Genomic profiling of cefotaxime-resistant Haemophilus influenzae from Norway and Sweden reveals extensive expansion of virulent multidrug-resistant international clones

Abstract

Cefotaxime-resistant Haemophilus influenzae (CRHI) are a global concern, but little is known about their molecular epidemiology. The goal of this study was to perform genomic profiling of 191 CRHI from Norway (n = 183) or Sweden (n = 8) (2006-2018) and assess clonal spread using core genome multilocus sequence typing (cgMLST)-based Life Identification Number (LIN) codes based on whole genome sequencing (Ion Torrent). Cefotaxime resistance was confirmed with broth microdilution minimal inhibitory concentration (MIC), interpreted with the European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints. 35.7% of isolates with cefotaxime gradient MIC of 0.25 mg/L were falsely resistant. All but two isolates (blood) were non-invasive, and all but two (serotype f) were non-typeable. Characterization included calling of resistance determinants, ftsI typing (penicillin-binding protein 3, PBP3), and classification of PBP3-mediated beta-lactam resistance (rPBP3), with assignment to rPBP3 stage and group. All isolates had rPBP3-defining substitutions, and 78.5% were stage 3 (L389F positive). Beta-lactam MICs correlated well with rPBP3 genotypes. Significant proportions of stage 3 isolates were cross-resistant to ceftriaxone (86.0%) and meropenem (meningitis breakpoints, 26.0%). The CRHI prevalence in Norway doubled during the study period and approached 1%. A shift from stage 2 to stage 3 rPBP3 in 2011-2012 led to emergence of CRHI with higher beta-lactam MICs and co-resistance to multiple non-beta-lactams, including extensively drug-resistant (XDR) strains. The shift was driven by transformation with two distinct variants of the transpeptidase region and multiclonal expansion. 66.0% of the isolates belonged to 27 clusters. Ten clusters or singletons belonged to international CRHI clones represented in the PubMLST database. The study provides new insight into CRHI evolution, resistance profiles, and clonal dynamics in a period when this phenotype went from exceptional to unusual in Europe. International CRHI clones are described for the first time, including eight high-risk clones associated with invasive disease, calling for enhanced genomic surveillance. LIN coding, supplemented with ftsI typing and rPBP3 staging, is well-suited for definition of CRHI clones. LIN9, defined by ≤ 10 allelic differences, offered the highest resolution level fully supported by maximum likelihood core genome phylogeny and is proposed as a global standard for genomic surveillance of H. influenzae.

Keywords: BLNAR; BLPACR; CRHI; LIN; PBP3; cefotaxime; cgMLST; ftsI.

Figure 1

Evaluation of cefotaxime gradient MIC with broth microdilution (BMD) MIC as gold standard. The evaluation included 211 isolates with cefotaxime gradient MIC > 0.125 mg/L by supplementary testing (two isolates did not grow in the BMD MIC assay) (Supplementary Figure S1). (A) MIC correlations. (B) Agreement rates. Full agreement, identical MIC (green shades). Essential agreement, ±1 dilution (light green shades). Categorical agreement, identical categorization as susceptible (S) or resistant (R). The clinical breakpoint (S ≤ 0.125/R > 0.125 mg/L) is shown as a dashed vertical line in (A) (EUCAST v. 14.0, 2024) (see text footnote two).

Figure 2

Antimicrobial susceptibility testing (broth microdilution MIC) of included isolates (n = 191) against nine beta-lactam and 12 non-beta-lactam antibiotics, with categorization as susceptible (S), susceptible increased exposure (I), or resistant (R) according to EUCAST clinical breakpoints (v. 14.0, 2024) (see text footnote two). Epidemiological cut-off values (ECOFFs) were used for assessment of susceptibility to agents without breakpoints (azithromycin, tigecycline, and gentamicin) (see text footnote ten). Horizontal lines separate antimicrobial categories (Supplementary Table S1). Vertical lines indicate MIC50 and MIC90, representing MIC values covering 50% and 90% of the isolates, respectively.

Figure 3

Timelines showing proportions of the included isolates (n = 191) by (A) cefotaxime MIC, (B) ceftriaxone MIC, and (C) extent of resistance to non-beta-lactams, expressed as the number of categories of non-beta-lactams affected by resistance (including categories without breakpoints) (Supplementary Table S1). Absolute numbers of isolates are shown in Supplementary Figure S2B.

Figure 4

Maximum likelihood phylogenetic tree based on 1,286 core genes identified by pangenome analysis of the included isolates (n = 191) and H. influenzae Rd KW20 (ATCC 51907 and the reference sequence GCA_000027305.1). H. haemolyticus ATCC 33390 (GCA_004368535.1) was used as outgroup. Colored nodes, sequence types (ST) with ≥3 representatives (labeled). Transparent nodes, STs with 1–2 representatives. Black nodes, Rd KW20 (ST47 and ST1621). Invasive (n = 2) and encapsulated (n = 2) isolates are marked with asterisks and letters indicating serotype (f), respectively. Color bands show Life Identification Number (LIN) code (13 levels) derived from core genome multi-locus sequence typing (cgMLST) profiles (PubMLST protocol) (see text footnote four), PopPUNK strain (see text footnote seventeen), and clusters (01–27). PopPUNK strains and LIN5 with ≥3 representatives are labeled. Lacking LIN (five study isolates and the reference sequence) and no cluster assignment (singletons) are shown as transparent bands.

Figure 5

Classification of penicillin-binding protein 3-mediated resistance (rPBP3), with assignment of the included isolates (n = 191) to rPBP3 stage and group based on amino acid substitutions in key positions (bold) (Table 1). Assignment to PBP3 types based on substitution patterns was in line with a previously established system (letters for stage 1, numbers for stage 2–3) (Skaare et al., 2014a,b), except that the fragment was extended from aa 338–573 to aa 327–610. Previously defined PBP3 types with variable substitution patterns in the extended fragment (red positions) were divided into subtypes, e.g., 2a and 2b. Blue positions are encoded by the PubMLST ftsI allele (see text footnote twenty-five). For positions with observed substitutions (n = 51), the corresponding amino acids in the reference sequences of H. haemolyticus and H. parainfluenzae are shown. Dots, amino acids identical to H. influenzae Rd KW20 (reference). Misc, miscellaneous.

Figure 6

Timelines showing proportions of the included isolates (n = 191) by (A) rPBP3 stage and group, (B) PBP3-ftsI types (rPBP3 stage in brackets), and (C) clusters (same colors as in Figure 4 and Supplementary Figure S7). PBP3-ftsI types occurring in single isolates and isolates not belonging to clusters (singletons) are shown as white background in (B) and (C). Absolute numbers of isolates are shown in Supplementary Figure S2B.

Figure 7

Maximum likelihood phylogenetic tree based on partial ftsI sequences (transpeptidase region, nt 979–1,833, aa 327–610) in the included isolates (n = 191) and H. influenzae Rd KW20 (ATCC 51907 and the reference sequence GCA_000027305.1). The most divergent sequence, encoding PBP3 type 20 (defined in Figure 5) was used as outgroup. Node colors and labels indicate combinations of PBP3 types and ftsI alleles (PubMLST protocol, nt 976–1,596) (see text footnote twenty-five), termed PBP3-ftsI types, present in ≥2 isolates (n = 15, range 3–35) (same colors as in Figure 6B and Supplementary Figure S8). Transparent nodes, PBP3-ftsI types present in single isolates (n = 34). Black nodes, Rd KW20. Color bands indicate PopPUNK strain (see text footnote seventeen), rPBP3 group and stage, and MICs for cefotaxime (CTX), ceftriaxone (CRO), piperacillin-tazobactam (TZP), and meropenem (MEM). MIC colors reflect categorization as susceptible or resistant according to EUCAST clinical breakpoints (legend, dashed lines) (see text footnote two). CSF, cerebrospinal fluid.

Figure 8

Core genome phylogeny and antibiotic resistance. Maximum likelihood phylogenetic tree based on 1,286 core genes identified by pangenome analysis of the included isolates (n = 191) and H. influenzae Rd KW20 (ATCC 51907 and the reference sequence GCA_000027305.1) (same tree as in Figure 4). H. haemolyticus ATCC 33390 (GCA_004368535.1) was used as outgroup. Node colors and labels indicate clusters (01–27) (same colors as in Figures 4, 6C, and Supplementary Figure S7). Singletons have transparent nodes. Black nodes, Rd KW20. Color bands show PopPUNK strain (see text footnote seventeen), rPBP3 stage, and antibiotic resistance profiles. Blue, beta-lactams; red, “epidemiologically significant” non-beta-lactams (breakpoints for therapeutic use); green, other non-beta-lactams (Supplementary Table S1). Darker colors indicate acquired antibiotic resistance genes. Gray, silent gene. Classification of multidrug resistance was based on nine antimicrobial categories (disregarding agents and categories in brackets). Pc, penicillin; BLI, beta-lactamase inhibitor; ES, extended-spectrum; CSF, cerebrospinal fluid; MDR, multidrug-resistant; XDR, extensively drug-resistant.

Figure 9

Minimum spanning trees (MST) of clones of cefotaxime-resistant H. influenzae (CRHI) with members from three or more countries in addition to study isolates. The clones were identified through a search in the PubMLST database (2024-11-29, 6.552 genomes) (see text footnote six) for isolates with Life Identification Number (LIN) code corresponding to a threshold of ≤ 10 allelic differences (LIN9) identical to clusters or singletons in the present study. The search was limited to clones with stage 3 rPBP3 types because of the strong association with cefotaxime resistance. (A–C) show clones consisting entirely of CRHI isolates, while (D) shows a CRHI subclone carrying ftsI-26 (dashed box) emerging from a non-CRHI clone carrying ftsI-27, which does not encode rPBP3-defining substitutions. The trees were calculated from cgMLST profiles (PubMLST protocol, 1,037 loci) (see text footnote four) using the GrapeTree plugin (see text footnote twenty-two). Node colors by country. Numbers along lines connecting nodes indicate allelic differences by MST analysis. Larger nodes comprise two or more isolates with identical cgMLST profiles. Invasive CRHI are marked with asterisks.