Imipenem heteroresistance but not tolerance in Haemophilus influenzae during chronic lung infection associated with chronic obstructive pulmonary disease

Antibiotic resistance is a major Public Health challenge worldwide. Mechanisms other than resistance are described as contributors to therapeutic failure. These include heteroresistance and tolerance, which escape the standardized procedures used for antibiotic treatment decision-making as they do not involve changes in minimal inhibitory concentration (MIC). Haemophilus influenzae causes chronic respiratory infection and is associated with exacerbations suffered by chronic obstructive pulmonary disease (COPD) patients. Although resistance to imipenem is rare in this bacterial species, heteroresistance has been reported, and antibiotic tolerance cannot be excluded. Moreover, development of antibiotic heteroresistance or tolerance during within-host H. influenzae pathoadaptive evolution is currently unknown. In this study, we assessed imipenem resistance, heteroresistance and tolerance in a previously sequenced longitudinal collection of H. influenzae COPD respiratory isolates. The use of Etest, disc diffusion, population analysis profiling, tolerance disc (TD)-test methods, and susceptibility breakpoint criteria when available, showed a significant proportion of imipenem heteroresistance with differences in terms of degree among strains, absence of imipenem tolerance, and no specific trends among serial and clonally related strains could be established. Analysis of allelic variation in the ftsI, acrA, acrB, and acrR genes rendered a panel of polymorphisms only found in heteroresistant strains, but gene expression and genome-wide analyses did not show clear genetic traits linked to heteroresistance. In summary, a significant proportion of imipenem heteroresistance was observed among H. influenzae strains isolated from COPD respiratory samples over time. These data should be useful for making more accurate clinical recommendations to COPD patients.


Introduction
Antimicrobial resistance (AMR) is one of the most serious threats to Public Health and health care.Recent studies based on predictive statistical models estimated 4.95 million deaths associated with bacterial AMR in 2019, including 1.27 million deaths attributable to bacterial AMR.Among those, lower respiratory infections accounted for more than 1.5 million deaths associated with resistance in 2019, making it the most burdensome infectious syndrome (Murray et al., 2022).Resistant bacterial cells survive antibiotic treatment by carrying resistance factor(s) that allow growth at high antibiotic concentrations regardless of treatment duration.However, mechanisms other than resistance contribute to therapeutic failure.Heteroresistance leads to a fraction of the bacterial population displaying a substantial increase in minimal inhibitory concentration (MIC) value, meaning that those cells are less susceptible to the antibiotic and can grow in its presence (El-Halfawy and Valvano, 2015;Andersson et al., 2019;Band and Weiss, 2019).Conversely, tolerance is the ability of antibioticsusceptible bacteria to survive during transient exposure to high concentrations of a bactericidal antibiotic, whereas persistence is the ability of a clonal bacterial subpopulation to survive transient exposure to high concentrations of antibiotic, i.e., tolerance at the subpopulation level (Brauner et al., 2016;Balaban et al., 2019;Dewachter et al., 2019).Such phenomena escape standardized procedures used for antibiotic treatment decision-making as they do not involve MIC changes.
Nontypeable Haemophilus influenzae (NTHi) is a pathobiont that commonly resides in the human nasopharynx, from which it can cause otitis media, conjunctivitis, sinusitis and lower respiratory infections in children; exacerbations of chronic obstructive pulmonary disease (COPD) and cystic fibrosis (CF) in adults; and invasive disease in neonates, immunocompromised adults, and the elderly (Duell et al., 2016;Ahearn et al., 2017;Jalalvand and Riesbeck, 2018;Su et al., 2018).This opportunistic microorganism often develops resistance to β-lactam antibiotics by several mechanisms including β-lactamase production and polymorphisms in the penicillin-binding protein 3 (PBP3) (Parr and Bryan, 1984).Cephalosporins have been used since the emergence of ampicillin-resistant strains, and carbapenems are currently used as an alternative to extended-spectrum cephalosporins for initial empiric treatment of severe infections (Cherkaoui et al., 2018a,b).To date, resistance of H. influenzae to imipenem remains low (García-Cobos et al., 2014;Lâm et al., 2020;Potts et al., 2022;Safari et al., 2022), and tolerance has not been reported.However, limitations of the broth microdilution method seem to underestimate the extent of imipenem heteroresistance in H. influenzae, as previously reported for invasive and respiratory isolates (Cerquetti et al., 2007;Cherkaoui et al., 2017Cherkaoui et al., , 2018a,b;,b;Lâm et al., 2020).Although not formally demonstrated, imipenem heteroresistance may be linked to a combination of altered PBP3 and drug fluxes (Cherkaoui et al., 2017).Moreover, antibiotic heteroresistance or tolerance could be pathoadaptive traits of H. influenzae, facilitating persistence in patients by decreasing antimicrobial susceptibility.This aspect is currently unknown.
Here, we used a previously sequenced longitudinal set of COPD strains (Moleres et al., 2018) to address open questions about antibiotic heteroresistance and tolerance during chronic H. influenzae infection.Imipenem tolerance was not detected among the tested H. influenzae respiratory isolates.However, imipenem heteroresistance seems to be a frequent, complex and likely multifactorial phenotype that should be taken into account in the clinical practice.

Haemophilus influenzae imipenem heteroresistance but not tolerance during COPD chronic lung infection
Imipenem susceptibility was assessed in a previously sequenced longitudinal set of NTHi respiratory strains (Moleres et al., 2018) by using the Etest and disc diffusion methods (Figure 1 step 1).As expected, most strains (93.3%) were susceptible according to EUCAST guidelines (Figure 2A, Table 1).However, growth of colonies inside the inhibition zone, suggesting the presence of heteroresistant populations, was observed in a significant proportion of susceptible strains, 78.57and 40.48%, based on Etest and disc diffusion data, respectively.Strains showing heteroresistance by disc diffusion also resulted heteroresistant by Etest.Heteroresistance was heterogeneous among strains, which led us to establish categories upon repeated visual inspection based on the amount and relative location of colonies grown in the inhibition area: H1 refers to strains where colony growth was observed at the edge of the inhibition zone; H2 refers to strains where colony growth was observed, to a higher or lower extent, inside the entire inhibition zone (Figure 2A).Only two strains were β-lactamase positive, P640 and P671, H1 and susceptible, respectively.
Extensive use of the disc diffusion assay in clinical procedures, in comparison to Etest, led us to use disc diffusion as the method of choice to establish strain categories for further analysis (Table 1, column "Disc diffusion 24 h", subcolumn "Phenotype", bold text).Phenotypic variation within patient and/or clonal type (CT) was next analyzed, data are summarized in Figure 3 and Table 1.Within-patient phenotypic variation (susceptible, S; heteroresistant, H; resistant, R) was observed for all patients except for patients 3 and 12 (all isolates were susceptible), and for patient 9 (clonal resistant isolates) (Figure 3).However, when looking at strains belonging to the same CT, phenotypic variation was less common.For clarity, CTs were previously established by applying goeBurst in PHYLOViZ to allele assignments made from core protein-coding genes shared by all genomes.For each single-copy core gene, unique nucleotide sequences were defined as distinct alleles, strains with <15 allelic differences across all core genes were clustered together, and each resulting connected component was assigned an arbitrary CT number (Moleres et al., 2018).In CTs 3,7,16,45,54,76,100,106 and 137, all isolates were susceptible; in CTs 9, 17, 18, 40 and 73, all isolates were heteroresistant; in CT44, all isolates were resistant.Only CTs 14, 48, 72 and 95 showed intra-CT phenotypic variation.Strains in CTs 40 and 73 showed different degrees of heteroresistance (H1 or H2 phenotypes); strains in CTs 14, 48 and 72 were susceptible or heteroresistant; strains in CT 95 were resistant o heteroresistant (Table 1).
Further heteroresistance characterization assays were carried out (Steps 2 and 3 in Figure 1).Step 2 consisted of collecting the colonies grown within the disc diffusion inhibition zone of heteroresistant strains, to be used as starting inoculum in a new disc diffusion assay.In all cases, the heteroresistance phenotype was maintained.Step 3 consisted of profiling imipenem heteroresistance by the quantitative method of 10.3389/fmicb.2023.1253623Frontiers in Microbiology 03 frontiersin.orgpopulation analysis (Figure 2B).Although considered to be reliable to identify heteroresistance, population analysis profiling (PAP) interpretation may be complex as several aspects are considered when defining this phenotype, including the presence of a resistant subpopulation with a MIC at least eightfold higher than the highest concentration of drug that does not affect the growth of the main population, the growth of the resistant subpopulation above the clinical breakpoint with frequency above 1×10 −7 , or a gradual decrease in the number of CFU instead of a single-step loss of survival (Andersson et al., 2019;Stojowska-swędrzyńska et al., 2022).Here, 19 isolates displaying a heterogeneous resistant profile by the disc diffusion method were selected for this analysis as representative of genomic heterogeneity (7 strains belonging to single-strain CTs, P593, P630, P639, P638, P640, P588 and P619; 12 strains belonging to multi-strain CTs, P653, P652, P604, P623, P616, P612, P600, P608, P607, P628, P594 and P590, see Table 1).RdKW20 was used as a susceptible control.Based on growth at the antibiotic concentrations tested, three strain groups were observed: subpopulations of cells growing on medium containing ≥8XMIC, including concentrations higher than 2 μg/mL imipenem (EUCAST clinical breakpoint), were present in the cultures of strains P612, P623, P653, P652, P640 and P593 (31.57%, 6/19), Figure 2B, left panel; subpopulations of cells growing on medium containing ≥4XMIC, including concentrations higher than 2 μg/mL imipenem were present in the cultures of strains P619, P590, P594, P588, P607, P608, P638, P639 and P630 (47.36%, 9/19), Figure 2B, middle panel; subpopulations of cells growing on medium containing ≥2XMIC with a gradual decrease in the number of CFU were present in the cultures of strains P628, P600, P616 and P604 (21.05%, 4/19), Figure 2B, right panel.H1 and H2 strains were distributed across these three groups.Intra-CT variation was not observed in this assay, strains P652 and P653 (CT73) belong to the >8XMIC group; strains P607 and P608 (CT40) belong to the >4XMIC group.Correlation between disc diffusion and PAP was observed as all tested strains contained subpopulations of cells growing on medium with imipenem concentrations higher than their respective MIC, but strain behavior was heterogeneous and not always fulfilling all PAP heteroresistance recommended criteria.Lastly, given that the disc diffusion method does not allow detecting antibiotic tolerance, additional testing was considered (Figure 1 step 4).The tolerance disc (TD) test (Gefen et al., 2017) was adapted to H. influenzae growth requirements (see Methods section).This method relies on tolerant bacteria surviving the transient exposure to the antibiotic, without growing in the inhibition zone In sum, under the conditions tested, H. influenzae heterogeneous heteroresistance, but not tolerance, was found in a collection of NTHi respiratory isolates.Next, seeking for possible gene to phenotype associations, we analyzed variation both in specific genes and at the genome wide level.
Allelic variation of the acrA and acrB genes rendered 20 and 29 different variants, respectively (Tables 3, 4).Please note that AcrA variant 1 and AcrB variant 1 were not present in the strain collection under study.From those, 4 AcrA and 5 AcrB variants were only present in heteroresistant strains.Regarding amino acid changes in AcrA, 51 are newly described, and the Met36Val, Ser114Asn, Ser115Phe, and Ser326Leu changes were only found in heteroresistant strains (Tables 3, 6).Likewise, the amino acid changes Val12Ile, Ala492Thr, Ile909Met, Leu945Ile and Ile996Val in AcrB were only found in heteroresistant strains (Tables 4, 6).AcrB allelic variation also involved loss of function due to a 7-nt insertion (5´-ACTATAT) leading to a truncated variant present in 5 isolates, with susceptible or heteroresistant phenotypes (Supplementary Table S1).
The acrAB gene expression is negatively regulated by AcrR (Dean et al., 2005), which also undergoes allelic variation.AcrR variation involved amino acid substitutions leading to 14 variants with 3 newly described changes (Table 5).Please note that AcrR variant 1 was not present in the strain collection under study.Three AcrR allelic variants, represented by the Ile12Ser, Ser110Pro and Ala159Val changes, were only present in heteroresistant strains (Table 6).Moreover, AcrR loss of function was found in 10 imipenem heteroresistant and 14 susceptible strains.Truncated variants were due to (i) a C-T transition in three different positions (P645, nucleotide 291; P591, nucleotide 4; P851 and P592, nucleotide 7) rendering an early stop codon in 4 strains; (ii) an AT insertion generating the same frameshift in 17 strains; (iii) three combined 138, 6 and 5 bp deletions in 3 strains (Supplementary Table S2).We next tested expression of the acrA gene across strains, considering both acrR allelic variation and strain phenotypic heterogeneity for strain set selection.As shown in Figure 4, the acrA gene highest expression was observed in strains with AcrR truncated variants (P591, P645, P621), but such gene inactivation did not necessarily lead to high acrA gene expression, and full-length variants led in some cases to high acrA gene expression (P607).Conversely, phenotypic clustering was not observed, as susceptible (S), heteroresistant (H1 or H2) or resistant (R) phenotypes did not correlate with acrR allelic variation and/or with acrA gene expression.Therefore, although we do not exclude an association to imipenem heteroresistance, the AcrAB-TolC efflux pump component and regulatory elements are not only contributors to the observed phenotypes.

Genomic variation across strain collection does not render gene to heteroresistant phenotype association
Next, taking advantage of this longitudinal set, we analyzed genomic variation among closely related strains, i.e., belonging to the same CT, showing intra-CT phenotypic heterogeneity in terms of imipenem effects.We selected four strain series, belonging to CTs 14, 48, 72 and 95 to examine intra-CT phenotypic variation involved in susceptible-heteroresistant, or heteroresistant-resistant changes.Among all SNPs detected along whole genome alignment, none of them did segregate susceptible and heteroresistant strains.Lastly, a pangenomic analysis was performed to look for differences in present and absent genes between phenotypically different strains.Such analysis did not reveal any candidate gene that was significantly present or absent in susceptible or heteroresistant strains explaining this phenotype.Together, genomic variation across this strain collection did not render gene to phenotype clear-cut associations.

Discussion
The carbapenems (e.g., imipenem) are a group of antibiotics that belong to the β-lactam class.Low toxicity and low prevalence of resistance make carbapenems a valuable alternative for initial empirical treatment of severe infections before a definitive microbiological diagnosis is available.Imipenem is not a first-line agent against NTHi infections and comprehensive data on imipenem resistance are scarce, but imipenem heteroresistance has been reported in NTHi isolates from an epidemiologically undefined collection of isolates (Cerquetti et al., 2007;Cherkaoui et al., 2017Cherkaoui et al., , 2018a,b;,b;Lâm et al., 2020).Following this notion, this study reports a high prevalence rate of imipenem heteroresistance in a longitudinal set of respiratory H. influenzae strains.Strains were isolated from COPD sputum samples recovered over time, i.e., persistent isolates from patients which, based on available clinical data, had not been administered imipenem (Moleres et al., 2018), thus excluding a direct correlation between the observed heteroresistance and imipenem exposure.This               In terms of methods, differences between Etest and disc diffusion assays were observed.Such differences were most likely due to the different way the antibiotic diffuses into the agar, but strains with heteroresistance by disc diffusion were also heteroresistant in the Etest assay.This, added to the extensive use of the disc diffusion assay in clinical procedures, led us to use it as the method of choice to establish strain categories in this study, and we strongly suggest that these differences should be considered in the clinical practice.Another aspect worth discussing is the heterogeneity observed in terms of heteroresistance intensity.We established two categories (H1 and H2), but it should be emphasized that this phenotypic visual interpretation requires training and should be made by double-blind assessment.Special attention should also be paid to procedures and interpretation of quantitative methods, i.e., population analysis profiling (PAP), as NTHi heterogeneity among strains may hinder inocula normalization when multiple unrelated strains are simultaneously assayed.This aspect was considered here, and it should apply to any phenotypic  We acknowledge that, for simplicity when performing the PAP assay, the same range of imipenem concentrations was used for all strains independently of their respective MIC, and subpopulations with an MIC x-fold higher that the highest antibiotic dose not affecting the growth of the main population were calculated afterwards.This is another aspect to consider when testing a wide strain number, as adjusting imipenem range to each respective MIC will increase assay complexity.In this regard, performing population analysis profiling in the clinical practice is unlikely as it is time consuming and requires trained personnel.However, heteroresistance frequency should not be eluded in clinical settings and methodical and thorough disc diffusion assay interpretation may contribute to do so.Following this notion and also aiming for practicality, TD-test implementation in the clinical routine may be worth considering, as hardly extra-material and manipulation are required upon disc diffusion, and it may provide key qualitative assessment of tolerance or persistence.
Here, we also tried to identify genomic traits that may account for the observed heteroresistance, but definitive conclusions were not reached.We used a bottom-up approach, addressing allelic variation in a small set of four genes previously associated to H. influenzae imipenem heteroresistance, followed by genome-wide allelic variation in available series of clonally related and phenotypically heterogeneous strains and, lastly, genome-wide differences in terms of gene distribution by sorting strains in two groups based on phenotyping, i.e., imipenem susceptible and heteroresistant isolates.The ftsI, arcA, arcB and arcR gene allelic variation is synthesized Table 6 by highlighting changes found only in heteroresistant strains, but overall variation is complex to read as shown in Tables 2-5.In the case of the ftsI gene encoding PBP3, Asn526Lys and Arg517His amino acid changes were previously associated to imipenem resistance (Cherkaoui et al., 2017), but also appeared in susceptible strains from our collection.Thirty-eight strains (42.2%, 38/90) had changes surrounding the Lys512-Thr-Gly (KTG) motif, being Ala502Thr (20.67%, 24/90), Asn526Lys (40%, 36/90) and Ala530Ser (7.78%, 7/90) the most commonly found.These changes were found across phenotypes.We also found the Met377Ile substitution, surrounding the Ser379-Ser-Asn (SSN) motif, in a heteroresistant strain, and four changes surrounding the Ser327-Thr-Val-Lys (STVK) motif in 36 strains, the most common Asp350Asn (34.44%, 31/90), unlikely related to heteroresistance as found across phenotypes.Conversely, acrB inactivation could lead to higher susceptibility; however, this did not seem to be the case as four strains carrying acrB truncated variants showed heteroresistance.Also, inactivation of AcrR could lead to higher acrAB gene expression and therefore, to higher drug efflux capacity and lower susceptibility; however, this was not necessarily the case as arcA gene expression was not necessarily higher in strains presenting acrR truncated variants, and 14 out of the 24 strains carrying acrR truncated variants showed imipenem susceptibility.
Heteroresistance is common for several different bacterial species and antibiotic classes, but it is often difficult to detect and study due to its phenotypic and genetic instability, and pose clinical concern as it could increase in frequency during antibiotic exposure and cause treatment failure (Andersson et al., 2019).Besides H. influenzae (Cerquetti et al., 2007(Cerquetti et al., , 2017)), heteroresistance to carbapenems has been detected in clinical isolates of Acinetobacter baumannii, Enterobacter cloacae, Klebsiella pneumoniae, Escherichia coli, Pseudomonas aeruginosa, and Salmonella typhimurium (Andersson et al., 2019;Stojowska-swędrzyńska et al., 2022).In terms of mechanisms, resistance mutations (SNPs, insertions and deletions) and increased copy number of tandem amplifications of genes that increase antibiotic resistance may lead to unstable heteroresistance (Andersson et al., 2019;Nicoloff et al., 2019).Other mechanisms of heteroresistance have been reported for which no information regarding stability is available.Studies have suggested that low expression of the OprD porin, marginal overexpression of some efflux pump encoding genes, and intrinsic β-lactamases, may contribute to imipenem-heteroresistant P. aeruginosa (Ikonomidis et al., 2008;Mei et al., 2015;Xu et al., 2020).An insertion sequence was found in the promoter region of a class C β-lactamase gene in imipenem-heteroresistant multi-drug resistant A. baumannii strains, and was associated to overexpression of such β-lactamase.In that case, the use of carbapenem seemed to be the only risk factor identified for the emergence of carbapenem-heteroresistance (Lee et al., 2011).Also, bla OXA-58-like , bla OXA-51-like , ISAba2 and ISAba3 have been associated with A. baumannii heterogeneous resistance to carbapenems (Fernández Cuenca et al., 2012).In the case of K. pneumoniae, heteroresistant colonies were shown to have significantly elevated expression of the bla KPC gene compared with the native populations but did not retain heteroresistance when subcultured in drug-free media (Pournaras et al., 2010); more recently, resistant and susceptible K. pneumoniae subpopulations have been shown to be genetically identical by genome sequencing (Sancak et al., 2022).In our strain collection, we could not establish associations between imipenem heteroresistance and overexpression of efflux pump encoding genes, β-lactamases, or imipenem use.Together, our observations suggest that imipenemen heteroresistance may be a frequent but also complex and likely multifactorial phenotype.These observations align with the notion that high prevalence of heteroresistance with the potential for treatment failure highlights the limitations of MIC as the sole criterion for susceptibility determination (Nicoloff et al., 2019).Finally, we advocate that the existing heteroresistance high rate is worth our attention, and support the need for facile and rapid protocols to identify heteroresistance in pathogens.

Bacterial isolates and growth conditions
We used a previously sequenced set of NTHi isolates serially collected from COPD patients (Moleres et al., 2018).Thirteen patient series with four or more longitudinally sampled NTHi isolates were included.This collection consists of 90 isolates collected over a period of 1 to 9 years, belonging to 40 clonal types (CTs) (Table 1 and Figure 3).H. influenzae RdKW20 (Fleischmann et al., 1995)  imipenem resistance (MIC >2 μg/mL; growth inhibition diameter < 20 mm) (Figure 1 step 1).In all cases, assays were performed in three independent occasions, using different plate batches and imipenem Etests and discs.Results were independently read by the first and last authors.The growth of colonies inside the growth inhibition halo in the presence of imipenem disc or Etest was considered as an indicator of heteroresistance.When colonies were found inside the growth inhibition halo, such colonies were collected and used as starting point to repeat the disc diffusion assay as described above (Figure 1 step 2).The β-lactamase activity was determined by the chromogenic cephalosporin test using nitrocefin as a substrate and following the manufacturer's directions (Becton Dickinson).

Population analysis profiling
Imipenem PAP (Figure 1 step 3) was determined by following previously described procedures (Cerquetti et al., 2007;Cherkaoui et al., 2017Cherkaoui et al., , 2018a,b),b).Briefly, H. influenzae strains were grown on PVX agar for 16 h.A suspension of bacteria was then generated in 30 mL PBS and adjusted to OD 600 = 1.Suspensions were centrifuged at 4,000 r.p.m. for 10 min, pellets were resuspended in 3 mL PBS rendering the starting inoculum dose, ~10 9 CFU/mL, and serially diluted up to 10 −7 .The microdilution plate count was performed by plating 10 μL droplets of each dilution onto a set of sHTM agar plates containing increasing concentrations of imipenem (concentration range, 0.25-64 μg/mL) and onto antibiotic-free sHTM agar.After incubation at 37°C with 5% CO 2 for 48 h, bacterial colonies were counted.Assays were carried out in triplicate and at least two independent occasions (n ≥ 6).

Tolerance disc test
The two-steps TD-test is a modification of the standard discdiffusion assay that enables the detection of tolerant and persistent bacteria by promoting the growth of the surviving bacteria in the inhibition zone, once the antibiotic has diffused away (Gefen et al., 2017).This test, originally described for Escherichia coli, was adapted to H. influenzae in terms of growth media requirements (Figure 1 steps 1 and 4).Briefly, a standard disc diffusion assay, using 10 mg imipenem discs on MH-F agar, was followed by an antibiotic disc replacement with a 2 mg glucose-containing disc after 24 h.Then, plates were incubated at 37°C with 5% CO 2 for 24 h.For preparation of glucose discs, 6 mm diameter filter paper discs (Whatman, ref. WHA2017006) were moistened with 5 μL of a 40% sterile glucose solution.Tolerance was assessed by the appearance of colonies within the growth inhibition zone.Controls were run in parallel by maintaining the antibiotic disc for 48 h.The growth of colonies inside the growth inhibition halo, after replacing the antibiotic disc with a glucose disc, was considered as an indicator of tolerance.

PBP3, AcrA, AcrB, and AcrR allelic identities for Haemophilus influenzae strain collection
Analysis of allelic variation of the ftsI, acrA, acrB and acrR genes was performed using Geneious Prime (Biomatters).A local database containing the 90 whole genome sequences was created and local blast tool was employed, with the genome of H. influenzae RdKW20 (NC_000907) as a reference.Subsequently, the nucleotide alignments were translated to detect amino acid changes.

Comparative pangenome and single nucleotide polymorphism variation analysis
The pangenomic analysis was performed to detect differences in presence or absence of genes when comparing susceptible and heteroresistant strains.Imipenem-resistant strains were excluded.A gene presence and absence matrix was created with Roary (Page et al., 2015), with a minimum percentage of identity of 70% for BLASTp, and the analysis of potential genes associated to heteroresistance was done using Scoary (github.com/AdmiralenOla/Scoary).An intra-CT SNPs analysis was performed in those CT presenting differences in terms of heteroresistance phenotype.A whole genome alignment for each CT was created by Snippy v4.4.0 (github.com/tseemann/snippy)using P595-8370 (GCA_003425765) as reference for CT14; P621-7028 (GCA_003425445) for CT48; P604-7629 (GCA_003415395) for CT72; and P625-8065 (GCA_003415195) for CT95.

Statistical analyses
In all cases, p < 0.05 value was considered statistically significant.Analyses were performed using Prism software, version 7 for Mac (GraphPad Software) statistical package and are detailed in the Figure Legends when needed.
may be important as previous work by our group showed acquisition of macrolide resistance in H. influenzae during persistent respiratory infection in COPD patients receiving long-term azithromycin treatment(Carrera-Salinas et al., 2023), maybe masking possible heteroresistance development.Also, no specific trends were observed among serially isolated and clonally related strains.When looking at strains belonging to the same CT, only four of them showed intra-CT phenotypic variation, i.e., CTs 14, 48 (see haplotypes C and D), 72 and 95.

TABLE 1
Summary of imipenem susceptibility testing (Etest and disc diffusion), and FtsI, ArcA, ArcB, and ArcR variant distribution across a NTHi longitudinal respiratory strain collection.

TABLE 2
FtsI allelic variation in the H. influenzae strain collection under study.

TABLE 2
AcrA allelic variation in the H. influenzae strain collection under study.

TABLE 4
AcrB allelic variation in the H. influenzae strain collection under study.

TABLE 4
AcrR allelic variation in the H. influenzae strain collection under study. ." indicates no amino acid substitution.We refer as variant 1 to the one present in reference strain RdKW20.I12S, R18G and K35N changes are newly described in this study. "

TABLE 6
Amino acid changes in FtsI, AcrA, AcrB, and AcrR that were only found in imipenem heteroresistant H. influenzae isolates.