Introduction

The global vaccination programme against Haemophilus influenzae type b (Hib) has substantially affected the epidemiology of Hib infections in most countries. Compelling observations provide concrete evidence for the thriving of non-typeable H. influenzae (NTHi) in the post-Hib vaccination era. NTHi is a major cause of otitis media, bacterial sinusitis, and bacterial conjunctivitis in children; moreover, it exacerbates chronic obstructive pulmonary disease and causes up to 5% of neonatal invasive bacterial infections [1]. National surveillance data from 14 European and other countries indicate that NTHi causes 97% of non-Hib infections in invasive H. influenzae disease [2]. The highest incidence of invasive NTHi diseases occurs in premature infants, elderly adults, the immunosuppressed, individuals with malignant disease, and those with chronic cardiovascular or respiratory conditions. The fatality rate is higher than 10% among such individuals [3, 4].

Empirical treatment for NTHi infection is akin to that for typeable H. influenzae infection. Ampicillin is the standard recipe for β-lactamase-negative ampicillin-susceptible NTHi infections, with chloramphenicol or broad-spectrum cephalosporins used as alternate regimens. In both typeable H. influenzae and NTHi, β-lactamase production, particularly by blaTEM-1, is the predominant resistance mechanism against penicillins and some cephalosporins. Studies have reported rates of β-lactamase-positive ampicillin-resistant (BLPAR) NTHi between 10 and 25% in South Africa, Europe, and the Americas [5,6,7,8]. In Taiwan, Vietnam, Japan, and South Korea, BLPAR may constitute up to 55% of NTHi [9,10,11,12]. The prevalence of β-lactamase-negative ampicillin-resistant (BLNAR) NTHi is also high in Taiwan, Vietnam, and South Korea [9, 11, 12]. Macrolides, quinolones, tetracycline, trimethoprim-sulfamethoxazole, and third-generation cephalosporins may provide coverage for ampicillin-resistant H. influenzae infection.

Multidrug-resistant (MDR) NTHi was infrequent in Western countries but has been sporadically reported in Asian areas [13,14,15]. Reported MDR NTHi strains were resistant to β-lactams, macrolides, quinolones, tetracycline, and trimethoprim-sulfamethoxazole, suggesting that NTHi is acquiring drug resistance through different ploys. Because the molecular epidemiology of MDR in NTHi remains unclear, we conducted whole genome and plasmid analyses to delineate the genes and genetic variations relevant to MDR NTHi and high-level drug resistance.

Results

Antibiogram of NTHi

Twenty-six NTHi isolates were included in this study. An analysis of β-lactam resistance revealed that 12 (46.2%) isolates were non-susceptible to ampicillin, 4 (15.4%) to amoxicillin-clavulanate, and 8 (30.8%) to cefuroxime (Fig. 1). For non-β-lactam resistance, 20 (76.9%) isolates were non-susceptible to trimethoprim-sulfamethoxazole, 1 (3.8%) to azithromycin, 15 (57.7%) to levofloxacin and ciprofloxacin, and 5 (19.2%) to tetracycline. All isolates were susceptible to cefotaxime, ertapenem, meropenem, and tigecycline. Regarding the antibiogram, 5 isolates (No. 01, 11, 12, 14, and 20) were susceptible to all drugs, whereas 8 were non-susceptible to 2 categories of agents (Table 1). Among 13 MDR isolates, 4 (No. 05, 08, 13, and 18) were non-susceptible to 3 categories of agents, 8 (No. 04, 06, 07, 09, 10, 15, 19, and 23) to 4 categories, and 1 (No. 22) to 5 categories.

Fig. 1
figure 1

MIC and susceptibility profile of non-typeable Haemophilus influenzae (NTHi). Percentage of MICs of different antimicrobial agents in 26 NTHi isolates are shown in a circle plot. Bold arc represents the drug-susceptible range according to the Clinical & Laboratory Standards Institute M100 31st edition. MICs of ertapenem, meropenem, and tigecycline in all isolates were ≤ 0.25 μg/mL. AM, ampicillin; AMC, amoxicillin-clavulanate; AZM, azithromycin; CIP, ciprofloxacin; CTX, cefotaxime; CXM, cefuroxime; ETP, ertapenem; LVX, levofloxacin; MEM, meropenem; MIC, minimum inhibitory concentration; SXT, trimethoprim-sulfamethoxazole; TE, tetracycline; TGC, tigecycline

Full size image
Table 1 Antibiogram of non-typeable Haemophilus influenzae isolates
Full size table

Plasmid-mediated drug resistance

Plasmid-encoded antibiotic resistance genes were assessed in each isolate. An intact blaTEM-1 was detected in 6 isolates (Table 2). Partial sequences of blaTEM-1 were detected in isolate No. 23. Moreover, genes with high sequence identity to blaROB-1 were discovered in isolates No. 22 and 23; blaROB-11 in isolate No. 22. Two of these BLPAR isolates (No. 06 and 22) were β-lactamase-positive amoxicillin-clavulanate resistant. Among 5 tetracycline-resistant isolates, tetR/acrR family transcriptional regulator gene was detected in isolate No. 02 and tet(B) in the other 4 isolates. Moreover, mef(A) was discovered in the azithromycin-non-susceptible isolate (No. 04). No plasmid-mediated quinolone-resistance genes or trimethoprim-sulfamethoxazole-resistance genes were characterized.

Table 2 Identification of plasmid-encoded drug-resistant gene in non-typeable Haemophilus influenzae isolates
Full size table

Genomic variations associated with drug resistance

In the analysis of β-lactam resistance, we focused on penicillin-binding proteins (PBPs) and penicillin-insensitive-related proteins. In ampicillin-resistant isolates, the group I ftsI mutation was detected in 2 BLPAR isolates, and both group II and III ftsI mutations were detected in 3 BLPAR isolates (Table 3). In BLNAR isolates, group II ftsI mutation was detected in isolate No. 05, and group III-like ftsI mutation was observed in isolates No. 09 and 13. A novel ftsI mutation type was found in isolate No. 19. Furthermore, genetic variations, namely 1070G > A (Ser357Asn), 1594A > T (Thr532Ser), and 1669T > C (Tyr557His) in ftsI, and 453G > C (Met151Ile) in lpoA (PBP1a), were associated with cefuroxime-resistance (Table 4). The area under the ROCs (AUROCs) for Ser357Asn, Thr532Ser, and Tyr557His substitutions in PBP3 and Met151Ile in PBP1a to discriminate the MIC of cefuroxime were 0.805, 0.910, 0.946, and 0.891, respectively. Other than ftsI, 3 genetic variations, namely 1552G > A in HI_0032 (mrdA) for Ala518Thr substitution in PBP2, 82C > A in HI_0197 (mepA) for Gln28Lys in penicillin-insensitive murein endopeptidase, and 674G > T in HI_0364 (pbpG) for Arg225Leu in PBP7, were correlated with amoxicillin-clavulanate non-susceptibility in NTHi (Table 4). The AUROCs for Ala518Thr substitution in PBP2, Gln28Lys in penicillin-insensitive murein endopeptidase, and Arg225Leu in PBP7 to discriminate the MIC of amoxicillin-clavulanate were 0.750, 0.767, and 0.859, respectively.

Table 3 Amino acid substitutions in FtsI in ampicillin-resistant non-typeable Haemophilus influenzae isolates
Full size table
Table 4 Genetic variations associated with cefuroxime, amoxicillin-clavulanate, ciprofloxacin/levofloxacin, and trimethoprim-sulfamethoxazole non-susceptibility in non-typeable Haemophilus influenzae isolates
Full size table

In isolates non-susceptible to fluoroquinolones and folate pathway antagonists, 12 genetic variations in gyrA, gyrB, parC, parE, and the parA-parC intergenic region and 6 variations in folA were associated with ciprofloxacin/levofloxacin and trimethoprim-sulfamethoxazole non-susceptibility, respectively (Table 4). The AUROCs for Ser84Phe, Asp88Gly, Glu433Asp, and Ile472Val substitutions in GyrA, Ala725Val in GyrB, and Ser145Asn in ParE to discriminate the MIC of levofloxacin were all higher than 0.73. Corresponding AUROCs for Lys20Arg and Ser84Ile substitutions in ParC and n.1600250A > C in the parC-parE intergenic region were higher than 0.82. Furthermore, AUROCs for Asn13Ser, Ile95Leu, and Lys107Gln substitutions in FolA to discriminate the MIC of trimethoprim-sulfamethoxazole were 0.954, 0.795, and 0.775, respectively.

We analyzed the sequence of related transporter and porin genes in the NTHi isolates (Table S1). No genetic variations in these genes were associated with the resistance to the agents we tested.

Protein substitutions associated with high-level drug resistance

Four NTHi isolates had a high-level cefuroxime resistance (MIC ≧64 μg/mL) (Fig. 2A). Moreover, 13 and 14 isolates had high-level resistance to levofloxacin (MICs ≧32 μg/mL) and trimethoprim-sulfamethoxazole (MICs ≧32 μg/mL), respectively. A logistic regression analysis revealed that Thr532Ser substitution in FtsI was an independent factor associated with high-level cefuroxime resistance (Table 5). Furthermore, Asp88Gly, Glu433Asp, and Asp740Glu substitutions in GyrA, Ala725Val in GyrB, Lys20Arg and Ser84Ile in ParC, Ser145Asn in ParE, and n.1600250A > C in the parC-parE intergenic region held up the relationship to high-level levofloxacin resistance. A network clustering showed the connection of these 8 variations (Fig. 2B). No single mutation was found to be associated with high-level trimethoprim-sulfamethoxazole resistance.

Fig. 2
figure 2

Isolates with high-level resistance to cefuroxime, levofloxacin, and trimethoprim-sulfamethoxazole. (A) A Venn diagram for NTHi isolates with cefuroxime MIC ≧64 μg/mL and both levofloxacin and trimethoprim-sulfamethoxazole MICs ≧32 μg/mL is shown. (B) A network clustering of mutation for high-level levofloxacin resistance is shown. The size of the circles and width of the lines are proportional to the frequency of each mutation and the connection of two mutations, respectively

Full size image
Table 5 Logistic regression analyses of genetic variations associated with high-level resistance to cefuroxime, levofloxacin, and trimethoprim-sulfamethoxazole in NTHi
Full size table

Molecular epidemiology of MDR NTHi

The molecular epidemiology of antimicrobial resistance in 13 MDR NTHi isolates is displayed in Fig. 3. Ampicillin resistance was detected in 12 isolates; of these 8 were BLPAR (blaTEM-1 in 6 and blaROB-1 in 2), and 4 were BLNAR. Amino acid substitutions in FtsI in ampicillin-resistant isolates are listed in Table 3. Thr532Ser, Ser357Asn, and Tyr557His substitutions in FtsI were detected in 7, 4, and 5 cefuroxime-resistant MDR isolates, respectively. Furthermore, Ala518Thr substitution in MrdA and Arg225Leu in PbpG were detected in all amoxicillin-clavulanate-non-susceptible MDR isolates, whereas Gln28Lys in MepA was absent in isolate No. 22. Regarding levofloxacin and ciprofloxacin non-susceptibility, Ser84Phe, Asp88Gly, Glu433Asp, Ile472Val, and Asp740Glu substitutions in GyrA were observed in 8, 4, 6, 7, and 3 isolates, respectively. Moreover, Ala606Val substitution in GyrB, Ser84Ile on ParC, and n.1600250A > C in the parC-parE intergenic region were discovered in more than 75% of MDR isolates. Asn13Ser substitution in FolA was found in all 12 trimethoprim-sulfamethoxazole-resistant MDR isolates, and Leu67Pro, Ile95Leu, and Lys107Gln in more than 9 MDR isolates. Finally, Tet(B), which confers tetracycline resistance was detected in 4 MDR isolates and Mef(A), which confers azithromycin resistance, in one isolate.

Fig. 3
figure 3

Schematic diagram of amino acid substitution panels in multidrug-resistant non-typeable Haemophilus influenzae isolates. Blue box, ampicillin-resistance; green box, amoxicillin-clavulanate resistance; beige box, cefuroxime resistance; red box, levofloxacin resistance; gray box, trimethoprim-sulfamethoxazole resistance; purple box, azithromycin resistance; orange box, tetracycline resistance. Substitutions associated with high-level drug resistance were marked with an asterisk

Full size image

Discussion

With a similar clinical spectrum to Hib, NTHi has become the most common cause of invasive H. influenzae diseases in all age groups since the global implementation of Hib-conjugate vaccination [2]. After the isolation of the first MDR H. influenzae strain in 1980 [16], infections caused by these bacteria have increased rapidly, thereby triggering enhanced vigilance worldwide. Yang et al. noted a nosocomial outbreak of MDR NTHi in the respiratory care ward of a community hospital in Taiwan [13]. Yamada et al. identified an NTHi strain resistant to ampicillin, amoxicillin-clavulanate, levofloxacin, clarithromycin, and tetracycline [14]. Li et al. also identified MDR NTHi strains with β-lactam/azithromycin/trimethoprim-sulfamethoxazole as the most frequent pattern of resistance [15]. The rising incidence of NTHi in difficult-to-treat respiratory infections and the drug-resistant evolution in this microbe warrant its surveillance. The World Health Organization included H. influenzae on the priority pathogen list for which new antibiotics are urgently required [17]. Our study is crucial for not only yielding a better understanding of the molecular epidemiology of antibiotic resistance in NTHi but also for providing critical information on therapeutic options to confront this emerging threat.

Ampicillin remains the first-line antibiotic in the treatment of H. influenzae infections. However, the widespread plasmid-mediated class A serine β-lactamases, most often TEM-1 and rarely ROB-1 [18, 19], drastically reduce the efficacy of ampicillin and other β-lactam agents against H. influenzae. Our previous 12-year survey demonstrated that the prevalence of BLPAR and BLNAR H. influenzae was 30.5% and 25.5%, respectively, in Taiwan [20]. Jean et al. reported the rate of BLPAR H. influenzae infection was 55% in patients treated at the intensive care units of 10 major teaching hospitals in Taiwan [11]. In this study, 30.8% of the NTHi isolates were BLPAR and 15.4% were BLNAR. TEM-1 β-lactamase was detected in 75% of the BLPAR isolates. Mutations in ftsI may increase ampicillin resistance in BLPAR strains. Two BLPAR isolates had type I FtsI mutations (Arg517His), 3 had type II (Asn526Lys), and 3 had type III (Ser385Thr with Asn526Lys) [21,22,23]. This may be the reason for the higher mean MIC of ampicillin in the BLPAR isolates (58.4 μg/mL) than in the BLNAR isolates (2.3 μg/mL). BLNAR H. influenzae strains are also less susceptible to other β-lactam agents, particularly cephalosporins. Cefaclor or cefuroxime might be better indicators than ampicillin for BLNAR strains [24, 25]. Straker et al. identified that Ser357Asn substitution in FtsI has a profound effect on cefuroxime resistance in H. influenzae, while D350N, A502T, and N526K do not directly attribute to cefuroxime resistance but may play an additive role in reduced cefuroxime susceptibility when present with other substitutions [26]. No mutations in dacA and dacB, encoding PBP5 and PBP4, respectively, were associated with cefuroxime resistance [26]. All 8 cefuroxime-resistant isolates had Ser357Asn substitution in FtsI except isolate No. 23, in which Arg517His, Thr532Ala, and Tyr557His substitutions in FtsI and Met151Ile in LpoA, a PBP activator, were found. Met151Ile substitution in LpoA was also detected in 2 cefuroxime-resistant isolates. The result from logistic regression showed that Thr532Ser substitution in FtsI but not Arg517His and Ser385Thr was correlated with high-level cefuroxime resistance. However, 5 NTHi isolates harbored this mutation and type III-like mutation in FtsI simultaneously. Accordingly, Thr532Ser substitution in FtsI may greatly augment cefuroxime resistance. Three genetic variations, causing missense mutations in MrdA, MepA, and PbpG, respectively, were linked to amoxicillin-clavulanate non-susceptibility in NTHi. MrdA (PBP2) is a D,D-transpeptidase. PbpG (PBP7) is a D,D-endopeptidase that hydrolyzes diaminopimelate-alanine bonds. MepA is a penicillin-insensitive murein endopeptidase that cleaves the D-alanyl-meso-2,6- diamino-pimelyl amide bond. These proteins are essential for peptidoglycan synthesis during cell wall elongation. Further investigations are needed to understand how mutations in these proteins contribute to amoxicillin-clavulanate non-susceptibility.

Regarding macrolide resistance, H. influenzae may acquire efflux pumps homologous to the AcrAB efflux machinery or to major facilitator superfamily transporters in Escherichia coli, thereby leading to the extrusion of macrolides from cells. Although the existence of macrolide efflux genes in H. influenzae has been a subject of debate, several studies have confirmed their presence. Roberts et al. identified azithromycin-resistant NTHi in children with cystic fibrosis, and mef(A), erm(B), and erm(F) were detected in 74%, 31%, and 29% of isolates, respectively [27]. Moreover, mef(A)-bearing NTHi was reported in Japan [28]. We identified one azithromycin-resistant isolate (MIC = 8 μg/mL) with mef(A). No genetic variations in the efflux pump genes were associated with azithromycin resistance in this isolate. We should be cautious regarding the expansion of macrolide-resistant NTHi because of a horizontal transfer of macrolide-resistance genes from other microbes over time. Four MDR NTHi isolates obtained conjugative plasmids with tet(B), which encodes an efflux protein leading to tetracycline resistance. Tet(B) can confer resistance to both tetracycline and minocycline, but not to the new glycylcyclines [29]. Therefore, 4 tet(B)-positive isolates exhibited MICs of tetracycline higher than 4 μg/mL but were all susceptible to tigecycline.

Because of toxicity, quinolones are not favored as therapeutic options in the treatment of H. influenzae diseases, particularly in children. Resistance to quinolones in H. influenzae is rare in the West but severe in East Asia. In Taiwan, as indicated by earlier surveys, the levofloxacin non-susceptibility rate in H. influenzae is 12.5% to 14.1% [20, 30]. Herein, 57.7% of the NTHi isolates and 61.5% of MDR strains were non-susceptible to ciprofloxacin and levofloxacin. Similar to the cause of quinolone resistance in other bacterial species, resistance in NTHi mainly results from mutations in the quinolone resistance-determining region of the genes encoding DNA gyrase or topoisomerase IV [31, 32]. Studies have demonstrated that in H. influenzae, Ser84 and Asp88 substitutions in GyrA, and Gly82, Ser84, and Glu88 in ParC were closely related to low quinolone susceptibility [33,34,35]. In addition to these mutations, we identified novel significant mutations that rendered high-level levofloxacin resistance. One point mutation in the parC-parE intergenic region was detected in 73% of levofloxacin/ciprofloxacin-non-susceptible isolates but was absent in all susceptible isolates. This mutation locates in the promoter region of parC and how it affects gene expression is unclear. These mutations were shown to be related to high-level levofloxacin resistance in a univariate but not a multivariate regression model (data not shown), indicating that these factors were tightly correlated, which can be evidenced by the network clustering model.

Trimethoprim-sulfamethoxazole, by blocking tetrahydrofolate synthesis in bacterial cells, is a relatively inexpensive drug, with wide use worldwide. Trimethoprim resistance is caused primarily by the low affinity of mutant dihydrofolate reductase to the drug but not to natural substrates, whereas sulfamethoxazole resistance generally arises from the acquisition of either sul1 or sul2, encoding different forms of dihydropteroate synthase that are not inhibited by the drug. In this study, neither sul1/sul2 nor genetic variations in folP and thyA were linked to trimethoprim-sulfamethoxazole resistance in the NTHi isolates. Six substitutions in FolA identified in this study have been characterized to be associated with low-level trimethoprim-sulfamethoxazole resistance [36, 37]. These 6 substitutions individually had no relevance in high-level resistance and seemed to act in an interconnected manner to cause the higher MIC values. Our earlier report demonstrated that more than half of H. influenzae strains in Taiwan were resistant to trimethoprim-sulfamethoxazole, making this drug no longer efficacious in the treatment of H. influenzae infections [20]. Therefore, clinicians should be aware of the local trimethoprim-sulfamethoxazole resistance situation and use this drug carefully with consideration of qualified susceptibility reports.

Conclusion

In closing, no extended-spectrum β-lactamases were detected in our NTHi isolates. Carbapenems and tigecycline are currently active against MDR NTHi strains. Studies monitoring trends in the antimicrobial resistance of H. influenzae should be standardized and continued. Furthermore, antibiotic stewardship at healthcare facilities should exercise standard operating procedures to manage H. influenzae infections, particularly those caused by MDR strains, and to avoid the acquisition of broad spectrum β-lactamases or other resistant mechanisms to last-line drugs.

Materials and methods

Isolation and serotyping of H. influenzae

The Institutional Review Board of E-Da Hospital approved this study (No. 2021009) and waived the informed consent requirement because this study used only bacterial isolates and did not have any negative impact on patients. No patients were under the age of 18 years. After the exclusion of H. influenzae isolates from unqualified sputum specimens, 26 NTHi isolates were collected from September 2017 to December 2019, as previously reported [38]. Eight isolates were obtained from blood cultures and 18 from sputum specimens. Suspected bacteria were re-isolated on chocolate agar (Creative Life Sciences, New Taipei city, Taiwan) and identified using matrix-assisted laser desorption ionization-time of flight mass spectrometry (VITEK MS, BioMérieux, Marcy-l’Étoile, France). Slide agglutination tests were conducted using Difco Haemophilus Influenzae Antisera (Becton, Dickinson and Company, Sparks, MD, USA) to determine the serotype of H. influenzae. Isolates were stored in skim milk containing glycerol at − 80 ℃ until use.

Antimicrobial susceptibility testing

The minimum inhibitory concentrations (MICs) of 12 antimicrobial agents, namely ampicillin (penicillins), amoxicillin-clavulanate (β-lactam combination agents), cefuroxime (cephems), cefotaxime (cephems), ertapenem (carbapenems), meropenem (carbapenems), trimethoprim-sulfamethoxazole (folate pathway antagonists), azithromycin (macrolides), levofloxacin (fluoroquinolones), ciprofloxacin (fluoroquinolones), tetracycline (tetracyclines), and tigecycline (glycylcycline), were examined using ETEST strips (BioMérieux). Disk diffusion tests using a BBL Sensi-Disc (Becton, Dickinson and Company) were also performed to confirm the drug susceptibility patterns. The ATCC 49247 NTHi strain was used as the control in antimicrobial susceptibility tests. The antimicrobial susceptibility breakpoints were interpreted in accordance with the Clinical & Laboratory Standards Institute, M100, 31st Edition guidelines. However, for tigecycline, the US Food and Drug Administration breakpoints were used. MDR is defined as non-susceptibility to at least one agent in three or more antimicrobial categories [39]. High-level resistance to cefuroxime, levofloxacin, and trimethoprim-sulfamethoxazole in NTHi were defined as MIC of ≧64 μg/mL, ≧32 μg/mL, and ≧32 μg/mL, respectively.

Genomic and plasmid DNA purification and sequencing

Each NTHi isolate was cultured on chocolate agar at 37 ℃ with 5% CO2 for 20 h, which was followed by harvesting. Bacterial genomic DNA and plasmid DNA were extracted using a Presto Mini gDNA Bacteria Kit (Geneaid, New Taipei City, Taiwan) and a QIAGEN Plasmid Midi Kit (QIAGEN, Germantown, MD, USA), respectively. DNA samples were concentrated using a Qubit dsDNA HS assay kit (Thermo Fisher Scientific). The DNA library was prepared using a Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, USA) and sequenced with the MiSeq system (Illumina) using the 250-bp paired-end read protocol.

Bioinformatic analyses

The quality of raw read data was analyzed using FastQC [40]. High-quality reads were trimmed using Trimmomatic [41] and assembled into genomes using BWA software with H. influenzae strain Rd KW20 as the reference (NC_000907.1). A genome subset was generated using PacBio, and Samtools was used to assess plot coverage. Variant calling and annotation were conducted using Bcftools and SnpEff, respectively. Drug-resistant genes on plasmids were assessed using BLASTx and compared with sequences available in the RefSeq database of the National Center for Biotechnology Information. Alignment of reads to plasmids was accomplished using the QIAGEN CLC Workbench software (CLC bio).

Statistical analysis

SPSS 18.0 for Windows was used for statistical analyses. Fisher's exact tests were used to search amino acid substitutions that were associated with drug non-susceptibility. A receiver operator characteristic (ROC) curve was used for different amino acid substitutions to differentiate the MIC of antimicrobial agents. Logistic regression assays were used to identify amino acid substitutions that were associated with high-level resistance to cefuroxime and levofloxacin. Significance is set at P < 0.05 (2-tailed).