Search for Potential Vaccine Candidate Open Reading Frames in the Bacillus anthracis Virulence Plasmid pXO1: In Silico and In Vitro Screening

Abstract

A genomic analysis of the Bacillus anthracis virulence plasmid pXO1, aimed at identifying potential vaccine candidates and virulence-related genes, was carried out. The 143 previously defined open reading frames (ORFs) (R. T. Okinaka, K. Cloud, O. Hampton, A. R. Hoffmaster, K. K. Hill, P. Keim, T. M. Koehler, G. Lamke, S. Kumano, J. Mahillon, D. Manter, Y. Martinez, D. Ricke, R. Svensson, and P. J. Jackson, J. Bacteriol. 181:6509-6515, 1999) were subjected to extensive sequence similarity searches (with the nonredundant and unfinished microbial genome databases), as well as motif, cellular location, and domain analyses. A comparative genomics analysis was conducted with the related genomes of Bacillus subtilis, Bacillus halodurans, and Bacillus cereus and the pBtoxis plasmid of Bacillus thuringiensis var. israeliensis. As a result, the percentage of ORFs with clues about their functions increased from ∼30% (as previously reported) to more than 60%. The bioinformatics analysis permitted identification of novel genes with putative relevance for pathogenesis and virulence. Based on our analyses, 11 putative proteins were chosen as targets for functional genomics studies. A rapid and efficient functional screening method was developed, in which PCR-amplified full-length linear DNA products of the selected ORFs were transcribed and directly translated in vitro and their immunogenicities were assessed on the basis of their reactivities with hyperimmune anti-B. anthracis antisera. Of the 11 ORFs selected for analysis, 9 were successfully expressed as full-length polypeptides, and 3 of these were found to be antigenic and to have immunogenic potential. The latter ORFs are currently being evaluated to determine their vaccine potential.

Bacillus anthracis is the causative organism of the potentially fatal disease anthrax. Although primarily a disease of animals, anthrax can also affect humans, and depending on the route of infection, the consequences are sometimes fatal. While the disease is well characterized, it is only in recent years that we have begun to understand the molecular basis of B. anthracis pathogenicity (22, 31).

Fully virulent forms of B. anthracis carry two large plasmids, pXO1 (∼181 kbp) and pXO2 (∼97 kbp), which are considered to be major virulence determinants, since strains lacking either one of these plasmids are attenuated in most animal models (38). These two plasmids also distinguish B. anthracis from the other members of the closely related Bacillus cereus group of bacteria (23). Despite the important role of pXO1 and pXO2 in B. anthracis virulence and the recent availability of their sequences (44; GenBank accession no. AF188935), only a few essential virulence genes encoded by these plasmids have been identified and characterized so far. These genes include the pagA, lef, and cya genes encoding the principal virulence factor (the tripartite toxin), which code for the protective antigen (PA) and the lethal and edema factors, respectively, all located on pXO1 (38, 44), and the pXO2 genes capBCAD.

The PA is the major immunogenic component of the cell-free vaccines licensed for use in humans (4, 11, 19, 61, 63). An inherent limitation of these vaccines is the requirement for numerous immunizations. Additional somatic antigens (spore antigens or antigens expressed during the vegetative stage) appear to improve vaccine potency (9, 11). Identification of such novel antigens is therefore essential for development of second-generation B. anthracis vaccines.

Whole-genome sequencing of a microbial pathogen together with in silico and cross-genomic analysis is a powerful tool that allows prediction of genes coding for virulence factors or immunogens, which may be novel drug targets or vaccine candidates (16-18, 20, 51, 52). The first step towards genomic characterization of B. anthracis was the recent sequencing, assembly, and partial annotation of the pXO1 plasmid (from an isolate of the live veterinary spore vaccine strain Sterne [44]). Recently, 143 open reading frames (ORFs) were identified in pXO1, and these ORFs comprised only 61% of the plasmid DNA sequence. For about 70% of the predicted ORF products (108 of 143 ORFs), putative functions could not be assigned (based on sequence similarity alone) as significant similarity to sequences available at that time in open databases was not detected (44).

In order to gain further insight into the role of pXO1 in B. anthracis pathogenesis, the bioinformatics analysis of pXO1-derived ORF products was extended to include searches against diverse sequence databases, as well as sequence cluster and sequence motif databases (databases of protein signatures of families, domains, active sites, and secretion and anchoring signals). Comparative genomics was used to facilitate identification of unique B. anthracis proteins. This study focused mainly on secreted and/or surface-exposed ORF products and ORF products similar to proteins involved in microbial pathogenesis. Such criteria are considered relevant for identification of targets for induction of protective immunity and for identification of proteins involved in various steps of the infection process. Our in silico analysis resulted in a significant reduction in the percentage of totally unknown ORFs, from ∼70% (44) to ∼40%, and allowed selection of candidate vaccine genes based solely on computational predictions.

The first step toward assessment of the immunogenic potential of in silico-selected candidate ORF products generally involves the tedious process of cloning, expression, and purification prior to in vivo functional screening. Such an approach has been used successfully for selection of vaccine candidates for human pathogens like Helicobacter pylori (24), Neisseria meningitidis type B (50), Streptococcus pneumoniae (66), and Chlamydia pneumoniae (39). Here we describe a simple method in which the first step relies on in vitro translation of linear full-length DNA of the selected ORFs, followed by analysis of the reactivity of the ORF products with hyperimmune anti-B. anthracis antisera.

MATERIALS AND METHODS

In silico analysis. (i) Amino acid sequences of pXO1-derived ORF products.

The sequences of the 143 pXO1-derived predicted ORF products determined by using GenMark were obtained from the study of Okinaka et al. (44).

(ii) Functional annotation of ORF products.

Preliminary annotation of the 143 ORF products was based on a Psi-BLAST analysis (five iterations [1, 2]) performed with the nonredundant protein database. Protein domains were assigned by searching against the PFAM (6), SMART (56), and Conserved Domain Database (CDD) (64) databases and clusters of orthologous groups (COGs) (National Center for Biotechnology Information [NCBI] [60]) and by sequence motif analysis (E-motif [26]). A cutoff E value of 10⁻⁴ was used for all database searches. A TMPRED analysis (25) (submitted to BCM launcher, Baylor College of Medicine, Houston, Tex.) was carried out to identify transmembrane segments, and SignalP (43) was used to predict signal peptide regions. Lipoproteins were identified by using PSORT (40, 41), as well as by searching for lipobox signals (27, 42). Gram-positive anchoring and sorting motifs were identified by using the pattern search module of SEALS (NCBI) (62) or in-house Perl scripts. Paralogs were identified by using BLAST analysis with the 143-ORF pXO1 data set, as well as a draft sequence of the B. anthracis chromosome (Ames isolate; The Institute for Genomic Reseach, Rockville, Md.) (5). For selected ORFs, searches for functional assignments were performed by using the ERGO database (Integrated Genomics, Chicago, Ill.), the GenQuiz server (3), and the fold recognition server GenThreader (28). Unique B. anthracis ORF products were identified by BLAST sequence similarity searches against the nonredundant database (NCBI) and the unfinished microbial genome database (NCBI), as well as the recently published sequence databases for the related bacilli Bacillus halodurans (NCBI) and B. cereus 14579 (Integrated Genomics) and the sequence of toxin-encoding plasmid pBtoxis of Bacillus thuringiensis var. israeliensis (Sanger Institute, Hinton, United Kingdom).

In vitro analysis. (i) PCR production of T7 promoter-containing linear DNA sequences of selected ORFs.

DNA from the B. anthracis Vollum strain was a gift from Z. Altboum's laboratory at the Israel Institute for Biological Research. Individual pXO1-derived ORFs were amplified by PCR. All 5′ primers contained a common sequence (CGGAGAATTCTAATACGACTCACTATAGGTACCACCATG) coding for the T7 promoter and a Kozak sequence upstream of the specific sequence for amplification of the selected gene. The 3′ end of each selected ORF's sequence included a stop codon (TAA) followed by three unique restriction sites (SmaI, NotI, and BglII). The specific primer sequences for each ORF are shown in Table 1. The Expand High Fidelity PCR system (for genes up to 2,000 bp long) or the Expand Long PCR system (for longer genes), both products of Roche Molecular Biochemicals USA, were used. PCR products were verified by resequencing by PCR.

TABLE 1.

Selected ORF candidates for antigenicity studies

ORF	Expected mol mass of product (kDa)	Function	5′ primer sequence	3′ Primer sequence
13	151	Sir2, ADP ribosylation	CAAGAGATACTAGAAGATGCAATCG	TTACTGATTTTGCGCTAAAAATAAA
53	15	Metal-dependent hydrolase	CAAAATACAGTAACAACAGCG	TTAACGGAAATAGATTTGTCCCAT
54	41	S-layer protein	ACTGTTAAGGCTGATAACACAGACTC	TTACTACTTGCTCTCGACTTGAACATAC
65	18	Putative adhesin	GCAACAACAGCTGATACAAAACTA	TTATTAAGGTCCTGTTACAATAGAATATGTCAG
67	66	Invasin-like	CTTTTAACTGCCACATGC	TTATCAACAGGTGTATTTGGAATCTG
76	36	M peptidase	TTGGAAATGATATATTTTGCG	TTACGCCACCCCATTATTTG
79	130	Cell wall, repeat protein	GATACTATTCCAGTTCTAGTCTTTG	TTAGTCTAAATCAGACATATAGAAGTGACC
85	26	CAAX-like protease	AGTAGTGGTATTAGATATAAATCGG	TCAAGAAATCTGGAGTATAACAACAAT
90	73	S-layer, DNA interaction	GAAAGTGATTTAAAAAAAATAGTTCCAG	TTAACTGTTTGGTATTCTAAGTTTAGG
91	32	Phosphatase B	TTTAATTTTGATGTATCATTCC	CTATCTATGAACATGAAAAATACGG
130	24	Adhesin, zinc protease	CAAGAGCAAACATCTGAAAGTCATAC	TTAATGCGCCATCATCTCG

(ii) In vitro translation of linear PCR products.

Each ORF DNA was translated individually by using a coupled rabbit reticulocyte lysate transcription and translation kit for in vitro assays (TNT T7 Quick for PCR DNA; Promega, Madison, Wis.). The reaction mixture contained T7 RNA polymerase and 1 mCi of [³⁵S]methionine per ml. The reactions were carried out at 30°C for 1.5 h. The products of each TNT reaction were analyzed by sodium dodecyl sulfate (SDS)-10 to 15% polyacrylamide gel electrophoresis (PAGE); the gels were fixed in a methanol-acetic acid-H₂O (20:7:63) solution, soaked in Enhance (Sigma), dried, and subjected to fluorography. The ¹⁴C-labeled methylated protein molecular mass markers included myosin (220 kDa), phosphorylase b (97 kDa), bovine serum albumin (66 kDa), ovalbumin (45 kDa), carbonic anhydrase (30 kDa), trypsin inhibitor (20 kDa), and lysozyme (14 kDa) (Rainbow; Amersham Pharmacia Biotech).

(iii) Hyperimmune antisera.

Rabbit and guinea pig anti-B. anthracis (Vollum strain) hyperimmune sera, obtained following multiple injections of live B. anthracis spores, were gifts from Z. Altboum's laboratory. The control sera consisted of 1:1 mixtures of naïve rabbit and guinea pig sera. The titers of anti-PA and anti-core antigens (total vegetative protein antigens) were as follows: for the hyperimmune guinea pig serum, 1:128,000 and 1:25,600, respectively; and for the hyperimmune rabbit serum, 1:102,000 and 1:256,000, respectively.

(iv) Immunoprecipitation of TNT products.

Three- to five-microliter portions of individual ³⁵S-radiolabeled TNT reaction mixtures were each diluted to obtain a final volume of 100 μl in RIPA buffer (20 mM Tris-HCl [pH 7.5], 150 mM NaCl, 5 mM EDTA, 0.5% sodium deoxycholate, 0.1% SDS, 1% Triton X-100) and were reacted with 10 μl of each of the sera. Following 1 h of incubation at 37°C, antibodies were precipitated by using protein A-Sepharose (Sigma) (50 μl in RIPA buffer for 30 min at 4°C). The immunoprecipitated radiolabeled TNT products were boiled and analyzed by SDS-PAGE, as described above.

RESULTS AND DISCUSSION

The bioinformatics analysis of the pXO1 ORF products was expanded to include extensive sequence similarity searches against unfinished microbial genome sequence databases, as well as the recently available sequences of related genomes. We also carried out domain and motif assignment studies, studies of functional assignments based on orthology, and cellular localization studies. The detailed analysis together with systematic evaluation and cross matching of the data resulted in a significant increase in the number of fully or partially annotated proteins, as well as identification of ORFs with putative relevance to vaccine design and/or B. anthracis pathogenicity (Table 2).

TABLE 2.

Functions of pXO1 ORF products

ORF	Length of product (amino acids)	Locationa	Function		Unique
			Putative (this study)	Determined by Okinaka et alb
1	167	S/M
2	192	M	MDR permeasec
3	211	S/M			+
4	209	C	Repeat motif protein bdr7ac
5	106	C
6	166	C
7	602	S/M	Reverse transcriptase	Reverse transcriptase
8	861	C	Helicasec
9	75	S/M
10	363	S/M	Type I restriction enzyme homolog, methyltransferase
11	117	C	Repeat proteinc		+
12	230	S/M
13d	1,320	S/M	Sir2 + regulatory protein	Rhoptry protein
14d	564	C	Clp ATPase, LacIc
15	80	C
16d	572	C	Bacterial transfer complexc
17	47	S/M
18	315	C	Integrase	Integrase, recombinase
19	73	C			+
20	498	C	Repeat proteinc
21	81	C
22	91	C	Adenine nucleotide hydrolase fragmentc
23	461	C	IepA truncated	Maturase-related protein
24	132	C	Nucleotidyl transferase
25	57	C
26	64	C			+
27	94	C			+
28	87	C			+
29	274	S/M	ATPase DNA repairc
30	155	C	Cystatin-likec		+
31	198	C			+
32	327	S/M	Ribosomal protein SIc
33	259	S/M	Spectinomycin 3′ adenyltransferase
34	68	S/M	Transcriptional regulator
35	478	S/M	Transposase	Transposase
36	484	S/M	Transposase	Transposase
37	193	C	Histone acetyltransferase
38	276	C	RecA homolog, AbiDc
39	325	S/M	Transposase	Transposase
40	65	C	Repressor protein, phage transcription regulation
41	108	S/M			+
42	452	S/M	Truncated stage III sporulation protein
43	128	C
44	428	C			+
45	435	S/M	Truncated cell division protein FtsZ	FtsZ
46	119	C			+
47	201	C	Myosin heavy chain (repeats)c
48	75	S/M
49d	227	C	Hydrolase, histone acetylasec
50	317	C			+
51	492	C	Nitrogenase, dethiobiotin synthasec
52	149	S/M	Topoisomerasec		+
53d	131	S/M	Protease, ATP dependentc		+
54d	393	S/M	S-layer protein (N terminus)	S-layer protein (N terminus)
55	790	S/M	Conserved membrane protein
56	45	S/M			+
57	142	C			+
58	272	C	RepB/ParA-nitrogenase
59	477	C	Secretory protein kinase GSPII-E	Secretory protein kinase
60	310	S/M			+
61	236	S/M			+
62	57	S/M			+
63	67	S/M			+
64	68	S/M
65d	202	S/M	Cellular recognitionc
66	361	S/M
67d	682	S/M	Cellular adhesionc
68	181	M			+
69	135	S/M
70	437	C			+
71	85	C			+
72	101	C
73	89	C
74	111	S/M			+
75	116	S/M			+
76	304	S/M	M1 peptidasec
77	91	C			+
78	405	C	Truncated TrsE homologc
79d	1,222	M	Cell wall protein, adhesinc	Hypothetical hydrophobic
80	89	S/M	Transcription regulatorc
81	424	S/M	Transposon-related NlpDc	Transposon related
82	171	C
83	127	S/M	Phage-related proteinc
84	101	S/M			+
85d	227	M	CAAX-like protease	Hypothetical
86	381	M	Transporter	Hypothetical
87	160	M	Thioreduxin isomerase	Putative thioreduxin
88	82	C	Conserved protein, transcription factor
89	158	C	Truncated RepA
90d	652	S/M	S-layer plus DNA repair, repeats	Hypothetical
91d	280	M	Phosphatidylglycerol phosphatase B	Hypothetical
92	92	S/M	N-terminus of Has-A
93	366	S/M	Has-A lacks N-terminus	Has-A
94	296	S/M	Has-C	Has-C
95	443	S/M	NDP-sugar dehydrogenase	NDP-sugar dehydrogenase
96	274	C	Transposase	Transposase
97d	150	C	BS phage SPBC2, hypotheticalc
98d	410	C	Neisseria-Clostridium imported proteinsc
99d	164	C	Ribosomal protein Zf-A20c
100	62	C
101d	170	C	DNA binding, RepAc
102	80	C
103	317	C	Integrase	Integrase
104	60	C
105d	67	C	Truncated AbrB
106	126	S/M			+
107	809	S/M	LF	LF
108	101	C	Hsp protein		+
109	99	C	PagR	PagR
110	764	S/M	PagA	PagA
111d	204	C	Domain 4 of PAc	Hypothetical protein in PA
112	288	C	GerXC	GerXC
113	492	C	GerXA	GerXA
114	359	M	GerXB	GerXB
115	193	C	Resolvase	Resolvase
116	925	S/M	Transposase	Transposase
117	132	C			+
118d	150	C	Histidine kinase (truncated)c	Unknown
119	475	S/M	AtxA	AtxA
120	190	S/M	Transposase	Transposase
121	57	S/M	Truncated Apt	Apt
122	800	S/M	EF	EF
123	37	S/M			+
124	55	C	C terminus of hemolysin II
125	125	C			+
126	151	C
127	214	S/M	Transposase	Transposase
128	84	C	Transposase
129	137	S/M	Transposase	Transposase
130d	237	S/M	Adhesin, enzyme	Hypothetical
131	349	C	Truncated GTP pyrophosphokinase
132	361	C	Integrase	Integrase
133	485	C	Truncated RECD, lectin-like foldc
134	125	C			+
135	149	C
136	364	C	Response regulator	Response regulator
137	61	S/M	Hfq	Hfq
138	97	S/M	PagR-like	PagR-like
139	96	C	Truncated DsdB	Hypothetical
140	95	S/M
141	214	S/M	Nuclease	Thermonuclease
142	887	C	Topoisomerase III	Topoisomerase
143	150	C			+

^aS/M, secretion signal or transmembrane; C, cytoplasm.

^bData from reference 44.

^cThe function assignment is partial.

^dORF discussed in the text.

Our selection of vaccine candidate ORF products was guided by several criteria. The primary criterion was cellular localization (the presence of anchoring and/or secretion signals). Additional criteria were the potential relevance to virulence based on location in the pXO1 pathogenicity islands (PAI) and/or similarity to proteins previously implicated as proteins that are relevant to microbial pathogenesis (on the basis of sequence similarity or the presence of sequence or structural motifs typical of such proteins). The results of the analyses are summarized in Table 2, and details concerning selected ORF products are described below.

ORFs encoding putative secreted and/or surface-exposed antigens.

Systematic screening of pXO1 for ORFs likely to encode surface-anchored proteins was carried out as described in Materials and Methods. These ORF products include proteins with up to three transmembrane segments, a lipoprotein signal, a signature sequence motif typical of secreted and/or surface-associated proteins, and a gram-positive specific anchoring motif (e.g., the LPXTG motif responsible for covalent binding to peptidoglycans, interaction with choline residues or teichoic acid via C-terminal repeats, or noncovalent membrane anchoring to peptidoglycan via S-layer homology [SLH] domains) (10, 13, 32, 33, 38, 42).

Of the 143 ORF products, about 50% contain either a secretion signal sequence (cleavable or uncleavable) and/or a transmembrane segment(s) (Table 2). The ORF45, ORF130, and ORF141 products contain lipoprotein signal sequences. As expected, the secreted or anchored proteins include the tripartite toxin gene products, S-layer proteins, hydrolytic enzymes, additional putative virulence factors, and several unknown proteins. Only a single pXO1-derived protein, the ORF79 product, seems to have a gram-positive LPXTG anchoring signal.

ORFs from the pXO1 PAI.

The pXO1 PAI was previously determined to reside between ORF96 and ORF127 (44), both of which encode hypothetical transposases. Although this PAI does not exhibit all of the classical features of a PAI (21, 29), it is still considered a PAI since it is bordered by insertion elements and carries the major known virulence determinants. Given the documented importance of a PAI's contribution to virulence, deciphering the role of the 16 unknown or hypothetical proteins (of the 31 putative proteins encoded in this region) could increase our understanding of the involvement of the pXO1 PAI in B. anthracis pathogenesis. Clues concerning the putative role or function of putative gene products, previously identified as hypothetical proteins, are presented below (Table 2).

The ORF97 product is a 150-amino-acid cytoplasmic protein. The only significant hits in BLAST searches against the nonredundant database were to a larger (172-amino-acid) unknown protein from Pasteurella multocida (Expect = 4e⁻¹⁰; 48% positives) and to a hypothetical protein from a Bacillus subtilis temperate phage (SPBc2; YokK-like; 192 amino acids; Expect = 7e⁻⁶). The recently published draft genome of B. cereus ATCC 14579 (Integrated Genomics) seems to contain an unknown protein (RZC03206) whose size is similar (151 amino acids) and whose sequence is almost identical (Expect = 5e⁻⁵⁹; 81% positives, extending over the whole sequence).

The ORF98 product is a 410-amino-acid cytoplasmic protein with no clear paralog in the B. anthracis genome or a B. cereus ortholog. However, it is not unique to B. anthracis. Its N terminus (residues 1 to 236) shows significant similarity to the N terminus of a hypothetical protein encoded by the B. anthracis chromosome (Expect = 1e⁻³⁶; 53% positives) and to an unknown protein encoded in the B. cereus genome (RZC06077; amino acids 98 to 270; 45% positive residues; Expect = 5e⁻¹⁷). Its central 130 residues exhibit sequence similarity to an uncharacterized small protein from Clostridium acetobutylicum (a homolog of B. subtilis YUKE/YFJA; Expect = 2e⁻⁰⁵; 47% positives) and to YeeF from B. subtilis (Expect = 2e⁻⁰⁵; 48% positives). A single WD-40 repeat domain (PF00400 [56]) was detected between residues 130 and 144. A BLAST search against the nonredundant protein database revealed that the C-terminal residues (amino acids 336 to 399) show significant sequence similarity (Expect = 7e⁻⁸; 64% positives) to an 81-amino-acid imported hypothetical protein from Neisseria meningitidis 0855 (with the identical sequence motif YTWHHHQ.GRMQL), as well as to the C terminus of a hypothetical protein from Helicobacter pylori (both N. meningitidis and H. pylori are gram-negative human pathogens). The same region of the C terminus exhibits significant sequence similarity (Expect = 4e⁻¹⁰) to the C terminus of a larger (1,232-amino-acid) hypothetical protein from another human pathogen, Corynebacterium diphtheriae (ERGO database). All of the information described above may indicate that the ORF98 product is involved in pathogenicity. Moreover, in a recent study, Pallen (45) identified the ORF98 product as a member of the novel ESAT-6/WX100 superfamily (based on similar motifs found in Mycobacterium tuberculosis proteins), members of which may constitute a new gram-positive secretion system (in spite of the absence of a signal or anchoring sequence). As the Mycobacterium orthologs have been reported to be relevant to the virulence of pathogenic mycobacteria, the same may be true for the B. anthracis orthologs, particularly in view of the location of ORF98 within the PAI.

The ORF99 product (164 amino acids) is a cysteine-rich cytosolic protein. The only weak functional hit was to ribosomal protein S27e (Pfam domain). The C terminus of this protein (residues 136 to 159) appears to contain an A20-like zinc finger (Pfam domain), a motif known to mediate interleukin-1 NF-κB activation.

The ORF101 product, a 170-amino-acid cytoplasmic protein, is lysine rich and thus hits many repeat-rich Plasmodium falciparum-derived proteins in BLAST searches (e.g., O1 protein, a 706-amino-acid protein [Expect = 9e⁻⁰⁴]). Its C terminus (residues 124 to 166) resembles that of the helix-turn-helix protein CopG family (also known as RepA) responsible for regulation of gram-positive plasmid copy number (it could also be a helix-turn-helix and β-sheet DNA-binding protein, which is normally larger) and shows weak sequence similarity to an ORF product in partial transposon ISC1778. COG analysis showed a weak hit to a zinc ribbon protein, possibly a nucleic acid-binding protein.

The ORF105 product (67 amino acids) is a truncated version of AbrB (a central pleiotropic transcriptional regulator) that is missing ∼30 amino acids at the C terminus. Orthologous proteins are present in other bacilli. The full-size paralog encoded in the B. anthracis chromosome (not the ORF105 product) was recently shown to be implicated in regulation of the timing of virulence gene expression (53).

The ORF111 product (204 amino acids), previously characterized as a hypothetical 21.6-kDa protein, resembles a stretch in the C terminus of PA (ORF110 product; residues 616 to 763; Expect = 3e⁻¹⁷; 56% positives). This region corresponds to the receptor-binding domain of PA (domain 4; residues 595 to 735 [49]). In spite of the extensive sequence similarity, it is not known whether this protein competes with PA for receptor binding. It has an insertion in its N terminus, lacks most of PA's β-sheet 7 (based on sequence structure alignment [49]), has mutations in residues previously shown to be critical for activity (including residues in the two surface loops involved in receptor interaction), and has a unique 80-amino-acid C terminus extending beyond the PA-aligned region.

The ORF118 product (150 amino acids), previously defined as an atxA gene product, has a shorter (136-residue) paralog that is encoded on pXO2 (pXO2-61). Both proteins resemble the N-terminal half of the larger (372-amino-acid) histidine protein kinase from B. thuringiensis (part of Tn5401; Expect = 3e⁻²¹; 57% positives) and the N terminus of an unknown B. cereus 14579 protein (RZC06056; 340 amino acids; Expect = 5e⁻¹⁷; 56% similarity). According to the ERGO database, BLAST analysis of the ORF118 product resulted in a hit to the N-terminal half of a histidine kinase from Bacillus stearothermophilus (Expect = 9e⁻¹⁴). The aligned regions probably do not correspond to the active center of the molecule (which is located in the C-terminal half of the molecule). ORF118 may represent a gene fragment.

In conclusion, we were able to assign putative functions to 8 of the 16 PAI-located ORFs analyzed (Table 2). It is worth noting that none of the products of these ORFs appear to have significant functions with immediate relevance to virulence.

Products of pXO1 ORFs with putative functions of biological interest outside the PAI.

In addition to the PAI ORF products, we noted the presence of products of multiple pXO1-derived ORFs with putative assigned functions and cellular localizations relevant to virulence and/or to vaccine design (Table 2). We chose the 11 ORF products listed below (and in Table 1) as candidates for experimental evaluation.

The ORF13 product is a 1,320-amino-acid membrane-anchored protein (probably anchored via its N-terminal alpha helical segment). The sequence of the ORF13 product is not unique to B. anthracis; it has partial sequence homologs in streptococcal, Borrelia, and P. falciparum repeat-containing proteins. The ORF13 product was initially identified as a protein similar to rhoptry protein (which participates in malaria erythrocyte invasion) by Okinaka et al. (44). Most of the ORF13 product sequence aligns with residues 783 to 2056 of the rhoptry protein (gi|7458799; Expect = 5e⁻¹⁰; 36% positives, mostly due to the presence of repeats). The N-terminal part (amino acids 14 to 647) aligns with residues of a hypothetical protein from Streptococcus pneumoniae R6 (gi|15903136; Expect = 8e⁻¹³; 41% positives). CDD analysis revealed the presence of a Sir2 domain in the N terminus (amino acids 23 to 260; Expect = 5e⁻¹⁹). Sir2 proteins are NAD-dependent histone deacetylases that mediate transcriptional silencing in bacteria (15). Functional Sir2 proteins require an NAD-binding domain with a Rossman fold and a smaller domain containing a zinc ribbon motif (30). The Sir2 domain in the ORF13 product seems to lack the zinc-binding domain; however, the NAD-binding domain seems to be intact, retaining the two strictly conserved glycine residues in site A (binding the adenine-ribose moiety of NAD), as well as other fully or partially conserved residues typical of Sir2 proteins (34). The ORF13 product also appears to have the hydrophobic residues necessary for the formation of the B-site, accommodating the nicotinamide-ribose moiety. It retains three of the four conserved polar residues at the polar C-site, with a tyrosine rather than a histidine as the second polar residue (34). The C terminus (residues 829 to 840) seems to contain a bacterial regulatory protein motif (Crp family; global regulators known to regulate transcription in response to anaerobiosis and carbon source in Escherichia coli). From amino acid 700 on, the protein's putative secondary structure resembles the secondary structures of the membrane-binding domains (Hc) of clostridial neurotoxins (GeneThreader). Based on the analyses described above and the multidomain nature of this ORF product, it is difficult to determine whether the protein may act as a molecular switch mediator in DNA metabolism (67) or may belong to the expanding family of virulence-associated ADP-ribosylating proteins (46). Based on the data mentioned above, functional evaluation by experimental methods may be useful in resolving the role of the ORF13 product.

The ORF53 product is a 131-amino-acid protein with an amino acid sequence unique to B. anthracis. It may function as a putative metal-dependent hydrolase. This protein is devoid of a signal sequence, yet it could be anchored to the membrane via an N-terminal helix. In view of its uniqueness and putative assigned function, which is often virulence related (37), this protein was selected for further evaluation.

The ORF54 product is a 404-amino-acid S-layer protein, and no clues concerning its function have been found. This ORF product harbors a secretion signal sequence and three SLH domains (residues 48 to 90, 110 to 151, and 168 to 208) followed by a region with low complexity (residues 221 to 250). The N-terminal SLH domains of the ORF54 product show sequence similarity to those of other gram-positive S-layer proteins (32, 33, 38), while this protein's C-terminal segment seems to be unique to B. anthracis. We note that S-layer proteins are thought to be involved in pathogenicity of gram-positive bacteria (55).

The ORF65 product is a 202-amino-acid secreted protein. Sequence-based analyses did not reveal any clues about its function. According to GenThreader, the secondary structure of this protein resembles (with >90% certainty) that of the Bordetella pertussis virulence factor cell adhesin p.69/pentactin (14, 65). Based on these findings, the ORF65 product could be important for virulence.

The ORF67 product is a 562-amino-acid membrane-anchored (via its N-terminal helices) protein that is probably involved in protein-protein interactions. Part of the sequence resembles a cell adhesion domain from C. acetobutylicum (gi|15896336; residues 252 to 393; Expect = 4e⁻¹¹; 38% positives), which harbors tetratricopeptide repeats and is involved in protein-protein interactions (3). The next 100 C-terminal amino acids resemble the N terminus of the ORF65 product. Based on sequence and structure alignments (GeneQuiz), this protein seems to contain a region similar to the Pseudomonas aeruginosa invasin that mediates bacterial internalization. Thus, this ORF product could be relevant to pathogenesis, as well as to vaccine design.

The ORF76 product is a 304-amino-acid membrane-anchored (via an N-terminal helix) putative M peptidase (Pfam domain). The only known ortholog is a protein from C. acetobutylicum whose function is not known. Based on its putative biological function, together with the location of the ORF in a unique B. anthracis region within the plasmid, this ORF product could be a potential vaccine candidate.

The ORF79 product is a 1,222-amino-acid alanine- and serine-rich membrane-anchored protein with a distinct signal sequence, which is followed by six N-terminal transmembrane segments, several repeat domains, and a gram-positive LPXTG anchor. This protein does not have an ortholog in the related organisms B. halodurans and B. cereus. Its N terminus (residues 54 to 350) shows sequence similarity to a 567-amino-acid Bacillus firmus hydrophobic protein (gi|1813499; residues 54 to 443; Expect = 9e⁻¹⁶; 39% positives), as well as to the C-terminal part of a larger protein from Streptococcus cristatus (SsrpA; gi|6984160; aligned with residues 350 to 1181; Expect = 3e⁻¹¹). This protein is also similar to an S. pneumoniae cell wall surface anchor family protein (C-terminal 765 amino acids; Expect = 2e⁻¹⁰; gi|15901602), which in turn is similar to domains present in proteins such as the Yersinia enterocolitica invasin. An ATP/GTP-binding site motif (P-loop) is present between residues 235 and 275. Amino acids 520 to 860 resemble a B. pertussis virulence factor repeat protein. According to the COG database, this protein belongs to the family of methyl-accepting chemotaxis proteins. Based on the data described above and in view of its relative uniqueness, this ORF product should be evaluated by genetic studies to determine its involvement in pathogenicity.

The ORF85 product is a 227-amino-acid membrane protein, probably a type II CAAX prenyl endopeptidase. Type II CAAX endopeptidases are membrane-bound metalloproteases that are potentially involved in protein and/or peptide modification and secretion or, alternatively, in protection of bacteria against bacteriocins. This family of enzymes normally found in eukaryotes was recently also identified in prokaryotes and plants (48). In terms of sequence similarity, the pXO1-derived enzyme resembles hypothetical proteins from other gram-positive organisms. Its N-terminal half harbors three transmembrane segments (positions 9 to 31, 5 to 71, and 84 to 106). The amino acids at positions 135 to 225 match those in PF02517, an Abi phage abortive infection protein belonging to the CAAX protease family (Expect = 1e⁻²⁹). In view of its putative roles and localization the ORF85 product should be tested further.

The ORF90 product is a 652-amino-acid membrane-anchored protein that carries its three SLH domains in the N terminus (residues 45 to 210). The SLH domains are followed by five coiled-coil segments (residues 227 to 329, 335 to 377, 393 to 444, 455 to 482, and 526 to 552), a lysine-rich region (around amino acids 225 and 582),and a six-leucine zipper motif (residues 249 to 382). The C-terminal part is rich in myosin-like repeats. Sequence similarity searches resulted in a hit to a hypothetical protein from P. falciparum (residues 215 to 633; Expect = 4e⁻²⁸; 47% positives), and the same region also aligned with purine nucleoside triphosphatase from Sulfolobus tokudaii (gi|15922434) and proteins from the slime mold Dictyostelium (mostly due to the presence of repeats). The secondary structure of the ORF90 product resembles that of a contractile prokaryotic protein containing spectrin repeats (GenThreader). Other Pfam hits for the non-SLH moiety include proteins associated with transcription factor activity or DNA repair and expression regulation (i.e., exonuclease SbcC, chromosome segregation protein smc2, DNA mismatch repair protein MUT5, DNA repair protein RECN, etc). Similarly, COG analysis indicated that residues 218 to 643 correspond to an ATPase involved in DNA repair (Fts [57]). However, the ATP-binding motif and the signature sequence of this family of phosphodiesterases (12) are missing. Given its cell surface location, this protein could be involved in surface contact. Therefore, the precise role of the ORF90 product and its putative mode of interaction with DNA, if any, are unclear. Nevertheless, in view of its characteristic gram-positive SLH anchoring feature and the fact that some SLH proteins are highly immunogenic, it would be interesting to evaluate the immunogenic potential of this protein.

The ORF91 product is a 280-amino-acid putative membrane-anchored (via its four transmembrane segments) protein. Its assigned function (phosphatidylglycerolphosphatase B) is based on sequence similarities to predicted phosphatases from C. acetobutylicum and Clostridium difficile and other bacterial enzymes (gi|15893894; Expect = 2e⁻²⁷; 49% positives with the C. acetobutylicum enzyme), as well as on sequence similarity to the eukaryotic sphingosine 1-phosphatase (gi|13540569). COG database analysis assigned this protein to the membrane-associated phosphatase family, and CDD analysis identified a PAP2 family signature (amino acids 35 to 170; Expect = 4e⁻¹³) involved in hydrolysis of phosphomonoesters at an acidic pH. This family of enzymes has been proposed to be involved in microbial pathogenesis and thus should be inspected.

The ORF130 product is a 237-amino-acid lipoprotein. COG analysis classified this protein in the general category of predicted periplasmic secreted proteins. The closest sequence neighbor is a hypothetical B. subtilis 251-amino-acid protein residing in the AADK (aminoglycoside 6-adenylyltransferase)-SIGZ (RNA polymerase sigma factor SigZ-like protein) intergenic region. The sequence is nearly identical but contains an insertion consisting of about 20 amino acids close to the N terminus of the molecule. The ORF130 product sequence is also significantly similar to the sequence of AGR-C from Agrobacterium tumefaciens (resembling part of an ABC transporter), to the C-terminal part (residues 311 to 510) of a larger putative adhesion protein from the virulent organism S. pneumoniae M1 GAS, and to other zinc-binding bacterial lipoproteins (e.g., the longer AdcA protein from S. aureus). According to ERGO analysis, it could be a copper-containing amine oxidase responsible for oxidation of primary amines, diamines, and histamines. In eukaryotes a similar enzyme (VAP1) controls lymphocyte migration, regulating physiological trafficking and inflammation (54). The ORF130 product has a distinct histidine-rich region (residues 39 to 55). Phi-BLAST searches against nonredundant sequences identified several proteins with very similar histidine-rich motifs in their N or C termini; several of these proteins are hypothetical proteins from the spirochete Borrelia burgdorferi and from eukaryotes. An interesting hit is FimA of P. multocida. FimA is encoded by a type IV fimbrial adhesin subunit gene, an essential virulence gene in oral pathogens, and plays a role as an adhesin in iron transport in streptococci (59). In the B. anthracis chromosome, the same motif appears in an ORF coding for an ABC transporter. In the B. cereus ATCC 14579 chromosome it is present, in part, in a gene coding for a manganese transport system. Based on the information described above and in view of the documented importance of ABC transporters and adhesins in virulence and as efficient vaccines against other gram-positive organisms (7, 8, 36), the ORF130 product may be immunogenic and may play a role in B. anthracis pathogenesis.

Unknown and unique B. anthracis ORFs.

There are around 50 pXO1 ORF products whose functions remain unknown, but this should not be too surprising since in spite of the progress in microbial genome sequencing and the increasing numbers of annotated genes, approximately 30% of microbial genomes still consist of so-called hypothetical or orphan genes (18, 20). The availability of additional bacterial genome sequences (particularly the sequences of organisms belonging to the B. anthracis family, including B. thuringiensis and additional B. cereus strains) may shed light on orphan gene function (for example, the unknown ORF14, ORF16, and ORF49 products appear to exhibit extensive similarity to putative proteins encoded by the recently published sequence of pBtoxis, the B. thuringiensis var. israeliensis toxin plasmid [Sanger]).

The unknown pXO1 ORFs are in most cases unique to B. anthracis (no sequence neighbors have been found in both nonredundant and unfinished microbial genome databases), and they are dispersed throughout the plasmid and not necessarily concentrated in the PAI. Certain unknown ORFs appear as clusters (e.g., ORF26 to ORF31 and ORF60 to ORF64), with no clear indication of horizontal transfer. Some ORFs encode very short peptides and could be disregarded as they may not correspond to real genes (58) or may represent nonfunctional gene remnants. Moreover, the seemingly relative abundance of nonfunctional genes or pseudogenes may be a result of plasmid decay (35).

Recently, Pannucci et al. (47) studied plasmid pXO1 sequence conservation (by PCR and hybridization) and also assigned unique pXO1 B. anthracis ORFs. However, some of the unique ORFs assigned by these authors do not correspond to those assigned in this study. This may be a result of methodological differences.

Evaluation of antigenic potential of selected gene products.

Of the putative secreted, anchored, or membrane proteins with assigned functions relevant to virulence (Table 2), 11 candidates, listed in Table 1, were selected for a preliminary functional genomics analysis.

Here we describe the development of a screening system based on coupled prokaryotic-eukaryotic in vitro transcription and translation of each selected ORF, in which in vitro-expressed linear DNA elements were used. The polypeptides obtained in vitro were then directly evaluated for immunogenicity by studying their reactions with hyperimmune anti-B. anthracis antisera. This method is fast and relatively simple and does not involve cumbersome cloning and expression of each ORF, thus facilitating high-throughput screening. Moreover, omission of the cloning step was a distinct advantage as gene expression in a heterologous genetic background may lead in some cases to biased results due to toxicity of expressed products, misfolding, etc.

The coding sequence of each of the 11 selected ORFs was expressed with omission of only the signal peptide sequence (Table 1). In addition, the pagA gene coding for the B. anthracis PA was added as a positive control representing a well-established immunogen. Of the 11 candidate ORF products (not including PA), 9 were successfully expressed. In Fig. 1, lanes labelled TNT show the results of in vitro translation of each ORF. The two genes which were not expressed, ORF85 and ORF91, were determined by the bioinformatics analysis to encode membrane proteins. Eight linear expression elements generated TNT translation products exhibiting molecular weights close to the expected values (ORF13, ORF54, ORF65, ORF67, ORF76, ORF90, ORF130, and the positive control PA [Fig. 1]). We noted, however, that some of the TNT translation products appeared as a ladder of bands, probably as a result of premature termination (seen also for the positive control PA [Fig. 1]), and in a few cases the full-length product appeared as a minor band (e.g., ORF13). The TNT product of ORF53 had an apparent molecular weight which is lower than the theoretical mass (Fig. 1), probably due to distortion caused by the presence of a high concentration of proteins of similar size in the rabbit reticulocyte lysate. While ORF79 appeared to be expressed, most of the translation product was retained close to the origin of the separating gel (Fig. 1). This observation could be explained by the fact that this protein is highly hydrophobic and harbors six transmembrane segments in its N terminus, properties that may lead to the formation of aggregates. To test this hypothesis, the DNA sequence coding for the N-terminal transmembrane stretch was deleted, and only the sequence coding for residues 423 to 1042 was expressed (ORF79tr). Indeed, as shown in Fig. 1, ORF97tr generated a polypeptide product of the expected size.

FIG. 1.

Translation and immunogenicity of selected pXO1 ORFs. [³⁵S]methionine-labeled proteins were obtained from PCR-derived DNA by using an in vitro transcription-translation TNT kit and were analyzed by SDS-PAGE followed by autoradiography (lanes TNT). The labeled TNT products were immunoprecipitated by using anti-B. anthracis hyperimmune antisera, including rabbit sera (lanes R), guinea pig sera (lanes G), and naöve sera (1:1 mixture of rabbit and guinea pig naöve sera) (lanes N). The apparent molecular masses determined for the ORF products are as follows: ORF13 product, ∼150 kDa (discrete band with the slowest migration); ORF53 product, 13 kDa; ORF54 product, 41 kDa; ORF65 product, 20 kDa; ORF67 product, 66 kDa; ORF76 product, 35 kDa; ORF79 product, no discrete band observed (see text); ORF79tr product (residues 423 to 1042 of ORF79 product), 65 kDa; ORF90 product, 73 kDa; ORF130 product, 28 kDa; and PA, 83 kDa. The predicted molecular weights of the full-length ORF products are shown in Table 1; the expected molecular mass of the ORF79tr product is 65 kDa.

Antigenicity of the selected in vitro protein products.

The protein products were reacted with the hyperimmune rabbit and hyperimmune guinea pig anti-B. anthracis antisera (Fig. 1). Both hyperimmune sera exhibited high titers against B. anthracis toxin components, as well as against a bacterial extract representing total vegetative proteins (see Materials and Methods). The control serum consisted of a mixture of rabbit and guinea pig naöve sera.

Of the 12 TNT products tested, 6 were found to be immunoreactive. As expected, the positive control PA reacted with both guinea pig and rabbit antisera. Two gene products (the ORF67 and ORF76 products) reacted with both antisera; however, they also showed weak cross-reactivity with the naöve serum and therefore could not be identified as B. anthracis-specific immunogens. Distinct positive results were obtained for the following three translation products: the ORF54 product, which reacted only with the guinea pig antiserum; the ORF130 product, which reacted only with the rabbit antiserum; and the ORF90 TNT product, which reacted strongly with both antisera.

Two of the immunoreactive proteins (the ORF54 and ORF90 gene products) are S-layer proteins. Since some of the S-layer proteins are highly immunogenic, it was interesting to determine whether the immunoreactivity was due to the SLH moiety (an outcome of cross-reactions with other B. anthracis S-layer proteins) or due to the non-SLH unique sequence. Since the ORF90 gene product reacted with both sera and appeared to be a very potent immunogen, it was chosen for further study. The ORF90 gene sequence was truncated by removal of the stretch coding for the N-terminal three SLH domains (residues 1 to 210). Translation of the truncated gene generated a polypeptide having the expected molecular weight. The immunoreactivity of the truncated gene product with both sera appeared to be similar to that observed with the full-length gene product (Fig. 2). Thus, we concluded that for the ORF90 product, the C-terminal segment (which is unique to B. anthracis) contributes in a significant manner to its immunogenicity. It is interesting that of the 143 pXO1-derived ORFs, only 2 encode S-layer proteins, both of which were identified as immunogens in this study. The third immunogen reported here, the ORF130 product, is a member of a family of ABC transporters and adhesins shown to be involved in the virulence of several human pathogens. Indeed, representatives of this family were recently reported to be S. pneumoniae vaccine candidates (e.g., PsaA [36]).

FIG. 2.

Effect of truncation of the SLH domains on the antigenicity of the ORF90 product: SDS-PAGE autoradiography of the immunoprecipitation reactions for the full-length ORF90 product and the ORF90tr product (an N-terminal SLH-truncated TNT product; residues 211 to 652; calculated and observed molecular mass, 53 kDa) with anti-B. anthracis hyperimmune rabbit sera (lanes R) and guinea pig sera (lanes G). The control serum consisted of 1:1 mixture of rabbit and guinea pig naöve sera (lanes N). Equivalent counts of [³⁵S]methionine-containing polypeptide TNT products were used for the immunoprecipitation reactions for the ORF90 and ORF90tr products and PA.

In conclusion, the combination of the bioinformatics analysis and the efficient screening system reported here facilitated the selection and identification of previously unknown highly antigenic proteins. It is worth noting that 30% of the in vitro-expressed genes selected by our bioinformatics analysis proved to be immunogenic. It is possible that if a larger battery of antisera were used, the percentage could be increased even further. We noted that the immunoreactivity of the three antigenic proteins (the ORF54, ORF90, and ORF130 products) appears to be essentially similar to that of the known and very immunogenic PA, a major constituent of B. anthracis vaccine. Accordingly, these antigens are currently being evaluated to determine their potential to induce protective immunity.