Modeling the Bacterial Protein Toxin, Pneumolysin, in Its Monomeric and Oligomeric Form*

Pneumolysin is a member of the family of related bacterial thiol-activated toxins, which share structural similarities and a proposed common cytolytic mechanism. Currently the molecular mechanism of membrane damage caused by these toxins remains a matter of con- troversy. Aprerequisite for defining this mechanism is a detailed knowledge of the monomeric and oligomeric pneumolysin structures. We present for the first time details of the monomeric structure of a thiol-activated toxin, pneumolysin. Electron microscope images of metal-shadowed pneumoly- sin monomers show an asymmetric molecule composed of four domains. We have studied the conformation of pneumolysin monomer by low resolution hydrodynamic bead modeling procedures. The bead model dimensions and shape are derived solely from the electron micro- graphs. The bead model has been evaluated in terms of the predicted solution properties, which in turn have been compared to the experimental values of the sedimentation coefficient, s&,,, obtained by analytical ultra- centrifugation and the intrinsic viscosity, [q]. Pneumolysin oligomers, observed as ring and arc-shaped structures, were also examined by electron microscopy. Metal shadowing and negative staining methods were used to establish the overall dimensions of the oligomer

The membrane-damaging toxin pneumolysin is an important virulence factor of the bacterium Streptococcus pneumoniae (Berry et al., 1989). This bacterium is a major pathogen of m a n and commonly causes pneumonia, meningitis, and otitis media (Austrian, 1981). Pneumolysin is one of the so-called thiolactivated toxins, which are produced by a broad range of Grampositive bacteria and which share a large number of structural characteristics and biological properties (Alouf and Geoffroy, 1991).
One of the most noted effects of pneumolysin is the cytolysis of eucaryotic cells, which contain cholesterol in their membranes. It has been suggested that cholesterol acts as the membrane receptor for these toxins and facilitates the concentration

525030.
and resulting oligomerization of toxin monomers in the membrane (Bhakdi et al., 1985). This results in the formation of the classical ring and arc structures that have been reported for many of the thiol-activated toxins (Dourmashkin and Rosse, 1966;Duncan and Schlegel, 1975;Cowell et al., 1978;Mitsui et al., 1979;Rottem et al., 1982;Niedermeyer, 1985). We have recently presented (Morgan et al., 1994) a preliminary study, which shows, for the first time, the formation of these ring and arc structures by pneumolysin. A number of previous papers have proposed models for the oligomeric structure of these rings and arcs (Bhakdi et al., 1985;Niedermeyer et al., 1985;Duncan and Schlegel, 1975;Sekiya et al., 1993), but all have lacked essential information concerning the size and conformation of the building block, the toxin monomer.
We present a detailed analysis of the oligomeric structures, obtained by use of electron microscopy techniques. The same techniques have been applied to the pneumolysin monomer, and based on these studies we are able to propose a hydrodynamic bead model for the monomer structure. The intrinsic viscosity of pneumolysin has been determined and, in common with the previously determined sedimentation coefficient (Morgan et al., 19931, is in close agreement with the value predicted for the model. This monomeric model provides the essential information necessary for constructing the assembled form of pneumolysin, and on this basis we are able to propose a model for the gross structure of the pneumolysin oligomer.

MATERIALS AND METHODS
Reagents-All chemicals used were analytical grade (Fisons) unless otherwise stated. All protein samples were dissolved in phosphatebuffered saline (8 m M Na,HPO,, 1.5 m M KH,PO,, 0.137 M NaCl, 2.5 mM KCl, pH 7.3).
Expression and Purification of Recombinant Wild-type Pneumolysin-Recombinant pneumolysin protein was purified from Escherichia coli JMlOl and purified as described previously (Mitchell et al., 1989). Sample purity was checked by SDS-polyacrylamide gel electrophoresis and hemolytic activity assayed as described previously (Mitchell et al., 1989).
Isolation of Pneumolysin Oligomers-Isolated oligomeric pneumolysin was prepared as described previously in Saunders et al., 1989. One milliliter of sheep red blood cells (lo9 cells) were incubated with 50 pg of purified wild-type pneumolysin for 5 min at 37 "C. Membranes were pelleted by centrifugation (25,000 x g , 15 min, 4 "C) and washed three times with 5 m M phosphate buffer (pH 8.0) to lyse any remaining intact cells. Membranes were solubilized by addition of sodium deoxycholate to a concentration of 250 mM. Samples of 1 ml were applied to linear 1040% (w/v) sucrose density gradients containing 6.25 m M sodium deoxycholate as described previously (Bhakdi et al., 1985). After centrifugation at 150,000 x g for 16 h at 4 "C (Sorvall AH-627 with swingout rotor), 10 fractions (1 ml each) were collected by puncturing the bottom of the tubes. Fractions were analyzed for the presence of pneumolysin by SDS-polyacrylamide gel electrophoresis.

Intcrac/ion rc-ith Rrv/hrocy/r .~~~~, , l h r n n r s -~(~g n t i v c~l v
stained p r e p a r a t i o n s o f s h e r p r r y t h r o c y t r s t r r a t c d w i t h p n r wmolysin reveal largr numhrrs of a r c and r i n g structurc*s on their surface (Fig. In 1. Thc n u m h r r o f arcs and rings apprnrrrl t o he in proportion to thr cnnccntration of toxin usrrl. Thr s t r u c t u r e s appeared to fall i n t o thrrr m a i n r :~t r g o r i r s : rings, arcs, and "douhle arcs" (Fig. In. src nrrorc's). Thr r i n g w i d t h (see Fig. l h ) was found to m r a s u r r 6 . 5 n m 1S.D. 2 0 . 3 nmk regardless of which catrgorv of s t r u c t n r c it formcd. Thc o u t w diameter of the r i n g s v a r v in sizr from 30 to 45 n m . T h r r a d i i of c u r v a t u r e of thc a r c s appcar t h r s a m e ns t h r r i n g struct urrs.
Iso/ntrr/ O/i~ro,ncrs-Oligomrrs wrrr isolatrd from pnrumo-Ivsin-trratcd m r m h r a n c s h v s u c r o s r d r n s i t y p x t l i c n t w n t r i f u - gation. The isolated structures were prepared by various electron microscopy techniques. Preparations of the isolated oligomers that were negatively stained with uranyl acetate or sodium phosphotungstate revealed ring and arc structures identical to those seen in membranes (Fig. 2).
The oligomers were linearly shadowed on mica with platinum and carbon and the replica viewed (Fig. 3a). As with the negatively stained isolated oligomers, the dimensions were the same as those found in membranes. By measuring the length of the shadows created by the oligomers, an estimate for the height of the oligomers was determined. Tobacco mosaic virus was used for calibration purposes to accurately determine the angle of shadowing and the oligomer height was calculated as 9.3 nm (S.D. 2 1.6 nm). The absence of shadowing metal inside the oligomeric ring confirms the presence of a central hole. At high magnification (Fig. 3h), the ring-and arc-shaped structures appear to be composed of subunits, presumably pneumolysin monomer.
Molecular Shape of Pneumolysin-High magnification micrographs of platinum/carbon-coated pneumolysin reveal the apparent domain structure of the monomeric protein (Fig. 4a). The pneumolysin molecules are asymmetric as opposed to globular and are slightly curved. The individual protein molecules appear to be composed of four domains, three of which are linearly associated, with the fourth observed at a range of angles. This may imply flexible linkage with respect to the other three. The center-to-center distance between domains was measured as 3.08 nm (S.D. 2 0.65 nm) and the overall projected length of the monomers was 12.56 nm (S.D. f 0.49 nm).
Viscometry-Extrapolating the kinematic reduced viscosity to infinite dilution yielded a value of 9.1 2 0.5 mug for the dynamic intrinsic viscosity, [ql, of pneumolysin monomer (see Fig. 5

) . [q]
is a particularly sensitive probe of solution conformation but is also highly dependent upon molecular hydration (6). The shape contribution to [q] is represented by the Simha factor ( v ) (Tanford, 1963), and the two are related through the following equation, where uo is the specific volume of pure solvent.

DISCL'SSIOS
We are able to confirm preliminay data presented prrtriously (Morgan et al.. 1994) to show that pneumolysin forms thr ring and arc structures traditionally associntrd with the thiolactivated toxins. Further detailed examination of these structures shows that the dimensions of the pneumolysin oligomers are very similar to those reported for the other toxins that have been investigated (Dourmashkin and Rosse, 1966; Duncan and Schlegel, 1975; Cowell rt 01.. 1978: Mitsui rt 01.. 1979: Rottem et al., 1982Niedermeyer, 1985). This is to he expected considering their primary structure and function similarities. The oligomeric structures display extraordinary stahilitv and remain intact and, as the micrograph in Fig. 2. reveals. have a tendency to aggregate.
Surprisingly little information is availahlc concerning the detailed structure of the oligomers from this family of toxins. The most comprehensive model for a thiol-activated toxin oligomer was presented recently for streptolysin 0 (Sekiya rt n l . , 1993). This model was derived from enhanced images of electron micrographs of negatively stained streptolysin 0 oligomers. These workers proposed that the oligomeric structure is composed of an outer and an inner laver of monomer suhunits and that a %rown" portion of the monomer projrcts out from the cell membrane. However, the presence of two concentric rings of monomer suhunits is unique to the streptolysin 0 model. This proposal does not conform to the hasic packing principle of organized structures, which states that "each suhunit must he bonded to its neighhour in exactly the same way; thus all subunits are in identical environments" (Klug, 19691. We can see that for the kind of model proposed for streptolysin 0, the monomer suhunits in the outer ring are not in the same environment as those in the inner ring and r9icr r'rrso. Enhanced image processing of electron micrographs was also employed (Olofsson et al., 1993) to study perfringolys~n oligomers. The final projection density map prrsentcd the width of the ring. Another study of perfringolysin (Harris et nl., 1991), in which the technique of fluorescence resonance energy transfer was used to monitor the kinetic aspects of aggregation, suggrsts that perfringolysin 0 aggregates in a linear fashion and would seem to preclude the two-ring model. A number of other models have been proposed (Rhakdi et nl., 1985;Niedermeyer et nl., 1985;Duncan and Schlegel, 1975). In general, the proposed models have been generated using an assumed thickness for the oligomeric structure. More importantly, to date there has been no structural information concerning the toxin monomer, the primary building block of the oligomeric structure.
For this reason we have constructed a bead model of the pneumolysin monomer (Fig. 4h) based on measurements taken from electron micrographs of low angle, finely metal-shadowed pneumolysin (seen in Fig. 4a). The molecules appear to be -5.0 nm in width, although this value is incompatible with the well characterized overall length and volume. It should he noted that measuring dimensions in the direction of the shadow can be very misleading due to metal deposition, which can vary considerably (as much as 1-4 nm), in which case the thickness of the deposited metal can be of a similar order to the structure being measured. However, dimensions perpendicular to the direction of shadow along the length of the molecule may he measured quite accurately by this method. The head model representation of the pneumolysin monomer presented in Fig.   4h is hased on thrsr measurrmrnts. Thr sizr of rach domain is based on the center-to-center distnncr hrtwrrn tlom:lins r3.08 nm) along thr Irngth of thr molrculr, and thc assumption has been made that they are sphcricnl. In this modrl, pnrumolysin is composed of four domains of q u a l s i z r whrrc onr rnd domain is flrxihlr with respect to t.hr axis of thr othrr thrrr..
The use of simplr equations (Rowr. 1984) allows thc c:~lculation of a molecular Wright, hasrd on thc volumr of t h r individual domains (as estimatrd from rlrctron microp-aphs 1 and the partial sprcific volumr of pnrumolysin. Thr partial sprcific volume of 0.737 ml/g was drrivrd from thr amino arid composition (Morgan f t nl., 1993). Thr cnlculatrd valur of 50.000 agrees well with the srqucncr molrcular weight of 52.900. suggesting that in terms of volume the modrl is a good rrprrsrntation of the molecule.
Using low rrsolution hrad modrling analysis (Garcia dr l a Torre, 1989). i t is possihle to prrdict throrrtical pnramrtrrs hased on thr proposrd brad rrprrsrntation, molrculnr Wright, and partial specific volume. Thr drprndrncr of both thr srdimentation coefficient and intrinsic viscosity upon modt~l hydration and flrxibility was prohrd for a rangr of hydrations (0.0-1.4 g of waterlg of protein) and intrrsrgmrnt nnglrs (1H0. 135, and go"), as illustrated in Fig. 6 ( n and h ). Wr h a w :~dtlrrssrd the issue of flexibility in a mannrr somrwhnt dinrrrnt than that of other workrrs in this arrn (Rocco ct nl., 19931. in that w r have not attrmptrd to genrratr a srrirs of modrls for thc mon- omer each with a randomly assigned intersegment angle. Instead we are asserting that a solution of monomer will contain molecules whose time-averaged conformation is characterized by the hydrodynamic parameters to which the head model is constrained. Thus, requiring the model to conform to a n s:~,,,,, = 3.35 ? 0.1 S (as previously measured for pneumolysin; Morgan d al. (1993)) and an intrinsic viscosity o f l a ] = 9.1 T 0.5 ml/g, we observe from Fig. 6 ( n and h ) that the hydration must lie hetween 0.4 and 0.6 g of solvrntlg of protein. It is evident that the sedimentation coefficient (s&,,,,) is relatively insensitive to the intersegment angle ( ( 1 ) . However, intrinsic viscosity ([ql) is a far more sensitive probe of conformation, as demonstrated by Fig.   6h, which illustrates the dependence of [TI upon a and hydration. We have also demonstrated that more compact conformations of the domains (exemplified by a square planar configuration for which predicted data are included in Fig. 6 ( a and h ) ) are unlikely, as these would require an unreasonably high level of hydration. Recently the crystal structure of proaerolysin (Parker rt nl., 1994), the toxin produced by the Gram-negative hacterium Arronwnas hydrophiln, has heen published. Pro-aerolysin has four domains, one of which is flrxihly linked to the other three in a fashion that closely rescmhlcs our proposed structure for pneumolysin. Although proaerolysin has littlr sequence similarity to the thiol-activated toxins, the active form of the protoxin, aerolysin, does share a numhcr of structural characteristics and hiological functions. In particular it has n similar molecular weight ( M , = 48,000) and causes cytolysis. The oligomeric form of aerolysin is proposed to contain swrn monomer suhunits (Wilmsen et nl., 19921 arrangtd to form a 17-A channel. This is significantly smaller than t h e oligomeric form of pneumolysin. On the hasis of the low rcbsolution images of the pneumolysin oligomer and the head mod(.l of t h r monomer we propose the oligomeric model shown in Fig. 7 In and h ). The oligomer consists of about 30 suhunits. nrrangcd so that three ofthe domains lie adjacent to each other to form a cylindrical column and the fourth flexihle domain forms a flangrb outside the cylinder. The resulting plan view nppcars as two concentric rings of subunits. This is consistent with the cnhanced images presented for perfringolysin (Olofsson r t a/.. 1993) and streptolysin 0 fSekiya rt ai,, 1993). which hoth indicate two concentric rings of suhunits. However, thv crucial difference is that we propose a single laver of s:uhunits \r.hnsr domain structure gives the appearance of a douhlr ring. This model is similar to the models proposed for the C5h-9 complex (Tranum-Jensen et nl., 1978, and the aerolysin mcmhranc channel (Wilmsen rt nl., 1992).
We recognize that the precise orientation of thr monomcr suhunits and their domains with rcspwt to thr mcmhranc is a b 6.5nm -Modeling Monomeric and Oligomeric Pneumolysin 4 -. 35nm -+ the proposed structures for pneumolysin will be relevant to other of the thiol-activated toxins.