Salmonella enterica serovar Typhimurium strain 14028s was used as wild-type in this study. The following knockout mutants derived from this wild type strain were also used: ΔbcsA:kan (Mills et al., 2015), ΔSTM14_1614::cam, ΔSTM14_5387::cam, ΔSTM14_5436::cam (ΔhsdS:cam), ΔSTM14_0418::cam, ΔSTM14_0191::cam, ΔSTM14_5316::cam, ΔSTM14_5327::cam, ΔSTM14_5319::cam, and ΔcsgB:kan (Porwollik et al., 2014). Finally, two Escherichia strains were used: Escherichia coli strain MG1655 and Escherichia fergusonii strain Yara015 (Poudel et al., 2022). Bacteria were grown overnight in Lysogeny Broth (LB) at 37°C shaking. OD₆₀₀ of overnight cultures was measured before dilution of bacteria into egg white.

The experimental setup of all the experiments is described in Figure 1. Unfertilized eggs were obtained from the hen house at the Robert H. Smith Faculty of Agriculture, Food and Environment on the day of laying and stored at 4°C for up to 3 days. The egg white was separated from the egg yolk into a test tube under sterile conditions and mixed in a stirrer for half an hour to homogenize the media. This was directly used for the experiment described in Figure 2. In all other experiments, egg whites were then diluted with autoclaved saline solution (PBS) in a 1:1 ratio and stirred for an additional half hour. This facilitated better mixing of the egg white layers, resulting in a more homogeneous and pipette-able solution. This produced 50% egg white that was used in all other experiments.

In most experiments, bacteria were diluted into egg white or 50% egg white at a concentration of 8*10³ cells per 1 mL. Only for the Transposon sequencing (TnSeq) experiment a concentration of 8*10⁴ cells per 1 mL was used. Compound stocks were prepared at 500 µM. Compounds were mixed at a 1:3 ratio with the bacteria pre-diluted into egg white, such that the final concentration of compounds was 125 µM. For the experiment described in Figure 2, growth in egg white was directly monitored in a plate reader set to 37°C. For all other experiments, samples were incubated with egg white for 24 h at 37°C. 37°C was used, as this is the optimal temperature for Salmonella growth. To determine the relative numbers of bacteria at the end of the 24 h incubation, 20 µL of sample was diluted into 180 µL of LB, and growth at 37°C was monitored by plate reader. To facilitate comparisons between samples, for some experiments the time in minutes till OD₆₀₀ = 0.2 was determined.

To convert differences in the time it took for a sample to reach OD₆₀₀ = 0.2 given in minutes to fold differences in the relative numbers of bacteria in the original sample, we performed a calibration curve experiment. One reason for this experiment is that while samples were diluted into LB it was possible that residual egg white was slowing bacterial growth. Thus, Salmonella wild type was grown over night, and 10-fold dilutions starting at OD₆₀₀ = 1.0 were created in a mixture containing egg white at 5% in LB which represents the final concentration of 50% egg white after the 1:10 dilution into LB performed for the other experiments. This calibration curve experiment was repeated three times. Because lag time is unknown, instead we calculated generation time using the time in minutes to OD₆₀₀ = 0.2 reached by the different 10-fold dilutions. We found that the average generation time in LB containing 5% egg white to be 25.589 ± 3.77 min. We used this to convert differences in the time till OD₆₀₀ = 0.2 in minutes to relative bacterial numbers in the original samples. For example, two samples which reached OD₆₀₀ = 0.2 51 min apart have a difference of two duplications or a 4-fold difference between them in the numbers of bacteria at the end of the 24 h incubation in egg white. Finally, it should be noted that differences in the time till OD₆₀₀ = 0.2 can either reflect differences in bacterial numbers in the original samples or differences in the lag time. For example, if the presence of an amino acid causes membrane damage it might take longer for bacteria to recover and start growing when diluted into LB.

Because a preliminary attempt to create a transposon library on the wild-type background resulted in a low number of transformants, we decided to use as a background a mutant in the host restriction system. A mutation in the hsdRMS system is known to result in high efficiency transformation (O’Callaghan and Charbit, 1990) and commercially available electrocompetent cells are often optimized by introducing mutations in the hsdRMS genes. Plasmid pBAD24 DNA was utilized for a preliminary electroporation experiment comparing several Salmonella mutants in known and putative host restriction systems, which included the following strains; ΔSTM14_1614 (putative DNA/RNA non-specific endonuclease), ΔSTM14_5387 (putative restriction endonuclease), ΔSTM14_5436 (type I restriction enzyme specificity protein, hsdS), ΔSTM14_0418 (DNA restriction enzyme, res), ΔSTM14_0191 (putative restriction endonuclease), ΔSTM14_5316 (putative endonuclease), ΔSTM14_5327 (putative endonuclease), and ΔSTM14_5319 (putative endonuclease). This analysis found the mutant in hsdS was by a large difference the most electrocompetent. Notably, a knockout mutation in hsdS gene increased transformation efficiency up to 220,000 transformants per reaction, compared to only 7,000 transformants obtained in the wild-type background. Of note, the growth inhibition in egg white upon arginine addition phenotype was confirmed for this hsdS mutant strain (data not shown).

To generate the transposon library the ΔhsdS::cam strain was grown overnight in 2 mL LB and then diluted 1:100 into 200 mL 2 x YT broth [16.0 g/L Tryptone, 10.0 g/L Yeast Extract, and 5.0 g/L NaCl]. The culture was incubated at 37°C with shaking at 250 rpm to an OD₆₀₀ of 0.4–0.5. Cells were harvested by centrifugation at 1,000 ×g for 8 min at 4°C and washed 3 times in ½ vol cold 10% glycerol. The final electrocompetent cells pellet was resuspended in 60 µL of cold 10% glycerol and used immediately. Cells were mixed with 0.5 µL of EZ-Tn5™ <KAN-2>Tnp Transposome™ Kit (Lucigen, Cat. No. TSM99K2). After incubation in ice for 10 min, the cells were electroporated in a 1 mm electrode gap cuvette using a Bio-Rad MicroPulser set to 2.5 kV. Electroporated cells were immediately resuspended in 500 µL of pre-warmed Super Optimal broth with Catabolite repression (SOC) medium [0.5% Yeast Extract, 2% Tryptone, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl2, 10 mM MgSO4, 20 mM Glucose]. After incubation for 90 min at 37°C with shaking, the volume was adjusted by a dilution factor of 20 in SOC medium to create 10 mL. 87 LB agar plates with kanamycin (100 μg/μL) were plated with 100 µL each. After 18 h of incubation, estimated colony yield was calculated. 1 mL of LB was added to each plate and colonies were scraped and pooled together into 50 mL tubes. Bacterial suspension from all tubes was vortexed, and transferred to a sterile bottle. 80% glycerol was added to a final concentration of 25%. Finally, the Tn5 transposon library was aliquoted to 1 mL microcentrifuge tubes and stored at −80°C. The library was estimated to contain 260,000 mutants.

The protocol was similar to the basic experiment described above with the following modifications. In short, a tube of the Tn5 library was thawed on ice and 100 µL were diluted into 10 mL LB with kanamycin (50 μg/mL) to a final bacterial concentration of 10⁸ CFU/mL. The library was incubated at 37°C shaking at 250 rpm for 1 h to initiate log phase. To prepare the library input stock, OD₆₀₀ was measured and the library was diluted to OD₆₀₀ = 1. The library was centrifuged and the pellet was resuspended in PBS and two 700 µL samples were taken into separate tubes and immediately placed in −80°C for safe keeping. These two samples are the T0 input - the initial population of the Tn5 library (before selection). Aliquots of 2 µL from the PBS resuspended library input stock were added to 20 mL PBS to create a library dilution of 8*10⁴ CFU/mL for 2 biological repeats. Four sterile flasks were prepared, representing 2 biological replicates for 2 treatments: DDW control 1, DDW control 2, L-arginine treatment 1 and L-arginine treatment 2. Each flask contained a 100 mL mixture composed of: (a) 7.5 mL Tn5 library–from library dilution 1 or 2. (b) 67.5 mL egg white + PBS (1:1) mixture. (c) 25 mL L-arginine for a final concentration of 125 µM or DDW carrier as a control. Flasks were aliquoted into separate 5 mL tubes with 1 mL mixture in each tube. The first replicate consisted of 100 tubes for the L-arginine treatment and another 100 for the control. In the second replicate, the pool consisted of 50 culture tubes, rather than 100, due to technical complications. A total of 300 tubes were incubated for 24 h at 37°C shaking at 250 rpm. This large number of tubes was required to allow testing of a broad number of the transposon mutants and avoid a bottleneck. After incubation, tubes originating from the same flask were pooled into 50 mL tubes and centrifuged at 7197 RCF (maximum speed) for 5 min to eliminate the egg white mixture. The pellet of each tube was resuspended in 10 mL of LB and incubated for 1 h at 37°C shaking. This step was required in order to differentiate live bacteria from dead bacteria. By allowing live bacteria to replicate their representation compared to the DNA of dead cells increased. This step also improved DNA yield. Finally, samples were washed, resuspended in 700 µL LB and stored at −80°C for DNA extraction.

The DNA extraction protocol was adapted from Stevenson and Weimer (2007). DNA was extracted from 6 samples: control DDW 1, control DDW 2, L-arginine treatment 1, L-arginine treatment 2, input T0 1 and input T0 2. Samples were disrupted with 0.1 mm glass beads in the presence of Tris-saturated phenol following phenol-chloroform extraction. DNA concentration was measured using a Qubit 2.0 fluorometer and dsDNA Quantification Assay Kit and samples were kept at 4°C.

The protocol for amplification of transposon junctions using the semi-arbitrary PCR strategy was adapted from Christen et al. (2011) with a few modifications. In a first-round of PCR, a transposon specific primer (Ez tn5 <KAN2> FP, Table 1) was used with a semi-arbitrary primer (Rnd1-PE-arb1). The semi-arbitrary primer contains a defined 3′ penta-nucleotide sequence that anneals on average every 233 bp in the Salmonella S14028 genome, an interspace of 10 bp arbitrary sequence and a 5’ PE 2.0 sequence that serves as a binding site for the subsequent PCR. The first round of PCR was performed with the following program: (1) 94°C for 3 min, (2) 94°C for 30 s, (3) 42°C for 30 s, slope −1°C per cycle, (4) 72°C for 1 min, (5) go to step 2, 6 times, (6) 94°C for 30 s, (7) 58°C for 30 s, (8) 72°C for 1 min, (9) go to step 6, 25 times, (10) 72°C for 3 min. The 50 µL PCR reaction consisted of 25 µL Platinum™ Hot Start PCR Master Mix (Invitrogen, Carlsbad, CA, United States), 1 µL Ez tn5 <KAN2> FP primer (10 µM), 1 µL Rnd1-PE-arb1 (10 µM), 10 µL Template DNA (50–80 ng), and 13 µL PCR grade dH2O. PCR products were purified using Zymo Clean and Concentrator-5 kit (Zymo Research, Irvine, CA) and eluted with 12 µL DDW. Products served as a template for a second round PCR reaction using Illumina paired-end primers.

The second reaction was composed of a flanking P5-Tn5 barcoded primer specific to the transposon, downstream to the transposon specific primer used in the 1st PCR reaction to limit amplification of non-specific products, and a PE P7 primer 2.0. To determine the optimal cycle number for amplification at the second PCR reaction and by that minimize bias introduction due to over amplification of specific fragments (Gallagher, 2019), suggests tracking amplification by using real-time PCR. Thus, prior to the second PCR reaction, a real-time PCR was carried using the following parameters: (1) 94°C for 1 min, (2) 94°C for 30 s, (3) 64°C for 30 s, (4) 72°C for 1 min, scan, (5) go to step 2, 20 times, followed by a dissociation curve reaction. The 25 µL qPCR reaction consisted of 12.5 µL 2x qPCRBIO Fast qPCR SyGreen Blue Mix (PCR Biosystems, London, UK), 0.5 µL P5 barcoded primer (10 µM), 0.5 µL Round2-PE-P7 primer (10 µM), 1 µL 1^st round PCR product, and 10.5 µL PCR grade dH2O. As a result of this qPCR the optimal number of cycles was determined as 14 for the DDW control and L-arginine treatment samples and 9 for the T0 input samples. Second-round PCR was preformed using the following parameters: (1) 94°C for 3 min, (2) 94°C for 30 s, (3) 64°C for 30 s, (4) 72°C for 1 min, (5) go to step 2, 9 or 14 times (as determined for each sample), (6) 72°C for 3 min. The 50 µL PCR reaction consisted of 25 µL Platinum™ Hot Start PCR Master Mix, 1 µL P5 barcoded primer (10 µM), 1 µL Round2-PE-P7 primer (10 µM), 10 µL 1^st round PCR product, and 13 µL PCR grade dH2O. Finally, second round PCR products were purified using Zymo Clean and Concentrator-5 kit and eluted in 25 µL DDW. Library molarity was calculated using Qubit measurements and average fragment size reported by TapeStation (Agilent, Santa Clara, United States). Equimolar amounts from each sample were pooled together to 4 nM and sent for single-read sequencing with 150 cycles on an Illumina MiSeq platform (Genomic Technologies Facility of the Hebrew University of Jerusalem, Givat Ram, Jerusalem).

Salmonella enterica subsp. Enterica serovar Typhimurium str. 14028S, assembly ASM2216v1, was downloaded from ENA database (Accession: GCA_000022165) and used as a reference genome. Raw fastq files were demultiplexed using Cutadapt (Martin, 2011) based on the sample inline barcode and 6 nucleotides from the transposon 5’ end sequence (TCAGGG), allowing no mismatches. Fastq files were processed through the BioTradis analysis pipeline (Barquist et al., 2016). The pipeline was executed on Ubuntu 20.04. Reads were filtered based on the presence of the transposon tag (TTGAGATGTGTA) in the beginning of the reads, allowing for 1 mismatch. Three comparisons were carried separately: (1) input T0 vs. control DDW, (2) input T0 vs. L-arginine treatment, and (3) DDW control vs. L-arginine treatment. Genes with low read counts were filtered using a threshold of 10. p-values were corrected for multiple testing using Benjamini–Hochberg method. Genes with corrected p-value (Q-value) ≤ 0.05 were considered significant.

As a first step, we confirmed that the strain used, S. enterica serovar Typhimurium strain 14028s (Salmonella), is able to survive and grow in the egg white environment. As a comparison to Salmonella growth, the growth in egg white of additional phylogenetically closely related bacteria, E. coli strain MG1655 and a poultry derived E. fergusonii (Yara15) strain (Poudel et al., 2022), was examined. Overnight cultures of all three bacteria were diluted into egg white at a concentration of 8*10³ cells per 1 mL. This low concentration was used so that the bacteria themselves would not substantially modulate the egg white environment for the duration of the experiment, i.e., by degrading egg white proteins or changing the pH through their own metabolism. Growth in egg white was measured overnight. Compared to egg white-alone background readings, it seemed that the two Escherichia species did display some growth (Figure 2). This limited growth might be a result of residual media carried over from the overnight culture. However, Salmonella displayed much more significant growth. This shows that Salmonella encodes systems enabling it to overcome at least some of the stressors that this environment uses to inhibit bacterial growth. That being said, compared to growth in LB which can reach OD₆₀₀ = 0.5 in the plate reader (results not shown) this represents limited growth. To conclude, the S. enterica serovar Typhimurium strain utilized here not only survives in the egg white environment but can also grow at least to a limited degree.

To test the hypothesis that adding compounds to an environment might limit bacterial growth, we chose the 20 L-amino acids. This is because L-amino acids are found in all living cells, and are taken up as nutrients by both animals and bacteria including Salmonella. For this set of experiments, Salmonella was incubated for 24 h in egg white at 37°C, each time with a different L-amino acid at a final concentration of 125 µM. After incubation, samples were diluted into LB and growth was recorded to determine how many bacteria survived the 24 h treatment. We chose OD₆₀₀ = 0.2 because at this OD Salmonella were in logarithmic growth. Differences in the time to reach OD₆₀₀ = 0.2 compared to the control are therefore indicative of differences in the numbers of surviving bacteria after egg white incubation. Among the amino acids, only L-arginine was found to delay Salmonella growth in a statistically significant manner (Figure 3). L-arginine exposure resulted in a 57.5 min delay in the time to reach OD₆₀₀ = 0.2 compared to the double distilled water (DDW) control. Considering our measured generation time of 25.589 min, this equals a 4.75-fold reduction in the number of bacteria surviving in egg white. Interestingly, L-cysteine displayed a much bigger inhibition, 117.5 min, which is equivalent to a 24.11-fold reduction. However, because of a large variation in L-cysteine measurements, this was not statistically significant (p = 0.0587). To conclude, these results show that the addition of L-arginine and likely also L-cysteine inhibits the growth of Salmonella in the egg white environment, even though these are nutrient type compounds.

Of note, environmental L-arginine is known as a signal used by Salmonella to modulate cellular cyclic-di-GMP (CDG) levels and downstream cellulose synthesis and secretion (Mills et al., 2015). CDG is a bacterial second messenger considered the switch controlling biofilm formation versus motility (Mills et al., 2011). Furthermore, cellulose is the major exopolysaccharide of Salmonella and a major component of the Salmonella biofilm. Thus, we hypothesized it was possible that biofilm formation was involved in the inhibition of Salmonella growth upon L-arginine addition. Because salicylic acid is known to effect CDG and cellulose levels in the opposite direction to L-arginine (Mills et al., 2015), we tested if salicylic acid would increase Salmonella’s growth in egg white. In addition, two knockout mutants of Salmonella were tested for growth and survival in the presence of L-arginine, a bcsA knockout strain unable to produce cellulose and a csgB knockout strain unable to produce curli fibers that are also a major component of the Salmonella biofilm (Jonas et al., 2007). On the one hand, the presence of salicylic acid did seem to have a small positive effect on the growth of Salmonella in egg white (Figure 4). However, L-arginine inhibition still occurred in the bcsA knockout strain background in a statistically significant manner. Due to large variation in measurements when using the csgB knockout strain background, only a trend was observed (Figure 4). To conclude, it seems that cellulose, but likely also curli and CDG, are not part of the mechanism of action of L-arginine inhibition on Salmonella growth in egg white.

We hypothesized that arginine was possibly transported into Salmonella cells where it was dysregulating gene expression causing decreased bacterial survival. Thus, a TnSeq analysis might identify genes which are required for arginine inhibition of Salmonella in the egg white environment. Furthermore, TnSeq might identify genes which are involved in whatever stress response is produced due to the presence of arginine. Therefore, a TnSeq analysis of Salmonella exposed to egg white either in the presence or absence of arginine was performed. Furthermore, both of these were compared to the input Salmonella transposon library. A comparison of transposon insertions after exposure to egg white or to egg white with arginine, to the input library, identified 226 and 220 genes respectively which were significantly differentially represented (Figure 5). Furthermore, the vast majority of these genes had a lower representation than in the input. This supports the notion that egg white is a harsh environment for Salmonella, and that many transposon insertions which are tolerated in rich media become essential in egg white. Importantly, a direct comparison of gene insertions between Salmonella exposed to egg white with arginine to those exposed only to egg white identified just three genes which were differentially represented in a statistically significant manner (Figure 5). This implies that either the inhibition by arginine was not a large stress on Salmonella in comparison to egg white itself, or that arginine inhibition could not be countered by knocking out Salmonella genes by transposon insertions or both.

As a first step in analyzing TnSeq data, genes required for Salmonella survival in egg white were characterized. To stringently identify transposon insertions into genes that effect survival in egg white the data from both arginine and control, DDW treated, samples were used. A gene was considered affected by egg white if it was significantly affected (q < 0.05) in both the arginine vs. input comparison as well as the control DDW vs. the input. This resulted in the identification of 82 high confidence genes. To expand the analysis a bit more, genes that were significant in one comparison, either arginine vs. input or control vs. input, and had a significant p < 0.05 (statistical significance before multiple comparison correction) in the other comparison were also included. This resulted in the identification of another 44 genes with lower confidence. We also verified that the effect was in the same direction, i.e., increased in both data sets in comparison to the input, or reduced in both data sets in comparison to the input. All of the high confidence genes were affected in the same direction. However, eight genes of the lower confidence genes were not affected in the same direction by egg white and were removed. This produced a set of 118 genes that were differentially represented after egg white treatment compared to the input library (Table 2). Notably, only one gene, stm14_4543, had a higher representation after egg white treatment compared to the input library whereas all the other 117 genes had a lower representation in egg white compared to the input. Among the genes identified as required for Salmonella growth are genes required during iron starvation, biotin metabolism genes, genes required for the transport and metabolism of amino acids, purine and other nutrients, stress response genes, genes required for outer membrane maintenance, and genes required for DNA repair and recombination. This further supported the notion that egg white is a harsh environment requiring the function of many genes.

An analysis of transposon insertions deferentially distributed between egg white with L-arginine and egg white control, identified only three genes that were statistically significant (Table 3). The three genes were citB, stm14_0668, and stm14_0670. However, the known functions of these genes, mannose transport, and citrate utilization do not seem to relate to arginine transport and metabolism nor to a possible stress acting on the bacterial cell.

Next, specific molecular functions were examined to better characterize the effect of arginine on Salmonella in egg white. One such function is arginine transport. It can be hypothesized that if arginine inhibition functions by causing dysregulation of bacterial gene expression, transposon insertions into genes involved in arginine transport should confer protection. One such insertion, into the gene argT was expended in the arginine treatment samples compared to the control but was not statistical significance. Moreover, insertions into other genes of the same transport machinery, hisQMP, seems to not have been affected by the presence of arginine (Table 3). Another function analyzed is genes involved in arginine metabolism. It was hypothesized that uptake and utilization of arginine as a nutrient source might affect insertions in genes involved in arginine metabolism. Three genes involved in arginine metabolism were affected by the presence of arginine, but below statistical significance. Other genes involved in arginine metabolism did not seem to have been affected by the presence of arginine (Table 3). To conclude, it was unclear if arginine was transported into Salmonella cells for its inhibitory function.

Next, the database was searched for gene insertions that might allow the characterization of the molecular mechanism of arginine inhibition. Interestingly, insertions in rcsF were deleterious in the presence of arginine, but not statistically significantly so. A search of other systems involved in cell integrity, such as lipopolysaccharide biosynthesis, peptidoglycan biosynthesis, exopolysaccharide biosynthesis, and cationic antimicrobial peptide (CAMP) resistance found that most genes were not affected by arginine presence (Table 3). Finally, it was confirmed that insertions into genes of the cellulose pathway were not affected by arginine. To conclude, apart from a possible involvement of rcsF, no stress related pathways were identified.

To determine if the mechanism of action of L-arginine and L-cysteine includes interaction with Salmonella biological systems or not, we turned to their D-isomers, D-arginine and D-cysteine. This is because most enzymes, receptors, and transporters have evolved to bind the L-isomers but not D-isomers of the 20 amino acids (Genchi, 2017). For example, proteins are made with L-isomers and not D-isomers. Both D-isomers, D-arginine and D-cysteine, were found to inhibit Salmonella growth in 50% egg white (Figure 6). These results imply but do not prove that the inhibition of Salmonella growth by cysteine and arginine do not include bacterial metabolism or signaling as part of their mechanism of action but is connected to chemical interactions of arginine and cysteine with the bacterial membrane or egg white environment itself.