Analysis of stress response in multiple bacterial pathogens using a network biology approach

Abstract

Stress response in bacterial pathogens promotes adaptation, virulence and antibiotic resistance. In this study, a network approach is applied to identify the common central mediators of stress response in five emerging opportunistic pathogens; Enterococcus faecium Aus0004, Staphylococcus aureus subsp. aureus USA300, Klebsiella pneumoniae MGH 78,578, Pseudomonas aeruginosa PAO1, and Mycobacterium tuberculosis H37Rv. A Protein-protein interaction network (PPIN) was constructed for each stressor using Cytoscape3.7.1 from the differentially expressed genes obtained from Gene expression omnibus datasets. A merged PPIN was constructed for each bacterium. Hub-bottlenecks in each network were the central stress response proteins and common pathways enriched in stress response were identified using KOBAS3.0. 31 hub-bottlenecks were common to each individual stress response, merged networks in all five pathogens and an independent cross stress (CS) response dataset of Escherichia coli. The 31 central nodes are in the RpoS mediated general stress regulon and also regulated by other stress response systems. Analysis of the 20 common metabolic pathways modulating stress response in all five bacteria showed that carbon metabolism pathway had the highest crosstalk with other pathways like amino acid biosynthesis and purine metabolism pathways. The central proteins identified can serve as targets for novel wide-spectrum antibiotics to overcome multidrug resistance.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-91269-5.

Introduction

Bacteria have developed sophisticated mechanisms to survive in extreme environmental conditions through years of evolution¹. The ability to screen the environment for physical and chemical disturbances is integral for bacteria to stay in a dynamic environment². Survival of the fittest is achieved as a result of enormous genetic elasticity in the bacteria. The capacity to deal with stressors is the determining factor for their survival. Bacteria fight against stress by adopting many mechanisms including proteolysis, formation of dormant structures such as spores, modification of molecules entering the bacteria, decreased antibiotic penetration or efflux, changes in the target site, changes in the composition of the stressor and modifications in metabolic pathways³.

Studies on the stress response have determined that different environmental conditions affect cell cycle, movement of signalling molecules, metabolism, the level of gene expression and protein activity in bacteria⁴. Sigma factor, the factor that bind to the RNA polymerase at transcription initiation site is the primary regulator for general stress response in many bacterial species⁵. Different sigma factors perform various tasks such as heat shock, motility of flagella, maintaining the stationary phase, cell envelop formation, sporulation etc⁶. Bacteria are simultaneously exposed to various stressors, such as changes in temperature, pH, availability of nutrients, and exposure to toxins⁷. Harmless bacteria are able to survive in the environment due to the alterations in gene expression caused by stress but pathogenic bacteria use this as an advantage to escape from the immune system and antibiotics⁶.The stress response mechanisms in bacteria are associated with increased virulence in bacteria, bacterial adaptation, physiological alterations, multidrug resistance, biofilm formation, quorum sensing, and antibiotic resistance^1,8.

A unified control of the stress response is suggested by cross-stress (CS) protection, the capacity of one stress situation to protect against other stressors⁹. As an example, keeping Escherichia coli in a medium containing acid provides protection against cold, antibiotics, heat and chemicals. In comparison, the cells that were not exposed to acid first were not resistant to other stresses¹⁰. CS protection has also been reported in fungus Metarhizium robertsii where treatment with heat shock induced protection against UV, osmotic and oxidative stress¹¹. CS protection against environmental stress and antibiotics promote bacterial resistance to therapeutic agents^7,12,13. In this way, CS protection has also resulted in increased virulence in bacteria. CS protection has also posed a threat to food safety. Exposure to acidic conditions could induce cross-protection for foodborne pathogens against subsequent stress or multiple stresses such as heat, cold, osmosis, antibiotic, disinfectant, and non-thermal technology¹⁴. When a cell is adapted to a specific condition and suddenly in it subjected to another condition there is a change in gene expression.

Large omics datasets can be analysed through machine learning, statistical analysis, pathway analysis, network studies, visualization methods such as scatter plots etc¹⁵. Networks are of different types, such as protein-protein interaction (PPI), isoform-isoform interactions, genetic interactions, neuron-neuron interactions, metabolic molecule interactions, and drug-target interactions¹⁶. Bacteria have numerous genes and interconnected pathways that benefit from a system level approach for their integrative analysis using network biology¹⁷. Network studies are highly effective in studying the biological processes and pathways as well as key genes and proteins, including drug targets¹⁸. As an example, Protein-Protein Interaction network (PPIN) analysis elucidates the association of one disease with another^19,20. Another study showing PPI analysis of human proteins and virus proteins was used to study the properties of human proteins targeted by viruses²¹. Analysis of host-pathogen PPIs was carried for prediction of cardiovascular diseases induced by microbes²². Network studies in association with machine learning was done to predict the drug targets in microbe associated cardiovascular diseases²³.

Genes involved in all the stress responses are likely to be differentially expressed under multiple conditions and in multiple pathogens. Genes with a prominent role in stress response are also likely to assume a central role in the PPIN of DEGs. Bacterial stress response has been modelled in E. coli using the DEGs in stress conditions such as pH, temperature and antibiotics. As a result, the role of 24 central proteins pathways such as ribosome, purine metabolism and ABC transporters in stress response was identified²⁴. In the present study, we hypothesized the presence of a set of central stress response genes and pathways that would be common to and necessarily required for response to all stressors as well as CS response in multiple pathogenic bacteria. The bacteria used in this study are Enterococcus faecium Aus0004, Staphylococcus aureus subsp. aureus USA300, Klebsiella pneumoniae MGH 78,578, Pseudomonas aeruginosa PAO1, and Mycobacterium tuberculosis H37Rv. These opportunistic and antibiotic resistant bacteria are responsible for a majority of hospital-acquired infections cause infections in the nose, skin, lungs, urinary tract, and in different regions undergoing surgery in hospitalized patients²⁵. The different types of stressors considered are antibiotics, drugs, chemicals, prebiotics, nutrients starvation etc. The DEGs in specific stress conditions were extracted from Microarray and RNA Seq datasets. PPINs were generated from the five bacteria in different stress conditions. Thirty-one central proteins were found to be common to all the stressors in each pathogen while 20 common pathways were identified to have a role in stress response in the bacteria studied. Cross-validation with an independent CS response dataset of Escherichia coli showed that the central stress responses genes coincided with central genes of the CS-PPIN. Our work indicates the existence of some common genes and pathways that are responsible for stress response. Identification of the central proteins and pathways involved in stress response in bacteria can be helpful for determining broad-spectrum antibiotic therapies in future.

Materials and methods

Identification of datasets from gene expression omnibus

The expression data was extracted from GEO, an international public repository which contain high-throughput gene expression and other functional genomics data²⁶. The GEO datasets was searched for emerging pathogenic bacteria. Specific strains were taken to avoid strain wise variations. The strains with the maximum number of available datasets of stress were selected. Studies with knockouts or other genetic manipulation were excluded. The GEO datasets used are shown in Table 1^27–36. The information about stressors considered is provided in Supplementary Table S1.

Table 1.

Details of datasets selected to find stress response DEGs in five pathogenic bacteria.

No.	Microorganism	Technique	Stress condition	Accession no.	References	No. of DEGs
1.	Enterococcus faecium Aus0004	RNA Seq	Daptomycin	GSE94924	27	752
2.	Staphylococcus aureus subsp. aureus USA300	RNA Seq	Nitric oxide	GSE99563	28	514
		Array	Benzbromarone	GSE55980	29	771
		Array	Thioridazine	GSE43759	30	396
3.	Klebsiella pneumoniae MGH 78,578	RNA Seq	NMP 1-(1-naphthylmethyl)-piperazine	GSE122651	31	1431
4.	Pseudomonas aeruginosa PAO1	RNA Seq	Tanreqing	GSE141753	32	707
		RNA Seq	Diallyl ether	GSE151292	33	559
		RNA Seq	Fructooligosaccharide	GSE124468	34	85
5.	Mycobacterium tuberculosis H37Rv	RNA Seq	Low iron	GSE213943	35	2030
		Array	Low zinc	GSE168513	36	725

Identification of DEGs

From the complete set of DEGs present in the datasets, genes showing significant change in gene expression having |Log2FC | ≥ 1 and Benjamini-Hochberg False Discovery Rate (FDR) ≤ 0.05 were taken as seed nodes^22,24. For the microarray datasets, DEGs were extracted through GEO2R²⁶. For RNA Seq data, the DEGs were extracted by processing the FPKM (fragments per kilobase of transcript per million reads mapped). Fold change (log₂FC) calculations were done and p-value was calculated by applying the t-test. FDR was calculated by sorting p-values in ascending order, ranking them and applying the formula³⁷:

where m is total number of hypotheses tested and i is rank of p-value.

The list of DEGs is given in Supplementary sheet S1.

PPIN generation

The individual PPIN in each stress condition was generated using STRING³⁸. The confidence score for network construction was taken as 0.775²⁴. The probability that a predicted link between two proteins in the same metabolic map in the KEGG database actually exists is represented by the confidence score. The PPIN of each bacterium in separate stress conditions was generated. The union of the different stressors of bacteria was generated using Cytoscape 3.7.1³⁹.

Validation and randomization of networks

To validate the network as biological network power law was fitted to the network and node degree distribution was plotted. The generated cross response network was also statistically validated using Simple randomization method preserving the degree option³⁹ of the network Randomizer 1.1.3 plugin in Cytoscape 3.7.1.

Identification of hubs and bottlenecks

The topological analysis of the PPIN was done by using several topological measures namely degree, betweenness centrality (BC) and clustering coefficient. These can be explained as follows:

1. A degree of a node in the network is defined as number of edges linked to it which means the number of interactions The nodes with degree exponent < 2 were considered as Hub nodes as previously used for calculation of significant hub nodes²².

2. The BC is a measure that tells how frequently a node functions as a bridge over the shortest pathways connecting two other nodes²². A node with high BC has greater control over the network⁴⁰. The top high BC nodes in this study were the same as the number of nodes identified as hubs for each network. The nodes that are both hubs as well as bottlenecks are called Hub-bottleneck nodes were considered central in the network.

3. The clustering coefficient is the measure used to infer the properties of the proteins in a cluster. It is the measure used in network analysis to define the level to which nodes tend to cluster together. The proteins in the similar clusters have similar properties. The value of clustering coefficient is between 0 and 1. The higher the value, higher the similarity/association between the proteins in the same cluster⁴¹.

Generation of CS-PPIN

In order to compare the central stress response proteins obtained from different pathogens with those under CS response, a PPIN of CS was generated. DEGs under two or more stress conditions were obtained from a dataset of CS protection study in which E. coli was grown for 500 generation under five different stressors (nutrient deprivation, 0.3 M NaCl, 0.6 n-butanol, 0.1 mMol H₂O₂, and pH 5.5)⁴². The DEGs with a cut-off score FDR < 0.05 were obtained and were used to construct a CS-PPIN. The hub-bottlenecks were computed.

Identification of stress response pathways

The pathways involved in stress response in five selected bacteria were identified using KOBAS 3.0, a webserver for annotation and identification of enriched pathways and gene ontologies. The enrichment module was used to find the pathways associated with the genes in stress response⁴³.

Results

Topological analysis of the generated networks

To study the stress response, ten stress conditions were analysed in five bacteria. For this, ten individual networks were constructed and merged to obtain five final stress response PPINs for five selected bacteria. The number of seed nodes used to generate the stress response PPIN of each bacterium are shown in Table 1. The details of networks constructed using the DEGs extracted in STRING v11 are shown in Supplementary Table S2.

A single network was generated using seed nodes in Daptomycin stress E. faecium (referred to as EF-PPIN) while three individual networks were constructed for stress conditions corresponding to Nitric oxide, Benzbromarone and Thioridazine in S. aureus. The merged networks of these conditions are then formed in Cytoscape (termed the SA-PPIN). For K. pneumoniae a network was constructed using seed nodes in 1-(1-naphthylmethyl)-piperazine stress, represented as KP-PPIN. For P. aeruginosa three independent stress response networks were constructed for stressors Tanreqing, Diallyl ether and Fructooligosaccharide. They were then merged to form a network (PA-PPIN). In case of M. tuberculosis, two networks were generated in Low Fe and Low Zn and then merged to construct the MT-PPIN. So, the final merged networks are represented as EF-PPIN, SA-PPIN, KP-PPIN, PA-PPIN, MT-PPIN. The CS-PPIN of E. coli was also generated.

The topological characteristics of the five constructed networks are shown in Table 2. The values of correlation coefficient, coefficient of determination and degree exponent indicates the scale free nature of all the networks, which in turn means that most nodes have only a few connections, while a small number of nodes have a large number of connections. The scale free nature of the biological nature is important as it allows the efficient coordination among components of the biological systems. Similarly, the calculated average clustering coefficient of real networks are much greater than the random networks as shown in Table 2 and indicate the biological significance of the network²². In network analysis, the clustering coefficient is a metric that measures how closely connected nodes are in a graph (Chalancon, Kruse, and Babu 2013). For comparison, the average clustering coefficients of 1000 randomized networks have also been added. The average clustering coefficient value of all five networks was found to be in the range of 0.345–0.571, and is similar to that reported for other PPI networks^44,45. The clustering coefficient of the PPI networks is also much higher in comparison with that of 0.027–0.161 in the random networks, further validating the node distribution and densely connected nature of the networks.

Table 2.

Topological characteristics of the PPINs.

Network	Nodes	Edges	Highest Degree	Highest BC	Correlation	R 2	Deg exp	Clustering coefficient of real networks	Clustering coefficient of randomized networks
EF-PPIN	995	8129	83	1	0.665	0.758	−1.142	0.459	0.096
SA-PPIN	2038	9125	105	1	0.748	0.838	−1.500	0.411	0.043
KP-PPIN	1299	8466	95	0.666	0.665	0.808	−1.296	0.481	0.064
PA-PPIN	2289	14,970	105	1	0.789	0.887	−1.477	0.422	0.027
MT-PPIN	2875	16,787	209	1	0.829	0.907	−1.428	0.345	0.034
CS-PPIN	715	9887	123	0.064	0.397	0.643	−0.879	0.571	0.161

Constructing the networks of the hub-bottlenecks also showed that they form highly connected subnetworks. The connected components of the subnetworks are shown in Table 3. Constructing the networks of these hub-bottlenecks showed they form highly connected subnetworks. The connected components of the subnetworks are shown in Table 3. In all networks, the hub-bottlenecks were fully connected while in KP-PPIN, all are connected except 1. This indicates the completeness of the constructed networks. The five bacterial stress response networks are shown in Supplementary Figure S1.

Table 3.

The number of Hub-bottlenecks and the connected components of their subnetworks in each network.

Microorganism	No of hubs	No. of bottlenecks	No. of hub bottleneck	Connected components
EF-PPIN	801	801	770	All
SA-PPIN	558	558	349	All
KP-PPIN	856	856	758	757
PA-PPIN	711	711	461	All
MT-PPIN	628	628	323	All
CS-PPIN	691	691	680	All

Central proteins in stress response

The number of Hubs, bottlenecks, and central proteins of the final generated networks were computed and are shown in Table 3. The overlapping of central proteins in the networks is shown in Fig. 1. Comparison of central proteins in all bacterial networks suggested the presence of thirty-one common proteins. These common central proteins across all the five bacterial stress response networks are listed in Table 4. These proteins have a reported role in DNA repair pathways, synthesis of some important biomolecules, activation of some other proteins in stress and performing cellular tasks, energy metabolism, and efficient signalling molecule production as shown in Table 4. The presence of these 31 central proteins was also checked in 10 individual stress response networks. It was observed that all of the 31 common proteins identified were also central in most of the individual stress response networks. As shown in Supplementary Table S3, a majority of the 31 proteins were central to the individual stress response networks, with the lowest number being found in lowest zinc conditions in M. tuberculosis. Additionally, Table S4 gives statistical information such as degree, betweenness centrality and closeness centrality of the 31 common central proteins in each of the five networks. This gives an overall importance of these central proteins in driving stress response in each of the individual bacterial network.

Fig. 1.

Overlapping central proteins modulating stress response in five pathogenic bacteria: The orange ellipse represents central proteins in EF-PPIN, green ellipse represents central proteins in SA-PPIN, blue ellipse represents central proteins in KP-PPIN, yellow ellipse represents central proteins in PA-PPIN, pink ellipse represents central proteins in MT-PPIN.

Table 4.

Role of the common central proteins in normal and stress condition in cells.

Gene name	Protein name	Role of protein in cell	Role of protein in stress	Regulon info
adk	Adenylate kinase	Cellular homeostasis. Energy metabolism46. ADP synthesis90.	Energy metabolism. Nucleotide synthesis46.	F, ns
atpD	ATP synthase subunit beta	ATP synthesis. Metabolic pathways of proteins and carbohydrates47.	Increases antibiotic susceptibility of pathogens91.	TU, ns
eno	Enolase	Glycolytic enzyme, role in catabolic glycolytic pathway57.	Oxidative stress response in P. aeruginosa92 .	TU, linear
gyrB	DNA gyrase subunit B	DNA replication, DNA binding, ATP hydrolysis65.	Develop resistance against fluoroquinolones66.	F, ns
pgk	Phosphoglycerate kinase	ATP production70.	Increase glycolytic flux in hypoxia. Triggers pathways involved in DNA repair and SOS response70.	F, linear
carB	Carbamoyl-phosphate synthase	Pyrimidine biosynthesis50.	Absence of this gene in S. aureus led to significant growth defect and increased sensitivity to hydrogen peroxide51.	TU, Ns
cmk	CMP kinase	Nucleotide biosynthesis52.	DNA repair52.	Ns
folD	Bifunctional protein FolD, Methylenetetrahydrofolate dehydrogenase	Folate cycle, synthesis of methionine, purine and thymidylate58.	Help bacteria in heat-stress response59.	F, Ns
guaA	GMP synthetase	Biosynthesise guanine nucleotides62.	Purine metabolism. Helps bacteria in low guanine63.	TU, In
purB	Adenylosuccinate lyase	Purine metabolism71.	Antibiotic and stress tolerance71.	TU, ns
purE	N5-carboxyaminoimidazole ribonucleotide mutase	Biosynthesis of purine nucleotides72.	Biosynthesis of purine nucleotides73.	F, ns
purH	Bifunctional purine biosynthesis protein	Purine biosynthesis74.	Purine biosynthesis74.	F, ns
ftsZ	Cell division protein FtsZ	Plays major role in cytokinesis48.	Interact with transcription factors. Relocates its Z -ring to other parts of the cell which help other proteins in stress response pathways49.	TU, ns
cysE	Serine acetyltransferase	Cysteine and serine metabolism53.	Cysteine metabolism is linked with antibiotic resistance and biofilm formation in some bacteria53.	F, ns
dnaK	Chaperone protein, Heat shock protein 70	Required for growth in absence of stress54.	Maintain homeostasis in thermal stress55.	F, linear
dnaN	DNA polymerase III beta sliding clamp subunit	DNA replication, maintenance of membrane structure56.	SOS response and DNA repair56.	TU, ns
glmS	Glutamine–fructose-6-phosphate aminotransferase	Protein glycosylation Growth and functional maintenance of the cell60.	Help in Cadmium, osmotic and pH stress61.	TU, ns
gyrA	DNA gyrase subunit A	Replication64.	Play an active role in quinolone treatment64.	F, ns
lepA	Elongation factor 4	Biogenesis of ribosome 30s subunit. Protein export functions67.	Released in the cytoplasm of high ionic strength or low temperature68.	F, ns
pgi	Glucose-6-phosphate isomerase	glycolysis and gluconeogenesis69.	Activated in oxidative, heat and nutrient stress69.	F, ns
pyrG	CTP synthase	Pyrimidine biosynthesis75.	Pyrimidine biosynthesis75.	F, linear
recA	Recombinase A	DNA repair76.	DNA repair pathways76. Swarming motility77.	F, ns
rplB	50 S ribosomal protein L2	Transcription. Formation of peptide bond78.	Helps in high hydrostatic pressure78.	TU, in
rplC	50 S ribosomal protein L3	Role in ribosome assembly. Transcription regulation79.	Resistance tolinezolid (an oxazolidinone) and tiamulin (a pleuromutilin)80.	F, ns
rpoA	DNA-directed RNA polymerase subunit alpha	Transcription81.	Regulation of Transcription in stress81. Inhibition of RpoS by oligomerization93.	TU, ns
rpoB	DNA-directed RNA polymerase subunit beta	RNA synthesis, Binding of sigma factor. Serve as a binding site for a antibiotics82.	Interacts with heat shock proteins to maintain the stability of RNA polymerase83. Inhibition of RpoS by oligomerization93.	TU, ns
rpoC	DNA-directed RNA polymerase subunit beta’	Synthesis of RNA. Catalytic Mg2 + coordination82.	Plays role in efficient transcription in stress response. Binding site for ppGpp which is a global gene regulator in stress. Sigma factor binding82.	TU, ns
rpsA	30 S ribosomal protein S1	Transcription and Translation processes84.	Downregulated in oxidative stress84. RNA mediated stress response85. Inhibition of RpoS by oligomerization93.	F, ns
rpsB	30 S ribosomal protein S2	Binding proteins to ribosome86.	Upregulation on exposure to silver nanoparticle. Ribosomal protein S2 regulates its own level by inhibition of messenger (m)RNA synthesis during stress conditions86.	F, ns
rpsE	30 S ribosomal protein S5	Translocation and translocation. mRNA binding87.	Low expression level in salt stress87.	TU, ns
thyA	Thymidylate synthase	Required for DNA synthesis94.	DNA replication and repair88.Decreased activity in oxidative and Nitrosative stress. Increased activity in acidic stress89,95.	TU, ns

Cross-validation using CS-PPIN

It was found that 31 central proteins are common to the stress response networks of all the pathogens and stressors studied. These 31 nodes were then compared with a CS-PPIN from E. Coli constructed as described in methodology. Out of the 31 central proteins found to be central to stress response in our study, 29 were found to central proteins in CS-PPIN also. Also, 14 out of 20 enriched pathways (common among stress response networks) identified in our study were found to be enriched in CS-PPIN. The central stress response nodes in common with those of the CS-PPIN are shown as bold in Table 4.

Pathways enriched in each bacterium for stress

The common pathways responsible for mediating the stress response in bacteria having p-value < 0.05 were identified. A Venn-diagram to represent overlapping pathways in the five bacteria is shown in Fig. 2. These common enriched pathways are listed in Table 5 which highlights their role in normal and stress conditions such as biofilm formation, signalling molecules formation, biosynthesis of amino acids, maintenance of redox balance, developing resistance against antibiotics etc. The percentage of proteins annotated by KOBAS 3.0 for the stress response network generation is shown in Supplementary Table S5. Twenty pathways were found to be common in all five bacteria. Additionally, Table S6 gives statistical significance in the form of p-value for the 20 common enriched pathways in each of the five networks. This information shows how enriched these common pathways are in each of the five bacterial stress response networks. A low p-value shows that each of the pathway in each network is highly enriched and is important for stress response mechanism.

Fig. 2.

Overlapping enriched pathways modulating stress response in five pathogenic bacteria: The orange ellipse represents the pathways enriched in EF-PPIN, green ellipse represents pathways enriched in SA-PPIN, blue ellipse represents pathways enriched in KP-PPIN, yellow ellipse represents pathways enriched in PA-PPIN, pink ellipse represents central proteins in MT-PPIN.

Table 5.

Common enriched pathways with their role in normal and stress conditions in bacteria.

Pathways	Role in normal conditions	Role in stress
Metabolic pathways	Energy generation, synthesis of biomolecules, cellular processes115.	ATP is the essential requirement in stress conditions115.
Carbon metabolism	Redistribute carbon groups to various amino acids to generate compounds which serve as the building blocks of the cell115.	Modification of enzymes in S. aureus in Cu stress by carbon metabolism116.
Biosynthesis of amino acids	Series of enzymatic reaction that converts precursor molecules into amino acids117.	Effect of cefotaxime stress in amino acid metabolism. Level of some amino acids increases in stress. Enhanced amino acid levels is indicator of stress response of S. pneumoniae to antibiotics117.
Purine metabolism	Energy transfer functions, cell signalling, production of GTP118.	Production of signalling molecules by purines have important role in formation of biofilm118.
Amino sugar and nucleotide sugar metabolism	Synthesis of various sugar derivatives and biomolecules. Synthesis of cell wall component and polysaccharides119.	High mucin environment adaptation of Akkermansia muciniphila119. Acid stress in S. thermophilus120.
Ribosome	Translation process121.	Formation of leaderless mRNA, it selectively induces under stress121.
Pyrimidine metabolism	Synthesis and degradation of pyrimidine molecules which are involved in DNA, RNA synthesis122.	Formation of biofilm. Formation of cell signalling molecules in stress123.
Pentose phosphate pathway	Source of biosynthetic intermediates for the synthesis of nucleotides, amino acids, vitamins, lipopolysaccharide and also fatty acids.  maintain carbon homoeostasis124.	Cell stabilizes NADPH level in oxidative and ROS stress by mediating this pathway125.
Alanine, aspartate and glutamate metabolism	Energy production, amino acid biosynthesis, nitrogen metabolism126.	Ciprofloxacin shows large decrease in Alanine, aspartate and glutamate metabolism127.
One carbon pool by folate	Synthesis of precursors for cellular functions, epigenetic maintenance128.	Mediate antibiotic stress, redox defense Biofilm formation129.
Cysteine and methionine metabolism	Sulphur metabolism130.	Play role in nutrient and oxidative stress reactive oxygen species stress131.
Glycine, serine and threonine metabolism	Interconvert through various enzymatic reactions. Have role in protein synthesis132.	Oxidative, osmotic stress, resistance to Kanamycin133.
Glyoxylate and dicarboxylate metabolism	Glyoxylate shunt allows bacteria to grow on acetate or fatty acids in Krebs cycle134.	Specific enzymes for this pathway have been shown to be upregulated in the stressed environment of biofilms135.
Pyruvate metabolism	Glycolysis, energy production, catabolism, anabolism136.	Fatty acid biosynthesis, and thereby affect membrane fluidity in acid stress137.
Oxidative phosphorylation	Generates ATP through electron transport chain and ATP synthase138.	Aerobic electron transfer chain components downregulated in stress138.
Arginine biosynthesis	Involved in urea cycle. Immune cell functions by NO production139.	Acid, oxidative and phosphate deprivation stress140.
Citrate cycle (TCA cycle)	energy production, carbon skeleton generation for biosynthesis, redox balance141.	TCA maintains the disrupted redox balance in stress142.
Starch and sucrose metabolism	Utilization and breakdown of carbohydrates as energy sources143.	Osmotic and nutrient deprivation stress143.
Nicotinate and nicotinamide metabolism	Glucose and lipid metabolism. Energy metabolism, DNA repair, epigenetic modification, inflammation, circadian rhythm144.	Enable cell to survive in changes such as nutrient oxidative stress deficiency145.
2-Oxocarboxylic acid metabolism	Metabolism of amino acids, carbohydrates, hydrocarbons and fats143.	Energy production and redox balance143.

A cross-talk analysis was carried out between the 20 common enriched pathways in all the five bacterial networks (Fig. 3). As observed from the figure, there were a high number of edges between the top enriched pathways with “carbon metabolism pathway” as the most critical mediator of the pathway crosstalk making interactions with all other common enriched pathways in all five bacterial networks. The carbon metabolism pathway makes the largest number of interactions with the biosynthesis of amino acids pathway followed by purine metabolism pathways in case of all 5 bacterial networks. This confirms the importance of interplay among these pathways in bacterial stress response. Also, majority of nodes involved in cross-talk are the central nodes of the stress-response networks.

Fig. 3.

Involvement of the 31 central proteins identified in various stress response pathways.

Discussion

A network biology approach was used to find the stress response in some emerging pathogenic bacteria. The microbes used in our study are pathogens commonly causing infections in hospital and community settings. This includes E. faecium, S. aureus, K. pneumonia, P. aeruginosa, and M. tuberculosis. A set of common proteins that have a central role in stress response in these five bacteria were determined. The previously reported roles of these proteins in physiological as well as stress conditions such as biofilm formation, developing resistance against antibiotics, SOS response generation, DNA repair, formation of other important biomolecules are shown in Table 4^46–95. Analysis using the Gene Ontology (GO) resource (https://geneontology.org/) showed that most of the central proteins identified were highly enriched in Transcription by Polymerase I-V and AMP biosynthetic processes. Other enriched processes were translesion synthesis, nucleoside monophosphate phosphorylation, DNA toplogical change, negative regulation of DNA transcription initiation, regulation of DNA transcripted transcription elongation, proton motive force-driven plasma membrane ATP synthesis, pyrimidine NTP biosynthesis, DNA templated transcription intiation, positive regulation of translation, glycolytic processes, cellular response to cell envelope stress, transcription anti-termination, and ribosomal small subunit assembly.

Central proteins of the stress response networks common to the stress response networks of all five pathogenic bacteria

• Several different distinct stress response systems have been described in bacteria like the SOS, generalized stress, temperature, acid, starvation, oxidative, envelope and osmotic stress responses. The generalized stress response of bacteria is triggered by many different stress signals like is often accompanied by growth cessation or reduction. It confers the ability to withstand the actual stress response encountered as well as stressors not yet encountered. In contrast to the response to specific stressors, this provides a preventive mechanism that renders the cell stress resistant and is also termed as cross stress protection⁹⁶. RpoS, the decisive master regulator of the generalized stress response, directs RNA polymerases to their promoters. RpoS controls over 1000 genes identified in E. coli by two separate studies^97,98. All the central proteins identified in this study are part of the RpoS regulon either as the first gene or as a part of the transcriptional unit (TU) as shown in Table 4.

A further CHIP-Seq investigation showed that expression of the genes in the regulon may vary dramatically in response to RpoS. Thus, sensitive genes are transcribed rapidly with the [production of only a little RpoS, those whose expression changes linearly with increase in RpoS while the expression of insensitive genes change very little⁹⁸. Only four of the central nodes identified were in the category of linear variation in concentration in response to RpoS levels, namely enolase, phosphoglycerate kinase, Heat shock protein 70 and CTP synthase. Many of the insensitive genes coincided with the glutamate-dependent acid resistance 2 system controlled by the GadEWX system that allows E. coli to survive at pH 2.0 and colonize the gastrointestinal tract⁹⁹. Though part of the regulon, the expression pattern of the central genes identified in this study largely did not significantly change in response to RpoS concentration. Also, they were not target genes for GadEWX regulon. These genes may be under additional control of other stress response systems.

• RecA, the main mediator of the DNA damage induced SOS response, is among the central proteins identified in this study. Apart from DNA damage, aging colonies, the SOS response is known to be activated by alkaline pH, cells reaching senescence in a medium, fluoroquinolone and β-lactam antibiotics. The RecA gene is important for controlling swarm motility, bacterial behaviour in biofilms and promotion of homologous recombination. It also upregulates integrase genes that in turn increase cassette rearrangement leading to spread of antibiotic resistance genes by horizontal gene transfer. RecA mediated SOS response is instrumental in promoting biofilm formation and development of persistence. Therefore, RecA has been proposed as a target for antibiotic resistance¹.

• The central proteins in our study also include dnaK, a key component of the heat shock stress response that is activated by the appearance of denatured or unfolded proteins in response to heat shock. DnaK is a chaperonin that has an additonal role in transport, transporting proteins for degradation, assembly of proteasome and temperature homeostasis. More recently, it has been shown to be involved in FliC mediated biofilm formation, cell adhesion to human plasminogen, bacterial growth, pathogenicity and antimicrobial resistance. DnaK also has a role in immunogenicity and evasion of host immune response with potential for adjuvant and immunomodulatory activity⁵⁴. The disruption of dnaK was shown to increase susceptibility of E. coli to fluroquinolones, methicillin and oxacillin¹⁰⁰. The methylene tetrahydrofolate encoded by FolD is part of the folate biosynthetic pathway, a common target of antibiotics¹⁰¹ and also involved in heat stress response⁵⁹.

• The adk gene codes for the adenylate kinase enzyme that catalyzes the conversion of AMP and ATP to two ADP molecules in the de novo ribonucleotide synthesis pathway. The transcriptional repression of adk has been reported to cause accumulation of (p)ppGpp, a stress alarmone of the stringent stress response. (p)ppGpp binds to RNA polymerase to achieve global rewiring of protein expression in stress such as amino acid starvation. with a marked reduction in cell size, reduced growth rates and partial dormancy.

• AtpD or the conserved beta subunit of the ATP synthase complex is an important catalytic component of the main energy transducing machinery¹⁰². Inhibitors that inhibit the ATP synthase complex have been studied in combination with various antibiotics to prolong their efficacy¹⁰³. As an example, reservatrol that between the beta and gamma subunit of ATP synthase improves the effectiveness of aminoglycoside antibiotics in S. aureus⁹¹.

• The adaptive oxidative stress response is activated in bacteria in response to high levels of reactive oxygen species that activates many efflux pump systems to achieve antibiotic resistance¹. Eno encodes a glycolytic enzyme that is also part of the RNA degradosome. Mutations of eno were shown to affect virulence and oxidative stress in P. aeruginosa. GyrB subunit belongs to the DNA gyrase that introduces negative supercoil into DNA to relieve mechanical stress and is a target for quinolone antibiotics. The SOS DNA damage response and the oxidative stress response is likely to be activated by inhibition of DNA gyrase¹⁰⁴.

• The LPS induced envelope stress response of bacteria is responsible for membrane homeostasis, sensing of external signals and cell damage repair. It is also the target site of many antibiotics. Osmotic stress predominantly affects bacterial cell structure and permeability. The osmotic stress response works at both the protein and gene transcription level, and includes efflux pump systems, Two component membrane bound histidine kinase systems and membrane bound chemoreceptors¹. CarB is a component of Carbamoyl phosphate synthase enzyme and is important for arginine and pyrimidine synthesis in E. coli¹⁰⁵. In other pathogenic bacteria, it has been implicated in maintaining cell wall integrity, responding to osmotic stress, virulence, motility and biofilm formation^106,107.

• The ribosome acts as a main checkpoint for maintaining the balance between protein synthesis, sequestration, turnover and degradation during stress response. Ribosomal biogenesis may be reduced or ribosomes may be silenced in order to protect them under stress conditions¹⁰⁸. RplA and rplB encode the protein L2 and L3 that are involved in the 50 S subunit of ribosomes. Both L2 and L3 are necessary for the functioning of the peptidyl transferase center^109,110. L2 additionally functions as a transcriptional modulator⁷⁸ while absence of methylation in L3 lead to a poor growth phenotype and cold sensitivity with accumulated ribosomal subunit precursors¹¹¹. Both L2 and L3 are targets for antibiotics.

• The cmk gene encodes a conserved cytidine monophosphate kinase that is essential for synthesis of nucleoside precursors. A Δcmk mutant in Yersinia pseudotuberculosis was growth deficient and highly attenuated while an antisense effect causes significant growth inhibition in S. aureus¹¹². However, the exact mechanisms underlying these effects have not yet been elucidated.

• GuaA codes for GMP synthetase, an enzyme that catalyzes the final step in the de novo synthesis of GMP. GTP has a key role in cellular processes and is an energy currency. The inverse relationship between (p)ppGpp has been proposed as a metabolic switch to control growth and survival¹¹³.

• PurB, PurE and PurH are related to de novo purine nucleotide biosynthesis and are regulated by PurR, a general transcription factor. Downregulation of the PuR regulated genes has been observed in stress conditions and may be associated with repression of purine biosynthesis during starvation¹¹⁴.

Systematic analysis of the 31 central proteins identified allowed us to categorize the involvement of these rpoS regulon proteins in various other types of stress responses as shown in Fig. 4. The stress response proteins found in our study were cross validated with the available dataset of CS response in E. coli and it was found that they coincide with each other. Twenty-nine proteins were found to have role in stress in this dataset also. The proteins are found to have role in transcription translational mechanisms, DNA repair pathways, cell division, energy metabolism, biosynthesis of important and amino acids.

Fig. 4.

Crosstalk between the common enriched pathways in the 5 bacterial stress response networks (a) SA-PPIN, (b) EF-PPIN, (c) MT-PPIN, (d) PA-PPIN and (e) KP-PPIN. The clusters represent the proteins in each numbered pathway, while edges show the interactions between the proteins in different pathways. The central nodes are depicted to be bigger than the other nodes. The pathway clusters are numbered in all the networks as follows: (1) 2-Oxocarboxylic acid metabolism; (2) Alanine, aspartate and glutamate metabolism; (3) Amino sugar and nucleotide sugar metabolism; (4) Arginine Biosynthesis; (5) Biosynthesis of amino acids; (6) Carbon metabolism; (7) Citrate cycle (TCA cycle); (8) Cysteine and methionine metabolism; (9) Glycine, serine and threonine metabolism; (10) Glyoxylate and dicarboxylate metabolism; 11. Nicotinate and nicotinamide metabolism; 12. One carbon by folate; 13. Oxidative Phosphorylation; 14. Pentose phosphate pathway; 15. Purine metabolism; 16. Pyrimidine metabolism; 17. Pyruvate metabolism; 18. Ribosome; 19. Starch and sucrose metabolism.

Common pathways in stress response in five pathogenic bacteria

Pathway enrichment analysis of all the networks identified 20 pathways common to all the stress response proteins of all the bacteria. The role of these common pathways in normal as well as stress conditions such as biofilm formation, signalling molecules formation, biosynthesis of amino acids, maintenance of redox balance, developing resistance against antibiotics etc., is shown in Table 5^115–146. Among these, some of the top enriched pathways are discussed below:

• 4.4.1Carbon metabolism- When the cell is in stress condition, it leads to depletion of vital nutrients from the cell. But activation of carbon metabolism pathway is responsible for the synthesis of important macromolecules in the cell. It starts with the uptake of glucose by phosphotransferase system and then a series of interconnected pathways comes into action such as glycolysis, gluconeogenesis, tricarboxylic cycle etc. This pathway performs the effective utilization of sources of carbon to energy¹⁴⁷.
• 4.4.2Biosynthesis of amino acids- This pathway involves many interconnected pathways involved in the synthesis of the amino acids which is very important for the synthesis of proteins e.g. in Tryptophan biosynthesis pathway secondary metabolites are synthesised in the cell which are not involved in the normal growth of the cell but provide selective advantage to them to sustain in the environment¹⁴⁸. They may be toxins which help in maintaining their chance of survival. Other pathways such as Aspartate pathway starts by an intermediate from the Citric acid cycle. During stress conditions some reactive oxygen species are produced which is metabolised by certain amino acids also some amino acids also help in cell wall synthesis¹⁴⁹.
• 4.4.3Purine and pyrimidine metabolism- The products of purine and pyrimidine metabolism are involved in processes like replication and transfer of signals in the cell. In stress conditions bacteria for its DNA repair acquire this pathway. Cell in stress sends signals for the production of toxins and other regulatory proteins¹⁵⁰.
• 4.4.4Oxidative phosphorylation- The process of oxidative phosphorylation is considered important process in organisms and it ensures the completion of all other vital processes in cell. The process of oxidative phosphorylation represents the formation of energy rich adenosine triphosphate. It involves energy generation by transfer of electrons from electron donors to electron acceptors in the presence of oxygen in electron transport chain. Oxidative phosphorylation is required in metabolism, growth, maintenance and active transport of electrons across cell membrane¹⁵¹.
• 4.4.5Pentose phosphate pathways- Pentose phosphate pathway has two phases oxidative and non-oxidative phase. It involves formation of pentose sugars for nucleotide synthesis in non-oxidative phase and formation of reducing equivalents such as nicotinamide adenine dinucleotide phosphate in oxidative phase¹⁵².

Cross validation with CS response dataset of E. coli showed that many of the central responsible for mediating stress response in different stress conditions as well as responsible for CS protection. This indicates that there are certain common stress response proteins in a wide range of different bacteria. However, to advance these targets into effective therapeutic strategies, further analysis is needed to clarify their precise role in bacterial pathogenesis and survival. The carbon metabolism pathway was shown to have a high significance in cross-talk analysis in our study, underscoring its potential impact on bacterial survival. This pathway has been considered as a promising target for antimicrobials as its disruption induces nutritional starvation stress¹⁵³.

Prioritized lists of drug targets have been generated for several diseases using high-quality interaction networks and analysis of differentially expressed genes¹⁵⁴. The general and specific bacterial stress responses have been proposed as potential targets for overcoming bacterial resistance to antibiotics^1,6. A recent study used network systems biology and transcriptomic analysis to identify critical stress-response genes in E. coli. The study revealed that bacterial responses to multiple stressors share overlapping genes and pathways, that regulate essential survival mechanisms like DNA repair and biofilm formation. By targeting these central nodes, researchers are developing adjuvant therapies that enhance antibiotic efficacy and reduce tolerance¹¹⁴. Stress adaptation systems like efflux pumps and biofilm formation contribute significantly to antibiotic resistance in bacteria. Thus, inhibiting these pathways and proteins could also lead to novel therapeutic strategies that work alongside antibiotics, effectively reducing resistance by compromising the bacteria’s adaptive capabilities¹. The identification of shared proteins and pathways in the bacterial stress response system paves the way for the development of broad-spectrum antibiotics or adjuvants that combat multidrug resistance.

Conclusion

A network biology approach was used to analyze stress response in five emerging pathogenic bacteria: E. faecium, S. aureus, K. pneumoniae, P. aeruginosa, and M. tuberculosis, which commonly cause infections in healthcare and community settings. Thirty-one core proteins were identified as mediators of cellular stress (CS) across these bacteria, with roles in DNA repair, biomolecule synthesis, protein activation, cellular processes, energy metabolism, and signaling. Cross-referencing with E. coli CS data validated 29 of these proteins, emphasizing their roles in transcription, translation, DNA repair, cell division, and biosynthesis. These conserved proteins may provide potential catalytic sites for developing new antibiotics and treatments against these pathogens. Pathway enrichment identified 20 shared pathways, including metabolism, antibiotic biosynthesis, carbon metabolism, and ribosome function. Further, cross-talk analysis highlighted the “carbon metabolism pathway” as the most critical mediator of the pathway crosstalk making interactions with all other common enriched pathways in all five bacterial networks. This study is important for effective targeting of multiple resistance, persistence and virulence determining phenomena as an antimicrobial strategy.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

GEO

Gene expression omnibus

DEGs

Differentially expressed genes

PPIN

Protein-protein interaction network

E. coli

Escherichia coli

CS

Cross- stress

PPI

Protein-protein interactions

FDR

False discovery rate

FPKM

Fragments per kilobase of transcript per million reads mapped

FC

Fold change

BC

Betweenness centrality

CS-PPIN

Cross-stress protein-protein interaction network

EF-PPIN

Enterococcus faecium Aus0004-PPIN

SA-PPIN

Staphylococcus aureus subsp. aureus USA300-PPIN

KP-PPIN

Klebsiella pneumoniae MGH 78578-PPIN

PA-PPIN

Pseudomonas aeruginosa PAO1-PPIN

MT-PPIN

Mycobacterium tuberculosis H37Rv -PPIN

Author contributions

A.S. and S.T. wrote the main manuscript text, A.S. prepared the figures. S.B. designed the work and revised it critically. All authors reviewed and approved the manuscript.

Funding

The authors would like to acknowledge funding from Department of Biotechnology grant number BT/PR40197/BTIS/137/68/2023.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its Supplementary information files. Additionally, the raw network files are available at https://github.com/sblab/Stress-response/tree/main.

Declarations

Competing interests

The authors declare no competing interests.