ABSTRACT
Two whole-genome screening approaches are described for studying the mode of action and the mechanisms of resistance to trimethoprim (TMP) in the Gram-positive Streptococcus pneumoniae. The gain-of-function approach (Int-Seq) relies on a genomic library of DNA fragments integrated into a fucose-inducible cassette. The second approach, leading to both gain- and loss-of-function mutation, is based on chemical mutagenesis coupled to next-generation sequencing (Mut-Seq). Both approaches pointed at the drug target dihydrofolate reductase (DHFR) as a major resistance mechanism to TMP. Resistance was achieved by dhfr overexpression either through the addition of fucose (Int-Seq) or by mutations upstream of the gene (Mut-Seq). Three types of mutations increased expression by disrupting a predicted Rho-independent terminator upstream of dhfr. Known and novel DHFR mutations were also detected by Mut-Seq, and these were functionally validated for TMP resistance. The two approaches also suggested that an increase in the metabolic flux from purine synthesis to GTP and then to folate can modulate the susceptibility to TMP. Finally, we provide evidence for a novel role of the ABC transporter PatAB in TMP susceptibility. Our genomic screens highlighted novel aspects on the mode of action and mechanisms of resistance to antibiotics.
INTRODUCTION
Antimicrobial resistance (AMR) has become a global threat, and a list of priority AMR pathogens for research purposes was recently established (1). Whole-cell phenotypic screens are becoming standard for the discovery of novel bioactive molecules. These have the benefit of revealing cell-permeable chemical entities, but this is at the cost of knowledge about their mode of action (MOA). However, the identification of the target(s) of bioactive molecules is facilitated by whole-genome approaches such as mutant selection and genome sequencing (2–4), gain-of-function screens (5–7), or loss-of-function screens using gene deletion (8) and transposon-based (9) or clustered regularly interspaced short palindromic repeat-based libraries (10). Plasmid-based multicopy suppressor screens have been used to pinpoint genes that provide a selective advantage in the presence of drugs. These screens were especially sensitive when coupled to next-generation sequencing (NGS) (5, 11).
Streptococcus pneumoniae is a Gram-positive bacterium responsible for respiratory and invasive diseases. It is on the WHO priority pathogens list for its levels of resistance to β-lactam antibiotics. However, plasmid-based screens are difficult with S. pneumoniae, because plasmids cannot be electroporated (12, 13) and natural transformation is less efficient with plasmid DNA. Thus, we developed a gain-of-function genomics screen (Int-Seq) for S. pneumoniae by designing a gene overexpression cassette integrated into the fucose operon (14, 15).
Point mutations contribute to AMR, and chemical mutagenesis coupled to NGS, a screening approach called Mut-Seq (16), has been successfully used to study drug resistance (17–20). Thus, we have also adapted Mut-Seq for S. pneumoniae and selected for clones resistant to trimethoprim (TMP).
We used Int-Seq and Mut-Seq for genes conferring resistance to TMP, an inhibitor of dihydrofolate reductase (DHFR), which converts dihydrofolate (DHF) to tetrahydrofolate (THF), a key carbon donor in metabolism. TMP was chosen to benchmark the approaches, as resistance is usually achieved by point mutations or changes in gene expression, namely, through DHFR coding mutations which decrease the affinity to TMP (21–27) or through dhfr overexpression (28, 29). While the I100L substitution in DHFR is a major driver of TMP unresponsiveness in S. pneumoniae, it was shown to be absent from a significant number of resistant clinical isolates (22), implying the presence of other unrecognized TMP-resistant genotypes. Both Mut-Seq and Int-Seq captured dhfr as the primary TMP resistance gene but with new mutations, a novel mechanism of dhfr overexpression, as well as novel resistance pathways, collectively leading to new findings related to TMP resistance and mode of action.
RESULTS
Generation of a Streptococcus pneumoniae Int-Seq genomic library.We cloned the S. pneumoniae R6 fucose operon promoter (Pfcsk) with the chloramphenicol acetyltransferase (cat) gene and a downstream EcoRV restriction site (Fig. 1A and B). Genes fcsR and fcsK were cloned at the 5′ and 3′ flanks of the cassette for its targeting to the fucose operon (Fig. 1B and C). To validate the cassette, the gene coding for the green fluorescent protein (GFP) was cloned into the EcoRV site, and the GFP signals of S. pneumoniae were monitored by fluorescence microscopy in the presence and absence of 0.5% fucose. Bright green signals were obtained in the presence of fucose, while only a faint background of autofluorescence was observed in the absence of induction (see Fig. S1 in the supplemental material).
Overview of the Int-Seq approach. (A) The native fucose operon with its promoter (PfcsK) in the genome of S. pneumoniae R6. (B) The fucose cassette with its fcsR and fcsK genes (on its 5′ and 3′ ends, respectively) for integration of the cassette at the native fucose operon by homologous recombination. The chloramphenicol resistance marker (cat) as well as NotI and EcoRV restriction sites were added to the construct in a pGEM-T Easy backbone. Genomic DNA fragments of 2 to 5 kb were cloned into the EcoRV site. The final Int-Seq library is recovered by NotI restriction digest followed by its transformation into S. pneumoniae R6. (C) The genomic library is integrated in the fucose operon. The Int-Seq inserts are PCR amplified by the Int_Fw and Int_Rv primers prior to their sequencing by NGS.
A genomic library or random 2- to 5-kb fragments derived from S. pneumoniae R6 was then cloned into EcoRV (Fig. 1B). The genomic library was linearized by NotI and transformed into S. pneumoniae R6 (Fig. 1C). More than 100,000 clones with an average insert size of 1.5 to 3 kb were obtained, leading to a library of >100× genome coverage. The library inserts were amplified by PCR from genomic DNA (gDNA) extracted from the pool of transformants (Fig. 1C). Sequencing this baseline Int-Seq library confirmed that the genome was well represented (Fig. S2).
The S. pneumoniae Int-Seq library cells were then selected with TMP (2 mg/liter) and either with 0.5% fucose (14) or without. Upon fucose induction, TMP selection, and NGS, two loci were enriched, covering genes spr1425 to spr1430 and genes spr0266 to spr0269, respectively (Table 1). The most enriched locus encoded DHFR (spr1429) (Table 1). Surprisingly, dhfr was also enriched in the noninduced control (Table 1). This insert contained dhfr with its native promoter in antisense orientation in the fucose cassette. Both the native and integrated dhfr were expressed, leading to increased TMP resistance. For the selection under the induced condition, dhfr was only found in the sense orientation in the fucose cassette.
Genomic loci enriched by the Int-Seq screen
The dhfr gene as well as spr0267 and spr0268, coding for dihydrofolate synthetase (SulB) and GTP cyclohydrolase (FolE), were integrated into the fucose operon. Reverse transcription-quantitative PCR (RT-qPCR) confirmed the overexpression of dhfr by 5.6- ± 1.2-fold in the presence of 0.5% fucose, and this conferred a 4-fold increase in MIC for TMP (Table 1). Neither spr0267 nor spr0268 conferred resistance when individually overexpressed. However, cloning both genes in the fucose cassette increased the MIC for TMP by 2-fold (Table 1). Other genomic loci were weakly enriched during the Int-Seq procedure, but none tested decreased the susceptibility to TMP (Table S1).
Chemical mutagenesis and selection for resistance to TMP.S. pneumoniae R6 was treated with ethyl methanesulfonate (EMS) and selected with TMP. The mutagen concentrations (8× and 16× MIC for EMS), exposure (20 min) and recovery (3 h) times, and TMP selection doses were optimized (see Materials and Methods). The mutagenized population was selected for TMP resistance under aerobic and anaerobic atmospheres, which are encountered during S. pneumoniae colonization and infection, respectively, to assess the impact of oxygen levels on TMP resistance. S. pneumoniae is a catalase-negative facultative anaerobe for which oxygen concentration was indeed shown to influence metabolism (30). TMP selection was done with 8 or 16 mg/liter for anaerobic clones and 8 mg/liter for aerobic ones. The mutants selected under aerobic conditions had various levels of TMP resistance, while those obtained under anaerobic atmosphere all had the same MIC (Table 2).
Resistance levels of Streptococcus pneumoniae TMP-resistant mutants selected under aerobic and anaerobic atmosphere after random mutagenesis by EMS exposure
The genomes of 47 mutants selected under aerobic atmosphere were sequenced and compared to the genome of the parental wild-type (WT) clone. A total of 893 single-nucleotide variants (SNVs) were identified, representing ∼20 SNVs/genome. Slightly more SNVs occurred in intergenic sequences than in coding regions, and 70% of the coding mutations were nonsynonymous (Fig. 2A). We also found two SNVs in rRNA genes as well as eight indels in coding sequences (Fig. 2A). The mutagen EMS is known to modify G-C basepairs to A-T basepairs, and most mutations were indeed G-to-A and C-to-T transitions (Fig. 2B). The genome of 50 mutants selected under anaerobic atmosphere were also sequenced, and the dynamics of their 887 SNVs were similar (Fig. 2A and B).
Properties of mutations induced by EMS in Streptococcus pneumoniae. (A) Number and type of mutations found in Mut-Seq mutants selected for TMP resistance under aerobic (black) or anaerobic (white) atmosphere. Syn, synonymous; Non-syn, nonsynonymous. (B) Distribution of intragenic mutations detected in Mut-Seq mutants selected for TMP resistance under aerobic (black) or anaerobic (white) atmosphere.
Coding and regulatory dhfr mutations confer TMP resistance.Gene candidates were selected on the basis of mutations in several mutants and in their diversity. The dhfr gene carried the most diverse set of mutations, with SNVs either in the coding region (CDS mutations), in its usptream intergenic sequence (regulatory mutations), or both (Table S2). Within dhfr (spr1429), eight different SNVs were detected in ten mutants (seven aerobic and three anaerobic) (Table 3 and Table S2), including the widespread I100L mutation, known to confer TMP resistance (21), found in three mutants (Table S2). When independently transformed into the S. pneumoniae R6 WT backgound, the I100L substitution was the most potent (32-fold TMP resistance), followed by P28S/L31F/M53I (8-fold TMP resistance), P24S (4-fold resistance), and W9C/Y98C (2-fold resistance) mutations (Table 3). F50L was the only DHFR mutation that did not decrease TMP susceptibility (Table 3).
Genes mutated in more than two mutants that were functionally characterized for their role in TMP resistance
Two SNVs and one deletion were found upstream of dhfr in most mutants (Table S2). The G-64A transition upstream of dhfr was found in 78 mutants, the G-81A mutation in 11 mutants, and the deletion in 3 mutants (Table 3 and Table S2). When any of these 3 mutations were independently transformed in R6 WT, it increased expression of dhfr by ∼10-fold and increased TMP resistance by 8-fold (Table 3). All three mutations mapped upstream of the dhfr transcription initiation site, which we located 23 nucleotides upstream of the start codon of dhfr by 5′ rapid amplification of cDNA ends (RACE) PCR (Fig. 3A and B and Fig. S3). Instead, these mutations colocalized with a putative Rho-independent transcription terminator downstream of the dpr gene (Fig. 3A and B). The gene dpr is located immediately upstream of dhfr in the genome of S. pneumoniae (Fig. 3A). The G-64A and G-81A mutations were located on opposite arms of the same stem-loop structure (Fig. 3B). The indel mutation deleted the entire stem-loop. No transcription terminator could be predicted in the presence of any of the regulatory mutations. RNA sequencing data revealed a higher expression for dpr than dhfr in S. pneumoniae R6 (31), and dhfr overexpression (Table 3) may result from the coexpression of both genes by transcriptional readthrough. Hybridization of a Northern blot with a dpr probe revealed a single hybridization signal of 0.6 kb in S. pneumoniae R6 WT (Fig. 3C). In contrast, a hybridization signal of 1.2 kb corresponding to a dpr-dhfr cotranscript was obtained with each of the three regulatory mutations (Fig. 3C). Additional bands of higher molecular weight were also observed, and these matched the size expected from the cotranscription of different sets of genes from the dhfr locus (Fig. 3A and C). RT-qPCR also confirmed the increased abundance of the dpr-dhfr cotranscript in the presence of the G-64A and G-81A mutations (Fig. 3D). Transcriptional readthrough also occurs in S. pneumoniae R6 WT but at a level that could only be detected by RT-PCR (Fig. 3C and E).
Mutations upstream of dhfr disrupt a predicted Rho-independent transcription terminator. (A) Schematic representation of the dhfr locus of S. pneumoniae R6 from gene dpr (spr1430) to spr1423. Genes are represented by open arrows. The dpr probe used for hybridizing the Northern blot is indicated. The dashed line indicates the dpr-dhfr region scaled up in panel B. (B) A scale-up of the dpr-dhfr region with its putative Rho-independent transcription terminator downstream of dpr, as predicted by ARNold software (http://rna.igmors.u-psud.fr/toolbox/arnold/). The sequence of the terminator is shown, and the arms of its stem-loop are underlined. The G-64A and G-81A mutations are shown in boldface in the terminator sequence. The −35 and −10 boxes (gray boxes) of the dhfr and dpr promoters, transcription terminator (hairpin), and the experimentally validated transcription start sites (arrows) (see Fig. S3) are indicated. Also indicated are the SP1-SP3 (dhfr) and SP4 (dpr) primers used for 5′ RACE PCR for mapping the transcription initiation site of dhfr and dpr. (C) Northern blot hybridization of total RNA (20 μg) derived from S. pneumoniae R6 WT (lane 1) and transformants (lane 2, R6spr1429G-64A; lane 3, R6spr1429G-81A; lane 4, R6spr1429del-53-113). Northern blots were hybridized with a P32-labeled DNA probe covering gene dpr (spr1430). One representative of two independent experiments yielding similar results is shown. (D) RT-qPCR for monitoring the expression of the dpr gene (1), the dpr-dhfr cotranscript (2), and the dhfr gene (3) in S. pneumoniae R6spr1429G-64A and R6spr1429G-81A compared to R6 WT. (E) Cotranscription of dpr and dhfr occurs in S. pneumoniae R6 WT. Reverse transcription PCR was performed with a forward primer annealing at the 3′ end of dpr and a reverse primer annealing at the 5′ end of dhfr. Amplification of the cotranscript was performed in the absence of reverse transcriptase (lane 2) and using cDNA derived from S. pneumoniae R6 WT (lane 3). Lane 1, molecular weight marker (GeneRuler 1-kb plus DNA ladder; Invitrogen).
Novel genes involved in TMP resistance.While the dhfr mutations explained TMP resistance of every anaerobic mutant, it explained resistance for only 11 of the 47 aerobic mutants (Tables 2 and 3 and Table S2). Besides dhfr, four other genes were mutated in more than one clone. Genes spr1885 and spr1887, coding for two subunits of the PatAB ABC transporter (32), as well as gene spr1417, coding for a phosphoglucosamine mutase, were mutated in five clones, and the pur operon repressor purR (spr1793) was mutated in two clones (Table 3 and Table S2).
Each of the five mutations in patAB (Table 3) decreased susceptibility to TMP by 2-fold when introduced into an otherwise S. pneumoniae R6 WT background (Table 4). The two patA mutations (Q35* and Q128* in spr1887) and one patB mutation (ΔAA67A in spr1885) are expected to lead to a nonfunctional transporter (Table 4); thus, resistance to TMP seems to result from the loss of PatAB efflux pump activity. This is supported by the observation that the MIC for TMP of S. pneumoniae R6 WT increased by 2-fold in the presence of the efflux pump inhibitor reserpine (Table 4). The two other patB mutations in M11 and M23, which are coding but nonsynonymous, also increased TMP resistance by 2-fold (Table 4), suggesting that these interfere with the activity of the transporter. The five mutations also sensitized S. pneumoniae to ethidium bromide (Table 4), a known substrate of PatAB (32). The overexpression of patAB, but not its inactivation, was previously shown to influence the susceptibility of S. pneumoniae to ciprofloxacin, linezolid, and tetracycline (4, 32–35). We reconfirmed here the role of PatAB overexpression in resistance to ciprofloxacin, tetracycline, or ethidium bromide using S. pneumoniae R6 patAB overexpressors (33, 34) (Table 4), but we found no change regarding TMP susceptibility (Table 4). None of the patAB mutations associated with TMP resistance increased sensitivity to ciprofloxacin or tetracycline (Table 4).
MIC for TMP, ethidium bromide, ciprofloxacin, and tetracycline in the presence of patA and/or patB mutations
The gene spr1417 codes for a phosphoglucosamine mutase with five different mutations (Table 3). Despite several attempts, R421H is the only spr1417 mutation that could be introduced into S. pneumoniae R6 WT, leading to a 2-fold decrease in TMP susceptibility (Table 3). The spr1417 SNVs coding for the G47E, G109R, Q274*, and D340N mutations could be introduced in the genome and increased the MIC for TMP by 2-fold, but only if a mutation responsible for dhfr overexpression (e.g., G-64A) was first introduced (Table 3).
The two mutations in purR (spr1793), when introduced by DNA transformation in S. pneumoniae R6, increased the MIC for TMP by 2-fold (Table 3). These cells overexpressed specifically the pur operon (Fig. 4), as the expression of additional genes involved in purine metabolism remained unchanged (Fig. 4).
Mutations in purR (spr1793) increase expression of the pur gene cluster. (A) Schematic representation of the pur gene cluster in S. pneumoniae R6. Arrows represent the genes from the pur gene cluster. 1, purC; 2, purL; 3, purF; 4, purM; 5, purN; 6, vanZ; 7, purH; 8, purD; 9, hypothetical protein; 10, purE; 11, purK; 12, hypothetical protein; 13, purB. Genes of the pur operon studied by RT-qPCR are shown in black. Flanking comB and strH genes are shown as gray boxes. (B) RT-qPCR analysis showing increased expression for genes purN (5), purH (7), and purK (11), part of the pur gene cluster (black bars) in R6spr1793M18 and R6spr1793M19 transformants. The expression of two other genes involved in purine metabolism but located outside the pur gene cluster was also measured (dark gray bars): hgt (spr0011), hypoxanthine phosphoribosyltransferase (14), and pnpA (spr0516), polyribonucleotide nucleotidyltransferase (15). The expression of gyrA (16), recA (17), and rpoB (18) was measured as additional control genes (pale gray bars). All RT-qPCR data were normalized according to the amplification signals of the housekeeping gene era mRNA.
Lastly, a data set of 200 S. pneumoniae clinical isolates (118 TMP-resistant and 82 TMP-sensitive) was searched for polymorphisms in genes identified by our screens. Interestingly, three resistant isolates had single-nucleotide frameshift mutations at distinct positions in spr1887 (Fig. S4A), resulting in altered versions of PatA (Fig. S5A). A fourth resistant isolate also had a single-nucleotide deletion but this time in spr1885/PatB (Fig. S4B and S5B). Every TMP-sensitive isolate coded for full-length transporters, similar to S. pneumoniae R6 (Fig. S5). This suggests that mutations in the PatAB transporter are of clinical relevance.
DISCUSSION
We have adapted whole-genome screening approaches coupled to NGS with S. pneumoniae. These were used to screen for genes involved in the MOA or resistance to TMP, leading to the identification of novel genes that contribute to TMP resistance in addition to DHFR-related mutations. TMP is part of the key access list of essential medicines of the WHO, but its efficacy against S. pneumoniae infections is hampered by the high rates of resistance (36). TMP is recommended by the WHO as prophylactic medication for the young with HIV in Africa (37), where it can select for resistance among nasopharyngeal isolates. Extensive cotrimoxaxole (trimethoprim-sulfamethoxazole combination therapy) prophylaxis was indeed shown to select for widespread TMP resistance among S. pneumoniae isolates in Malawi (22).
The DHFR I100L mutation is the most prevalent in clinical isolates and was shown to confer resistance to 32 mg/liter TMP (21), the same level as that observed here (Table 3). Additional DHFR mutations have been described, most often in combination with I100L, but these either were not investigated or did not specifically segregate among resistant isolates (21–25, 27, 38), with the possible exception of the L31F and M53I susbtitutions (23). These mutations were found by our Mut-Seq screen, and we validated their role in TMP resistance (up to 8 mg/liter) (Table 3). Our screen revealed that mutation at four other positions (amino acids 9, 24, 28, and 98) can also lead to resistance. None of the other DHFR mutations previously reported in combination with I100L in clinical isolates were detected here, suggesting that these are bystanders or act as compensatory mutations, a phenomenon well described in drug-resistant pneumococci (39). DHFR mutations were indeed reported to influence the stability of the protein (40), with positions equivalent to L31 and I100 having the greatest impact on DHFR stability.
Target overexpression in TMP resistance has been described in S. pneumoniae (28). This is overlooked in surveillance studies, even if resistance levels frequently exceed those provided by the I100L mutation. Experimental evolution of TMP resistance in Escherichia coli found that regulatory mutations often precede CDS mutations (29). Regulatory mutations indeed predominate in our mutants. Because cells are resilient to DHFR inhibition, for instance, >97% inhibition only modestly affects the growth of E. coli (41), overexpression may suffice to ensure survival in the presence of TMP up to a certain concentration before CDS mutations are needed to decrease the affinity of the target. We have identified three novel SNVs located in a predicted Rho-independent transcription terminator upstream of the gene. Rho-independent transcription terminators consist of an RNA stem-loop structure that arrests transcription elongation, followed by a uracil-rich region that promotes polymerase dissociation (42). Disruption of the stem-loop results in readthrough transcription of the downstream gene. A similar phenomenon was reported in fluoroquinolone-resistant S. pneumoniae isolates, for which SNVs disrupting the stem-loop of a Rho-independent terminator located upstream of patAB were responsible for the overexpression of this multidrug efflux pump (43). While patAB overexpression is known to be associated with resistance to a number of antibiotics (4, 32–35), here we found that the opposite is true for TMP, and it is the loss of function of PatAB that correlates with resistance, a finding also observed with resistant clinical isolates. PatAB may have a physiological substrate in addition to drugs. While the putative substrate is unknown, it is possible that it alters TMP action and that increasing its intracellular concentration through patAB inactivation increases S. pneumoniae survival in the presence of TMP. The loss of the OmpA porin, known to alter antibiotic susceptibility through its interaction with efflux pump systems (44), similarly decreases susceptibility to TMP in the Gram-negative bacteria Acinetobacter nosocomialis (45) but not in Acinetobacter baumannii (44). Thus, the loss of drug efflux pump components may affect membrane protein composition, which in turn alters its structure or integrity, ultimately influencing antibiotic susceptibility. Antifolates were indeed shown to alter membrane properties in S. pneumoniae (46).
Our screens revealed that increasing metabolic flux from purine synthesis to GTP and then to folate increases the MIC for TMP. Purine biosynthesis was also shown to be the bottleneck in the growth of TMP-treated Bacillus subtilis (47). Mutations in the PurR repressor increased the expression of the entire pur operon, which should result into an overproduction of IMP from 5-phospho-α-d-ribosyl 1-pyrophosphate (PRPP). Within a few steps, IMP becomes GMP and then GTP. GTP then enters folate metabolism, where it is converted to dihydroneopterin by FolE and then to DHF in a series of reactions, the latter of which is performed by SulB. FolE and SulB were highlighted in our Int-Seq screen, and their coexpression was required for TMP resistance (Table 1). This suggests that FolE and SulB are involved in rate-limiting steps for the conversion of GTP to DHF in S. pneumoniae. Interestingly, a machine learning approach searching for associations between genome sequences and TMP resistance found folE and sulB to be predictive factors associated with TMP resistance (48).
Most mutations in phosphoglucosamine mutase (spr1417) were shown to increase the MIC for TMP only in a background of dhfr overexpression. Phosphoglucosamine mutase converts glucosamine-6-phosphate to glucosamine-1-phosphate toward UDP-N-acetylglucosamine for the synthesis of peptidoglycan. Identifying the link with resistance and the requirement for dhfr overexpresion, or of another coexpressed gene from the operon (Fig. 3C), will require further work. Intriguingly, phosphoglucosamine mutase was shown to interact with high confidence with dihydropteroate synthase, an enzyme involved in the production of folate precursors (49).
By sequentially transfecting back mutations recurring in the same genes, we could recontruct the MIC of the parent mutants for all anaerobic mutants and for half the aerobic ones. For example, the MIC of mutant 32 is 16 mg/liter, while the introduction of the DHFR W9C mutation only accounts for 2-fold resistance (i.e., 2 mg/liter). The remaining mutations involved in resistance are likely mutant specific. Despite the fact that our strategy to focus on recurring genes has paid off, we failed in demonstrating any role for TMP resistance, with some other genes mutated in at least 2 mutants, such as spr0011, spr0516, or spr2033, involved in purine metabolism (Table S3). Int-Seq and Mut-Seq each identified the target of TMP but also revealed additional genes linked to its MOA as well as novel resistance genes. Clinical levels of TMP resistance often exceed what is conferred by DHFR mutations alone, implying additional contributors, and further work with additional clinical isolates is warranted to truly assess the role of spr1885 and spr1887 in resistance to TMP in the clinic.
In principle, even without prior knowledge about the nature of the cellular target, the combination of Int-Seq and Mut-Seq screens has the potential to expedite a better understanding of novel antimicrobial molecules. TMP and S. pneumoniae experiments revealed DHFR as the most likely target. These also suggested harnessing purine metabolism or blockade of drug pumps as a possible means for potentiating TMP action.
MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions.Plasmids and bacterial strains are listed in Table S4 in the supplemental material. E. coli was grown at 37°C in LB broth. S. pneumoniae R6 was grown at 35°C with 5% CO2 in brain heart infusion (BHI) broth or on Trypticase soy agar with 5% sheep blood (TSAΙΙ; BD). S. pneumoniae R6 culture under anaerobic conditions was performed in an anaerobic chamber. MICs were determined by microdilution in 96-well plates according to CLSI guidelines in 0.1 ml cation-adjusted Müller-Hinton broth with 5% lysed sheep blood and from at least three independent biological replicates. Point mutations often conferred a 2-fold increase in MIC for TMP, i.e., within the experimental error of microdilution. However, these were consistent between replicate measurements and were further confirmed by the macrodilution technique, which was performed in triplicate according to the CLSI guidelines for at least one clone per gene. All drugs were purchased from Sigma-Aldrich.
Genomic and plasmid DNA purification and amplification.gDNA was extracted using a Wizard genomic DNA purification kit (Promega). Plasmid DNA was purified from E. coli using the GenElute plasmid miniprep kit (Sigma-Aldrich). PCRs were performed using primers described in Table S5.
DNA transformation.S. pneumoniae transformation was performed as previously described (50). Transformants were selected on casein tryptone (CAT) agar with 5% (vol/vol) sheep blood.
Int-Seq.Gene fcsR (spr1974) and the downstream fucose promoter (Pfcsk) were amplified from S. pneumoniae R6 gDNA and cloned into the pGEM-T Easy vector (Promega). The cat gene, coding for a chloramphenicol acetyltransferase and amplified from pEVP3 (51), was cloned downstream of Pfcsk along with an EcoRV restriction site. Finally, a PCR fragment covering the fcsK gene (spr1973) was cloned downstream of EcoRV.
S. pneumoniae R6 gDNA was randomly sheared by nebulization. Fragments of 2 to 5 kb were size selected on 1.3% agarose gel and end repaired using Klenow polymerase. Fragments were cloned in the EcoRV site of the cassette (Fig. 1). A total of 100,000 E. coli clones (mean insert size, 1.5 to 3 kb) were obtained, representing ∼100-fold genome coverage. Plasmids were purified from the pool of clones. The fucose library was linearized by NotI digestion (Fig. 1), transformed into S. pneumoniae R6, and selected on plates with 0.5% fucose and 4 mg/liter chloramphenicol. Clones were pooled and kept at −80°C. PCR primers at the 3′ end of cat (Int_Fw) and the 5′ end of fcsK (Int_Rv) (Table S5) were used for amplification of the inserts.
Mut-Seq.A single S. pneumoniae R6 colony was grown in 100 ml of BHI until an optical density at 600 nm (OD600) of 0.2, and then the culture was split into 10-ml cultures. Ethyl methanesulfonate (Sigma-Aldrich) was added at 8× or 16× its MIC. Cultures were incubated for 20 min at 35°C before 10 ml of cold BHI medium was added. An 800-μl aliquot was incubated in 7.8 ml of BHI for 3 h. Bacteria were harvested by centrifugation (10 min, 4,000 rpm), resuspended in 200 μl of 1× phosphate-buffered saline (PBS), and plated with 8 or 16 mg/liter TMP. Nonmutagenized 10-ml cultures were similarly processed and used as a control. No clones were obtained from these controls upon TMP selection.
Illumina sequencing.Illumina Nextera XT sequencing libraries were prepared from gDNA (Mut-Seq) or amplified inserts (Int-Seq) according to the manufacturer’s instructions. The size distribution of Nextera XT libraries was validated using a 2100 Bioanalyzer and high-sensitivity DNA chips (Agilent Technologies). Sequencing was performed using an Illumina HiSeq2500 system (101-nucleotide paired-end sequencing) at a final concentration of 8 pM. The NGS data have been deposited in the Sequence Read Archive (SRA) (Table S6).
Sequence reads were aligned to the S. pneumoniae R6 genome using the software bwa-mem (52). The maximum number of mismatches was 4, the seed length was 32, and 2 mismatches were allowed within the seed. Read duplicates were marked using Picard (http://broadinstitute.github.io/picard), and we applied GATK for single-nucleotide polymorphism and indel discovery (53). Several python and bash scripts were created to further analyze the data.
Reconstruction of resistance for Int-Seq and Mut-Seq.Genes of interest were cloned in the EcoRV site of the fucose cassette (Fig. 1). These cassettes were linearized by NotI, transformed in S. pneumoniae R6, selected with chloramphenicol at 4 mg/liter with 0.5% fucose, and tested for TMP resistance.
PCR fragments (5 kb) amplified from Mut-Seq mutants and covering the mutation of interest were transformed in S. pneumoniae R6 and selected using TMP (∼1× to 8× MIC). Mutations that allowed growth on higher TMP concentrations than the control were further validated by microdilution.
RNA extraction and RT-qPCR.RNA extraction and RT-qPCR were performed as described previously (31). All RT-qPCR data were normalized according to the amplification signals of era mRNA. The RT-qPCR primers are listed in Table S5.
5′‐RACE PCR.Total RNAs were extracted from S. pneumoniae grown to an OD600 of 0.3 in BHI medium, and the 5′ ends of dhfr and dpr mRNAs were determined using a 5′/3′ RACE kit (Roche Applied Science). Three dhfr (spr1429) antisense-specific primers (SP1, SP2, and SP3) and one dpr (spr1430) antisense-specific primer (SP4) were used (Table S5) for cDNA production. Reactions without reverse transcriptase were performed in parallel to control for DNA contamination.
Northern blot analysis.Fifteen μg of S. pneumoniae RNA was electrophoresed through a 1.2% agarose–0.66 M formaldehyde gel. Electrophoresed RNAs were transferred to a Hybond-N+ membrane (GE Healthcare Life Sciences) and hybridized to an spr1430 probe.
Data availability.The NGS data have been deposited in the SRA database under accession numbers SAMN09690270 to SAMN09690366 (Table S6).
ACKNOWLEDGMENTS
This work was supported by the Canadian Institutes of Health Research (grant number 81266 to M.O.). M.O. is the holder of a Tier 1 Canada Research Chair in Antimicrobial Resistance. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. We acknowledge the Bioimaging Platform of the Infectious Disease Research Centre, funded by an equipment and infrastructure grant from the Canadian Foundation for Innovation, for its assistance with fluorescence microscopy. We have no conflicts of interest to declare. Authors made the following contributions to the manuscript: designed the experiments, H.G., K.P., A.B., P.L., and M.O.; performed the experiments, H.G. and K.P.; analyzed the data, H.G., K.P., and P.L.; wrote the manuscript, P.L., K.P., and M.O.; provided critical comments and expertise, A.B.
FOOTNOTES
- Received 7 November 2018.
- Returned for modification 28 November 2018.
- Accepted 8 February 2019.
- Accepted manuscript posted online 19 February 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AAC.02381-18.
- Copyright © 2019 American Society for Microbiology.
