A New Mutant Class, Made by Targeted Mutagenesis, of Phage PRD1 Reveals That Protein P5 Connects the Receptor Binding Protein to the Vertex

Abstract

Phage PRD1 and adenovirus share a number of structural and functional similarities, one of which is the vertex organization at the fivefold-symmetry positions. We developed an in vitro mutagenesis system for the linear PRD1 genome in order to make targeted mutations. The role of protein P5 in the vertex structure was examined by this method. Mutation in gene V revealed that protein P5 is essential. The absence of P5 did not compromise the particle assembly or DNA packaging but led to a deficient vertex structure where the receptor binding protein P2, in addition to protein P5, was missing. P5⁻ particles also lost their DNA upon purification. Based on this and previously published information we propose a spatial model for the spike structure at the vertices. This resembles to the corresponding structure in adenovirus.

Bacteriophage PRD1 is the type virus of the Tectiviridae family. It infects gram-negative bacterial hosts provided they harbor an IncP, IncN, or IncW conjugative plasmid that codes for the virus receptor. The characteristic PRD1 features are the linear double-stranded DNA genome with 5′ covalently linked replication priming proteins and a membrane that resides inside the icosahedral protein coat (2). The similarities of the PRD1 major coat protein fold, capsid geometry (T=25), vertex structure, and replication strategy to those of human adenovirus (10, 11, 15, 16, 20, 21, 34) have led to the surprising conclusion that these viruses share a common ancestry. An additional intriguing feature of PRD1 is the function of the internal membrane as a device to deliver the genome into the host cell in a process in which the spherical membrane transforms into a tube (tail) that is proposed to penetrate the cell envelope (3, 20, 24, 34).

The 15-kbp viral genome encodes about 25 structural protein species (8, 18). Disruption studies with guanidine hydrochloride (3) revealed that the major coat protein (P3) and a minor protein (P5) were released as soluble multimers, the rest of the proteins being precipitated with the viral membrane. The high number of membrane-associated proteins was surprising, and further studies have revealed that the structural proteins can be divided into three categories (in addition to the genome terminal protein): (i) those forming the outer icosahedral coat, (ii) those responsible for the integrity of the viral membrane, and (iii) those that are associated with the DNA delivery to the host cell (34). It seems that proteins in the last category are not involved in the particle assembly and that the number of these protein species is close to 10. The adsorption-DNA delivery proteins seem to be located at or near the icosahedral vertices (the fivefold symmetry axes) (20, 33, 34).

Extensive attempts have been carried out to select suppressor-sensitive PRD1 mutants to saturate the genome (26, 34). These mutants have been invaluable in assigning functions to the corresponding viral proteins. However, we have not been able to obtain amber mutations to all of the PRD1 genes, leading to a situation wherein ca. 10 genes are without mutations. Targeted in vitro mutagenesis is an obvious approach to address the rest of the genes. However, due to the linear form and the difficulties in dealing with the hydrophobic terminal proteins of the genome, there has been no success thus far in these attempts.

One of the genes for which no mutants have been isolated is gene V. It is a 1,023-bp gene encoding a 34-kDa protein P5. The gene sequence revealed that the protein contained an internal collagen-like domain (gly-x-y)₆, which could be cleaved with collagenase (6). This observation suggested an ancient evolutionary history for the collagenous protein architecture widely utilized in eukaryotic cells (1). The purified recombinant P5 protein (as well as the one isolated from the virion) is a soluble trimer. It is composed of an N-terminal smaller domain that has sequence similarity to the pentameric vertex protein P31 and a C-terminal larger domain that is responsible for the trimerization of the protein (J. Caldentey, R. Tuma, and D. H. Bamford, submitted for publication). These domains are connected by the collagen-like region, followed by a glycine-rich stretch of amino acids. In solution purified P5 trimers slowly form higher-order multimers and a heterocomplex with protein P31 (Caldentey et al., submitted). Also, yeast two-hybrid analyses have revealed interactions between the smaller N-terminal domain of P5 and protein P31 (M. Aalto, personal communication).

In this study we describe a site-directed mutagenesis system for PRD1. Using this technology we have generated amber mutations in gene V. P5⁻ mutants are not compromised in particle formation or DNA packaging but are deficient in DNA retention and delivery as well as in assembly of the receptor binding protein P2 to the fivefold vertex.

MATERIALS AND METHODS

Bacteria, phages, and plasmids.

The bacterial strains used in this study are listed in Table 1. Cells were grown in Luria-Bertani (LB) medium (35). When appropriate, tetracycline (10 μg/ml), streptomycin (100 μg/ml), or chloramphenicol (25 μg/ml) was added. PRD1 was propagated on Salmonella enterica serovar Typhimurium strain DS88 or on a suppressor strain, PSA(pLM2). The purification of the virus was done as described earlier (7). Wild-type and mutant viruses were labeled with ¹⁴C-labeled amino acids and purified as described previously (23). The plasmids and phages used are listed in Table 2.

TABLE 1.

Bacterial strains used in the study

Strain	Genotype/phenotype	Source or reference
E. coli K12
HB101	recA1 endA hsdS pro leu thi lacY galK xyl mtl1 ara14 supE44	35
JM107	endA gyrA96 relA1 hsdR17 supE44 thi Δ(lac-proAB) [F−traD36 proAB lacIqZΔM15] λ−	37
DH5α	F−recA1 endA gyrA96 relA1 hsdR17 supE44 thi Δ(lacZYA-argF) U169 φ80d lacZΔM15 λ−	35
HMS174	F−recA hsdR	17
HMS174(pLM2)
S. enterica Typhimurium LT2
SL5676	ΔH2 H1-i::Tn10(Tcs)	B. A. D. Stocker, Stanford University, Palo Alto, Calif.
DS88	SL5676(pLM2)	8
PSA(pLM2)	supE	28

TABLE 2.

Plasmids and phages used in the study

Plasmid or phage	Selective marker	Description (nt coordinate[s]) in PRD1 genome)a	Source or reference
pLM2	Kmr	Encodes PRD1 receptor	28
pJB15	Tcr	Encodes PRD1 receptor	22
pSU18	Cmr	Cloning vector	9
pJB500		pSU18+XXXI+V (4894–6309)	This study
pJB501		pJB500 BamHI in gene V (5305)	This study
pJB504		pJB501+lacZα (5305)	This study
pJB515		pJB500 TAG (5299)	This study
pJB521		pJB500 TAG (5734)	This study
pJB523		pJB500 TAG (5836)	This study
pJB525		pJB500 TAG (5896)	This study
PRD1 wt			31
PRD1[lacZα]-1	Blue on X-Gal and IPTG	lacZα fragment (6309)	This study
PRD1[lacZα]-5	Blue on X-Gal and IPTG	lacZα fragment from pJB504	This study
PRD1 sus690		4Qam, from pJB515	This study
PRD1 sus691		S150am, F151I, from pJB521	This study
PRD1 sus692		Q184am, from pJB523	This study
PRD1 sus693		S204am, L295I, from pJB525	This study
PRD1 sus539		Amber in gene II	34

^aAccording to Bamford et al. (8). 

DNA cloning.

Standard DNA cloning techniques were performed according to the method of Sambrook et al. (35). The genes and DNA fragments were amplified using Pfu polymerase (Stratagene) and specific primers with designed restriction enzyme recognition sequences. The region of PRD1 genome (nucleotide coordinates 4894 to 6309 [8]) containing the structural genes XXXI and V was amplified using the viral DNA as a template. The fragment was purified from low-melting-point agarose gel and ligated into pSU18 vector under the lac promoter, resulting in plasmid pJB500. Using pJB500 as a template, a BamHI site was generated using the QuikChange site-directed mutagenesis kit (Stratagene) at the beginning of gene V (position 5305 in the PRD1 genome), resulting in plasmid pJB501. Chromosomal DNA from Escherichia coli JE2571 (14), carrying the wild-type β-galactosidase gene, was isolated and used as a template to amplify the lacZα fragment. The fragment was cloned into the BamHI site of pJB501, and the resulting plasmid was named pJB504. Using plasmid pJB500, carrying the wild-type gene V, as a template the in vitro mutagenesis method was used to introduce an amber mutation into the beginning of gene V (5299) corresponding to the Q4 residue in protein P5. The resulting plasmid was named pJB515.

Construction of recombinant phages.

PRD1 genomic DNA with the covalently bound terminal proteins was isolated as previously described (25). The DNA was digested with PacI restriction enzyme cleaving the genome once, between genes V and XVII, at position 6306. The amplified lacZα fragment was ligated with the PacI-cleaved genome and transfected by electroporation (25) into JM107(pJB15) cells and plated on LB agar with 30 μg of 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-Gal) per ml and 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG). Three blue plaques were obtained, one of which was purified and named PRD1[lacZα]-1.

In another construction, the lacZα fragment was introduced into the virus genome by homologous recombination between the wild-type (wt) genome and plasmid pJB504. To obtain the recombinant virus, strain JM107(pJB15)(pJB504) was infected with wt PRD1 to make semiconfluent plates. The soft agar layer was transferred into a beaker, and growth was continued for 4 h at 37°C. The lysate was cleared by centrifugation, and the supernatant was plated on complementing strain JM107(pJB15)(pJB500). Three blue plaques were obtained; one was further purified on DH5α(pJB15)(pJB500) and named PRD1[lacZα]-5.

Homologous recombination was further used to exchange the lacZα fragment in the genome of PRD1[lacZα]-5 to an amber mutation constructed into the recombinant plasmid. The suppressor strain JM107(pJB15) containing plasmid pJB515 was infected with PRD1[lacZα]-5, and homologous recombination was allowed to occur during growth as described above. The cleared lysate was plated on the serovar Typhimurium PSA suppressor strain, and plaques were screened for a suppressor-sensitive phenotype by picking them on the lawns of PSA and DS88 hosts. One of the obtained amber mutant phages was named sus690. Amber codons were also introduced into three other locations in gene V: TAG codons were generated by in vitro mutagenesis in plasmid pJB500 in positions 5734, 5836, and 5896 in the PRD1 genomic sequence, resulting in plasmids pJB521, pJB523, and pJB525, respectively, and the mutations were introduced into the PRD1[lacZα]-5 genome. The corresponding recombinant phages were named sus691, sus692, and sus693, respectively.

Phage adsorption test.

The adsorption assay was, in principle, carried out as previously described (20, 23). Receptor-less (SL5676) and receptor-containing (DS88) cells were grown in LB medium to approximately 2 × 10⁹ CFU/ml (the optimal adsorption phase [23]). The binding of the wt PRD1, sus539 (P2⁻), and sus690 (P5⁻) particles to the cells was measured by using ¹⁴C-labeled particles. About 6,000 cpm (3.5 × 10⁻⁵ cpm/PFU) of labeled wt virus particles corresponding to a multiplicity of infection (MOI) of 1 were used. The number of the labeled noninfectious mutant viruses was calculated by using the protein concentration of the preparations. The 6,000 cpm of mutant particles used in the assay corresponded to an MOI of ca. 30. Bacterial cells (100 μl) were mixed with the labeled virus and incubated for 15 min at room temperature. The cells were collected (Heraeus Biofuge pico; room temperature, 3 min, 20,000 × g), washed twice with 100 μl of LB medium, and finally resuspended into 100 μl of LB medium. The supernatants and the pellet fraction were collected, and the radioactivity was measured by liquid scintillation counting.

Electron microscopy methods.

For thin-section electron microscopy serovar Typhimurium DS88 was grown to a density of 10⁹ CFU/ml and infected with sus690 phage stock (grown on PSA) using an MOI of 9. Samples were taken at 30, 55, 70, and 100 min postinfection and fixed with 3% (vol/vol) glutaraldehyde in 20 mM potassium phosphate buffer (pH 7.2). After 30 min of incubation at room temperature, the cells were collected by centrifugation, washed twice, and prepared for transmission electron microscopy as described elsewhere (4). The micrographs were taken with a JEOL 1200 EX electron microscope (at the electron microscopy unit, Institute of Biotechnology, University of Helsinki) operating at 60 kV. For negative-staining electron microscopy, the purified virus particles were stained on carbon-coated grids with 1% phosphotungstic acid (pH 6.5) for 15 s, blotted dry, and examined under electron microscopy as described above.

Analytical methods.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed as described earlier (30). The proteins were either stained with Coomassie brilliant blue or transferred from the gel onto a polyvinylidene difluoride membrane (Millipore) and then detected with antibodies. Polyclonal antisera recognizing proteins P5 (22), P2 (20), and P31 (34) and the monoclonal antibodies 6T56, 7T41, and 18T56 (detecting proteins P6, P7, and P18, respectively [22]) were used. The chemiluminescent detection of peroxidase-conjugated secondary antibodies was performed using the EZ-ECL Detection Kit (Biological Industries). The protein concentration of the virus preparates was determined according to the method of Bradford (13) using bovine serum albumin as a standard.

RESULTS

PRD1 can package extra DNA.

The ability of the PRD1 genome to express the lacZα fragment (393 bp) and to package this extra DNA was tested by inserting it into the unique PacI site between the genes V and XVII. The lacZα fragment DNA, including the promoter, was amplified by PCR, and inserted into the PacI-cleaved genome. When plated on strain JM107(pJB15) producing the complementing O fragment of LacZ, this phage was able to make blue plaques, revealing that the promoter of lacZ was functional in the virus genome. However, the titer of the recombinant phage was about 1 order of magnitude lower than the titer of the wt phage (1.6 × 10¹⁰ and 2.5 × 10¹¹, respectively). Restriction enzyme analysis of the resulting recombinant phage DNA, PRD1[lacZα]-1, confirmed that the fragment was inserted into the genome oriented in the opposite direction from the rest of the PRD1 genes (data not shown).

P5 is essential for virus infectivity.

We tested whether P5 is essential for virus infectivity. This was done by disrupting gene V by a lacZα insertion. The fragment was first cloned close to the beginning of gene V in plasmid pSU18 and then introduced into the virus genome by in vivo homologous recombination between the linear virus genome and the circular plasmid. Blue plaques were obtained, but the color was lost during plaque purification. Either the fragment was not stably maintained in the genome or the expression of the fragment was abolished due to mutations. Restriction enzyme digestions of the isolated recombinant virus genome revealed that the fragment was still in gene V (data not shown), favoring the second alternative. The titer of the recombinant phage, PRD1[lacZα]-5, was about 500 times lower on the noncomplementing DH5α(pJB15) strain than on the corresponding strain containing the complementing plasmid pJB500, suggesting that protein P5 is essential.

The lacZα insertion, which inactivated gene V, was further used as a marker when introducing the amber codon from plasmid pJB515 to the virus genome by in vivo homologous recombination. During the growth of PRD1[lacZα]-5 on a suppressor strain containing pJB515, the resulting recombinant phages representing the wt phenotype were enriched. The virus pool obtained was plated on strain PSA to suppress the amber mutation, and the plaques were screened by picking them on DS88 and PSA lawns. Three mutants out of 150 plaques tested (one of which was sus690 [see below]) were obtained, revealing that the recombination frequency and the rate of enrichment were fairly good. The titers of the mutants were determined on the nonsuppressing strains HMS174(pLM2) and HMS174(pLM2)(pJB500). The mutant phenotype could be rescued when protein P5 was provided in trans, the titers being approximately 2 × 10⁵ and 8 × 10¹¹ on the noncomplementing and complementing strains, respectively. This finding confirmed that the defect was in gene V. Accordingly, protein P5 was missing when the sucrose-gradient-purified mutant virus was analyzed by SDS-PAGE with Coomassie staining (data not shown). However, a small amount of P5 was detected on Western blots with polyclonal anti-P5 serum (Table 3 and Fig. 1). For this reason, decreasing amounts of purified wt and mutant virus particles were analyzed by Western blotting using polyclonal P5 antiserum. On the basis of this analysis the amount of P5 was about 50 times lower in the amber mutant than in the wt virus (Table 3). Except for P2, all other PRD1 proteins were present in wt amounts, as judged from SDS-PAGE gels stained with Coomassie brilliant blue or detected with available specific antibodies. The amount of P2 (62 kDa) was found to be about 50 times lower in the mutant compared to the wt virus using Western blotting and polyclonal P2 antiserum (Table 3).

TABLE 3.

Proportion of P5, the amber fragment of P5 (P5^*), P2, and P31 in mutant virus particles compared to the wt particle^a

Phage strainb	Proportion of:
	P5	P5*	P2	P31
wt	1		1	1
sus690	0.02		0.02	1
Phages expressing fragments of P5
sus691 (149)	0.01	0.1	0.05	1
sus692 (183)	0.01	0.01	0.01	1
sus693 (203)	0.01	0.1	0.01	1

^aThe amounts of the proteins were estimated from three independent Western blots of serial dilutions of the virus samples, as shown in Fig. 1. 

^bThe length of the P5 N-terminal fragments (in amino acids) is given in parentheses. 

FIG. 1.

Immunological identification of proteins P2 and P5 from serial dilutions of purified virus particles analyzed by SDS-PAGE. The highest amount applied to the gel was 20 μg of protein.

P5 stabilizes the interaction of the receptor-binding protein P2 with the particle.

Since the amount of the spike protein P2 in gene V amber mutant sus690 was diminished, the adsorption efficiency of the mutant was tested. wt and mutant particles missing protein P2 (sus539) were used as controls. The adsorption test was carried out by binding ¹⁴C-labeled virus particles both to the receptor-less SL5676 and to receptor-containing DS88 cells. The adsorption level of the wt particles to DS88 cells was ca. 50% and to the SL5676 cells was ca. 1.3%, corresponding well with the earlier reported values (20, 23). sus690 particles clearly displayed no binding to the cells (adsorption level of 0.6% to DS88 cells), which was also the case with the control mutant sus539 (adsorption level of 2.6% to DS88 cells). The results support the finding that P5 is needed for infectivity, most probably due to its role in stabilizing P2 interaction with the particle.

Gene V mutations do not interfere with particle formation or packaging of the DNA.

Nonsuppressing DS88 cells were infected with sus690, grown on the suppressor strain PSA, and samples were taken for thin-section electron microscopy at different time points postinfection. The mutant infection followed that of the wt one, yielding a large number of DNA filled particles at the end of the infection cycle (Fig. 2). Although almost all of the intracellular particles appeared to contain DNA, as seen in the thin-section electron micrographs (Fig. 2), only about 50% were filled after purification by rate zonal sucrose gradient centrifugation. This was estimated from the intensities of the slow- and fast-moving light-scattering bands in sucrose gradients representing the empty and filled particles, respectively (data not shown). In the case of the wt virus the proportion of filled particles is about 80% (5). Negative staining of the purified mutant virus revealed that the appearance of both the filled and the empty particles was identical to the corresponding wt particles (Fig. 3). Surprisingly, no tail structures were detected, although these structures are readily seen in the case of the mutant particles missing the spike protein P2 (20). The tail formation is associated with membrane transformation upon DNA release (3, 20, 34).

FIG. 2.

Electron micrographs of thin-sectioned DS88 cells infected with sus690. (A) Image obtained 30 min after infection. The arrows point to an adsorbed and an intracellular empty particle. (B) Image obtained 70 min postinfection showing a large number of peripheral filled particles. Bar, 300 nm.

FIG. 3.

Negative-staining electron micrograph of purified sus690 viruses showing filled (A) and empty, DNA-less (B) particles. Bar, 100 nm.

The amino-terminal domain of P5 is attached to the particle.

Genes XXXI (381 nucleotides [nt]) and V (1,023 nt) are organized as an operon (OL1 [19]) in the PRD1 genome. Gene V can be divided into two parts, which are separated by a linker region encoding a collagen-like helix, a flexible area, and a glycine-rich stretch (reference 6 and Fig. 4). Comparison of the two genes at the nucleotide level showed that gene XXXI has 48.1% identity to the sequence encoding the amino-terminal part and 35.7% identity to the sequence encoding the carboxy-terminal part of P5. The sequences encoding the amino- and carboxy-terminal parts of protein P5 display 43% identity with each other. This shows that these two regions of gene V could be originating from gene XXXI or vice versa. In a recent study, the carboxy-terminal domain of P5 has been proposed to be involved in the trimerization of the protein (Caldentey et al., submitted). In that study it was shown that, after digesting the purified protein with collagenase into two parts, the 135-residue amino-terminal part appears to be monomeric and the 205-residue carboxy-terminal part retains its trimeric form. We made DNA constructions with amber mutations in gene V producing three amino-terminal fragments of different lengths and recombined them into the virus genome in order to see whether the truncated molecules will be degraded or rescued by the subsequent assembly to the particle. The shortest truncated polypeptide (mutation S150am) was designed to be 149 amino acids long, containing the N-terminal domain and the collagen-like helix; the second (Q184am) ended before the glycine-rich stretch, and the third (S204am) contained the entire linker region (Fig. 4).

FIG. 4.

Late operon OL1 of PRD1 (nt 4894 to 6309 [8, 19]). Transcription is from left to right. The horizontal positions of the genes indicate the different reading frames. Gene XXXI encodes the penton protein, and gene V encodes the vertex protein. The function of open reading frame d (orf d) is not known at the moment. Gene V has 5′- and 3′-terminal parts which are separated by a linker region. The linker region can be divided into three parts on the basis of the amino acids it encodes (gray boxes). The first (on the left) contains the collagen-like helix, the one on the right has eight glycines in a row, and the area between these two parts seems to be flexible, with many glycine and serine residues. The positions of the created amber mutations are indicated with arrows.

The different truncated P5 polypeptides were overexpressed from plasmid constructions in HMS174 (nonsuppressor) cells. These constructions produced large amounts of recombinant proteins, as seen in SDS-PAGE gels with Coomassie brilliant blue staining. The apparent molecular weights of the amber fragments corresponded to those calculated from the truncated polypeptide sequence (data not shown). We examined the multimeric state of the amber fragments in conditions in which P5 retains its trimeric form. This was done by omitting the boiling step in SDS prior to loading the samples on top of an SDS-PAGE gel (22, 27). The amber fragments were detected only in monomeric and not in higher-molecular-weight forms, in contrast to the full-length P5, which was seen in this assay as migrating in both forms (data not shown).

The ability of the truncated P5 polypeptides to assemble into the virus particle was determined by analyzing the purified recombinant viruses sus691, sus692, and sus693 (S150am, Q184am, and S204am, respectively). Surprisingly, in sucrose gradients the mutants behaved like the wt virus, yielding 80% of filled particles, in contrast to sus690 (Q4am), which only yielded 50%. This indicates that the shortened P5 fragments are able to seal the particle, thus preventing DNA leakage upon purification. It was shown by Western blotting that the truncated P5 molecules assembled into the virus particle (Table 3). Low amounts of full-length P5 were also associated with the particles due to the leakiness of the amber mutations. The amount of the full-length polypeptide was determined to be the same as the amount of the truncated molecule in the case of sus692, encoding the middle-length truncation (Table 3). The shorter and the longer amber fragments (sus691 and sus693, respectively) were able to compete with the full-length P5 upon assembly, as can be seen in Table 3. The amount of the full-length P5 in sus691, sus692, and sus693 was clearly diminished compared to sus690r, meaning that the particle-sealing effect was achieved upon assembly of the fragments alone and not by the coassembly of the full-length polypeptide with the amber fragment. These results also suggest that P5 is oriented so that the amino-terminal portion is associated with the virus capsid.

DISCUSSION

Current progress in establishing methods for creating site-directed amber mutations within the PRD1 genome opens up the possibility for testing the essentiality of the genes as well as for obtaining information about the functions of genes for which no classical mutations are available. The general usefulness of the targeted mutagenesis method in PRD1 will depend on (i) the recombination frequency in different genomic regions and (ii) the phenotype of the resulting mutant, since the enrichment of the mutant viruses during recombination is based on the selective advantage of the suppressed phenotype over the mutant one. We are also developing methods to delete entire genes. This will in the future extend the genetic analysis of PRD1 to applications where the leakiness of amber mutations presents a major obstacle. It has been shown with LacI-LacZ fusion proteins containing amber codons that the wt background in a nonsuppressing Salmonella strain can be as high as 2% (12). This finding is in agreement with the amount of the full-length P5 found in the mutant sus690 particle (1/50 of that found in the wt).

The high sequence similarity between the DNA regions encoding the amino- and carboxy-terminal parts of P5 and the vertex penton protein P31 suggests a single gene origin for all of these elements. Both P31 and P5 form homomultimers, and these interact with each other (6, 34; Caldentey et al., submitted). Using the same structural elements utilizing gene multiplication to conserve interactions when building increasingly complex structures is an intriguing phenomenon.

We have previously used the PRD1-adenovirus analogy successfully in revealing the PRD1 penton protein (P31) location at the fivefold-symmetry positions of the capsid (34). The adenovirus spike protein (IV) is an elongated trimeric structure with a distal receptor-binding knob (32). It extends from the pentameric penton base protein (III), creating a symmetry missmatch considered to be important in forming a metastable structure utilized in receptor binding, virus entry, and DNA delivery (36, 29). The vertex structure of PRD1 is much less characterized. Mutational analysis has revealed that in the absence of the penton protein P31, proteins P2 (the adsorption protein) and P5 are also absent, in addition to the peripentonal ring of coat protein trimers (34). Protein P5 was previously shown to be accessible on the virus surface (22). Isolated protein P2 was shown to be monomeric and readily bound to the PRD1 receptor on the host cell surface (20).

We have demonstrated here that in the absence of protein P5 (sus690), the penton protein P31 is present and the adsorption protein P2 is absent and that there is no effect on the virion assembly or on DNA packaging. It has been shown earlier that, in the absence of protein P2, the viral particles are assembled and packaged but that upon storage release the DNA and that this release is related to the membrane transformation into a tube (20). sus690 mutant phage appeared to lose the DNA upon purification, resulting in reduced amounts of filled particles. With negative-staining electron microscopy of the purified particles, we showed here that the membrane transformation to a tail structure is not enhanced in the mutant (as is the case with the P2-deficient mutant), indicating that the loss of DNA probably occurs through an opening in the vertex and not via membrane transformation. Taking all of this information into account, we propose a model wherein the trimeric P5 is associated with the pentameric P31 and in which P2 is the most distal component of the spike structure connected to P5. Thus, P5 would be functionally analogous to the adenovirus spike shaft, and P2 would correspond the adenovirus spike knob (and maybe part of the shaft). In both viruses there is a symmetry missmatch at the vertex that is important in the infection process. We have previously demonstrated that part of protein P2 is associated with the virus membrane fraction after removal of the major coat protein P3 and protein P5 with guanidine hydrochloride (3). Whether there is also a direct contact between P2 and the penton base protein P31 or whether this result is due to unspecific interactions of these proteins upon denaturation awaits the ongoing structure determinations of the virion and its spike components.

All of the N-terminal P5 fragments obtained here were found to be associated with the virion and alleviated the unstable DNA packaging phenotype of the P5 (and P2)-deficient particles. This indicates that the N-terminal portion of P5 is proximal to the virus surface. The leakage of DNA from P5-deficient particles also implies that P5 sits at the fivefold-symmetry position, preventing DNA release. We can assume a model for the DNA release wherein the process is initiated by the binding of P2 to the primary receptor followed by binding of P5 to a possible secondary receptor, leading to the subsequent triggering of the membrane tail tube formation. Again, there is a resemblance to the two-step adenovirus infection mechanism, in which the fiber protein first mediates attachment to cells via interaction with CAR (adenovirus and coxsackievirus receptor). The second interaction between the penton base and the cell integrins promotes the virus internalization.