J. Mol. Biol. (1990) 215, 567-588

Structure o f A d e n o v i r u s Fibre I. Analysis of Crystals of Fibre from Adenovirus Serotypes 2 and 5

by Electron Microscopy and X-ray Crystallography Christiane D e v a u x t , Marc A d r i a n , , Carmen Berthet-Colominas Stephen Cusack§ and Bernard Jacrotll European Molecular Biology Laboratory, Grenoble Outstation c/o I.L.L., 156X, 38042 Grenoble Cedex, France (Received 13 March 1990; accepted 20 June 1990) An analysis by electron microscopy in amorphous ice and X-ray diffraction of four types of three-dimensional crystals of adenovirus fibre is presented. Fibre from adenovirus type 2 (Ad2) crystallizes in two forms depending on whether it is native or cleaved near the N terminus at Tyrl7. Fibre from Ad5 also crystallizes in two forms, both of which contained fibre cleaved at Tyrl 7. Analysis of the packing of the fibres in each of these crystals suggests that the overall length of the fibre may be considerably longer (about 350 to 370 A) than previously reported. Crystals of cleaved Ad2 fibre are of sufficient quality to be characterized by X-ray diffraction. They are of space group C2 and cell dimensions a ----134"4 A, b = 77"6 A, c = 539"4 A, fl = 92"7°. These crystals are remarkable in that, despite being monoclinic, the ab plane forms a perfect hexagonal lattice. This is explained by a trigonal packing of the trimeric fibre heads in the crystal. A similar feature is found for one type of Ad5 crystal, although the hexagonal lattice is 12°/o smaller. The crystals of cleaved Ad2 show very strong meridignal intensity at a Bragg spacing of 4-4 A and weaker diffuse intensity corresponding to layer-lines of spacing 26.4 A. This must reflect the quasiperiodicity of the structure of the fibre shaft, which is apparent in the primary sequence. The occurrence of these features combined with the new determination of the length of the fibre (see also the accompanying paper) require a reappraisal of the cross-fl model of the fibre shaft proposed by Green et al.

1. Introduction

of the icosahedral adenovirus. Here, we report the results of structural investigations on the fibre that were undertaken with the goal of understanding the interaction between a non-enveloped virus and its receptor at the atomic level. In electron micrographs, the isolated adenovirus fibre appears as a long, thin shaft with a diameter of 2 to 3 nm terminated by a globular head of about 5 nm in diameter. The basal end of the fibre is inserted into the penton base of the virus; the distal knob presumably carries the receptor-binding site. The overall length of the fibre varies according to serotype (Norrby, 1969). Published values are 27 to 31 rim for the fibre of adenoviruses type 2 (Ad2¶) and type 5 (Ad5), and 9 to I 1 nm for type 3 (Ad3). The sequence of Ad2 fibre polypeptide (which comprises 582 residues) shows a remarkable quasiperiodicity in the appearance of hydrophobie and proline residues between amino acid residues 45 and 401. A consensus motif of 15 residues can be defined. On the basis of this periodicity, Green et al. (1983) assigned this region to the shift and proposed for it a structural model based on the cross-fl fold

The first step in the infection by adenoviruses is the recognition of the virus by a cellular receptor. A specific viral protein, the fibre, is involved in this recognition (for a review, see Pettersson, 1984). Antibodies against the fibre neutralize the virus, possibly by inhibiting its attachment to the cellular receptor (Philipson et al., 1968). The fibre is an elongated protein that projects from the 12 vertices

t Present address: Laboratoire de Biochimie, Facult~ de M~decine Nord, 13326 Marseille, Cedex 15, France. :~ Present address: Universit~ de Lausanne, Faeuit~ des Sciences, Laboratoire de Microscopie Electronique, Lausanne-Dorigny, CH-1015, Suisse. § Author to whom all correspondence should be addressed. H Present address: Institut Laue-Langevin, 156X, 38042 Grenoble Cedex, France. ¶ Abbreviations used: Ad, adenovirus; SDS/PAGE, sodium dodecyl sulphate/polyacrylamide gel electrophoresis; PEG, polyethylene glycol; EM, electron microscopy; IEF, isoeleetric focusing. 0022-2836/90/200567-22 $03.00/0

567

© 1990 AcademicPressLimited

568

C. Devaux et al.

(Geddes et al., 1968). In this model, 44 short fl-strands connected by fl-bends would form a long, narrow amphipathic sheet. The fibre could then be a dimer in which the two sheets would be associated by their hydrophobic faces. More recently, the gene of the Ad5 fibre has been sequenced (Chroboczek & Jacrot, 1987). The homology between the polypeptides of Ad2 and Ad5 fibre is only 6 9 % , but the features of the sequences t h a t are the basis of the model for the Ad2 fibre are fully conserved in Ad5. This is also the case for the fibre of human adenovirus types 40 (Kidd & Erasmus, 1989) and 41 (Pieniazek et al., 1989), which interestingly differ in length by exactly one 15-residue motif in the tail region. O n e of the consequences of the model proposed by Green et al. (1983) is t h a t the N terminus of the polypeptide chain should be at the basal (viral) end of the fibre. This was in contradiction to a previous determination (Boudin & Boulanger, 1982) of the polarity of the pol)Teptide chain in the fibre. This point was reinvestigated b y Devaux et al. (1987), who established t h a t the N-terminal end is directly involved in the interaction of the fibre with the penton base, the major protein forming the vertices of the icosahedron of the virion, in agreement with the polarity predicted by Green et al. (1983). I t was found also t h a t the polypeptide chain is very easily cleaved after Tyr17, a fact t h a t can explain the previous incorrect assignment of the polarity of the fibre. The first crystals of adenovirus fibre were microcrystals of Ad5 fibre obtained by Mautner & Pereira (1971), but these were unsuitable for examination with X-rays. Small, single crystals of Ad2 fibre were described by Devaux et al. (1984), and a model for the packing of the molecules in the crystal was proposed on the basis of electron micrographs and X - r a y powder diffraction. However, the d a t a were insufficient to make an unambiguous determination of the space group, upon which depended the deduction of the degree of oligomerization of the fibre (dimer or trimer). Although the stoichiometry of the fibre has been uncertain for some time, an analysis of the virion composition by Van Oostrum & B u r n e t t (1985) has clearly established t h a t the fibre of Ad2 is a trimer. This is in contradiction to the dimerie molecule favoured by the model described by Green et al. (1983). I t is reasonable to assume t h a t because of the sequence similarities Ad5, Ad3 and Ad40 all have the same basic trimeric structure. We have now obtained four different crystal forms of adenovirus fibre, from native and cleaved (after T y r l 7 ) fibres of Ad2 and two different forms of cleaved Ad5 fibre, all of which have been observed by electron microscopy in amorphous ice. The crystals of cleaved Ad2 fibre and one form of Ad5 fibre grow to macroscopic size and, although they are not suitable for structure determination, X - r a y diffraction d a t a have been obtained. These, combined with electron microscope images, provide detailed information on the packing of the fibres in the crystals and suggest some new features for the structure of the shaft of the fibre. In the aceom-

panying paper, Ruigrok et al. (1990) demonstrate t h a t the length of the fibre of Ad2 and Ad5 is most likely 37 nm and t h a t of Ad3 is 16 nm, significantly longer than previous measurements. T h e y are, however, consistent with the analysis of the crystals presented in this paper and imply the need to propose an alternative model to t h a t proposed by Green et al. (1983) for the fold of the polypeptide. This will be the subject of a forthcoming paper.

2. Materials and Methods (a) Purification of Ad2 and Ad5 fibres Fibres were purified from the pool of cellular" and nonassembled viral proteins as described by Devaux el al. (1987), precipitated by 8 % (w/v) polyethylene glycol 6000 (PEG), dissolved in 40mM-Tris'HCl (pH7"65) and directly subject to DEAE-Sepharose chromatography without dialysis. Fibres were eluted with 70 mM-NaCI. (b) Chymotryptic cleavage Digestions by chymotrypsin were performed at 37°C in 40 mM-Tris-HC1 (pH 7-65}. To explore the kinetics of digestion, the reaction was stopped by addition of chymostatin. For preparative purposes, the chymotrypsin was removed by the following procedure. Immediately after the incubation, the solution was loaded onto a DEAE-Sepharose column equilibrated in the same Tris buffer'. Chymotrypsin was not retained and fibres were eluted by 150 mM-NaCI in Tris buffer. Times of digestion and the enzyme to substrate ratio varied according to serotype and concentration of the fibre solutions. (c) Crystallization Proteins were concentrated by vacuum dialysis in a Prodicon system, Crystallization was carried out by the hanging drop method, using drop volumes of 4pl for screening assays and 20p[ when appropriate conditions were found. (d) SDS/polyacrylamide gel electrophoresis ( S D S / P A G E ) SDS/PAGE was performed as described by l)evaux et al. (1987). The concentration of acrylamide was 13 or 17% (w/v). (e) Analytical isoelectric focussing ( I E F ) Analytical IEF under native conditions was carried out on an LKB Instruments Multiphor flat-bed gel apparatus according to the LKB application note. Portions (15 pl) of proteins (3 mg/ml) were applied to a 1"5 mm thick 5% polyacrylamide gel containing 1"6°/o (v/v) Pharmalyte (pH 5 to 8; Pharmacia Fine Chemicals), 13% (v/v) glycerol and 0"023% (w/v) ammonium persulphate. Gels were pre-run at a constant power of 15 W for l h and then at 30 W with a maximum voltage of 1500 V for 3 h. (f) Sequencing Sequencing was performed either on purified fragments of Ad2 or Ad5 fibre or after electro-elution from SDS/PAGE in an ISCO apparatus as described by Bhown & Bennett (1983). Limited manual sequence analysis was performed according to Tarr (1977). The phenyl thiohy-

Crystals of Adenovirus Fibre. I dantoin (PTH) amino acid derivatives were identified with an isocratic system (Lottspeich, 1985). (g) Titration of accessible -8H gTvups Free -SH groups were determined using Elhnan's reagent 5.5'-dithio-bis(2-nitrobenzoic acid) (NbS2; Boehringer-Mannheim France). The reaction was performed in 50raM-sodium phosphate (pH7.2), 10mMEDTA. A total of 300 pl of solution containing 0"5, l and 2 nM-protein were mixed with 150/~1 of NbS2 (40rag NbS_, in l0 ml phosphate buffer). The yellow colour was developed for 15 rain at room temperature and read at 412 n m. A molar absorption coefficient of 13.600M -I cm -I was used tbr the calculation of the sutphydLvl content. Protein concentration was calculated Dora absorbance at 278 nm (using absorption coefficients of el,,,~ (! era)=-8"5 for fibre, 14"5 for hexon (Devaux et al., 1982) and 25"5 for lysozyme). All manipulations were performed trader nitrogen. A reagent blank was subtracted fi'om the measured vahte. (h) Electron microscopy ( EM ) EM observations were made on frozen-hydrated samples (Adrian et al.. 1984) with a JEOL CX II. Some observations were made on samples stained with uranyl acetate. Filtered images were obtained from well-ordered images by pseudo-optical filtering or correlation averaging using the Seml)er image processing package (Saxton el al.. 1979). (i) X-ray diffraction Powder diffraction patterns fl'om microc15,stals were collected on an Elliot GX20 rotating an(ide tube with a camera using a double focusing system consisting of a Fra,lks mirror and a quartz monochromator. Several sample:to-film distances between 200 anti 700 nm were used. Single-crystal diffraction was observed in the laboratory using a precession camera on an Elliot GX13 rotating anode generator with double-mirror tbcusing optics. Other measurements were made on rotation cameras at the synchrotron sources at LURE (station D41) and Daresbury (Station 7-2. )~= 1"488A (l A =0'! nm), specimen-detector distance 145 ram). The thin plate-like crystals were difficult to handle without breaking; best results were achieved by mounting the crystals in flattened capillaries.

3. Results (a) Purification and crystallization (i) Adenovirus type 2 fibre (i)(a) Purification As described by Devaux et al. (1987), the fibre solutions purified by the three-step method of Boulanger & Puvion (1973) (55% saturated ammonium sulphate precipitation, D E A E - S e p h a d e x and h y d r o x y a p a t i t e chromatography) were very sensitive to proteolysis. I n t a c t and cleaved fibres gave two different types of crystals when observed b y electron microscopy (Devaux et al., 1987). F o r this work, a simplified procedure (see

569

Materials and Methods) was used to obtain a stable pure fibre. S D S / P A G E (Fig. l(a)) showed t h a t after elution from the DEAE-Sepharose column, the fibre was present as a single band in a major peak followed by a shoulder. F o u r fractions were pooled and examined by I E F (Fig. l(b)). Fraction 1 contained a major band at pI 6-2, with a very weak band at pI 6"1; fractions 2 and 3 contained about equal amounts of the pI 6.2 and pI 6.1 species, whereas in fraction 4 the pI 6.2 species had completely disappeared and a pI 6"0 species was visible. All the bands reacted in immunoblotting with antiAd2 fibre and anti-N-terminal peptide sera, and with wheat-germ agglutinin lectin, which recognizes the N-acetyl glucosamine residues t h a t are known (Ishibashi & Maizel, 1974; Caiilet-Boudin et al., 1989) to be present in the Ad2 fibre (and probably in the Ad5 fibre). (i)(b) Proteotysis Kinetics of cleavage by chymotrypsin (fibre concentration, 2 mg/ml; ratio of enzyme to substrate, 1:10), followed by analysis by SDS/PAGE, showed t h a t the 62,000 M r polypeptide chain of the fibre was rapidly converted into a 60,000 M r species. This conversion was completed in two hours, but two other bands appeared at 4 5 , 0 0 0 M r and at 15,000M r. The 60,000M r band had completely disappeared after 20 hours of incubation (Fig. 2). The cleavage t h a t produces the 45,000 and 15,000 Mr products is clearly much slower than what cleaves 2000 of the initial 62,000 M r intact fibre. The polypeptide chain of the fibre contains a blocked N terminus (acetyl-methionine; Anderson & Lewis, 1980). Sequencing of the N terminus of the c h y m o t r y p t i c fragments showed t h a t the first cleavage occurred after T y r l 7 (Devaux et al., 1987). The 60,000 and 45,000 M r fragments showed the same N-terminal sequence: Asp-Thr-Glu-Thr, which follows T y r l 7 . The N-terminal sequence of the 15,000 M r fragment was found to be Thr-Gly-. This dipeptide appears five times in the Ad2 fibre sequence, but only the one t h a t follows Met448, a potential site of action for the chymotrypsin, can give rise to a 15,000 M r fragment. According to our previous results on the polarity of the fibre, we could say t h a t the 45,000 Mr fragment was part of the shaft and the 15,000 Mr p a r t of the C-terminal end t h a t constitutes the head. (i)(c) Crystallization The largest and most homogeneous fraction 1 of the DEAE-Sepharose c h r o m a t o g r a p h y (Fig. l) was concentrated to a b o u t 3 mg/ml (some aggregation occurred beyond this concentration) and was assayed for crystallization. Aggregates of microcrystals were obtained using 4~/o P E G 6000 as precipitating agent in 10 m~-Tris-maleate or sodium phosphate at p H 6-1. These were unsuitable for s t u d y by X - r a y diffraction, but they could be observed by electron microscopy (see below).

570

C. Devaux et al.

(o)

131

I

=gllP .,=mm~

-

T''

I

2

3

4

tct

2ct

3ct

4ct

(b)

(A

6"2'

6"1 6"0

+

Figure 1. (a} Analysis by SDS/PAGE of the proteins eluted from the DEAE-Sepharose chromatography. (b) Four pooled fractions were gathered, concentrated and analysed by isoelectric focusing before {lanes 1, 2, 3 and 4) and after (lanes lct, 2ct, 3ct and 4ct) digestion with chymotrypsin (ct). The proteins were visualized by Coomassie blue staining. Ad2 proteins dissociated from the virus were used as markers in the SDS/PAGE (lane V). The pH gradient in the IEF gel was determined with a pH meter after elution in water of some pieces of gel.

Crystals of Adenovirus Fibre. I 2Z

2

5

4

571 ~

5

6

11

tTr

)"~{] ,,,,,,, rv

62K 60K

;i

.

.

.

.

.

.

.

.

.

.

.

I~1], --45K

Figure 2. SDS/PAGE analysis of the kinetics of cleavage of Ad2 fibre by chymotrypsin (ratio of enzyme to substrate of 10) at 37 °C; after t h (lane 2), 3 h (lane 3), 6 h (lane 4) and 20 h (lane 6). (CT, ehymotrypsin). Proteins dissociated from Ad2 virus were used as markers (lane V). Lanes 1 and 5 showed intact fibre polypeptide (62,000 Mr). K, x 10a Mr. 1 :

Conditions for the growth of macroscopic crystals of intact fibre have not been found. Fraction 1 (3 mg/ml) was digested for two hours with chymotrypsin. SDS/PAGE showed that all the 62,000 M r polypeptide was converted into 60,000 Mr, but the two other bands (45,000 and 15,000 Mr) were detected at a level of about 10%. It was not possible to find conditions where only the 60,000 M, species was present. The cleaved protein appeared as a single band at pI 6-15 in I E F (Fig. l(b)). It was highly soluble and could easily be concentrated up to 16 mg/ml. The best crystals were grown at about 9 mg/ml, 5 % PEG in l0 mM-sodium phosphate (pH 6"0), in about three weeks. They were thicker than those obtained previously (Devaux et al., 1984) and could be used for X-ray analysis, although they appeared to be twinned. In

electron microscopy, they give one type of image only (see below). SDS/PAGE and sequencing of washed crystals revealed that all the fragments ((60+45+ 15) x 10~ Mr) were present in the crystals in the same proportion as that of the starting solution of cleaved fibre. (i)(d) Stability of the fibre The oligomeric structure of the fibre is extremely stable. When the intact fibre is treated with 8 Murea (60 min at 37 °C) and analysed by SDS/PAGE, no band appears at the position of the monomer (Fig. 3, lane 3). To observe the monomer of the: intact fibre (62,000 Mr} by SDS/PAGE, the samples of fibres must be dissociated by 2 % SDS and 8 Murea (10 min at 37°C; Fig. 3, lane 1). This dissociation does not require the presence of a reducing

C. Devaux e t al.

572

"(T

I

2

5

4

5

6

7

vrr

62K

'fl'Fo

60g

w-

45K

,/?;'~/:)i!~!??,:

o-r'r

15K m"

F i g u r e 3. SDS/PAGE analysis of the fibre treated with urea in the presence or absence of SDS. After t r e a t m e n t with chymotrypsin for 20 h the fibre was incubated for 60 min at 37°C with various amounts of urea (0 M, lane 3; 2 M, lane 7; 4 M, lane 6; 6 M, lane 5; 8 M, lane 4) in the absence of SDS or in the presence of 2 °/o SDS and 8 M urea (lane 2). The intact fibre incubated for l0 min at 37°C with 8 M-urea and 2O/o SDS is in lane l. All the samples were deposited onto the gel without further treatment. The stacking gel, separating gel a n d the separating gel buffer contained 0"l ~o SDS. In the absence of SDS, all fibre samples showed several high molecular weight species indicated at the top of the gel. K, x l 0 3 Mr.

Crystals of Adenovirus Fibre. 1 Table 1 Data for the titration of -SH groups A

B

Ad2 libre

0'9 I '0

3"O 2"9

Ad5 fibre Lysozyme

0

I "4 0'3

Ad2 hexon

|'15 1 "7

0-1 5-0 6"5

C 4 6 6"6

A. Samples in native conditions (50 raM-sodium phosphate (pH 7-2) = buflbr A). B. Samples in denaturing conditions without reducing agent (buft;er A with 2 % SDS in 8 M-urea). (!, Samples in denaturing conditions its iu B but in the presence of dithiothreitol. Each value is the average of 3 assays. When 2 v~lues appear in the Table, they result from 2 independent experi,nents.

reagent (this work; and Caillet-Boudin & Lemay, 1986). The same results are obtained with the cleaved fibre containing the 60,000/1/~ polypeptide (not shown). Fibre solutions containing only the 45,000 and 15,000 M~ polypeptides can be obtained after 20 hours of incubation with chymotrypsin. Attempts to separate these two species by gel filtration (Ultrogel AcA 54; IBF France) were unsuccessfnt. On electron micrographs ot" negatively stained molecules, the fibres made of the 60,000 M~ polypeptides and those containing only the 45,000 and 15,000 Mr |)'agments were indistinguishable ~P~ob FCuigrok, l)ersonal communication). However, the behaviour of these latter fibres is somewhat different in denaturing gels. More precisely, when the sample is incubated in 2 M-urea (without SI)S) and analysed by SDS/PAGE, a 15,0()0 AIr band is observed. The intensity of this baud increases with the amount of urea (Fig. 3, lanes 7 to 4). On the other hand, the 45,000 M t band is observed only after incubation in SDS (compare the 62,000 and 60,000 M~ species). This result suggests that the extreme stability of the fibre, which is demonstrated by its non-dissociation in urea alone, should be attributed to its shaft. (i)(e) Accessibility of the -SI! groups of the fibre There are three cysteine residues per chain tbr Ad2 fibre (H6rriss6 et al., 1981) of which only two are present in Ad5 fibre (Chroboczek & Jacrot, 1987). Their accessibility and possible involvement in disulphide bridges were investigated using NbS: (see Materials and Methods). Ad2 hexon (7 cysteine residues/chain) and lysozyme (8 cysteine residues engaged in 4 disulphide bridges) were used as positive and negative controls, respectively. Grutter & Franklin (1974) and Jornvall et al. (1974) found one fi'ee -SH group per chain in the native hexon and six when the molecule was denatured by SDS without reduction. Jornvall & Philipson (1980) suggested that there was no evidence for disulphide bridges. The results are summarized in Table 1. Titration with NbS 2 under native conditions revealed one free

573

-SH per chain for the Ad2 fibre and hexon and none for the Ad5 fibre. After denaturation by SDS and urea, but without reducing agent, three -SH groups were titrated for each chain of"Ad2 fibre, 1-4 for Ad5 fibre, five to seven for hexon and none for lysozyme. Six to seven were revealed when lysozyme was reduced by dithiothreitol. These results suggest that all the cysteine residues of the fibre are free, but probably differ' in accessibility. This is consistent with the observation that the monomer of the fibre is obtained in SI)S/PAGE even in the absence of reducing agent. It is likely that the most accessible cysteine in Ad2 is Cys419, which is absent in Ad5. (ii) Adenovirus type 5 fibre (ii)(a) Purification and proteolysis Ad5 fibre was purified in the same way as Ad2 fibre. The elution curve was similar" and SDS/PAGE showed a single band at an apparent molecular weight of 61,000, but several bands in IEF. Although the molecular' weights of Ad2 and Ad5 fibres calculated fi'om their" sequences differ by the equivalent of only one amino acid, the migration of their monomers is slightly different in SDScontaining gels (62,000 and 61,000 apparent molecular weight, respectively). Ad5 fibre is also sensitive to chymotrypsin. The band at 61,000 M r is converted into a single band with an apparent molecular weight of 59,000 (fibre, ] mg/ml; ratio of enzyme to substrate, 1 : 5, 5 h at 37 °C). Sequencing of the 59,000 Mr species gave the same N-terminal sequence as 60,000 M r fl'agment of Ad2 fibre. Sequences of Ad2 fibre (H6riss6 et al., 1981) and Ad5 fibre (Chroboczek & Jacrot, 1987) show that the first 50 amino acids at the N-terminal end are identical, but that in the C-terminal part, the second site of cleavage in Ad2, Met448-Thr-Gly- is replaced by Ile448-Ser-Gly-. Isoleucine is not a substrate for chymotrypsin. (ii)(b) Crystallization Ad5 fibre from DEAE-Sepharose was difficult to concentrate to more than 1.5 to 2 mg/ml. Microcrystallization occurred very quickly (overnight) by dialysis against l0 mM-sodium phosphate (pH 7 or 6) at 4°C. In the electron microscope two different crystal forms were observed (Figs 8 and 13), sometimes on the same grid. The two forms are similar to those found by Mautner & Pereira (1971) and Pereira et al. (1975). Following these authors we call them, respectively, the Pisa and sawtooth types (personal communication). In these first assays the spontaneous proteolysis of the Ad5 fibre was not controlled, but in all crystals that were analysed by SDS/PAGE the fibre was found to be cleaved. Further assays were performed using fibre solutions pretreated with chymotrypsin as described above. The results can be summarized as follows. Both Pisa and sawtooth crystals can be obtained with chymotrypsin-treated fibre. Furthermore, no crystals have ever been obtained with intact fibre. The largest

C. Devaux et al.

574

Table 2

Lattice parameters for each crystal type Crystal type

a (A)

b (A)

c (A)

fl (°)t

Method

Ad2 (cleaved) Ad2 (uncleaved) Ad5 (sawtooth)

134'4 135"8 135"6

77"6 78-2 78-2

539"4 540 615

92'7 92-0 92'5

X-ray (single crystal) EM (ice) EM (ice)

73'9 74"2

---

186"1 175"1

-95-2

74"6 118"5 118"6

-68"3 68"6

184"0 376"5 370

94"7 95 95

X-ray (powder) EM (low dose, negative stain) EM (ice) X-ray (singlecrystal) EM (ice)

Ad2 (Pisa)

t The angle fl is chosen to be as near as possible to 90°. crystals are obtained in ammonium sulphate, as described below, and are always of the Pisa type. The problem of precipitation during concentration was overcome by replacing the dialysis under vacuum step by absorption onto microcolumns of DEAE-Sepharose followed by elution at 150 mMNaCI. The cleaved fibre (59,000 Mr) could then be concentrated to 2-5 to 3"5 mg/ml in 20 raM-sodium phosphate (pH 7-2) containing 75 mM-NaCI. The best crystals were obtained by the hanging drop method and grown in one week at room ~mperature in 20 mM-sodium phosphate (pH 6-7 or 6-5) containing 75 mM-NaCI with 23% ammonium sulphate. Electron microscopy showed that they were of the Pisa type (Fig. 8). SDS/PAGE and sequencing of washed crystals showed the presence of only one kind of polypeptide (59,000 Mr) whose N-terminal sequence begins at Aspl8. (b) Electron micrographs of adenovirus fibre crystals Four types of crystals have been obtained and observed by electron microscopy in amorphous ice. They will be referred to hereinafter as uncleaved Ad2 (Fig. l l), cleaved Ad2 (Fig. 9), Ad5 Pisa (Fig. 8) and Ad5 sawtooth (Fig. 13) crystals. Both types of Ad5 fibre crystals were originally observed by Pereira et al. (1975) using standard negative staining techniques. In the case of Ad2 fibre, the two types of crystals can be unambiguously related to the chemical form of the molecule. When the fibre preparation is carried out in such a way that proteolytic cleavage does not occur, "uncleaved" Ad2 crystals are always obtained. If, on the other hand, the fibre is treated with chymotrypsin, it crystallizes as "cleaved" Ad2 crystals. In previous studies (Devaux et al., 1984), the fibre was prepared without special care against proteolysis and cleaved Ad2 crystals were observed. Analysis with SDS/ PAGE and sequencing of the N terminus revealed (Devaux et al., 1987) that in these crystals the fibre was cut, presumably by a cellular protease, after Tyrl7, at the same position as the cleavage achieved in a controlled way by chymotrypsin. Although in this case cleaved Ad2 crystals were the usual form, small uncleaved Ad2 crystals were

sometimes observed on the same grid. A possible explanation could be that the small fraction of uncleaved fibre that is still present separates from the cleaved fibre during the crystallization process and forms a microcrystal. With Ad5 fibre there is no such clear correlation between crystal type and the proteolysis of the molecule (see above), and again it is possible to observe both types on the same grid. These images were obtained with samples in amorphous ice without stain, and dark regions correspond to protein-rich parts of the crystals. Therefore, it is reasonable to assume that the very dark bands (observed in both types of Ad2 fibre crystals and the Pisa type of Ad5) or elongated spots (sawtooth type of Ad5) result from the fibre heads and that the thin vertical striations {visible in all but the Pisa type) correspond to the shift of the fibre. Optical diffraction patterns were obtained for each crystal type, from which lattice parameters have been derived (Table 2). With the exception of the Ad5 sawtooth, each crystal type was observed in two different orientations, both with the long axis parallel to the plane of the grid (see e.g. Fig. 11 for the uncleared Ad2 form). Calibration of the electron microscope magnification was done by two methods: firstly, using the value of 540 A for the long axis parameter as determined by single-crystal X-ray diffraction for the cleaved Ad2 fibre crystals; and secondly, using a catatase single crystal. Both methods give the same results, indicating the absence of significant distortions in crystals observed in amorphous ice.

(c) X-ray diffraction on single crystals of

adenovirus fibre (i) Crystal morphology and quality Crystals of cleaved Ad2 fibre grow as fiat plates up to 1 mm in lateral dimensions, but only about 20 to 100 #m thick. The crystals do not have a regular shape, but usually have one smoothly curved edge {sometimes nearly straight), the other edges being highly irregular {Fig. 4). They usually give the appearance of being layered. The crystals are very

Crystals of Adenovirus Fibre. I

575

Figure 4. Photograph of relatively thick (100 pro) crystals of cleaved Ad2 fibre. fragile and have to be mounted in, flattened capillaries to minimize the possibility of breakage or bending. Ad5 Pisa crystals show interesting changes in morphology depending on their sensitivity to the amount of ammonium sulphate used as precipitant. Crystals grown in 22 to 23°/o ammonium sulphate are very similar in morphology to those described above for cleaved Ad2, and give oriented diffraction patterns, but are invariably multiply twinned (see below). At 24% ammonium sulphate the crystals are very clearly multilayered, but still fiat, and give diffraction patterns with ahnost complete disorder about the axis perpendicular to the plate. At 25o/o ammonium sulphate "hexagonal whirls" grow, which have an overall hexagonal shape built up of layers that are tilted to give a spiral pattern. These crystals are again disordered. (ii) Unit cell dimensions and space group Table 2 gives a summary of the unit cell dimensions obtained for the four" types of adenovirus fibre crystals by X-ray analysis or electron microscopy. The agreement between the values obtained by X-ray diffraction on single crystals and electron microscopy in amorphous ice is very good, with less than 1% difference. (ii)(a) Cleaved Ad2 fibre crystals While all cleaved Ad2 fibre crystals appear to be twinned to a greater or lesser extent, often the dominant component gives a clearly distinguishable

diffraction pattern. Precession photographs (both screened and unscreened, not shown) of the principal zones have enabled an unambiguous determination of the space group for' these crystals; it is C2, with cell parameters a = 134"4/~, b = 77"6 A (unique 2-fold), c--539.4 A, fl=92-7 °, implying a unit cell volume of 5619 nm 3. Twinning is usually observed in such photographs as a splitting of the c*-axis row lines by up to several degrees. Physically, this corresponds to the macroscopic crystals being made up of a small number of thin plates that are not quite parallel. Because of the long c-axis, the relatively high mosaic (maybe due to crystal bending) and twinning, still X-ray photographs with the beam along to the c-axis, have the appearance expected of a screenless precession photograph (Fig. 5(a)) and it is extremely difficult to achieve accurate alignment in this orientation. Figure 5(a) shows almost exact mirror symmetry as expected ['or this orientation, but also two other remarkable features. Firstly, the a'b* plane is, within experimental error, a perfect hexagonal lattice, i.e. a = bx//3; and secondly, although there is no overall 3- or 6-fold symmetry, the diffraction pattern in Figure 5(a) shows pseudo-6-fold symmetry (i.e. local mirror symmetry) at high resolution. The best screened precession photograph of the hkO plane (not shown) shows almost perfect cram symmetry (as it should for C2), but also only small deviations from 6-fold symmetry. The presence of a true 3- or 6-fold axis in the crystals (which is simply not apparent due to crystal misalignment)

576

C. Devaux et al.

(a) Fig. 5.

is completely ruled out by the observation of a non90 ° fl* (Fig. 5(b)). These results are compatible with the crystals being composed of close-packed long trimeric molecules whose 3-fold axis is nearly parallel to the c-axis. Some suggestion as to the tilt of the fibre axis to the c-axis can be obtained from the a'c* screened precession photograph (Fig. 5(b)). Here, it can be observed that the 10/and 20l row lines sample strong intensity that is inclined at an angle of about 9 ° to the a*-axis. If this strong intensity corresponds to the transform of the cylindrically averaged fibre, it would imply an angle of 9 ° between the fibre axis and the c-axis. As discussed below (section (d)(ii)(b)), the hexagonal lattice and pseudo-trigonal symmetry probably arise from a trigonal packing of the fibre heads. These crystals are identical to, but of much larger size than, those studied by Devaux et al. (1984). In this work, the determination of the space group was tentative and in the absence of single-crystal diffraction patterns the hexagonal lattice was taken as an indication of a trigonal space group. With this space group and the measured crystal density, the fibre was argued to be a dimer. Now it is known that the space group is monoclinic and from packing arguments (see section (d)(ii)(b), below) it seems more likely that the unit cell accommodates 16

trimeric fibres rather than 24 dimeric fibres. More direct evidence for the trimeric nature of the fibre come from quantative gel analysis (Van Oostrum & Burnett, 1985) and cross-linking experiments (Chatellard & Chroboczek, 1989; Ruigrok et al., 1990). Crystals of Ad2 fibre diffract to at least 3"3 A resolution, although the 1/540A spacing is not always well resolved at high angles. This, combined with the difficulty of handling the crystals and the twinning, makes them unsuitable for a highresolution structural analysis. Figure 6 shows one of the best rotation photographs obtained at l)aresbury, which shows other features described in section (c)(iii), below. (ii)(b) Ad5 Pisa crystals X-ray diffraction patterns of Ad5 crystals always showed twinning, but the cell dimensions could nevertheless be determined (see Table 2). The parameters are a = l l 8 ' 5 A , b--68"3A, c----376"5A, fl (estimated)--94 °. This gives a volume for the unit cell of 3040 nm 3. It is remarkable that, as for Ad2, the a'b* plane is, within experimental error, a perfect hexagonal lattice, i.e. a = b~/3, but the Ad5 lattice is smaller by 12~o. From the X-ray patterns

Crystals of Adenovirus Fibre. I

577

(b)

Figure 5. (a) Still X-ray photograph of cleaved Ad2 fibre crystal with the beam along the c axis. The diffraction pattern has overall mirror symmetry about the a* direction corresponding to b* being a crystallographic 2-fold axis. Note the hexagonal lattice and the local 2-fold axes at 60° and 120° to a*. (b) Central part of a 2° unscreened precession photograph of cleaved Ad2 fibre showing the a'c* plane (crystal-to-film distance 150 mm). A fl* angle of 87"3° was measured from this photograph. The broken line shows strong intensity inclined at 9 ° to the a* axis (see the text).

and the cr~To-EM images it is evident t h a t the space group must again be C2 with the unique 2-fold axis as b. However, unlike Ad2, it has not been possible to obtain unambiguous pl~cession photographs of the principal zones. Indeed, still photographs with the crystal aligned as well as possible with the X-ray beam along the long axis show a hexagonal lattice with intensities showing almost perfect 6m s y m m e t r y ; however, tile spots are clearly multiple. This we interpret as arising fi'om the macroscopic crystal being composed of several layers, which can be disoriented with respect to each other by some

multiple of 60 ° a b o u t the c-axis. This is confirmed by the appearance of some diffraction patterns taken with the X - r a y beam perpendicular to the long axis. These show the meridian (00l) with a welldefined spacing of 376 A. However, the row lines 20l and - 2 0 l both show more closely spaced spots, i.e. an apparently larger spacing. Close inspection shows t h a t these spots are not regularly spaced (and hence do not arise from a longer unit cell axis, with a screw axis along c) and probably arise from a superposition of the 20l and - 2 0 l (and possibly 1 1 / a n d - 1 ll) row lines due to the proposed 60 ° rotational

578

C. Devaux et al.

Figure 6. Oscillation photograph (A~b= 2 °) of cleaved Ad2 fibre crystal with the beam about 6° away from perpendicular to the c axis and with b* close to the spindle axis (taken at Daresbury synchrotron radiation laboratory, Station 7-2, wavelength 1-488 A, crystal-to-film distance 145 ram). Reflections separated by 1/540 A are resolved to the edge of the film (3-9 A). The broken white lines indicate 7cry strong (near-) meridional intensity in the region of 4";~ to 4"4 A, and at 6'6 A and a diffuse streak sampled by Bragg reflections at 13"3 A. These correspond to layer-lines I = 2, 4 and 6 of an axial repeat of 26"4/1 (see also Fig. 9). Note extra rows of reflections coming from differently orientated satellite crystals to the main crystal.

disorder about the long axis. The reason why this kind of disorder can arise in Ad5 and not Ad2 is apparent from the packing models described below. I f we assume roughly the same packing density as in the cleaved Ad2 crystals (16 fibres/5619 nm 3) the Ad5 unit cell could contain eight or nine fibres, only eight being consistent with space group C2. (iii) Other features in the X-ray pattern,s of cleaved

Ad2 fibre c~'ystals Two particular features of the X-ray diffraction patterns of the cleaved Ad2 crystals require special attention. Firstly, all pictures taken with the X-ray beam more or less perpendicular to the long axis show particularly intense meridional or near-meridional reflections in the proximity of Bragg spacings of 4"43 to 4"27A, of 6"6A and of 13.1 to 13.3A (see Figs 6 and 7). This feature is observed also with Pisa crystals of Ad5. This observation is consistent with the crystal lattice sampling layer-lines with a spacing of about 26"4 A. In overexposed films or

photographs in which the collimation was not good enough to resolve the 1/540/1 spacing, the appearance of the diffraction pattern is t h a t of a fibre pattern, with a particularly intense meridional reflection at 4"4 A (1 = 6 ) and continuous diffuse layer-lines corresponding to orders / = l (offmeridional), 2 and 4 of the fundamental spacing of 26-4 A (Fig. 7). Indeed, intense spots are also visible at 3"8 A and 3"3 A, which correspond to the seventh and eighth orders of this periodicity (Fig. 7). It seems likely t h a t these layer-lines arise from a helical motif along the length of' the fibre shaft. In the cross-fl model proposed by Green et al. (1983), strong meridional or near-meridional intensity is predicted at a spacing of 4-7 A on the c*-axis originating from the separation of successive fl-strands. The authors claim to have observed this intensity in electron diffraction images of Ad5 sawtooth crystals. Earnshaw et al. (1979) also observe strong diffraction at 1/4-7 A in their study of the tail fibre of bacteriophage T4, which t h e y a t t r i b u t e to a cross-fl structure. We have never observed strong meridional intensity at a spacing of 4'7 A in any

Crystals of Adenovirus Fibre. I

579

F i g u r e 7. Still X-ray photograph of cleaved Ad2 fibre crystal with the beam a b o u t 14° away from perpendicular to the c axis (taken with rotating copper anode generator, crystal-to-film distance 60 mm). This tilt brings axial reflections in the region of 3 to 4 A- ~ onto the Ewald sphere. The white broken lines correspond to orders of an axial repeat of 26-4 A. There are very strong reflections corresponding to orders l - - 6 {4-40 A), I = 7 (3.77 A) and 1-~ 8 (3"30 A). Weak diffuse bands are also visible corresponding to layer lines I = 1 (off-meridional), I = 2 (meridional) and l = 3 (off-meridional). The arrow indicates a very strong diffuse peak at about 15 A-1 on the equator.

580

C. Devaux et al. 376

0=

I~

!

Am.

,0, 'O

~=::=:=

I

Model A

V

@

J,O v

I

M o d e l El

Figure 8. Electron mierograph in amorphous ice (bottom) and alternative models for the longitudinal packing of Ad5 Pisa crystals. The edge of the crystal (E) clearly shows that packing model B is correct, and further strongly suggests that the length of the fibre is close to the unit cell length (376 A).

X-ray diffraction pattern and cannot explain this observation reported by Green et al. (1983). The continuous diffuse features just described show that a weak fibre diffraction pattern is superimposed on the crystalline diffraction pattern. This suggests that there is a small fraction of the crystal that is rotationally disordered. The l = 0 layer-line (equator) has a broad diffuse maximum at a spacing of about 15 A (Fig. 7). This is observed also in films of the a ' b * plane, as a diffuse ring (Fig. 5). A simple explanation for this feature would be that it corresponds to the first maximum in the Fourier transform of the projection of the shaft of the fibre. If the fibre tail is approximated by a solid cylinder of radius R, then the

maximum would correspond to that of {Jl(x)/x} 2 (at x = qR = 5-3), and this cylinder would then have a diameter of about 25 A, which is reasonable (see Discussion).

(d) Crystal packing (i) Longitudinal packing Most of the micrographs obtained with fibre crystals show crystallites with the fibres parallel to the plane of the grid. We interpret the main features of the micrographs by making the assumption that the heads of the fibres are all in the planes that are seen in the various images as dark bands (which in the

Crystals of Adenovirus Fibre. I

581

Figure 9. Top: Electron micrograph in amorphous ice of cleaved Ad2 fibre crystals. The axial unit cell dimension is 540 A. Bottom right: Filtered image of well-ordered region (F) obtained by correlation averaging. Bottom left: Axial density obtained by projecting scanned image perpendicular to the long axis, showing the 4 dark bands of heads (H1, H2, H3 and H4) and the 4 weaker bands presumed to be the N-terminal ends (N1, N2, N3 and N4). The crystal end (E) was used for the I-dimensional projection shown in Fig. 10.

case of the cleaved and uncleaved Ad2, and Ad5 Pisa crystals are perpendicular to the long axis of the crystal). The fibre tails extend perpendicularly (or nearly so) from these planes of heads in one or the other direction. This interpretation is supported by images in which the dark bands are clearly resolved into ellipsoids about 40 A apart and 55 A long, fi'om which originate thinner lines that are presumably the shafts of the fibre (Figs 9 and 11). A closer examination reveals that, for each type of crystal, two principal projections are observed, which are characterized by different spacing between the thin striations. Figure l l shows a typical micrograph obtained with crystals of uncleared Ad2 fibres in which these different projections are easily recognizable. Since the lattice parameters deduced from these micrographs are nearly identical to those measured by X-ray diffraction (see Table 2) there must be very little distortion of the crystals when observed in amorphous ice. This justifies tlm use of electron micrographs to determine crystal packing of the fibres. We shall now discuss the longitudinal packing in each of the four crystal types.

(i)(a) Ad5 Pisa crystals This type of crystal has a longitudinal period of 376 A, comprising two rows of heads perpendicular to the long axis and separated by 100 A with finer striations parallel to the long axis (Fig. 8). Two longitudinal packing models can explain these profiles, but only one of them (B) is consistent with the occurrence of a single band of heads at the end of the crystal, without apparent projecting material beyond. Such a model suggests that the length of the fibre in those crystals is roughly the same as the long axis of the crystal, i.e. about 370 A. (i)(b) Cleaved Ad2 fibre crystals Figure 9 shows an image of an ice-embedded crystal of cleaved Ad2 fibre together with the filtered image. The unit cell of this projection corresponds to that of the bc plane, i.e. is consistent with that of the [100] projection (with the 2-fold axis in the plane of the paper perpendicular to the long axis of the crystal) or because of the peculiarity of the hexagonal ab lattice, also of a projection along Ill0]. The difference would be that the former

582

C. Devaux et al.

{1"~=,=,,,,,,,,,=~

~

Model

A

~

Model B

Crystal '

'

'

'

ond

Figure 10. Axial density profile of cleaved Ad2 fibre crystal obtained fi-om the end (E) of the crystal shown in Fig. 9. Two alternative longitudinal packing models are shown that are both compatible with the profile. Model A implies a fibre length of 300 A and model B a length of 355 h.

should have a 90 ° angle and true mirror symmetry, while the latter would have an angle of about 92-3 ° (assuming fl= 92-7°). In fact, the measured angle from the image is 91-3 ° , so that the exact identity of this projection remains ambiguous. Also shown in Figure 9 is the one-dimensional projection of the density onto the long axis of the crystal, which is very useful for visualizing the longitudinal packing. These crystals are characterized by a set of four dark bands 55 A apart, which in the filtered image are clearly resolved into rows of fibre heads, laterally spaced by 39 A. The bands are numbered consecutively as H1, H2, H3 and H4. Between two sets of four lines (separated by the 540 A periodicity), two weaker bands are clearly visible, each of which can be resolved by filtering into two bands separated by 30 A. These bands are also visible in negatively stained images (Devaux et al., 1984). We interpret these as due to the basal (N-terminal) ends of the fibre and will be referred to as Nl, N2, N3 and N4. This is supported by images in which the extremity of a microcrystal is visible (see Fig. 10), where only two of the four weak bands remain; namely, N2 and N4. The separation of these two bands is the same as that between H1 and H3 (or equivalently H2 and H4). Thus, there are two possible solutions to the longitudinal packing both illustrated in Figure 10: either (A) heads H2 and H4 connect with ends N2 and N4 giving an overall fibre length of 300 A or (B) heads H1 and H3 connect with ends N2 and N4, respectively, giving an overall fibre length of 355A. This ambiguity can be resolved only by consideration of three-dimensional packing models (see below), by data from other sources on the fibre length or by analogy with the packing model of the uncleaved Ad2 crystals. In the accompanying paper, Ruigrok et al. (1990) find by negative staining a bimodal distribution of lengths

(310 A and 370 A) for the uneleaved adenovirus fibre and a unimodel distribution for the cleaved fibre (365 A). Only the short form of the uncleared fibre is observed to have an N-terminal biob. They argue that the true length is 370 A but flexibility in the N terminus can lead to shorter lengths under certain staining and grid conditions. Unfortunately, this does not resolve the question of the length of the fibres in the crystals and hence determine the correct longitudinal packing model, as it could be that the crystal environment also modifies the confbrmation of the N terminus. A test of the proposed model is its ability to explain observed crystal defects in the longitudinal packing. In one case, the distance between two successive sets of four dark bands (heads) rises to 650 A with tile appearance of a third weak line between. In another case, the distance is reduced to 440 A with only one weak band remaining. Both these defects can be explained by either packing model. In the one-dimensional projection shown in Figure 9 there is an additional reproducible feature; namely, another weak band between rows H4 and N1. This, we think, is most likely attributable to a bulge somewhere along the length of the shaft, possibly, for" instance, in the region of residues 300 to 350 where the model described by Green et al. suggests the existence of extra loops. (i)(c) Uncleaved Ad2 fibre crystals Figure 1 ] shows a projection perpendicular to the long axis of the uncleared Ad2 fibre crystals, a filtered image and a one-dimensional projection. This image has important similarities with Figure 9 of the cleaved fibre crystals except that the unit cell length is 615 A instead of 540 A and the four" planes of heads are separated into two pairs of two, separated by 230 A. However, in both crystal forms the lateral spacing between heads is 39 A, the separation of the closely spaced planes of heads is 55 A, and in this projection (which is probably the [100] projection, as the measured angle is 90-2 ° ) the heads in different planes are aligned along the c-axis (cf. Fig. 9). In the uncleared crystals there are two single weaker bands just either side of the two pairs of planes of heads (marked Nl and N2 in Fig. ll). From this comparison it seems likely that the packing in the two crystal forms is related. Indeed, both longitudinal models A and B can easily be adapted to explain the density profile (Fig. 12); this can be achieved simply by sliding apart planes H2 and H3 of heads. In doing this one keeps the assumption that the weak bands have their origin in N-terminal blobs. However, consideration of the crystal ends where it is apparent that bands N2 and H2 occur, but not H1, strongly supports packing model B (as shown in Fig. 12) and an implied fibre length of about 370 A. (i)(d) AdS sawtooth crystals A plausible interpretation of the packing of fibres in Ad5 sawtooth crystals has been given by Green et

Crystals of Adenovirns Fibre. I

583

Figure 11. Top: Electron micrograph in amorphous ice of uncleaved Ad2 fibre crystals. The axial unit cell dimension is 615 A. Bottom right: Filtered image of the well-ordered region (F) obtained by correlation averaging. Bottom left: Axial density obtained by projecting the scanned image perpendicular to the long axis, showing the 2 pairs of 2 dark bands of heads (H1, H2 and H3, H4) and the 2 weaker bands presumed to be the N-terminal ends (N1 and N2). The crystal end (E) was used for the l-dimensional projection shown in Fig. 12.

al. (1983) using negatively stained images. Images in amorphous ice of these crystals (Fig. 13) confirm the main features described and give reliable values for the a and c unit cell edges (see Table 2), but in

~

t

I

e--,,,,- 4

!

I

I

Crystal end

Figure 12. Axial density profile of uncleaved Ad2 fibre crystal obtained from the end (E) of the crystal shown in Fig. ll. The longitudinal packing model shown implies a fibre length of about 370 A.

general do not show ,as high resolution detail as negatively stained images. In particular, the fine striations 30 A apart that form a very prominent "ladder" in negatively stained images are only exceptionally observed in amorphous ice images and then only weakly. Images obtained in negative stain have been found to be very sensitive to radiation dose; overexposure causes shrinkage and distortion of the lattice. Figure 13 shows a filtered image of a low-dose, negatively stained specimen with cell parameters very close to those obtained in amorphous ice. A reproducible feature of these images is a nearly regular beading pattern along the length of the fibre tails, the beads being about 32 A apart. It should be noted that the distance between heads composing the antiparallel pair of fibres is about 360 h. (ii) Three-dimensional crystal packing (ii)(a) The dimensions of the fibre molecule According to the model described by Green et al. (1983), the heads of Ad2 fibre comprise three times 181 residues (401 to 581), which would occupy a volume of about 7-3 x 104 A3 (assuming a partial

584

C. Devaux et al. By a similar argument, the tails (3 x 400 residues) would occupy a volume of about l'5x l05 A 3, implying a uniform cross-section of 600A 2 (i.e. equivalent diameter of 28 A) for longitudinal model A (shaft length 250 A) or 500 A z (i.e. equivalent diameter of 25 A) for longitudinal model B (shaft length 300 A). These values for the shaft diameter are consistent with the transverse radius of gyration of the fbre of 10(_+l)A measured by solution neutron scattering (data not shown) and the value of 28(_+5) A measured by electron microscopy (Ruigrok et al., 1990).

Figure 13. Images of Ad5 sawtooth crystals. Top: Crystal in amorphous ice. Middle: Crystal negatively stained with uranyl acetate. Bottom: Filtered image of low-dose negatively stained crystal (not the same as in the middle panel) obtained by correlation averaging. Note the oeeun~nce of beads (size about 32 A) along the length of the fibre tails in the filtered image.

specific volume of 0"73 cma/g). Since the heads have a maximum length of 55 A (this is the separation between layers of heads in, for instance, the cleaved Ad2 fibre crystals), this implies a uniform crosssection of at least 1327 A 2 equivalent to a circle of diameter 41 A or equilateral triangle of side 55 A. In the accompanying paper, Ruigrok et al. report values of 56(+_4)A for the diameter of the fibre head and 49(_ 5) for its length as determined from negative-strain electron microscopy.

(ii)(b) Three-dimensional packing in cleaved Ad2 fibre crystals The space group C2 implies that there is a multiple of four (4n) complete fibres in the unit cell. Furthermore, the one-dimensional projections (Figs 9 and 10) suggest that the planes of heads have identical packing densities. This, combined with the space group operations, implies a multiple of eight complete fibres in the unit cell as the centring operation cannot relate heads in different planes. Previous measurements of the density of cleaved Ad2 fibre crystals and the specific volume of the fibre imply that there are 40 to 49 polypeptides in the unit cell, i.e. 20 to 25 dimers or 13 to 16 trimers (Devaux et al., 1984). Thus, the only reasonable possibilities are 24 dimers ( n = 6 ) or 16 trimers (n = 4), both corresponding to 48 polypeptides. It is now known that the fibre is a trimer (Van Oostrum & Burnett, 1985) but, on the basis of the crystals alone, this conclusion can be reached only by elimination of the dimer possibility because of the extreme difficulty of being able to pack together so many fibres. Indeed, the packing density of fibres in the unit cell is exceptionally high. The fl'action of the unit cell volume occupied by protein is about 65% in the case of cleaved Ad2 fibre crystals (VM = 1"95 ha/dalton) and 60% (Vu = 2"11 ha/dalton) in the case of Ad5 Pisa crystals (see below). Here, Vu is the parameter defined by Matthews (1977) as the ratio of the volume of the unit cell to the molecular weight of protein. Values as low as that of the fibre are fairly common for small proteins but not for oligomeric proteins of molecular weight higher than 100,000 (the lowest reported by Matthews is 2.14 Aa/dalton for a form of pyruvate kinase). The packing density of the fibres in the crystals is thus unusually high and this is probably related to the long fibrous nature of the molecules, which permit rod-like close packing. In order to solve the three-dimensional packing problem for cleaved Ad2 fibre crystals, we have to place four independent fibres in the asymmetric unit. On the assumption that each plane of heads has the same packing density, this requires two independent fibres in each of planes H1 and H2. A further assumption makes use of the observed regular hexagonal lattice of the ab plane. The most likely reason why this should arise is as a result of hexagonal packing of identical objects. Thus, we

Crystals of Adenovirus Fibre. I propose that each plane of heads is based on a trigonal lattice such that each trimeric head has three neighbours (Figs 14 and 15). This arrangement is consistent with the centring operation of the space group and accounts for two of the independent fibres. The unit cell dimension then implies a centre-to-centre distance of the heads of 44"8 A. The next step is to relate the hexagonal lattice of' heads in plane H1 with that in plane H2. First, we note that images of the kind shown in Figure 9 have the heads in different planes in register forming axial columns of four heads separated laterally by b/2 = 39 A. (For the purposes of this discussion, we assume that the projection shown in Fig. 9 is, in fact, the [100] projection, see discussion in section (d)(i)(b), above). From this we conclude that the hexagonal lattice of plane H2 is related to that of H1 by a translation of 55 A along c (the interplanar spacing) and a shift along a (to be determined), but no displacement along b. In other words, the assumption of open-hexagonal packing of heads in the ab plane leaves only one remaining degree of freedom in the complete determination of the arrangement of the heads; namely, the relative shift between layers HI and H2 along a. This will be determined by the need to accommodate the tails that interpenetrate the layers of heads. It turns out to be remarkably difficult to devise realistic packing arrangements of the heads and tails fulfilling all the constraints if one assumes simple geometrical forms (e.g. cylinders and ellipsoids). In fact, both components may have distinctly triangular crosssections (in the case of the tail, tiffs may twist around due to supercoiling) and thus permit tighter packing. For the longitudinal packing model A (total fibre length 300 A), each layer contains four heads and four tails. For model B (total fibre length 355 A), each layer contains fbur heads and eight tails. From the estimates of the cross-sectional areas of the heads and tails given in section (d)(ii)(a), above, this would imply occupation of 74% and 90% of the unit cell cross-sectional area for models A and B, respectively. Figure 14 shows schematically the packing in the layers of heads (assumed to have a triangular shape) for uncleared Ad2 fibre crystal models A and B as well as tbr Ad5 Pisa crystals (see section (d)(ii)(c), below). Attempting to connect the heads and tails in a full threedimensional packing model is very difficult to envisage for model B, because of the very high lateral packing density, so much so that from packing considerations this model seems less likely. Figure 15 shows a possible scheme assuming longitudinal model A, for the packing in each of the principal planes, with the four independent fibres marked (A, B, C and D). Overlap of tails in the middle is avoided by both tilting (i.e. tails not perpendicular to the layel~ of beads) and by having a nonorthogonal ac cell, both of which lead to the breaking of the trigonal symmetry. In Figure 15 it has also been assumed that all fibres tilt in the same way.

585

(b)

Figure 14. Schematic packing of heads and tails in the ab plane for (a) cleaved Ad2 fibre crystals (longitudinal model A), (b)cleaved Ad2 fibre crystals (longitudinal model B); and (c) Ad5 Pisa crystals. In all diagrams the tails (filled circles) and heads (filled or open triangles) are the same size, but the hexagonal cell parameter of the Ad5 Pisa crystals is 12% smaller. The cent,re-to-centre distance is 44.8 A for the cleaved Ad2 fibre crystals and 39"4 A for the Ad5 Pisa crystals, and the side length of the head is 56 A (see Results section (d)(ii)(a)).

586

C. Devaux

et al. i

.t

d

J N\K

~f

\f

~f

\f

J~,.

Ha

",,[ ~f

"xf

H3

J\

Nf

~,f

H4

=

[

I

:

N2

(b)

(o) N3

N4

( ( (

(

OC

"N/ "Nf J\

J\

~/

Nf

J\

J\ ~f

J\ ~f

J\ ~f

2\ ~f

J\ ~f

ix, ~f

J\

J\

b e plone

I)|0 ne

(c)

::

o b plone

Figure 15. A diagram showing a 3-dimensional packing model for cleaved Ad2 fibre crystals satisfying longitudinal packing model A (overall fibre length 300 A). Fibres in the ac plane are shown tilted with respect to the c axis and with an N-terminM blob. (a) Packing of a single layer in the ac plane. The 4 independent fibres in the asymmetric unit are labelled A, B, C and D, and the complete unit cell can be envisaged as corresponding to 2 such layers related by the centring operation of the C2 space group. (b) Projection along the a axis (this corresponds to the filtered image shown in Fig. 9. (c) Pseudo-hexagonal packing of heads and tails in the ab plane (e.g. layer HI).

(ii)(c) Three-dimensional packing in AdS Pisa crystals

Very similar arguments to those in the preceding section can be applied to the packing in Pisa crystals. Since the volume of the Pisa unit cell is only 3040 nm 3, smaller than the volume of 16 trimers (about 3680nm3), the cell must contain eight

trimers giving a volume fraction of" protein of 6 0 % and a VM value of 2-11 A3/dalton. Again, assuming a trigonat packing of heads in the ab plane, the centreto-centre distance of the heads is now only 39"5 A (compared to 44-8 A for cleaved Ad2) due to the smaller hexagonal lattice parameter. The sequence homologies between Ad2 and Ad5 fibre are such

Crystals of Adenovirus Fibre. I

that it is very likely that the two heads have similar shape and size. A head with a cylindrical shape would not allow different centre-to-centre distances for Ad2 and Ad5 fibres, but it is possible tbr heads of a distinctly triangular cross-section (Fig. 14). With the longitudinal packing shown in Figure 8 one has to accommodate in each of the two planes of heads as many tails as heads (Fig. 14). Again, using the cross-sectional estimates in section (d)(ii)(a), above, this would imply that in the layers of heads, 91% of the unit cell cross-sectional area is occupied. This is an unusually high packing density but very similar to that implied by longitudinal model B for cleaved Ad2 fibre crystals. In the preferred longitudinal packing model for Ad5 Pisa crystals (Fig. 8, model B) fibres are essentially contained within one unit cell with very little penetration into neighbouring unit cells along the c direction. It may be possible, therefore, for new growth in the c direction to occur by nucleation of a layer of heads on a previous layer but with a misorientation of 60 ° or 120°, thus preserving the pseudo-trigonal symmetry but not the orientation of the true monoclinic cell. This could explain the kind of disorder observed in X-ray diffraction patterns of the Pisa type of crystals (see section (c)(ii)(b), above). 4. Discussion

In this paper we have presented a number of biochemical results on the stability 9 f the adenovirus fibre and an analysis by electron microscopy and X-ray diffraction of four types of adenovirus fibre crystals. A simplified and rapid two-step procedure comprising selective precipitation with polyethylene glycol, followed by DEAE-Sepharose chromatography, has been used to obtain preparations of both Ad2 and Ad5 fibre that are homogeneous in both SDS/PAGE and isoelectric focusing. Previously, when less care was taken in the purification, the fibre was often cleaved after Tyrl7, or was heterogeneous in isoelectric focusing even when apparently homogeneous in SDS/PAGE. The adenovirus fibre is particularly resistant to denaturation since it remains trimeric in 8 M-urea at 37°C. It is also generally resistant to proteolysis except for cleavage by chymotrypsin at Tyrl7. Only Ad2 fibre chains have a second cleavage site at Met448 after a prolonged treatment with chymotrypsin at high concentration. Despite this second cleavage, the oligomeric structure of the fibre remains intact and, in particular, the stability of the shaft remains. We have shown that there are no disulphide bridges in the fibre molecule. These observations lend support to the idea that the shaft of the fibre is stabilized by inter-chain hydrogen bonds as well as hydrophobic interactions, as proposed in the triple-helical model to be described in a forthcoming paper. Four types of adenovirus fibre crystals, two from Ad2 fibre (cleaved and uncleaved) and two from

587

Ad5 fibre (Pisa and sawtooth, both from cleaved fibre) have been characterized by electron microscopy in amorphous ice. Cleaved Ad2 fibre crystals and Ad5 Pisa crystals can give macroscopic crystals that diffract to at least 3"5 A, but are invariably twinned and thus unsuitable for structure determination. Analysis of the longitudinal molecular packing in the crystals suggests that the overall fibre length is between 350 and 370 A, although in the case of the Ad2-cleaved crystals a length of 300 A would be possible. However, in view of the length measurements obtained by Ruigrok et al. (1990) on single fibre molecules, we think that the evidence is very strong that the fibre has, indeed, a length of about 370 A, considerably longer than previously thought, although there is also evidence for some conformational flexibility in the N-terminal part of the fibre (Ruigrok et al., 1990). The packing density of the fibres in the crystals is very high (VM= 1"95 A3/dalton for the cleaved Ad2 crystals), so much so that one has to assume that the fibre heads are of distinct triangular crosssection in order to see how the unit cell can accommodate the required number of fibres while satisfying the space group symmetry and other packing considerations. The hexagonal lattice observed in the ab plane for cleaved Ad2 and Pisa Ad5 crystals is presumably related to the trimeric nature of the fibre heads. In the X-ray diffraction pattern from single crystals of cleaved Ad2 fibre, the most prominent feature is a very intense meridional intensity in the region of a Bragg spacing of 4.40 A. This must correspond to an important structural repeat along the fibre shaft. Furthermore, there are additional strong reflections and diffuse layer-lines, which suggests that the 4.4 A reflection corresponds to the sixth order of a repeat of 26"4 A, which again must represent some important helical repeat along the fibre shaft. Evidence from electron microscope measurements of the length of Ad3 fibre compared with that of Ad2 fibre suggest that the pseudorepeating motif in the fibre sequence corresponds to a distance of 13-2 A (Ruigrok et al., 1990). A model has been proposed by Green et al. (1983) for the structure of the shaft of the fibre. In this model, each polypeptide chain is folded in 44 short fl-strands perpendicular to the axis of the fibre connected by Ê-turns. The fibre would then be a dimer with the two monomers connected by their hydrophobic faces. In this model, a reflection at 1/4"7 A-1 is predicted, originating from the distance between fl-strands. This reflection is absent in our diffraction patterns and, furthermore, the fibre is a trimer rather than a dimer. A further deficiency of the model described by Green et al. (1983) concerns the length of the fibre shaft. The model would predict a maximum length of about 44x 4-7 A = 2 0 7 A (less if there was supercoiling), whereas data presented in this and the accompanying paper suggest a shaft length of nearer 300 A. This is possible only in a model in which the individual chains engage in intra-chain hydrogen bonds

588

C. Devaux et al.

rather than inter-chain hydrogen bonds, in which case the observed 4"40 A reflection could arise from the projection of the inter-chain distance on the axis. Such a triple-helical model will be investigated in detail in a future paper. We are indebted to B. Cortollezzis and E. Truche for excellent technical assistance. D. Petite did most of the cell culture for us in the Laboratoire de Virologie Mol~culaire (Lille) with strong support from P. Boulanger and J. C. D'Halluin. We thank V. Mautner for communication of unpublished data and discussions, J. Gagnon (C.E.N.G.) for help in sequencing and R. Wade (C.E.N.G.) for low-dose images of Ad5 sawtooth fibre crystals. We are grateful to the staff of Daresburv Laboratory (in particular Andrew Thompson) and LURE (in particular Thierry Prang~) for use of synchrotron radiation facilities. Finally, we thank R. Ruigrok tbr a critical reading of the manuscript.

References Adrian, M., Dubochet, J., Lepault, J. & McDowall, A. W. (1984). Nature (London), 308, 32-36. Anderson, C. W. & Lewis, J. B. (1980). Virology, 104, 27-41. Bhown, A. S. & Bennett, J. C. (1983). In Methods in Enzymology (Him, C+ H. W. & Timasheff, S. N., eds), vol. 91, pp. 451-455, Academic Press, New York. Boudin, M.-L. & Boulanger, P. (1982). Virology, 116, 589-604. Boulanger, P. & Puvion, F. (1973). Eur. J. Biochem. 39, 37-42. Caillet-Boudin, M. L+ & Lemay, P. (198.6). Electrophore.~is. 7. 309-315. Caillet-Boudin, M. L., St recker, G. & Michalski, J. C. (1989)+ Eur. J. Biochem. 184, 205-211. Chatellard, C. & Chroboczek, J. (1989). Gene, 81,267-274. Chroboczek, J. & Jacrot, B. (1987). Virology, 151. 549-554. Devaux, C., Zulauf, M., Boulanger, P. & Jacrot. B. (1982). J. Mol. Biol. 156, 927-939. Devaux, C., Berthet-Cotominas, C., Timmins, P+ A., Boulanger, P. & Jacrot, B. (1984). J. Mol. Biol. 174, 729-737.

Devaux, C., Caillet-Boudin, M.-L., Jacrot, B. & Boulanger, P. (1987). Virology, 161, 121-128. Earnshaw, W. C., Goldberg, E. B. & Crowther, R. A. (1979). J. Mol. Blot+ 132, 101-131. Geddes, A. J., Parker, K. D., Atkins, E. D. T. & Beighton. E. (1968). J. Mol. Biol. 32, 343-358. Green, N. M., Wrigley, N. G., Russell, W. C., Martin, S. Ft. & McLachlan, A. D. (1983). EMBO J. 2, ]357-1365. Grutter, M. & Franklin, R. M. (1974). J. Mol. Biol. 89, 163-178. H~rriss~, J., Rigolet, M., de Dinechin. S. !). & Galibert, M. (1981). Nncl. Acids Res. 9. 4023-4042. Ishibashi, M. & Maizel, J. V+ (1974). Virology, 58, 345-361. Jornvall, H. & Philipson, L. (1980). Ear. J. Biochem. 104, 237-247. Jornvall, H., Pettersson, U. & Philipson, L. (1974). Eur. J. Biochem. 48, 179-192. Kidd, A. H. & Erasmus, M. J. (1989). Virology, 172. 134-144. Lottpeich, F. (1985). J. Chromalogr. 336, 321-37 I. Matthews. B. W. (1977). In The Proteins (Neurath, H. & Hill, i%. L., eds). vol. 3, Academic Press, New York. Mautner. V. & Pereira, H. (]971). Nature (London), 230, 456-457. Norrby, E. (1969). J. Gen. Virol. 5, 221-236+ Pereira, H., Wrigley, N+ & Mautner, V. (1975). In Proceedings of the International Sympo+'ium on Macromolecules (Mano, E., ed.), pp. 451-463. Elsevier Publishing Company, Amsterdam. Pettersson, U. (1984). In The Adeno~,iruses (Ginsberg, H+ S+, ed+), pp. 205-270, Plenum Press. New York. Philipson, L., Lonberg-Holm, K. & Pettersson. U. (1968). J. Virol. 2, 1064-1075. Pieniazek, N. J+, Dlemenda, S. B., Pieniazek+ D., Velarde. J. & Luftig, R. 13. (1989). Nucl. Acids Res. 17, 9474. Ruigrok. R. W. H., Barge, A., Albiges-Rizo, C. & Dayan. S+ (1990). J+ Mol. Biol. 215. 589-596. Saxton, W. O., Pitt, T. J. & Horner, M. (1979). Ultramieroscopy, 4, 343-354. Tarr, G. E. (1977). In Methods in Enzymology (Hits, C. H. W. and Timasheff, S. N., eds), vol. 47. pp. 335-357, Academic Press, New York. Van Oostrum, J. & Burnett. R. M. (1985). J. Virol. 56. 43-)-448.

Edited by A. Klug

Structure of adenovirus fibre. I. Analysis of crystals of fibre from adenovirus serotypes 2 and 5 by electron microscopy and X-ray crystallography.

An analysis by electron microscopy in amorphous ice and X-ray diffraction of four types of three-dimensional crystals of adenovirus fibre is presented...
13MB Sizes 0 Downloads 0 Views