A Streptococcus pneumoniae pathogenicity island encoding an ABC transporter involved in iron uptake and virulence

Abstract

Restricted iron availability is a major obstacle to growth and survival of pathogenic bacteria during infection. In contrast to Gram-negative pathogens, little is known about how Gram-positive pathogens obtain this essential metal. We have identified two Streptococcus pneumoniae genetic loci, pit1 and pit2, encoding homologues of ABC iron transporters that are required for iron uptake by this organism. S. pneumoniae strains containing disrupted copies of either pit1 or pit2 had decreased sensitivity to the iron-dependent antibiotic streptonigrin, and a strain containing disrupted copies of both pit1 and pit2 was unable to use haemoglobin as an iron source and had a reduced rate of iron uptake. The pit2⁻ strain was moderately and the pit1⁻/pit2⁻ strain strongly attenuated in virulence in mouse models of pulmonary and systemic infection, showing that the pit loci play a critical role during in vivo growth of S. pneumoniae. The pit2 locus is contained within a 27 kb region of chromosomal DNA that has several features of Gram-negative bacterial pathogenicity islands. This probable pathogenicity island (PPI-1) is the first to be described for S. pneumoniae, and its acquisition is likely to have played a significant role in the evolution of this important human pathogen.

Introduction

The availability of free iron in extracellular fluid and on mucosal surfaces of mammals is restricted as a result of chelation by iron-binding proteins (Wooldridge and Williams, 1993). This provides a major obstacle to the growth and survival of pathogenic bacteria, and high-affinity iron uptake mechanisms are essential for the virulence of several Gram-negative pathogens, including Neisseria spp. (Schryvers and Stojiljkovic, 1999), Salmonella typhimurium (Janakiraman and Slauch, 2000), Pseudomonas aeruginosa (Takase et al., 2000), Legionella pneumophila (Viswanathan et al., 2000), Yersinia pestis (Bearden and Perry, 1999) and Helicobacter pylori (Velayudhan et al., 2000). The mechanisms by which Gram-negative pathogens obtain iron from the host are, in general, relatively well understood. They include the secretion of low-molecular-weight iron chelators, called siderophores, which scavenge iron from host iron-binding proteins such as transferrin, and secreted haemophores, which acquire iron from haemoglobin and haemin (Wooldridge and Williams, 1993; Wandersman and Stojiljkovic, 2000). Alternatively, proteins containing iron or haem can bind to specific receptors on the bacterial outer membrane (Wooldridge and Williams, 1993; Cornelissen and Sparling, 1994). Independent of the source of the captured iron, transport into the bacterial cytoplasm is often dependent on cytoplasmic membrane ABC transporters (Fetherston et al., 1999).

Despite the wealth of data on iron uptake by Gram-negative pathogens, little is known about the mechanisms and importance during infection of iron acquisition by Gram-positive pathogens. Of the Gram-positive pathogens, Streptococcus pneumoniae is second only to Mycobacterium tuberculosis as a cause of mortality worldwide. S. pneumoniae frequently colonizes the nasopharynx, and invasive infection can develop in a variety of body compartments, including the middle ear, the lungs, the blood and cerebrospinal fluid. The organism must use iron sources in each of these environments but, at present, how S. pneumoniae acquires iron and from which substrate(s) is poorly understood. Potential iron sources in the respiratory tract include lactoferrin, transferrin, ferritin (released from dead cells shed from the mucosal epithelium) and possibly small amounts of haemoglobin and its breakdown products (LaForce et al., 1986; Thompson et al., 1990; Schryvers and Stojiljkovic, 1999). In addition, siderophores produced by other nasopharyngeal commensals may provide an alternative iron source. S. pneumoniae growth in iron-deficient medium can be supplemented by haemin, haemoglobin and FeSO₄ (Tai et al., 1993). Although an S. pneumoniae protein, PspA, binds human lactoferrin (Hammerschmidt et al., 1999), in contrast to other mucosal pathogens, neither transferrin nor lactoferrin can support the growth of S. pneumoniae in iron-deficient medium (Tai et al., 1993; Cornelissen and Sparling, 1994; Schryvers and Stojiljkovic, 1999). Chemical and biological assays suggest that S. pneumoniae does not produce siderophores (Tai et al., 1993), but a haemin-binding polypeptide has been isolated, and an undefined mutant unable to use haemin as an iron source was reduced in virulence (Tai et al., 1993; 1997). However, the molecular basis for iron uptake by S. pneumoniae has yet to be characterized.

Virulence determinants of Gram-negative bacteria, including iron transporters, are frequently encoded in defined areas of chromosomal DNA thought to be acquired by horizontal transmission and termed pathogenicity islands (PAIs) (Hacker and Kaper, 2000). Characteristically, PAIs have a different G+C content from host chromosomal DNA, frequently have tRNA or insertion sequences at their boundaries, contain components of mobile genetic elements and are not present in less pathogenic but related strains of bacteria (Hacker et al., 1997). The acquisition of PAIs has probably been a major influence in the evolution of Gram-negative pathogens (Hacker and Kaper, 2000; Ochman et al., 2000). In contrast, only a few PAIs have been described for Gram-positive pathogens, and they rarely have the characteristic features of Gram-negative PAIs (Lindsay et al., 1998; Chakraborty et al., 2000). Furthermore, no PAIs have been identified in S. pneumoniae.

In this paper, we report the phenotypic characterization of S. pneumoniae strains containing defined mutations in two loci, pit1 and pit2, each consisting of four genes whose products have high degrees of identity to ABC transporters involved in iron uptake. Strains with mutations in either locus were less sensitive to the iron-dependent antibiotic, streptonigrin, and a pit2⁻ mutant strain was attenuated in virulence in mouse models of pneumonia and systemic infection. A strain containing insertional mutations in both pit1 and pit2 was unable to use haemoglobin as an iron source and was highly attenuated in virulence. Furthermore, the pit2 locus is contained within a region of chromosomal DNA with many of the characteristics of a Gram-negative PAI. This is the first S. pneumoniae PAI to be described and provides further evidence that the acquisition of PAIs has influenced the evolution of Gram-positive as well as Gram-negative pathogens.

Results

Identification and sequence analysis of the pit1 and pit2 loci

A signature-tagged mutagenesis screen of S. pneumoniae strain in a mouse model of pneumonia (Lau et al., 2001) identified a strain attenuated in virulence containing a mutation in a gene (smtA) whose predicted amino acid product has 31% identity and 53% similarity to CeuC, a component of a Campylobacter coli iron uptake ABC transporter (Richardson and Park, 1995). The predicted product of smtA has high degrees of identity to transmembrane permeases of iron uptake ABC transporters and is the second gene of a four-gene locus (Fig. 1A). This locus was renamed pit1BCD and A (pneumococcal iron transporter 1). Searches of the S. pneumoniae genome using IRP1, an iron-regulated Corynebacterium diphtheriae lipoprotein, identified a gene, pit2A, the predicted amino acid sequence of which has 43% identity and 62% similarity to IRP1 (Lee et al., 1997). pit2A is the first gene of a second four-gene locus, pit2ABC and D, whose products have high degrees of identity to iron uptake ABC transporters (Fig. 1C). Both the pit1 and the pit2 loci contain one gene encoding a putative ATPase, one gene encoding a putative lipoprotein iron receptor and two genes encoding putative transmembrane permease proteins (Fig. 1A and C). In both loci, all four genes are transcribed in the same direction and either have short intergenic sequences (maximum 135 bp between pit1C and pit1D) or overlapping open reading frames (ORFs), suggesting that they are single transcriptional units. Reverse transcriptase–polymerase chain reaction (RT–PCR) analysis confirmed that the four genes of both pit1 and pit2 are transcribed as single operons, with transcription terminating between pit1A and pit2D and the start codons of the next downstream ORFs (Fig. 1B and D). The putative metal-binding receptors Pit1A and Pit2A have closest similarity to the amino acid sequences of different iron uptake lipoproteins and have relatively low degrees of identity and similarity to each other (22% and 53% over 324 amino acids) (Table 1). Both Pit1A and Pit2A contain motifs matching the consensus sequence for the lipoprotein signal peptide cleavage site (Fig. 1E) (Sutcliffe and Russell, 1995). The derived amino acid sequences of the putative ATPases, Pit1D and Pit2D, contain motifs commonly found in ATP-binding proteins (Fig. 1F) (Linton and Higgins, 1998).

Figure 1. A. (pit1) and C. (pit2). Genetic organization of the pit loci. Key: line, chromosomal DNA; clear boxes, ORFs flanking pit loci encoding I, a putative UDP galactose epimerase; II, unknown function; III, a putative RNA methyltransferase; IV, a putative recombinase; grey boxes, pit genes (A, putative iron-binding receptor gene; B and C, putative permease genes; D, putative ATPase gene); arrows, site of insertions in mutant strains.

B and D. Ethidium bromide-stained agarose gels showing transcriptional analysis of the pit loci. Each box contains products with the same primer pairs for PCR using S. pneumoniae chromosomal DNA as a template (on the left) and RT–PCR using S. pneumoniae RNA as the template (on the right). RT–PCRs containing no reverse transcriptase gave no products. Bars represent the corresponding target segments for each pair of primers used.

E. Putative lipoprotein signal peptidase cleavage sites of Pit1A and Pit2A. Consensus sequence as defined by Sutcliffe and Russell (1995).

F. Conserved ATPase sequence motifs from Pit1D and Pit2D as defined by Linton and Higgins (1998). x, any residue; h, hydrophobic residue; +, charged residue.

Table 1.  Proteins to which Pit1A and Pit2A have close identity and similarity^a.

	Protein name/ accession no.	Function	Organism	% Identity/ similarity	Length of amino acids compared
a . Excludes proteins whose genes are uncharacterized ORFs identified by whole-genome sequencing.
Pit1A	CeuE	Enterochelin uptake	Campylobacter coli	36/52	336
	FetB	Putative enterobactin periplasmic-binding protein	Neisseria gonorrhoeae	34/52	310
	FatB	Ferric anguibactin binding protein precursor	Vibrio anguillarum	29/50	314
	SirA	Putative siderophore uptake receptor	Staphylococcus aureus	25/42	327
	FeuA	Iron uptake system	Bacillus subtilis	25/44	260
	FecB	Iron (iii) dicitrate periplasmic binding protein	Escherichia coli	27/46	186
	Pit2A	Putative lipoprotein iron receptor	S. pneumoniae	22/53	324
Pit2A	IRP1	Iron-regulated lipoprotein	Corynebacterium diphtheriae	43/62	339
	T36412	Probable siderophore-binding lipoprotein	Streptomyces coelicolor	28/47	305
	FxuD	Putative periplasmic iron transporter	Mycobacterium smegmatis	32/45	224
	SirA	Putative siderophore uptake receptor	S. aureus	30/45	275
	Yfmc	Ferrichrome ABC transporter homologue	B. subtilis	26/44	296
	FecB	Iron (iii) dicitrate periplasmic-binding protein	E. coli	24/44	283
	FepB	Ferrienterobactin-binding periplasmic protein	E. coli	24/41	275
	FeuA	Iron uptake system	B. subtilis	25/44	220
	FatB	Ferric anguibactin-binding protein precursor	V. anguillarum	25/39	294
	CeuE	Enterochelin uptake	C. coli	22/42	312

To assess the role of the pit loci in iron uptake and virulence, strains carrying mutations in pit1B and pit2A were constructed by insertion–duplication mutagenesis (Fig. 1A and C). Insertions were placed in the first genes of each locus to ensure complete inactivation of operon function. A third mutant containing insertions in both pit1B and pit2A was also constructed. To ensure that the pit⁻ mutant phenotypes were not the result of polar effects of the insertions on genes downstream of the pit loci, two further mutants, PPC4 and PPC29, were constructed containing insertions 73 bp and 100 bp downstream of the stop codons of pit1A and pit2D respectively (Fig. 1A and C).

Growth of pit⁻ mutants in iron-deficient medium

The growth rates of the pit⁻ mutant strains were compared with that of the wild-type parental strain in undefined complete medium THY and in THY that had been treated with Chelex-100 to remove cations. There were no discernible differences in the growth rates of each of the mutant strains and the wild-type strain in THY (Fig. 2A). All strains had impaired growth in Chelex-THY and, in this medium, the double mutant pit1B⁻/pit2A⁻ strain had decreased growth compared with the wild-type and single pit⁻ strains (Fig. 2B). The wild-type and single pit1B⁻ and pit2A⁻ strains' growth defects in Chelex-THY were reversed by supplementation with 10 µM FeCl₂ or FeCl₃(Fig. 2C and D), partially reversed by supplementation with 10 µM haemoglobin (Fig. 2E), but not by 5 µg ml⁻¹ lactoferrin or 5 µg ml⁻¹ ferritin (data not shown). The final optical density (OD) of a culture of the pit1B⁻/pit2A⁻ strain in Chelex-THY supplemented with FeCl₂ or FeCl₃ attained levels similar to that of the wild-type strain, suggesting that the pit1B⁻/pit2A⁻ strain's growth defect in Chelex-THY was the result of iron depletion (Fig. 2C and D). However, supplementing Chelex-THY with 10 µM haemoglobin did not restore growth of the pit1B⁻/pit2A⁻ strain (Fig. 2E). These findings were confirmed by investigating the growth of the pit⁻ strains in a defined medium containing no added iron, RPMIm. Growth of the wild-type strain in Chelex-RPMIm was delayed and reduced compared with growth in Chelex-THY (OD₅₈₀ of 0.19 after 24 h), but was markedly improved by the addition of 10 µM haemoglobin (OD₅₈₀ of 0.18 after 9 h), FeCl₂ (OD₅₈₀ of 0.20 after 9 h) or 20 µM MnCl₂ and 20 µM ZnCl (OD₅₈₀ of 0.29 after 17 h). The pit1B⁻/pit2A⁻ strain was unable to grow in Chelex-RPMIm without supplementing the medium with 10 µM FeCl₂ or FeCl₃(Fig. 2F). Supplementation with 5 µM haemoglobin or haemin had only a minimal effect on growth (Fig. 2F).

Figure 2. Growth of pit mutants measured by optical density in (A) THY broth, (B) Chelex-THY broth, (C) Chelex-THY broth + 10 µM FeCl₂, (D) Chelex-THY broth + 10 µM FeCl₃, (E) Chelex-THY broth + 10 µM haemoglobin. Key for A to E: diamonds, wild-type strain; squares, pit1B⁻ strain; circles, pit2A⁻ strain; filled triangles, pit1B⁻/pit2A⁻ strain.

F. Growth of the pit1B⁻/pit2A⁻ strain in Chelex-RPMIm. Key: crosses, no supplementation; squares, supplemented with 10 µM FeCl₃; circles, supplemented with 10 µM haemoglobin; triangles, supplemented with 25 µM MnCl₂ and 25 µM ZnCl₂.

pit1B⁻ and pit2A⁻ mutants have decreased sensitivity to streptonigrin

The bactericidal effect of the antibiotic streptonigrin requires intracellular iron (Yeowell and White, 1982), and decreased sensitivity to streptonigrin has been used to identify strains with mutations in iron transporter genes (Braun et al., 1983; Pope et al., 1996). We compared the proportional survival of pit⁻ strains with wild-type strains when incubated with streptonigrin. Both the pit1B⁻ and the pit2A⁻ strains were approximately 10-fold more resistant to streptonigrin than the wild-type strain (Fig. 3A), demonstrating that the pit⁻ cells contain less iron than wild-type cells. Disrupting both pit loci had a clear additive effect, resulting in a strain highly resistant to streptonigrin (Fig. 3A). Mutant strains containing insertions immediately downstream of the pit loci, PPC4 and PPC29, had streptonigrin sensitivities similar to that of the wild-type strain, confirming that the loss of streptonigrin sensitivity in the pit⁻ strains results from the mutations in the pit loci (Fig. 3A). The effect on streptonigrin sensitivity of growth in cation-depleted media was assessed by measurement of the radius of growth inhibition surrounding a streptonigrin disk for bacteria cultured on RPMIm plates with or without supplementation with the cation chelator 2,2′-dipyridyl (DIP). The radius of growth inhibition was smaller for the pit⁻ strains compared with the wild-type strain, confirming that these strains are less sensitive to streptonigrin, and these differences were accentuated by supplementation with 400 µM DIP (Fig. 3B). Strikingly, in the presence of DIP, the pit1B⁻/pit2A⁻ double mutant was completely resistant to 5 µg streptonigrin disks (Fig. 3B). The addition of 25 µM FeCl₃(Fig. 3B), but not 25 µM MnSO₄ (data not shown), to plates supplemented with DIP restored bacterial streptonigrin sensitivity, confirming that the action of streptonigrin is iron dependent (Fig. 3B).

Figure 3. Sensitivity to streptonigrin of the pit⁻ strains.

A. Results expressed as the log₁₀ ratio of mutant bacteria compared with wild-type bacteria surviving after incubation with 2.5 µg ml⁻¹ f streptonigrin for 40 or 60 min. The results are from three separate experiments. Error bars indicate standard error of the means. Statistical comparisons at 40 min: pit1B⁻ versus PPC4, P = 0.07; pit2A⁻ versus PPC29, P = 0.04; pit1B⁻/pit2A⁻ versus pit2A⁻, P = 0.0003. Statistical comparisons at 60 min: pit1B⁻ versus PPC4, P = 0.06; pit2A⁻ versus PPC29, P = 0.005; pit1B⁻/pit2A⁻ versus pit2A⁻, P = 0.003.

B. Results expressed as the radius of growth inhibition (mm) surrounding an antibiotic disk impregnated with 5 µg of streptonigrin for almost confluent cultures on RPMIm plates (RPMIm), on RPMIm plates containing 400 µM DIP (RPMIm + DIP) and on RPMIm plates containing 400 µM DIP and 50 µM FeCl₃ (DIP + FeCl₃). Error bars represent the 95% confidence intervals.

⁵⁵FeCl₃ uptake by pit⁻ mutant strains

Direct evidence for a role in iron uptake of the pit loci was obtained by measuring the uptake of ⁵⁵FeCl₃ by the mutant and wild-type strains. After 15 min incubation in RPMIm medium containing 0.2 µCi of ⁵⁵FeCl₃ no differences were detected between the wild-type and the single pit1B⁻ and pit2A⁻ strains. However, the ⁵⁵FeCl₃ level was significantly lower for the pit1B⁻/pit2A⁻ strain (Fig. 4A). To demonstrate that the lower level of ⁵⁵FeCl₃ for the pit1B⁻/pit2A⁻ strain at 15 min was caused by a reduced rate of iron uptake, the levels of ⁵⁵FeCl₃ for the wild-type and pit1B⁻/pit2A⁻ strains were compared after 15 and 30 min incubation with ⁵⁵FeCl₃. Between 15 and 30 min, the ⁵⁵FeCl₃ content of the wild-type strain increased by 158%, whereas the ⁵⁵FeCl₃ content of the pit1B⁻/pit2A⁻ strain increased by 42%, confirming that the pit1 and pit2 loci function as iron transporters (Fig. 4B).

Figure 4. A. ⁵⁵FeCl₃ uptake by the wild-type, pit1B⁻, pit2A⁻ and pit1B⁻/pit2A⁻ strains after 15 min. Data presented from three assays per strain. For the differences between the wild-type and the pit1B⁻/pit2A⁻ strain at 15 min, P = 0.05 (Student's t-test).

B. ⁵⁵FeCl₃ uptake by the wild-type and pit1B⁻/pit2A⁻ strains after 15 and 30 min. Data presented from two assays per strain. For the differences between the wild-type and the pit1B⁻/pit2A⁻ strains grown in THY at 30 min, P = 0.02 (Student's t-test). Error bars represent the standard deviation.

Virulence of pit⁻ strains in mouse models of pneumonia and systemic infection

The effect on virulence of mutations in pit1B or pit2A or both genes was investigated in mouse models of pulmonary (intranasal inoculation, IN) and systemic (intraperitoneal inoculation, IP) infection. The ability of the strains to cause fatal disease was assessed using survival curves. No significant differences were found between the survival of mice inoculated with the wild-type and the single pit1B⁻ or pit2A⁻ strains in pulmonary infection (inoculum of 5 × 10⁶ cfu, giving 80–100% mortality after 5 days) or between the wild-type and the single pit2A⁻ strains in systemic infection (inoculum of 50 cfu, giving 100% mortality after 48 h) (Fig. 5A and B). In contrast, for the pulmonary infection model after a transient illness with mild pilo-erection and decrease in mobility during the first 36 h after inoculation, all mice inoculated with the pit1B⁻/pit2A⁻ strain recovered and survived (Fig. 5A). Systemic infection with the double mutant pit1B⁻/pit2A⁻ strain resulted in a mortality of 90%, but the time of death was significantly delayed compared with the wild-type strain (P = 0.002, log rank test) (Fig. 5B).

Figure 5. Survival of groups of 10 mice inoculated with pit⁻ mutant strains.

A. IN inoculation (5 × 10⁶ cfu).

B. IP inoculation (5 × 10¹ cfu).

Diamonds, wild-type strain; squares, pit1B⁻ strain; triangles, pit2A⁻ strain; circles, pit1B⁻/pit2A⁻ strain. For the differences in survival between the wild-type strain and the pit1B⁻/pit2A⁻ strain after IN inoculation, P < 0.0001. For the differences in survival between the wild-type strain and the pit1B⁻/pit2A⁻ strain after IP inoculation, P = 0.002.

Although single pit⁻ mutants were capable of causing lethal infections, the attenuation of the double mutant clearly showed that these loci are important for virulence. Therefore, we also investigated the virulence of the pit⁻ strains using mixed infections and the competitive index (CI), which provides a more sensitive measure of virulence attenuation (Beuzón et al., 2000). Three separate mixed infection experiments demonstrated a consistent attenuation in the virulence of the pit2A⁻ strain compared with the wild-type strain in both the pulmonary and the systemic infection model (Table 2). The pit1B⁻ strain was mildly attenuated in the pulmonary infection model, but was not attenuated in the systemic infection model (Table 2). Hence, the pit loci seem to have different roles during infection, with pit2 being of greater importance for both pulmonary and systemic infection. The mutant strain containing an insertion immediately downstream of the pit1 locus, PPC4, was not reduced in virulence (Table 2), confirming that the reduced virulence of the pit1B⁻ strain was not caused by a polar effect on genes downstream of the operon. The mutation created immediately downstream of the pit2 locus was unstable, and this mutant strain, PPC29, was not tested in vivo. The double pit1B⁻/pit2A⁻ strain was profoundly attenuated in virulence compared with the wild-type strain in both the pulmonary and the systemic models of infection (Table 2). No pit1B⁻/pit2A⁻ colonies were recovered from the spleen 24 h after IP inoculation in a mixed inoculum with the wild-type strain, and ≈ 10³ fewer pit1B⁻/pit2A⁻ colonies were recovered from the lungs than from wild-type colonies after IN inoculation (1.3 × 10⁴ compared with 6.8 × 10⁷, respectively, n = 3). These results indicate that mutation of both pit1B and pit2A has a synergistic effect on the virulence attenuation of S. pneumoniae.

Table 2.  Virulence in mice of pit⁻ strains assessed by mixed infections.

Strains tested	IN inoculation			IP inoculation
	CI lungs (SD)	CI spleen (SD)	n^a	CI spleen (SD)	n^a
a .n, number of mice. For each experiment, results are from two to five mice. b .P = 0.003. c .P = 0.04. d .P = < 0.0001. e. For experiments with the pit1B^−/pit2A^− strain, statistical significance for CIs could not be calculated because of the low or non-recovery of pit1B^−/pit2A^− colonies from mice.
pit1B^− versus wild type	0.65^b (0.08)	0.4^c (0.35)	6	1.13 (0.34)	9
pit2A^− versus wild type	0.13^d (0.14)	0.14^d (0.15)	11	0.28^d (0.11)	12
pit1B^−/pit2A^− versus wild type^e	< 0.001	< 0.001	7	< 0.001	4
PPC4 versus wild type	1.03	0.94	3	1.45	4

Pit2 is encoded on a pathogenicity island

The closest homologue of the ORF immediately 3′ to the pit2 locus (ORF 1) is a putative recombinase carried by the Staphylococcus aureus mec mobile genetic element (25% identity over 551 amino acids; Table 3). mec is an ‘antibiotic resistance island’ that confers methicillin resistance, and the recombinase may catalyse the recombination of mec into S. aureus chromosomal DNA (Ito et al., 1999). Analysis of the S. pneumoniae genome sequence showed that the pit2 locus has a markedly lower G+C content than the upstream DNA region (32.0% versus 40.1%), with a striking drop occurring within the 3′ terminus of the ORF immediately upstream of the pit2 locus (ORF B). We therefore investigated the possibility that pit2 may be part of a pathogenicity island (PAI) using the available genome sequence from a serotype 4 S. pneumoniae strain.

Table 3.  Results of blast searches for the ORFs within and adjacent to PPI-1 (excluding pit2A, B, C and D).

ORF	Size	BLAST search results				Closest streptococcal homologue
		Closest homologue^a	Probable function	Organism	% Identity/similarity^b (residues)	Organism	% Identity/similarity^b (residues)
a . Excluding incomplete bacterial genomes. b . At the amino acid level.
A	299	ypuA	Unknown	Bacillus subtilis	27/45 (281)	S. equi	62/76 (320)
B	543	yefA	RNA methyltransferase	B. subtilis	42/62 (449)	S. equi	72/84 (503)
1	559	ABO14436	Recombinase	S. aureus	25/46 (551)	S. pyogenes	24/45 (518)
2	195	None	–	–	–	None	–
3	252	None	–	–	–	None	–
4	252	None	–	–	–	None	–
5	294	yerQ	Unknown	B. subtilis	32/52 (297)	S. mutans	32/51 (303)
6	395	yvdF	Glucan-1,4-maltohydrolase	B. subtilis	53/68 (362)	S. pyogenes	37/59 (381)
7	105	None	–	–	–	None	–
8	253	Protein zeta	Unknown	Plasmid pBT233 (S. pyogenes)	42/65 (213)	None (within chromosomal DNA)	–
9	257	None	–	–	–	None	–
10	607	L29324	Relaxase	S. pneumoniae	62/78 (448)	S. pyogenes	42/64 (469)
11	271	plcR	Transcription factor	Bacillus thuringiensis	23/49 (198)	S. mutans	22/42 (191)
12	448	None	–	–	–	None	–
13	305	ytgB	ATPase	B. subtilis	34/52 (216)	S. pyogenes	33/51 (210)
14	266	None	–	–	–	None	–
C	157	AF125554	Transposase	Enterococcus faecium	70/80 (149)	S. equi	60/74 (150)
D	409	ftsW	Cell division protein	Enterococcus hirae	31/49 (366)	S. equi	59/72 (357)
E	841	D89668	PEP carboxylase	Rhodopseudomona spalustris	37/55 (864)	S. mutans	66/78 (848)

Analysis of the mean G+C content of consecutive 800 bp segments of 43.2 kb of DNA including the pit2 locus identified an area of 27 200 bp with a mean G+C content of 32.6 ± 4.0%. This area was termed PPI-1 (pneumococcal pathogenicity island 1). The mean G+C contents of consecutive 800 bp segments for 3.2 kb to the left of PPI-1 is 40.1 ± 2.1%, and for 12.8 kb to the right of PPI-1 is 39.4 ± 2.7%. These are both significantly different from the mean G+C content of PPI-1 (P < 0.001, analysed using contrast coefficients) (Fig. 6A and B). The mean G+C content for 44 kb and 30 kb of chromosomal DNA flanking the right and left boundaries of PPI-1 are 39.0% and 40.1%, respectively, similar to the values for a 250 kb length of chromosomal DNA containing pit1 (38.9%) and to the estimate calculated for the S. pneumoniae genome by chemical methods (38.5%; Hardie, 1986). There are no regions longer than 1.5 kb with a low G+C content within the 44 kb and 30 kb of chromosomal DNA flanking PPI-1. The boundaries of PPI-1 are marked by sharp decreases in G+C content (Fig. 6A and B) and were defined to within 50 bp by G+C content analysis of 200 bp lengths of DNA overlapping by 10 bp. The left-hand boundary lies within the C-terminus of the first ORF 5′ to the pit2 locus, which is likely to encode an RNA methyltransferase (ORF B, 42% identity over 449 amino acids to yefA of Bacillus subtilis) (Fig. 6; Table 3). The right-hand boundary of PPI-1 lies between an ORF with no close homologues in the databases and a probable transposase gene that belongs to the insertion sequence family IS605 (ORF C, 70% identity over 149 amino acids to AF125554 of Enterococcus faecium;Fig. 6; Table 3) (Mahillon and Chandler, 1998). IS605 transposons have been reported in S. pneumoniae (Oggioni and Claverys, 1999) and are characterized by 5′ hairpin loops and a low frequency of transposition, but do not contain terminal inverted repeats (Beuzón and Casadesús, 1997; Mahillon and Chandler, 1998). There is a hairpin loop 159 bp 5′ to the transposase (AAAAGTTGGGTCATCCAACTTTT, ΔG 10⁻¹²), and visual analysis of the DNA sequence at the PPI-1 boundaries did not identify inverted or direct DNA repeats. In addition, PPI-1 also contains two ORFs associated with mobile genetic elements, the DNA recombinase gene described above and a DNA relaxase gene (ORF 11, 62% identity over 448 amino acid residues to L29324 of S. pneumoniae). Two regions within PPI-1 of ≈ 4.0 kb and 3.2 kb have a mean G+C content close to that of the S. pneumoniae chromosome (37.9% ± 1.4%, compared with mean G+C content of PPI-1 P < 0.001) (Fig. 6A and B). No BOX or RUP sequences (repeated extragenic S. pneumoniae elements sometimes associated with insertion sequences) were identified within PPI-1 (Martin et al., 1992; Oggioni and Claverys, 1999). In view of the evidence suggesting that pit2 is carried within a PAI, we analysed the G+C contents for several genes encoding known S. pneumoniae virulence determinants and for the pit1 locus. However, in each case, the G+C content is similar to the mean value for the S. pneumoniae chromosome (cbpA 41.4%, ply 42.0%, psaA 37.3%, pspA 40.1%, and pit1BCDA 40.1%).

Figure 6. Structure of PPI-1.

A. G+C content plot for PPI-1 and surrounding sequence calculated with 400 bp windows overlapping every 50 bp using windows application (Genetics Computer Group, University of Wisconsin). Dashed line represents mean G+C content of S. pneumoniae DNA (38.9%).

B. Representation of the G+C content of PPI-1 and the surrounding DNA (calculated using the percentage G+C content of successive 800 bp blocks).

C. ORFs flanking and contained within PPI-1 and results of Southern analysis of DNA from S. pneumoniae and related streptococci probed with internal fragments of the ORFs under non-stringent conditions (plus sign, positive hybridization signal; minus sign, no hybridization signal). All ORFs are transcribed from left to right. Shading of ORF boxes represents percentage identity of the derived amino acid sequence of that ORF to corresponding sequences from ORFs of S. pyogenes, S. mutans and S. equii. Black shading, > 59% identity; dark grey shading, 35–45% identity; speckled shading, 20–30% identity; white, no identity to available non-S. pneumoniae streptococcal genomes; diagonal shading, pit2 genes.

In addition to pit2, there are 14 ORFs with predicted amino acid products longer than 100 residues within PPI-1 (Fig. 6C). Incomplete genome sequences are available for three species distantly related to S. pneumoniae (Streptococcus pyogenes, Streptococcus mutans and Streptococcus equi) (Kawamura et al., 1995), allowing investigation of the distribution among streptococci of the ORFs present within and adjacent to PPI-1. The five ORFs flanking PPI-1 have > 58% amino acid identity to ORFs from at least two non-S. pneumoniae streptococcal species, whereas eight of the 14 ORFs within PPI-1 have no significant similarity to chromosomal ORFs from non-S. pneumoniae streptococcal species (Fig. 6C; Table 3). The predicted amino acid sequences of the remaining ORFs have less than 42% identity to amino acid sequences of ORFs from streptococcal species, similar to or lower than their level of identity to non-streptococcal ORFs (Table 3). Furthermore, these ORFs include parts of mobile elements (as discussed above) or encode members of families of proteins that have multiple representatives within a given genome (ORF 13 has 34% identity over 216 amino acids to an ATPase of B. subtilis, and ORF 11 has 23% identity over 198 amino acids to a transcription factor of Bacillus thuringiensis) (Table 3). Pit2A has high degrees of similarity only to predicted proteins from non-streptococcal species, whereas Pit1A has high degrees of identity to a predicted protein from S. mutans (35% identity over 332 amino acids).

The distribution of the pit loci and the ORFs within and surrounding PPI-1 in streptococcal species more closely related to S. pneumoniae (Kawamura et al., 1995) was investigated by Southern analysis using internal fragments of the pit genes and PPI-1 ORFs as probes under non-stringent conditions (Figs 6C and 7). The pit1A probe hybridized to genomic DNA fragments from Streptococcus mitis, Streptococcus sanguis, Streptococcus oralis and Streptococcus milleri, whereas the pit2A probe only hybridized to S. pneumoniae DNA (Fig. 7). Hence, pit1 is widely distributed among streptococcal species closely related to S. pneumoniae, whereas pit2 is restricted to S. pneumoniae. PCR with primers specific for internal fragments of pit1B, pit2A, ORF 1 and ORF 3 of PPI-1 amplified identical fragments from 11 S. pneumoniae strains (representing 10 different serotypes), showing that both genes and at least part of PPI-1 are present in all the strains tested (data not shown). The results of Southern analysis of DNA of streptococcal species closely related to S. pneumoniae probed with internal fragments of ORFs within and adjacent to PPI-1 under non-stringent conditions are represented in Fig. 6C. Hybridization signals were obtained from other streptococcal species for ORFs B and D, which flank the 5′ and 3′ ends of PPI-1. However, the majority of ORFs within PPI-1 gave no hybridization signals, confirming that PPI-1 is not present in the streptococcal species most closely related to S. pneumoniae.

Figure 7. Southern analysis of the distribution of the pit1 and pit2 loci among streptococcal species. Each lane contains EcoRI-digested genomic DNA from a different Streptococcus species, probed under non-stringent conditions with an internal portion of the pit1A gene (A) or an internal portion of the pit2A gene (B).

Discussion

The mechanisms of iron acquisition by Gram-positive pathogens remain largely uncharacterized, and virtually nothing is known of the molecular basis of iron uptake by S. pneumoniae. We have identified two S. pneumoniae loci, pit1 and pit2, which encode ABC transporters involved in iron uptake. Mutations in both pit1 and pit2 had a markedly greater effect on growth in iron-depleted medium, streptonigrin sensitivity and ⁵⁵FeCl₃ uptake compared with single mutations. The effect of both mutations on attenuation of virulence was dramatic and greater than the predicted effects of combining the single mutations. The simplest explanation for these results is that both ABC transport systems operate independently in acquiring iron and that loss-of-function of one system can be partially compensated for by the other, but loss of both systems drastically reduces the ability of the organism to acquire exogenous iron and survive in vivo. We have not directly linked the loss of virulence of the pit1B⁻/pit2A⁻ strain to a reduced ability to acquire iron in vivo. However, the synergistic effect of the double mutation on iron transport in vitro and on virulence strongly suggests that the attenuation of virulence is caused by a failure to acquire iron. The restoration of growth of the double mutant in cation-depleted medium by micromolar amounts of free iron suggests that S. pneumoniae may also have low-affinity uptake mechanism(s) distinct from Pit1 and Pit2.

Although Pit1A and Pit2A do not contain the conserved motifs present in Gram-negative outer membrane haem receptors (Bracken et al., 1999) and their highest degrees of similarity are to periplasmic iron-siderophore receptors, the data show that Pit1 and Pit2 can acquire iron from haemoglobin. Several mechanisms for acquiring iron from haemin and haemoglobin have been described and include haem capture by secreted haemophores, direct binding of host haem-containing proteins to bacterial cell surface receptors and degradation of host haem-containing proteins by extracellular proteases (Genco and Dixon, 2001). Whether Pit proteins bind haemoglobin directly, conforming to a proposed model for a C. diphtheriae haemin receptor (Drazek et al., 2000), or indirectly through an unidentified secreted or extracellular receptor is not known and requires further investigation. Our data suggest that the Pit iron transporters are required for high-affinity ferrous or ferric iron uptake as well as iron capture from haemoglobin, but which iron ligand is transported into the cytosol is unclear. Possibly, both iron and haem are transported into the cytosol by the Pit ABC transporters, or iron may be removed from haem at the bacterial surface by an unknown mechanism and transported into the cytosol as ferric or ferrous iron. As there is an influx of red blood cells into alveoli in the early phases of pneumonia, haemoglobin and/or haem will be available as iron sources for S. pneumoniae during invasive infection. Whether haemoglobin or haemin would be available in sufficient amounts to support growth of S. pneumoniae during nasopharyngeal colonization is unclear.

Many important virulence genes of Gram-negative pathogens are contained within distinct regions of chromosomal DNA termed PAIs that have been acquired by horizontal transmission (Hacker and Kaper, 2000; Ochman et al., 2000). In contrast, only a small number of Gram-positive PAIs have been identified, and most of the described virulence genes of Gram-positive pathogens do not appear to be within PAIs. The pit2 locus is in a chromosomal region designated PPI-1, which has most of the characteristic features of Gram-negative PAIs (Hacker et al., 1997), namely (i) a significantly different G+C content from the overall G+C content of the chromosome; (ii) contains genes encoding proteins likely to be associated with the transfer of DNA between bacteria (a recombinase, a relaxase and a transposase); (iii) is not present in less pathogenic but closely related species; and (iv) contains genes that encode a function essential for full virulence (as demonstrated by the mixed infection data for the pit2A⁻ strain). Like many Gram-negative PAIs (Shea et al., 1996; Hacker et al., 1997), but unlike SaPI1 of S. aureus (Lindsay et al., 1998), PPI-1 seems to be integrated into the chromosome in a stable fashion, and pit2 is present in all the S. pneumoniae strains investigated. Genes encoding iron uptake systems are often associated with PAIs and are occasionally the only virulence determinants within the PAI (Buchrieser et al., 1998; Vokes et al., 1999; Hacker and Kaper, 2000). Whether other ORFs within PPI-1 are also involved in virulence is unknown at present. The low degree of identity between pit2 and its closest S. pneumoniae homologue pit1 makes gene duplication an unlikely explanation for the origin of pit2, and PPI-1 was probably acquired by horizontal transfer from a non-streptococcal species. The closest known homologue to pit2A is IRP1, an iron-regulated C. diphtheriae gene (Lee et al., 1997). C. diphtheriae is also a respiratory tract pathogen and could have acquired IRP1 from the same source as S. pneumoniae. PPI-1 contains two regions of higher G+C content, suggesting that it has a mosaic structure that may reflect multiple integration events at this locus.

PAIs are thought to have contributed significantly to the evolution of distinct species of Gram-negative pathogens (Hacker and Kaper, 2000; Ochman et al. 2000) but, as few PAIs have been identified in Gram-positive pathogens, their influence on Gram-positive bacterial evolution is much less apparent. S. pneumoniae is naturally transformable with partially homologous DNA, and the incorporation of DNA from related streptococci results in significant genetic variation in S. pneumoniae (Claverys et al., 2000). PPI-1 appears to be a PAI, the first to be described for S. pneumoniae, and this suggests that, in addition to homologous recombination, S. pneumoniae can also acquire large regions of heterologous DNA from unrelated species. PPI-1 may have inserted by transposition or possibly by homology-directed illegitimate recombination promoted by a short length of homology between PPI-1 and S. pneumoniae DNA (Claverys et al., 2000). The identification of PPI-1 shows that S. pneumoniae can obtain specific virulence functions through the acquisition of PAIs, and there may be additional PAIs of this type in S. pneumoniae and other Gram-positive pathogens. As S. pneumoniae strains with a functioning pit2 locus have a survival advantage during invasive infections, the acquisition of pit2 may be one reason why S. pneumoniae is a more virulent pathogen than the closely related viridans streptococci.

Experimental procedures

Bacterial strains, media and culture conditions

S. pneumoniae strains used for this work are listed in Table 4. A type 3 S. pneumoniae strain, 0100993, isolated from a patient with pneumonia and obtained from SmithKline Beecham plc was used for construction of the mutant strains, phenotype analysis and in vivo studies. Strains for Southern analysis and PCR were obtained from W. Hanage (S. mitis, S. oralis, S. sanguis, Streptococcus gordonii and S. milleri), S. Sriskandan (S. pyogenes H305) and J. Paton (10 S. pneumoniae clinical isolates representing serotypes 2, 4, 7F, 17, 18C, 19A, 19F, 20 and 22). S. pneumoniae strains were cultured at 37°C and 5% CO₂ on Columbia agar supplemented with 5% horse blood, in Todd–Hewitt broth supplemented with 0.5% yeast extract (THY) or using a modified version of RPMI medium (RPMIm; Cockayne et al., 1998). RPMIm was made by the addition to the tissue culture medium RPMI (type 1640; Gibco) of 0.4% BSA factor V (Sigma), 1% vitamin solution (Sicard, 1964) and 2 mM glutamine. Medium was cation depleted by adding 2% (THY) or 6% (RPMIm) Chelex-100 (Bio-Rad) to broth medium for 8 (THY) or 24 (RPMIm) h under continuous agitation. The Chelex-100 was removed by filter sterilization, and the medium was supplemented with 100 µM CaCl₂ and 2 mM MgSO₄. When necessary, the following supplements were added to the medium: chloramphenicol (cm) 4 µg ml⁻¹; erythromycin (ery) 0.4 µg ml⁻¹; 10–50 µM FeCl₃; 10–50 µM FeSO₄; 1–5 µg ml⁻¹ lactoferrin (Sigma); 1–5 µg ml⁻¹ ferritin (Sigma); 5–10 µM human haemoglobin (Sigma); 1–10 µM haemin (Sigma); 5–25 µM MnCl₂; 5–25 µM ZnCl₂; and 400 µM 2,2′-DIP (Sigma). Broth culture growth was monitored by measuring optical density (OD) at 580 nm. To minimize Fe contamination, disposable plasticware was used to store stock solutions (made with MilliQ-purified water). Strains were stored at −70°C as aliquots of THY broth culture (OD₅₈₀ of 0.3–0.4) containing 10% glycerol. Plasmids were amplified in Escherichia coli strain DH5α, grown at 37°C on Luria–Bertani (LB) medium with appropriate selection (Sambrook et al., 1989).

Table 4.  Strains, plasmids and primers constructed and used in this study.

Name	Description/sequence
Strains
0100993	Serotype III clinical isolate
pit1B^−	0100993 containing an insertion in pit1B made with plasmid pPC5: cm^r
pit2A^−	0100993 containing an insertion in pit2A made with plasmid pPC12: cm^r
pit1B^−/pit2A^−	0100993 containing an insertion in pit1B and in pit2A made with plasmid pPC12 and pPC25: ery^r cm^r
PPC4	0100993 containing an insertion 73 bp downstream of pit1B made with plasmid pPC4: cm^r
PPC29	0100993 containing an insertion 100 bp downstream of pit2D made with plasmid pPC29: cm^r
Plasmids
pID701	Disruption vector for S. pneumoniae derived from pEVP3 (Lau et al., 2001): amp^r cm^r
pACH74	Disruption vector for S. pneumoniae (gift from J. Paton): cm^r ery^r
pPC4	pID701 with Smt3.3/Smt3.4 PCR product ligated into the XbaI site: amp^r cm^r
pPC5	pID701 with Smt6.1/Smt6.2 PCR product ligated into the XbaI site (pit1B^− disruption vector): amp^r cm^r
pPC12	pID701 with IRP1.1/IRP1.2 PCR product ligated into the XbaI site (pit2A^− disruption vector): amp^r cm^r
pPC25	pACH74 with Smt6.3/Smt6.4 PCR product ligated into the KpnI site (pit1B^− disruption vector): cm^r ery^r
pPC29	pID701 with IRP1.7/IRP1.8 PCR product ligated into the XbaI site: amp^r cm^r
Primers
Smt6.1	GCTCTAGACCCACAAGATGCTCTTCG
Smt6.2	CGCTCTAGACGCTTGTCGTTAGCGCCACC
Smt6.3	GGGGTACCCACAAGATGCTCTTCG
Smt6.4	CGGGGTACCGCTTGTCGTTAGCGCCACC
Smt3.3	GCTCTAGAAGCTATCGCCGCCCTTGAG
Smt3.4	CGCTCTAGAGCAACCTGCGGCTAGTTTCC
IRP1.1	GCTCTAGAGTTTTAGATCATGCTTTCGG
IRP1.2	CGCTCTAGATTTGTATGCTGCTACAGGAGC
IRP1.7	GCTCTAGATTGGGTCAAATGGTTGTGG
IRP1.8	CGCTCTAGAACTAGTCGTTGTACTTTC
ORFB.1	CACACCGTAATCAAGATC
ORFB.2	CTGGTCCGTAATATAGTC
ORFD.1	GGTCTATTCGACCACCAG
ORFD.2	CTGGTGACCTGCATCAGC
ORF1.1	CACGAGTATCTACGTC
ORF1.2	CAACTGCCTCTTCCTC
ORF3.1	GCTCTAGAGTAAATTACCAAGTGAGG
ORF3.2	CGCTCTAGATCACCTGTATAGGGTCG
ORF8.1	GCTCTAGAGCAAGGAATCTTCGTTCACTG
ORF8.2	CGCTCTAGAGGCCAATTGTACTTCATATC

DNA isolation and manipulation

Plasmid DNA was isolated from E. coli using Qiagen plasmid kits, and S. pneumoniae chromosomal DNA was isolated using Wizard genomic DNA isolation kits (Promega). Standard protocols were used for cloning, transformation, restriction digests and ligations of plasmid DNA (Sambrook et al., 1989). Nylon membranes for Southern hybridizations were prepared and probed using ³²P-dCTP-labelled probes made using the RediPrime random primer labelling kit (Amersham International) as described previously (Holden et al., 1989). Non-stringent hybridizations were performed at 47°C using 0% formamide in the hybridization fluid and washed at 37°C. Probes for PPI-1 were made by amplifying internal portions of PPI-1 ORFs by PCR using primers listed in Table 4.

RNA isolation and RT–PCR

Total RNA was isolated from S. pneumoniae cells cultured in THY medium to an OD₅₈₀ of 0.4 using the SV Total RNA isolation system (Promega). Before RNA isolation, cells were lysed by incubation with 0.1% deoxycholic acid (Sigma) at 37°C for 10 min. Prepared RNA was protected from degradation by the addition of 0.5% RNasin (Promega) and storage as single-use aliquots at −70°C. cDNA was derived from RNA and amplified using the Access RT–PCR system (Promega) and target-specific primers.

Computer analysis of nucleic acid sequences

Preliminary S. pneumoniae sequence data were obtained from The Institute for Genomic Research website (http://www.tigr.org) and analysed and manipulated using macvector (International Biotechnologies). blast2 searches of available nucleotide and protein databases and of incomplete microbial genomes were performed using the NCBI website (http://www.ncbi.nlm.nih.gov/blast/). DNA G+C content was analysed using artemis (Genome Research), and graphs of G+C content were made with the windows application of the Wisconsin sequence analysis package (Genetics Computer Group).

Construction of mutant strains

Plasmids, primers and S. pneumoniae strains constructed and used in this work are described in Table 4. S. pneumoniae mutant strains were constructed by insertion–duplication mutagenesis. Internal portions of the target genes were amplified by PCR using primers designed from the available genomic sequence and ligated into pID701 (a suicide vector derived from pEVP3 carrying cat for selection in S. pneumoniae and E. coli) (Claverys et al., 1995; Lau et al., 2001) or pACH74 (a suicide vector carrying erm for selection in S. pneumoniae and cat for selection in E. coli, a gift from J. Paton). To make a pit1B⁻ disruption vector, pPC5, an internal portion of pit1B from bp 320 to 711 was amplified using primers Smt6.1 and Smt6.2, digested with XbaI and ligated into the XbaI site of pID701. To make a pit2A⁻ disruption vector, pPC12, an internal portion of pit2A from bp 84 to 428 was amplified using primers IRP1.1 and IRP1.2, digested with XbaI and ligated into the XbaI site of pID701. To make a pit1B⁻ disruption vector, pPC25, for construction of the double mutant, an internal portion of pit1B from bp 320 to 711 was amplified using primers Smt6.3 and Smt6.4, digested with KpnI and ligated into the KpnI site of pACH74. Vectors designed to insert plasmid DNA 73 bp downstream of the stop codon of pit1A (plasmid pPC4) and 100 bp downstream of sit2D (plasmid pPC29) were constructed by ligating PCR products generated by primer pairs Smt3.3/Smt3.4 and IRP1.7/IRP1.8, respectively, into the XbaI site of pID701. Plasmid insert identities were confirmed by DNA sequencing.

S. pneumoniae strains were transformed using a modified protocol requiring induction of transformation competence with competence-stimulating peptide 1 (CSP1) (Håvarstein et al., 1995; Lau et al., 2001). S. pneumoniae strains containing a single mutation were made by transformation of strain 0100993 with pPC4, pPC5, pPC12 or pPC29, and the double pit1B⁻/pit2A⁻ strain was constructed by transformation of the pit2A⁻ strain with pPC25. Mutations were confirmed by PCR and Southern hybridization. All mutations were stable after two 8 h growth cycles in THY broth without antibiotic selection, except PPC29 in which 35 of 100 colonies lost resistance to chloramphenicol.

Streptonigrin sensitivity assays and ⁵⁵Fe transport assays

For streptonigrin assays, thawed stocks of S. pneumoniae strains were pelleted by centrifugation at 20 000 g at 4°C, resuspended in THY containing 2.5 µg ml⁻¹ streptonigrin (Sigma) and incubated at 37°C. Aliquots of the reaction cultures were diluted and plated after 40 and 60 min, and the cfu from each time point was expressed as a percentage of the cfu before adding streptonigrin. Mean data from triplicate experiments are presented as a log₁₀ ratio of the wild-type strain's results. To assess sensitivity to streptonigrin disks, S. pneumoniae strain stocks cultured in Chelex-THY broth were plated on RPMIm plates with or without 50 µM FeCl₃ and 400 µM DIP supplementation at a density of several thousand colonies per plate. Antibiotic disks impregnated with 5 µg of streptonigrin were placed onto the plates, and they were incubated for 20 h at 37°C in 5% CO₂. The width of the zone of growth inhibition surrounding each disk was measured, and confidence intervals for three separate results were calculated.

⁵⁵Fe transport assays were modified from previously described protocols (Bearden and Perry, 1999). Approximately 5 × 10⁷ cfu from −70°C stocks of S. pneumoniae strains were subjected to iron stress by incubation for 1 h at 37°C in 3 ml of RPMIm before the addition of 0.2 µCi ml^{−1 55}FeCl₃ (NEN). The reactions were incubated for 15 or 30 min at 37°C before being filtered through 0.45 µM nitrocellulose filters (Millipore). The membranes were washed with 10 ml of RPMI, allowed to dry and placed in 10 ml of Optisafe scintillation fluid (Wallac). The radioactivity was counted using a Beckman LS 1801 scintillation counter. To reduce background radioactivity, the filters were prewashed with 40 µM FeCl₃, and the ⁵⁵FeCl₃ medium was filtered through 0.2 µM membranes (Sartorius) before use.

In vivo studies using mouse models of S. pneumoniae infection

Outbred male white mice (strain CD1; Charles Rivers Breeders) weighing 18–22 g were inoculated with defrosted and appropriately diluted (in 0.9% saline) stocks of S. pneumoniae strains. For mixed infections, inocula consisted of approximately equivalent numbers of cells from two strains. For the pneumonia model, mice were anaesthetized by inhalation of halothane (Zeneca), and a 40 µl inoculum containing between 5 × 10⁵ and 5 × 10⁶ bacterial cfu was administered intranasally (IN). For the systemic model, mice were given a 100 µl inoculum containing 5 × 10¹ (for survival curves) or 1 × 10³ (for mixed infections) bacterial cfu by intraperitoneal (IP) injection. Mice were sacrificed after 24 h (IP inoculations) or 48 h (IN inoculations), and target organs were recovered and homogenized in 0.5 ml of 0.9% saline. Dilutions of the homogenized organs were plated on non-selective and selective medium for calculation of the CI (the ratio of mutant to wild-type strain recovered from the mice divided by the ratio of mutant to wild-type strain in the inoculum) (Beuzón et al., 2000). For survival curves, mice were sacrificed when they exhibited the following signs of disease: hunched posture, poor mobility, weight loss and (for IN inoculation only) coughing and tachypnoea. CIs were compared with 1.0 (the predicted CI if there is no difference in virulence between the two strains tested) using Student's t-test. Survival curves were compared using the log rank method.

Acknowledgements

We are grateful to Professor James Paton, Professor Paul Williams and Dr Miguel Camara for their helpful advice, and to Javier Ruiz-Albert for his considerable help with G+C content analysis. This work was supported by a Wellcome Trust Advanced Research Fellow grant awarded to J. S. Brown.