Transcriptomic Response of Listeria monocytogenes and Salmonella enterica Typhimurium to Power Ultrasound and Chlorine Treatments

Abstract

Listeria monocytogenes and Salmonella enterica Typhimurium are leading causes of foodborne illness in the United States and frequently implicated in produce outbreaks. Conventional decontamination methods, such as cold-water washes with chlorine, have limited antibacterial efficacy and environmental sustainability. Power ultrasound has emerged as a promising non-thermal alternative, but the molecular mechanisms remain insufficiently elucidated. This study evaluated transcriptomic responses of L. monocytogenes and S. enterica Typhimurium to (i) ultrasound (20 kHz), (ii) chlorine (50 ppm), and (iii) combined ultrasound + chlorine treatments. RNA-seq analysis identified differentially expressed genes, as well as enriched Gene Ontology and KEGG terms. Results showed that ultrasound and chlorine triggered distinct transcriptomic responses. L. monocytogenes exhibited broad transcriptional shifts under ultrasound, including significant upregulation of phosphotransferase system components and central metabolism. Chlorine alone induced a narrower response, with fewer differentially expressed genes clustering into limited functional categories. In contrast, the combined ultrasound + chlorine treatment elicited the strongest response in S. enterica Typhimurium, with enrichment of multiple energy- and metabolism-related pathways, including the citrate cycle, carbon metabolism, and microbial metabolism in diverse environments. These findings provide new insights into ultrasound-triggered responses in foodborne pathogens and may inform development of optimized ultrasound-based hurdle sanitization strategies for produce safety.

1. Introduction

It is estimated that over 48 million cases of foodborne illness occur annually in the United States due to the consumption of contaminated food [1]. Common symptoms include diarrhea, fever, nausea, abdominal pain, and vomiting [2]. While many cases are self-limiting, certain foodborne infections can lead to severe complications or death, particularly among vulnerable populations such as expectant mothers, older adults and the immunocompromised [3]. Foodborne pathogens such as Listeria monocytogenes (L. monocytogenes) and Salmonella enterica (S. enterica) are among the top five foodborne pathogens causing death [4]. L. monocytogenes causes approximately 1600 infections and 260 deaths, while S. enterica is responsible for an estimated 1.35 million infections, 26,500 hospitalizations, and around 420 deaths in the United States [5,6,7].

Between 2020–2025, L. monocytogenes and S. enterica were collectively responsible for 45 out of 66 (~68%) reported foodborne illness outbreaks in the U.S. [8]. Out of those 45 outbreaks, 18 (~40%) were linked to contaminated fresh produce [8]. Notably, fresh produce commodities such as herbs, seeded vegetables, and sprouts were disproportionately prevalent in human foodborne outbreaks based on their consumption rates by the U.S. population [9].

The safety of fresh produce has become an imminent concern due to the lack of a “killing” step to completely inactivate pathogens during the post-harvest minimal processing prior to retail. Other food categories may include validated microbial inactivation steps such as pasteurization in dairy products or thermal sterilization in canned foods. However, fresh fruits and vegetables are typically subjected only to basic washing with chlorinated water. This minimal processing is designed to maintain the desirable sensory and nutritional qualities of produce. However, it has limited efficacy in controlling and preventing pathogen contamination in the finished products. Throughout the production continuum, from cultivation and harvest to postharvest handling and distribution, produce can be exposed to contaminated soil, irrigation water, equipment, and handling practices, increasing the risk of contamination by foodborne pathogens such as L. monocytogenes and S. enterica [10,11,12,13].

Current industry-standard decontamination practices for fresh produce primarily rely on chemical sanitizers, particularly chlorine-based solutions and peracetic acid (PAA) [14,15]. While these sanitizers offer moderate antimicrobial efficacy, they have several disadvantages. For instance, chlorine-based treatments can generate potentially hazardous by-products and require large volumes of water, raising environmental and public health concerns [16,17]. In some regions, the use of chlorine has been restricted or prohibited due to safety considerations related to residual compounds [18,19,20]. Moreover, the continued occurrence of foodborne outbreaks and product recalls, despite the routine use of these chemical sanitizers, emphasizes their limited effectiveness and highlights a need for improved or new decontamination technologies.

One particular non-thermal approach that has been gaining in popularity as a supplement to traditional washing methods is power ultrasound technology [21,22,23,24,25,26]. By using high-frequency sound waves, this technology can physically disrupt bacterial cell membranes and cause leakage of cytosols through a process known as acoustic cavitation [27]. The localized areas of high pressure and temperature generated by the acoustic cavitation can also facilitate physical disruptions of microbial biofilms and weaken bacterial adhesion and attachment of fresh produce surfaces, thereby rendering attached pathogen cells more vulnerable to disinfection [27]. Previous studies have demonstrated that power ultrasound, particularly when combined with sanitizers such as chlorine or organic acids, can achieve greater microbial reductions on produce surfaces than sanitizers alone [24,28]. This technology is environmentally friendly, requires less water and produces fewer harmful byproducts than traditional chemical sanitizers, making it a promising option for sustainable food safety practices [18,28,29,30].

Beyond several known physical effects of acoustic cavitation on bacterial cells, the “mode of action” of power ultrasound at the molecular level on bacterial pathogens such as L. monocytogenes and S. enterica remains largely unexplored. Bacterial pathogens can deploy stress response mechanisms to endure environmental challenges, including oxidative damage generated by chlorine and power ultrasound. Such mechanisms may involve the regulation of genes related to stress protection, biofilm formation, and metabolic adaptation, which collectively play a role in enabling bacterial transmission and persistence on fresh produce surfaces.

In this study, we used high-throughput RNA sequencing (RNA-seq) to investigate the transcriptomic responses of L. monocytogenes and S. enterica to ultrasound, chlorine and the combination of ultrasound + chlorine treatments. By comparing differentially expressed genes under sublethal treatments, we were able to identify several important regulatory and metabolic pathways associated with L. monocytogenes and S. enterica in response to power ultrasound treatment. Our findings are valuable for the optimization of ultrasound treatment parameters and designing more effective hurdle interventions for pathogen control and prevention to improve fresh produce safety.

2. Materials and Methods

2.1. Strains and Culture Conditions

L. monocytogenes LS810 (isolated from cantaloupe, serotype 1/2b), and S. enterica Typhimurium LT2 were used in this study. Each strain was first individually cultured in 5 mL of Tryptic Soy Broth (TSB; Becton, Dickinson and Co., Sparks, MD, USA) at 37 °C for 16–18 h. Sub-cultures were then prepared by adding 100 µL of each culture to 25 mL of TSB, in duplicate (50 mL total), and incubating at 37 °C for approximately 4–5 h until an optical density of 0.4 at a wavelength of 600 nm (OD₆₀₀) was achieved, which corresponds to the cells in mid-logarithmic growth phase. The OD₆₀₀ was measured using a spectrophotometer (Genesys 30 Visible Spectrophotometer, Thermo Scientific, Waltham, MA, USA). Once cultures reached an OD₆₀₀ of 0.4, cells were pelleted via centrifugation at 8000 × g for 10 min at 4 °C. Cell pellets (from each of the 25 mL cultures) were resuspended in 2.5 mL of Butterfields’s Phosphate Buffer (BPB, pH 7.4) and combined (5 mL total) to achieve concentrations of approximately 9–10 log colony forming unit (CFU)/mL. The concentrations of the inocula were determined by serial dilution in BPB and plating onto Tryptic Soy Agar (TSA; Becton, Dickinson and Co.). Agar plates were incubated at 37 °C for 24 h prior to enumeration.

2.2. Treatment Conditions

Each bacterial strain was subjected to four separate treatments: (i) water (control), (ii) chlorine, (iii) power ultrasound, and (iv) a combination of chlorine and power ultrasound. To prepare the chlorine treatment, sodium hypochlorite (12% available chlorine; Spectrum Chemical, New Brunswick, NJ, USA) was added to sterile water to reach a final concentration of 50 ppm; the concentration of free chlorine was measured immediately prior to treatment using a pocket colorimeter (DR300, Hach Co., Loveland, CO, USA) according to the manufacturer’s instructions. A chlorine concentration of 50 ppm was selected to represent the minimum level permitted for fresh produce washing in the U.S. [26], and this concentration also provided sublethal stress conditions suitable for transcriptomic profiling. The water and chlorine solution were adjusted to 4 °C prior to all treatments. For the water and chlorine treatment, 5 mL of the L. monocytogenes or S. enterica Typhimurium inocula was added to 45 mL of sterile water or the 50 ppm chlorine solution, respectively, and mixed by hand using a sterile tongue depressor. After 1 min of treatment, 2 mL of the inoculum-treatment solution was immediately removed and added to 4 mL of RNAprotect Bacteria Reagent (Qiagen Inc., Germantown, MA, USA) and vortexed to mix to stabilize the bacterial RNA. A 1 min treatment was chosen based on prior studies to ensure that the treatment remained sublethal and could yield sufficient RNA for downstream transcriptomic analysis [26,31]. For the power ultrasound (ultrasound) and the chlorine + ultrasound combination treatment, 5 mL of the L. monocytogenes or S. enterica Typhimurium inocula was added to 45 mL of sterile water or the 50 ppm chlorine solution, respectively, mixed by hand using a sterile tongue depressor, and treated with a probe power ultrasound unit at a frequency of 20 kHz (F550 Sonic Dismembrator Ultrasonic Homogenizer ((500 W, 100% amplitude), Fisher Scientific, Waltham, MA, USA) on ice. Ultrasound was applied using a probe unit at 20 kHz, the lowest frequency classified as ultrasound, which has previously been shown to induce cavitation-driven bacterial stress in food safety applications [24,26,31,32,33]. After 1 min of treatment, 2 mL of the inoculum-treatment solution was immediately removed, and the RNA was stabilized as previously described. One trial consisted of both pathogens undergoing the four different treatments. Four trials for each pathogen were conducted, each with independent inocula, and a total of four biological replicates were conducted for each pathogen-treatment combination (32 total samples).

2.3. Bacterial RNA Extraction

Total RNA was extracted from all treated samples using the RNeasy Plus Mini Kit with gDNA Eliminator Spin Columns (Qiagen) according to the manufacturer’s instructions. Total RNA was eluted in 30 µL of RNase-free water. The RNA concentration was determined using a fluorometer (Qubit 3.0, Life Technologies, Carlsbad, CA, USA) with the RNA HS Assay Kit (Life Technologies). The quality of the RNA was examined using the Bioanalyzer with the RNA 6000 Pico Kit (2100 Bioanalyzer Instrument, Agilent, Santa Clara, CA, USA) according to the manufacturer’s instructions. Samples with RIN scores > 7 and electrophoretic profiles that showed clear rRNA bands with no evidence of degradation were used for subsequent library construction. All samples were processed in a single batch to minimize potential batch effects.

2.4. Library Construction and RNA-Sequencing

The library for sequencing was constructed using the Stranded Total RNA Prep with Ribo-Zero Plus (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The RNA UD Indexes Set A (Illumina) were used for indexing. The concentration of each sample was determined as previously described using a fluorometer (Qubit 3.0, Life Technologies) with the High Sensitivity DNA Kit (Life Technologies). The quality of the DNA was examined using the Bioanalyzer (2100 Bioanalyzer Instrument, Agilent, Santa Clara, CA, USA) with the DNA 1000 Reagent Kit according to the manufacturer’s instructions. To sequence the 32 samples, eight libraries consisting of four pooled samples each were prepared; each library contained one trial for one pathogen. Each library was normalized to 4 nM and the concentration was verified using a fluorometer (Qubit 3.0, Life Technologies) with the High Sensitivity DNA Kit (Life Technologies). Each library was denatured and diluted to 8 pM according to the Denature and Dilute Libraries for the MiSeq System documentation (Illumina). Denatured 20 pM PhiX (Illumina) library was added at 5% to each library prior to sequencing on a MiSeq (Illumina) using 600 cycles of V3 chemistry.

2.5. Pre-Processing and Quality Control of RNA-Sequencing Data

FastQC v0.12.1 [34] and MultiQC v1.16 [35] were used to evaluate the quality of raw Illumina paired-end reads (using default settings; Supplementary Tables S1 and S2, Supplementary Data S1 and S2). Each set of paired-end Illumina reads underwent trimming and quality filtering. For (i) L. monocytogenes, reads were trimmed using Trim Galore v0.6.10 [36] with the “--fastqc”, “--paired”, “--gzip”, and “--stranded_illumina” options enabled (to run Trim Galore’s implementation of FastQC on the resulting trimmed reads, to specify that the reads were paired-end reads, to compress the resulting trimmed reads via gzip, and to trim Illumina stranded mRNA/Total RNA adapters, respectively; default settings were used for all remaining parameters) [37]. MultiQC was used to aggregate all FastQC reports produced by Trim Galore (default settings). The resulting trimmed paired-end L. monocytogenes reads were determined to be of adequate quality (Supplementary Table S3, Supplementary Data S3) and were used in subsequent steps.

For (ii) S. enterica Typhimurium, raw reads underwent trimming and quality control with Trim Galore and FastQC/MultiQC as described above for L. monocytogenes, respectively; however, the quality of the S. enterica Typhimurium reads were poor, even after trimming (particularly for trailing bases; Supplementary Table S4, Supplementary Data S4). As such, S. enterica Typhimurium reads underwent two separate rounds of (pre-)trimming and quality filtering: (a) first, S. enterica Typhimurium reads were pre-trimmed using fastp v0.23.4 [38] with the following settings: “--detect_adapter_for_pe” (to auto-detect adapters for paired-end reads), “--length_required 36” (to discard reads shorter than 36 bp), “--trim_poly_x” (to enable polyX trimming for 3′ ends), “--correction” (to enable base correction in overlapped regions), “--cut_front”, “--cut_front_window_size 4”, “--cut_front_mean_quality 20” (to drop bases in a front-to-tail sliding window of 4 bp if mean quality < 20), “--cut_right”, “--cut_right_window_size 4”, “--cut_right_mean_quality 20” (to drop bases in/to the right of a front-to-tail sliding widing of 4 bp if mean quality < 20); (b) following trimming with fastp, a second round of trimming was performed using TrimGalore as described previously for L. monocytogenes. FastQC and MultiQC were used as described above to assess the quality of S. enterica Typhimurium reads after (a) fastp pre-trimming and (b) Trim Galore trimming (Supplementary Tables S5 and S6, Supplementary Data S5 and S6). The final trimmed paired-end S. enterica Typhimurium reads were deemed to be of adequate quality (Supplementary Table S6, Supplementary Data S6) and were used in subsequent steps.

2.6. Read Mapping and Quantification

The genomes of (i) L. monocytogenes str. MOD1_LS810 (NCBI GenBank Assembly accession GCA_016306305.1) and (ii) S. enterica str. LT2 (NCBI RefSeq Assembly accession GCF_000006945.2) were used as references for the L. monocytogenes and S. enterica Typhimurium datasets, respectively. For both the L. monocytogenes and S. enterica Typhimurium datasets, the “bwa index” command in BWA v0.7.17-r1188 [39,40] was used to index the reference genome; each set of trimmed paired-end reads was then mapped to the respective reference genome using “bwa mem” (default settings). The “samtools sort” command in SAMtools v1.12 [41] was used to create a sorted BAM file (i.e., a BAM file sorted by leftmost coordinates, created using default settings), and the resulting sorted BAM file was indexed using “samtools index” (default settings). The total number of mapped reads was obtained for each sample using the following Bash command: “samtools view −F 0 x 4 sorted.bam | cut −f 1 | sort | uniq | wc −l” (where “sorted.bam” corresponds to the sorted BAM file; Supplementary Tables S7 and S8).

The “featureCounts” function in the Rsubread v2.18.0 package [42] was used for read quantification, using R v4.4.0 [43], each sorted BAM file as input, and the following parameters: (i) the respective reference genome’s GFF file provided via NCBI (supplied via the “annot.ext” parameter, with “isGTFAnnotationFile” set to “TRUE”; Supplementary Tables S9 and S10, Supplementary Data S7 and S8); (ii) “gene” features used as the GTF feature type (GTF.featureType = “gene”); (iii) “locus_tag” elements used as the GTF attribute type (GTF.attrType = “locus_tag”); (iv) fractional counts assigned to multi-mapping reads (countMultiMappingReads = TRUE, fraction = TRUE); (v) paired ends used (isPairedEnd = TRUE); (vii) strand-specific read counting, with reverse-strandedness specified (strandSpecific = 2). This resulted in a matrix of read counts, with genetic loci as rows and samples as columns (Supplementary Tables S11–S14). Plots and tables summarizing read mapping and quantification results were constructed in R using the following packages: ggplot2 v3.5.1 [44], stringr v1.5.1 [45], reshape2 v1.4.4 [46], ape v5.8 [47], khroma v1.14.0 [48] (Supplementary Figures S1–S4, Supplementary Tables S15 and S16). For all subsequent steps, raw read counts mapped to protein-coding genes (i.e., “CDS” features per the reference genome’s GFF file) were used unless otherwise specified (referred to hereafter as “raw count matrices”; Supplementary Tables S12 and S14).

2.7. Exploratory Data Analyses

The variance stabilizing transformation (VST) in DESeq2 v1.44.0 [49] was applied to each of the L. monocytogenes and S. enterica Typhimurium raw count matrices (see Section 2.6 “Read mapping and quantification”). Briefly, each raw count matrix was supplied as input to the “DESeqDataSetFromMatrix” function, alongside (i) a matrix of conditions/treatments (e.g., whether a sample was treated with chlorine, ultrasound, both chlorine and ultrasound, or water), and (ii) no specified design (design = ~1); the resulting object was passed to the “DESeq” function, followed by the “vst” function to obtain a matrix of DESeq2-transformed, normalized counts for all protein-coding genes (referred to hereafter as the “VST-transformed count matrix”), using default parameters.

Each VST-transformed count matrix was supplied as input to the pvclust function (from pvclust v2.2-0) [50] to construct a dendrogram, using average linkage (method.hclust = “average”), Euclidean distances (method.dist = “euclidean”), and 1000 bootstrap replicates (nboot = 1000). Plots were constructed using dendextend v1.18.1 [51]. To assess the extent to which choice of linkage method affected clustering results, the “hclust” function in R’s stats package was used to cluster each transposed VST-transformed count matrix, using one of eight linkage methods (“average”, “complete”, “centroid”, “median”, “single”, “ward.D”, “ward.D2”, “mcquitty”), and the “corrplot” function in corrplot v0.95 was used to plot results (Supplementary Figures S5 and S6) [52].

In addition to the dendrograms, all samples were visualized in two-dimensional space using (i) uniform manifold approximation and projection for dimension reduction (UMAP) and (ii) principal component analysis (PCA). For (i) UMAP and (ii) PCA, the VST-transformed count matrix was supplied as input to the (i) “umap” and (ii) “prcomp” functions from R’s (i) umap v0.2.10.0 [53] and (ii) stats packages, respectively (default settings for umap; parameters “scale = FALSE” and “center = TRUE” for prcomp). The resulting (i) embedding coordinates (UMAP) and (ii) scores (the first three columns of the “x” matrix produced by prcomp) were plotted in R using the ggplot2 v3.5.1, ggpubr v0.6.0 [54], ggrepel v0.9.6 [55], viridis v0.6.5 [56], khroma v1.14.0, and gridExtra v2.3 [57] packages (Supplementary Figures S7–S10).

2.8. Identification of Differentially Transcribed Genes

For each data set (i.e., L. monocytogenes and S. enterica Typhimurium), differentially transcribed genes were identified for each condition (i.e., whether a sample was treated with chlorine, ultrasound, or both chlorine and ultrasound) relative to controls (samples exposed only to water) using DESeq2. Briefly, the “DESeqDataSetFromMatrix” function was used to create an object as described above (see Section 2.7 “Exploratory data analyses” above), but with “design = ~group” (where “group” corresponds to the four aforementioned conditions/controls); the resulting object was passed to the “DESeq” function, and the “results” function was used to identify differentially transcribed genes for each of the three treatments (i.e., ultrasound, chlorine, and both chlorine and ultrasound) relative to the control (i.e., water with no treatment applied). Genes were considered to be differentially transcribed if (i) the adjusted p-value produced by DESeq2 was <0.05 and (ii) the absolute value of the base-2 log fold change reported by DESeq2 was ≥1.0 (Supplementary Tables S17–S22). Volcano plots were constructed using ggplot2, while Venn diagrams and heatmaps were constructed using the ggVennDiagram v1.5.2 [58] and pheatmap v1.0.12 [59] packages, respectively (Supplementary Figures S11 and S12). For heat maps (Supplementary Figures S11 and S12), raw count tables were centered and scaled using the “scale” function in R (i.e., with “center = TRUE, scale = TRUE”).

2.9. Gene Ontology Enrichment Analysis

For each of the L. monocytogenes and S. enterica Typhimurium datasets, a Gene Ontology (GO) [60] enrichment analysis was performed for each (i) condition (i.e., ultrasound, chlorine, ultrasound + chlorine) relative to the control (i.e., water with no treatment), for upregulated genes (identified via DESeq2), using the topGO v2.54.0 [61] package in R. Gene-to-GO term mappings were obtained by supplying each reference genome’s CDS region sequences (i.e., DNA sequences of each protein-coding gene) to the eggNOG mapper web server (http://eggnog6.embl.de/#/app/emapper, accessed on 21 January 2024). For eggnog mapper, the following parameters were specified (all remaining parameters were set to their default values): (i) “CDS” as the data type, and (ii) “Gene Ontology evidence” set to “transfer all annotations (including inferred from electronic annotation)” [62].

The gene-to-GO term mappings produced via eggNOG mapper were filtered to remove genes without any linked GO terms. Genes were ordered by raw p-value (produced via DESeq2; see Section 2.8 “Identification of differentially transcribed genes”) in increasing order (i.e., from lowest DESeq2 p-value to highest). This ordered vector was converted to a vector of ones and zeroes, where a gene was considered to be a member of the “significant” gene set (i.e., it was assigned a value of “1”) if its DESeq2-adjusted p-value was <0.05 and if the absolute value of its DESeq2 log2-scaled fold change was ≥1.0; if either of these conditions were not met, the gene was considered to not be significant and was assigned a value of “0”. The resulting vector and eggNOG gene-to-GO term mappings were used to create a topGO object (i.e., as the “allGenes” and “annotationFun” attributes, respectively), using the GO biological process (BP) ontology (ontology = “BP”) and a node size of 3 (nodeSize = 3). This object was supplied as input to topGO’s “runTest” function, which was used to perform Fisher’s exact test (statistic = “fisher”), using the “weight01” algorithm (to account for GO graph topology). This was repeated for the GO molecular function (MF) and cellular component (CC) ontologies, for both up- and down-regulated genes, for each condition (i.e., ultrasound, chlorine, ultrasound + chlorine), for both data sets (i.e., L. monocytogenes and S. enterica Typhimurium; Supplementary Tables S23 and S34). GO terms were considered to be significantly enriched if the raw topGO p-value was <0.05 (per the topGO manual [61], raw p-values produced by the weight01 algorithm can be interpreted as “corrected” or “not affected by multiple testing”). However, as an additional precaution, R’s “p.adjust” function was used to apply a second correction to raw p-values produced by topGO (i.e., a false discovery rate correction with ‘method = “fdr”’). Results were plotted in R version 4.5.2 using the ggplot2 and gridExtra packages.

2.10. KEGG Over-Representation and Enrichment Analyses

For each of the L. monocytogenes and S. enterica Typhimurium datasets, over-representation and enrichment of KEGG (a) pathways and (b) modules was performed for each condition (i.e., ultrasound, chlorine, ultrasound + chlorine) relative to controls (i.e., water with no treatment), using the clusterProfiler v4.12.6 package [63] in R. For L. monocytogenes, KEGG pathway-to-gene and module-to-gene mappings were obtained from eggNOG mapper (see Section 2.9 “Gene Ontology enrichment analysis” above), as the strain used here was not part of KEGG’s databases. These mappings, along with results from DESeq2 (see Section 2.8 “Identification of differentially transcribed genes” above) were supplied as input to clusterProfiler’s “enricher” function to perform KEGG pathway and module over-representation analyses; specifically: (i) KEGG pathway-to-gene and module-to-gene mappings were treated as the “gene universe” (i.e., supplied to enricher’s “TERM2GENE” argument); (ii) “significant genes” were supplied to enricher’s “gene” argument (genes were considered to be “significant” if their DESeq2-adjusted p-value was <0.05 AND the absolute value of their log2-transformed fold-change was ≥1.0; up-regulated and down-regulated genes were tested separately); (iii) minimum and (iv) maximum numbers of genes in a set were set to arbitrarily low and high thresholds (“minGSSize = 1” and “maxGSSize = 1e9”, respectively); p-values were considered to be significant using a (v) 0.05 threshold (pvalueCutoff = 0.05) after (vi) controlling for the false discovery rate (pAdjustMethod = “fdr”). In addition to performing KEGG pathway and module over-representation analyses, clusterProfiler’s “GSEA” function was used to perform KEGG pathway and module gene set enrichment analyses (GSEA). For each GSEA, all genes were ranked in decreasing order according to their log2-transformed fold change value; this ranked list was supplied to the “GSEA” function’s “geneList” argument. The same parameters used with “enricher” were used with “GSEA” (i.e., the same gene universe supplied to “TERM2GENE”, “minGSSize = 1”, “maxGSSize = 1e9”, “pvalueCutoff = 0.05”, “pAdjustMethod = fdr”), along with “eps = 0” (setting the boundary for p-value calculations to 0; Supplementary Tables S35–S40).

For S. enterica Typhimurium, KEGG pathway and module over-representation and GSEA analyses were performed as described above for L. monocytogenes, but with the following changes: (i) because the S. enterica Typhimurium str. LT2 genome was included in KEGG’s database, pathway-to-gene and module-to-gene mappings were obtained directly from KEGG (i.e., using organism = “stm”, per https://www.genome.jp/kegg/catalog/org_list.html, accessed on 8 November 2024); (ii) instead of “enricher”, “enrichKEGG” and “enrichMKEGG” were used to perform KEGG pathway and module over-representation analyses, respectively; (iii) instead of “GSEA”, “gseKEGG” and “gseMKEGG” were used to perform KEGG pathway and module GSEA, respectively (Supplementary Tables S41–S46). Plots were constructed in R using the ggplot2 package.

2.11. Data Availability

Illumina sequencing reads generated for this project are available in the NCBI Sequence Read Archive (SRA) database under BioProject accessions PRJNA1344886 (BioSample accessions SAMN52656871-SAMN52656886) and PRJNA1344789 (BioSample accessions SAMN52654005-SAMN52654020). RNA-seq analysis code is available via GitHub: https://github.com/lmc297/fda_rna_seq_og_lissal. Supplementary Material is available via Zenodo (https://doi.org/10.5281/zenodo.17351413).

3. Results

3.1. Treatments with Power Ultrasound Triggered Strong Transcriptomic Shifts in L. monocytogenes but a Relatively Less Diverse Response in S. enterica Typhimurium

L. monocytogenes and S. enterica Typhimurium underwent four different treatments for 1 min: (i) 20 kHz ultrasound, (ii) 50 ppm chlorine, (iii) 20 kHz ultrasound + 50 ppm chlorine treatment and (iv) water to serve as the control. After the treatments, RNA was extracted, sequenced and analyzed. Differentially expressed genes of both strains were identified and visualized using volcano plots to highlight the expression gene changes (Figure 1 and Figure 2). The L. monocytogenes genome contained a total of 2910 protein-coding genes. The 20 kHz ultrasound treatment led to significant (DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0) upregulation of 462 genes and downregulation of 430 genes compared to the control (Figure 1A). The 50 ppm chlorine treatment led to significant (DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0) upregulation of 5 genes and 34 downregulated genes (Figure 1B). For the combined treatment of 20 kHz ultrasound + 50 ppm chlorine, L. monocytogenes showed the highest numbers of differentially expressed genes with a significant (DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0) upregulation of 551 genes and 466 downregulated genes when compared to the control (Figure 1C). Of the three treatments, chlorine incited the lowest number of differentially expressed genes for L. monocytogenes with lower fold changes (log₂ fold change < 1.25) compared to ultrasound alone and ultrasound + chlorine (log₂ fold change > 2.5) (Figure 1). For the downregulated genes, there was only one gene that was moderately downregulated while others were mildly downregulated (log₂ fold change < −1.75) for chlorine (Figure 1B). For ultrasound alone and ultrasound + chlorine treatments, there was a higher number of genes that were more moderately downregulated (log₂ fold change < −4.0) compared to chlorine (Figure 1). This suggests that chlorine treatment for 1 min did not trigger as significant of a transcriptomic response for L. monocytogenes compared to the other two treatments that included ultrasound.

The S. enterica Typhimurium genome contained a total of 4554 protein-coding genes. The 20 kHz ultrasound treatment induced significant (DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0) upregulation of 38 genes and downregulation of 82 genes compared to the control (Figure 2A). In comparison, the 50 ppm chlorine treatment triggered significant (DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0) upregulation of 94 genes and 22 downregulated genes (Figure 2B). For the combined treatment of 20 kHz ultrasound + 50 ppm chlorine, S. enterica Typhimurium showed significant (DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0) upregulation of 85 genes and 98 downregulated genes compared to the control (Figure 2C). Overall, the three treatments led to similar numbers of differentially expressed genes in S. enterica Typhimurium compared to L. monocytogenes (Figure 1 and Figure 2). Chlorine may have incited a weaker transcriptomic shift in L. monocytogenes, but there were four upregulated genes in S. enterica Typhimurium that were more highly expressed (log₂ fold change > 4) compared to ultrasound alone and ultrasound + chlorine (log₂ fold change < 3.75). For the downregulated genes, ultrasound alone led to more highly downregulated genes (log₂ fold change > −3.75) while the genes were moderately downregulated (log₂ fold change > −3.0) for chlorine and ultrasound + chlorine (Figure 2). These results suggest that the transcriptional response from S. enterica Typhimurium was comparatively less diverse across all three treatments when compared to L. monocytogenes.

The overlap patterns of differentially expressed genes between all three treatments continued to support the differences in stress response between L. monocytogenes and S. enterica Typhimurium. The number of differentially expressed genes of L. monocytogenes shared among the three treatments were visualized in a Venn diagram (Figure 3). Among a total of 623 upregulated expressed genes, most of them (395 genes, 63%) were found in both ultrasound and ultrasound + chlorine treatments (Figure 3A). Individually, 156 genes (25%) and 67 genes (11%) were only highly expressed after ultrasound + chlorine and ultrasound treatment, respectively (Figure 3A). In contrast, only 5 genes (1%) were shown to be differentially expressed under the chlorine treatment and there were no overlaps in genes highly expressed between chlorine and the two other treatments (ultrasound and ultrasound + chlorine) (Figure 3A). A similar trend was observed for downregulated genes in L. monocytogenes, where a majority (369 out of 560, 66%) of differentially expressed genes were found in both ultrasound and ultrasound + chlorine treatments (Figure 2B); 96 (17%) and 61 differentially expressed genes (11%) were found unique under ultrasound + chlorine and ultrasound treatment, respectively (Figure 3B). Thirty-three (6%) downregulated genes were identified under chlorine treatment with one gene shared between chlorine and ultrasound + chlorine treatment. However, there was no overlap of genes between all three treatments and none between chlorine and ultrasound treatments alone (Figure 3B). This indicates that ultrasound was the leading factor to the observed transcriptional response in L. monocytogenes; the addition of chlorine to the combined treatment had a minimal impact on the overall transcriptomic responses.

For S. enterica Typhimurium, out of 191 upregulated expressed genes, 19 (10%) were found in both ultrasound and chlorine + ultrasound treatments (Figure 4A). In addition, 2 differentially expressed genes (1%) were found in both chlorine and ultrasound, 2 genes (1%) in all three treatments and 1 gene (1%) in both chlorine and chlorine + ultrasound (Figure 4A). Most upregulated expressed genes were found uniquely in one treatment: 89 genes (47%) after chlorine, 63 genes (33%) in ultrasound + chlorine and 15 genes (8%) in ultrasound treatments (Figure 4A). For downregulated expressed genes, 57 genes (40%) out of 141 were found in both ultrasound and ultrasound + chlorine treatments (Figure 4B). There were only 2 downregulated genes (1%) found in both chlorine and ultrasound treatments and 1 gene (1%) between all three treatments (Figure 4B). The remaining downregulated genes were uniquely in one treatment: 19 genes (13%) for chlorine, 22 genes (16%) for ultrasound and 40 genes (28%) for ultrasound + chlorine treatments (Figure 4B). Even though S. enterica Typhimurium appeared to exhibit a similar response across all treatments compared to L. monocytogenes, the majority of overlapping upregulated and downregulated genes were shared specifically between ultrasound and ultrasound + chlorine treatments. This pattern suggests that ultrasound may potentially be the primary driver of transcriptional changes in the combined treatment.

For each of L. monocytogenes and S. enterica Typhimurium, Gene Ontology (GO) term enrichment analyses were used to identify GO biological processes (BPs), molecular functions (MFs), and cellular components (CCs) enriched among upregulated genes (i.e., DESeq2-adjusted p-value < 0.05 & log₂ fold change ≥ 1.0) under ultrasound (A), chlorine (B) and ultrasound + 50 ppm chlorine (C) treatments. For L. monocytogenes exposed to ultrasound, a total of 51 GO terms were enriched among upregulated genes (topGO raw p-value < 0.05; Supplementary Table S23). The most significantly enriched L. monocytogenes GO terms (false discovery rate [fdr]-corrected topGO p-value < 0.05) were biological processes (Figure 5), specifically cellular catabolic processes (GO:0044248), followed by glycerol catabolic processes (GO:0019563), Mo-molybdopterin cofactor biosynthesis process (GO:0006777), cobalamin biosynthetic processes (GO:0009236) and glyceraldehyde-3-phosphate metabolic processes (GO:0019682).

Following chlorine treatment, only two GO terms, extracellular region (GO:0005576), a cellular component, and isomerase activity (GO:0016853), a molecular function, were significantly enriched among L. monocytogenes upregulated genes (topGO raw p-value < 0.05; Supplementary Table S25). After ultrasound + chlorine treatment, 55 GO terms were significantly enriched (topGO raw p-value < 0.05; Supplementary Table S27). Among the top enriched GO terms (fdr-corrected topGO p-value < 0.05), four were the same terms that were significantly enriched in ultrasound treatment: Glycerol catabolic processes (GO:0019563), cellular catabolic processes (GO:0044248), cobalamin biosynthetic processes (GO:0009236) and Mo-molybdopterin cofactor biosynthesis process (GO:0006777), which are all biological processes (Figure 5). The two top GO terms that were uniquely significantly enriched (fdr-corrected topGO p-value < 0.05) were carbohydrate metabolic process (GO:0005975), a biological process, and intramolecular oxidoreductase activity (GO:0016861), a molecular function (Figure 5). The stark contrast between ultrasound and ultrasound + chlorine treatment compared to chlorine continues to suggest that chlorine triggers a more limited and focused transcriptional adjustment in L. monocytogenes.

In contrast to L. monocytogenes, only 18 GO terms were enriched among upregulated S. enterica Typhimurium genes under ultrasound treatment (topGO raw p-value < 0.05; Supplementary Table S29). This was understandable, as there were fewer upregulated S. enterica Typhimurium genes after the ultrasound treatment, the lowest amount compared to the other treatments against S. enterica Typhimurium (Figure 4A). This is similar to chlorine treatment against L. monocytogenes, with only two enriched GO terms (topGO raw p-value < 0.05; Supplementary Table S25) from only five upregulated genes, which was also the lowest amount compared to the other treatments against L. monocytogenes (Figure 3A). After chlorine treatment of S. enterica Typhimurium, 32 GO terms were enriched among upregulated genes (topGO raw p-value < 0.05; Supplementary Table S31). Two of the top enriched GO terms (fdr-corrected topGO p-value < 0.05) included one molecular function (structural constituent of ribosome, GO:0003735) and one biological process (viral tail assembly, GO:0098003; Figure 6). When ultrasound and chlorine treatments were combined, there were 28 significantly enriched GO terms (topGO raw p-value < 0.05; Supplementary Table S33), including four highly enriched biological processes (fdr-corrected topGO p-value < 0.05; Figure 6). For GO terms enriched among upregulated genes, it appeared that ultrasound and ultrasound + chlorine triggered a more significant response in L. monocytogenes, whereas chlorine induced more differentially expressed genes in S. enterica Typhimurium, as reflected by the higher number of GO terms.

3.2. Distinct KEGG Pathways Were Identified Under Power Ultrasound-Based Treatments

clusterProfiler was additionally used to identify KEGG modules and pathways that were over-represented among differentially expressed genes of L. monocytogenes and S. enterica Typhimurium (i.e., DESeq2-adjusted p-value < 0.05 & |log₂ fold change| ≥ 1.0), as well as to perform a gene set enrichment analysis (GSEA). KEGG module enrichment analysis revealed distinct transcriptional patterns in L. monocytogenes to ultrasound, chlorine, and combined ultrasound + chlorine treatments (Figure 7, Supplementary Tables S37, S38 and S40). Using GSEA, ultrasound treatment alone resulted in positive enrichment of modules such as the cellobiose-, fructose-, mannose- and galactitol-specific PTS systems, as well as the pentose phosphate pathway (clusterProfiler GSEA-adjusted p-value < 0.05 & Normalized Enrichment Score [NES] > 0; Figure 7A, Supplementary Table S40). In contrast, KEGG modules with the lowest NES values for L. monocytogenes after ultrasound treatment (clusterProfiler GSEA-adjusted p-value < 0.05 & NES < 0) included glycine betaine/proline transport system, uridine monophosphate biosynthesis, putative polar amino acid transport system and iron complex transport system (Figure 7A, Supplementary Table S40). The most positively enriched KEGG modules identified via the GSEA (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0) were also identified via KEGG module overrepresentation analysis among upregulated genes under ultrasound treatment (clusterProfiler overrepresentation-adjusted p-value < 0.05; Figure 7B, Supplementary Table S37). This overrepresentation analysis supports the GSEA enrichment results, reinforcing the consistency of the findings by confirming that these KEGG modules were not only ranked highly in the gene set, but are also statistically overrepresented among the differentially expressed genes.

In contrast, under chlorine treatment, KEGG modules such as galactitol-, cellobiose-, and fructose-specific PTS systems and reductive pentose phosphate cycles (glyceraldehyde-3P → ribulose-5P and the Calvin cycle) were among the most significantly downregulated modules as indicated by negative GSEA NES values (clusterProfiler GSEA-adjusted p-value < 0.05 & NES < 0; Figure 7A, Supplementary Table S40). As expected, none of the KEGG modules mentioned above were significantly overrepresented among upregulated genes, but overrepresentation analysis did identify significant (clusterProfiler overrepresentation-adjusted p-value < 0.05) enrichment of a single KEGG module, namely, the catechol meta-cleavage pathway. These trends were mirrored at the KEGG pathway level for L. monocytogenes, where all eight pathways were suppressed under chlorine stress (clusterProfiler GSEA-adjusted p-value < 0.05 & NES < 0; Figure 8A, Supplementary Table S39). In the overrepresentation analysis of KEGG pathways, four pathways under chlorine stress were significantly (clusterProfiler overrepresentation-adjusted p-value < 0.05) enriched among upregulated genes: Dioxin degradation, xylene degradation, benzoate degradation, and degradation of aromatic compounds (Figure 8B, Supplementary Table S35).

KEGG module enrichment of L. monocytogenes’ differentially expressed genes after ultrasound + chlorine was very similar to ultrasound treatment, as many of the same modules were among the most significantly enriched/suppressed via GSEA, including the same phosphotransferase system components (mannose-, cellobiose-, and galactitol-specific II components) and pentose phosphate pathway (non-oxidative phase), as well as the same modules with negative GSEA NES values (glycine betaine/proline transport system, putative polar amino acid transport system and iron complex transport system, clusterProfiler GSEA-adjusted p-value < 0.05; Figure 7A, Supplementary Table S40).

KEGG pathway analysis reinforced the KEGG module findings, as pathways with positive GSEA NES (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0) and significant overrepresentation among upregulated genes (clusterProfiler overrepresentation-adjusted p-value < 0.05) were similar to significant KEGG modules (e.g., modules/pathways related to PTS, fructose and mannose metabolism; Figure 7 and Figure 8). As was the case with KEGG modules, KEGG pathways enriched and overrepresented among upregulated genes were similar for L. monocytogenes treated with ultrasound and ultrasound + chlorine (Figure 8A,B). These results suggest a consistent transcriptional response in L. monocytogenes in carbohydrate transport and metabolism between ultrasound and ultrasound + chlorine treatments (Figure 7 and Figure 8).

For the KEGG module analysis of ultrasound-treated S. enterica Typhimurium, there were no KEGG modules identified as statistically significant by GSEA (Figure 9A, Supplementary Table S46), suggesting that there was not a module that showed a clear overall increase or decrease in activity. However, overrepresentation analysis revealed one module (dTDP-L-rhamnose biosynthesis) that was significantly (clusterProfiler overrepresentation-adjusted p-value < 0.05) enriched among upregulated genes (Figure 9B), indicating that while this module did not show a consistent directional trend across all genes in GSEA, a subset of its genes was significantly affected by ultrasound treatment with overrepresentation.

In chlorine-treated S. enterica Typhimurium, the only KEGG module to display a positive NES after GSEA analysis is cobalamin biosynthesis (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0; Figure 9A). However, there were four modules (cobalamin biosynthesis, ornithine-ammonia cycle, ornithine biosynthesis, and arginine biosynthesis) that were significantly overrepresented among upregulated genes (clusterProfiler overrepresentation-adjusted p-value < 0.05; Figure 9B), suggesting that while only cobalamin biosynthesis showed consistent changes across the full gene set (as captured by GSEA), additional metabolic modules were also significantly affected, as shown by their overrepresentation among upregulated genes.

After ultrasound + chlorine treatment, GSEA revealed positive enrichment of modules such as citrate cycle (TCA cycle, Krebs cycle), citrate cycle (second carbon oxidation), methylcitrate cycle, and reductive citrate cycle (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0; Figure 9A). Overrepresentation analysis significantly (clusterProfiler overrepresentation-adjusted p-value < 0.05) identified the same modules, along with several additional modules (Figure 9B, Supplementary Table S46).

KEGG pathway analyses in S. enterica Typhimurium revealed many similar pathways that were positively enriched between two or all three treatments (Figure 10A). For example, using GSEA, ribosome and bacterial chemotaxis were positively enriched between all three treatments (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0; Figure 10A). Comparatively, overrepresentation analysis did not identify any KEGG pathways enriched among upregulated genes for S. enterica Typhimurium treated with ultrasound (Figure 10B). These results indicate that S. enterica Typhimurium exhibited both common and treatment-specific transcriptional responses. Core pathways such as ribosome and bacterial chemotaxis were enriched across all treatments, while others showed enrichment in specific treatment combinations, reflecting overlapping but distinct transcriptional adjustments to ultrasound, chlorine, or ultrasound + chlorine treatments.

Under ultrasound treatment, KEGG pathway GSEA in S. enterica Typhimurium revealed positive enrichment for ten pathways (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0; Figure 10A, Supplementary Table S45); however, none reached statistical significance (clusterProfiler overrepresentation-adjusted p-value < 0.05) in the overrepresentation analysis among upregulated genes (Figure 10B, Supplementary Table S41). This was likely due to moderate gene expression changes spread across many genes rather than concentrated in a small number of highly differentially expressed, upregulated genes, which is generally used for overrepresentation detection.

After chlorine treatment, only one KEGG module for S. enterica Typhimurium was identified as statistically significant after GSEA analysis (clusterProfiler GSEA-adjusted p-value < 0.05; Figure 9A, Supplementary Table S46). However, KEGG pathway analysis identified several pathways with positive NES values, as well as several with negative NES values (clusterProfiler GSEA-adjusted p-value < 0.05; Figure 10A, Supplementary Table S45). While most of the pathways with positive NES values did not reach statistical significance in the overrepresentation test among upregulated genes, ribosome and arginine biosynthesis did (clusterProfiler overrepresentation-adjusted p-value < 0.05; Figure 10B, Supplementary Table S41), potentially indicating that these two pathways not only showed coordinated expression changes captured by GSEA, but also contained a significant concentration of upregulated genes, reinforcing their role in the response to chlorine treatment. In contrast, pathways such as PTS, microbial metabolism in diverse environments and tyrosine metabolism were downregulated (clusterProfiler GSEA-adjusted p-value < 0.05 & NES < 0; Figure 10A).

At the KEGG pathway level of S. enterica Typhimurium after ultrasound + chlorine treatment, numerous pathways were enriched (clusterProfiler GSEA-adjusted p-value < 0.05 & NES > 0; Supplementary Table S45) and two of those pathways (microbial metabolism in diverse environments and carbon metabolism) were also overrepresented among upregulated genes (clusterProfiler overrepresentation-adjusted p-value < 0.05; Figure 10A,B, Supplementary Table S41). In addition to those two pathways, overrepresentation analysis also identified several other pathways that were significantly (clusterProfiler overrepresentation-adjusted p-value < 0.05) enriched such as fatty acid degradation, fatty acid metabolism, pinene, camphor, and geraniol degradation and other carbon fixation pathways (Figure 10B, Supplementary Table S41). The overlap between GSEA and overrepresentation analyses, such as carbon metabolism and microbial metabolism in diverse environments, may suggest that the treatment affected both basic metabolism and more specific stress-related pathways.

4. Discussion

The transcriptomic analyses revealed that both L. monocytogenes and S. enterica Typhimurium exhibited some distinct responses to the ultrasound, chlorine and combined ultrasound + chlorine treatments. For L. monocytogenes, there was a consistent upregulation of multiple PTS system components in the KEGG pathways under power ultrasound treatments (Figure 7). Modules encoding sugar-specific PTS transporters for cellobiose, fructose, mannose, and galactitol were significantly enriched or overrepresented, indicating a transcriptional shift toward increased carbohydrate uptake. These sugar-specific PTS are transport systems that catalyze their specific sugars/carbohydrates during transport into their respective phosphoesters [64,65,66]. The high number of PTSs enriched (for ultrasound alone and for ultrasound + chlorine) suggests that L. monocytogenes utilizes a diverse range of carbon sources to support survival and adaptation under adverse stress conditions. This is consistent with the findings of Stoll and Goebel (2010), who examined PTS systems for L. monocytogenes and found L. monocytogenes to rely on glucose, mannose and cellobiose-specific PTSs for survival and cellular growth [67]. The strong activation of multiple sugar-specific PTS systems under ultrasound (with and without chlorine) suggests that L. monocytogenes enhances carbon uptake and energy production to sustain repair and survival under mechanical stress.

In contrast, chlorine treatment alone led to a different L. monocytogenes transcriptional pattern. The only significantly overrepresented KEGG module among upregulated genes under chlorine treatment was catechol meta-cleavage (clusterProfiler overrepresentation-adjusted p-value < 0.05; Figure 7B), a pathway involved in aromatic compound degradation [68]. The significant overrepresentation of the catechol meta-cleavage pathway under chlorine treatment in L. monocytogenes indicates that the bacterium was actively engaging in aromatic compound degradation under these conditions. While the precise triggers for activation of this pathway under chlorine stress are unclear, it is well established that bacterial cells often undergo shifts in metabolic activity during stress, which can include turning on pathways such as aromatic compound degradation. This difference in PTS regulation under chlorine stress contrasts with the consistent PTS upregulation observed under ultrasound and combined treatments. A study by Chen et al. (2023) using diamide and copper ions to induce oxidative stress in L. monocytogenes also reported downregulation of PTS genes (lmo1997–lmo2004) under stress [69], aligning with the chlorine treatment results in this study. Chlorine, similar to the diamide and copper agents used in that study, also induced oxidative stress. However, the effects of chlorine can be quite different from copper [70,71]. Fan et al. (2013) evaluated the effectiveness of copper sulfate and chlorine on cyanobacterial cell integrity and found chlorine to be the most effective at lysing cyanobacterial cells, but copper sulfate continued to inhibit Microcytosis aeruginosa growth over the 7-day treatment [72]. Ultrasound generates cavitation-induced mechanical stress that disrupts bacterial cell membranes, a mechanism that is distinct from the oxidative stress caused by chlorine, copper, and diamide [73,74]. Variations in the modes of action across different stressors could potentially account for the contrasting patterns of PTS gene expression observed between treatments with and without ultrasound. In contrast to ultrasound and ultrasound + chlorine treatments against L. monocytogenes, chlorine treatment resulted in comparatively reduced PTS activity and enrichment of aromatic-degradation pathways. This may indicate a metabolic shift towards oxidative-stress mitigation rather than nutrient acquisition.

For S. enterica Typhimurium, modules related to the citrate cycle were enriched prominently under the combined ultrasound + chlorine treatments (Figure 9). The citrate cycle is a core component of central carbon metabolism, providing energy for microbial processes, including survival and repair in S. enterica [75]. The upregulation of several citrate cycle-related modules suggests that S. enterica Typhimurium may have required increased energy production following ultrasound + chlorine treatment, compared to ultrasound or chlorine treatments alone, to support cellular repair and survival. Under ultrasound treatment alone, only one KEGG module (dTDP-L-rhamnose biosynthesis) was significantly (clusterProfiler overrepresentation-adjusted p-value < 0.05) overrepresented, and this module did not appear in the GSEA analysis (Figure 9). L-rhamnose is a sugar incorporated into cell wall polysaccharides of various bacteria, including S. enterica [76]. dTDP-L-rhamnose serves as the activated form required by glycosyltransferases to incorporate L-rhamnose into growing polysaccharides, where it functions to strengthen, stabilize, and protect the bacterial cell wall under stress conditions such as those induced by ultrasound [76,77,78]. The targeted upregulation of this module highlights a potentially distinct cell wall adaptation response specific to ultrasound-induced stress, where the bacterium may alter or reinforce its cell wall structure to better withstand the physical damage caused by ultrasound, which differs from the broader metabolic pathways activated under chlorine treatment. Under chlorine treatment alone, four modules were significantly overrepresented among upregulated genes (ornithine-ammonia cycle, ornithine biosynthesis, arginine biosynthesis, and cobalamin biosynthesis), but only one (cobalamin biosynthesis) showed positive enrichment in GSEA (Figure 9). A potential explanation for this discrepancy is that while certain genes within these modules exhibited strong differential expression, the overall gene expression shifts across the entire module were insufficiently coordinated to produce a significant NES in GSEA [79,80].

Following chlorine treatment, cobalamin biosynthesis was the only KEGG module that was both significantly overrepresented (Figure 9B) and detected in GSEA (Figure 9A). S. enterica is known to upregulate cobalamin synthesis under anaerobic conditions [81,82]. This may indicate that chlorine, as a strong oxidant [83], impaired S. enterica aerobic respiration, leading the bacterium to upregulate cobalamin biosynthesis to maintain survival pathways that are independent of oxygen metabolism. The other overrepresented modules after chlorine treatment (ornithine-ammonia cycle, ornithine biosynthesis, and arginine biosynthesis) (Figure 9B) are involved in amino acid biosynthesis and nitrogen metabolism, potentially supporting the synthesis of amino acids for protein repair and turnover following cellular damage induced by chlorine exposure [84]. Overall, the patterns of KEGG module overrepresentation across all three treatments in S. enterica Typhimurium suggest that this organism experienced cellular damage and upregulated pathways essential for stress adaptation and survival. Notably, the greater number of energy-related KEGG modules overrepresented after ultrasound + chlorine treatment suggests that this combined treatment may have caused greater cellular damage than ultrasound or chlorine alone. The increased need for energy production likely reflects the energetic demands of membrane repair, protein synthesis and metabolic recovery required to restore homeostasis following compounded mechanical and oxidative stress from ultrasound + chlorine.

For S. enterica Typhimurium under ultrasound treatment, an interesting divergence was observed between the KEGG module and pathway enrichment analyses (Figure 9 and Figure 10). Specifically, one KEGG module was significantly overrepresented among upregulated genes, while no modules showed significant enrichment by GSEA NES (Figure 9). In contrast, for KEGG pathways, ten pathways displayed positive GSEA NES values, yet none were significantly overrepresented among upregulated genes (Figure 10). This pattern could potentially be explained by the distinct statistical characteristics of the two approaches. Overrepresentation analysis detects pathways or modules in which a large proportion of genes are classified as significantly differentially expressed. The one KEGG module that was overrepresented may have contained a subset of genes with strong expression changes, but these changes were not sufficiently coordinated across the entire module to produce a significant NES in GSEA. Conversely, GSEA NES is sensitive to subtle but consistent trends across all genes in a pathway, even when only a small number of those genes meet strict differentially expressed gene thresholds. The ten KEGG pathways enriched in GSEA likely reflect broad, modest transcriptional adjustments across many metabolic pathways in response to ultrasound, which did not generate a concentrated differentially expressed gene signal strong enough to reach significance in overrepresentation analysis. Biologically, this may be due to the effects of ultrasound stress, which primarily induces cell membrane disruption and cell envelope stress instead of activating oxidative stress responses [24,26,85]. As a result, S. enterica Typhimurium may respond by modestly adjusting metabolism and energy pathways broadly without strongly upregulating classic stress-related genes.

Between the two bacterial species, L. monocytogenes exhibited broader and more pronounced transcriptional shifts, particularly under ultrasound and combined ultrasound + chlorine treatments, with consistent upregulation of multiple PTS components and central carbon metabolism pathways (Figure 7 and Figure 8). In contrast, S. enterica Typhimurium exhibited more modest metabolic adjustments, particularly under ultrasound alone (Figure 9 and Figure 10). The most substantial response in S. enterica Typhimurium was observed under the combined ultrasound + chlorine treatment, where extensive activation of citrate cycle-related modules and energy production pathways was detected, suggesting increased energy demands for cellular repair and survival. These patterns suggest that L. monocytogenes may experience greater membrane perturbation and metabolic stress under ultrasound-based treatments, leading to stronger activation of nutrient acquisition and carbon metabolism pathways. In contrast, S. enterica Typhimurium appears to engage in a more targeted and energy-focused survival strategy, particularly under compounded mechanical and oxidative stress from ultrasound + chlorine. Overall, the results suggest that L. monocytogenes was more metabolically affected by ultrasound-based treatments, while S. enterica Typhimurium exhibited its strongest response to the combination of ultrasound + chlorine treatment. These differences likely reflect species-specific variations in stress tolerance mechanisms, membrane structure, and metabolic flexibility, which merit further investigation in the future.

5. Conclusions

These transcriptomic findings provide valuable mechanistic insights into how L. monocytogenes and S. enterica Typhimurium respond at the molecular level to power ultrasound, chlorine, and combined treatments. For L. monocytogenes, the strong upregulation of sugar-specific PTS systems and central metabolic pathways in response to ultrasound suggests the pathogen may be initiating energy-intensive repair and survival processes. This highlights an opportunity to optimize ultrasound-based decontamination strategies by either enhancing ultrasound exposure conditions, such as extending treatment duration, fine-tuning parameters to amplify cavitation effects, or combining ultrasound with metabolic inhibitors to disrupt these compensatory pathways. In contrast, S. enterica Typhimurium required combined ultrasound and chlorine treatment to activate energy pathways such as the citrate cycle, indicating that stronger or multiple stressors to be used in combination with ultrasound + chlorine are needed to disrupt its cellular defenses. Findings from this study may shed new light on fine-tuning hurdle treatment designs for the industry to allow for more effective, targeted, and sustainable fresh produce safety interventions.