Uncovering the Important Genetic Factors for Growth during Cefotaxime-Gentamicin Combination Treatment in bla_CTX-M-1 Encoding Escherichia coli

Abstract

Due to the rapid spread of CTX-M type ESBLs, the rate of resistance to third-generation cephalosporin has increased among Gram-negative bacteria, especially in Escherichia coli, and there is a need to find ways to re-sensitize ESBL E. coli to cephalosporin treatment. A previous study showed that genes involved in protein synthesis were significantly up-regulated in the presence of subinhibitory concentration of cefotaxime (CTX) in a CTX-M-1-producing E. coli. In this study, the interaction between CTX and gentamicin (GEN), targeting protein synthesis, was evaluated in MG1655/pTF2, and the MIC of CTX was strongly reduced (128-fold) in the presence of this combnation therapy. Since the underlying mechanism behind this synergy is not known, we constructed a saturated transposon mutant library in MG1655/pTF2::bla_CTX-M-1 containing 315,925 unique transposon insertions to measure mutant depletion upon exposure to CTX, GEN, and combination treatment of CTX and GEN by Transposon Directed Insertion-site Sequencing (TraDIS). We identified 57 genes that were depleted (log₂FC ≤ −2 and with q.value ≤ 0.01) during exposure to CTX, 18 for GEN, and 31 for combination treatment of CTX and GEN. For validation, we deleted eight genes that were either uniquely identified in combination treatment, overlapped with monotherapy of GEN, or were shared between combination treatment and monotherapy with CTX and GEN. Of these genes, we found that the inactivation of dnaK, mnmA, rsgA, and ybeD increased the efficacy of both CTX and GEN treatment, the inactivation of cpxR and yafN increased the efficacy of only CTX, and the inactivation of mnmA, rsgA, and ybeD resulted in increased synergy between CTX and GEN. Thus, the study points to putative targets for helper drugs that can restore susceptibility to these important drugs, and it indicates that genes involved in protein synthesis are essential for the synergy between these two drugs. In summary, the study identified mutants that sensitize ESBL-producing E. coli to CTX and a combination of CTX and GEN, and it increased our understanding of the mechanism behind synergy between β-lactam and aminoglycoside drugs. This forms a framework for developing new strategies to combat infections caused by resistant bacteria.

1. Introduction

Drug-resistant infections account for the death of around 700,000 people globally every year, and this number is estimated to grow to more than 10 million and cost the global economy up to 100 trillion USD by 2050 if the problem is not solved [1]. The World health organization (WHO) has published a list of resistant bacteria, which should be prioritized when searching for novel antibiotics and Enterobacteriaceae with resistance to third generation cephalosporins are among the top priorities on this list [2].

β-lactam antibiotics are bactericidal agents that inhibit peptidoglycan synthesis by binding to Penicillin-binding proteins (PBPs), thereby preventing bacterial cell wall synthesis. Their high use has caused widespread resistance among Gram-negative bacteria. A particular concern is Extended Spectrum β-lactamase (ESBL) producing E. coli, which are capable of hydrolyzing the β-lactam ring in third and fourth generation cephalosporins [3]. ESBL genes are primarily carried on plasmids and include enzymes of the TEM, SHV, and CTX-M classes, and among these classes, the broadest dissemination has been detected for the CTX-M family [3,4,5]. To date, more than 190 variants of CTX-M-β-lactamases have been identified [6]. They are grouped into five gene families (CTX-M-1, CTX-M-2, CTX-M-8, CTX-M-9, and CTX-M-25) [7,8].

One of the possible strategies to overcome antimicrobial-resistant organisms is to use combination treatment, especially with drugs that have synergistic interaction [9,10]. Another possibility is to revitalize the existing antimicrobials by identifying helper drugs that can inhibit the resistant bacteria from growing at full capacity while expressing their resistance mechanisms [11]. A recent study characterized the global transcriptomic response in a CTX-M-1-producing E. coli growing in the presence of subinhibitory concentrations of the third-generation cephalosporin, cefotaxime (CTX) and found that a high number of genes, which were unrelated to the resistance mechanism, were significantly regulated [12]. The study demonstrated that genes involved in protein synthesis were enriched among the up-regulated genes, and in accordance with this, combination treatment with low concentrations of chloramphenicol (CHL), which inhibits protein synthesis, significantly reduced the MIC of CTX [12]. Since CHL is rarely used for systemic treatment due to its hematologic toxicity [13], the combination of CTX with aminoglycosides, such as gentamicin (GEN), is one of the most commonly prescribed combination treatments in clinical practice [14,15]. However, the molecular mechanism behind how these combination treatments work together remains surprisingly elusive.

There is no direct link between regulation and essentiality of genes, and thus it may be cumbersome to identify the most promising targets to re-sensitize ESBL bacteria to treatment from transcriptomic data. Jana et al. [11] demonstrated a more suited approach relying on libraries of random transposon mutants in their study of the secondary resistome (SR) of Klebsiella pneumoniae toward colistin, and in the current study, we have used this approach to investigate the SR of ESBL E. coli towards CTX. To increase the understanding of the synergistic effect between CXT and GEN, we further determined the SR to GEN treatment in the ESBL strain as well as the SR to the combination treatment. Thus, the aims of the study were to assign fitness values to all non-essential genes in the CTX-M-1-producing E. coli strain for the ability to grow in the presence of ½ MIC of CTX, GEN, and combination of CTX with GEN, and to investigate whether inactivation of genes with a high fitness effect would re-sensitise the bacteria to the drugs.

2. Materials and Methods

2.1. Bacterial Strains

Bacterial strains and plasmids used in this study are listed in Table S1. The main strain used was E. coli MG1655 carrying a bla_CTX-M-1 gene on the IncI1 plasmid pFT2 [16]. Strains were grown in Difco™ lysogeny broth (LB), Lennox (Becton, Dickinson, Albertslund, Denmark), and on LB agar plates (Becton, Dickinson, Albertslund, Denmark) overnight at 37 °C, except for the strains harboring the temperature-sensitive plasmid, pKD46, which were grown at 30 °C. Where needed, media were supplemented with antibiotics (Sigma, Copenhagen, Denmark), including 20 mg/L CTX, 50 mg/L kanamycin (Kan), 20 mg/L gentamicin (GEN).

2.2. Determination of MIC and Interaction between CTX and GEN

The minimum inhibitory concentrations (MICs) of CTX, GEN, and other generation cephalosporins, including cefazolin (CFZ), cefoxitin (FOX), ceftazidime (CAZ), cefepime (FEP), and ceftaroline (CPT) against the MG1655/pTF2 and its mutant-derivatives strains were determined by broth microdilution method following CLSI guidelines [17]. The interaction between CTX and GEN was determined by the chequerboard assay as described [18,19] with some modifications. Briefly, CTX and GEN (Sigma, Copenhagen, Denmark) were used in the ranges 0–512 and 0–4 mg/L, respectively. The first antibiotic of the combination was two-fold serially diluted in the horizontal line of the panel, while the second antibiotic was serially diluted in the vertical line of the panel. The inoculum for each strain was prepared by using a McFarland 0.5 standard (1–2 × 10⁸ CFU/mL) and a Sensititre Nephelometer (Thermo Scientific^TM, Roskilde, Denmark). All the wells containing 100 μL of Mueller–Hinton broth II (MHB-II) (Sigma, Copenhagen, Denmark) were inoculated with 100 μL of inoculum to a final cell density of around 5 × 10⁵ CFU/mL and incubated for 18–22 h. The fractional inhibitory concentration (FIC) and FIC index (FICI) were calculated using the formula: FIC of drug A = MIC of drug A in combination/MIC of drug A alone, FIC of drug B = MIC of drug B in combination/MIC of drug B alone, and FICI = FIC of drug A+ FIC of drug B [20]. The interpretation of synergism was defined as an FICI ≤ 0.5, additive or indifference was defined as an FICI > 0.5–1 or >1–4, respectively, and antagonism was defined as an FICI > 4 [21,22]. The values were obtained from triplicate biological replicates per condition.

2.3. TraDIS Library Construction and Validation

Electrocompetent cells of E. coli MG1655/pTF2 were made as previously described [23,24,25]. Briefly, the bacterial cells were grown overnight in LB medium, then 1:100 diluted in sterilized LB and grown to an OD₆₀₀ of 0.5–0.6. The cells were then centrifuged and washed in 1 × volume of cold Milli-Q water, 0.5 × volume of cold Milli-Q water, 0.05 × volume of 10% cold glycerol, respectively. After washing, the cells were resuspended in 0.002 × volume of 10% cold glycerol, and aliquots of 60 μL electrocompetent cells were stored at −80 °C until use. For electroporation, a 60 μL electrocompetent cells aliquot was mixed with 1 μL EZ-Tn5 < KAN-2 > Tnp Transposome (Epicentre) and electroporated in a 1 mm cuvette using a BioRad MicroPulser set to 1.8 kV. The mixture was immediately resuspended in 1 mL Super Optimal broth with Catabolites repression (SOC) medium and incubated at 37 °C for 2 h. Then, 100 μL of aliquots was spread on nine large LB agar plates, 15 cm, supplemented with 50 mg/L Kan. After incubation overnight at 37 °C, the kanamycin-resistant colonies from each plate were resuspended in 2–3 mL sterilized LB supplemented with glycerol 20% using a sterilized bacteriological spreader and transferred into 50 mL tube and stored at −80 °C. Each tube contained an estimated 10,000–50,000 mutants. The final input library contained 3 × 10⁵ mutants that were generated by pooling 13 mutant batches, resulting in a cell suspension of 1.3 × 10¹¹ CFU/mL.

2.4. TraDIS Library Antimicrobial Exposure and Sequencing

In duplicate, 100 μL of 1 mL aliquots of the input library, containing approximately 2 × 10⁹ mutants, were diluted into four falcon tubes containing 9.9 mL MHB-II, 9.9 mL MHB-II supplemented with 128 mg/L CTX, 9.9 mL, MHB-II supplemented with 0.5 mg/L GEN, and 9.9 mL MHB-II supplemented with 1 mg/L CTX plus 0.25 mg/L GEN, corresponding to ½ MIC of each treatment (Figure 1). After 24 h of incubation at 37 °C, 1 mL from each culture (output library) was centrifuged at 12000 rpm for 2 min, and the resulting pellet was used for DNA isolation using GenElute™ Bacterial Genomic DNA kit (Sigma-Aldrich, Soeborg, Denmark) following manufacturer’s instructions. In addition, 500 μL from the 1 mL aliquot of the input library, containing approximately 2 × 10⁹ mutants, was also used for DNA extraction. The quality of DNA was evaluated by NanoDrop (Thermo Fisher Scientific, Roskilde, Denmark). A ratio of OD260/OD280 between 1.8 and 2.0 and a ratio of OD260/OD230 between 2.0 and 2.2 were set as criteria for samples to be used for sequencing (Table S2). The DNA concentration was measured using Qubit dsDNA HS Assay Kit (Thermo Fisher) (Table S2). For TraDIS sequencing, 2–4 μg of DNA was fragmented using Covaris M220 (Covaris, Woburn, MA, USA) into approximately 300 bp fragments, and each library was prepared for sequencing following the protocols previously described [25,26], using primers (Table S3) specific for PCR enrichment of the DNA fragments tagged with transposon. PCR amplified fragment libraries were then pooled and sequenced on a MiSeq instrument using a MiSeq reagent kit V2 (50 cycles) (Illumina, San Diego, CA, USA) following the recipe previously described [26].

2.5. Analysis of TraDIS Data

The analysis of sequences containing the transposon was performed as described [26] using the Bio::TraDIS pipeline (https://github.com/sanger-pathogens/Bio-Tradis, accessed on 26 June 2022) with some modifications previously described [25]. Briefly, the processing of raw reads was performed using fq2bam.pl script, which relocates the tag from the read name to the front of the read and converts the files to SAM format. Then, the generated SAM files were converted to BAM format by using samtools. The generated BAM format files were then used for analyses with check_tradis_tags and add_tradis_tags scripts. Afterwards, the output BAM files were converted to FASTQ format using samtools. Next, the bacteria_tradis mapping pipeline was used, and the obtained FASTQ files were filtered and trimmed for ten bases matching the 3′ end of the transposon. The generated files were then mapped against the MG1655 reference genome (GenBank:LR881938) using SMALT short-read mapper (https://www.sanger.ac.uk/tool/smalt-0/, accessed on 26 June 2022). After mapping, the precise transposon insertion site was assessed, and unique insertion sites (UISs) and read counts were determined per gene. Artemis version 18.1.0 [27] was used to visualize the UISs and read counts, and DNAPlotter version 18.1.0 [28] was used to create the circular genome diagram. Subsequent analysis steps were done using R scripts (tradis_essentiality.R and tradis_comparisons.R) included in the Bio::Tradis pipeline to assess the log₂ fold change (log₂FC) of read counts and q.value for each gene between test and control samples. To identify the important genes for survival in presence of antimicrobials, the significant genes were defined as genes with a log₂FC ≤ −2 and q.value ≤ 0.01. The complete output from R scripts showing the log₂FC and q.value of each gene appears in Table S4. The TraDIS raw reads of this study were deposited on the European Nucleotide Archive (ENA) under accession number (PRJEB52919). STRING (https://www.string-db.org, accessed on 30 June 2022) analysis was used to detect interactions between genes where mutation caused ≤ −2 log₂FC between the challenge and control condition and to highlight the enrichment of KEGG terms and Gene Ontology (GO) terms based on these genes.

2.6. Deletion of Selected Genes in MG1655/pTF2

Mutants in MG1655/pTF2 were constructed by Lambda Red recombination system using the plasmids pKD46 and pKD4, essentially following the protocols described [29,30]. Correct insertion of the antibiotic gene cassettes in the target genes was confirmed by PCR. Primers used for gene knockout and for PCR control of gene insertion are listed in Table S5.

2.7. Growth Experiments

Bacterial growth experiments were performed in two biological replicates with two technical replicates for each strain on Bioscreen C (Thermo Labsystems, Helsinki, Finland) for 24 h at 37 °C. Briefly, 250 μL of MHB-II was inoculated with cells from LB agar plates to a final cell density of approximately 10⁶ CFU/mL, using a Sensititre™ Nephelometer (Thermo Fisher Scientific, Roskilde, Denmark) with a 0.5 McFarland standard (1–2 × 10⁸ CFU/mL). The cultures were then grown without antibiotic or supplemented with either 128 mg/L CTX, 0.5 mg/L GEN, or a combination of CTX and GEN of 1 + 0.25 mg/L. The OD₆₀₀ was measured every 20 min with continuous shaking, and the growth curves were obtained using GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA).

2.8. Morphology of Bacteria

The WT strains MG1655/pTF2 and mutant strains (ΔcpxR, ΔdnaK, ΔmnmA, ΔrsgA, ΔyafN, and ΔybeD), which were identified to reduce the MIC of CTX, were inoculated into LB. After overnight growth at 37 °C, the strains were inoculated into sterilized LB to a starting OD₆₀₀ of 0.05 and further incubated at 37 °C with continuous shaking to reach an OD₆₀₀ of 0.3–0.4. To minimize the overnight culture residues and ensure that most of the cells were in early exponential phase, the strains were re-inoculated into new LB to a starting OD₆₀₀ of 0.05, followed by incubation at 37 °C with continuous shaking to reach an OD₆₀₀ of 0.3. CTX was then added to the cultures to a final concentration of 128 mg/L, and the treated and not-treated cultures were incubated for an additional hour. A volume of 900 μL of each culture was mixed with 100 μL of 37% formalin, resulting in cells fixed with 3.7% formalin. Cells were then washed two times with 1 × DPBS and then 1:10 diluted in 1 × DPBS. A volume of 50 μL was spread on Nunc™ Glass Bottom Dishes (Thermo Scientific™) coated with poly-L-lysin solution (Sigma, Copenhagen, Denmark). Cells were observed by differential interference contrast (DIC) using Zeiss LSM 780 laser scanning confocal microscope equipped with AxioCam camera. Bacteria segmentation was performed by training a custom Cellpose [31] model using cyto2 as a base model. The custom model was trained on 12 images corrected manually to segment correctly bacteria of various sizes and bending. The predictions were then exported into text files and converted into ROI in Fiji to measure the Feret diameter and extract the bacteria length. The experiment was repeated two times per condition, and no intravariability was noted during the analysis.

2.9. Time-Kill Assays

The activity of CTX against WT and its mutant-derivatives strains, which were identified to increase susceptibility to CTX, was evaluated by measuring the reduction of CFU/mL in the presence of antibiotic. Briefly, 10 mL MHB-II containing 10⁶ CFU/mL for each strain was prepared as described above, and 128 mg/L CTX was added to each tube and incubated at 37 °C for 24 h with continuous shaking. The CFU/mL was determined after several time points (4, 8, and 24 h). Two independent biological replicates were performed, and the time-kill curves for WT against each mutant strain were depicted using GraphPad Prism 9.

2.10. RNA Extraction and RT-qPCR

Single colonies of the WT and mutant strains that enhance the efficacy of CTX were grown separately in MHB-II with continuous shaking overnight at 37 °C. The cultures were then diluted to a starting OD₆₀₀ of 0.05 and grown with different concentrations of CTX, corresponding to ½ MIC of the strains, to OD₆₀₀ of 0.5–0.6. Total RNA for each sample was extracted by mechanical disruption with a FastPrep-24 homogenizer (MP Biomedicals, Solon, OH, USA) and RNeasy Mini Kit (Qiagen, Sollentuna, Sweden). The quality and quantity of extracted RNA was evaluated by NanoDrop. 600 ng of the RNA samples were purified by DNA digestion using TURBO™ DNase (2 U/μL) (Ambion^®, Naerum, Denmark), and the treated RNA was reverse transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies, Naerum, Denmark). Reverse-transcribed-quantitative real time polymerase chain reaction (RT-qPCR) was performed using FastStart Essential DNA Green Master (Roche, Hvidovre, Denmark) and a LightCycler 96 (Roche, Hvidovre, Denmark). The expression of bla_CTX-M-1 in mutant strains was calculated compared to expression in the WT strain in the presence of ½ MIC CTX. The expression data were normalized to a validated reference gene gapA [16]. The results were calculated by 2^−ΔΔCT method [32]. Two independent biological replicates were performed using two technical replicates. Primers are listed in Table S5.

2.11. In Silico Homology Study

To evaluate whether the targets that increased susceptibility to CTX upon deletion share similarity to host proteome, the protein sequences of CpxR, YafN, DnaK, MnmA, RsgA, and YbeD were obtained from Artemis, and the sequence homology to human proteins was searched using NCBI Homo sapiens Protein BLAST (BLASTp) (http://blast.ncbi.nlm.nih.gov/Blast.cgi, accessed on 16 February 2023). The proteins that showed no hits at the E-value cutoff of 10 × 10⁻¹⁰ were considered non-homologous proteins [33].

2.12. Statistical Analysis

The Kruskal–Wallis rank sum test followed by Dunn’s test was used to recognize significant changes in bacterial cell length of mutants compared to the WT (control) in the absence and presence of CTX. Statistical analysis for time-kill assay was performed by comparing the mean log₁₀ CFU/mL of WT and mutant strains using multiple t-tests. For qPCR, the significant changes in fold change of bla_CTX-M-1 in the WT to the mutant strains in the presence of ½ MIC CTX was evaluated using one-way ANOVA. A p-value < 0.05 was considered significant.

3. Results

3.1. Validation of Synergy between CTX and GEN

The MIC of CTX and GEN for MG1655/pTF2 were 256 and 1 mg/L, respectively. The synergistic interaction between CTX and GEN was identified by chequerboard assays, and the MIC of CTX was found to be reduced significantly from 256 to 2 mg/L (128-fold reduction) in the presence of 0.5 mg/L GEN with an FICI of 0.5 (Figure S1). Therefore, a combination of 1 mg/L CTX and 0.25 mg/L (1/2 of the synergy concentration) was used as treatment concentrations for the transposon library to elucidate the mechanism of synergy.

3.2. Generation of a High-Density Mutant Library and Sequencing Using TraDIS Protocol

A high-density Tn5 mutant input library was constructed in MG1655/pTF2, and the library of Kan^R mutants was exposed to three antimicrobial selection conditions (Figure 1). The reads of input and output libraries were mapped to K-12 MG1655 reference genome, with the percentage of mapped reads ranging from 84.38% to 92.03% (

Table 1. Mapping of Tn5 insertions to K-12 MG1655 reference genome.

Library	Total Reads	Reads Mapped (%) 1	Total UIS 2	Total Seq Len/Total UIS
Input
MG1655_pTF2_input_1	10,795,077	90.79	234,898	19.69
MG1655_pTF2_input_2	11,001,585	85.73	279,308	16.56
Input 1 + Input 2 (combined)	21,796,662	88.23	315,925	14.64
Output
MG1655_pTF2_without_antibiotic_1	11,238,753	89.03	229,617	20.14
MG1655_pTF2_without_antibiotic_2	12,879,322	89.20	233,968	19.77
MG1655_pTF2_128mg_CTX_1	11,456,775	88.42	219,223	21.10
MG1655_pTF2_128mg_CTX_2	9,862,543	85.08	227,102	20.37
MG1655_pTF2_0.5mg_GEN_1	11,632,252	84.38	224,341	20.61
MG1655_pTF2_0.5mg_GEN_2	9,398,550	89.74	205,101	22.55
MG1655_pTF2_CTX+GEN_1	12,080,082	92.03	194,803	23.74
MG1655_pTF2_CTX+GEN_2	9,614,653	89.80	194,408	23.79

¹ Percentage of mapped sequence reads against K-12 MG1655 reference genome. ² Unique Insertion Sites.

). When merging the duplicates of the input library, the saturation of Tn5 insertions across the MG1655 genome consisted of 315,925 unique transposon insertions (Figure S2) with an average of one Tn5 insertion every 14.6 base pairs (Table 1). Furthermore, 367 genes were identified to be essential for growth on LB agar plates supplemented with kanamycin, judged from the absence of insertions in these genes in the input library (Table S6).

3.3. Identification of Genes Relevant for Growth with and without Antibiotics

The comparison of Tn5 insertions between the input library and the untreated library (control) resulted in the identification of 103 genes that were important for growth in MHB-II (Table S4), and comparison between the untreated library (control) and antibiotic-treated libraries identified 57, 18, and 31 genes which were important for growth with CTX, GEN, and CTX+GEN combination treatments, respectively (Figure 2, Table S4). Of the 31 genes identified for the combination treatment, 6 and 14 genes overlapped with genes significantly affected in the libraries exposed to monotherapy of CTX and GEN, respectively, and one gene (cpxR) was part of the SR in all three libraries. Thus, of the 31 genes identified as SR to combination treatment, only ten genes were unique to the library exposed to the combination treatment.

3.4. The Secondary Resistome to CTX

The top 20 genes, according to Log₂FC after CTX treatment, including genes involved in many different cell systems (

Table 2. Top 20 genes according to fitness defect in the SR to CTX in E. coli MG1655/pTF2.

Gene	Function	128 mg/L CTX
		Log2FC	q.Value
lpoB	penicillin-binding protein activator LpoB	−10.10	2.31 × 10−17
gatB	PTS galactitol transporter subunit IIB	−8.85	1.40 × 10−11
minE	cell division topological specificity factor MinE	−7.99	3.06 × 10−8
tolQ	Tol-Pal system protein TolQ	−7.33	6.65 × 10−23
pal	peptidoglycan-associated lipoprotein Pal	−6.96	1.63 × 10−22
tolA	cell envelope integrity protein TolA	−6.92	8.42 × 10−19
tolB	Tol-Pal system protein TolB	−6.62	1.08 × 10−23
mrcB	bifunctional glycosyl transferase/transpeptidase	−6.28	1.40 × 10−80
dicC	dicB transcriptional regulator DicC	−5.88	0.00303
tolR	colicin uptake protein TolR	−5.38	1.04 × 10−27
rlmH	23S rRNA (pseudouridine(1915)-N(3))-methyltransferase RlmH	−4.89	7.91 × 10−05
sdhC	succinate dehydrogenase cytochrome b556 subunit	−4.86	2.88 × 10−10
ldcA	muramoyltetrapeptide carboxypeptidase	−4.83	9.77 × 10−14
sucD	succinate--CoA ligase subunit alpha	−4.68	5.13 × 10−22
plsX	phosphate acyltransferase PlsX	−4.48	4.72 × 10−50
dapB	4-hydroxy-tetrahydrodipicolinate reductase	−4.30	0.0005
dacA	D-alanyl-D-alanine carboxypeptidase DacA	−4.14	2.22 × 10−23
rfaD/hldD	ADP-glyceromanno-heptose 6-epimerase	−4.13	1.31 × 10−5
rsmH	16S rRNA (cytosine(1402)-N(4))-methyltransferase RsmH	−3.91	2.52 × 10−6
sltY	murein transglycosylase	−3.88	5.43 × 10−103

Genes where growth was significantly affected by transposon insertions (Log₂FC ≤ −2 and q.value ≤ 0.01) in the presence of 128 mg/L CTX. The differences in transposon insertion abundance compared to the control (without antibiotic) are shown as Log₂FC. The false discovery risk is indicated by the q.value. The full list of genes identified as SR to CTX is provided in Table S4.

) and insertion of transposons into the gene lpoB, which is involved in peptidoglycan synthesis [34,35], caused the highest fitness defect (Log₂FC = −10.10). To further understand the cellular responses, STRING analysis was performed based on genes showing Log₂FC ≤ −2 and q.value ≤ 0.01. A graphic presentation of the interactions between the genes is given in Figure S3. The significantly affected genes corresponded to the enrichment of two KEGG pathways, the TCA cycle and carbon metabolism. Analysis of GO terms showed that 12 GO terms were significantly enriched in the biological process category, with cell cycle, cell division and bacteriocin transport having the highest scores. No terms were enriched in the category Molecular Functions, while 14 terms were significantly enriched in the category Cellular Components. The top enriched terms were cell division site, cellular anatomical entity, and cell envelope (Table S7).

3.5. The Secondary Resistome to GEN

The SR genes to GEN treatment in the CTX-resistant, GEN-sensitive MG1655/pTF2 are listed in

Table 3. Top 20 genes according to fitness defect in the SR to GEN in E. coli MG1655/pTF2.

Gene	Function	0.5 mg/L GEN
		Log2FC	q.Value
atpF	F0F1 ATP synthase subunit B	−8.30	2.65 × 10−16
atpG	F0F1 ATP synthase subunit gamma	−6.49	1.19 × 10−23
ybeD	YbeD family protein	−5.60	0.005
atpD	F0F1 ATP synthase subunit beta	−5.55	6.69 × 10−41
atpH	F0F1 ATP synthase subunit delta	−5.50	0.009
plsC	1-acylglycerol-3-phosphate O-acyltransferase	−5.48	0.010
atpC	F0F1 ATP synthase subunit epsilon	−4.87	5.94 × 10−11
atpA	F0F1 ATP synthase subunit alpha	−4.06	1.70 × 10−48
atpB	F0F1 ATP synthase subunit A	−3.99	0.0003
yajC	preprotein translocase subunit YajC	−3.84	5.57 × 10−8
gmhB	D-glycero-beta-D-manno-heptose 1,7-bisphosphate 7-phosphatase	−3.22	5.11 × 10−6
sdaC	HAAAP family serine/threonine permease SdaC	−3.03	8.89 × 10−148
hflC	protease modulator HflC	−2.86	2.14 × 10−155
ackA	acetate kinase	−2.79	2.53 × 10−9
cpxR	envelope stress response regulator transcription factor CpxR	−2.66	1.40 × 10−134
galU	UTP--glucose-1-phosphate uridylyltransferase GalU	−2.57	6.36 × 10−16
hflK	FtsH protease activity modulator HflK	−2.55	8.42 × 10−133
pta	phosphate acetyltransferase	−2.02	5.07 × 10−16

Genes where growth was significantly affected by transposon insertions (Log₂FC ≤ −2 and q.value ≤ 0.01) in the presence of 0.5 mg/L GEN. The differences in transposon insertion abundance compared to the control (without antibiotic) are shown as Log₂FC. The false discovery risk is indicated by the q.value.

. Genes belonging to the F₁F₀-ATP synthase subunits [36], such as atpF, atpG, atpD, atpH, atpC, atpA, and atpB, were among the highest attenuated genes following exposure to GEN treatment. STRING analysis was used to analyze interactions between the genes in the SR, and the encoded functions were elucidated by the enrichment of KEGG pathway and GO terms. A graphical illustration of the interactions between genes is shown in Figure S4. The significantly affected genes corresponded to the enrichment of three KEGG pathways, the oxidative phosphorylation pathway, metabolic pathways, and taurine and hypotaurine metabolism. Analysis of GO terms showed that 13 GO terms were enriched in the biological process category, among which purine ribonucleotide biosynthetic process, nucleobase-containing small molecule metabolic process, and carbohydrate derivative biosynthetic process showed the highest significant enrichment values. Proton-transporting ATP synthase activity, proton-transporting ATPase activity, inorganic molecular entity transmembrane transporter activity, and ion transmembrane transporter activity were the enriched GO terms for Molecular functions. Ten GO terms related to Cellular components were enriched after the treatment of MG1655/pTF2 with 0.5 mg/L GEN, with the proton-transporting ATP synthase complex, membrane protein complex, and protein-containing complex having the highest significant scores (Table S8).

3.6. The Secondary Resistome to CTX-GEN Combination Treatment

To elucidate the genes and, therefore, indirectly the mechanism(s) associated with the synergy between CTX and GEN, the transposon input library was grown in the presence of 1 mg/L CTX and 0.25 mg/L GEN, which corresponded to ½ MIC detected under synergy conditions according to the chequerboard assay results (Figure S1). Based on the TraDIS analysis, 10 genes out of the 31 genes identified in the combination treatment were unique to this condition, while one gene (cpxR), as mentioned, was shared between all libraries, and the remaining genes were shared with the CTX or GEN output libraries (

Table 4. The genes significantly depleted during combination treatment of E. coli MG1655/pTF2 with CTX and GEN, including illustration of the overlapping SR genes between CTX or GEN monotreatments and the combination treatment.

Gene	Function	1 + 0.25 mg/L CTX+GEN		128 mg/L CTX		0.5 mg/L GEN
		Log2FC	q.Value	Log2FC	q.Value	Log2FC	q.Value
atpD	F0F1 ATP synthase subunit beta	−10.12	1.90 × 10−40	0.11	0.87	−5.55	6.69 × 10−41
atpG	F0F1 ATP synthase subunit gamma	−8.97	5.70 × 10−23	−0.27	0.74	−6.49	1.19 × 10−23
atpF	F0F1 ATP synthase subunit B	−8.36	1.25 × 10−15	1.21	0.04	−8.30	2.65 × 10−16
atpC	F0F1 ATP synthase subunit epsilon	−8.21	1.54 × 10−12	0.05	0.98	−4.87	5.94 × 10−11
yajC	preprotein translocase subunit YajC	−7.71	1.22 × 10−10	1.48	0.01	−3.84	5.57 × 10−8
atpB	F0F1 ATP synthase subunit A	−6.47	5.39 × 10−5	2.58	7.83 × 10−5	−3.99	0.0003
mnmA	tRNA 2-thiouridine(34) synthase MnmA	−6.04	0.0009	1.98	0.012	−0.14	0.94
rsgA	small ribosomal subunit biogenesis GTPase RsgA	−5.86	0.002	1.23	0.15	0.24	0.94
ybeD	YbeD family protein	−5.65	0.006	0.83	0.41	−5.60	0.005
atpH	F0F1 ATP synthase subunit delta	−5.55	0.006	2.49	0.01	−5.50	0.009
IEU92_RS07640	Pseudogene	−5.54	0.010	3.31	0.0003	−0.33	0.81
ackA	acetate kinase	−3.98	3.69 × 10−11	5.49	2.02 × 10−98	−2.79	2.53 × 10-9
tolA	cell envelope integrity protein TolA	−3.90	9.54 × 10−33	−6.92	8.42 × 10−19	0.2	0.48
tolB	Tol-Pal system protein TolB	−3.75	1.30 × 10−36	−6.62	1.08 × 10−23	0.13	0.59
sdaC	HAAAP family serine/threonine permease SdaC	−3.73	2.00 × 10−138	0.03	0.88	−3.03	8.89 × 10−148
pal	peptidoglycan-associated lipoprotein Pal	−3.66	3.35 × 10−27	−6.96	1.63 × 10−22	0.27	0.2
atpA	F0F1 ATP synthase subunit alpha	−3.64	1.47 × 10−15	0.77	0.08	−4.06	1.70 × 10−48
IEU92_RS06395	septation protein A	−3.49	0.010	1.91	0.04	−1.05	0.28
yafN	type I toxin-antitoxin system antitoxin YafN	−3.44	1.94 × 10−5	1.06	0.14	−0.42	0.56
hflC	protease modulator HflC	−3.43	1.78 × 10−102	−0.61	0.0008	−2.86	2.14 × 10−155
tolQ	Tol-Pal system protein TolQ	−3.36	3.31 × 10−35	−7.33	6.65 × 10−23	0.1	0.65
cpxR	envelope stress response regulator transcription factor	−3.33	1.14 × 10−177	−2.24	3.62×10−20	−2.66	1.40 × 10−134
ydeI	YdeI family stress tolerance OB fold protein	−3.12	3.90 × 10−5	−0.86	0.33	−0.58	0.37
tusD	sulfurtransferase complex subunit TusD	−3.05	0.0005	3.75	3.03 × 10−10	−1.28	0.06
rsmH	16S rRNA (cytosine(1402)-N(4))-methyltransferase	−3.01	8.78 × 10−9	−3.91	2.52 × 10−6	−1.62	0.0002
dnaK	molecular chaperone DnaK	−2.95	8.72 × 10−5	3.41	1.98 × 10−9	0.98	0.02
tolR	colicin uptake protein TolR	−2.89	2.37 × 10−53	−5.38	1.04 × 10−27	0.29	0.04
hflK	FtsH protease activity modulator HflK	−2.88	5.39 × 10−173	−0.08	0.62	−2.55	8.42 × 10−133
pflA	pyruvate formate lyase 1-activating protein	−2.36	2.19 × 10−57	1.33	7.91 × 10−12	−1.94	9.74 × 10−47
epmA	elongation factor P--(R)-beta-lysine ligase	−2.18	1.88 × 10−8	1.17	0.0002	−1.47	2.69 × 10−13
pta	phosphate acetyltransferase	−2.17	1.39 × 10−12	4.2	1.70 × 10−99	−2.02	5.07 × 10−16

Genes were significantly affected by transposon insertions (Log₂FC ≤ −2 and q.value ≤ 0.01) when the input library was incubated in the presence of 1 mg/L CTX with 0.25 mg/L GEN. The differences in transposon insertion abundance compared to the control (without antibiotic) are shown as log₂FC. The false discovery risk is indicated by the q.value. The genes shown in bold were selected for validation by site-specific knockout followed by characterization of growth in the presence of the combination treatment. The genes highlighted in green were exclusively identified as fitness genes when the input library was exposed to the combination treatment, the genes in blue and red were also significantly affected when the input library was exposed to CTX or GEN monotreatment, respectively, the genes highlighted in grey is fitness genes common to all conditions tested.

). As for GEN treatment, most of the fitness genes during exposure to combination treatment belonged to the ATP synthase subunits. However, the common genes showed a higher fold change in the combination treatment than in the GEN treatment despite the lower concentrations of the two drugs when combined, except for the atpA gene (Table 4).

The graphical illustration of the interactions between genes in the SR to the combination treatment is indicated in Figure S5. The STRING analysis showed enrichment of only one KEGG pathway, the oxidative phosphorylation, which was also found for the GEN treatment. Fourteen GO terms for biological processes were enriched, with purine ribonucleotide biosynthetic process, bacteriocin transport, and protein import showing the highest significant scores. In addition to the four GO terms for Molecular functions that were enriched in GEN treatment, ligase activity was uniquely found as a term in the presence of combination treatment. Eight GO terms for Cellular components were enriched, of which six terms were also enriched after treatment with GEN (e.g., ATP synthase complex, membrane protein complex). In addition, the cell division site term was listed when the strain was treated with CTX alone. The term protein-containing complex was found to be enriched in all three libraries. GO terms such as cellular response to bacteriocin and organic substance, transport, tRNA wobble position uridine thiolation, and ligase activity were uniquely enriched in the combination treatment. The list of all terms enriched in combination treatment is shown in Table S9.

3.7. Validation of Genes Identified by TraDIS

Since one of the aims of the current study was to understand the synergistic effect of CTX and GEN in MG1655/pTF2, a total of eight fitness genes that were either uniquely identified in the library exposed to combination treatment, overlapped between the libraries exposed to GEN and combination treatment, or which were shared between all three libraries were individually knocked-out by site-directed mutagenesis to investigate their role for growth in the absence and presence of three antibiotic-related conditions. The functional classifications of these genes were transcription regulation (cpxR, yafN) [37,38], stress response (ydeI and dnaK) [39,40], translation, ribosomal structure and biogenesis (mnmA and rsgA) [41,42,43], protein transport (yajC) [44], and hypothetical protein (ybeD) [45].

The growth of mutant strains compared to the WT strain without the presence of antibiotics is shown in Figure 3. The analyses showed that the deletions of cpxR, yafN, yajC, and ydeI did not cause growth effects compared to the WT strain in the absence of antimicrobials. In contrast, the deletions of dnaK, mnmA, rsgA, and ybeD resulted in a slower growth of the mutants compared to the WT strain, mainly caused by slightly prolonging the lag phase.

In the presence of 128 mg/L CTX, only a slight growth defect was seen in the cpxR mutant compared to WT strain, whereas remarkable growth defects were found in dnaK, rsgA, ybeD, mnmA, and yafN mutant strains, respectively. No notable growth defects were found in yajC and ydeI mutant strains compared to the WT (Figure 4).

In the presence of 0.5 mg/L GEN, a minor growth defect was also seen in cpxR mutant strain compared to the WT. No growths were detected after the deletion of dnaK, mnmA, rsgA, and ybeD during 24 h. Surprisingly, the yafN and ydeI mutant strains were found to grow faster than the WT in the presence of 0.5 mg/L GEN. No apparent difference was observed in yajC mutant compared to the WT strain (Figure 4).

In the presence of the combination treatment of 1 mg/L CTX plus 0.25 mg/L GEN, the mutant strains cpxR, dnaK, and yajC were found to grow slightly slower than the WT strain. On the contrary, the deletions in yafN and ydeI were growing faster than the WT. No growths were identified in mutant strains with deletion of mnmA, rsgA, and ybeD in the presence of combination treatment (Figure 4).

Based on the results above, chequerboard assays were performed with mutants with deletion of mnmA, rsgA, and ybeD to determine whether these mutations were associated with increased synergy between CTX and GEN compared to the WT strain. Based on the chequerboard assays, the mnmA and rsgA mutant strains were found to be fully sensitive to CTX according to CLSI breakpoints [46], with a MIC of 0.5 mg/L in the presence of only 0.25 mg/L GEN with an FICI of 0.53. For the ybeD mutant strain, the strain was also susceptible to CTX with a MIC of 1 mg/L in the presence of only 0.125 mg/L GEN with an FICI of 0.56. Since these FICIs values were between 0.51–1, they were considered additive effects. However, synergy (FICI ≤ 0.5) was found between 4 mg/L CTX and 0.125 mg/L GEN, 2 mg/L CTX and 0.125 mg/L GEN, and 4 mg/L CTX and 0.06 mg/L GEN for mnmA, rsgA and ybeD mutant strains, respectively. Chequerboard assays with all FICIs are shown in Figure S6.

3.8. Increased Susceptibility to Antimicrobials

To investigate whether growth attenuation would lead to increased susceptibility of the mutant strains to treatment, MICs for mutant strains were obtained using broth microdilutions (

Table 5. MICs of CTX, GEN, and other generation cephalosporins in the mutant strains.

Strain	CTX (mg/L)	GEN (mg/L)	CFZ (mg/L)	FOX (mg/L)	CAZ (mg/L)	FEP (mg/L)	CPT (mg/L)
MG1655/pTF2 (WT)	256	1	1024	4–8	2	2	1024
ΔcpxR	32	0.5–1	512	4	2	2	1024
ΔdnaK	8	0.5	32	0.5	0.25	1	256
ΔmnmA	16	0.5	256	2	0.25	0.25	512
ΔrgsA	8	0.5	128	0.5	0.125	0.06	512
ΔyafN	32	1–2	512	2	2	2	1024
ΔyajC	128–256	0.5–1	-	-	-	-	-
ΔydeI	256	1–2	-	-	-	-	-
ΔybeD	16	0.25	512	2	0.25	0.5	1024

Values in bold represent a reduction in MIC compared to WT MG1655/pTF2. -: represents mutants that do not change the MIC to CTX.

). The mutations of cpxR, dnaK, mnmA, rsgA, yafN, and ybeD were found to decrease the MIC to CTX with 8-fold, 32-fold, 16-fold, 32-fold, 8-fold, and 16-fold, respectively, compared to the WT. In addition, the mutant strains of dnaK, mnmA, rsgA, and ybeD also showed increased susceptibility to GEN with 2-fold or 4-fold reduction of the MIC compared to the WT.

Since the deletions of cpxR, dnaK, mnmA, rsgA, yafN, and ybeD were found to reduce the MIC to CTX compared to the WT strain, we extended the analysis for these mutants by testing the MICs of five additional cephalosporins: CFZ (first generation), FOX (second generation), CAZ (third generation), FEP (fourth generation), and CPT (fifth generation). Mutations of dnaK, mnmA, and rsgA resulted in increased susceptibility to all five antibiotics (Table 5). Mutations of cpxR, yafN, and ybeD exhibited increased susceptibility to at least one cephalosporin antibiotic: Mutation of cpxR led to increased susceptibility to CFZ, mutation of yafN led to increased susceptibility to CFZ and FOX, and mutation of ybeD led to increased susceptibility to CFZ, FOX, CAZ, and FEP.

3.9. Determination of Filament Formation in the Mutant Strains

Many β-lactam antibiotics, including CTX, induce filament formation in susceptible Gram-negative bacteria [47,48]. This filamentation depends on the antibiotic concentration [47,49]. Therefore, a fixed concentration of CTX, which corresponds to ½ MIC of the WT, was used for WT strain and cpxR, dnaK, mnmA, rsgA, yafN, and ybeD mutant strains were identified to restore susceptibility to CTX. These strains were then evaluated for morphological changes using fixed-cell imaging. To ensure that the morphology changes were caused by the exposure to CTX, the microscopic appearance of the same strains was evaluated in the absence of CTX.

The microscopic analysis showed no obvious morphological changes in the mutant strains compared to the WT strain in the absence of CTX (Figure 5). In statistical terms, however, the average cell lengths for all the mutants, except ΔybeD, were significantly higher compared to the WT, with the difference in average length being between 0.16–1.79 µm (Figure 6). On the contrary, longer filament formations were clearly seen in the mutants compared to the WT in the presence of 128 mg/L CTX (Figure 5), and the average cell length for all mutants was significantly larger than the WT in the presence of CTX with increases ranging from 3.2 to 11.9 µm (Figure 6).

3.10. CTX Efficacy for Mutants Showing Reduced MIC to CTX

The reduction of MIC to CTX associated with mutations of cpxR, dnaK, mnmA, rsgA, yafN, and ybeD could be caused by either an increased efficacy of CTX or by a prolongation of the lag phase in the presence of CTX, which could limit their potential as helper drug targets. Therefore, a time-kill kinetics assays were used to evaluate the activity of CTX against these mutants and the WT in the presence of 128 mg/L CTX, which corresponds to ½ MIC of the WT (Figure 7). The analysis showed that the CFU/mL of all these mutants was significantly reduced compared to WT strain within 24 h in the presence of 128 mg/L CTX. In addition, strains with mutation of dnaK and rsgA were completely killed within 8 h and 24 h, respectively, in the presence of CTX.

3.11. bla_CTX-M-1 mRNA Levels in the Mutants Identified to Increase Efficacy of CTX

To investigate whether the mutant strains that were found to revitalize CTX affected the expression of bla_CTX-M-1 in the presence of CTX, we exposed each mutant strain to CTX at a concentration of ½ MIC and measured the relative expression ratio of bla_CTX-M-1 for the WT strain in the presence of ½ MIC. The expression analysis showed that the bla_CTX-M-1 expression in all the mutant strains, except in mutations of rsgA and yafN, were significantly down-regulated relative to the WT strain in the presence of CTX. The most significant down-regulation of bla_CTX-M-1 was found in mutant strain with deletion of dnaK, ybeD, mnmA, and cpxR, respectively (Figure 8). Considering the filament formations observed in these mutants upon the exposure to 128 mg/L CTX (Figure 5), the comparisons of CFU/mL between the WT and mutant strains in the presence of different CTX concentrations, which correspond to their ½ MIC, at an OD₆₀₀ of 0.5 were calculated to make sure that roughly the same number of cells were compared in the RT-qPCR (Figure S7). The analysis showed no significant change in CFU/mL between WT and mutant strains. Therefore, this suggests that the numbers of cells used in the qPCR were comparable.

3.12. Homology to Human Proteome

To evaluate if the identified targets could be a promising candidate for helper drug development, the protein sequence of CpxR, YafN, RsgA, YbeD, DnaK, and MnmA were compared to human proteome using NCBI-BLAST tools to minimize unwanted side effects on the host. In silico analysis did not identify any homology of CpxR, YafN, RsgA, and YbeD from MG1655/pTF2 to human protein, whereas DnaK and MnmA were found to share homology to human 70 kilodalton (kDa) heat shock proteins (HSPs) and mitochondrial tRNA-specific 2-thiouridylase 1, respectively (Table S10).

4. Discussion

Although combination treatments have been used as a strategy to fight infections caused by resistant bacteria, the understanding of molecular mechanisms of the enhanced effect between antibiotics remains elusive. Since protein synthesis genes are up-regulated in the presence of CTX treatment [12], we previously hypothesized that a combination treatment with CTX and CHL would result in a synergistic effect, and indeed this was confirmed experimentally [12]. Due to the serious side effects of treatment with CHL, such as aplastic anemia [13], in the current study, we went on to evaluate the interaction between CTX and GEN, which inhibits protein synthesis by binding to the 30S ribosomal subunit instead of the 50S subunit as CHL does. This antibiotic was chosen instead of CHL because combination treatment between ß-lactams and aminoglycosides is well known and due to the fact that GEN is more relevant in human medicine [50,51,52,53]. Our results showed that the MIC of CTX could be reduced from 256 to 2 mg/L by the addition of just 0.5 mg/L GEN in E. coli MG1655 expressing bla_CTX-M-1. Therefore, a high-throughput TraDIS screening approach was used to identify genes required for growth in the presence of CTX and GEN, and to increase our understanding of why there is synergy between these two drugs in this ESBL strain. The genes shown to be important for growth in the presence of CTX (n = 57), GEN (n = 18), and combination treatment (n = 31) constitute the SR in this strain towards these treatments, according to the definition of the SR given by Jana et al. [11].

Previously, global gene expression by RNA-seq was used to study responses in the same strain to treatment with CTX. This revealed that genes that were important for peptidoglycan biosynthesis, TCA and respiratory chain were among the most enriched functional classes [12]. In the current study, analysis of genes where insertion of transposon(s) caused significant growth attenuation confirmed the importance of the TCA cycle, and it showed that carbon metabolism was enriched in the KEGG analysis of these genes. When GO terms were analyzed, results also revealed the expected importance of peptidoglycan-based cell wall synthesis, while the protein synthesis apparatus per se was not suggested by the transposon analysis as important for growth in the presence of CTX. The large overlap between transcriptomic and genomic approaches gives high confidence in the results. The genomic approach used here is a more direct indicator of fitness effects, and we were able to add genes involved in cell division, bacteriocin transport, response to radiation, and outer membrane to the list of significantly enriched factors. With respect to TCA, the analysis showed that this was related to the succinate-CoA ligase and the succinate dehydrogenase complex. Previously, a similar approach was used to identify fitness genes for growth in the presence of CTX in the bla_CMY-23 encoding uropathogenic E. coli strain EC958 of ST131. In that study, GO terms, such as cell wall organization, cell wall biogenesis, response to radiation, cell division site, outer membrane-bounded periplasmic space, peptidoglycan-based cell wall, and TAT protein transport complex, were enriched after CTX treatment [54]. These terms were also enriched in the current study, suggesting a large overlap in the SR to CTX treatment, despite different classes of ESBL genes being responsible for the resistance.

The KEGG analysis of SR genes to GEN treatment showed the enrichment of oxidative phosphorylation, metabolic pathways, and taurine and hypotaurine metabolism. Moreover, the GO analysis revealed the importance of different terms related to energy metabolism, such as the F₁F₀-ATP synthase complex, membrane protein complex, and protein-containing complex. These findings were in agreement with previous studies showing that ATP synthase mutants were more susceptible to aminoglycosides (e.g., GEN) than WT strains in E. coli and Staphylococcus aureus [55,56]. Furthermore, the knock-out of the gene encoding the inner membrane protein HflK, which was identified in this study as an SR gene to GEN, was also found previously to restore the susceptibility to GEN in E. coli [57]. Therefore, we inferred that the SR to GEN identified in this study was highly precise. For SR to combination treatment of CTX and GEN, only the oxidative phosphorylation pathway was enriched in the KEGG analysis, which we suggest is related to the GEN treatment. According to the GO terms analysis, the majority of enriched terms overlapped with terms enriched after CTX or GEN monotherapy treatment. However, GO term such as tRNA wobble position uridine thiolation, which is important for translation efficiency [58], was among the terms exclusively enriched in the combination treatment. These data add to our understanding of the importance of genes involved in protein synthesis for the synergy between the CTX and GEN.

To validate whether the identified genes in the SR to combination and single drug treatment affected the susceptibility to CTX or GEN, eight genes that were either uniquely identified in the combination treatment (mnmA, rsgA, dnaK yafN, and ydeI), overlapped with GEN treatment (yajC and ybeD), or identified as common SR genes to CTX, GEN, and combination treatment (cpxR) were individually deleted in MG1655/pTF2. These genes were selected because they have not previously been reported to restore susceptibility to CTX or GEN upon deletion. Four mutants (mnmA, rsgA, dnaK, and ybeD) were found to have reduced MIC to both CTX and GEN compared to WT strain, two mutants (yafN and cpxR) had reduced MIC to CTX, and the remaining two mutants (ydeI and yajC) did not show reduced MIC to any tested antibiotic. Therefore, mutants with a deletion in mnmA, rsgA, dnaK, yafN, ybeD, and cpxR were further studied to evaluate their ability to increase the efficacy of CTX.

Because CTX binds to FtsI (PBP3), which is an essential protein for cell division [59], it causes filament formation in sensitive strains [47,48]. Thus, we used fixed-cell imaging to evaluate the filament formation for the mutant strains (dnaK, yafN, cpxR, mnmA, rsgA, and ybeD) in the absence and presence of 128 mg/L CTX. As expected, only minor variations in the cell length were found between the mutant strains and the WT strain when no CTX was present. However, in the presence of CTX, apparent filament formations with variations in the cell length between the mutant strains and the WT strain were identified. Whether this includes reduced binding of CTX to FtsI or reduced levels of CTX in the cell cytosol is unknown, and further studies should focus on this.

In agreement with the microscopic analyses, the time-kill kinetics assays showed that these mutations also caused significant reduction in CFU/mL compared to WT strain in the the presence of 128 CTX. Consequently, we infer that by these mutations, we were able to increase the efficacy of CTX in the strains. As this could be attributed to either a direct effect on the bla_CTX-M-1 by lowering its expression, or an indirect effect such as a disruption in the cell wall synthesis, we went on to determine whether there was an effect on the expression of the resistance. This showed that expression of bla_CTX-M-1 was reduced in the dnaK, ybeD, mnmA, and cpxR mutant strains compared to the WT strain. Therefore, the re-sensitization towards CTX associated with these mutant strains might be linked to a direct effect on the bla_CTX-M-1 expression, allowing a higher build-up of drug in the cell and hence the higher effect on cell wall synthesis, as illustrated by the microscopy analysis.

Some of the mutations that changed the susceptibility to CTX caused slower growth phenotype in the absence of antibiotic (dnaK, mnmA, rsgA, and ybeD) while others did not have any growth effect (cpxR and yafN). Using the latter proteins as helper drug targets is preferrable since this would not have an effect on the growth of the normal E. coli population. Most of the genes are well characterized, and based on current knowledge and the results observed in the study, it is possible to suggest likely mechanisms for their effect on CTX and CTX+GEN susceptibility (see Supplementary discussion S1). However, no proof of such mechanisms was performed in relation to the current study. Studies of gene-expression in the presence of the challenges in both WT strain and mutant strains would possibly enlarge our understanding of the reasons behind the observed phenotypes, and multiple methodological approaches are available for such studies [60].

Combining CTX with low concentrations of a protein-synthesis-inhibitor-antibiotic is a known principle. The synergy between such drugs was previously hypothesized to be caused by the effect of protein-synthesis-inhibitor-antibiotics (in that study CHL) on protein synthesis [12]. In the current study, we provide support for this hypothesis by identifying a high number of genes involved in protein-synthesis in the SR to combination treatment with CTX and GEN. The mutations of these genes, such as mnmA and rsgA, which are involved in translation, ribosomal structure and biogenesis, were found not only to increase the efficacy of CTX and GEN but also to increase the synergy between CTX and GEN, i.e., causing synergy at lower doses. Mutation of the gene encoding the hypothetical protein YbeD also increased the synergy between CTX and GEN. However, for this protein, the reason for its effect is unknown.

In summary, our results show that the added effect of combining CTX treatment with tre the aminoglycoside GEN is most likely linked to the effect of GEN on the protein synthesis, as we found that the interference with genes involved in protein synthesis restored the susceptibility to CTX. The study also revealed that several genes in the SR to CTX and to CTX combination treatment are promising targets for helper drugs to restore susceptibility to CTX and, for some of them, also other cephalosporin drugs.