Peptides were synthesized by an automated peptide synthesizer liberty blue (CEM, Matthews, NC, United States) on rink amide 0.68 mmol/mg MBHA resin, using the Fmoc solid phase strategy. The resin-bound peptide was washed thoroughly with dimethylformamide and then dichloromethane, dried, and cleaved. Lipid groups were conjugated at this stage. Cleavage was done by addition of 95% trifluoroacetic acid (TFA), 2.5% water, and 2.5% triethylsilane. The crude peptides were purified by reverse-phase high-performance liquid chromatography. Purification of the peptides was done using a C4 or a C8 column (Grace Discovery Sciences, Columbia, MD, United States) in acetonitrile in water gradient (both containing 0.1% TFA) for 40 min.

All S. typhimurium mutant strains used were derived from the ATCC 14028s background. Generally, cells were grown overnight in LB and before every experiment were regrown by a dilution of 1:100 into fresh sterile LB medium and incubated at 37°C with agitation for 2 h to reach a mid-log phase (OD_{600 nm} = 0.5–0.7), unless mentioned differently.

The experimental selection assay was carried out in 96-well plates at a final volume of 100 μL as previously described (Lofton et al., 2013) with several alterations. The 4DK5L7 peptide was dissolved in double distilled water (DDW) to the appropriate concentration, and 5 μL of the peptide was added to 45 μL of Mueller-Hinton broth (MHB) in a flat-bottom 96-well plate (four replicates). Five microliters of DDW was used only for the wild type (WT) as a negative control. Fifty microliters of WT S. typhimurium culture (approximately 10⁵ bacteria/mL) in MHB was added to the peptide-containing 96-well plate. The plate was incubated at 37°C with shaking for approximately 24 h. Every day, 10 μL of bacteria from each lineage was transferred to 90 μL of a fresh medium containing peptides, in a new 96-well plate. After transfer, 10 μL of 70% glycerol was added to each culture in the old plate, and the 96-well plates were kept at −80°C. If the bacteria in a well did not grow after transfer, the lineage was restarted from the most recent passage that showed successful growth. When a lineage survived three subsequent transfers, the peptide concentration was increased by 50% for approximately 30 passages. RL1 was isolated from the well in which the maximal concentration of a peptide was achieved. Bacteria were spread on LB agar plate, and a single colony was collected. The colony was grown in liquid LB for three passages without a peptide, and then the minimal inhibitory concentration (MIC) was determined.

Mid-log cultures (OD₆₀₀ = 0.5–0.7) were adjusted to final OD₆₀₀ of 0.005 in MHB. Peptides or antibiotics were dissolved in DDW and added to the MHB for the appropriate final concentration. Plates were incubated for 18 h at 37°C with agitation. Inhibition of growth was determined by absorbance measurements at 600 nm using a microplate autoreader. Alternatively, absorbance at 600 nm was measured at 37°C with agitation, every 20 min for 18 h in Cytation 5 plate reader (BioTek, Winooski, VT, United States) to evaluate the growth curves. Growth rates were calculated using the R package “growth rates” according to Kahm et al. (2010). Data were filtered for growth (OD₆₀₀ > 0.2), and growth rates for data sets not reaching that threshold were set to zero.

DNA was purified from the control strain and RL1 overnight cultures using the DNeasy Blood & Tissue Kits (Qiagen, Germany). Sequencing was performed on an Illumina MiSeq platform, as previously described (Blecher-Gonen et al., 2013) with the following modifications: 300 to 600 ng of gDNA was sheared using the Covaris E220X sonicator (Covaris, Inc., Woburn, MA, United States). End repair was performed in 80-μL reaction at 20°C for 30 min, followed by Agencourt AmPURE XP beads cleanup (Beckman Coulter, Inc., Indianapolis, IN, United States) in a ratio of 0.75 × beads/DNA volume. A-bases were added to both 3′ ends, followed by adapter ligation in a final concentration of 0.125 μM. An SPRI bead cleanup in a ratio of 0.75 × beads/DNA volume was performed, followed by eight polymerase chain reaction (PCR) cycles using 2 × KAPA HiFi ready mix (Kappa Biosystems, Inc., Wilmington, MA, United States) in a total volume of 25 μL with the following program: 2 min at 98°C, eight cycles of 20 s at 98°C, 30 s at 55°C, and 60 s at 72°C followed by 72°C at 10 min. Sequencing QC was performed using fastqc website. Reads were mapped to S. typhimurium strain 14028s genome version ASM2216v1 using BWA-MEM v0.7.12-r1039 (Li and Durbin, 2009), sorted using SAMtools v1.2 website. Polymerase chain reaction duplicate reads were identified and marked using Picard MarkDuplicates. GATK HaplotypeCaller v3.5 (McKenna et al., 2010) was used to discover single-nucleotide polymorphisms (SNPs) and insertions/deletions, using the setting-ploidy 10 in order to increase sensitivity to subclonal variants. The resulting variants were hard filtered using standard GATK filters for base quality, strand bias, and so on. Variant allele fractions for each variant and each sample were computed from the allelic depths field in the VCF. Variants were annotated for predicted effect using snpEFF (Cingolani et al., 2012).

RNA was purified from cells at the mid-log stage. Cells were centrifuged at 4,000 revolutions/min for 10 min. The pellet was resuspended with 150 μL of fresh 2 mg/mL lysozyme (Sigma, Germany), Tris 1M, EDTA 0.5M, pH 8.0. Following incubation of 3 to 5 min at 37°C, 1 mL of Tri-reagent (Sigma, Germany) was added. The mixture was vortexed for 10 s at maximum speed and then incubated at room temperature for 5 min. After addition of 200 μL chloroform, the mixture was vortexed until homogeneity. The mixture was incubated at room temperature for 5 min until phase separation has appeared and subsequently centrifuged at 12,000 g for 15 min at 4°C. The upper phase was collected into a new tube, and 500 μL of isopropanol was added. The tube was inverted gently and incubated at room temperature for 10 min and then at −20°C for 30 min. The mixture was centrifuged at maximum speed for 25 min. The pellet was washed twice with 70% ethanol and left to dry. The pellet was resuspended in 30 μL of DDW and incubated at 55°C for 10 min until completely dissolved. One microgram of the RNA was reverse transcribed using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, United States), dNTPs, and random hexamers. Quantitative reverse transcriptase (qRT)–PCR was performed in 7300 Real time PCR system (Applied Biosystems) with Platinum SYBR Green qPCR Super Mix-UDG with ROX (Invitrogen). The qRT-PCR was performed in duplicates, three biological replicates in one plate. The values were normalized to the gyrB gene.

Point mutations and HA epitope tagging in the bacterial genome were performed using the pWRG730 and pWRG717 constructs, as described previously (Hoffmann et al., 2017). Briefly, the ISceI + kanamycin cassette was amplified from the pWRG717 plasmid with primers containing homology parts to the target gene (see primers information in Supplementary Table S1). Salmonella typhimurium carrying the pWRG730 were grown to an OD₆₀₀ = 0.4–0.5, and λ red recombinase expression was induced by incubation at 42°C for 12.5 min and then transferred to ice for 15 min. The cells were prepared for electroporation by three washes in 10% ice-cold glycerol. The pellet was resuspended in 100 μL of 10% glycerol, and 50 to 100 ng of the PCR product was added. The cells were electroporated and were recovered in LB at 30°C for 1 h. Then, they were selected on kanamycin plates, and positive clones for homologous recombination were chosen. A 300-bp fragment was amplified from RL1 genome, designed to contain the point mutation or the epitope tag. Expression of the λ red recombinase was induced in the positive clones as described above. Next, the clones were electroporated with the 300-bp fragment containing the point mutation or the HA epitope tag. Cells were selected on chloramphenicol plates containing 100 ng/mL anhydrotetracycline. Positive clones were verified by colony PCR and sequencing.

Cells were grown to mid-log stage, and OD_{600 nm} was checked. For OD_{600 nm} of less than 0.5, 1 mL of culture was collected; for 0.5 < OD_{600 nm} < 1, 0.5 mL of culture was collected, and for OD_{600 nm} > 1, 0.25 mL was collected. Cells were centrifuged at 6,000 g for 10 min, at 4°C. The supernatant was removed, and the pellet was resuspended in 1 × sample buffer (Tivan-Biotech, Kfar Sava, Israel) according to the following formula: sample buffer vol (mL) = 0.5 × vol sample × OD_{600 nm}. The samples were heated to 95°C for 10 min, cooled down on ice, and were loaded on 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis gel, 110 V. Proteins were transferred to a polyvinylidene difluoride membrane using wet transfer overnight at 30 V. The blots were blocked for 1 h in TBST + 5% milk, washed with TBST, and incubated overnight at 4°C with an anti-HA antibody (C29F4) rabbit Mab #3724 (Cell Signaling Technology, Danvers, MA, United States). The blots were washed three times with TBST and incubated for 1 h with goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (Jackson Immuno Research Laboratories, Inc., West Grove, PA, United States) The treated blots were developed on an enhanced chemiluminescence detection system with Luminata^TM Crescendo Western HRP substrate (Merck Millipore, Burlington, MA, United States). For the housekeeping gene, the blots were incubated with anti GroEL antibody produced in rabbit (Sigma). Quantification of bands intensity was performed in ImageJ software¹.

Cells were grown to mid-log phase, centrifuged at 4,000 g for 5 min, and pellet was frozen on dried ice. The samples were subjected to lysis and in solution tryptic digestion using the S-Trap method [by ProtiFi (Huntington, New York, NY, United States)]. The resulting peptides were analyzed using nanoflow liquid chromatography (nanoAcquity) coupled to high-resolution, high-mass-accuracy mass spectrometry (Q-Exactive HFX). Each sample was analyzed on the instrument separately in a random order in discovery mode. Raw data were processed with MaxQuant (Cox and Mann, 2008). The data were searched with the Andromeda search engine against the S. typhimurium 14028s proteome database appended with common laboratory protein contaminants and the following modifications: fixed modification—cysteine carbamidomethylation; variable modifications—methionine oxidation, asparagine, and glutamine deamidation. For the single-gene analysis, the quantitative comparisons were calculated using Perseus (Tyanova et al., 2016). Decoy hits were filtered out. Only proteins that had at least two valid values in at least one experimental group were kept. Pathway enrichment analyses were done on the basis of the current Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation of the S. typhimurium ATCC 14028s genome (“seo”) using the package “clusterProfiler” (Yu et al., 2012) in “R” v3.51².

RL1 was isolated in our laboratory using experimental evolution. The WT S. typhimurium, with a MIC of 12.5 μM (Table 1) were exposed to a sub-inhibitory concentration (0.5 × MIC) of the 4DK5L7 peptide. Following three successful passages, we increased the concentration by 50%. In total, 30 passages were conducted, and the maximal concentration in which bacteria exhibited growth was 160 μM (12.8 × MIC) (Table 1). As a control, we passaged the WT bacteria with DDW at the equivalent volume to the peptide for the same amount of passages. From the last passage, we isolated a single colony, which was then grown in the absence of the peptide for 3 days. The isolate that was passaged with 4DK5L7 was named RL1, and the isolate that was passaged with DDW instead of the peptide and will be used in our further experiments was named “control strain.” Next, we tested the MIC of the control strain and RL1 for the 4DK5L7 peptide and to other AMPs: seg5D, C8-K5L7, and colistin (Table 2). We found that RL1 exhibited a high resistance level toward 4DK5L7, as the MIC was 200 μM compared to 12.5 μM in the control strain (Table 1). Moreover, RL1 showed cross-resistance toward another AMP, seg5D, a synthetic AMP that was designed in our laboratory (Papo et al., 2002). RL1 showed a MIC of 50 μM compared to 6.25 to 12.5 μM in the control strain (Table 1). RL1 showed hyposensitivity toward the cyclic lipopeptide colistin of the polymyxin family and to the synthetic lipopeptide C8-K5L7. RL1’s MIC to colistin was detected as 6.25 versus 1.56 to 3.125 μM in the control strain and to C8-K5L7 6.25 to 12.5 versus 6.25 μM in the control strain (Table 1).

In order to investigate the mutations that were acquired in RL1, DNA was purified from the control strain and RL1 bacteria, and whole genome sequencing was performed. While no mutations occurred in the control strain, RL1 carried five mutations in five loci: hlpA (skp), acrB, ramR, yeiU, and rfaY (Table 3). hlpA (histone-like protein A) encodes the Skp chaperone that facilitates the translocation of membrane proteins from the inner to the outer membrane, preventing protein aggregation in the periplasmic space (Schafer et al., 1999). RL1 carried a deletion mutation in position 265,722, resulting in a frameshift that eventually led to a premature stop codon within hlpA (Table 3). The acrB gene encodes AcrB, the substrate-binding domain of the AcrAB-TolC system, a multidrug efflux pump. This system is mostly known to protect bacteria from bile salts, dyes, and antibiotics such chloramphenicol and tetracycline (Adhya et al., 1995; Yu et al., 2003). There is evidence that several AMPs are substrates of the system. Two studies reported that deletion of this system increased K. pneumoniae and E. coli susceptibility to AMPs, respectively (Padilla et al., 2009; Warner and Levy, 2010). The AcrB mutation in RL1 is located at arginine 620, which was replaced by serine (R620S). Based on previous structural studies, this amino acid is located within a switch loop that allows the conformational change of AcrB monomers and is important for substrate access and binding (Eicher et al., 2012). We compared the conservation of the switch loop sequence among ∼1,000 Gram-negative bacterial strains, which include ∼240 different species. We found that, generally, this domain is highly conserved (Figure 1A). The R/Q620 position is well-conserved and might indicate a functional importance (Figure 1A). ramR encodes a transcriptional regulator which activity was previously described in S. typhimurium LT2. Accordingly, RamR represses the transcription of ramA, an activator of the acrA and acrB expression (Abouzeed et al., 2008). In RL1, the mutation is located at position 639,370, which led to substitution of threonine 19 with proline (T19P). This amino acid is located in the alpha helix α1, a part of the protein DNA-binding domain (Yamasaki et al., 2013). Because proline is known to disrupt alpha helical structures (Li et al., 1996), it could potentially affect RamR’s ability to bind the ramA promoter. Quantification of ramR transcripts indicated a 7.8-fold increase in RL1 compared to the control strain (Figure 1B). Yet, the acrB transcript levels were higher in RL1 than in the control strain (Figure 1B), implying that the repressive activity of RamR might be impaired in RL1.

The gene products of yeiU and rfaY are involved in LPS biosynthesis. yeiU encodes the inner membrane protein LpxT, which transfers a phosphate group to lipid A, forming the 1-diphosphate species (Touze et al., 2008). RL1 carries an SNP mutation in yeiU, which resulted in a stop codon (Table 3). Inactivation of this protein was shown to confer resistance to AMPs due to reduction of the net negative surface charge (Herrera et al., 2010; Kato et al., 2012). rfaY encodes WaaY, which is involved in phosphorylation of the HepII of the LPS core (Yethon et al., 1998). RL1 carries a deletion mutation in this locus, which resulted in a frameshift effect (Table 3). A frameshift mutation in this gene was previously reported to evolve in experimental evolution that induced AMP resistance. It was suggested that the reduction in the LPS net negative charge weakens the interaction with positively charged AMPs as a mechanism of resistance (Lofton et al., 2013). Because mutations with similar impact both in yeiU and rfaY were already reported to confer resistance to AMPs, we focused on the mutations in acrB, ramR, and skp and their possible contribution to AMP resistance.

To investigate whether the mutations in AcrB and RamR contribute to resistance toward the 4DK5L7 peptide, we generated mutants carrying RL1’s point mutations in acrB (acrB mut), in ramR (ramR mut), and both of these mutations (double mutant) (Table 3). Next, we compared the growth of the control strain, RL1, acrB mut, double mutant, and ramR mut in the presence of 4DK5L7 at the MIC of the control strain (12.5 μM). As expected, RL1 showed a remarkable optimal growth, whereas the control strain did not exhibit growth (Figure 2A). The double mutant, having a MIC twofold higher than the control strain (Table 1), showed a moderate growth, which was significantly higher than the control strain but lower than RL1 (Figure 2A). The acrB mut and the ramR mut bacteria displayed a similar MIC to the control strain (Table 1), and no significant differences between their growth curves were observed (Figure 2A). As a negative control, bacteria were grown without peptide and showed no significant differences in growth (Figure 2B). Furthermore, we checked whether the mutations contribute to resistance against seg5D, a peptide that RL1 showed cross-resistance toward (Table 1). The MIC of the acrB mut, ramR mut, and the double mutant was similar to that of the control strain (Table 1). Accordingly, at the MIC of the control strain (12.5 μM), RL1 exhibited an optimal growth, whereas none of the mutants grew (Figure 2C). In the negative control, the strains showed similar growth, whereas RL1 reached a slightly lower OD₆₀₀ value (Figure 2D). Based on the growth curves, we calculated the growth rates of the strains in the presence of these AMPs. We found that at 12.5 μM of 4DK5L7 the double mutant and the acrB mut had similar low growth rates. However, the double mutant was significantly different from the control strain, whereas the acrB mut was not (Figure 2E). In addition, RL1’s growth rate was only slightly reduced at 12.5 μM of 4DK5L7 (Figure 2E) and 12.5 μM of seg5D (Figure 2F) compared to the growth rate in the absence of AMPs.

Next, we aimed to validate the protein expression levels of the mutated AcrB and to test the impact of the mutation in ramR on the protein level. For this purpose, an epitope tag was added to the C-terminal end of the protein (Table 1). The addition of an HA epitope tag to the AcrB protein did not affect the MIC toward 4DK5L7 among the mutants and the control strain (Table 1 and Supplementary Figure S1). Using Western blot analysis, we found that the acrB mut and the control strain bacteria had similar AcrB-HA protein levels (Figure 2G). Bacteria carrying the point mutation in the ramR gene (RL1, double mutant, and ramR mut) showed a significant higher levels of AcrB than the control strain (a fold change of 2.2–2.9) (Figure 2G). This fold change is similar to the elevation in the acrB transcripts that were measured in RL1 (Figure 1B).

Because the AcrAB-TolC system was mostly studied in the context of resistance to antibiotics, we evaluated the impact of RL1’s mutations on the MIC of chloramphenicol and ampicillin compared to the control strain. We found that the mutation in ramR increased the MIC of chloramphenicol and ampicillin, whereas the mutation in acrB reduced the MIC of ampicillin (Table 4). More specifically, at 5.8 μM (1 × MIC of the control strain), no growth was observed for the control strain, RL1 and acrB mut. This is in contrast to the double mutant and the ramR mut (Figure 3A). At 2.9 μM (0.5 × MIC of the control strain), RL1 and the acrB mut were hyposensitive to chloramphenicol compared to the other strains (Figure 3B). At 1.45 μM (0.25 × MIC of the control strain), all the strains grew, while RL1 was yet hyposensitive to chloramphenicol than the others (Figure 3C). No significant differences in growth were observed in the absence of chloramphenicol (Figure 3D). In the case of ampicillin, we found more significant impact of the mutation in acrB on the MIC values (Table 4). At 33.6 μM (1 × MIC of the control strain), ramR mut was the only strain that exhibited growth (Figure 3E). At 16.8 μM (0.5 × MIC of the control strain), the control strain, double mutant, and ramR mut exhibited growth, whereas RL1 and acrB mut did not (Figure 3F). At 8.4 μM (0.25 × MIC of the control strain), all the strains showed growth, whereas RL1 and acrB mut were slightly more susceptible (Figure 3G). In the absence of ampicillin, all strains grew similarly, whereas the control strain exhibited a minor advantage (Figure 3H).

To investigate whether the mutation in skp contributes to resistance to AMPs, we introduced this point mutation in the control strain background (skp mut) and in the double-mutant background (triple mutant) (Table 1). We found that the MIC to 4DK5L7 was increased by twofold in the skp mut and by fourfold in the triple mutant. A comparison between the growth curves of the different strains revealed that at 25 μM (2 × MIC of the control strain) RL1 showed an optimal growth, and the triple mutant showed a suboptimal growth (Figure 4A). The other bacterial strains did not exhibit any growth at this concentration (Figure 4A). At 12.5 μM (1 × MIC of the control strain), RL1 and the triple mutant showed a similar optimal growth (Figure 4B). The skp mut and the double mutant showed a similar moderate growth, whereas the control strain did not grow (Figure 4B). At 6.25 μM (0.5 × MIC of the control strain) and without peptide, no significant differences were observed between the five strains (Figures 4C,D, respectively). The growth rates of RL1 and the triple mutants were similar at the different peptide concentrations (Figure 4E). The skp mut and the double mutant exhibited similar growth rates that were lower compared to RL1 and the triple mutant at 12.5 μM (Figure 4E). Next, we tested the impact of the mutation in skp on the MIC of the seg5D peptide. We found that the MIC of the skp mut and the triple mutant increased by fourfold (Table 3). Comparison between the growth curves of the strains demonstrated that the skp mut and the triple mutant showed a moderate growth at 25 μM (2 × MIC of the control strain) (Figure 5A). RL1 showed an optimal growth, whereas the control strain and the double mutant did not grow (Figure 5A). At 12.5 μM (1 × MIC of the control strain) RL1, skp mut and the triple mutant showed similar optimal growth, whereas the control strain and the double mutant did not exhibit growth (Figure 5B). At 6.25 μM (0.5 × MIC of the control strain) RL1, skp mut and the triple mutant reached higher OD₆₀₀ levels compared to the control strain and the double mutant (Figure 5C). No significant differences between the growth curves of the five strains were observed without peptide (Figure 5D). The growth rates at the different concentrations of seg5D were similar between the control strain and the double mutant and between the skp mut and the triple mutant (Figure 5E). The growth rates of RL1 were mostly similar to the skp mut and the triple mutant except of the growth rate at 25 μM, which was higher for RL1 (Figure 5E). We further tested the growth in the presence of other AMPs and observed variations in growth at sub-inhibitory concentrations. In the presence of 3.125 μM C8-K5L7 peptide (0.5 × MIC of the control strain) RL1, skp mut and the triple mutant had growth advantages due to a shorter lag phase and higher maximal OD₆₀₀ values compared to the control strain and the double mutant (Figure 6A). We also found differences in growth in the presence of 1.56 μM colistin (0.5 × MIC of the control strain). RL1 showed an optimal growth, and skp mut showed a slightly slower growth than RL1 (Figure 6B). The triple mutant and the control strain showed similar growth (Figure 6B). The double mutant showed a negligible growth under these conditions, suggesting colistin hypersusceptibility of that strain (Figure 6B). Once more, no differences were observed between the strains in the absence of peptides (Figure 6C). The growth rates of the control strain were significantly reduced at 1.56 μM of colistin or 3.125 μM of C8-K5L7, opposed to RL1, skp mut, and the triple mutant (Figure 6D). Moreover, the growth rate of the double mutant was significantly reduced at 1.56 μM of colistin compared to all the other strains (Figure 6D). Overall, the skp mutation induced hyposensitivity toward 4DK5L7 and C8-K5L7 and resistance toward seg5D. The double mutant showed hyposensitivity toward 4DK5L7 and hypersusceptibility to colistin. A combination of the three mutations in ramR, acrB, and skp resulted in a significant resistance to the 4DK5L7 peptide.

In Gram-negative bacteria, 35% of the proteome is directed to the inner and the outer membrane (Mas et al., 2019). Since Skp plays a key role in OMP translocation, we hypothesized that its mutation would have a significant impact on the proteome. Moreover, we aimed to shed light on the resistance mechanism to AMPs of this genotype. We analyzed the proteome of the control strain, RL1, and the skp mut strains. Single-protein analysis and pathway analysis, based on the KEGG database, were conducted. First, we found that the Skp protein was expressed in the control strain, but not in RL1 or the skp mut bacteria (Supplementary Table S2). This suggests that the mutation led to the loss of this protein. Unlike in the control strain and skp mut strains, we did not detect the LPS biosynthesis gene products of yeiU and rfaY in RL1 (Supplementary Table S2). The RamR protein was found to be expressed similarly among the three strains, suggesting that the ramR mutation did not influence protein stability but rather impaired its activity. According to the KEGG database, we found that proteins of the cationic antimicrobial peptide (CAMP) resistance pathway were enriched in RL1 and skp mut proteomes compared to the control strain (Figure 7). This was reflected by elevation in HtrA (DegP) both in RL1 and the skp mut, compared to the control strain. This protein is known to play a role along with Skp in OMP translocation through the periplasm (Sklar et al., 2007). Additional proteins in the CAMP resistance pathway that were up-regulated in RL1 but not in the skp mut are ArnD, YcfS, AcrA, AcrB, and TolC (Supplementary Table S3). The up-regulation of AcrB in RL1 due to the mutation in ramR is consistent with our Western blot analysis (Figure 2B) and could also explain the elevation in AcrA levels, as acrA and acrB are cotranscribed from a single operon (Yamasaki et al., 2013). A depletion in β-lactam resistance pathway was observed in RL1 and skp mut (Figure 7), which was enunciated by the reduction of OMPs. A significant reduction in OmpC was observed for skp mut, whereas in RL1 additionally OmpF was significantly depleted (Supplementary Table S3). According to the single-gene analysis, we observed a more extensive effect of the skp mutation on OMPs. We found a significant reduction of OmpC, OmpD, and OmpF levels both in RL1 and in the skp mut compared to the control strain (Supplementary Table S2). However, some OMPs exhibited higher levels in RL1 and skp mut compared to control strain. Amounts of OmpX and OmpA were higher in the skp mut compared to the control strain, and OmpR and OmpN were found to be more abundant in RL1 compared to the control strain (Supplementary Table S2). Other cellular pathways that were enriched in RL1 and skp mut in comparison to the control strain are as follows: microbial metabolism in diverse environments, purine metabolism, selenocompounds metabolism, and sulfur metabolism, which were mostly reflected in up-regulation of the genes in the L-cysteine synthesis pathway (Supplementary Table S3). Compared to the control strain and the skp mut, RL1 showed depletion of proteins that are part of the bacterial chemotaxis, flagellar assembly, TCS, and Salmonella infection pathways. RL1 showed enrichment in the proteins of the β-lactam resistance, CAMP resistance, and starch and sucrose metabolism pathways (Figure 7). Proteins of the propanoate metabolism pathway were depleted in the skp mut compared to the control strain and to RL1 (Figure 7). The enrichment of the bacterial secretion system in RL1 compared to the skp mut was reflected by elevation in TolC levels (Supplementary Table S3). To summarize, the skp mutation led to the loss of Skp and had a widespread influence on the proteome. Importantly, Skp loss led to a significant reduction in OMPs and to elevation in DegP. Compared to the skp mut, RL1 differs more from the control strain in the cellular pathways, apparently, due to additional mutations it carries.