†Present address: Department of Microbiology, University of Texas Health Science Center, 7703 Floyd Curl Drive, San Antonio TX 78229–3900, USA.
The pavA gene of Streptococcus pneumoniae encodes a fibronectin-binding protein that is essential for virulence
Article first published online: 7 JUL 2008
Volume 41, Issue 6, pages 1395–1408, September 2001
How to Cite
Holmes, A. R., McNab, R., Millsap, K. W., Rohde, M., Hammerschmidt, S., Mawdsley, J. L. and Jenkinson, H. F. (2001), The pavA gene of Streptococcus pneumoniae encodes a fibronectin-binding protein that is essential for virulence. Molecular Microbiology, 41: 1395–1408. doi: 10.1046/j.1365-2958.2001.02610.x
- Issue published online: 7 JUL 2008
- Article first published online: 7 JUL 2008
- Accepted 4 July, 2001.
Streptococcus pneumoniae colonizes the nasopharynx in up to 40% of healthy subjects, and is a leading cause of middle ear infections (otitis media), meningitis and pneumonia. Pneumococci adhere to glycosidic receptors on epithelial cells and to immobilized fibronectin, but the bacterial adhesins mediating these reactions are largely uncharacterized. In this report we describe a novel pneumococcal protein PavA, which binds fibronectin and is associated with pneumococcal adhesion and virulence. The pavA gene, present in 64 independent isolates of S. pneumoniae tested, encodes a 551 amino acid residue polypeptide with 67% identical amino acid sequence to Fbp54 protein in Streptococcus pyogenes. PavA localized to the pneumococcal cell outer surface, as demonstrated by immunoelectron microscopy, despite lack of conventional secretory or cell-surface anchorage signals within the primary sequence. Full-length recombinant PavA polypeptide bound to immobilized human fibronectin in preference to fluid-phase fibronectin, in a heparin-sensitive interaction, and blocked binding of wild-type pneumococcal cells to fibronectin. However, a C-terminally truncated PavA′ polypeptide (362 aa residues) failed to bind fibronectin or block pneumococcal cell adhesion. Expression of pavA in Enterococcus faecalis JH2–2 conferred > sixfold increased cell adhesion levels to fibronectin over control JH2–2 cells. Isogenic mutants of S. pneumoniae, either abrogated in PavA expression or producing a 42 kDa C-terminally truncated protein, showed up to 50% reduced binding to immobilized fibronectin. Inactivation of pavA had no effects on growth rate, cell morphology, cell-surface physico-chemical properties, production of pneumolysin, autolysin, or surface proteins PspA and PsaA. Isogenic pavA mutants of encapsulated S. pneumoniae D39 were approximately 104-fold attenuated in virulence in the mouse sepsis model. These results provide evidence that PavA fibronectin-binding protein plays a direct role in the pathogenesis of pneumococcal infections.
The gram-positive bacterium Streptococcus pneumoniae is a leading cause of sepsis, meningitis and middle ear infections (otitis media) in children, and of pneumonia and sepsis in the elderly or immunocompromised (Tuomanen et al., 1995). The pneumococcus colonizes the nasopharynx in about 10% of healthy adults and 40% of healthy children (Austrian, 1986), and can spread locally to colonize the alveolar epithelium and cause respiratory tract infection. Bacterial cells may invade the bloodstream from the nasopharynx via the cervical lymphatics or by crossing the alveolar capillaries, leading to bacteraemia, penetration of the vascular cell layer of the blood–brain barrier and meningitis.
The S. pneumoniae polysaccharide capsule, of which there are at least 90 serotypes, is antiphagocytic and a major virulence determinant (AlonsoDeVelasco et al., 1995). However, the current pneumococcal polysaccharide vaccine comprising 23 capsular serotypes is of limited clinical efficacy because of poor immunogenicity in the high-risk groups, especially infants (Paton, 1998). Accordingly, there is much interest in the possibility of including pneumococcal common antigens as partners in capsular-based or conjugate vaccines, or in producing a polypeptide-only based vaccine (Paton, 1998). In these respects, a number of pneumococcal polypeptide antigens have been suggested as potential vaccine candidates. These include hyaluronate lyase (Hyl), pneumolysin (Ply), neuraminidases (NanA and NanB), autolysin (LytA), choline-binding protein A (CbpA), pneumococcal surface antigen A (PsaA), pneumococcal surface protein A (PspA) (reviewed by Jedrzejas, 2001), and complement component C3-degrading protease (PhpA) (Zhang et al., 2001). At present, the more promising candidates are considered to be Ply, LytA or PspA, because inactivation of the genes encoding these proteins significantly reduces pneumococcal virulence in intraperitoneal challenge experiments in mice, while inactivations of nanA, hyl or cbpA do not have significant effects on virulence (Berry and Paton, 2000). Pneumolysin activates the classic complement pathway in the absence of specific antibody (Paton, 1996) and initiates nitric oxide production from macrophages contributing to tissue damage (Braun et al., 1999). During the first few hours of bacteraemia in the mouse model, it is thought that pneumolysin plays a role in preventing inflammation-based immunity, thereby allowing unchecked growth of pneumococci (Benton et al., 1995). PspA is a cell-surface-associated protein, the N-terminal domain of which binds lactoferrin (Hammerschmidt et al., 1999; Hakansson et al., 2001). PspA functions as an inhibitor of factor B-mediated complement activation (Tu et al., 1999), interfering with deposition of C3b and therefore reducing susceptibility of pneumococci to opsonophagocytosis. PspA is a member of the family of choline-binding proteins (CBPs) that are non-covalently attached to the pneumococcal cell surface (Yother and White, 1994). PspC (Brooks-Walter et al., 1999), also designated SpsA and CbpA, is structurally related to PspA and mediates adhesion of pneumococci to activated human lung cells and endothelial cells (Rosenow et al., 1997), binds the secretory component of secretory immunoglobulin A (S-IgA) (Hammerschmidt et al., 1997), and binds complement factor H (Dave et al., 2001). The major cell wall hydrolase LytA (autolysin) is also a CBP that is required for the separation of daughter cells during cell division, and for stationary-phase-associated or penicillin-induced cell lysis (Ronda et al., 1987).
As colonization of the nasopharynx by S. pneumoniae is a prerequisite for the development of pneumococcal disease, the mechanisms by which pneumococci adhere to and penetrate host tissues are important to define. A two-step model for pneumococcal adhesion has been proposed, involving recognition of different host cell glycoconjugates at each step (Cundell et al., 1995), but the bacterial adhesins remain largely uncharacterized. In other gram-positive cocci such as Staphylococcus aureus and Streptococcus pyogenes, binding to fibronectin has been shown to be important for adhesion to epithelial cells (Joh et al., 1994; Okada et al., 1997) and for subsequent invasion of host cells (Molinari et al., 1997; 2000; Okada et al., 1998; Dziewanowska et al., 1999). Although pneumococcal cells do not appear to bind soluble fibronectin, they adhere to immobilized fibronectin via a trypsin-sensitive mechanism (van der Flier et al., 1995). This preference for adhering to surface-bound fibronectin is also displayed by the related mitis-group oral bacterium Streptococcus gordonii, in which adhesion is mediated to a greater extent by a 259 kDa cell wall-anchored polypeptide designated CshA (McNab et al., 1996). We have been unable to demonstrate the presence of cshA-like coding sequences within the S. pneumoniae genome, but we have identified an orthologue of the fbpA gene in pneumococcus, which is located immediately downstream of cshA in S. gordonii. This gene, designated pavA in S. pneumoniae, encodes a protein with high sequence similarity to the S. pyogenes fibronectin-binding protein Fbp54 (Courtney et al., 1994; 1996). In this paper we show that inactivation of pavA in S. pneumoniae significantly reduces binding of bacteria to fibronectin and provide evidence that pavA expression plays a role in the pathogenesis of pneumococcal infections.
Genetic analysis of the S. pneumoniae pavA locus
Immediately downstream of the cshA fibronectin-binding protein gene in S. gordonii is a gene that we have designated fbpA (GenBank accession number X65164) encoding a polypeptide with 70% identity to Fbp54 fibronectin-binding protein of S. pyogenes (Courtney et al., 1994). To determine whether this arrangement of fibronectin-binding protein genes was present also in S. pneumoniae, a 1.6 kb fragment of S. gordonii DNA encoding a segment of the C-terminal amino acid repeat block of CshA (McNab et al., 1996) and a fragment (802 bp) encoding the C-terminal portion of FbpA were used as hybridization probes on blots of restriction enzyme-digested S. pneumoniaeR800 genomic DNA. The results obtained indicated that S. pneumoniae contained sequences hybridizing with fbpA but not with cshA (data not shown).
In order to isolate the pneumococcal fbpA orthologue, oligonucleotide primers were synthesized corresponding to well-conserved regions within the fbpA and fbp54 genes and an internal fragment (671 bp) was amplified by polymerase chain reaction (PCR) from S. pneumoniae R800 DNA template. Inverse PCR (Ochman et al., 1990) amplification of DraI- or BsmI-digested chromosomal DNA, circularized by ligation, was then utilized to isolate DNA 5′ or 3′ to this fragment. Sequence data were subsequently obtained for a 3250 bp region of pneumococcal DNA defined by DraI and BsmI sites. Within this locus were four open reading frames, the largest being of 1653 bp encoding a 551 amino acid residue protein with a predicted molecular mass of 63 297 Da and pI of 8.78 (see Fig. 1). The deduced amino acid sequence of the protein, which was designated PavA because of its inferred role in determining pneumococcal adhesion and virulence, was 74% identical to S. gordonii FbpA and 67% identical to S. pyogenes Fbp54. The polypeptide contains 62 Leu residues, 52 Glu residues and 44 Lys residues. The high preponderance of Leu residues at specific regions was previously noted for Fbp54 (Courtney et al., 1994) and a weakly repetitive β-strand–α-helix structure could be invoked. There were no particularly remarkable features, such as conserved amino acid repeat blocks, within the primary sequence. In addition, there was no conventional hydrophobic leader and signal peptidase cleavage site to direct export via the general secretory pathway, no Gram-positive bacterial cell wall anchorage motif LPxTG at the C-terminal end (Navarre and Schneewind, 1994), and no choline-binding signature sequences to promote cell surface retention (Yother and White, 1994).
The genetic structure of the pavA locus is shown in Fig. 1. Upstream of the pavA gene are a consensus −10 promoter sequence and Gram-positive −16 sequence (Voskuil and Chambliss, 1998), common to many pneumococcal gene promoters (Sabelnikov et al., 1995), and a potential −35 sequence. The pavA promoter −35 region appears to overlap the putative promoter region of a divergently transcribed operon comprising orf1 (product uncharacterized) and dgk encoding a protein with 78% identity to the diacylglycerol kinase of Streptococcus mutans (Chen et al., 1998). Downstream of pavA is a strong rho-independent transcriptional terminator and a convergently transcribed gene, lytB, encoding a novel murein hydrolase (Garcia et al., 1999). Northern blots of S. pneumoniae R800 RNA probed with a PCR-derived 32P-labelled pavA fragment of 1.65 kb (see Fig. 2) revealed that pavA was expressed as an approximately 1.7 kb mRNA (data not shown). This transcript size was in good agreement with the genetic structure of the locus, and with predicted sites of transcriptional initiation and termination of pavA(Fig. 1).
The R800 pavA sequence was 98% identical to the nucleotide sequence of the pavA gene homologue identified within the unfinished genome sequence of S. pneumoniae type 4 strain that is available for blast search at The Institute for Genomic Research (TIGR) web site (http://www.tigr.org). A total of 64 clinical isolates of S. pneumoniae representing 21 different capsular serotypes were examined for the presence of the pavA gene using PCR amplification with primers designed to amplify the complete coding region (see Experimental procedures). A single fragment of 1.65 kb was amplified from every strain and serotype tested (Fig. 2). Restriction fragment length polymorphism (RFLP) analysis of these fragments with a variety of enzymes demonstrated identical patterns for BsmI, BspHI, DdeI and HhaI cleavage (results not shown). These data indicated that PavA protein is probably highly conserved.
PavA is a pneumococcal cell-surface protein
To detect expression of PavA protein by pneumococcal cells, we utilized antibodies that were generated in rabbits to recombinant PavA. Western immunoblot analysis of proteins present within soluble (cytoplasmic) or cell envelope fractions prepared from wild-type non-capsulated (Cps–) strain R800, revealed a major 62 kDa band that localized to a greater extent within the cell envelope fraction (Fig. 3, lanes 1 and 2). Similar immunoblot patterns were seen with extracts obtained from capsulated (Cps+) strains D39 (type 2) and NCTC10319 (type 35 A) (data not shown). Immuno-electron microscopy of cells, reacted pre-embedding with affinity-purified PavA antibodies and protein A-gold particles, revealed localization of PavA to the bacterial cell surface (Fig. 4A and B), with binding of the anti-PavA antibodies to the region of the capsule. Ultra-thin sections of these specimens demonstrated relatively uniform distribution of the gold particles around the cell surface (Fig. 4C and D). Immunogold labelling of samples post embedding indicated the presence of PavA protein on the cell surface, but also within the cytoplasm of the bacteria (Fig. 4E and F).
PavA mediates adhesion of pneumococci to fibronectin
A pavA mutant of S. pneumoniae R800 was generated by allelic exchange (double cross-over) of a 1073 bp cassette containing a chloramphenicol acetyltransferase (cat) gene (Claverys et al., 1995) inserted into the ApaI site present within the pavA gene (Fig. 1). Representative chloramphenicol-resistant mutants of R800 (Cps–) and D39 (Cps+), designated strains UB1339 and UB1341, respectively, were isolated and characterized in further studies. Mutant UB1339 cell protein extracts, analysed by Western immunoblot analyses, were devoid of PavA polypeptide (Fig. 3, lanes 5 and 6), as were extracts of UB1341 (not shown). To investigate the role of PavA in adhesion of pneumococci to immobilized fibronectin we compared the binding abilities of wild-type and mutant cells. Cells of UB1339 pavA1::cat were approximately 50% reduced in their numbers binding to fibronectin compared with R800 (Table 1). Re-introduction of the pavA gene in the merodiploid strain UB1340 restored expression of PavA (Fig. 3, lanes 7 and 8) and resulted in restored adhesion levels to fibronectin (Table 1). Strain D39 Cps+ cells showed approximately 10-fold lower levels of binding to fibronectin, confirming previous observations (van der Flier et al., 1995), and inactivation of pavA in UB1341 resulted in further reduced levels of adhesion (Table 1).
|Pneumococcal strain||No. cells bound to fibronectina (× 105) ± SDb|
|R800 wild-type Cps–||2.96 ± 0.46|
|UB1339 pavA1::cat Cps–||1.52 ± 0.28|
|UB968 pavA2′::cat Cps–||1.65 ± 0.32|
|UB1340 pavA1::cat::pSF143 tet pavA+ Cps–||2.78 ± 0.39|
|D39 wild-type Cps+||0.38 ± 0.08|
|UB1341 pavA1::cat Cps+||0.20 ± 0.07|
While the data indicated that PavA was involved in mediating adhesion to fibronectin, we also compared wild-type and mutant cells for a range of properties to determine if pavA gene inactivation had other phenotypic effects. Mutants UB1339 or UB1341 were not significantly different from isogenic wild-type R800 Cps– or D39 Cps+ cells, respectively, in growth rate (R800 td = 40 ± 3 min), extent and rate of high temperature- or detergent-induced lysis, in haemolysin (pneumolysin) production or capsule production (UB1341 and D39). Mutant UB1339 cells demonstrated similar surface hydrophobicity to wild-type R800 cells, as measured by contact angles (57 ± 6.0° of water, 37 ± 4.0° of formamide, 55 ± 4.0° of methyleneiodide) and similar zeta potentials (a measure of surface charge) at a range of pH values (results not shown). The patterns of proteins extracted with deoxycholate and reacted with a monoclonal antibody to surface antigen PspA were identical in wild-type and mutant cell extracts (data not shown). Production of PsaA lipoprotein, detected on Western blots of cell envelope proteins with S. gordonii ScaA antibodies (see Experimental procedures) was unaffected in UB1339 and UB1341 (not shown). Taken collectively, these results strongly suggested that inactivation of pavA did not affect expression of other proteins or phenotypic properties known to be associated with pneumococcal adhesion or virulence.
PavA protein binds immobilized fibronectin
Full-length recombinant x6His N-tagged PavA protein was purified from Escherichia coli(Fig. 5, lane 1) and was shown to bind avidly to human fibronectin immobilized onto microtitre plate wells (Fig. 6A). Binding of PavA to fibronectin was saturable and maximal binding levels were dependent upon the amounts of fibronectin immobilized (Fig. 6A). Addition of > 100-fold excess of soluble fibronectin was found to be completely without effect on PavA binding to immobilized fibronectin (Fig. 6B). Pre-incubating immobilized fibronectin with gelatin did not affect PavA binding, but preincubating with heparin reduced binding by 60% (Fig. 6B). These results indicate that PavA recognizes the heparin binding domains within human fibronectin.
A set of C-terminally truncated recombinant PavA polypeptides of 41.5 kDa, 32.6 kDa and 16.7 kDa were then generated (Fig. 5) and tested for their abilities to bind to immobilized fibronectin. None of these polypeptides were able to bind fibronectin (Fig. 5), indicating that removal of the C-terminal 189 aa residues was sufficient to abrogate binding (Fig. 6A). To confirm this, isogenic mutants of R800 Cps– and D39 Cps+, designated UB968 and UB970, respectively, were generated in which the cat cassette was inserted by allelic exchange into the BsmI site within the C-terminal coding region of pavA(Fig. 1). In cytoplasmic and envelope protein extracts of these mutants, a major PavA antibody-reactive band was evident with an estimated molecular mass of 42 kDa (Fig. 3, lanes 3 and 4). Production of a C-terminally truncated protein of this approximate molecular mass would be entirely consistent with the site of cat gene insertion within the pavA gene (Fig. 1). Cells of strain UB968 pavA2′::cat were reduced in their ability to adhere to immobilized fibronectin to about the same extent as cells of the knockout mutant UB1339 pavA1::cat(Table 1).
Purified full-length recombinant PavA polypeptide inhibited the binding of wild-type R800 cells to fibronectin in a dose-dependent manner, and to a maximal level of about 40% inhibition (Fig. 7). Recombinant PavA polypeptide did not inhibit the residual binding of UB968 mutant cells to fibronectin (Fig. 7). Conversely, the purified 362 aa residue C-terminally truncated PavA polypeptide (Fig. 5) failed to inhibit binding of R800 cells to fibronectin (Fig. 7), confirming the notion that the C-terminal region of PavA is essential for binding to fibronectin. We also investigated whether PavA protein would confer the ability to bind to fibronectin upon a heterologous organism. Enterococcus faecalis JH2–2, which binds only weakly to fibronectin (McNab et al., 1999) was transformed with replicative plasmid pAM401 carrying the entire pavA gene. Western immunoblot analyses of cell envelope proteins prepared from E. faecalis UB1332 (pAM401 pavA+) revealed the presence of a strongly reactive 62 kDa band (Fig. 3, lane 10) that was absent from JH2-2 control cell envelope extracts (Fig. 3, lane 9). E. faecalis UB1332 cells expressing PavA bound to immobilized fibronectin at > sixfold levels over and above those of JH2-2 control cells (Fig. 7). Adhesion of enterococci was > 90% inhibited by preincubation of fibronectin with full-length recombinant PavA (Fig. 7), but was not inhibited by C-terminally truncated 41.5 kDa recombinant PavA (Fig. 7). These data provide compelling evidence for PavA polypeptide as a cell surface-associated component that mediates, at least in part, binding of pneumococci to immobilized fibronectin.
Expression of pavA is essential for virulence
Utilizing the systemic infection model of pneumococcal virulence, groups of five or six BALB/c mice were inoculated intraperitoneally (i.p.) with exponential growth phase cells of pneumococci. All mice succumbed to challenge with 105 cfu wild-type S. pneumoniae D39, which is approximately 102 times the reported LD50 for this strain (Berry et al., 1989). For mice i.p. inoculated with between 104 and 107 cfu of UB1341 pavA1::cat or UB970 pavA2′::cat, both the median survival time (animals killed at 14 days post injection) and the survival rate (19 out of 20) were virtually identical to a control group of 11 mock-inoculated mice. These values were significantly greater (P < 0.0001 by the Mann–Whitney U-test) than those for groups of mice inoculated with between 104 and 107 cfu of D39, in which 20 out of 21 mice died within 3 days (Fig. 8). Only when the challenge doses of UB1341 or UB970 were increased to 108 cfu (i.e. > 105 times the LD50 of D39) did all mice succumb to infection (Fig. 8). No pneumococci were detected in blood or spleen cultures from any of the mice i.p. inoculated with < 106 cfu UB1341 or UB970, or from a control mock-inoculated group of 12 mice, whereas 11 out of 12 mice injected with 105 or 106 cfu D39 had bacteraemias of > 1 × 105 cfu ml−1 (P < 0.05 using Student's paired t-test) and 12 mice out of 12 had positive spleen cultures.
Streptococcal species that colonize mucosal surfaces are now recognized as being able to express a wide spectrum of cell-surface proteins that act as adhesins (Jenkinson and Lamont, 1997). These provide for diversity of substrate binding in a competitive environment and for tissue tropism in colonization of the host. The ability to bind fibronectin, either in fluid phase or immobilized onto a surface, is a common property of streptococci. This reaction has been postulated to assist adhesion of group A streptococci to epithelial cells (Hasty et al., 1992) and to promote binding of viridans streptococci in thrombotic vegetations associated with infective endocarditis (Baddour, 1994). Moreover, the ability to bind fibronectin has recently been shown to be associated with the invasion properties of S. pyogenes. The matrix form of fibronectin enhances the binding of S. pyogenes to host cells (Okada et al., 1997) and a major fibronectin-binding protein PrtF1 mediates invasion of human epithelial cells by group A streptococci (Molinari et al., 1997; 2000; Ozeri et al., 1998). It is possible that fibronectin-binding proteins may have a more general role in bacterial–host interactions by stimulating uptake of bacteria or products by host cells. This might account, at least in part, for the observation that a group A streptococcal fibronectin-binding protein has activity as an adjuvant for antigens delivered by the mucosal route (Medina et al., 1998).
In this paper we have identified PavA as a fibronectin-binding protein in S. pneumoniae. This protein mediates adhesion of pneumococci to the immobilized form of human fibronectin, and inactivation of the gene attenuates virulence. These properties are in keeping with recent observations suggesting that fibronectin-binding proteins in other streptococci are able to mediate adhesion and are associated with the course of disease pathogenesis (Courtney et al., 1999). Similar proteins to PavA are found in S. pyogenes and in S. gordonii, with the S. pyogenes orthologue Fbp54 having been previously characterized as a polypeptide that binds soluble fibronectin (Courtney et al., 1994). However, the contribution of Fbp54 to the overall adhesion of S. pyogenes cells to fibronectin is not clear and may indeed be difficult to evaluate given that individual strains may express at least five different polypeptides with fibronectin-binding activities (Rocha and Fischetti, 1999).
There are several features of PavA polypeptide that distinguish it from many of the other fibronectin-binding proteins characterized to date. One of these is that the primary sequence does not show a characteristic N-terminal leader peptide that might direct export of the polypeptide from the cell via the general secretion (Sec-dependent) pathway (Izard and Kendall, 1994). A second feature is that the polypeptide does not appear to have a typical C-terminal gram-positive bacterial cell-wall anchorage sequence (Navarre and Schneewind, 1994) or choline-binding sequences that might otherwise determine cell–surface association in pneumococci (Yother and White, 1994). Despite the lack of these features, our data show conclusively that PavA is presented on the pneumococcal cell envelope and that the recombinant polypeptide inhibits bacterial cell binding to fibronectin. The Fbp54 protein in S. pyogenes is also reported to be cell-surface localized (Courtney et al., 1996). Other streptococcal cell-surface proteins that do not contain recognized primary sequence signatures associated with Sec-dependent export and cell-surface linkage include streptococcal surface dehydrogenase (SDH), which exhibits glyceraldehyde-3-phosphate dehydrogenase activity (Pancholi and Fischetti, 1992), group A streptococcal enolase (SEN) (Pancholi and Fischetti, 1998) and pneumococcal enolase (Eno) (Bergmann et al., 2001). It seems probable therefore that streptococci may utilize novel and as yet uncharacterized mechanisms for export and surface retention of polypeptides such as these. SEN and Eno proteins bind plasminogen, which may facilitate bacterial penetration of the basement membrane (Bergmann et al., 2001), while SDH binds a variety of mammalian proteins and activates host intracellular signalling pathways (Pancholi and Fischetti, 1997) that may be important for bacterial invasion and proliferation during pharyngeal infection. The possibility that PavA may play a crucial role in the pathogenesis of pneumococcal infections by modulating host cell functions is currently under investigation.
Another structural feature of PavA that distinguishes it from many other fibronectin-binding proteins is that it does not contain sequences that closely match recognized motifs involved in binding fibronectin. A number of studies have suggested that a repeated GGX3−4I/VDF motif is responsible for fibronectin-binding activity in streptococcal and staphylococcal cell wall-anchored proteins (Jaffe et al., 1996; Sun et al., 1997). In addition, a second non-repeated region present in PrtF1 and PrtF2 polypeptides of S. pyogenes is implicated in fibronectin-binding (Jaffe et al., 1996; Ozeri et al., 1996). None of these sequences are found within PavA, suggesting that the mode of recognition of fibronectin by PavA might be different from that of the other proteins. Indeed, PavA binds to the immobilized form of fibronectin even in the presence of a vast excess of fluid-phase fibronectin. This suggests that PavA may contain unique binding region sequences that are able to specifically recognize the conformation of surface-bound fibronectin. As PavA binding to fibronectin was heparin-inhibitable, it is probable that PavA recognizes, at least in part, the heparin-binding domain(s) within fibronectin. This is consistent with previous data showing that pneumococcal cell adhesion to fibronectin was heparin sensitive (van der Flier et al., 1995).
The major surface components promoting adhesion of pneumococci to host cells and tissues are largely unknown. However, important factors promoting pneumococcal virulence are well understood and are generally accepted as being capsular polysaccharide, pneumolysin production and autolysin activity. These factors are believed to come into play only after the bacterial cells have become disseminated from the mucosal site of infection. The antiphagocytic and cytotoxic effects of capsular polysaccharide and pneumolysin, respectively, prevent immune cell-mediated clearance, while autolysis results in the release of cell-wall fragments that provoke intense inflammation. Surface protein PspA has recently been shown to also affect clearance of pneumococci by inhibiting deposition of complement C3b onto cells (Tu et al., 1999). In identifying PavA as an adhesin and as a virulence factor, we suggest that this polypeptide may function at more than one stage in the infection process. Firstly, by acting as an adhesin mediating binding of pneumococci to epithelial cells via fibronectin bridging, PavA might assist S. pneumoniae to efficiently colonize the oropharyngeal and nasopharyngeal mucosa. Although the production of capsule appears to reduce levels of fibronectin binding, recent evidence suggests that modulation of capsule production may be an integral factor in the colonization process that enhances adhesion (Magee and Yother, 2001). Secondly, the massive attenuation of virulence observed for the pavA mutant in the mouse sepsis model indicates a direct and critical function for PavA in determining pneumococcal survival and growth in vivo. Indeed, recent data from signature-tagged mutagenesis experiments have independently identified pavA as a virulence determinant in pneumococcal infection (Lau et al., 2001). No evidence was obtained to indicate that inactivation of the pavA gene affected production of capsule, pneumolysin, autolysin, PspA or PsaA. Anyway, because of the genetic structure of the pavA locus (Fig. 1), disruption of pavA would not be predicted to have a co-ordinated effect on expression of adjacent genes.
In summary, we have identified and characterized a protein PavA, produced by all strains of S. pneumoniae tested, that mediates adhesion of pneumococcal cells to immobilized fibronectin. The C-terminal portion of PavA appears critical to the fibronectin-binding and virulence-associated properties of the protein. The pavA mutant strains characterized nevertheless retained approximately 50% of wild-type binding levels to fibronectin. Therefore it is probable that other fibronectin-binding components, in addition to PavA, are present on the pneumococcal cell surface. Despite the close genetic relatedness of S. pneumoniae and S. gordonii, and the similarities displayed by these bacteria in preferentially binding to immobilized fibronectin, pneumococcus has not acquired the cshA-like gene that encodes a major fibronectin-binding protein in S. gordonii (McNab et al., 1996). Thus, the nature of the other fibronectin-binding components on the pneumococcal cell surface is open for investigation. It is of significance to this report that recent work has demonstrated that a protective immune response against S. pyogenes can be developed following immunization with fibronectin-binding proteins SfbI (PrtF1) (Guzman et al., 1999) or Fbp54 (Kawabata et al., 2001). It remains to be determined whether or not immunization with PavA can elicit protective antibodies and therefore whether PavA might be considered as a new pneumococcal vaccine candidate.
Bacterial strains and growth media
S. pneumoniae strains utilized were D39 (Cps+ type 2), R36A (rough colony variant of D39), NCTC 10319 (Cps+ type 35A) and R800 (Cps–, derived from R36A). Isogenic pavA mutants generated from the wild-type strains were designated UB1339 pavA1::cat Cps–, UB1341 pavA1::cat Cps+, UB968 pavA2′::cat Cps– and UB970 pavA2′::cat Cps+. Fifty-nine fresh clinical isolates of S. pneumoniae representing 17 different capsular serotypes were kindly provided by M. Smith, UK Public Health Laboratory Service, Wessex region. Serotype 3, 18c and 22 strains were from M. Brett, ESR, Porirua, New Zealand. Escherichia coli DH5α was routinely used for cloning while M15pREP4 (Qiagen) was host for expression x6His–PavA fusion proteins. Enterococcus faecalis JH2–2 was grown at 37°C in TY-glucose medium (Jenkinson et al., 1993), and for strains harbouring pAM401 (Wirth et al., 1986) the medium contained 5 µg ml−1 chloramphenicol. E. faecalis cells were transformed with plasmid DNA as previously described (McNab et al., 1999). Pneumococci were grown in THY medium (Todd–Hewitt broth (Difco) supplemented with 0.5% yeast extract) or in THY containing 10% heat inactivated fetal calf serum as stationary cultures in screw-capped tubes or bottles at 37°C. Pneumococcal colonies were grown in THY agar pour plates containing 0.5% horse red blood cells, overlaid with THY blood agar containing 10 µg ml−1 chloramphenicol if appropriate, and incubated aerobically at 37°C. E. coli cells were grown on Luria–Bertani (LB) medium (Sambrook et al., 1989) containing the following antibiotics when required: 50 µg ml−1 ampicillin; 10 µg ml−1 chloramphenicol; 25 µg ml−1 kanamycin; 10 µg ml−1 tetracycline.
Plasmid DNA was prepared from E. coli using Wizard minipreps (Promega), while S. pneumoniae chromosomal DNA was purified as described by Pearce et al. (1993). Restriction endonuclease digestion, 32P-labelling of DNA and other nucleic acid techniques were performed according to standard protocols (Sambrook et al., 1989). Synthetic primers FB2 (5′-GACAAGGCTGAGCGCGA, bp 900–916 pavA locus, GenBank accession AF181976) and FB3 (5′-GCTAGCTCGGCTGCGTC, bp 1575–1591) were utilized in polymerase chain reaction (PCR) amplification to first generate internal coding sequence of the pavA gene from S. pneumoniae R800. Conditions for PCR amplification were as follows: 94°C for 15 s; 50°C for 30 s; 72°C for 60 s repeated 30 times in a total volume of 50 µl containing dNTPs (10 µM), primers (0.25 µM), 50 ng of template DNA and 2 U of Taq polymerase. A 671 bp amplimer was cloned into pGEM-T vector (Promega) and the insert confirmed by automated sequencing.
Chromosomal DNA flanking the 671 bp fragment was obtained by inverse PCR (Ochman et al., 1990). Suitable upstream and downstream restriction enzyme sites were first identified by blot hybridization analysis of restricted R800 DNA with the 671 bp fragment as probe. Genomic DNA was then digested with either BsmI or DraI, ligated under dilute conditions to promote circularization, and amplified using outward facing primer pairs. An upstream 2.1 kb DraI fragment and an overlapping downstream 1.1 kb BsmI fragment (see Fig. 1) were cloned into pGEM-T and sequenced.
The entire pavA coding region together with upstream and downstream sequences (2.7 kb) was PCR-amplified from R800 DNA using Expand PCR System (Roche Diagnostics) with primers N1 (5′-GTAAGAATTCGTGTACTGCCAAGAAGCCCA, EcoRI site underlined) and C2 (5′-ATTTAAGCTTGGCTACAAACTACAGTGCTG, HindIII site underlined) (see Fig. 1) and cloned into pCR4-TOPO (Invitrogen BV). Conditions for PCR amplification were as follows: 94°C for 25 s; 54°C for 40 s; 72°C for 120 s repeated 30 times with a final run-off at 72°C for 10 min. To clone this fragment for heterologous gene expression in E. faecalis, it was excised from pCR4-TOPO by digesting with EcoRI and HindIII and ligated into EcoRI/HindIII-cut pBluescript SK (Stratagene). The resultant plasmid was then digested with BamHI and SalI, the pneumococcal DNA was gel-purified and ligated into BamHI/SalI-cut pAM401, and chloramphenicol resistant colonies were selected following transformation of E. coli DH5α.
The entire pavA coding region was also obtained with only 30 bp downstream sequence by PCR amplification from R800 DNA as described above, but utilizing primers N1BAM (5′-GTAAGGATCCGTGTACTGCCAAGAAGCCCA) and PR2XBA (5′-TATCTAGATAAAAATCCTGATTTCAAG). This 2.1 kb fragment was cloned into pGEM-T, excised by digesting with BamHI and XbaI (sites within the primers), and ligated into BamHI/XbaI-cut pSF143 (Tao et al., 1992) with transformation into E. coli and selection for tetracycline resistance.
Screening for pavA across a range of serotypes
Primers used for amplification of the pavA coding region (bp 108–1760) were as follows: upstream 5′-GATATACATATGTCATTTGACGGATTTTTTTTAC (NdeI site underlined) and downstream 5′-GGGCTCGAGGGATTTTTTCATGGATGCAA (XhoI site underlined). Conditions for PCR amplification were as described above for FB2/FB3 primer pair PCR except that template DNA was provided by adding 1 µl of washed cell suspension containing 105 cells directly to the PCR mix and heating at 94°C for 10 min before addition of Taq polymerase.
Plasmid pGEM-T containing the 671 bp internal fragment of pavA was linearized by digesting with BsmI and was ligated with the 1073 bp SmaI–XhoI fragment (blunt-ended using Klenow) excised from pR326 (Claverys et al., 1995) and containing a chloramphenicol acetyltransferase (cat) gene. Plasmid pCR4-TOPO containing the entire pavA gene within the 2.7 kb fragment was linearized with ApaI, incubated with T4 DNA polymerase to end-flush and ligated with the blunt-ended cat gene fragment. Resultant plasmids were purified, linearized with KpnI or EcoRI and transformed into competent cells of S. pneumoniae R800 or D39. To prepare competent cells, 0.5 ml exponential phase cells in THY medium were added to prewarmed THY medium (9.5 ml) and incubated at 37°C for 30 min A portion (1 ml) of culture was removed into a tube containing 20 ng of competence stimulating peptide (CSP) (Havarstein et al., 1995). After a further incubation at 37°C for 15 min, 0.2 ml portions were removed into fresh tubes containing approximately 0.1 µg of DNA (10 µl) and cultures were incubated at 37°C for 2 h. Cells were plated into THY blood agar, overlaid with THY blood agar and plates incubated at 37°C for 24 h.
Insertions of the cat cassette in mutant strains were confirmed by PCR amplifications from genomic DNAs using N1/C2 or FB2/FB3 primer pairs (Fig. 1). Amplimers of expected sizes were then reacted in blot hybridizations with 32P-labelled pavA or cat sequence probes. The chloramphenicol-resistant phenotype of pneumococcal isogenic mutants was completely stable and was maintained over prolonged passages in antibiotic-free medium. To construct a merodiploid in which the pavA1 mutation in UB1339 pavA1::cat was complemented, pSF143 carrying a 2.1 kb pavA gene fragment (see above) was transformed into UB1339 with selection for resistances to 5 µg ml−1 tetracycline and chloramphenicol. The resultant strain UB1340 pavA1::cat::pSF143 tet pavA+ was generally grown with tetracycline selection to maintain the plasmid insertion.
RNA was extracted from S. pneumoniae by sodium deoxycholate lysis of mid-exponential phase cells as described by Cheng et al. (1997) and purified using RNeasy (Qiagen). Samples containing 10 µg of RNA were electrophoresed through 1.2% agarose–formaldehyde and RNA was transferred by vacuum blotting onto N+ nylon membrane (Amersham Pharmacia Biotech). Blots were incubated with 32P-labelled pavA DNA probe under conditions described by Church and Gilbert (1984), and radioactive bands were detected by autoradiography.
Purification of recombinant PavA polypeptide
To produce N-terminal x6His-tagged PavA the entire coding region was PCR-amplified with Pfu polymerase from R800 DNA using the following primers: 5′-GATATAGCATGC TCATTTGACGGATTT (SphI site underlined) and 5′-AATGCGCTGCAGTCAGGATTTTTTCAT (PstI site underlined). To generate a series of sequentially truncated products, the N-terminal primer was fixed as above and the following C-terminal primers were utilized (PstI site underlined in each case): PAV11, 5′-AATGCGCTGCAGAAAATAGCGTTGGGC bp 1173; PAV15, 5′-AATGCGCTGCAGATTTTCAACACGACG bp 948; PAV17, 5′-AATGCGCTGCAGTGATAACTTCGAGGA bp 523. The products were digested with SphI and PstI, cloned into pQE30 (Qiagen) and transformed into E. coli M15. Expression of His-tagged PavA polypeptides was induced by addition of 1 mM IPTG to exponential phase cultures. After incubation at 30°C for 4 h, cells were harvested by centrifugation and lysed by sonication. Proteins were extracted with 8 M urea, and His-tagged polypeptides were purified on a nickel-NTA resin according to the manufacturer's guidelines, followed by dialysis step-wise against decreasing concentrations of urea to assist refolding.
Sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblot analysis
Pneumococci from late exponential-phase cultures (30 ml) were harvested by centrifugation (4000 g, 4°C, 10 min), washed once in 5 ml of TP buffer (50 mM tris-HCl pH 7.8 containing 5% (v/v) protease inhibitor cocktail (Sigma; for use with bacterial cell extracts), and suspended in 0.5 ml TP buffer. Cells were disrupted by sonication (4 × 15 s at 280 W cm−2 with cooling on ice between pulses) using an Ikasonic U50 hand-held sonicator fitted with a 3 mm-diameter probe (IKA-Werke GMBH). The suspension was then incubated at 37°C for 10 min with mutanolysin (50 U ml−1) and lysozyme (0.2 mg ml−1), which resulted in > 95% cell lysis. The suspension was fractionated by centrifugation (12 000 g, 4°C, 10 min) and the supernatant (designated the soluble or cytoplasmic fraction) was carefully removed. The cell envelope pellet was washed twice by suspension and centrifugation in TP buffer, and proteins were solubilized from fractions with 1% SDS (final concentration) by heating at 90°C for 10 min, mixed with loading dye (Jenkinson et al., 1993) and subjected to electrophoresis. Extracts of mid-exponential phase pneumococcal cells for detection of PspA were prepared using 0.1% sodium deoxycholate as described by Yother et al. (1992).
Polypeptides were transferred to nitrocellulose membrane by electroblotting. To detect PavA polypeptide, blots were blocked with Tris-buffered saline (TBS) (50 mM tris-HCl, 0.15 M NaCl, pH 8.0) containing 10% skimmed milk powder and 0.3% (v/v) Tween 20 (TBS-MT) and then incubated at 4°C for 16 h with affinity purified PavA antibody diluted 1:1 with TBS. The PavA antibodies were raised in rabbits to recombinant PavA by routine immunological techniques, and antibodies were affinity purified by incubating antiserum with a nitrocellulose strip containing recombinant PavA. Bound antibodies were eluted with 0.2 M glycine-HCl, pH 2.5, as described elsewhere (Jenkinson and Easingwood, 1990). After reacting the Western blots with primary antibodies, the blots were washed with TBS-MT (once for 15 min, twice for 5 min) and incubated at 20°C for 1 h with HRP-conjugated swine anti-rabbit secondary antibodies (Dako) diluted 1:2000 in TBS-MT. Antibody binding was detected by enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech) according to manufacturer's instructions.
To detect PspA or PsaA on Western blots the membranes were blocked with 0.5% gelatin in TBS and then incubated with either 1:10 diluted monoclonal antibody R278 (Crain et al., 1990) kindly provided by L. D. McDaniel (University of Alabama at Birmingham) or with 1:200 diluted rabbit antiserum to ScaA polypeptide from S. gordonii, kindly supplied by P. E. Kolenbrander (NIDCR, NIH) (Holmes et al., 1996). Bound primary antibodies were detected with the appropriate peroxidase-conjugated swine secondary antibodies, usually diluted 1:1000, and developed with 4-chloro-1-napthol as described previously (Jenkinson and Easingwood, 1990).
Binding of recombinant PavA to immobilized fibronectin
Binding of N-terminal x6His-tagged PavA proteins to fibronectin immobilized onto microtitre plate wells was measured essentially as described previously (McNab et al., 1996). Briefly, wells were coated with fibronectin in the range 0.12–2.5 µg per well, and blocked with 1% BSA. Amounts of up to 0.5 µg of recombinant PavA protein in diluted in PBS (50 mM K2HPO4-KH2PO4, 0.15 M NaCl, pH 7.0) were incubated with fibronectin-coated or BSA-coated (control) wells at 37°C for 2 h, washed with PBS and incubated with 1:1000 diluted tetra-His antibody (Qiagen) at 20°C for 1 h. After rinsing the wells with PBS, bound antibody was detected with 1:1000 diluted horseradish peroxidase (HRP)-conjugated anti-mouse IgG (Dako) as described elsewhere (Holmes et al., 1996). Enzyme-linked immunosorbent assay (ELISA) (A492) values for protein bound to fibronectin were corrected for background binding of protein to BSA-blocked wells. In separate experiments, we could detect no specific binding of an irrelevant N-terminally x6His-tagged protein (PMA1 from yeast) to fibronectin, confirming that the x6His tag does not bind to fibronectin. To measure the effects of fibronectin, gelatin or heparin on PavA binding to immobilized fibronectin, PavA was added to fibronectin-coated wells in the presence of 100 µg ml−1 soluble fibronectin, or to fibronectin-coated wells that had been preincubated with gelatin (1 mg ml−1) or heparin (100 U ml−1) at 37°C for 1 h. Results were expressed as percentage inhibition of binding compared with untreated control immobilized fibronectin.
Pneumococci were harvested from mid-log phase cultures in THY medium, washed twice with PBS and incubated (109 cfu ml−1) with a 10 µl suspension of anti-PavA antibodies, affinity-purified on Protein A Sepharose 4B (Amersham Pharmacia Biotech). The cells were washed and incubated with 1:100 diluted Protein A-10 nm gold particle colloidal suspension to label the primary antibodies. The samples were washed in PBS, post-fixed with 1% (v/v) formaldehyde in PBS at room temperature for 30 min, washed in TE buffer (20 mM tris-HCl, 1 mM EDTA, pH 6.9) and adsorbed onto carbon-coated formvar grids. These were washed in TE buffer, then distilled water, air dried and examined in a Zeiss field emission SEM (DSM962 Gemini). For post-embedding labelling, bacteria were fixed in a solution containing 0.2% glutaraldehyde and 0.5% formaldehyde in PBS on ice for 1 h. Aldehyde groups were blocked with 10 mM glycine and bacteria were taken into 2% agar, dehydrated and embedded in LR White resin. Ultra-thin sections were cut and incubated on drops of anti-PavA serum (100 µg IgG ml−1) at 4°C for 12 h. Sections were washed with PBS, incubated with 0.4% skimmed milk suspension for 5 min, dry blotted onto filter paper and incubated with 1:100 diluted protein A-colloidal gold suspension for 10 min. Sections were counterstained with 4% aqueous uranyl acetate for 5 min and then examined using a Zeiss TEM.
Haemolysin production was determined as described by Benton et al. (1997). Briefly, washed sheep red blood cells (2% in PBS) were dispensed into round-bottomed microtitre plate wells containing serially diluted pneumococcal culture supernatants, or cell lysates, in 0.1% BSA containing 10 mM dithiothreitol. Haemolysin activity was scored after incubation at 37°C for 30 min as the reciprocal of highest dilution causing haemolysis.
Determination of autolysis
Temperature-induced autolysis of exponentially growing cells in THY medium was determined by measuring the rates of decrease in optical density at 600 nm (OD600) of triplicate samples shifted to 47°C. Deoxycholate-induced autolysis was determined as described by Diaz et al. (1992). Exponentially growing cells were suspended in 100 mM sodium phosphate pH 8 and equilibrated at 37°C for 2 min in a thermostatically controlled cuvette in a recording spectrophotometer. Decrease in OD600 was then monitored continuously following addition of 0.1% (final concentration) sodium deoxycholate.
Contact angle measurements and microelectrophoresis
Pneumococcal cells were grown in THY medium at 37°C to early stationary phase. Contact angle measurements of water, formamide and methylene iodide were determined using the sessile drop technique (Van Oss and Gillman, 1972). Zeta potentials were determined by microelectrophoresis as described by James (1991). Briefly, cells were suspended at a density of 3 × 107 ml−1 in 10 mM potassium phosphate buffer adjusted to cover a range between pH 2 and pH 9 with HCl or KOH. Electrophoretic mobilities were measured using a Lazer Zee Meter 501 apparatus (Penkem) and these were converted to zeta potentials using the Helmholtz–Smoluchoski equation (Hiemenz, 1977).
Binding of bacteria to immobilized human fibronectin
Adhesion of S. pneumoniae or E. faecalis cells to fibronectin was performed as described previously for S. gordonii (McNab et al., 1996). Human fibronectin (Boehringer) (50 µl) was applied to microtitre plate wells (Nunc Immunosorp) in coating buffer (15 mM Na2CO3/35 mM NaHCO3, pH 9.6) and additional binding sites were blocked with 0.1% BSA in PBS. Cells of S. pneumoniae, radioactively labelled by growth in THY medium containing [methyl-3H]-thymidine (0.44 MBq (12 µCi) ml−1; 85 Ci mmol−1) to a specific activity of between 1 × 10−3 and 2 × 10−3 cpm cell−1, were added to each well (between 5 × 106 and 1 × 107 cells in 50 µl) and incubated at 37°C for 2 h with gentle shaking. E. faecalis cells were radioactively labelled to similar range of specific activities by growing in TY-glucose medium. Unattached cells were then aspirated, the wells were washed three times with PBS and numbers of cells bound were then determined by scintillation counting (Jenkinson et al., 1993).
To test inhibition of pneumococcal or enterococcal cell binding to fibronectin by PavA polypeptides, fibronectin was deposited onto ScintiStrip microwells (in which the plastic is impregnated with scintillant) and blocked as before with BSA. The wells were then either left untreated or were preincubated at 37°C for 1 h with up to 0.25 µg per well of purified recombinant PavA protein. The wells were then washed with PBS, radioactively labelled bacterial cells (1 × 107) were added, incubated as described above, washed with PBS, and numbers of cells bound estimated from radioactive counts.
Mouse sepsis model of virulence
To prepare bacteria for inoculation into mice, cultures of S. pneumoniae were grown in THY medium containing 10% fetal calf serum to a density of approximately 1 × 108 cells ml−1. Bacteria were harvested by centrifugation, suspended in PBS and diluted at a range of densities that were determined subsequently from viable counts of cfu. In four separate experiments, groups of 12 week-old BALB/c mice were inoculated intraperitoneally with 0.1 ml of cell suspension and monitored for up to 14 days. Blood or spleen cultures were prepared from mice that had died or that had been humanely killed, to detect the presence of pneumococci. Experimental procedures for all animal challenge studies were reviewed and approved by the University of Otago Animal Ethics Committee (Permit no. 44/95).
We thank J.-P. Claverys, D. B. Clewell, J. C. Paton and L. Tao for providing strains and plasmids, D. A. Morrison for the gift of synthetic competence stimulating peptide, P. E. Kolenbrander and L. C. McDaniel for providing antibodies, and M. Smith for supplying pneumococcal clinical isolates. We thank Tina Blair for excellent technical assistance, N. Pan for help with inverse PCR, J. Lewthwaite for expertise with ECL detection, and K. Cartwright, S. Funnell and R. Heyderman for helpful discussions. This work was supported in part by the Health Research Council of New Zealand.
- 191995) The pathogenesis of pneumococcal disease. A molecular model. Clin Infect Dis 21: 5204–5212, , (
- 301977) Electrophoresis and other electrokinetic phenomena. In Principles of Colloid and Surface Chemistry, Lagowsky, J.J. (ed.) New York: Marcel Dekker, pp. 452–487(
- 341991) Charge properties of microbial cell surfaces. In Microbial Cell Surface Analysis: Structural and Physicochemical Methods. Mozes, N., Handley, P.S., Busscher, H.J., and Rhouxhet, P.G. (eds). New York: VCH Publishers, pp. 221–262(
- 491990) Amplification of flanking sequences by inverse PCR. In PCR Protocols: a Guide to Methods and Applications. Innis, M.A., Gelfand, D.H., Sninsky, J.J., and White, T.J. (eds). San Diego: Academic Press, pp. 219–227, , , (
- 641989). Molecular Cloning: a Laboratory Manual. New York: Cold Spring Harbor Laboratory Press., , (