Amino Acid Substitution in the Major Multidrug Efflux Transporter Protein AcrB Contributes to Low Susceptibility to Azithromycin in Haemophilus influenzae

ABSTRACT

Clarithromycin-resistant Haemophilus influenzae strains with a nonsense mutation in acrR generally exhibited susceptibility to azithromycin, although one strain was found to be nonsusceptible; we aimed to clarify the differences. This strain had an amino acid substitution, Arg327Ser, in AcrB. Introduction of this substitution into H. influenzae Rd caused an increase in the MIC of azithromycin, suggesting that this substitution contributed to nonsusceptibility. These findings indicate that azithromycin-nonsusceptible isolates could occur through stepwise mutation in the acr region.

TEXT

Macrolide-, β-lactam-, and fluoroquinolone-resistant Haemophilus influenzae strains have been isolated worldwide (1–4). In particular, clarithromycin (CLR)-resistant strains have been increasing in Japan (1). We reported previously that these strains harbored a nonsense mutation in the negative regulator gene acrR that resulted in overexpression of the chromosomal efflux pump AcrAB, which likely contributed to the CLR resistance (MICs of ≥32 μg/ml) (5). Although all of the isolates with a nonsense mutation in acrR showed resistance to CLR (MICs of ≥32 μg/ml), almost all of the isolates showed susceptibility to azithromycin (AZM) (MICs of ≤4 μg/ml) and the AZM MICs were even slightly reduced, indicating that this mutation could not be involved in resistance to AZM. However, one of the CLR-resistant isolates with the acrR mutation showed nonsusceptibility to AZM (5). Therefore, the aim of the present study was to clarify the difference between AZM-nonsusceptible and -susceptible strains.

For this study, we used the AZM-susceptible strain 2011-70 (CLR MIC of 32 μg/ml and AZM MIC of 4 μg/ml) and the AZM-nonsusceptible strain 2012-42 (CLR MIC of 32 μg/ml and AZM MIC of 8 μg/ml), which were isolated at Hachioji Medical Center of Tokyo Medical University and have mutations at the same acrR site (Table 1) (5). These two strains have the same genetic background of sequence type 156 (ST156), as determined by multilocus sequence typing (1). H. influenzae ATCC 49247 and H. influenzae Rd were used as control strains in the antimicrobial susceptibility tests, and H. influenzae Rd was also used as a recipient for transformation.

TABLE 1.

Characteristics and MICs of bacterial strains used in this study

Strain	Serotype	ST	Deletion site in acrR	MIC (μg/ml)a
				CLR	CLR + CCCP		AZM	AZM + CCCP		EB
					0.75 μg/ml CCCP	1.5 μg/ml CCCP		0.75 μg/ml CCCP	1.5 μg/ml CCCP
2012-42	NTb	156	Nucleotide 141c	32	16	4	8	4	1	≥16
2011-70	NT	156	Nucleotide 141c	32	16	1	4	1	0.25	8
Rd	NT	47	Nonec	8	4	0.25	1	0.5	≤0.06	1
Rd42acr	NT	47	Nucleotide 141c	32	16	2	8	4	0.5	8
Rd70acr	NT	47	Nucleotide 141	32	16	1	4	2	0.13	4

^aCLR, clarithromycin; AZM, azithromycin; EB, ethidium bromide; CCCP, carbonyl cyanide m-chlorophenylhydrazone.

^bNT, nontypeable.

^cResults of a previous study by Seyama et al. (5).

Antimicrobial susceptibilities were measured in accordance with Clinical and Laboratory Standards Institute (CLSI) guidelines (6, 7). AZM (LKT Laboratories, St. Paul, MN, USA), CLR (Tokyo Chemical Industry Co., Tokyo, Japan), and ethidium bromide (EB) (Wako Pure Chemical Industries, Osaka, Japan) were used as the antimicrobial agents. Carbonyl cyanide m-chlorophenylhydrazone (CCCP) (Wako Pure Chemical Industries) was used as the efflux pump inhibitor. The antimicrobial susceptibility breakpoints were set to ≤4 μg/ml for AZM and ≤8 μg/ml for CLR, according to the CLSI guidelines. In addition, to detect slight changes in susceptibility, we tested a more detailed range of concentrations than recommended by the CLSI, which were established through serial dilutions from 2 μg/ml for CLR and 1 μg/ml for AZM. These experiments were performed four times on independent occasions.

The transcription levels of the chromosomal multidrug efflux pump acrAB were determined by semiquantitative reverse transcription (RT)-PCR, as described previously (5). The acr operons of isolates 2011-70 and 2012-42 were analyzed according to the sequencing results obtained in the previous study (5).

The efflux pump activity was determined based on the intercalation fluorescence of EB, which is a substrate of AcrAB, according to the method described by Hayashi and colleagues (8, 9), with minor modifications. The isolates were treated with EB, CLR, and AZM at a final concentration of 10 μM. The fluorescence was measured every 5 min with a Varioskan Flash 2.4 microplate reader (Thermo Fisher Scientific, Kanagawa, Japan), with excitation at 530 nm and emission at 630 nm. The experiment was performed at least four times on independent occasions.

Comparison of acr operon regions, from 100 bases upstream of acrR to 30 bases downstream acrB (5,021bp), between the strains showed that 2012-42 had a specific amino acid substitution in AcrB, in which arginine was replaced by serine at position 327 (Arg327Ser) (see Fig. S1 in the supplemental material). No other CLR-resistant and AZM-susceptible strains (n = 6) tested had this substitution (data not shown) (5).

To investigate whether this amino acid substitution contributed to AZM nonsusceptibility, the acr operon regions, from 134 bases upstream of acrR to 258 bases downstream of acrB (5,282 bp), of strains 2012-42 and 2011-70 were amplified by PCR and introduced into H. influenzae Rd by electroporation, as described previously (5). Transformants (Rd_42acr and Rd_70acr) were selected by plating on chocolate agar containing 16 μg/ml CLR. The acr operon regions, from 100 bases upstream of acrR to 30 bases downstream acrB (5,021 bp), in the transformants were confirmed by DNA sequencing, which showed that the acr operons of the transformants obtained were identical to those of the donor strains (strains 2012-42 and 2011-70). The MICs of CLR and AZM for the transformants increased to the same levels as observed for the donor strains (Table 1). In addition, the acrB transcriptional levels of the transformants increased significantly, compared to that of Rd (Table 2).

TABLE 2.

Detailed MICs of H. influenzae strain Rd and strains with acr mutations

Strain	Relative acrB mRNA levela	MIC (μg/ml)		Ratiob
		CLR	AZM	CLR	AZM
Rd	0.82 ± 0.24	8	1	1	1
Rd42acr	1.86 ± 0.35c	30	8	3.8	8
Rd70acr	1.45 ± 0.09c	26	3.5	3.3	3.5

^aThe transcription levels were calculated in comparison with the gyrB transcription level. The measurements of acrB transcription levels were performed at least three times on independent occasions.

^bRatio of each MIC to that of H. influenzae Rd.

^cP < 0.05, Rd versus Rd_42acr and Rd versus Rd_70acr.

The antimicrobial susceptibility test recommended by the CLSI was designed as a serial 2-fold dilution method; this method is concise and highly reproducible but would not detect subtler changes. Therefore, the detailed antimicrobial susceptibilities of the transformants were tested (Table 2). The detailed MICs of CLR and AZM in Rd_42acr were 1.2 and 2.3 times higher, respectively, than those in Rd_70acr.

We hypothesized that this amino acid substitution would alter the ability of the efflux pump AcrAB to recognize the substrate. To test this hypothesis, the macrolide efflux activities were measured with an EB efflux competition assay (Fig. S2), according to the gradient of the curve of the graph indicating the change in fluorescence over time. After AZM addition, the initial velocity at 5 to 15 min, which might reflect efflux speed, was slightly but significantly larger in Rd_42acr than in Rd_70acr, by two-tailed Student's t test. In contrast, no significant differences between the strains were detected following EB or CLR addition (Fig. S2).

Overall, a nonsense mutation of acrR in H. influenzae appears to confer resistance to CLR but not to AZM. The present study further demonstrates that a subsequent missense mutation in acrB could result in AZM nonsusceptibility. Although the mechanism by which this substitution confers nonsusceptibility to AZM remains unclear, the results of the EB efflux assay showed that AZM addition competed with EB efflux at the early stage, suggesting that this amino acid substitution contributed to alterations in initial recognition and efflux. AcrB is known to form a trimer that acts as a complex with the outer membrane channel TolC and the membrane fusion protein AcrA (10–12). This efflux system discharges the agents from both the inner membrane and the periplasm (10). Several studies with efflux pump inhibitors demonstrated that the sensitivity of the inhibitors was determined by slight steric effects involving the amino acids around the inhibitor-binding pit (13, 14). These findings suggest that the amino acid substitution in the efflux pump influences the activity of the efflux pump. Therefore, we attempted to analyze the precise location of the amino acid 327 substitution in AcrB. The transmembrane protein structure prediction software CCTOP (http://cctop.enzim.ttk.mta.hu), which is a transmembrane region prediction tool (15), predicted that amino acid 327 of AcrB was located outside the outer membrane (data not shown). Amino acid 327 of AcrB in Escherichia coli was also reported to lie outside the outer membrane (16). This configuration implies that the amino acid could be involved in substrate recognition.

AZM shows relatively stronger antibacterial activity for H. influenzae, with a superior tissue distribution effect, compared to conventional macrolides (17–21). Furthermore, AZM was reported to be effective even when administered over a short period with a small dose, owing to its extended half-life (22, 23). Therefore, the possibility of increases in AZM-nonsusceptible strains could represent a crucial clinical problem, because of the loss of therapeutic options.

Our findings strongly suggest that the amino acid 327 substitution of AcrB contributes to AZM nonsusceptibility. Moreover, it appears that AZM-nonsusceptible isolates have emerged from stepwise mutations in the acr region; H. influenzae acquired a nonsense mutation in acrR as the first step, followed by a missense mutation in acrB as the second step. As we reported previously, CLR-resistant strains with a nonsense mutation in acrR are already prevalent in Japan. Our present study indicates that these strains could easily become AZM nonsusceptible through a single mutation in AcrB.

Accession number(s).

The analyzed nucleotide sequences were submitted to GenBank under accession numbers LC269307 (strain 2012-42) and LC269308 (strain 2011-70).