Inactivation of deoxyadenosine methyltransferase (dam) attenuates Haemophilus influenzae virulence


  • Michael E. Watson Jr,

    1. Seattle Biomedical Research Institute, 307 Westlake, Suite 500, Seattle, WA 98109-5219, USA.
    2. Department of Molecular Microbiology and Immunology, School of Medicine, University of Missouri-Columbia, Columbia, MO 65212, USA.
    Search for more papers by this author
  • Justin Jarisch,

    1. Seattle Biomedical Research Institute, 307 Westlake, Suite 500, Seattle, WA 98109-5219, USA.
    Search for more papers by this author
  • Arnold L. Smith

    Corresponding author
    1. Seattle Biomedical Research Institute, 307 Westlake, Suite 500, Seattle, WA 98109-5219, USA.
    Search for more papers by this author


Mutants in deoxyadenosine methyltransferase (dam) from many Gram-negative pathogens suggest multiple roles for Dam methylase: directing post-replicative DNA mismatch repair to the correct strand, guiding the temporal control of DNA replication and regulating the expression of multiple genes (including virulence factors) by differential promoter methylation. Dam methylase (HI0209) in strain Rd KW20 was inactivated in Haemophilus influenzae strains Rd KW20, Strain 12 and INT-1; restriction with Dam methylation-sensitive enzymes DpnI and DpnII confirmed the absence of Dam methylation, which was restored by complementation with a single copy of dam ectopically expressed in cis. Despite the lack of increased mutation frequency, the dam mutants had a 2-aminopurine-susceptible phenotype that could be suppressed by secondary mutations in mutS, suggesting a role for Dam in H. influenzae DNA mismatch repair. Invasion of human brain microvascular endothelial cells (HBMECs) and human respiratory epithelial cells (NCI-H292) by the dam mutants was significantly attenuated in all strains, suggesting the absence of a Dam-regulated event necessary for uptake or invasion of host cells. Intracellular replication was inhibited only in the Strain 12 dam mutant, whereas in the infant rat model of infection, the INT-1 dam mutant was less virulent. Dam activity appears to be necessary for both in vitro and in vivo virulence in a strain-dependent fashion and may function as a regulator of gene expression including virulence factors.


The Gram-negative bacterium Haemophilus influenzae is an obligate commensal of the human nasopharynx and an occasional pathogen, particularly of children less than 5 years of age. Encapsulated strains, especially serotype b, are capable of causing invasive infections such as sepsis and meningitis. Non-encapsulated, or non-typeable, H. influenzae (NTHi) strains are more common causes of infections including otitis media, sinusitis, conjunctivitis and lower respiratory tract infections, particularly in patients with chronic obstructive pulmonary disease (COPD) or cystic fibrosis (CF) (Murphy and Apicella, 1987; Tang et al., 2001). Since the introduction of the type b polysaccharide vaccine (Hib) in 1985, the incidence of invasive infections resulting from type b strains has decreased by more than 95% (Bisgard et al., 1998). However, the Hib vaccine provides no specific protection against NTHi strains, and the search for vaccine candidates for these strains continues.

Dam methylase, or deoxyadenosine methyltransferase, enzymatically transfers a methyl group from S-adenosylmethionine to adenine at N6 in the sequence 5′-GATC-3′ in double-stranded (ds)DNA (Hattman et al., 1978; Bergerat and Guschlbauer, 1990). This methylation functions as a mechanism for strand discrimination and helps to ensure replication fidelity. Mutations inactivating dam, or overproduction of Dam in Escherichia coli, increase the mutation frequency by approximately 20-fold (Marinus and Morris, 1974; Glickman, 1979; Herman and Modrich, 1981). In E. coli, dam mutants are also susceptible to, and hypermutable by, mutagens including UV irradiation and the adenine base analogue 2-aminopurine (2-AP) (Glickman et al., 1978). 2-AP is an adenine base analogue that can occasionally pair with cytosine (Ronen, 1980), a mismatch that initiates the methyl-directed mismatch repair system. Serratia marcescens dam mutants have an increased mutation frequency with UV radiation and are slightly more sensitive to inhibition of growth by UV irradiation than dam+ strains (Ostendorf et al., 1999). In contrast, Salmonella typhimurium dam mutants do not show increased UV sensitivity (Torreblanca and Casadesus, 1996). In dam+ strains, the MutSLH complex is directed to hemimethylated GATC sequences within newly replicated DNA strands. If there is a mismatch, the MutSLH complex nicks the single non-methylated strand of DNA removing the misincorporated base causing the mismatch, and then repair continues to replace the edited DNA sequence. Growth of dam mutants is slowed by 2-AP, presumably because of uncontrolled mismatch repair-mediated double-strand cleavage at unmethylated GATC sequences. This growth restriction in dam mutants is suppressed by additional mutations inactivating mutS, mutL or mutH reducing repair-mediated double-strand breaks (Glickman and Radman, 1980).

A second function of Dam methylation is the regulation of DNA replication in the origin region: E. coli strains that lack or overproduce Dam methylase have aberrant control and timing of DNA replication (Messer et al., 1985; Boye and Lobner-Olesen, 1990). A third function of Dam methylation seems to be gene regulation at the transcriptional level; GATC sequences are unequally distributed within the chromosome and tend to cluster in promoter regions or within binding sequences for global regulators such as CRP, Fnr and IHF (Henaut et al., 1996; Oshima et al., 2002). This methylation has been proposed to affect protein–DNA interactions by modifying the recognition sequences of transcriptional regulators or RNA polymerases (Marinus, 1996). Indeed, microarray analysis of damE. coli detected numerous gene dysregulations compared with parent strains: significant differences were noted in metabolism, respiration, motility and the stress response pathways (Oshima et al., 2002).

The virulence of dam mutants has been studied most extensively with S. typhimurium.Heithoff et al. (1999) found that dam-deficient S. typhimurium were proficient in colonization of the gastrointestinal tract, but were unable to invade the mucosa; these avirulent mutants were effective as live vaccines against murine typhoid fever, providing cross-protective immunity to other Salmonella serovars (Heithoff et al., 2001). Garcia-Del Portillo et al. (1999) reported that S. typhimurium dam mutants are attenuated for virulence in BALB/c mice. The same study also reported that dam mutants had reduced invasion into non-phagocytic cells, although intracellular replication in macrophages and non-professional phagocytes was preserved. Reduced invasion of the dam mutants into non-phagocytic cells correlated with reduced type III secretion of Salmonella pathogenicity island (SPI)-1-encoded effector proteins (Garcia-Del Portillo et al., 1999).

Invasive disease resulting from H. influenzae involves colonization of the respiratory mucosa, invasion across epithelium, traversing the submucosa and penetrating the endothelium from the adventitial side; invasive strains are invariably resistant to the bactericidal activity of human blood (Tang et al., 2001; St Geme, 2002). Infection occurs after NTHi colonize the upper respiratory tract using multiple mechanisms to avoid mucociliary clearance. Characterized adherence mechanisms include pili encoded by the hif gene cluster (van Ham et al., 1994; Gilsdorf et al., 1997), the protein Hap (St Geme et al., 1994), Hia and the homologous protein Hsf (Barenkamp and St Geme, 1996; St Geme et al., 1996), the high-molecular-weight adhesins HMW1 and HMW2 (Barenkamp and Bodor, 1990; Barenkamp and Leininger, 1992; Barenkamp and St Geme, 1994) and several additional factors including outer membrane proteins P2, P5 and lipooligosaccharide (LOS) (St Geme, 2002). Adherence to and invasion of H. influenzae into cultured epithelial cells (St Geme and Falkow, 1990) is probably through macropinocytosis (Ketterer et al., 1999) in a process involving the interaction of several bacterial and host factors including the binding of a phosphorylcholine (ChoP) moiety of LOS with the host cell platelet activating factor (PAF) receptor (Swords et al., 2000). In addition to cell invasion, paracytosis between NCI-H292 epithelial cells has been demonstrated as another mechanism for H. influenzae to invade the subepithelial space (van Schilfgaarde et al., 1995). Invasion of human endothelial cells by H. influenzae involves host cell-dependent uptake of bacteria and their translocation in membrane-bound vacuoles from the apical to the basal surface (Virji et al., 1991). H. influenzae bacteraemia and subsequent meningitis have been characterized extensively using the infant rat model of infection (Smith et al., 1973; Moxon et al., 1977), and the presence of the polysaccharide capsule in type b strains provides serum resistance by inhibiting binding of complement protein C3, thereby reducing engulfment by phagocytes and bacteriolysis (Noel et al., 1990). Certain NTHi are also serum resistant as a result of a non-capsule-mediated mechanism; strain INT-1 was found to evade normal human serum bactericidal activity by inhibiting or delaying C3 deposition on the cell surface (Williams et al., 2001).

A mutation in dam methylase was constructed in three strains of H. influenzae to gain insight into the role of Dam in pathogenesis: the sequenced, avirulent strain Rd KW20 (Wilcox and Smith, 1975), the virulent blood isolate invasive non-typeable-1 (INT-1) (Nizet et al., 1996) and the disease-associated non-typeable otitis media isolate Strain 12 (Barenkamp and Leininger, 1992). A dam mutant in Rd KW20 was generated previously by Bayliss et al. (2002) and, unexpectedly, did not have a hypermutable phenotype, suggesting that mismatch repair in H. influenzae may not follow the E. coli model. We have found a strain-dependent decrease in adherence, invasion and intracellular replication in H. influenzae dam mutants, suggesting that Dam methylase activity is required for virulence.


Construction of H. influenzae dam mutants

The sequenced H. influenzae Rd KW20 genome (Fleischmann et al., 1995) contains nine putative methylases including dam (HI0209), hindII methylase (HI0513), hindIII methylase (HI1392) and others. The DNA sequence of dam (b3387) from E. coli K-12 (Blattner et al., 1997) was compared with the H. influenzae database using blastx, and the only significant match was dam, HI0209, with 55% identity and 69% similarity (Altschul et al., 1997). Using the sequence of strain Rd KW20 (R652), the dam (HI0209) open reading frame (ORF) was cloned and insertionally inactivated with antibiotic resistance cassettes. H. influenzae strains Rd KW20 (R652), INT-1 (R2866) and Strain 12 (R2846) were transformed with the inactivated R652 dam construct to produce dam mutants. Transformants were confirmed by Southern analysis, and each mutant strain contained only one copy of the inactivated gene and thus was not a merodiploid (data not shown). These mutants lacked functional Dam methylation activity as shown by restriction analysis of genomic DNA with the enzymes Sau3AI, DpnI or DpnII. Figure 1 shows that genomic DNA isolated from dam+ parent strains was digested by the enzymes Sau3AI (digests all DNA) or DpnI (only acts on Dam-methylated DNA), but not by DpnII (blocked by Dam methylation). In contrast, genomic DNA isolated from the dam::TSTE mutant strains was digested by Sau3AI or DpnII, but not by DpnI, which confirms that GATC methylation was absent in the dam::TSTE mutants. The complemented Rd KW20 dam::TSTE, HI1018::dam mutant (R3547) displays a reversion back to a methylated GATC DNA phenotype indicated by the lack of cutting with DpnII (Fig. 1). These data suggest that dam (HI0209) is a Dam methylase and is responsible for methylation of GATC sequences in these three H. influenzae strains.

Figure 1.

Dam methylation profile of parent and mutant H. influenzae strains. Chromosomal DNA from each strain was digested with Sau3A1 (cuts all DNA), DpnI (only cuts Dam-methylated DNA) or DpnII (blocked by Dam methylation), and the restriction products were analysed by electrophoresis on a 0.8% agarose gel and visualized by ethidium bromide staining. M is molecular weight standards of 1 kb ladder (Promega).

No difference was noted in the in vitro growth rate of the parent strains and dam mutants tested by measuring optical density over time in sBHI broth and in liquid, chemically defined Coleman's media at 37°C in room air (data not shown). Similarly, no differences in growth rate were noted in either media when incubated anaerobically (data not shown). Morphological examination of the parent and mutant strains by Gram stain revealed a trend towards more cellular filamentation in the dam::TSTE mutants, although the extent of filamentation was not more than 10% of the total cells within any field examined. Filaments up to 14 µm in length were observed, with the average normal single cell being ≈ 1 µm in length.

Haemophilus influenzae dam mutants are not hypermutable

Bayliss et al. (2002) found that an H. influenzae Rd dam mutant was not hypermutable when tested for the development of resistance to nalidixic acid, suggesting that mismatch repair in H. influenzae may not follow the E. coli model system. In the present study, we have confirmed this surprising finding that the H. influenzae dam::TSTE mutants of Rd KW20 (R3526), Strain 12 (R3528) and INT-1 (R3549) do not have an increased frequency of mutation to rifampicin and nalidixic acid resistance (data not shown). Mutation frequencies for parent and dam mutant strains were statistically similar and were ≈ 20- to 50-fold less than the hypermutable mutS mutants also generated in each parental strain.

Haemophilus influenzae dam mutants are hypersusceptible to 2-AP

To investigate further the role of Dam methylase in DNA mismatch repair in H. influenzae, the growth of dam mutants on Coleman's media containing 2-AP was examined. We reasoned that, if dam mutants were susceptible to 2-AP, but not dam mutS double mutants, then a role for Dam in directing H. influenzae mismatch repair would be supported. Figure 2 shows the growth of Rd KW20 (R652), Rd KW20 dam::TetR (R3545) and Rd KW20 dam::TetR; mutS::TSTE (R3546) strains on Coleman's agar  around  a  6 mm  paper  disk  saturated  with  25 mg ml−1 2-AP. The presence of 2-AP resulted in a relatively large zone of growth inhibition for R3545 compared with a smaller zone of growth inhibition for R652 or R3546. The complemented Rd KW20 dam::TSTE; HI1018::dam (R3547) strain also became resistant to 2-AP with growth inhibition similar to the dam+ parent strain R652 (data not shown). Within the zone of growth inhibition of R3545 were numerous 2-AP-resistant colonies, which may be spontaneous mutants in one or more components of mutSLH as has been reported by others (Glickman and Radman, 1980). In liquid Coleman's media, we found that all H. influenzae strains, including the dam+ strains, had slower growth rates with 100 µg ml−1 2-AP: the growth of all strains in liquid media was inhibited by 2-AP concentrations at 500 µg ml−1 (data not shown). The mutation frequency to rifampicin and nalidixic acid resistance after growth in 0, 50 or 100 µg ml−1 2-AP during growth in Coleman's liquid media was not increased in the 2-AP concentrations tested (data not shown).

Figure 2.

Dam mutant susceptibility to 2-AP. Approximately 107 cfus of H. influenzae strains suspended in PBSg were spread on to Coleman's media plates and allowed to dry for 30 min. Sterile 6 mm filter paper discs were saturated with 2-AP (25 mg ml−1 in Coleman's media), placed on the plate surface, and the plates were incubated at 37°C for 24 h.

Haemophilus influenzae dam mutants are not hypersusceptible to UV irradiation

Haemophilus influenzae strains Rd KW20 (R652), Rd KW20 dam::TSTE (R3526) and DB117, a rec-1 Rd mutant, were tested for susceptibility to UV irradiation and UV irradiation-induced mutability. The survival of strains R652 and R3526 was not significantly different over the range of UV irradiation dosages tested (0–2800 µJ cm−2). At the highest UV irradiation dose, the survival of both strains was three logs greater than that of strain DB117 (data not shown). Strain DB117 was killed rapidly by the UV irradiation as would be expected for a rec-1 mutant (Beattie and Setlow, 1971). No significant increases in mutation frequency to rifampicin or nalidixic acid were noted after exposure to UV irradiation for any of the strains tested (data not shown).

Haemophilus influenzae dam mutants have a strain-specific attenuation in adherence and invasion

Adherence of H. influenzae dam and parent strains was assessed in human brain microvascular endothelial cells (HBMECs), while invasion was assessed in both HBMECs and human respiratory epithelial cells (NCI-H292s) using gentamicin protection assays. Previous work has demonstrated a time-dependent increase in H. influenzae invasion into HBMECs and NCI-H292 cells with maximal invasion achieved at 3–4 h after inoculation (unpublished data); a 4 h invasion followed by a 1 h gentamicin treatment (100 µg ml−1) was chosen as the time point most likely to identify potential differences. Results for adherence and invasion assays are shown in Table 1. Significant reductions in adherence to HBMEC cells were noted for Strain 12 dam::TSTE (R3528) compared with Strain 12 (R2846) (P < 0.01), and for INT-1 dam::TSTE (R3549) compared with INT-1 (R2866) (< 0.05). No significant reduction in HBMEC adherence was noted for Rd KW20 dam::TSTE (R3526) compared with Rd KW20 (R652). Invasion of HBMECs was greatest for Strain 12 (R2846) at 0.85 ± 0.19% inoculum invaded, whereas Rd KW20 (R652) and INT-1 (R2866) were less invasive, with frequencies of 0.40 ± 0.08% and 0.11 ± 0.02% respectively. Significant reductions in HBMEC invasion frequency were found for all dam mutants compared with parent strains: R3528 versus R2846 (P < 0.01); R3549 versus R2866 (P < 0.01); and R3526 versus R652 (P < 0.05). The complemented Rd KW20 dam mutant (R3547) had an invasion frequency similar to that of the Rd KW20 parent (R652). Only Strain 12 dam::TSTE (R3528) compared with the Strain 12 parent (R2846) had a significant decrease (< 0.01) in intracellular replication. For NCI-H292 cells, invasion frequency was greatest for Strain 12 (R2846) at 42.51 ± 20.08% inoculum invaded, whereas INT-1 (R2866) and Rd KW20 (R652) were less invasive with frequencies of 9.13 ± 1.83% and 0.35 ± 0.10% respectively. Significant reductions in NCI-H292 cell invasion frequencies were found for all dam mutants compared with parent strains: R3526 versus R652 (P < 0.01); R3549 versus R2866 (P < 0.05); and R3528 versus R2846; (P < 0.05). The cis-complemented Rd KW20 dam mutant (R3547) had an NCI-H292 invasion frequency similar to that of the Rd KW20 parent (R652). Measurement of intracellular replication in NCI-H292 cells revealed a significant reduction only for Strain 12 dam::TSTE (R3528) compared with the Strain 12 parent (R2846) (< 0.05).

Table 1. Adherence and invasion of human cells by dam mutant and parent strains.
Bacterial strainHBMEC (endothelial cells)NCI-H292 (epithelial cells)
% Cell
% InvadedbIntracellular
% InvadedbIntracellular
  • a. 

    The cell-associated bacteria represent all adherent and invaded bacteria present at 4 h after inoculation of non-gentamicin-treated wells, and are expressed as the mean percentage of inoculum ± standard error of the mean.

  • b. 

    The invaded bacteria represent all bacteria surviving 1 h of 100 µg ml−1 gentamicin treatment (5 h after inoculation), expressed as the mean percentage of inoculum ± standard error of the mean.

  • c. 

    Intracellular replication represents the ratio of gentamicin-resistant bacteria at 24 h versus 5 h after inoculation, expressed as mean ± standard error of the mean.

  • The data presented are geometric means of at least six experiments performed in duplicate.

  • ND indicates not determined. Significantly different from parent strain (*P < 0.05; **P < 0.01).

R652 (Rd KW20) 16.17 ± 3.330.40 ± 0.08 8.52 ± 0.49 0.35 ± 0.1013.50 ± 6.05
R3526 (R652 dam::TSTE) 16.00 ± 2.890.27 ± 0.07*10.24 ± 2.02 0.14 ± 0.06**18.66 ± 8.85
R3547 (R3526 dam+)ND0.42 ± 0.06 8.90 ± 1.62 0.34 ± 0.0712.46 ± 8.64
R2866 (INT-1) 20.00 ± 1.900.11 ± 0.0215.41 ± 3.97 9.13 ± 1.83 8.48 ± 2.74
R3549 (R2866 dam::TSTE)  8.70 ± 2.20*0.02 ± 0.01**16.72 ± 3.09 1.03 ± 0.15* 9.94 ± 3.72
R2846 (Strain 12)154.14 ± 44.400.85 ± 0.19 4.48 ± 0.2642.51 ± 20.08 1.67 ± 0.36
R3528 (R2846 dam::TSTE) 61.05 ± 14.80**0.19 ± 0.03**0.002 ± 0.001**23.55 ± 3.01* 0.63 ± 0.18*

Haemophilus influenzae dam mutants are not hypersusceptible to peroxide

To determine whether an increased susceptibility of dam mutants to intracellular peroxide may account for the decreased invasion phenotype, the susceptibilities of Rd KW20, INT-1 and Strain 12 parent and their dam::TSTE mutants were tested in a disc diffusion assay using hydrogen peroxide. Over the range of hydrogen peroxide concentrations tested (0–25%), no significant difference was observed in growth inhibition between parent and dam mutant strains at any peroxide concentration (data not shown). In addition, no significant difference was observed in growth inhibition between parent and its mutS mutant at any peroxide concentration tested, suggesting that mismatch repair was not required for the response to oxidative stress (data not shown).

Virulence gene expression

To gain insight into the decreased invasion phenotype, a limited survey of transcript expression was assessed for H. influenzae genes previously demonstrated to be involved in adherence and invasion (hmw1, hap, iga), intracellular survival (rpoE, htrA) or induced in vivo (purE) (St Geme et al., 1994; Vitovski et al., 1999; Pedersen et al., 2001; Craig et al., 2002; Mason et al., 2003). H. influenzae total RNA was isolated from Strain 12 (R2846) and Strain 12 dam::TSTE (R3528) from HBMECs, after invasion, using both extracellular and intracellular bacteria. Two-step, reverse transcription real-time polymerase chain reaction (PCR) was used to assess the relative expression of hmw1, hap, iga, rpoE, htrA and purE, with normalization to the expression of DNA gyrase (gyrA). Table 2 depicts the relative amount of each transcript in strain R3528 (dam) in comparison with the parent (strain R2846) for bacteria grown in contact with HBMECs and those harvested from within the eukaryotic cell. Although there is a tendency for the adhesins to be expressed when in contact with the HBMECs, the differences are not statistically significant. We could not assess transcripts of Strain 12 or the dam mutant in tissue culture media alone as, in the absence of eukaryotic cells, bacterial viability is lost in a time-dependent manner. The trend towards increased transcription of hmw1 and hap, when in contact with HBMECs, is consistent with the role of the products of these genes in adherence.

Table 2. Relative transcript expression of strain R3528 (dam) compared with strain R2846 in HBMEC assaysa.
TranscriptSupernatantIntracellular P b
  • a. 

    Mean and standard error of the mean ( ) are depicted.

  • b. 

    Two-tailed t-test.

  • Data are relative values normalized to gyrA expression and calculated using the 2–ΔΔCT relative quantification method.

hmw1 4.14 (1.80)0.93 (1.7)0.089
hap 4.50 (1.70)1.40 (0.36)0.143
iga 1.70 (0.82)3.26 (0.82)0.763
htrA 3.26 (1.02)1.57 (0.54)0.146
purE 0.83 (0.27)1.67 (0.46)0.242
rpoE 1.08 (0.44)1.96 (0.56)0.366

Haemophilus influenzae dam mutants are attenuated in the infant rat model

To determine whether the decreased invasiveness of the H. influenzae dam mutants noted with human cells cultured in vitro would extend to decreased virulence in vivo, we used the infant rat model of infection. Strains Rd KW20 (R652), INT-1 (R2866) and INT-1 dam::TSTE (R3549) were inoculated intraperitoneally into 5-day-old Sprague–Dawley rats, and bacteraemia was assessed after 48 h. Strains INT-1 and Rd KW20 were shown previously to be virulent and avirulent, respectively, in this model (Nizet et al., 1996). Figure 3 shows the results of a 1 × 105 cfu inoculum of bacteria per rat pup. Statistical comparison by a two-tailed Mann–Whitney test identified a significant difference in bacteraemia for rats inoculated with Rd KW20 (R652) (n = 10) compared with rats inoculated with INT-1 (R2866) (n = 9) (< 0.0001), and a significant difference for rats inoculated with INT-1 dam::TSTE (R3549) (n = 19) compared with rats inoculated with INT-1 (R2866) (n = 9) (< 0.005). Rats inoculated with INT-1 dam::TSTE (R3549) had an ≈ 2 log reduction in the geometric mean density in blood compared with rats inoculated with INT-1 (R2866). An additional experiment testing a 1 × 106 cfu inoculum per rat pup also identified a statistically significant difference in bacteraemia for rats inoculated with Rd KW20 (R652) (n = 10) compared with rats inoculated with INT-1 (R2866) (n = 10) (< 0.0001), and a significant difference for rats inoculated with INT-1 dam::TSTE (R3549) (n = 10) compared with rats inoculated with INT-1 (R2866) (n = 10) (< 0.005), with an ≈ 4 log reduction in the geometric mean density of bacteria in blood of INT-1 dam::TSTE (R3549) compared with INT-1 (R2866) (data not shown).

Figure 3.

Dam inactivation reduces virulence in the infant rat model of bacteraemia. Five-day-old Sprague–Dawley rat pups were inoculated intraperitoneally with 1 × 105 cfu in PBSg of Rd KW20 (R652, n = 10), INT-1 (R2866, n = 9) or INT-1 dam::TSTE (R3549, n = 19). After 48 h, pups were sacrificed, and aliquots of blood were plated on chocolate or sBHI agar for assessment of bacteraemia. Each point represents the number of H. influenzae (cfu ml−1 blood) for a single pup, with bars representing the geometric mean for each set. Animals without bacteraemia were grouped as 1 cfu ml−1 and are indicated on the graph by the number in parentheses. Statistical comparison by a two-tailed Mann–Whitney test identified a significant difference in bacteraemia production for R652 compared with R2866 (< 0.0001), and a significant difference for R3549 compared with R2866 (< 0.005).


Mutants of E. coli, S. typhimurium and S. marcescens lacking dam activity have a hypermutable phenotype (Glickman, 1979; Torreblanca and Casadesus, 1996; Ostendorf et al., 1999). This observation led to the conclusion that Dam activity directs DNA mismatch repair to discriminate properly between template and newly replicated DNA by hemimethylation at GATC sites (Glickman et al., 1978; Glickman, 1979; Glickman and Radman, 1980). The absence of a hypermutable phenotype in H. influenzae dam mutants found in this study and that reported by Bayliss et al. (2002) requires that the role of Dam in mismatch repair be re-examined. However, we found that dam mutants were hypersusceptible to the base analogue 2-AP and that this susceptibility was suppressed by either complementation in cis with a functional copy of dam or secondary mutations in mutS eliminating mismatch repair. This evidence supports a role for Dam in directing mismatch repair in H. influenzae as the presence of the hemimethylated GATC sequences was sufficient to restore the activity of the MutSLH complex. If Dam is a necessary component of H. influenzae mismatch repair, why are dam mutants not hypermutable? Dam mutants of other pathogens such as E. coli (Glickman, 1979) and S. marcescens (Ostendorf et al., 1999) can exhibit up to 30-fold increased mutation frequency, while mutations in other components of the mismatch repair system, including mutS, mutL or mutH, have even higher mutation frequencies, often 100- to 1000-fold greater than wild type (LeClerc et al., 1996). We found that mutS mutants of three H. influenzae strains had mutation frequencies ≈ 20- to 50-fold greater than wild type or dam mutants. Although this difference is significant, it is not as great as the corresponding increase in mutation frequency seen in E. coli; therefore, a corresponding smaller difference in the mutation frequency in H. influenzae dam mutants may be undetectable using our methods. As noted previously, the Rd KW20 genome contains multiple putative methylases, but only one copy of dam (HI0209). The finding that mutation of dam alone produced susceptibility to 2-AP is evidence that Dam is the only methylase functioning to direct mismatch repair and, therefore, the lack of an increased mutation frequency is not likely to result from mismatch repair directed by an alternative methylation process.

We also found that H. influenzae dam mutants are not hypersusceptible to either UV radiation or hydrogen peroxide, two DNA-damaging exposures to which E. coli dam mutants are susceptible (Glickman et al., 1978; Wyrzykowski and Volkert, 2003). Therefore, the response to these two DNA-damaging agents does not require Dam methylation in H. influenzae. It seems that the DNA repair pathways in H. influenzae are less complex than those in the E. coli model, perhaps as a reduction of ‘excess’ genomic content as H. influenzae adapted to its confined human environment where chemical and UV irradiation-induced DNA damage is not as common as in an environmental organism such as E. coli.

As in E. coli, HI0209 (dam) appears to have regulatory functions. H. influenzae dam mutants were less virulent in both in vitro and in vivo models of infection: this finding is similar to S. typhimurium dam mutants that had reduced ability to invade cultured non-phagocytic cells, and virulence was attenuated in the BALB/c mouse infection model (Garcia-Del Portillo et al., 1999). The defects in human cell invasion seemed to be multifactorial as different invasion steps were affected between the three H. influenzae strains. The Rd KW20 dam mutant (R3526) was attenuated in cell invasion, but not in adherence to HBMECs or intracellular replication. The INT-1 dam mutant (R3549) had decreased adherence and invasion to HBMECs, but not intracellular replication. The Strain 12 dam mutant (R3528) was attenuated for all steps including adherence to HBMECs, invasion and intracellular replication. Reduction of invasion at the point of cell entry was a phenotype of all three dam mutants and could be explained if a critical bacterial factor required for efficient host cell invasion was not expressed or accessible. Invasion of E. coli into HBMECs occurs through a host cell-mediated uptake mechanism involving the interaction of E. coli OmpA with the Ecgp host cell surface glycoprotein (Prasadarao, 2002). Invasion of H. influenzae into monolayers of NCI-H292 cells has been shown to be chiefly by a paracellular route (van Schilfgaarde et al., 1995) and, therefore, the gentamicin-resistant bacteria measured by this study in NCI-H292 cells may not have been truly intracellular, but shielded from the gentamicin by residing between the human cells. However, H. influenzae dam mutants had decreased cell entry into monolayers of both cell types. If Dam functions as a regulator of virulence gene expression, decreased adherence and invasion of INT-1 and Strain 12 dam mutants may be the result of decreased expression of one of several adhesins or invasins (see Introduction), but further studies are needed. Strain 12 (R2846) has 5140 potential Dam methylation sites, strain INT-1 (R2866) has 4907, and strain Rd KW20 (R652) 4974; in contrast, E. coli has 18 000 potential sites. Examination of the methylation of the promoters of putative H. influenzae virulence factors may provide insight into the regulatory roles of dam in this species. Decreased intracellular replication of the Strain 12 dam mutant was probably not caused by increased susceptibility to oxidative stress as the mutants were not hypersensitive to hydrogen peroxide.

A limited survey of gene expression found no significantly reduced relative expression of rpoE, htrA, iga, hap, hmw1 or purE, transcripts by the Strain 12 dam mutant (R3528) compared with the parent strain (R2846) replicating on or within eukaryotic cells. This suggests that decreased expression of these genes alone is not likely to be responsible for the reduced invasion phenotype. A larger survey of gene expression by microarray analysis will be beneficial in determining the exact molecular defect. H. influenzae grown on the surface of the HBMECs had increased expression of hmw1 and hap mRNA, consistent with their role in promoting adherence; thus, the reduction in adherence of the Strain 12 dam mutant (R3528) may result from post-transcriptional mRNA modification or increased degradation of the adhesins.

Reduction in the magnitude of the bacteraemia in the infant rat model with strain R3549 (INT-1 dam) was an unexpected finding, but one that is consistent with a report of reduced virulence of S. typhimurium dam mutants in the BALB/c mouse model (Garcia-Del Portillo et al., 1999). These data suggest that the mechanism of evasion of C3 used by strain INT-1 (R2866) is downregulated in the dam mutant. It is uncertain whether H. influenzae dam mutants could be used to provide cross-protective immunity against other invasive non-typeable strains for consideration as a vaccine candidate, as has been proposed for other bacterial pathogens.

Reduction of H. influenzae virulence both in vitro and in vivo was achieved by eliminating Dam methylation. In H. influenzae, Dam methylase may represent an ideal target as a potential vaccine candidate or for specific inhibitors to reduce the incidence of invasive disease.

Experimental procedures

Bacterial strains and growth conditions

The H. influenzae strains used in this study are described in Table 3. Strain R652 is Rd KW20 for which the complete sequence is available (Fleischmann et al., 1995), is an unencapsulated type d isolate and was provided by Dr H. O. Smith (Johns Hopkins University) (Wilcox and Smith, 1975). Strain R2846 is a non-typeable otitis media isolate also known as Strain 12 from which the hmw1 and hmw2 adhesin genes were originally cloned (Barenkamp and Leininger, 1992), and it was provided by Dr S. Barenkemp (St Louis University). Strain R2866 is the invasive non-typeable 1 (INT-1) strain, isolated from the blood of a child with meningitis (Nizet et al., 1996). H. influenzae strains were grown on chocolate agar supplemented with 1% isovitalex or in brain–heart infusion broth supplemented (sBHI) with haemin (10 µg ml−1) and β-NAD+ (10 µg ml−1) and agar for solid sBHI plates. Where indicated, experiments also used Coleman's chemically defined media for H. influenzae growth (Coleman et al., 2003). H. influenzae strains were incubated at 37°C in 5% CO2, which enhanced the growth of the clinical isolates. For Haemophilus strains, selective antibiotic concentrations used included ribostamycin (15 µg ml−1), tetracycline (5 µg ml−1) and chloramphenicol (2 µg ml−1). All chemicals were purchased from Sigma Chemical unless otherwise specified. E. coli DH5α was obtained from Gibco BRL and was made competent by the CaCl2 method and stored at −70°C before use (Sambrook et al., 1989). E. coli strains were incubated at 37°C in air in Luria–Bertani broth or agar with one or more of the following selective antibiotics: ampicillin (100 µg ml−1), chloramphenicol (40 µg ml−1), kanamycin (50 µg ml−1) and tetracycline (12 µg ml−1).

Table 3. Bacterial strains and plasmids used in this study.
Strain or plasmidRelevant genotype or descriptionSource or reference
H. influenzae strains
 R652Rd KW20, Unencapsulated, originally type d Wilcox and Smith (1975)
 R3526R652 dam::TSTE, ribostamycinRThis study
 R3544R652 mutS::TSTE, ribostamycinRThis study
 R3545R652 dam::Tet, tetracyclineRThis study
 R3546R652 dam::Tet; mutS::TSTEThis study
 R3547R3526 dam+, chloramphenicolRThis study
 R2846Strain 12, otitis media isolate Barenkamp and Leininger (1992)
 R3528R2846 dam::TSTE, ribostamycinRThis study
 R3548R2846 mutS::TSTE, ribostamycinRThis study
 R2866INT-1, blood isolate Nizet et al. (1996)
 R3549R2866 dam::TSTE, ribostamycinRThis study
 R3237R2866 mutS::TSTE, ribostamycinRThis study
 DB117Rd rec-1 Beattie and Setlow (1971)
E. coli strains
 DH5αStandard cloning strainGibco BRL
 pUC19AmpR, E. coli cloning vectorNew England Biolabs
 pUC19::TSTEAmpR, KanR, TSTE cassette inserted at BamHI Sharetzsky et al. (1991)
 pMW022pUC19 with dam::TSTEThis study
 pGJB103AmpR, TetR, E. coli–H. influenzae shuttle vector Tomb et al. (1989)
 pSU2718CmR, lacZα and polycloning site from pUC18 Chandler (1991)
 pMW053pSU2718 with dam::tetRThis study
 pACYC177AmpR, KmR, Low-copy E. coli cloning vectorNew England Biolabs
 pBR322AmpR, TetR, E. coli cloning vectorNew England Biolabs
 pMW033pBR322 with mutS::TSTEThis study
 pMW055HI1018 insertional cloning vector, CmR, lacZαThis study

DNA manipulations, cloning and sequencing

Bacterial genomic DNA was prepared using the DNeasy tissue kit from Qiagen, and plasmid DNA was prepared using the Qiaquick plasmid miniprep kit (Qiagen). All PCRs and restriction enzyme reactions were partially purified using the Qiaprep PCR purification kit (Qiagen). All restriction and modifying enzymes were purchased from New England Biolabs. PCR was performed using the High-fidelity polymerase blend (Roche) or the Fail-safe amplification system (Epicentre). Oligonucleotide primers were purchased from Integrated DNA Technologies. Southern blotting to Nytran nylon membranes was performed with the Turboblotter kit (Midwest Scientificl) with probes labelled using the DIG DNA labelling kit (Roche) and chemiluminescent detection according to the manufacturer's recommendations. DNA sequencing was performed by the DNA core facility at the Seattle Biomedical Research Institute.

Construction of H. influenzae mutants

The dam gene HI0209 appears to be the final gene of a three ORF operon starting with HI0207 shikimic acid kinase (aroK), HI0208 3-dehydroquinate synthase (aroB) and then dam at HI0209. The starting ATG for dam is located 1.6 kb downstream of the putative promoter region for the operon. A 2044 bp PCR fragment containing dam methylase (HI0209) was amplified from R652 genomic DNA using primers Dam F (5′-cgggatcccgtccaacaacattgctttcac-3′; Rd KW20 co-ordinates 223139–223158) and Dam R (5′-cgggatcccgtgag cacaaatagggcagtg-3′; Rd KW20 co-ordinates 225183–225164). The 5′ underlined nucleotides incorporate BamHI sites into the PCR product. This PCR product was spin-column purified and digested with BamHI and then ligated with BamHI-digested pUC19 using T4 DNA ligase. The resulting ≈ 4.7 kb plasmid was digested with EcoRV at residue 199 of 861 within the dam ORF, and this was ligated with the ≈ 2.4 kb BamHI TSTE fragment blunted with mung bean nuclease. The TSTE cassette contains an aminoglycoside phosphotransferase (aph) providing kanamycin resistance in E. coli or ribostamycin resistance in Haemophilus and is flanked by Haemophilus-specific uptake signal sequences to aid DNA uptake (Sharetzsky et al., 1991). The resulting ≈ 7.1 kb plasmid was identified as pMW022 and was linearized by digestion with KpnI and used to transform H. influenzae made competent by the M-IV method (Steinhart and Herriott, 1968). Dilutions of transformed cells were plated on chocolate agar plates containing ribostamycin and incubated at 37°C for 48 h in air before picking colonies. This protocol was followed to make dam::TSTE mutants of Rd KW20 (R3526), Strain 12 (R3528) and INT-1 (R3549). An additional construct was produced to inactivate dam with the tetracycline resistance cassette from pGJB103 (Tomb et al., 1989). The PCR product containing dam was digested with BamHI and ligated into BamHI-digested pSU2718 (Chandler, 1991). This construct was linearized with EcoRV and ligated with a blunted ≈ 2.7 kb HindIII–XbaI fragment from pGJB103 containing the tetracycline resistance cassette. The resulting plasmid containing dam::TetR was called pMW053 and was linearized with PstI for transformation into H. influenzae strains with selection on chocolate agar plates with tetracycline. This protocol was followed to make dam::TetR mutants of Rd KW20 (R3545) and an Rd KW20 double mutant with mutS::TSTE and dam::TetR (R3546).

Transformants were screened for insertion of the cassette into the genomic dam locus and the loss of functional dam methylase activity. Colony PCR and Southern blotting were used to identify and confirm transformants that had an insertion into the dam locus and that the mutant dam was in single copy within the genome. Functional dam methylase activity was assessed by digesting 10 µg of genomic DNA from each of the mutant and parent strains with 10 units of Sau3AI, DpnI or DpnII restriction endonucleases, which cut GATC DNA sequences differently based on methylation status: Sau3A1 cuts all GATC sequences regardless of methylation; DpnI only cuts methylated GATC sequences; and DpnII only cuts non-methylated GATC sequences. After 4 h of incubation at 37°C, the digestion reactions were visualized by agarose gel electrophoresis.

To complement the dam mutation in the Rd KW20 dam::TSTE mutant (R3526), the putative upstream promoter region was PCR amplified and fused to the ATG start codon of dam, and this construct was inserted into a pACYC177-derived vector flanked by the HI1018/IS1016 (Daines and Smith, 2004) sequence to produce the plasmid pMW056. The HI1018 locus in H. influenzae Rd KW20 (R652) is a remnant of the lost encapsulation locus and was chosen as a site likely to accept transformed DNA without causing polar effects on surrounding genes. Complementation with one copy of dam expressed ectopically in cis was chosen rather than expression in trans on a multicopy plasmid to reduce any confounding effect that overexpression of Dam methylase may produce on any phenotype, as has been reported with E. coli (Herman and Modrich, 1981). The promoter of the operon containing dam was PCR amplified using the primers Dam Promoter F (5′-ctagctagctagctaaaaatagcgcgtatttaacg-3′, NheI site is underlined; Rd co-ordinates 221931–221952) and Dam Promoter R (5′-gttttttcggacgtaacattgttctttatct-3′; Rd co-ordinates 222202–222188). The dam ORF was PCR amplified using the primers ATG Dam F (5′-atgttacgtccgaa aaaacaatctttaaaacc-3′; Rd co-ordinates 223852–223883) and Dam NheI R (5′-ctagctagctagtgagcacaaatagggcagtg-3′, NheI site is underlined; Rd co-ordinates 225180–225162). The primers were designed to make the 3′ end of the promoter fragment complementary to the 5′ end of the ORF fragment. These PCR products were purified and linked together by PCR extension with the primers Dam Promoter F and Dam NheI R, and the resulting fragment was digested with NheI and cloned into the plasmid pMW055 that contains the HI1018 sequence interrupted with the chloramphenicol acetyltransferase cassette and the lacZα multiple cloning sites from pSU2718 on a pACYC177-derived backbone. This plasmid, pMW056, was linearized with BseRI and transformed into strain R3526, and transformants were selected for chloramphenicol resistance. Transformants were verified by PCR for insertion into the HI1018 locus: this strain was identified as R3547. Resistant colonies were tested for dam methylase activity by isolating genomic DNA and subjecting it to restriction digestion with methylation-sensitive enzymes as described above. Complementation was attempted for the dam mutants of Strain 12 (R3528) and INT-1 (R3549) but was unsuccessful, perhaps due to absence or modified HI1018 locus in these strains.

PCR primers XbaI MutS F (5′-gctctagagctataaaggtaaatca gtggatc-3′; Rd KW20 co-ordinates 750979–751002) and XbaI MutS R (5′-gctctagagccaatccatataaacctttg-3′; Rd KW20 co-ordinates 753790–753770) were used to amplify an ≈ 2.8 kb fragment containing mutS (HI0707). The underlined nucleotides in each primer incorporate XbaI sites into the PCR product. This PCR product was cleaned and ligated into pBR322 previously digested with NheI to produce the ≈ 7.2 kb plasmid pMW032. This plasmid was then digested with NheI and blunted to open the plasmid within the mutS ORF. The blunt plasmid was ligated with the blunt ≈ 2.4 kb BamHI TSTE fragment to produce the ≈ 9.6 kb plasmid pMW033. This plasmid was linearized by digestion with PvuII and used to transform competent H. influenzae to produce mutS mutants of Rd KW20 (R3544) and Strain 12 (R3548). The INT-1 mutS::TSTE mutant (R3237) was produced similarly by B. Williams (unpublished). Insertions interrupting mutS were confirmed by Southern blotting, and the hypermutable phenotype was confirmed as described below.

Growth and viability of dam mutants

Growth was assessed by measuring OD600 versus time for mutant and parent strains performed in sBHI broth and Coleman's chemically defined media (Coleman et al., 2003). Strains were subcultured from overnight growth on chocolate agar. Each strain was inoculated at a density of 107 cfu ml−1 into 12.5 ml of media in a 125 ml flask and grown at 37°C with shaking at 200 r.p.m. in air. Absorbance at 600 nm was measured on aliquots in plastic cuvettes on a Hitachi model U-2000 spectrophotometer. Cell morphology was assessed microscopically by Gram stain with visualization with a Nikon Eclipse E600 microscope at 100× magnification.

Determination of spontaneous mutation frequency

Mutation frequency to rifampicin resistance (25 µg ml−1) and nalidixic acid resistance (14 µg ml−1) was assessed on sBHI agar plates. Fluctuation assays (Luria and Delbruck, 1943) were performed by inoculating 10 ml of sBHI broth with 107 cfu ml−1H. influenzae and growing for 8–10 h shaking at 200 r.p.m. in 37°C in room air. These cultures were then pelleted and resuspended in 500 µl of phosphate-buffered saline (pH 7.0) with 0.1% gelatin (PBSg). From these concentrated cultures, 100 µl was spread in duplicate on to sBHI plates with rifampicin or nalidixic acid to determine the number of drug-resistant mutants ml−1 culture. In addition, another 100 µl of the same culture was serially diluted in PBSg and plated in duplicate on antibiotic-free sBHI plates to determine the total number of viable cells present ml−1 concentrated culture: the plates were incubated for 48 h at 37°C before colony counting. Mutation frequency was calculated as the number of drug-resistant cells ml−1 culture divided by the total viable count ml−1 culture. Because the mutation frequency fluctuates between individual cultures, at least eight independent measurements were performed for each mutant and parent strain.

Treatment of cells with 2-AP, UV irradiation or hydrogen peroxide

Experiments were performed with colonies grown overnight on chocolate agar plates at 37°C in room air. Each experiment was replicated three or more times.

2-AP.  For experiments with 2-AP, we found it necessary to use Coleman's defined media as opposed to sBHI media. Coleman's media provided consistent results from batch to batch, which was not achievable with sBHI, presumably because of BHI's variable lot-to-lot composition. 2-AP was dissolved in Coleman's media, and growth curves were performed in both the presence or the absence of the drug. Spontaneous mutation frequency after growth in 2-AP was tested by growing H. influenzae cultures in the presence of 0, 50 or 100 µg ml−1 2-AP in Coleman's media for 8 h and then harvesting 5 ml of cells by centrifugation and washing twice in PBSg before serial dilution and plating on sBHI plates containing no antibiotic, rifampicin (25 µg ml−1) or nalidixic acid (14 µg ml−1).

UV irradiation.  Bacteria were resuspended to an OD600 of 0.2 in PBSg. Serial dilutions of this suspension were plated onto sBHI agar plates and allowed to dry for 30 min. The plates were then exposed to UV light by introducing them into a Fisher Biotech UV cross-linker (FB-UVXL-1000), removing the plate lid and briefly pulsing with UV light at the following energies: 0, 400, 800, 1200, 1600, 2000, 2400 and 2800 µJ cm−2. After the UV pulse, the plates were covered with aluminium foil and incubated in the dark at 37°C for 36 h in room air. Survival was plotted as the viable count versus the energy of the UV dose. Mutation frequency after UV exposure was tested by growing H. influenzae cultures in 5 ml of sBHI broth to an OD600 of 0.2 and then transferring the total culture to a sterile 100 mm plastic Petri dish, which formed a thin layer of culture on the bottom. The plate, without the lid, was introduced into the UV cross-linker and irradiated as described above. After the UV exposure, the cultures were transferred back into their original 50 ml conical growth tubes and covered with aluminium foil to keep them in the dark for an additional 6 h of outgrowth before serial dilution and plating on either plain sBHI or plates containing rifampicin or nalidixic acid.

H 2 O 2 . Bacteria were resuspended to an OD600 of 0.2 in PBSg. Of this suspension, 100 µl (107 cfu) was spread on the surface of a plain sBHI agar plate and allowed to dry for 30 min. An aliquot of 30% H2O2 (Sigma) was diluted with sterile ddH2O to 25%, 10% and 5% concentrations. A sample of 10 µl was allowed to absorb on to sterile 6 mm filter paper discs, and these discs were then applied to the surface of the sBHI plates spread with bacteria. Plates were incubated for 24 h at 37°C in room air before reading. Measurements of the maximum diameter (mm) of growth inhibition surrounding each disc were taken as an index of peroxide susceptibility. Alternatively, log phase H. influenzae in sBHI broth were centrifuged, washed twice in PBSg and resuspended in sBHI broth containing 0.1%, 0.5%, 1.0%, 5.0% or 10% H2O2, or an equal concentration of water as a control, and incubated at 37°C for 30 min. Survival of H. influenzae was compared by viable counts of matched pairs in sBHI/H2O2 versus sBHI/water.

Adherence to and invasion of cultured human cell lines

Human brain microvascular endothelial cells (HBMECs) were a gift from Dr Kwang-Sik Kim (Johns Hopkins University School of Medicine) (Prasadarao, 2002) and were maintained in HBMEC media [76% RPMI 1640 supplied with 25 mM Hepes and 2 mM l-glutamine, 10% heat-inactivated fetal calf serum, 10% heat-inactivated NuSerum V (Becton Dickinson) and the following added to the indicated final concentrations: 2 mM l-glutamine, 1× MEM non-essential amino acid solution, 1× MEM vitamin solution and 1 mM MEM sodium pyruvate solution]. NCI-H292 human respiratory epithelial cells (pulmonary mucoepidermoid carcinoma cells ATCC CRL-1848) were grown in H292 media (RPMI 1640 with 25 mM Hepes and 2 mM l-glutamine, with 10% fetal bovine serum and the following added to the indicated final concentrations: 1 mM MEM sodium pyruvate solution, 20 mM sodium bicarbonate and 25 mM glucose). Monolayers of each cell line were grown in 1.0 ml of media on collagen-1-coated 12-well plates (BioCoat; Becton Dickinson) at 37°C with 5% CO2 to ≈ 90% confluency before use. The invasion protocol described here was adapted and modified from that described by Daines et al. (2003). H. influenzae bacteria from a fresh chocolate agar plate incubated overnight at 37°C in room air were scraped and resuspended in PBSg to an OD600 of 0.2 (≈ 1.0 × 108 cfu ml−1). Aliquots of this suspension were used to inoculate each well with ≈ 1.5 × 106 cfu of bacteria, a multiplicity of infection (MOI) of ≈ 10:1. Aliquots of the bacterial suspension were also serially diluted in PBSg and plated on plain sBHI plates to define the inoculum. The 12-well plates were incubated at 37°C in 5% CO2 for 4 h. After incubation, the monolayers were washed twice with 1.0 ml of Dulbecco's phosphate-buffered saline (D-PBS), and half the wells for each strain were harvested for total cell-associated counts, and the remaining wells were given 1.5 ml of fresh medium containing 100 µg ml−1 gentamicin. After 1 h at 37°C, the gentamicin-containing media were removed, and the wells were washed three times with 2.0 ml of D-PBS without magnesium or calcium to facilitate subsequent detachment of the monolayers. To the monolayers, 1.0 ml of 1% saponin in PBS without magnesium or calcium was added and allowed to incubate for 10 min before scraping the cells out of the wells and vortexing vigorously to lyse the eukaryotic cells and release the invaded bacteria. Before the adherence and invasion assays, all strains were tested for susceptibility to 1% saponin and gentamicin; no differences in percentage survival after 1 h in 1% saponin were noted (data not shown), and all strains had a gentamicin MIC of 1.0 µg ml−1. The resulting supernatant containing released bacteria was serially diluted in PBSg, and aliquots were spread on plain sBHI plates for determination of the gentamicin-resistant, invaded bacteria. Invasion was calculated as the percentage of gentamicin-resistant cells of the initial inoculum. Each experiment was performed in duplicate at least three times. For certain experiments, half the wells were harvested for invaded cells after the 4 h invasion and 1 h gentamicin kill while the remaining wells were incubated overnight after gentamicin removal and addition of fresh media to determine the viable intracellular bacteria density and the ratio of viable intracellular bacteria at 24 h versus 5 h.

Infant rat model of bacteraemia assays

The procedure was adapted from Smith et al. (1973). All animal protocols were submitted to and approved by the Institutional Animal Care and Use Committee (IACUC) of the Seattle Biomedical Research Institute. All personnel involved with laboratory animal experiments received prior training in laboratory animal policies and procedures administered by the University of Washington-Seattle IACUC.

Animals were obtained from the Charles River Laboratories. Pregnant rats usually delivered 7 days after receipt with an average litter size of 8–12 pups: 24 h after delivery, all pups were collected, randomized and redistributed with 10 pups per mother. At day 5 after birth, each neonatal rat was inoculated intraperitoneally with 0.1 cc of H. influenzae resuspended in PBSg to a concentration depending on target inoculum. After inoculation, pups were immediately returned to cages for nursing. Pups were checked visually at 24 h for signs of infection (i.e. shivering, hypoactivity, decreased feeding, grey skin colour or death) and, at 48 h after inoculation, pups were sacrificed by decapitation, and blood cultures were obtained to estimate the magnitude of the bacteraemia. Blood (100 µl) was plated directly on to a chocolate agar plate, and an additional 10 µl of blood was diluted 1:100 in 990 µl of PBSg, of which 100 µl was plated on to a second chocolate or sBHI agar plate for a 10−3 dilution. All rats not used in experiments were euthanized by inhalation of halothane (Halocarbon Laboratories) in a bell jar, with subsequent cervical dislocation before disposal.

Two-step reverse transcription PCR and real-time sequence detection assays

For quantification of gene expression by intracellular H. influenzae, bacterial RNA from human cell cultures was isolated after 4 h of infection as in a standard invasion assay described previously with the modification that human cells were grown to confluency in a collagen-coated flask (T-75) instead of a 12-well plate. Human cells were lysed by addition of 10 ml of 1% saponin in PBS without divalent cations, and the released intracellular bacteria were collected by centrifugation at 3000 r.p.m. for 15 min. The resulting cell pellet was saved and immediately resuspended in 300 µl of RNA Later (Ambion) RNA protection solution to prevent RNA degradation. RNA was isolated using the SV Total RNA system (Promega) and was eluted into 200 µl of RNase-free water. This dilute RNA was further concentrated to 14 µl by the RNeasy MinElute Cleanup kit (Qiagen), which resulted in a final RNA concentration of 2–5 µg µl−1. RNA quality was assessed by 1.0% agarose gel electrophoresis using RNA sample loading buffer with ethidium bromide (Sigma).

Initial first-strand synthesis of cDNA was performed using the MasterAmp RT-PCR kit (Epicentre) with a modified RT-PCR protocol of first-strand extension at 60°C for 20 min, followed by a heat inactivation step at 95°C for 6 min. In each 50 µl cDNA reaction, 7 µl of concentrated bacterial RNA was reverse-transcribed using the gene-specific reverse primers to be used for subsequent quantitative PCR. Real-time quantitative PCR was performed with the Applied Biosystems sequence detection system 7000 using the Quantitect SYBR Green PCR kit (Qiagen) for detection. For each 25 µl real-time PCR, 1 µl of cDNA product diluted 1:50 in water was mixed with 1 µl (22.5 pmol) of each forward and reverse primer, 9.5 µl of water and 12.5 µl of 2× SYBR Green PCR Mastermix. Each individual gene for both parent and mutant strains was measured in triplicate. Dissociation curves were performed after each real-time amplification to confirm that each reaction produced a single peak corresponding to a specific PCR product. The cDNA for DNA gyrase (gyrA) was used as a ‘housekeeping’ normalization control and was previously determined to be constitutively expressed under various growth conditions (Mason et al., 2003). Relative gene expression was determined between mutant and parent strains using the 2ΔΔCT relative quantification method (Livak and Schmittgen, 2001). Before using this method, all primers were tested to ensure the specificity of amplification from DNA, and a validation experiment was performed to demonstrate equal efficiencies of target and reference (gyrA) gene amplifications. Primer sequences for real-time PCR were as follows: gyrA F (5′-gcttctgctcgtcctgatgagtta-3′; Rd KW20 co-ordinates 1343083–1343060), gyrA R (5′-cttcttcaaacgcaat gcccgt-3′; Rd KW20 co-ordinates 1342951–1342972), rpoE F (5′-cgcgtaacgatattcctgatgtgg-3′; Rd KW20 co-ordinates 667949–667972), rpoE R (5′-ggcgaccttgagccgttaaata-3′; Rd KW20 co-ordinates 668077–668098), purE F (5′-ggctcatcg tacgcctgataaact-3′; Rd KW20 co-ordinates 1683084–1683107), purE R (5′-gctgcgatcataccgggtaaatgt-3′; Rd KW20 co-ordinates 1683200–1683177), htrA F (5′-tcgttcaacaggttct gacagtgg-3′; Rd KW20 co-ordinates 1336471–1336448), htrA R (5′-caccgcttggagaaataattgcgg-3′; Rd KW20 co-ordinates 1336326–1336349), hap F (5′-tctatttcggctttcgtggtg gtc-3′; not in Rd KW20), hap R (5′-agccacttgagtggtattatggt tcac-3′; not in Rd KW20), iga F (5′-tgttagccagcctcaagagacttc-3′; Strain 12 sequence), iga R (5′-ttatgtggagccgagctaacatcc-3′; Strain 12 sequence), hmw1a F (5′-gttgacgccaaagagtggt tgttagac-3′; not in Road KW20) and hmw1a R (5′-ctattcccg gatcccgtgtattcatc-3′; not in Rd KW20).

Methods for statistical analysis

Statistical analyses were performed using the statistical analysis functions of Microsoft excel or by the VassarStats website for statistical computation by Richard Lowry ( For most comparisons of data with a normal distribution, the Student's t-test was used, and P-values of < 0.05 were considered to indicate statistically significant differences. Where appropriate, the geometric mean rather than the arithmetic mean was reported for analysis of non-uniform distributions. For adherence and invasion assays, the Wilcoxon matched-pairs signed rank test was used to compare matched pairs (parent and mutant) of invasion results between different experiments with non-uniform differences. For infant rat assays, the Mann–Whitney test was used to compare bacteraemia results between groups. P-values of < 0.05 were considered to indicate statistically significant differences.


This work was supported by the National Institutes of Health grant AI44002.