Molecular Analysis of Rising Fluoroquinolone Resistance in Belgian Non-Invasive Streptococcus pneumoniae Isolates (1995-2014)

We present the results of a longitudinal surveillance study (1995–2014) on fluoroquinolone resistance (FQ-R) among Belgian non-invasive Streptococcus pneumoniae isolates (n = 5,602). For many years, the switch to respiratory fluoroquinolones for the treatment of (a)typical pneumonia had no impact on FQ-R levels. However, since 2011 we observed a significant decrease in susceptibility towards ciprofloxacin, ofloxacin and levofloxacin with peaks of 9.0%, 6.6% and 3.1% resistant isolates, respectively. Resistance to moxifloxacin arised sporadically, and remained <1% throughout the entire study period. We observed classical topoisomerase mutations in gyrA (n = 25), parC (n = 46) and parE (n = 3) in varying combinations, arguing against clonal expansion of FQ-R. The impact of recombination with co-habiting commensal streptococci on FQ-R remains marginal (10.4%). Notably, we observed that a rare combination of DNA Gyrase mutations (GyrA_S81L/GyrB_P454S) suffices for high-level moxifloxacin resistance, contrasting current model. Interestingly, 85/422 pneumococcal strains display MICCIP values which were lowered by at least four dilutions by reserpine, pointing at involvement of efflux pumps in FQ-R. In contrast to susceptible strains, isolates resistant to ciprofloxacin significantly overexpressed the ABC pump PatAB in comparison to reference strain S. pneumoniae ATCC 49619, but this could only be linked to disruptive terminator mutations in a fraction of these. Conversely, no difference in expression of the Major Facilitator PmrA, unaffected by reserpine, was noted between susceptible and resistant S. pneumoniae strains. Finally, we observed that four isolates displayed intermediate to high-level ciprofloxacin resistance without any known molecular resistance mechanism. Focusing future molecular studies on these isolates, which are also commonly found in other studies, might greatly assist in the battle against rising pneumococcal drug resistance.


Introduction
Streptococcus pneumoniae is a major cause of community-acquired respiratory infections including otitis media and pneumonia, as well of serious invasive infections like septicaemia and meningitis [1]. Penicillins and macrolides were mainstay in the treatment of respiratory diseases for decades [2], but the worldwide spread of drug-resistant clones translated into increased usage of fluoroquinolones [3,4]. Fluoroquinolones are synthetic, broad-spectrum antibiotics targeting the DNA gyrase (GyrA/B) and topoisomerase IV (ParC/E) enzymes, which are critically involved in DNA supercoiling and chromosome segregation, respectively [5]. The early fluoroquinolones ciprofloxacin (CIP) and ofloxacin (OFL) target ParC and display poor potency against pneumococci, rapidly leading to emergence of resistance [6]. In the late 1990s, they were replaced by the so-called "respiratory fluoroquinolones levofloxacin (LVX; the active isomer of ofloxacin) and moxifloxacin (MXF) that acts on both enzymes [2]. In Belgium, this has been reflected by steadily declining sales of OFL and norfloxacin while, in contrast, the use of MXF has markedly increased since 2009 (Fig 1) and will probably further expand as its patent has recently expired. Since the global switch to LVX and MXF was established, the worldwide prevalence of fluoroquinolone resistance (FQ-R) in S. pneumoniae remained below 2% [7] Moreover, it seems unrelated to the serotype switches that were observed upon the introduction of 7-and 13-valent pneumococcal conjugate vaccination [8]. Genetic analyses showed that a first mechanism of FQ-R is through stepwise accumulation of spontaneous mutations in the quinolone resistance determining regions (QRDR) of gyrA and parC, and rarely gyrB and parE [9]. The effect of a given mutation depends on the genetic context and the type of fluoroquinolone used [10]. ParC mutations at positions 79 and 83 are most frequently found among pneumococci and are associated with CIP and LVX usage [11]. These first-step mutations lead to a dramatic increase in mutant prevention concentration of all fluoroquinolones [12], enabling prompt selection of secondary and tertiary QRDR mutations in GyrA (mainly positions 81 and 85) required for the FQ-R phenotype [3]. Unlike β-lactam and macrolide resistance mechanisms, QRDR mutations do not appear to be clonally spread and only a minor fraction (0.5-10%) stems from recombination with co-habiting commensal streptococci of the viridans group [13].
In recent years, the role of efflux in low-level S. pneumoniae FQ-R has become more and more appreciated. Beyond causing a moderate increase in MIC, increased efflux is indeed associated with rising mutational frequencies in the QRDRs [14]. Gene disruption experiments, expression analyses and susceptibility testing in the presence of the efflux pump inhibitors led to current consensus that two distinct transporters, PmrA and PatA/B, are capable of fluoroquinolone efflux [15][16][17][18][19]. The Major Facilitator Superfamily pump PmrA, however, is reported as intrinsically inactive and non-inducible under CIP pressure [20]. More clinical relevance is therefore attributed to the reserpine-sensitive heterogenic ABC efflux pump PatAB. Deletion of this pump in a laboratory strain led to hypersusceptibility to CIP [21], and its expression is induced in the presence of CIP [20]. Moreover, constitutive overexpression of patA/B was observed in roughly one-third of clinical isolates with low-level FQ-R [17], and is linked to gene duplication and disruptive mutations in the transcriptional attenuator upstream patA [22][23][24]. Recently, point mutations in PatA were associated with increased CIP resistance by putative enhanced substrate binding [25].
Although most S. pneumoniae surveillance studies focus on bacteraemia, recent work estimated that for every adult bacteraemic case there are three non-invasive infections [26]. In this paper, we present data on FQ-R in non-invasive pneumococci from a longitudinal surveillance program across Belgian clinical laboratories (1995-2014), spanning the world-wide transit era between the use of early (CIP, OFL) and newer (LVX, MXF) fluoroquinolones. We noted that resistance against the early drugs are markedly on the rise since 2011. By studying the molecular background to dissect the relative contribution of target site mutations versus drug efflux, we identified interesting pneumococcal isolates which confer FQ-R through yet uncharacterized mechanisms.

Bacterial strains
Non-invasive respiratory clinical isolates of S. pneumoniae were collected during winter seasons between 1995 and 2014 in 15 clinical laboratories throughout Belgium by members of The Belgian Streptococcus pneumoniae Study Group. The access to patient information was encrypted. All isolates were kept at −70°C in Brain Heart Infusion Broth (Difco) containing 10% (v/v) glycerol until transfer to the Scientific Institute of Public Health for susceptibility testing and downstream molecular analyses. The identification of each isolate made by the participating laboratories was confirmed using PCR targeting the autolysin encoding gene lytA [26], slide agglutination (Slidex pneumo Kit™, BioMérieux, Marcy-l'Étoile, France) and Optochin (Opto-F, bioMérieux) tests, all performed according to the manufacturer's instructions. For selected strains, capsular sequence typing (CST) was performed by sequence analysis of the wzh gene using a dedicated web application (http://www.rivm.nl/mpf/spn/cst/) [27].

Antibiotic susceptibility testing
For each isolate, the minimal inhibitory concentration (MIC) was determined by broth microdilution as recommended by the US Clinical and Laboratory Standards Institute (CLSI; called National Committee for Clinical Laboratory Standards (NCCLS) at the onset of the study in 1997. The following fluoroquinolones were provided as laboratory standards with known potency by the manufacturers of the original products: levofloxacin and ofloxacin from Aventis Pharma (Mumbai, India), CIP and MXF from Bayer (Leverkusen, Germany). All antibiotics were tested for 16 serial twofold dilutions (0.001-32 μg/mL), with S. pneumoniae ATCC 49619 [28,29], S. pneumoniae TPN 881, Staphylococcus aureus NCTC 11561 and S. aureus ATCC 29123 being included as quality control organisms in each series (S1 Table). Interpretation of the results was based on the breakpoints set by the European Committee on Antimicrobial Susceptibility Testing (EUCAST; http://www.eucast.org/). To assess possible synergy between fluoroquinolones and efflux pump inhibitors, commercial E-tests of CIP and MXF (bioMérieux) were applied on MH Blood agar plates containing 0 and 20 μg/mL reserpine. This method was devised after observing that reserpine causes turbidity of broth, preventing a correct reading of the results of the microdilution assay.

Determination of FQ-R related sequences
The DNA sequences of the QRDRs in gyrA, gyrB, parC and parE genes, and of the regulatory regions and coding sequences of patA and patB were determined by PCR sequencing using the primers listed in S2 Table. All sequences were screened for SNPs in comparison to corresponding regions of FQ-sensitive clinical strains using Clustal Omega (http://www.ebi.ac.uk/Tools/ msa/clustalo/); the stability of the patA transcriptional attenuator was assessed using the MFold web server (http://unafold.rna.albany.edu/?q=mfold).

Quantitative real-time PCR
All tested S. pneumoniae strains were grown overnight in duplicate at 35°C and 5% CO 2 on Mueller-Hinton agar plates supplemented with 5% defibrinated sheep blood (Bio-Rad Laboratories, Hercules, CA, USA). Bacteria were collected using a sterile loop and suspended in 15 mL Todd-Hewitt Broth medium to an OD 620 nm of 0.1-0.2. These samples were incubated at 35°C with occasional stirring to late mid-late exponential phase (OD 620~0 .5-0.6), at which point 4 mL of the culture was sampled and cells were harvested by centrifugation (8,000 x g for 10 min at 4°C). Cell pellets were rapidly frozen at −80°C until further processing.
Total RNA extraction was performed using the InviTrap 1 Spin Cell RNA Mini Kit (Stratec Biomedical, Birkenfeld, Germany) according to the manufacturer's instructions and stored at -80°C. Next, the samples were treated two consecutive times with 2 units TURBO™ DNase (Thermo Fisher, Waltham, USA) for 30 min at 37°C, followed by inactivation of the enzyme. To confirm removal of genomic DNA, the pmrA gene of S. pneumoniae was amplified as described elsewhere [19], and RNA concentrations were determined using Qubit fluorescence (Thermo Fisher).
cDNA was synthesized from 150 ng total RNA using the SuperScript 1 III First-Strand Synthesis System for RT-PCR (Life Technologies) according to the manufacturer's instructions and using random hexamer primers. Residual RNA was removed using RNase III for 30 minutes at 37°C. Finally, real-time PCR was performed in an iQ cycler (Bio-Rad) in 25 μL reaction mixtures containing 12.5 μL of iQ SYBR Green Supermix (2×), 400 nM of forward and reverse primers and 5 μL of cDNA in RNase/DNase-free water. Primers used for amplification of pmrA, patA and patB are listed (S2 Table), and conditions were used as previously described [19]. Differential gene expression was calculated from the two replicates, as described in Pfaffl et al. [30] and using rpoD and proC genes as references to normalize transcript levels, as specified by PrimerDesign (Southampton, UK).

Strain collection
A total of 5,602 unduplicated clinical isolates of S. pneumoniae were included in this study. Isolates were obtained from both ambulatory and hospitalized patients presenting a non-invasive respiratory clinical picture. These strains were collected during the winter seasons in 16 surveys spanning the period 1995 and 2014 by 15 participating clinical laboratories, determinedly selected to obtain a representative sampling of the country (Fig 1). Overall, 47.6% (varying between 40.9% and 53.9%) of the isolates were collected in the Southern part of the country, 44.7% (varying between 39.3% and 49.1%) in the Northern part and 7.6% (varying between 4.7% and 10.1%) in the Brussels area.
Next, we investigated the influence of the role of efflux in fluoroquinolone resistance using the efflux pump inhibitor reserpine. Hereto, we selected 422 pneumococcal isolates displaying varying MIC CIP and repeated the MIC testing of CIP and MXF in the presence of reserpine (MIC CIP/MXF+R ). We observed that for 85 (20.1%) isolates, at least a fourfold decrease in MIC-CIP was achieved upon addition of the efflux pump inhibitor (Fig 2), which is a common threshold for the definition of an efflux phenotype [31,32]. For 57 (13.5%) isolates there was no effect, while in 16 cases (3.7%) this reduction was very drastic and caused a decrease to up to nine MIC CIP dilutions, accounting for the entire resistance phenotype. In contrast, MXF MICs were much less decreased by the addition of reserpine as the maximal effect was a two-fold reduction in 45 (10.7%) strains (Fig 2). Of note, we used E-tests for all analyses which included reserpine, and recorded MICs which were generally higher than with the corresponding microdilution method: 52.8% one or less, and 90.2% two or less dilutions difference in CIP MICs, and 56.7% one or less, and 89.2% two or less dilutions difference for MXF (S4 Table). Comparable deviations have been reported elsewhere for other Gram-positive bacteria [33][34][35], and could be attributed to a conservative interpretation due to insufficient growth of the bacterial lawn.

Analysis of QRDR regions and serotypes
For the same set of 422 isolates, the QRDR of all gyrA, gyrB, parC and parE genes were sequenced (Table 2 and S4 Table). To dissect the influence of QRDR from efflux-mediated <1 μg/mL, respectively. Secondly, we identified signatures of recombination with members of the S. mitis group, judged by the presence of ParC S52G, N91D and/or GyrA S114G substitutions [13], in 10.4% of the strains. These recombinant genes were already identified at the onset of our study, but no significant increase in topoisomerase recombination was noted by 2014. Classical QRDR mutations were retrieved in GyrA at positions 81 (n = 24) and 85 (n = 1), ParC positions 78 (n = 1), 79 (n = 38) and 83 (n = 7), and ParE position 435 (n = 3). These topoisomerase mutations were found in varying combinations, arguing against clonal expansion of FQ-R (Table 2). To investigate this hypothesis, we performed CST typing on 54 FQ-R isolates which showed a wide variety of associated serotypes (S3 Table). In concordance to previous studies, isolates carrying mutations in both topoisomerases unequivocally displayed high-level resistance to CIP (MIC CIP+R > 12 μg/mL). In contrast, strains with sole mutations in ParC (50.9% of isolates with mutated QRDR) or GyrA (13.7%) display more variable MIC CIP +R values. For example, four strains carrying a GyrA S81F/G mutation displayed a MIC CIP+R of 32 μg/mL (e.g., 13C28), whereas other strains with the same mutation (e.g., 05A05 and 04L17) only reached 0.5-1 μg/mL.
Finally, some interesting genotypes were observed. For example, we identified a very rare GyrB mutation (P454S) in isolate 05A20 which, in combination with GyrA S81L and wild-type Topoisomerase IV, was correlated with full resistance to MXF (MIC MXF+R = 32 μg/mL) ( Table 2). An even more noticeable result was that some isolates with complete wild-type QRDRs (e.g., 13C24, 11I08) were nevertheless fully resistant to CIP, even in the presence of reserpine (Table 2).

Efflux pump expression analysis
Next, we assessed the contribution of PmrA and PatAB efflux transporters to FQ-R by measuring early-log expression levels of patA, patB and pmrA in 94 S. pneumoniae isolates grown in the absence of antibiotics, in comparison to the control strain S. pneumoniae ATCC 49619 [27,28]. Strains were selected based on various susceptibilities to reserpine, i.e. displaying either    Table 3.
Since putative highly different genetic backgrounds preclude reliable strain-to-strain comparison, we performed Kruskal-Wallis testing (with Dunn's multiple comparison test) on three sample groups of strains either susceptible, intermediate or resistant to CIP (S1 Fig). This nonparametric method showed no statistically significant differences in pmrA expression among the three groups. In contrast, both patA and patB expression was significantly upregulated in CIP-resistant, but not in CIP-intermediate S. pneumoniae strains. Although this correlated with the observed susceptibility to reserpine, the levels of reserpine-mediated MIC reductions varied strongly among strains with similar transcript levels. For example, strains 99J16 and 13L23 both strongly overexpressed patA (6.0±1.3 and 14.8±4.4, resp.) and patB (56.5±13.6 and 48.4±10.5, resp.), but showed a sixteen-vs. threefold reduction in MIC CIP in the presence of the efflux pump inhibitor.
Conversely, strain 10N11 also overexpressed patA and patB but showed no reserpinedependent reduction of MIC CIP (Table 3). Notably, in 12 and 19 isolates only patA or patB were overexpressed, respectively, arguing against uniform operon coregulation for these genes. Finally, in 17.0% of the reserpine-susceptible strains both patA and patB were downregulated. Since the heterogeneous PatAB pump requires both functional subunits to be functional [36], these two last observations strongly indicate the presence of other reserpine-sensitive systems involved in FQ-R.
Constitutive induction of patA has very recently been correlated to disruption of the transcriptional terminator of the upstream hexA gene [22,24]. We therefore sequenced this upstream region in 103 isolates. Although none of previous described mutations were retrieved, we identified six novel mutations: C(-41)T, G(-40)A, G(-46)T, G(-48)A, G(-49)A and C(-28)T. Each of these mutations could be related to decreased hairpin stability (ΔG increases > 3.2 kCal/mol), leading to increased transcription. The A(-52)G mutation was found not to play a role in patA regulation. It is important to note that these mutations were found in only a fraction (15.5%) of the isolates which overexpress patA, implying that terminator disruption is only a minor regulatory mechanism in the isolates under study.

Discussion
At the introduction of the respiratory fluoroquinolones LVX and MXF in the treatment of (a) typical pneumonia, there was concern that while treatment success in the short term could be enhanced, highly FQ-R S. pneumoniae strains would emerge by accumulation of additional QRDR mutations [37]. The continued high use of CIP for specific respiratory indications, such as the treatment of bronchial infections in cystic fibrosis patients [13], poses an additional risk factor to select for first-step ParC mutations which precede the ones in GyrA under CIP selective pressure. Moreover, the continuous exposure to sub-MIC levels of CIP and levofloxacin has been shown to select for efflux overexpression [38].
In our surveillance data on FQ-R among Belgian non-invasive S. pneumoniae isolates (1995-2014), some evidence points in this direction. From 2011 onwards, we observe a trend towards increased resistance to CIP and ofloxacin, and (although only visible at the MIC 50 level) also for LVX. Our data from CST typing clearly indicates no clonal spread of CIP-R isolates, and thereby suggests there is no direct influence of vaccination campaigns on FQ-R in non-invasive pneumococci. The preference of first-step mutations in ParC is reflected by the 4:1 ratio of single QRDR mutations in the Topoisomerase IV subunits compared to the DNA Gyrase. Although a similar increase in CIP resistance was reported in Canada [10], this was not confirmed in other surveillance studies covering Europe, North America or Asia [8,9]. In Table 3. Expression analyses of a selection of S. pneumoniae strains, with inclusion of the QRDR sequences and phenotypic FQ-R analyses.  contrast, resistance to MXF was only sporadic and globally minimal. As these trends are still very recent, it is critical that resistance rates and their changes are continuously monitored in the near future.
The exceptional case of high-level resistance to MXF (MIC MXF+R of 32 μg/ml) was associated with a GyrB P454S mutation, combined with a mutated GyrA (S81L) subunit but a wildtype ParC/E. Notably, a Chinese group recently reported both ParE_P454S as GyrB_P454S to be associated with MXF resistance in combination with dually mutated GyrA and ParC residues [39,40]. However, our observation of a wild-type Topoisomerase IV QRDR region is important, as it contrasts with the current model which states that mutations in both topoisomerases are a prerequisite for high-level MXF resistance.
Analysis of the contribution of efflux pumps to pneumococcal FQ-R revealed no significant upregulation of the Major Facilitator PmrA in CIP resistant strains. We did observe varying constitutive expression of pmrA among clinical isolates, which has been shown before [16], but its contribution to drug resistance in non-invasive pneumococcal strains remains unclear. In contrast, non-parametric analyses showed marked higher expression of the ABC efflux pump PatAB associated with decreasing CIP susceptibility. Unfortunately, the underlying regulatory mechanism behind this upregulation remains unexplored. Although we found novel disruptive mutations in upstream transcriptional terminator sequences [22][23][24], this mechanism seems rather rare among clinical isolates as the large majority has wild-type upstream regions. A patA/B repressor has not been found or seems deleted in comparison to similar operons in related bacteria [41]. Various other levels of regulation can be envisioned at the post-transcriptional, translational or post-translational level.
Another important finding of this study is related to strains which seem deprived of known molecular FQ-R mechanisms, but yet display a resistant phenotype. For example, isolate 11I08 displays a wild-type QRDR yet a MIC CIP+R of 24 μg/mL. A possibility is the involvement of chromosomally encoded qnr-like proteins, which shield topoisomerases from invading fluoroquinolones [42]. In contrast, many reserpine-susceptible strains did not express PatAB pumps, implying the involvement of other reserpine-sensitive efflux mechanisms in pneumococcus. This can be either novel efflux pumps, like the recently identified DinF transporter [43], or any of the five transporter genes found to be consistently induced by fluoroquinolones [44]. In any case, isolates with elevated MICs but without defined resistance mechanism are also commonly reported in other studies [3,17,25], and deserve more attention in the future.
We acknowledge limitations of the presented patAB expression studies. First of all, putative gene duplication of patA [23] could not be detected with the applied methods. Moreover, although we assessed constitutive gene expression, it has been shown that expression is quickly upregulated upon exposure to CIP, with patA being more strongly upregulated than patB [19,40]. This might level out the difference we observed in basal expression levels between both genes.
In conclusion, 15 years after the introduction of respiratory fluoroquinolones, we observe a concerning rise in resistance among non-invasive pneumococci. MXF remains a very potent drug with minimal level of resistance, but a combination of rare mutations in the DNA Gyrase was associated with full resistance to this compound. While target topoisomerase mutations and efflux pump (over)expression clearly contribute to FQ-R, we add novel isolates to the existing collection of strains deprived of known molecular mechanisms of fluoroquinolone resistance. It would be of great value to bring these clinical isolates together, and unravel their resistance mechanisms through a profound, comparative molecular characterization at the genomic, transcriptomic and proteomic level.  Van Nimmen, L., Ide, L. A.Z. Jan Palfijn, 9000 Gent (Louis.ide@janpalfijngent.be). Frans, J., Van Noyen, R. Imeldaziekenhuis, 2820 Bonheiden (Johan.frans@imelda.be).