The structural proteome of Pseudomonas aeruginosa bacteriophage {phi}KMV

Differential mass spectrometry (MS) techniques after 1-D or 2-D polyacrylamide protein fractionation have been applied to the identification and characterization of proteins in bacterial lysates (Goodacre et al., 1998; Zhang et al., 2004), and they have recently been applied to study viruses (Roberts et al., 2004). With the advent of high-throughput sequencing, providing over 300 complete genome sequences of phages infecting different bacterial hosts, MS protein identification allows comparative genomics based on experimentally determined structural proteins and proper genome sequence annotation.

Bacteriophage φKMV is a strongly lytic, T7-like bacteriophage infecting Pseudomonas aeruginosa. In silico analysis of the 42 519 bp double-stranded DNA sequence reveals that this phage shares genome architecture with Escherichia coli phage T7 and contains 11 ORFs similar to those of the latter phage. Apart from the major capsid protein, which was identified by N-terminal protein sequencing, some putative structural proteins, including tail, connector and core proteins, have been hinted at by sequence similarity searches (Lavigne et al., 2003). In addition, phylogenetic analysis based on the putative RNA polymerase and on the DNA ligase revealed SP6 as the phage most closely related to φKMV (Lavigne et al., 2003, 2005). The completed nucleotide sequences of Podoviridae SP6 and K1-5 (phages infecting Salmonella typhimurium and E. coli K1-5, respectively) have allowed more in-depth comparisons (Dobbins et al., 2004; Scholl et al., 2004). These sequences and phage tail morphology seem to place SP6 and K1-5 in a distinct subgroup within the T7 supergroup. However, evidence such as the late localization of the φKMV RNA polymerase suggests a distinct replication strategy and gene regulation, compared to members of the SP6 group.

We used different approaches to analyse the structural proteome of φKMV by MS, to verify sequence annotation and to gain insights into the phage particle structure.

Phage propagation and purification.
Bacteriophage particles were purified using sequential step CsCl-gradient centrifugation as described by Sambrook & Russell (2001) in which φKMV has a buoyant density of about 1·50 g cm³. Before disruption of the phage particles by a triple freezethawing cycle, the Microcon-procedure (Millipore) was used to concentrate the sample 12-fold.

To visualize bands, gels were silver-stained as described by Shevchenko et al. (1996a) and subsequently scanned and analysed for band size using Un-Scan-It gel v5.1 software (Silk Scientific Corporation), compared to a standard low-molecular-mass protein ladder (GE Healthcare). Subsequently, heat-denatured samples (5 min at 95 °C) were loaded onto a standard 12 % SDS-PAGE gel. Samples for trypsin treatment (Shevchenko et al., 1996b) and MS analysis were picked from Coomassie-stained gels (Simply Blue Safestain; Invitrogen).

Liquid chromatography-MS (LC-MS) and data analysis.
Protein digests were analysed by electrospray ionization (ESI)-MS/MS on an LCQ Classic (ThermoFinnigan) equipped with a nano-LC column switching system. The trapped sample was separated over 55 min on an analytical column (Biosphere C18, 200 mmx0·05 mm i.d., 5 µm; NanoSeparations) using a linear gradient from 5 to 60 % (v/v) acetonitrile in water containing 100 mM acetic acid. The eluate from the analytical column was introduced by a nanoelectrospray device (ThermoFinnigan) and sprayed from a gold-coated fused silica emitter (5 µm i.d.; NanoSeparations). The mass spectrometer was operated in a data-dependent acquisition mode to automatically switch between MS (m/z 3001500 Thompson in centroid mode at a maximum injection time of 150 ms) and MS/MS acquisition on the three most intense precursor ions, controlled by Xcalibur 1.3 software. All MS/MS data were searched using Mascot (Matrix Sciences) and Sequest (ThermoFinnigan) against the GenBank non-redundant protein database (31 March 2005) and against a local database of all possible φKMV gene products.

For the Sequest parameters, a cross correlation value (Xcorr) was set at 1·8, 2·5 or 3·5 for singly, doubly or triply charged peptide ions, respectively. The delta correlation value (C_n) was >0·1, while the parent and fragment ion mass tolerance allowed ±3·0 and ±1 Da variation, respectively. Possible static and chemical modifications included were cysteine carbamidomethylation and oxidation of methionine, histidine and tryptophan.

Search parameters included all organisms, with a parent and peptide ion mass tolerance of ±3 and ±0·5 Da, respectively. The significance threshold was set at P0·05 and one missed trypsin cleavage was allowed.

To ensure validity of the identified proteins, single- and double-peptide protein identifications were re-examined with the de novo sequencing algorithm Lutefisk1900 v.1.3.2. (Taylor & Johnson, 1997), utilizing the database sequence option. In doing so, the de novo-derived sequence candidates were finally evaluated against the peptide sequence (as returned by Sequest and Mascot) entered in the Lutefisk database file. In case the program evaluated the database sequence as being as good as or better than the de novo sequences, then this constituted evidence that the corresponding single- and double-peptide protein identification was correct.

Whole-phage shotgun analysis (WSA).
Phage particles were disrupted by freezethawing in 6 M urea, 50 mM Tris/HCl, pH 8, and 5 mM dithiothreitol. Proteins were denatured and reduced by incubation at 60 °C for 1 h and subsequently alkylated by addition of an equal volume of 100 mM iodoacetamide in 50 mM ammonium hydrogen carbonate plus 3 additional volumes of 50 mM ammonium hydrogen carbonate for incubation at room temperature in the dark for 45 min. The protein mixture was then digested with trypsin (Promega) and the resulting peptide mixture was fractionated by reversed-phase HPLC coupled to the LCQ. The generated peptide ions were further mass-fractionated in the gas phase in the LCQ according to Spahr et al. (2001). The mass spectrometer was operated in a data-dependent MS/MS mode with a full scan mass window of m/z 50. Consecutive analyses were performed to cover a total mass range from m/z 400 to 1500.

Cloning, recombinant expression and lysozyme assay.
The gp36 ORF was amplified using tailed primers containing a BglII site and cloned into the pBAD-ThioTOPO expression vector (Invitrogen) containing a C-terminal His₆-tag for expression in E. coli (TOP10). gp36 expression in 50 ml exponentially growing E. coli cells (OD₆₀₀=0·6) was induced by adding 0·2 % arabinose followed by 4 h culture. Protein purification was performed by His₆-based affinity chromatography. Protein bands on SDS-PAGE were quantified by Un-Scan-It software (v. 5.1). Enzymic activity towards P. aeruginosa at pH 6 and I=0·1 in potassium phosphate buffer was performed and calculated exactly as described by Lavigne et al. (2004).

Two approaches were followed for identification of structural proteins by MS. First, phage particle proteins were separated by standard denaturing gel electrophoresis (Fig. 1). Visible bands were subjected to in-gel trypsin digestion and the resulting peptides were identified by LC-ESI-MS/MS on an LCQ system (ThermoElectron). In a second approach, whole phage particles were digested with trypsin and the resulting peptides were separated by reversed-phase HPLC and gas-phase fractionation in the mass spectrometer prior to MS/MS analysis as described in the WSA section of Methods.

(71K): Fig. 1. SDS-PAGE analysis of disrupted φKMV particles. Proteins derived from disrupted φKMV particles were separated on SDS-PAGE (12 %) after different steps of purification. After silver-staining, each band was excised and identified by MS. For the structure-related φKMV proteins, the relative amount of protein between bands remains constant in the different gels. In contrast, the disappearance of OprF after a second CsCl purification step exemplifies the relative decrease in the amount of contaminants. While a single purification step allowed identification of five φKMV proteins, purification and concentration allowed visualization and identification of additional structural proteins (gp37 and 40 after two CsCl steps and gp29 after further Microcon concentration).

While the SDS-PAGE approach separates proteins for analysis of multiple peptide mixtures (each from one individual protein), the WSA approach creates a single complex mixture of peptides from all structural proteins, which is separated/fractionated in two dimensions (based on charge and size) prior to identification.

The combined data of the two identification approaches allowed us to assign 12 φKMV phage structural proteins to annotated ORFs (Table 1) and to verify the corresponding ORF predictions, since the peptide coverage is spread throughout previously predicted protein sequences. In this manner, protein annotation of six in silico-predicted structural proteins is confirmed. In addition, five gene products (gp29, 38, 40, 47 and 48) having little or no similarity to known phage proteins can now be classified as structural proteins (Fig. 2).

Table 1. Mass spectrometry data for φKMV The proteins detected by MS are listed with their predicted (Mol. mass) and observed (Obs. mol. mass) molecular mass. For both the SDS-PAGE and WSA approaches, the number of identified peptides in each protein and the corresponding protein sequence coverage are indicated. gp29 was found at the electrophoretic front (EF), while other gene products could not be determined (ND) from the SDS-PAGE analysis.The identification analyses of the proteins (Mascot and Lutefisk files) can be verified in the supplementary table of data available with the online version of this paper.

(21K): Fig. 2. Comparison of the Class III genome organization in T7 and φKMV. The structure/assembly and release genes from φKMV and T7 are compared to each other, showing their relative positions in the genome (in bp). Light-grey boxes with black arrowheads indicate the structural protein genes from φKMV and the known structure-related genes from T7, while other ORFs are shown as white boxes. The structure-related lysin domains from both phages are shown in dark grey. The large frames highlight the differences between T7 and φKMV genome organization in this region. These include, from left to right: the absence of ribosomal frameshift in the capsid protein, the domain-switch of the lysin domain, the putative heteropolymeric tail fibres and two minor structural proteins at the right terminus of the genome.

During data analysis, peptide MS/MS spectra were compared to all theoretical peptide spectra deduced from all possible ORFs in the six frames of the φKMV genome sequence. This did not reveal previously unpredicted gene products, nor translational frameshifts like those observed in the major capsid protein of T7 (Condron et al., 1991).

Identification of proteins after SDS-PAGE separation
Phage particles were analysed by SDS-PAGE after one or two CsCl gradient purification steps and Microcon concentration using a 100 kDa filter (Fig. 1). Subsequently, visible protein bands were analysed by MS. It should be noted that LC-MS does not allow quantification of relative amounts of the identified proteins. Hence, visible protein bands on SDS-PAGE were used to verify the relative decrease of contaminating proteins during subsequent purification steps, while phage particle proteins remain in constant relative amounts. As shown in Fig. 1, MS analysis identified several P. aeruginosa proteins of the outer-membrane porin family. These include OmpC, F, G, H1, OprI and OprL and a type B flagellin protein. The number of different P. aeruginosa proteins decreased gradually in the consecutive purification cycles, exemplified most clearly by the disappearance of the OprF band after the second CsCl purification. Although the number of contaminating proteins decreased drastically, final phage samples still contained contaminating porins OprE and OprF, as well as the presumed outer-membrane protein PA2760. This is presumably due to the presence of P. aeruginosa membrane vesicles (between 50 and 250 nm), which are continuously discharged from the cell surface during bacterial growth (Mayrand & Grenier, 1989; reviewed by Beveridge, 1999) and probably some fractions may have a similar density/size as intact φKMV phage particles during CsCl centrifugation. Alternatively, aspecific association between the phage and cellular debris or a specific phage receptor/host interaction cannot be excluded.

In Fig. 1, the proteins identified after SDS-PAGE separation, including tail protein B, the headtail connector, the capsid and internal core proteins, are shown. Table 1 displays the number of identified peptides and their amino acid coverage within each protein. The most abundant protein in the SDS-PAGE gel, corresponding to the capsid protein (gp32), was previously identified by N-terminal sequencing (Lavigne et al., 2003) and is now confirmed by MS analysis. Because of its abundance, gp32 peptides are present throughout the gel (Fig. 1), most probably due to aspecific adsorption to the gel (in bands of higher molecular mass) and to partially degraded protein (in bands of lower molecular mass).

The recombinant C-terminal domain of gp36 (M₇₃₇ and E₈₉₈) was recently described as a functional peptidoglycan-degrading lysozyme with strong thermostable properties (Lavigne et al., 2004). For gp36, one of the phage proteins migrating around 94 kDa, MS analysis revealed a total of 38 significant peptides, originating from throughout the entire gp36 sequence with an amino acid coverage of 51·9 %. Additionally, to verify the enzymic activity of the entire gene product, ORF36 was cloned into pBAD Thio-TOPO (Invitrogen), expressed in E. coli and purified as a thioredoxin/gp36/V5/His6 fusion protein by His₆-based affinity chromatography. A mean yield of approximately 0·3 µg (gp36) soluble protein (ml cell culture)¹ was obtained. Enzymic activity of the purified recombinant protein was verified by turbidimetric assays on P. aeruginosa protoplasts. At optimal conditions, an activity of approximately 2820 U mg¹ was measured. Compared to the enzymic activity of the C-terminal end of this protein (gp36C, 20·9 kDa) described previously (Lavigne et al., 2004), the expressed gp36 protein (115 kDa) has a threefold higher activity, taking into account the molecular mass of both proteins and comparing the activity per molecule of enzyme.

WSA
In Table 1, WSA is compared to protein identifications from gel bands. Interestingly, the number of identified peptides within individual proteins bands (and thus amino acid coverage) in SDS-PAGE bands is considerably higher than in the WSA approach. Using WSA, all proteins seen in the SDS-PAGE-based approach analysis were found, but supplemented by three other protein species, i.e. the tail protein A (gp33) and two minor polypeptides, gp47 (10·6 kDa) and gp48 (6·9 kDa). Corresponding ORFs 47 and 48 are located at the extreme right end of the genome, downstream of the lysis cassette, and have no similarity to any protein present in the NCBI database. Since neither these proteins nor gp33 (±21 kDa) has a visible gel band on SDS-PAGE, it is clear that the optimized WSA allows the identification of peptides from small, low-abundance proteins. These identified proteins do not appear to be artefacts, since they are found after all three purification steps. Conversely, the disappearance of gp31 after a second CsCl centrifugation indicates the absence of this protein in the mature phage particle. However, this observation could be consistent with the possible role of this protein as a scaffolding protein, especially in view of its genomic localization compared to T7 (Fig. 2).

Two gene products (gp35 and gp39) are located within the cluster encoding structural proteins, but neither was identified by either technique in the final phage samples used for analysis. If gene annotation is presumed to be accurate, either the absence of these proteins in the mature phage particle or the presence therein at very low abundance is possible. For gp35 and gp39 the latter is expected to be the case.

Phage morphology as visualized by electron microscopy has so far served as a strong basis for classification, while DNA-based classification has proved difficult for diverse bacteriophage genomes. However, an alternative classification method based on protein similarity has been suggested (Rohwer & Edwards, 2002), and this method indicates that SP6 is the closest relative of φKMV, based on structural proteins. First, SDS-PAGE and MS analyses indicate that φKMV has no ribosomal slippage within the mRNA sequence of the capsid gene, as observed in SP6 and K1-5 as well. Indeed, sequence data also suggest this absence (Dobbins et al., 2004; Scholl et al., 2004). This either implies morphological differences in the phage head, compared to T7, or raises questions about the function of the minor capsid protein in T7 and other phages displaying ribosomal slippage during capsid gene expression (Condron et al., 1991). Indeed, T7-like phage gh-1 also does not exhibit ribosomal slippage (Kovalyova & Kropinski, 2003). Elucidation of the corresponding gp36 of φKMV as a structural protein containing a functional C-terminal lysozyme domain strongly suggests that gp36 has a structural role in the internal core/injection needle. Corresponding proteins gp35 of SP6 and K1-5 also have a proper lysozyme domain at the C-terminal. This element of the injection needle is a second feature by which the φKMV/SP6 group differs from T7, since other T7-like phages have a functional transglycosylase domain at the N-terminal end of gp16 (Moak & Molineux, 2000, 2004). In addition, both φKMV and SP6 seem to lack a corresponding protein to gp13 of T7, a protein essential for T7 infection, presumably acting as a degradable plug that keeps the DNA transfer channel closed until infection (Molineux, 2001). These elements imply differences in the expulsion mechanism of the internal core proteins of φKMV, compared to T7. Differences between φKMV and T7 may also be present in the tail fibre protein since this gene is usually located in front of the DNA maturase genes (φKMV ORF43; DNA maturase B) and lysis cassette (φKMV ORF44 and 45; holin and lysin, respectively). However, φKMV exhibits three smaller genes in this region, two of which were shown to be structural proteins (gp38 and gp40), hinting at a potential heteropolymeric tail fibre protein, responsible for adsorption. Although tail fibres of T7-like phages vary extensively, heteromeric tail fibres have not yet been observed. While phage K1-5 encodes two different tail fibre proteins, specific for K1 and K5 antigen recognition in E. coli (Scholl et al., 2001), SP6 contains a P22-like tail-spike protein in its genome.

Despite obvious features indicative of the relatedness of phages in the T7 supergroup, the unique genomic features of φKMV (e.g. genome structure and the absence of consensus phage promoter sequences) and the structural differences stated above suggest that φKMV should be considered as a separate entity within the T7-like Podoviridae.

Conclusion
MS analysis of purified phage particles allows genome-wide analysis of the structural phage proteins and provides useful complementary information to the available genome sequence. LC-MS/MS after SDS-PAGE separation allowed identification of the dominant structural proteins in φKMV with high amino acid coverage. Tryptic digestion of whole phage particles followed by reversed-phase LC and a subsequent fine-tuned gas-phase chromatography prefractionation step before MS/MS additionally revealed smaller, less abundant structural proteins. These methods have allowed us to experimentally establish the structural proteome in bacteriophages, allowing classification based on structural proteins, and to provide preliminary insights into phage particle architecture. For φKMV, the MS and sequence data suggest subtle structural differences to T7. These differences, also present in phage SP6, include domain reorganizations in the internal core and injection needle, as well as the absence of ribosomal slippage in the capsid protein of φKMV. The flexibility of the localization of the structural lysin, present on either the C terminus of gp15 or at the N terminus of gp16 (in T7, corresponding to gp36 in φKMV) and observed in other T7-like phages, such as gh-1, suggests that both proteins (gp15 and 16 in T7, and gp35 and 36 in φKMV) interact to form a protruding mechanism of injection.

K. H., B. R. and Y. B. are FWO postdoctoral and TWO IWT predoctoral fellows, respectively. V. V. M. is supported by the Howard Hughes Medical Institute and Human Frontiers Science Program. This work was financed by the Research Council of the K.U. Leuven (grant OT/04/30) and by the IQR-fund of the Laboratory of Gene Technology.