FEMS MicrobiologyImmunology105 (1992) 5-12 © 1992 Federationof European MicrobiologicalSocieties0920-8534/92/$05.00 Published by Elsevier
FEMSIM00225
Oligomer formation of staphylococcal a-toxin analyzed by electron microscopy and image processing Hans Hebert
a,
Anders Olofsson
a,
Monica Thelestam b and Elisabeth Skriver c
, Center for Structural Biochemistry, Karolinska Institutet, NOVUM, Huddinge, Sweden, b Department of Bacteriology, Karolinska Institutet, Stockholm, Sweden, and c Department of Cell Biology at the Institute of Anatomy and the Biomembrane Research Center, University of Aarhus, Aarhus, Denmark
Key words: S t a p h y l o c o c c u s aureus a-toxin; Pore-forming toxin; Electron microscopy; Image processing; Freeze-fracture
1. SUMMARY
smaller particle is consistent with the presence of a penetrating pore in this structure.
The 12S oligomeric form of Staphylococcus aureus a-toxin has been studied with electron microscopy after incubation of the toxin with membrane preparations or liposomes. The target material originated from human platelets. Different electron microscopic preparation techniques were used including negative staining, freezefracture and vitrification in liquid ethane. Analysis of micrographs with image processing methods revealed two groups of ring-like structures corresponding to a-toxin oligomers. One form measured 75 ,~ in diameter and had a high stain density in the central protein deficient part while the other was larger with a diameter of 100 A and less stain accumulation in the center. The conditions under which the latter were formed suggest that this correponds to an inactive loosely-bound form of the toxin. The high stain density in the
to: H. Hebert, Center for Structural Biochemistry, Karolinska Institutet, NOVUM, S-141 57 Huddinge, Sweden. Correspondence
2. INTRODUCTION The ability of Staphylococcus aureus a-toxin to cause damage to cell membranes is thought to be coupled to an oligomerisation of the protein from a 3S form to a hexameric 12S form. Structural and functional studies suggest that this oligomeric structure defines a membrane pore through which ions and small molecules can diffuse, a-Toxin has been classified as belonging to a family of amphipathic proteins called 'channel forming' or 'pore forming' proteins [1]. This group contains not only bacterial toxins but also proteins of eucaryotic origin such as complement and the lymphocytic pefforins. The amino acid sequence of the 33-kDa a-toxin protein has been determined by Gray and Kehoe [2]. Spectroscopic analyses suggest a secondary structure with a high/3-sheet content [3,4]. Early electron-microscope studies of membranes or liposomes exposed to a-toxin revealed annular structures with a central protein deficient part
3. MATERIAL AND METHODS
[5,6]. Tetragonal two-dimensional crystals of the toxin induced on lipid layers have recently been used to reconstruct a three-dimensional model of the protein at 25 A resolution [7]. In this model a deep cavity rather than a completely penetrating pore is delineated by the stain in the center of the structure. Different electron microscopic methods were used in the present work, in combination with image processing, to further analyse a-toxin interaction with membranes or lipids. It is of particular interest to compare different structures since oligomers of a-toxin which are not membranedamaging can be formed on membranes under certain conditions, such as at low temperature or in the presence of Ca2+-ions [8].
3.1. a-Toxin purification
o
Purification of a-toxin from Staphylococcus aureus strain Wood 46 was made by isoelectric focusing and gel filtration as described previously [9]. The specific activity of the purified toxin preparations was 30 000 hemolytic U / m g protein. Human platelets were obtained by mixing fresh blood with 3.8% sodium citrate (4 : 1) followed by centrifugation for 15 min at 150 x g and 22°C. The supernatant was further centrifuged at 1800 x g, 22°C for 30 min producing a pellet of platelets which was mixed with Tris-buffered saline, pH 7.0 (TBS). The latter centrifugation step was repeated. One part of the platelet sus-
Fig. 1. Liposomes induced by iipids from human platelets and subsequently incubated with Staphylococcus aureus a-toxin. Bar: 100 rim.
ing human and rabbit red blood ceils and cultured mouse adrenocortical Y1 cells.
3.2. Lipid extraction
Fig. 2. Correlation averaged projection map of a-toxin oligomers randomlydistributed on cell membranes. Bar: 20 ,~.
Lipids were extracted from platelet membranes by dissolving the platelet suspension in c h l o r o f o r m / m e t h a n o l (1:1.2, v/v). The lower phase, obtained after centrifugation at 3000 X g for 1-2 min, was withdrawn and brought to dryness under a stream of nitrogen. The residual lipid was suspended in TBS at 1 m g / m l . These preparations were used either for direct incubation with a-toxin at the same conditions used for the platelet suspensions or for preformation of lipid layers on carbon coated grids as described [7]. The toxin-treated membrane or lipid preparations were applied to carbon coated grids and negative staining was performed with a 1% (w/v) solution of Na-PTA, pH 7.2.
3.3. Freeze-fracture extraction pension was incubated with two parts of a-toxin (1.15 m g / m l ) for 30 min to several days. Initital tests were also made with other cell types includ-
For freeze-fracture the platelet suspension was incubated with a-toxin (0.15 mg) for 36 h at room temperature and then pelleted in a Beckman airfuge for 10 min and resuspended in 5 /zl of
Fig. 3. Formation of hexagonal and tetragonal crystals after incubationof a-toxin with human platelets. Bar: 100 nm.
Reichert-Jung H F 80 Cryofixation System. The specimens were fractured at - 1 0 0 ° C and immediately shadowed with platinum either at an angle of 45 ° without rotation or at an angle of 25 ° with rotation and replicated with carbon in a Balzer BAF 300 freeze-fracture apparatus. The replicas were cleaned in sodium hypochlorite before mounting on grids. Blood platelets without atoxin were processed in the same way. Frozen hydrated specimens were obtained after application of specimens to naked hexagonal 600-mesh grids. Vitrification of the suspensions was m a d e in liquid ethane and the grids were transferred to a G a t a n cryo-specimen holder.
3. 4. Electron microscopy Fig. 4. Correlation averaged projection map from a hexagonal crystalline domain. Bar: 20 ,~.
TBS. Specimens were mounted between two copper plates and rapid-frozen by plunging into liquid propane cooled by liquid nitrogen with a
Electron microscopy was performed under controlled dose conditions with a Philips 301, Philips 420 or a Jeol 100CX microscope. Optical density measurements of selected micrographs were made with an Eikonix 1412 scanner at 0.02 m m spot size. The electron optical magnifications were controlled with orthorombic catalase crystals.
Fig. 5. Tetragonal crystals formed after incubating a-toxin with human platelets. Bar: 100 nm.
3.5. Software The software used for the image analysis was the 'EM' program system [10] implemented on VAX computers and a SUN Spare2 work station. Correlation averaging, essentially as described by Saxton and Baumeister [11], was used to produce projection maps of the structures. In one case, the tetragonal crystals described below, a rigorous treatment was made including phase restoration and lattice straightening on a large dataset as described by Olofsson et al. [12].
4. RESULTS Among the cell types investigated the human platelets showed the highest density of the characteristic ring structures seen after incubation with a-toxin. Liposomes induced with lipids extracted from the platelet membranes and mixed with a-toxin produced an even higher number of annular structures (Fig. 1). Controls made from membranes or lipid vesicles not treated with the toxin did not show such structures. Single particle correlation averaging of the ring structures showed a protein domain with six peaks
Fig. 6. Correlation averaged projection map merg.ed from a set of tetragonal crystalline areas. Bar: 20 A.
(Fig. 2) [13]. The diameter of the annulus was pH dependent but at neutral pH it was 75 ,~. Decreasing the pH increased the size until a complete collapse occurred at pH 4. The stain-filled pore or cavity in the center had a diameter of about 25 ,~. The density of stain was similar in regions surrounding the oligomer and in the center. Two different types of two-dimensional crystals of a-toxin could be observed after incubation with platelet suspensions or liposomes (Figs. 3, 5). The symmetry of one type was hexagonal while the other was tetragonal. As can be seen in Fig. 3, the two types could be observed simultaneously from the same preparation and even on the same grid. The unit cell dimension of the hexagonal type of crystal was 87 ,~ for both platelet membranes and liposomes. The tetragonal unit cell parameter showed a constant value of 109 ,~ among the different preparations. The tetragonal crystals were more frequently observed and they also formed on carbon coated grids covered with lipid as described [7]. The projection structures obtained after correlation averaging of the two crystal types showed pertinent differences (Figs. 4, 6). The diameter of the annulus in the hexagonal crystals was 75 ~, while the tetragonal c~stals showed larger rings with a diameter of 100 A. The stain density in the center of the structures was higher in the hexagonal rings as compared to the tetragonal ones. The substructure of the protein in the tetragonal unit cell could be described as consisting of four major symmetry-related domains each containing two peak regions. The hexagonal form on the other hand did not show any significant protein density variation in the ring. The tetragonal crystals were also investigated from freeze-fractured and frozen hydrated specimens (Figs. 7, 8). A common observation was that the unit cell was shorter, 93 and 94 /~ respectively, in these preparations as compared to the negatively stained specimens. In freeze-fracture the tetragonal crystals were seen on both concave and convex fracture faces and consisted of short intramembrane particles. No complementary impressions were observed. At certain areas the membrane showed a tendency of stacking as also
10
Fig. 7. Tetragonal crystalline area after incubating a-toxin with human platelets following freeze-fracture and rotary shadowing at an angle of 25°. The inset shows a 125 × 125 ,~2 cutout of the correlation-averaged structure. For comparison with the negatively stained specimens, the contrast has been reversed so that the light areas correspond to protein. Bar: 100 nm.
Fig. 8. Frozen hydrated unstained specimen of a-toxin incubated with human platelets containing a tetragonal crystalline domain (arrow) and the corresponding correlation averaged projection map. For comparison with the negatively stained specimens, the contrast has been reversed in the map so that the light areas correspond to protein. Bars: 100 nm (micrograph), 2 nm (map).
11 obse~ed by negative staining and thin sectioning [7]. The height of the steps was 150-180 A, corresponding to two layers of membranes. Blood platelets without a-toxin had only few intramembrane particles on both P- and E-faces, as observed earlier [14]. The form of the particles from freeze-fracture and frozen hydrated specimens was slightly asymmetric, mainly due to crystal tilt and had diameters of about 70 A. o
5. DISCUSSION In the present work it is demonstrated that two different types of annular structures can be observed in specimens containing Staphylococcus aureus a-toxin together with platelet membranes or liposomes. Since these were not observed in controls without the toxin, we conclude that they correspond to different oligomeric forms of the protein. Analysis of the toxin alone in the watersoluble form did not show any formation of rings. One of the oligomeric types, observed as single non-crystalline particles or in hexagonal.two-dimensional crystals, has a diameter of 75 A and a high stain density in the central pore or cavity. The average structure from the single particles clearly shows six identical protein domains while the projection from the hexagonal crystal only shows small variation in the density within the protein domain. This can be explained by rotational variance of the oligomers in the hexagonal crystal. In that case a conventional crystallographic approach will result in a cylindrical average. The tetragonal two-dimensional c~stals show oligomers with a larger diameter, 100 A and little stain penetration in the center. We have recently reconstructed a three-dimensional model from this form which demonstrates that the negative stain does not delineate a penetrating pore in the center but rather a deep cavity [7]. The annular protein structure is composed of four symmetryrelated domains. We have now confirmed that the negative stain used, Na-PTA at pH 7.2, does not influence the formation of the tetragonal crystals since both freeze-fracture and the frozen
hydrated specimens show crystals with similar properties. We have recently investigated the kinetics of staphylococcal a-toxin oligomerisation in relation to membrane permeabilisation using different biochemical and biophysical approaches [8]. The results show that a-toxin oligomers can be present under conditions where no membrane damage can be detected. This is the case, for instance, at low temperature and in the presence of Ca2+-ions. This suggests that a conformational change is likely to occur to form the fully active form of the protein after membrane association. Such a conformational change has also been predicted based on spectroscopic measurements [4]. The results presented in this work give additional support to the suggestion that the toxin changes during transfer from membrane binding to membrane permeabilisation. The form observed in the tetragonal crystals may correspond to the state of the protein which binds to the membrane. This oligomer does not contain a penetrating pore and it can be found also on grids with a preformed lipid layer where no extensive membrane penetration can occur. An inactive trypsin-treated form of the toxin has recently been shown to arrange in tetragonal crystals with a very similar protein domain structure [12]. Moreover, the freeze-fracture electron microscopy of tetragonal crystals revealed one type of crystals and no complementary impressions. We therefore suggest that this oligomer form is attached to the outer leaflet of the lipid bila.yer. The relatively small size of the particle, 70 A as compared to 100 A in negative stain, is consistent with the structure of the 3-D model [7], which can be described as cupshaped with its compressed bottom attached to the lipid bilayer. The particles in the crystals are only slightly elevated above the fracture plane indicating that the oligomer only extends a short distance into the inner leaflet of the bilayer. This is in accordance with side views in negative stain of a-toxin attached to membranes [15]. The other type of a-toxin oligomer, found either as single particles or in hexagonal crystals, with smaller diameter and higher stain density in
12
the center probably corresponds to the permeabilising form. The asymmetric protein unit has a smaller projected surface in the plane of the membrane consistent with a conformational change from a folded state to a more extended structure. Unfortunately, large well-ordered hexagonal crystals have not yet been frequent enough to form the basis for a 3-D analysis.
ACKNOWLEDGEMENTS This study was supported by Grants Nos. 03X00144 and 16X-05969 from the Swedish Medical Research Council as well as grants from the funds of the Karolinska Institute. Cryo-electron microscopy of frozen hydrated specimens were performed at the Max-Planck-Institut fiir Biochemie, Martinsried, FRG, during a uisit made possible through a fellowship from the Alexander von Humbolt Stiftung (HH).
REFERENCES [1] Bhakdi, S. and Tranum-Jensen, J. (1987) Rev. Physiol. Biochem. Pharmacol. 107, 147-223.
[2] Gray, G.S. and Kehoe, M. (1984) Infect. Immun. 46, 615-618. [3] Ikigai, H. and Nakae, T. (1987) J. Biol. Chem. 262, 2156-2160. [4] Tobkes, N., Wallace, B.A. and Bayley, H. (1985) Biochemistry 24, 1915-1920. [5] Arbuthnott, J.P., Freer, J.H. and Bernheimer, A.W. (1967) J. Bacteriol. 94, 1170-1177. [6] Freer, J.H., Arbuthnott, J.P. and Bernheimer, A.W. (1968) J. Bacteriol. 95, 1153-1168. [7] Olofsson, A., Kav6us, U., Hacksell, I., Thelestam, M. and Hebert, H. (1990) J. Mol. Biol. 214, 299-306. [8] Thelestam, M., Olofsson, A., Blomqvist, L. and Hebert, H. (1991) Biochim. Biophys. Acta 1062, 245-254. [9] Thelestam, M. (1983) Biochim. Biophys. Aeta. 762, 481488. [10] Hegerl, R. and Altbauer, A. (1982) Ultramicroscopy 9, 109-115. [11] Saxton, W.O. and Baumeister, W. (1982) J. Microsc. 127, 127-138. [12] Olofsson, A., Hebert, H. and Thelestam, M. (1991) J. Struct. Biol. 106, 199-204. [13] Olofsson A., Kav6us, U., Thelestam, M. and Hebert, H. (1988) J. Ultrastruc. Mol. Struct. Res. 100, 194-200. [14] Feagler, J.R., Tillack, T.W., Chaplin, D.D. and Majerus, P.W. (1974) J. Cell Biol. 60, 541-553. [15] Tranum-Jensen, J. and Bhakdi, S. (1987) Electron Microscopy of Proteins, Vol. 6 (Harris, J.R. and Horne, R.W., Eds), pp. 445-492, Academic Press Inc., London.
FEMSIM00225
Oligomer formation of staphylococcal a-toxin analyzed by electron microscopy and image processing Hans Hebert
a,
Anders Olofsson
a,
Monica Thelestam b and Elisabeth Skriver c
, Center for Structural Biochemistry, Karolinska Institutet, NOVUM, Huddinge, Sweden, b Department of Bacteriology, Karolinska Institutet, Stockholm, Sweden, and c Department of Cell Biology at the Institute of Anatomy and the Biomembrane Research Center, University of Aarhus, Aarhus, Denmark
Key words: S t a p h y l o c o c c u s aureus a-toxin; Pore-forming toxin; Electron microscopy; Image processing; Freeze-fracture
1. SUMMARY
smaller particle is consistent with the presence of a penetrating pore in this structure.
The 12S oligomeric form of Staphylococcus aureus a-toxin has been studied with electron microscopy after incubation of the toxin with membrane preparations or liposomes. The target material originated from human platelets. Different electron microscopic preparation techniques were used including negative staining, freezefracture and vitrification in liquid ethane. Analysis of micrographs with image processing methods revealed two groups of ring-like structures corresponding to a-toxin oligomers. One form measured 75 ,~ in diameter and had a high stain density in the central protein deficient part while the other was larger with a diameter of 100 A and less stain accumulation in the center. The conditions under which the latter were formed suggest that this correponds to an inactive loosely-bound form of the toxin. The high stain density in the
to: H. Hebert, Center for Structural Biochemistry, Karolinska Institutet, NOVUM, S-141 57 Huddinge, Sweden. Correspondence
2. INTRODUCTION The ability of Staphylococcus aureus a-toxin to cause damage to cell membranes is thought to be coupled to an oligomerisation of the protein from a 3S form to a hexameric 12S form. Structural and functional studies suggest that this oligomeric structure defines a membrane pore through which ions and small molecules can diffuse, a-Toxin has been classified as belonging to a family of amphipathic proteins called 'channel forming' or 'pore forming' proteins [1]. This group contains not only bacterial toxins but also proteins of eucaryotic origin such as complement and the lymphocytic pefforins. The amino acid sequence of the 33-kDa a-toxin protein has been determined by Gray and Kehoe [2]. Spectroscopic analyses suggest a secondary structure with a high/3-sheet content [3,4]. Early electron-microscope studies of membranes or liposomes exposed to a-toxin revealed annular structures with a central protein deficient part
3. MATERIAL AND METHODS
[5,6]. Tetragonal two-dimensional crystals of the toxin induced on lipid layers have recently been used to reconstruct a three-dimensional model of the protein at 25 A resolution [7]. In this model a deep cavity rather than a completely penetrating pore is delineated by the stain in the center of the structure. Different electron microscopic methods were used in the present work, in combination with image processing, to further analyse a-toxin interaction with membranes or lipids. It is of particular interest to compare different structures since oligomers of a-toxin which are not membranedamaging can be formed on membranes under certain conditions, such as at low temperature or in the presence of Ca2+-ions [8].
3.1. a-Toxin purification
o
Purification of a-toxin from Staphylococcus aureus strain Wood 46 was made by isoelectric focusing and gel filtration as described previously [9]. The specific activity of the purified toxin preparations was 30 000 hemolytic U / m g protein. Human platelets were obtained by mixing fresh blood with 3.8% sodium citrate (4 : 1) followed by centrifugation for 15 min at 150 x g and 22°C. The supernatant was further centrifuged at 1800 x g, 22°C for 30 min producing a pellet of platelets which was mixed with Tris-buffered saline, pH 7.0 (TBS). The latter centrifugation step was repeated. One part of the platelet sus-
Fig. 1. Liposomes induced by iipids from human platelets and subsequently incubated with Staphylococcus aureus a-toxin. Bar: 100 rim.
ing human and rabbit red blood ceils and cultured mouse adrenocortical Y1 cells.
3.2. Lipid extraction
Fig. 2. Correlation averaged projection map of a-toxin oligomers randomlydistributed on cell membranes. Bar: 20 ,~.
Lipids were extracted from platelet membranes by dissolving the platelet suspension in c h l o r o f o r m / m e t h a n o l (1:1.2, v/v). The lower phase, obtained after centrifugation at 3000 X g for 1-2 min, was withdrawn and brought to dryness under a stream of nitrogen. The residual lipid was suspended in TBS at 1 m g / m l . These preparations were used either for direct incubation with a-toxin at the same conditions used for the platelet suspensions or for preformation of lipid layers on carbon coated grids as described [7]. The toxin-treated membrane or lipid preparations were applied to carbon coated grids and negative staining was performed with a 1% (w/v) solution of Na-PTA, pH 7.2.
3.3. Freeze-fracture extraction pension was incubated with two parts of a-toxin (1.15 m g / m l ) for 30 min to several days. Initital tests were also made with other cell types includ-
For freeze-fracture the platelet suspension was incubated with a-toxin (0.15 mg) for 36 h at room temperature and then pelleted in a Beckman airfuge for 10 min and resuspended in 5 /zl of
Fig. 3. Formation of hexagonal and tetragonal crystals after incubationof a-toxin with human platelets. Bar: 100 nm.
Reichert-Jung H F 80 Cryofixation System. The specimens were fractured at - 1 0 0 ° C and immediately shadowed with platinum either at an angle of 45 ° without rotation or at an angle of 25 ° with rotation and replicated with carbon in a Balzer BAF 300 freeze-fracture apparatus. The replicas were cleaned in sodium hypochlorite before mounting on grids. Blood platelets without atoxin were processed in the same way. Frozen hydrated specimens were obtained after application of specimens to naked hexagonal 600-mesh grids. Vitrification of the suspensions was m a d e in liquid ethane and the grids were transferred to a G a t a n cryo-specimen holder.
3. 4. Electron microscopy Fig. 4. Correlation averaged projection map from a hexagonal crystalline domain. Bar: 20 ,~.
TBS. Specimens were mounted between two copper plates and rapid-frozen by plunging into liquid propane cooled by liquid nitrogen with a
Electron microscopy was performed under controlled dose conditions with a Philips 301, Philips 420 or a Jeol 100CX microscope. Optical density measurements of selected micrographs were made with an Eikonix 1412 scanner at 0.02 m m spot size. The electron optical magnifications were controlled with orthorombic catalase crystals.
Fig. 5. Tetragonal crystals formed after incubating a-toxin with human platelets. Bar: 100 nm.
3.5. Software The software used for the image analysis was the 'EM' program system [10] implemented on VAX computers and a SUN Spare2 work station. Correlation averaging, essentially as described by Saxton and Baumeister [11], was used to produce projection maps of the structures. In one case, the tetragonal crystals described below, a rigorous treatment was made including phase restoration and lattice straightening on a large dataset as described by Olofsson et al. [12].
4. RESULTS Among the cell types investigated the human platelets showed the highest density of the characteristic ring structures seen after incubation with a-toxin. Liposomes induced with lipids extracted from the platelet membranes and mixed with a-toxin produced an even higher number of annular structures (Fig. 1). Controls made from membranes or lipid vesicles not treated with the toxin did not show such structures. Single particle correlation averaging of the ring structures showed a protein domain with six peaks
Fig. 6. Correlation averaged projection map merg.ed from a set of tetragonal crystalline areas. Bar: 20 A.
(Fig. 2) [13]. The diameter of the annulus was pH dependent but at neutral pH it was 75 ,~. Decreasing the pH increased the size until a complete collapse occurred at pH 4. The stain-filled pore or cavity in the center had a diameter of about 25 ,~. The density of stain was similar in regions surrounding the oligomer and in the center. Two different types of two-dimensional crystals of a-toxin could be observed after incubation with platelet suspensions or liposomes (Figs. 3, 5). The symmetry of one type was hexagonal while the other was tetragonal. As can be seen in Fig. 3, the two types could be observed simultaneously from the same preparation and even on the same grid. The unit cell dimension of the hexagonal type of crystal was 87 ,~ for both platelet membranes and liposomes. The tetragonal unit cell parameter showed a constant value of 109 ,~ among the different preparations. The tetragonal crystals were more frequently observed and they also formed on carbon coated grids covered with lipid as described [7]. The projection structures obtained after correlation averaging of the two crystal types showed pertinent differences (Figs. 4, 6). The diameter of the annulus in the hexagonal crystals was 75 ~, while the tetragonal c~stals showed larger rings with a diameter of 100 A. The stain density in the center of the structures was higher in the hexagonal rings as compared to the tetragonal ones. The substructure of the protein in the tetragonal unit cell could be described as consisting of four major symmetry-related domains each containing two peak regions. The hexagonal form on the other hand did not show any significant protein density variation in the ring. The tetragonal crystals were also investigated from freeze-fractured and frozen hydrated specimens (Figs. 7, 8). A common observation was that the unit cell was shorter, 93 and 94 /~ respectively, in these preparations as compared to the negatively stained specimens. In freeze-fracture the tetragonal crystals were seen on both concave and convex fracture faces and consisted of short intramembrane particles. No complementary impressions were observed. At certain areas the membrane showed a tendency of stacking as also
10
Fig. 7. Tetragonal crystalline area after incubating a-toxin with human platelets following freeze-fracture and rotary shadowing at an angle of 25°. The inset shows a 125 × 125 ,~2 cutout of the correlation-averaged structure. For comparison with the negatively stained specimens, the contrast has been reversed so that the light areas correspond to protein. Bar: 100 nm.
Fig. 8. Frozen hydrated unstained specimen of a-toxin incubated with human platelets containing a tetragonal crystalline domain (arrow) and the corresponding correlation averaged projection map. For comparison with the negatively stained specimens, the contrast has been reversed in the map so that the light areas correspond to protein. Bars: 100 nm (micrograph), 2 nm (map).
11 obse~ed by negative staining and thin sectioning [7]. The height of the steps was 150-180 A, corresponding to two layers of membranes. Blood platelets without a-toxin had only few intramembrane particles on both P- and E-faces, as observed earlier [14]. The form of the particles from freeze-fracture and frozen hydrated specimens was slightly asymmetric, mainly due to crystal tilt and had diameters of about 70 A. o
5. DISCUSSION In the present work it is demonstrated that two different types of annular structures can be observed in specimens containing Staphylococcus aureus a-toxin together with platelet membranes or liposomes. Since these were not observed in controls without the toxin, we conclude that they correspond to different oligomeric forms of the protein. Analysis of the toxin alone in the watersoluble form did not show any formation of rings. One of the oligomeric types, observed as single non-crystalline particles or in hexagonal.two-dimensional crystals, has a diameter of 75 A and a high stain density in the central pore or cavity. The average structure from the single particles clearly shows six identical protein domains while the projection from the hexagonal crystal only shows small variation in the density within the protein domain. This can be explained by rotational variance of the oligomers in the hexagonal crystal. In that case a conventional crystallographic approach will result in a cylindrical average. The tetragonal two-dimensional c~stals show oligomers with a larger diameter, 100 A and little stain penetration in the center. We have recently reconstructed a three-dimensional model from this form which demonstrates that the negative stain does not delineate a penetrating pore in the center but rather a deep cavity [7]. The annular protein structure is composed of four symmetryrelated domains. We have now confirmed that the negative stain used, Na-PTA at pH 7.2, does not influence the formation of the tetragonal crystals since both freeze-fracture and the frozen
hydrated specimens show crystals with similar properties. We have recently investigated the kinetics of staphylococcal a-toxin oligomerisation in relation to membrane permeabilisation using different biochemical and biophysical approaches [8]. The results show that a-toxin oligomers can be present under conditions where no membrane damage can be detected. This is the case, for instance, at low temperature and in the presence of Ca2+-ions. This suggests that a conformational change is likely to occur to form the fully active form of the protein after membrane association. Such a conformational change has also been predicted based on spectroscopic measurements [4]. The results presented in this work give additional support to the suggestion that the toxin changes during transfer from membrane binding to membrane permeabilisation. The form observed in the tetragonal crystals may correspond to the state of the protein which binds to the membrane. This oligomer does not contain a penetrating pore and it can be found also on grids with a preformed lipid layer where no extensive membrane penetration can occur. An inactive trypsin-treated form of the toxin has recently been shown to arrange in tetragonal crystals with a very similar protein domain structure [12]. Moreover, the freeze-fracture electron microscopy of tetragonal crystals revealed one type of crystals and no complementary impressions. We therefore suggest that this oligomer form is attached to the outer leaflet of the lipid bila.yer. The relatively small size of the particle, 70 A as compared to 100 A in negative stain, is consistent with the structure of the 3-D model [7], which can be described as cupshaped with its compressed bottom attached to the lipid bilayer. The particles in the crystals are only slightly elevated above the fracture plane indicating that the oligomer only extends a short distance into the inner leaflet of the bilayer. This is in accordance with side views in negative stain of a-toxin attached to membranes [15]. The other type of a-toxin oligomer, found either as single particles or in hexagonal crystals, with smaller diameter and higher stain density in
12
the center probably corresponds to the permeabilising form. The asymmetric protein unit has a smaller projected surface in the plane of the membrane consistent with a conformational change from a folded state to a more extended structure. Unfortunately, large well-ordered hexagonal crystals have not yet been frequent enough to form the basis for a 3-D analysis.
ACKNOWLEDGEMENTS This study was supported by Grants Nos. 03X00144 and 16X-05969 from the Swedish Medical Research Council as well as grants from the funds of the Karolinska Institute. Cryo-electron microscopy of frozen hydrated specimens were performed at the Max-Planck-Institut fiir Biochemie, Martinsried, FRG, during a uisit made possible through a fellowship from the Alexander von Humbolt Stiftung (HH).
REFERENCES [1] Bhakdi, S. and Tranum-Jensen, J. (1987) Rev. Physiol. Biochem. Pharmacol. 107, 147-223.
[2] Gray, G.S. and Kehoe, M. (1984) Infect. Immun. 46, 615-618. [3] Ikigai, H. and Nakae, T. (1987) J. Biol. Chem. 262, 2156-2160. [4] Tobkes, N., Wallace, B.A. and Bayley, H. (1985) Biochemistry 24, 1915-1920. [5] Arbuthnott, J.P., Freer, J.H. and Bernheimer, A.W. (1967) J. Bacteriol. 94, 1170-1177. [6] Freer, J.H., Arbuthnott, J.P. and Bernheimer, A.W. (1968) J. Bacteriol. 95, 1153-1168. [7] Olofsson, A., Kav6us, U., Hacksell, I., Thelestam, M. and Hebert, H. (1990) J. Mol. Biol. 214, 299-306. [8] Thelestam, M., Olofsson, A., Blomqvist, L. and Hebert, H. (1991) Biochim. Biophys. Acta 1062, 245-254. [9] Thelestam, M. (1983) Biochim. Biophys. Aeta. 762, 481488. [10] Hegerl, R. and Altbauer, A. (1982) Ultramicroscopy 9, 109-115. [11] Saxton, W.O. and Baumeister, W. (1982) J. Microsc. 127, 127-138. [12] Olofsson, A., Hebert, H. and Thelestam, M. (1991) J. Struct. Biol. 106, 199-204. [13] Olofsson A., Kav6us, U., Thelestam, M. and Hebert, H. (1988) J. Ultrastruc. Mol. Struct. Res. 100, 194-200. [14] Feagler, J.R., Tillack, T.W., Chaplin, D.D. and Majerus, P.W. (1974) J. Cell Biol. 60, 541-553. [15] Tranum-Jensen, J. and Bhakdi, S. (1987) Electron Microscopy of Proteins, Vol. 6 (Harris, J.R. and Horne, R.W., Eds), pp. 445-492, Academic Press Inc., London.
