Ribosomal Protein L4 of Lactobacillus rhamnosus LRB alters resistance to macrolides and other antibiotics

Abstract

Lactobacillus rhamnosus is an important lactic acid bacterium that is predominantly used as a probiotic supplement. This bacterium secretes immunomodulatory and antibacterial peptides that are necessary for the probiotic trait. This organism also occupies diverse ecological niches, such as gastrointestinal tracts and the oral cavity. Several studies have shown that L. rhamnosus is prone to spontaneous genome rearrangement irrespective of the ecological origins. We previously characterized an oral isolate of L. rhamnosus, LRB, which is genetically closely related to the widely used probiotic strain L. rhamnosus LGG. In this study, we isolated a nontargeted mutant that was particularly sensitive to acid stress. Using next generation sequencing, we further mapped the putative mutations in the genome and found that the mutant had acquired a deletion of 75 base pairs in the rplD gene that encodes the large ribosomal subunit L4. The mutant had a growth defect at 37°C and at ambient temperature. Further antibiotic sensitivity analyses indicated that the mutant is relatively more resistant to erythromycin and chloramphenicol; two antibiotics that target the 50S subunit. In contrast, the mutant was more sensitive to tetracycline, which targets the 30S subunit. Thus, it appears that nontargeted mutations could significantly alter the antibiotic resistance profile of L. rhamnosus. Our study raises concern that probiotic use of L. rhamnosus should be carefully monitored to avoid unintended consequences.

INTRODUCTION

Lactobacillus rhamnosus is commonly present as a commensal organism in the human oral cavity and it also inhabits the gastrointestinal tracts and vagina. This lactic acid bacterium (LAB) exhibits a diverse range of ecological niches that vary from animals to plant materials and fermented dairy products. This LAB is often used in the dairy industry due to its lactic acid production and cheese-ripening traits. However, in recent years, various strains of L. rahamnosus are extensively used as a probiotic supplement because many of the isolates exhibit immunomodulatory and antimicrobial properties, and thus it is used for treatment of various human ailments (for reviews see, (Banna et al., 2017; Segers & Lebeer, 2014)). Among the probiotic strains that are currently used, L. rhamnosus GG (LGG; ATCC 53103) is perhaps the best characterized strain, which was originally isolated from human feces (Lebeer, Verhoeven, Perea Velez, Vanderleyden, & De Keersmaecker, 2007; Silva, Jacobus, Deneke, & Gorbach, 1987). Two additional strains, a food isolate HN001 and a vaginal isolate GR-1, are also widely used as a probiotic supplement to treat various human illnesses (Aljewicz, Siemianowska, Cichosz, & Tonska, 2014; Kim, Shynlova, & Lye, 2019; Petrova, van den Broek, et al., 2018; K. Wickens et al., 2012; K. L. Wickens et al., 2017; Yeganegi et al., 2010; Yeganegi et al., 2009). Despite the fact that L. rhamnosus is generally considered as a commensal and beneficial organism, sometimes this organism can cause serious illnesses such as bacteremia and sepsis (Gouriet, Million, Henri, Fournier, & Raoult, 2012; Robin et al., 2010).

The average genome size of L. rhamnosus is 3.0 ± 0.2 Mb in length and it is one of the largest among the lactobacilli group (Canchaya, Claesson, Fitzgerald, van Sinderen, & O’Toole, 2006). While the small size of the core-genome (~2100 genes) is typical for a lactobacillus (~2100 genes), the size of the pan-genome (>4700 genes) is relatively large as compared to other lactobacilli strains (Ceapa et al., 2016; Douillard et al., 2013). Several comparative genome studies have revealed that L. rhamnosus isolates of different origins contain nearly 20 highly variable regions on the genome that are necessary for functions related to the lifestyle and adaptation for a given niche. These functions are typically related to carbohydrate transport and utilization, capsule production, pili production, altered transcriptional regulators, and bile salt resistance (Douillard et al., 2013; Nadkarni, Chen, Wilkins, & Hunter, 2014; Petrova, Macklaim, et al., 2018). All the L. rhamnosus isolates could be genotypically classified into eight different clades with isolates from similar environmental niches grouped together. However, when both the phenotype and the genomic data of various isolates are assimilated, only two geno-phenotype groups, named A and B, have emerged (Douillard et al., 2013). The strains belonging to gene-phenotype A lack pili production and display altered carbohydrate metabolism suggestive of a dairy-like environment. On the other hand, geno-phenotype B isolates produce pili, display bile salt resistance and exhibit a carbohydrate metabolism pattern suitable for adaptation to intestinal tracts (Douillard et al., 2013). Because of the nature of the lifestyle, all the probiotic strains such as LGG and GR-1 belong to the geno-phenotype B group.

One of the concerns using L. rhamnosus isolates as probiotic organisms is that often the strains undergo spontaneous genome rearrangements. Most of the genome rearrangements occur due to the resident IS elements. For example, the L. rhamnosus LGG strain contains nearly 70 IS elements and recombination between any two IS elements results in deletion of internal segments (Sybesma, Molenaar, van, Venema, & Kort, 2013). An experimental evolution study with the L. rhamnosus LGG strain involving 1000 generations not only produced large genomic rearrangements mediated by the IS element but also insertions/deletions (indels) and single nucleotide polymorphisms (SNPs) (Douillard et al., 2016). Interestingly, the LGG strain also generated frequent genome rearrangements when grown under stress conditions (bile-salt)(Douillard et al., 2016). Another laboratory evolution study generated SNPs that made the LGG strain resistant to freeze-thaw and improved growth fitness only after 150 generations (Kwon, Bae, Kim, & Han, 2018). These alterations cause both genome reduction and gene decay and thus the suitability of LGG as a probiotic strain is lost. Thus, further investigations are required to evaluate whether a strain is suitable for use as a probiotic organism.

We have previously isolated a L. rhamnosus strain, LRB, from the oral cavity of a child (S. Biswas & Biswas, 2016). The LRB strain exhibited the highest homology with strains HN100 and LGG (S. Biswas & Biswas, 2016; S. Biswas, Turner, & Biswas, 2018). Upon comparison of the genomes of LRB and LGG, we observed that the LRB genome is nearly 75-kb shorter than the LGG genome. We found that the lost ~75 kb segment is flanked by the IS5 elements; which might be responsible for the loss of the segment due to recombination between the IS elements. The lost segment encoded genes that are required for carbohydrate metabolism and pili formation. The LGG genome also contains a ~285 Kb segment with the replication origin that appeared to be translocated in the LRB genome (S. Biswas & Biswas, 2016; S. Biswas et al., 2018). LRB secretes antimicrobial factors that inhibits various streptococci and a few gram-negative pathogens (S. Biswas et al., 2018). During our previous studies, we noticed that LRB also frequently generates spontaneous mutants and we previously characterized one such mutant that was sensitive to various stresses, including acid stress, oxidative stress, and osmotic stress (S. Biswas, Keightley, & Biswas, 2019). In the current study, we have isolated a nontargeted mutant (RB3S) that is specifically sensitive to acid stress. Whole genome sequencing indicated that the mutant acquired a partial in the rplD gene that encodes a large ribosomal subunit, L4. Subsequently, we found that this mutant showed altered antibiotic sensitivities due to the mutated L4 subunit.

MATERIALS AND METHODS

Bacterial strains and growth media:

Lactobacillus rhamnosus LRB and its derivatives were routinely grown in Mann Rogosa Sharpe (MRS; Difco). L. rhamnosus strains were also grown in Todd-Hewitt medium supplemented by 0.2% yeast extract (THY). Strains were incubated at 37°C, grown under microaerophilic conditions (no oxygen) using candle jar. For some stress studies, strains were incubated at different temperatures (where indicated).

Isolation of mutants:

Mutants were isolated as described previously (S. Biswas et al., 2019). Briefly, electrocompetent cells were prepared and electroporated using ~500ng of pGhost9TR plasmid DNA. Electroporated cells were recovered in MRS+ broth containing 2mM CaCl₂ and 20mM MgCl₂ for an hour at 37°C and then plated onto MRS+ agar containing 5 μg/ml erythromycin (Em) and incubated at 37°C in candle jar. Em resistant colonies was selected, grown at 37°C in the presence of Em. About 1000 colonies were replica patched on MRS+ agar and MRS+ agar adjusted to pH 4.5 with citrate-phosphate buffer as described previously (I. Biswas, Drake, Erkina, & Biswas, 2008). Sensitive colonies were recovered from the MRS+ agar plate and verified again using fresh media. For growth kinetics, overnight cultures were diluted 20-fold in fresh MRS media and grown at 37°C water bath. Growth was monitored by using a Klett-Summerson colorimeter with a red filter, as previously described (I. Biswas & Scott, 2003)

Acid and thermal stress tolerance response:

Cultures were grown to exponential phase in MRS broth, harvested by centrifugation and washed twice with 0.85% NaCl before resuspending in 0.85% NaCl. The optical density (A600) of the cultures were adjusted to 2.0 and 10-fold serially diluted, and 4 μl of each dilution was spotted onto MRS agar or MRS agar with pH adjusted to pH 4.5. Plates were incubated overnight at 37°C or at various temperatures, under microaerophilic conditions (candle jar), and bacterial growth was evaluated as previously described (S. Biswas & Biswas, 2011).

Antibiotic sensitivity assays.

Disk diffusion assays were performed to evaluate antibiotic susceptibility of the L. rhamonosus mutants as described previously with the following modifications (S. Biswas et al., 2019). Fresh cultures were grown in MRS broth and the initial optical density (OD₆₀₀) was adjusted to 0.1 with 0.85% NaCl. Cultures were spread onto MRS plates with a cotton swab and an automatic plate rotator. Antibiotic disks (6 mm in diameter; BD) were then placed on the inoculated plates. The zones of inhibition were measured after overnight incubation under microaerophilic conditions. For some antibiotics, the desired amount of antibiotics was added to an empty disc and placed on the agar plate with the cultures. The antibiotic discs used are listed in Table 1.

Table 1:

Zone of inhibition (ZOI) for various antimicrobials

Antibiotics$	Zone of inhibition diameter (mm)
	LRB	RB3S
Bacitracin (10 mcg)	18.0 ± 1.0	18.0 ± 1.0
Chloramphenicol (5 mcg) *	17.5 ± 1.2	11.8 ± 0.4
Ciprofloxacin (5 mcg)	18.5 ± 0.5	18.0 ± 1.0
Clindamycin (2 mcg) *	25.8 ± 1.6	20.5 ± 0.5
Doxycycline (30 mcg)	36.3 ± 5.4	37.0 ± 5.8
Erythromycin (15 mcg) *	34.0 ± 2.2	26.8 ± 1.7
Kanamycin (125 mg/ml) *	15.0 ± 1.0	13.5 ± 1.0
Linezolid (30 mcg) *	32.5 ± 2.2	30.0 ± 2.3
Meropenem (10 mcg) *	19.5 ± 1.1	16.2 ± 0.9
Mezlocillin (75 mcg) *	41.2 ± 2.0	46.3 ± 3.8
Neomycin (30 mcg) *	9.8 ± 0.7	8.8 ± 0.7
Nisin (40 mg/ml)	12.0 ± 0.5	13.0 ± 1.0
Puromycin (10 mg/ml)	14.0 ± 2.0	14.2 ± 1.8
Rifampin (5 mcg)	27.5 ± 0.5	27.0 ± 0.0
Spectinomycin (50mg/ml) *	20.5 ± 0.5	17.5 ± 0.5
Streptomycin (50mg/ml)	20.0 ± 0.5	21.5 ± 0.5
Tetracycline (10 mg/ml) *	34.0 ± 1.0	40.0 ± 2.0
Tigecycline (15 mcg) *	35.1 ± 3 .0	38.0 ± 3.2

^$, per disc (mcg);

^*, significant difference

Minimum inhibitory concentrations (MIC) were measured following a protocol that is specifically adjusted to lactobacilli (Mayrhofer et al., 2014). Briefly, MIC was determined by the microdilution method using a 96-well plate and each well contains 100μl of ~5×10⁵ cfu/ml of fresh overnight grown cultures and 2-fold serial dilutions of antibiotics. Plates were incubated at 37°C in a CO₂ incubator. After 24 hrs of incubation, the wells with no visible growths were noted for MIC determination.

Whole genome sequencing to map mutations.

Whole genome sequencing was done at the Institute for Genome Sciences (Univ. Maryland) using the Illumina HiSeq2000 platform as described previously (S. Biswas et al., 2019). Briefly, genomic DNA was purified from L. rhamnosus and the purity was assessed with a NanoDrop ND-1000 spectrophotometer. A paired-end 151 bases (PE150) DNA library was generated following the manufacturer’s protocol (Illumina). The average genome coverage provided by the library was nearly 60X. Geneious Prime software was used to map the Illumina reads onto the reference genome sequence of L. rhamnosus LRB (CP016823) with the default alignment parameters. Short indels and SNPs were called based on the read mapping of the SNP calling procedures following the default setting. Once the putative genes were identified, the sequences were verified by PCR amplification followed by Sanger sequencing.

Comparative proteomic studies using level free quantitation (LFQ).

LFQ studies were done essentially as described previously (S. Biswas et al., 2019). In short, whole cell protein lysates were made in PBS from exponentially grown cultures using PBS. Lysates were subjected to proteolytic digestion using Filter-Aided Sample Preparation (FASP) as previously described (McDowell, Gaun, & Steen, 2013). Approximately 100μg protein was reduced (10mM TCEP), then diluted with 8M urea, and filtered through a microcon filter (10k MWCO). The retained protein was alkylated with 50mM chloracetamide for 30 minutes and then rinsed with 8M urea. The buffer was exchanged to 100mM Tri-Ethyl-Ammonium Bicarbonate (TEAB) and 0.5μg Trypsin/LysC (Promega V5073) was added in the same buffer and incubated at 37C overnight. Peptidomes were harvested by centrifugation, dried down and resuspended in pure water. Peptidomes were analyzed in a single long gradient (140 min.) with standard acidic (0.2% formic acid) reversed phase on a C18 column. A QExactive Plus HRMS system (Thermo) was used to acquire data dependent runs at 35k MS, and 17.5k MS/MS resolution, with 18 dependent scans per survey. Each sample was run in quadruplicate, generating 4 data files each for the pairwise comparison (LRB vs RB3S). MaxQuant was used for the database search and LFQ analysis against the complete proteome of Lactobacillus rhamnosus (ATCC 53103) plus Saccharomyces cerevisiae (baker’s yeast) proteome for additional complexity (8,874 total protein sequences), enabling decoy (reversed sequence) database for False Discovery Rate calculations (FDR). FDR was set at 1% for identification purposes within MaxQuant. LFQ was enabled with no changes to default settings and the resulting protein identifications and quantitative data was further analyzed using Perseus (Perseus reference).

RESULTS

Isolation and characterization of a nontargeted mutant of L. rhamnosus LRB.

We previously isolated a spontaneous acid-sensitive mutant of L. rhamnosus LRB when we electroporated the strain (S. Biswas et al., 2019). Therefore, we wondered whether we could reproduce the phenomenon. For this, we used plasmid pGhost9-TR, which contains the pWV01 backbone and an erythromycin resistance marker (Maguin, Duwat, Hege, Ehrlich, & Gruss, 1992). We electroporated the plasmid into the LRB strain and selected erythromycin resistant colonies. We then screened the transformants for sensitivity to low pH (pH 4.5) by replica patching. Among ~1000 colonies, we found only one clone that displayed growth defect on an MRS agar plate buffered at pH 4.5. We selected the clone, named it RB3R, and then cured the plasmid pGhost9-TR from the cells by culturing without erythromycin for a few generations (Maguin, Prevost, Ehrlich, & Gruss, 1996). One such erythromycin sensitive clone was selected, PCR verified for the loss of the plasmid, and named RB3S for further analysis. We first confirmed whether RB3S is sensitive to low pH (pH 4.5). As shown in Fig. 1A, RB3S, like the original clone RB3R, was also highly sensitive to acid stress response. We found that while RB3S grew well on an MRS agar plate at 37°C, it had a growth defect when grown in liquid. The doubling time of RB3S was ~3.3 times slower than the wild type strain in the mid-log phase of growth (5– 6 hrs). The final biomass for RB3S was nearly 50% less than the wild type; typical Klett unit values for the overnight cultures were approximately 400 and 200, for LRB and RB3S, respectively (Fig. 1B). While RB3S did not display any noticeable growth defect at 42°C when grown on an MRS agar plate, the mutant showed a significant growth defect when grown at ambient temperature on an MRS agar plate (Fig. 1A, data not shown).

Figure 1: Growth characteristic of the mutant.

A. Acid and cold sensitive phenotypes of the mutant. Fresh overnight cultures were adjusted to a starting optical density (A₆₀₀) of 5.0 made in 0.85% NaCl, and a 10-fold dilution series were made. 5μl from the dilution series were spotted onto MRS agar adjusted to pH 4.5 with sodium citrate buffer. As a control, cultures were also spotted on MRS agar plates. Strains used for the assays are wild type L. rhamnosus LRB; and RB3R, an isogenic mutant derived from LRB containing an erythromycin-resistant plasmid; RB3S, RB3R that is cured of the plasmid; and RBM1, an ftsH mutant as control. Experiments were repeated at least three times, and relevant areas of representative plates are shown. B. Growth kinetics of LRB and RB3S. All strains were grown in MRS broth at 37°C in Klett flasks, and growth was monitored at 60-min intervals for 26 hours. Growth curves were plotted using three replicates for each strain, and the experiment was repeated twice independently. The data shown are the means ± standard deviations from one such experiment.

The previously characterized acid sensitive mutant (RBM1) was also sensitive to multiple stresses including osmotic and superoxide stresses (S. Biswas et al., 2019). Thus, we tested RB3S for osmotic stress generated by NaCl (5%) and superoxide stress generated by viologens. We found that RB3S behaves very similar to the wild type LRB strain for both the stresses. In addition, we also found that RB3S behaves similar to LRB when exposed to acriflavine, chlorohexidine gluconate and potassium tellurite (data not shown). Thus, RB3S seems to be only sensitive to acid stress and not to other stresses.

L. rhamnosus LRB produces antimicrobial activities that inhibit a wide range of streptococci (S. Biswas, Turner, & Biswas, 2018). To confirm that the mutant strain, RB3S, still secretes antimicrobial activities, we performed a plate-based diffusion assay by spotting the overnight grown cultures as described previously (S. Biswas, Turner, & Biswas, 2018) except that the final cell numbers were adjusted to equal since RB3S and LRB grow differently. We used three oral streptococci (S. sobrinus, S. oralis, and S. mutans) that commonly share the same environmental niche with L. rhamnosus. We observed no significant reduction in secretion of antimicrobials by the RB3S mutant strain as compared to the wild type LRB (data not shown).

RB3S acquired a deletion in the rplD gene.

To identify the site of the nontargeted mutation(s), we sequenced the RB3S genome using the Illumina platform with a genome coverage of ~60x. As described above, we aligned the sequence reads on the LRB genome. We found a large deletion in the rplD gene that encodes for the 50S ribosomal subunit protein L4. To confirm that RB3S indeed harbors the deletion, we PCR amplified the entire gene including the flanking regions and performed Sanger sequencing. We found that RB3S had lost a 75-bp segment in the rplD gene, which corresponds to deletion of codon 61 through codon 85 (Fig. 2). We noticed that the deleted region is flanked by an eight-base repeat sequence GCGGCGGT. Apart from this deletion we found no other significant mutations in the RB3S genome that altered the coding sequences. Thus, it appears that the observed phenotypes in the RB3S strain are probably due to the mutated L4 protein only.

Figure 2: Multiple sequence alignment of L4 proteins from diverse bacteria.

The sequences shown are: Escherichia coli (ECL), Bacillus subtilis (BSU), Streptococcus pneumoniae (SPN), Streptococcus mutans (SMU), Lactobacillus rhamnosus (LRB), and the L. rhamnosus mutant RB3S (RB3). The alignment was performed using Clustal-Omega. Degree of shading is indicative of sequence homology and was done using BoxShade. The deleted region of the L4 protein in the RB3S mutant strain is shown with a bar above the alignment.

RB3S displays altered antibiotic sensitivity.

Since we found that the RB3S strain encodes a mutated L4 protein that lacks a conserved motif important for binding of macrolide antibiotics, we tested the mutant strain for erythromycin sensitivity. We also included antibiotics that target both the 50S and 30S ribosomal subunits. In addition, we also tested a few other antibiotics that target other biosynthetic pathways. As shown in Fig. 3, RB3S displayed increased resistance to both erythromycin (macrolide) and chloramphenicol antibiotics as compared to the wild type LRB. The zone of inhibition (ZOI) that was observed with RB3S for erythromycin was smaller (26.8 ± 1.7 mm) whereas the ZOI was larger for LRB (34.0 ± 2.2 mm). Similarly, the ZOIs were 11.8 ± 0.4 mm and 17.5 ± 1.2 mm, respectively, for RB3S and LRB strains for chloramphenicol. RB3S was also more resistant to other antibiotics that target the 50S ribosomal subunit such as clindamycin and linezolid (see Table 1). However, we did not observe any noticeable difference for puromycin, that also targets the 50S subunit (ZOI was ~14.0 mm). On the other hand, we observed different results for antibiotics that target the 30S ribosomal subunit. It appears that RB3S was slightly more resistant to aminoglycosides such as kanamycin and neomycin as well as spectinomycin, which is chemically related to aminoglycosides (Table 1). In contrast, RB3S showed increased sensitivity to tetracycline and its derivative tigecycline as compared to the wild type LRB (Table 1). We also determined the minimum inhibitory concentrations (MIC) for some of the antibiotics that target either the 30S or the 50S subunits. We found that MICs for erythromycin and chloramphenicol were nearly four- and two-fold higher, respectively, for RB3S as compared to LRB (Table 2). In the case of tetracycline, the MIC for RB3S was four-fold lower as compared to LRB.

Figure 3: Antibiotic sensitivity assay.

Disk diffusion assay was used to measure the susceptibility of LRB and RB3S against different antibiotic discs. MRS agar containing bacterial lawns of LRB and RB3S strains were prepared as described in the text. Antibiotics discs were placed and the plates were incubated further. Symbols are: EM, erythromycin; CC, clindamycin; and MZ, mezlocillin. The inhibitory-zone diameters (indicated by arrow) for LRB and RMB1 were measured and compared. Experiments were repeated no fewer than three times, and relevant areas of representative plates are shown.

Table 2:

Minimum inhibitory concentrations (MIC) for protein synthesis inhibitors

Antibiotics	MIC (μg/ml)
	LRB	RB3S
Apramycin	>64	>64
Chloramphenicol	8	16
Erythromycin	0.5	2
Gentamycin	>64	>64
Kanamycin	>64	>64
Spectinomycin	256	256
Streptomycin	256	512
Tetracycline	1	0.25

As for the cell-wall targeting antibiotics, we also obtained mixed results. For the carbapenem class of drugs, such as meropenem, we found that RB3S was more resistant as compared to LRB (ZOIs were 19.5 ± 1.1 mm vs 16.2 ± 0.9 mm, respectively). However, for mezlocillin (a beta-lactam derivative), we observed RB3S was more sensitive as compared to LRB (ZOIs were 41.2 ± 2.0 mm vs 46.3 ± 3.8 mm, respectively). Taken together, we found that RB3S was differentially sensitive to various antimicrobials.

RB3S displays altered proteome.

To gain an overview at the molecular level for this stress sensitive phenotype, we compared the proteomes by performing LFQ analysis using MaxQuant. We grew the cells (RB3S and LRB) in THY broth at 37°C to the mid-logarithmic growth phase and collected the samples for LFQ analysis as described in the method section. L. rhamnosus strains including LRB encode nearly 2000 proteins (Koskenniemi et al., 2011; Laakso et al., 2011; Savijoki et al., 2011). But under these rich media growth conditions where many anabolic pathways would be repressed, we were able to identify approximately 700 proteins using only single LCMS gradient runs. Interestingly, we identified 26 proteins in the wild type strain that were not detected in the mutant, though possibly present at lower, undetectable levels (Table 3A). These proteins are in general involved in metabolic and regulatory functions. A few were hypothetical proteins with no known functions. In addition, we also found that 15 proteins were identified in LRB in at least 3 replicates, but only weakly detected in RB3S (in only one replicate, and at lower relative abundance). We identified three SAM-dependent methyltransferase enzymes (AXI94200, AXI95119, and AXI94979) that were more abundant in the wild type strain, or not detected at all in RB3S. On the other hand, we found 20 proteins that were only identified in the RB3S strain, while 9 other proteins were identified in at least three runs in the RB3S strain, but only one, and at lower relative abundance in wild type LRB strain (Table 3B). Several of these proteins were either ribosomal proteins such as L28 and L30 or associated with ribosomal function. For LFQ analysis, MaxQuant uses the average MS peak intensity corresponding to the identified peptides for the quantitative comparison. A log base 2 transformation of the raw intensities is performed prior to volcano plot (Figure 4). It should be noted that peptides present at lower abundance can be missed in data dependent MS acquisition, whereas peptides from more abundant proteins are more likely to be isolated for MS/MS, and identified. Also, peptides present at lower abundance may produce poor MS/MS data that do not meet identification criteria. All of the proteins listed in Tables 3A and 3B were excluded from Fig. 4, since they were not identified in all four replicate runs in both strains with stringent scoring criteria of 1% FDR (or not detected at all in one strain).

Table 3A:

Proteins expressed at relatively higher level in LRB by LFQ analysis

Accession	Description (Gene name)	# Unique Peptides	Sequence coverage (%)	Score
Identified only in LRB, all 4 replicate runs, or in 3 (*)
AXI94710	IMP cyclohydrolase PurH (DU507_09415)	14	48.9	125.54
AXI95264	NAD-dependent epimerase (DU507_12610)	07	60.1	87.82
AXI94993	β-hydroxyacyl-ACP dehydratase (DU507_11075)	02	22.1	5.64
AXI95459	Histidine phosphatase (DU507_13760)	02	13.1	58.06
AXI94334	LysR family regulator (DU507_07430)	05	25.8	99.43
AXI95425	Polyphosphate kinase (DU507_13565)	06	11.9	21.35
AXI94967	DNA-methyltransferase, PglX (DU507_10940)	02	2.3	2.93
AXI94719	Ribonucleotide mutase, PurE (DU507_09460)	02	15.4	36.82
AXI94989	FabK reductase (DU507_11055)	02	7.4	11.75
AXI94983	Acetyl-CoA carboxylase (DU507_11025)	02	9.6	49.41
AXI94985	Acetyl-CoA carrier protein, AccB (DU507_11035)	04	60.1	49.61
AXI93150	N-acetyltransferase, DapD (DU507_0580)	04	18.8	12.55
AXI95905	Hsp20 family protein (DU507_14615)	04	37.0	27.01
AXI93646	HAD family hydrolase (DU507_03425)	02	9.10	23.24
AXI94200	SAM-Methyltransferase, RsmH (DU507_06700)	04	17.3	42.79
AXI93364	Glycosyltransferase (DU507_01785)	02	4.1	6.15
AXI93102	Dehydrogenase family protein (DU507_00310)	06	15.6	16.51
AXI94787	Hypothetical protein (DU507_09935)	02	10.9	5.20
AXI93153	Diaminopimelate epimerase (DU507_00610)	02	7.8	7.67
AXI94630	RNA methyltransferase (DU507_08980)	01	8.3	6.26
AXI94986 *	Beta-ketoacyl-synthase, FabF (DU507_11040)	05	27.4	41.76
AXI93317 *	Oxidoreductase (DU507_01510)	05	36.5	48.86
AXI94988 *	ACP S-malonyltransferase (DU507_11050)	01	5.9	2.86
AXI93363 *	Glycosyltransferase (DU507_01780)	03	7.5	4.92
AXI93708 *	Hypothetical protein (DU507_03780)	01	14.8	2.97
AXI93059 *	DNA topoisomerase, GyrB (DU507_00025)	02	2.8	2.50
Identified: in LRB, in all 4 replicate runs, or in 3 (*); in RB3S, 1 replicate, and at lower abundance
AXI95119	SAM- methyltransferase (DU507_11795)	03	23.2	8.21
AXI93816	CutC (DU507_04540)	02	19.3	12.35
AXI94228	N-acetyltransferase (DU507_06850)	01	2.5	3.60
AXI93870	YvcK (DU507_04840)	02	6.1	3.07
AXI95791	Hydrolase (DU507_03460)	05	28.9	13.23
AXI94021	Hypothetical protein (DU507_05690)	03	12.0	10.51
AXI94794	Hydrolase (DU507_09985)	03	19.0	12.48
AXI94133	SufB (DU507_06315)	04	12.3	46.28
AXI95434	tRNA-ligase, TrpS (DU507_13610)	06	26.6	21.24
AXI93198	PTS sugar transporter (DU507_00850)	05	60.6	46.06
AXI93676	Nitroreductase (DU507_03605)	03	14.6	11.51
AXI94984	Dehydratase, FabZ (DU507_11030)	06	43.2	71.44
AXI94979 *	SAM-Methyltransferase (DU507_11005)	02	6.9	24.3
AXI94361 *	Fibrinogen-binding protein (DU507_07565)	01	3.9	20.56
AXI94970 *	BrxC (DU507_10955)	03	3.6	7.92

Table 3B:

Proteins expressed at relatively higher level in RB3S identified by LFQ)

Accession	Description (Gene name)	# Unique Peptides	Sequence coverage (%)	Score
Identified only in RB3S all 4 replicate runs, or in 3 (*)
AXI93596	Ribosomal protection protein (DU507_03130)	06	15.4	18.34
AXI95252	Hypothetical protein (DU507_12550)	04	54.5	54.50
AXI94936	ABC transporter (DU507_10780)	02	5.4	3.15
AXI95307	Ribosomal protein L30 (DU507_12850)	02	50.8	45.80
AXI93974	PHB domain protein (DU507_05405)	03	11.0	7.00
AXI95670	N-acetyltransferase (DU507_14915)	04	33.5	31.30
AXI94116	F0F1 ATP synthase (DU507_06210)	01	23.1	42.52
AXI93353	Beta-galactosidase (DU507_01720)	04	6.7	8.31
AXI94637	DUF177 protein (DU507_09025)	02	19.2	3.56
AXI94581	Primosomal protein N (DU507_08725)	03	5.3	5.65
AXI95292	Transcriptional regulator (DU507_12775)	01	7.9	8.34
AXI94294	rRNA pseudouridine synthase (DU507_07220)	02	10.5	4.49
AXI94566	ABC transporter (DU507_08650)	03	17.0	21.52
AXI94206 *	FtsA (DU507_06735)	01	2.9	3.00
AXI94779 *	Hypothetical protein (DU507_09860)	02	14.9	11.37
AXI95002 *	Transcriptional regulator (DU507_11130)	01	15.5	21.89
AXI94267 *	Pyrophosphohydrolase (Du507_07080)	03	39.6	20.72
AXI94077 *	Endonuclease (DU507_05990)	01	3.5	31.74
AXI93523 *	Hydrolase (DU507_02735)	02	16.8	5.94
AXI95101 *	pTs sugar transporter (DU507_11685)	02	24.1	6.08
Identified: in RB3S, 4 replicate runs, or in 3 (*); in LRB, 1 replicate, and at lower abundance
AXI93697	NAD-dependent malic enzyme (DU507_03725)	05	17.5	21.56
AXI94076	Monooxygenase (DU507_05985)	01	18.9	10.75
AXI94282	Phosphodiesterase (DU507_07160)	02	4.3	3.86
AXI94573	Ribosomal protein L28 (DU507_08685)	02	24.6	3.28
AXI94564	Permease (DU507_08640)	01	3.0	4.05
AXI95668	ABC transporter (DU507_14900)	02	18.7	8.18
AXI94619	Iron-sulfur cluster biosynthesis protein	02	24.2	5.38
AXI94680	Methyltransferase TrmB (DU507_09245)	03	12.6	6.44
AXI94277 *	Ribonuclease Z (DU507_07130)	02	10.3	34.97

Figure 4: Label-free quantitative analysis of differentially expressed proteins.

Volcano plot representing the logarithmic ratio of protein LFQ intensities in the RB3S/LRB experiments plotted against negative logarithmic p-values of the t-test performed from quadruplicates (FDR threshold = 0.1, S0 = 0.5). A hyperbolic threshold curve separates significant differentially expressed proteins from background. Some differentially expressed ribosome associated proteins are indicated. Dashed line indicates a more stringent p-value cutoff (1.5). Only proteins that were identified in all replicates in both strains were included in the plot.

We found over 230 proteins were readily identified in all replicates, but in addition, were found to be differentially regulated in the RB3S mutant strain as compared with the wild type LRB strain, based on the LFQ analysis (Fig. 4). Most of the proteins that were upregulated in the RB3S strain were ribosomal proteins. We observed that most of the ribosomal L-proteins (nearly 25) and S-proteins (about 14) were more abundant in the mutant. We also found 30S ribosome binding factor RbfA, ribosome maturation factors RimP and RimM, and ribosome silencing factor RsfS were all upregulated in RB3S. In addition, translational initiation factors (IF2 and IF3), elongation factors (G, P, Ts) were also seen at higher levels in the mutant. Two other factors, the ribosome recycling factor (RCF) and the ribosome splitting factor HflX, which is a GTPase, were both more abundant in the mutant. Thus, the results of the proteomic studies provide evidence for a significantly altered proteome in the RB3S mutant strain compared to LRB. Most of the ribosome associated proteins and proteins involved in translation were expressed at relatively higher levels in the RB3S mutant strain.

DISCUSSION:

L. rhamnosnus strains are widely used as probiotics yet various studies have shown that these strains are highly susceptible to genomic rearrangements and accumulate mutations that often abolish the beneficial traits of the probiotic strains. We previously isolated a spontaneous mutant of the L. rhamnosus strain that was highly sensitive to various stresses including acid stress. This mutant encodes a defective FtsH protein, which is an important intracellular protease (S. Biswas et al., 2019). In this study, we further isolated another nontargeted mutant that is sensitive to only acid stress.

The mutant was generated when we transformed a plasmid bearing an erythromycin resistant gene and selected for resistant colonies. We are not sure whether electroporation or the introduction of the plasmid had an effect on the mutant generation. However, upon further investigation, we unraveled that the mutant, RB3S, encodes a deleted L4 ribosomal protein that lacks a conserved motif important for macrolide binding. It appears that the deletion was created due to a recombination event between two tandem 8-bp repeat sequences that flanked the deleted region. Recombination between short tandem repeat (TR) sequences are infrequent in bacteria, particularly when they are present on the chromosome as opposed to the plasmid (Janniere & Ehrlich, 1987). While the number and the length of TR vary drastically among various bacterial species, it seems many pathogenic bacteria encodes numerous TRs that are important for their survival (Moxon, Bayliss, & Hood, 2006). These TRs play an important role in bacterial adaptation, tissue tropism, phase variation and stress tolerance response (Moxon et al., 2006; van Belkum, Scherer, van Alphen, & Verbrugh, 1998; Zhou, Aertsen, & Michiels, 2014). Furthermore, TR sequences are also known to regulate antibiotic resistance. Recently, it has been shown that in Burkholderia thailandensis TRs present in the β- lactamase gene are responsible for change of the substrate spectrum (Yi et al., 2014). TRs are also often use for genotyping pathogenic bacteria, such as Yesinia pestis and Salmonella enterica, which are otherwise genetically homogeneous (Le Fleche et al., 2001; Ramisse et al., 2004).

We found that the mutant grew slowly as compared to the wild type strain. This slow growth is consistent with the phenotypes that are associated with the defects in the L4 protein. Zaman and colleagues have isolated several erythromycin resistant mutants in E. coli that are in the L4 protein (Zaman, Fitzpatrick, Lindahl, & Zengel, 2007). These mutants are all mapped in the highly conserved extended loop region (Fig. 5). They observe that all the L4 mutants grow slowly with the doubling time up to four times that of the wild type doubling time (Zaman et al., 2007). In our case, we only observed about 3.3 times slower growth rate than the wild type, consistent with the previous studies. Moreover, we found that the final biomass was significantly lower in the mutant, suggesting that L. rhanmosus ceases to grow optimally when the L4 protein is mutated. The extended loop of the L4 protein contributes to the lining of the peptide exit channel and functions as a gating mechanism for the exit of the nascent peptides. While the extended loop is not required for proper ribosome assembly (Zengel, Jerauld, Walker, Wahl, & Lindahl, 2003), L4 mutants of the extended loop region often produce atypical ribosome profiles with an additional peak between the standard 30S and 50S peaks (Zaman et al., 2007). This peak, which is 45S, is thought to be incomplete ribosomal particles and accumulation of these aberrant ribosomal particles is responsible for the slow-growth phenotypes (Zaman et al., 2007).

Figure 5: Putative three-dimensional structures of the wild type and the mutant L4 proteins of L. rhamnosus.

The models were generated using the web-based SWISS-MODEL program (Waterhouse et al., 2018). The loop in the wild type L4 protein (LRB) is involved with the binding of erythromycin and this loop is truncated in the RB3S strain.

In addition to the slow-growth phenotype, we found that our mutant was also defective in acid stress response. The mutant also displayed growth defect at low temperature. This growth defect at low temperature could be attributed to the aberrant ribosome biogenesis and assembly (Brandi et al., 2019; Herold, Nowotny, Dabbs, & Nierhaus, 1986; Kaczanowska & Ryden-Aulin, 2007; VanBogelen & Neidhardt, 1990). However, to the best of our knowledge, the acid-sensitive phenotype due to a defective L4 protein has not been previously reported. Currently, we do not know whether this acid-sensitive phenotype is specific for lactobacilli or Gram-positive bacteria. The phenotype could also be a widespread phenotype, which needs to be experimentally determined. We speculate that at low pH, proper ribosomal assembly is drastically affected due to a defective L4 protein that needs to interact with rRNA and other ribosomal proteins during assembly (Kaczanowska & Ryden-Aulin, 2007; Lawrence et al., 2016). Additionally, it is also possible that at lower pH, the peptide transfer reaction is severely affected due to the defective L4 (Johansson et al., 2011).

We noticed that L4 mutant RB3S behaves differently when exposed to antibiotics that target cell wall biosynthesis. When we used a penicillin derivative (mezlocillin), the mutant was more sensitive as compared to the wild type. In contrast, the mutant was more resistant to carbapenem (meropenem) as compared to the wild type. This is surprising since both the antibiotics target penicillin binding protein (PBP) to inhibit cell wall synthesis. We believe that the observed differential results could be due to the changes in the cell wall structure that inhibited permeability of the drugs. Since the mutant strain exhibits growth defect, it should be more sensitive to cell wall targeting inhibitors. However, meropenem was less active in the mutant due to the expression of some factors that interfered with antibiotic action. More studies are needed to investigate whether this assumption is valid or not.

We identified several proteins that are differentially expressed in the mutant. A significant number of proteins that were upregulated in the mutant were related to ribosomal proteins or proteins that interact with the ribosome (Table 3B). This upregulation suggests that the cells are trying to compensate for the defective L4 by inducing the ribosome associated proteins to minimize the effect. On the other hand, proteins that were down regulated in the mutant appeared to be related to metabolism and growth, consistent with the slow-growth phenotype of the mutant. We also found several SAM-methyltransferases whose expressions were down regulated in the mutant. The reason for this observation is currently unclear. At present we do not know whether the differential expression of proteins occurs at the transcriptional or at the translational level. This is because L4 and a few other ribosomal proteins are special types since they display extraribosomal function (Warner & McIntosh, 2009). L4 is specifically extraordinary. It regulates the transcription of the S10 operon, which encodes for 11 ribosomal proteins including L4 (Yates & Nomura, 1980; Zengel, Mueckl, & Lindahl, 1980). This transcriptional regulation is accomplished by L4 stimulation of an attenuator structure at the mRNA upstream of the first initiation codon (Lindahl, Archer, & Zengel, 1983). Furthermore, the L4 protein also regulates RNA degradation by binding to RNase E, leading to substantial changes in mRNA composition in the cell in response to stress (Singh et al., 2009). Thus, it is possible that some of the altered proteins could be due to mRNA degradation and not due to translation.

The L. rhamnosus genome is nearly 47% G+C content. When we searched for the eight-nucleotide sequence, GCGGCGGT, we found that the sequence occurs 78 times and the complement sequence ACCGCCGC occurs 116 times in the LRB genome. When we searched for the seven-nucleotide sequence, GCGGCGG, we found that the frequency of occurrence is 268 (348 for the complement sequence). These frequencies are higher than the random occurrence, indicating that the sequence is over represented in the genome. Over represented sequences are common in prokaryotic genomes and they often play a role in genome stability and organization (Halpern et al., 2007) (Touzain, Petit, Schbath, & El Karoui, 2011). No systematic studies have been done in lactobacilli or other related organisms about the over represented sequences and their role in the genome stability.

L. rhamnosus is widely used as a probiotic organism. As a probiotic strain, L. rhamnosus encounters harsh conditions in the gastrointestinal tract that it constantly needs to fight. Strains that are defective in stress tolerance will have fitness defects and will rapidly be eliminated from the environment. We have isolated two mutants that were defective in stress tolerance response. Two earlier studies have shown that the widely used probiotic strain LGG exhibits a high degree of polymorphism and genome instability (Douillard et al., 2016; Sybesma, Molenaar, van Ijcken, Venema, & Kort, 2013). We found that some of the spontaneously generated mutants display enhanced antibiotic resistance. Therefore, it further raises concerns whether probiotic strains should be closely monitored not only for their efficacy and fitness but also for antibiotic resistance.