Proc. Null Acud. Sci. U S A , 73. 3695- 3699.

(1975) E.x/I. Crll. R ~ . s . Y2, 259-270.

(Land.) 257, 802 - 804.

Chum. 251,7083 - 7087.

Nuti Acad. Sci. USA 74, 3782 - 3786.

337, 107-117.

hoch, H. (1979) Biochemistr.y, I X , 1119-1124.

Biochim. Biophys. Actn, 296, 160- 170.

botti, V. (1976) J . Neurochem. 27, 511 -515.

J . Neurochem. 31, 947-956.

manti,G. (1980) Chem. Pliys. L.ipirf.s, 26, 225-238.

Biochem. 23, 320- 328.

Biophys. 109, 446- 479.

(1951) J . Biol. Chem. 193, 265-275.

18-30.

11, 1809-1817.

York.

303 - 324.

35. Borgstrom, B. (1975) J . Lipid Res. 16, 41 1-417. 36. Reynolds, J . A. & Tanford, C. (1970) Proc. Nut1 Acutl. Sci.

37. Rauvdla, H. (1979) Eur. J . Biochmz. Y7, 555-564. 38. Mrar, W., Schwdrzmann, G., Sattler, J., Seeman, B.. Momoi,

T. & Wiegandt, H. (1 979) Glycoconjugatrs, Proeredings i$

the 5/11 In/emationnl S.ymposium, pp. 1 12 - 1 13, Georg Thie- me, Stuttgart.

39. Vincent, J.-P., Balerna, M. & Lazdunski, M. (1978) F E B S Let/ . 85, 103- 108.

USA, 66, 1002-1010.

M. Tomasi, C. Ausiello and G. D'Agnolo, Laboratorio di Biologia Cellulare e Immunologia, lstituto Superiore di Srzniti, Vide Regina Elena 299, 1-00161 Roma, Italy

L. G. Roda, Laboratorio di Farmacologia, Istituto di Medicina Sperimentale e Clinica, Universiti degli Studi di Ancona, Via delle Grotte di Posatora 2, 1-60100 Ancona, Italy

B. Venerando, R . Ghidoni, S. Sonnino and G. Tettamanti, Istituto di Chimica Biologica, Facolti di Medicina e Chirurgia. Universiti degli Studi di Milano, Via Saldini 50, 1-20133 Milano, Italy