https://orcid.org/0000-0003-3865-1039
Figures
Abstract
Many pathogenic bacteria form biofilms to survive under environmental stresses and host immune defenses. Differential expression (DE) analysis of the genes in biofilm and planktonic cells under a single condition, however, has limitations to identify the genes essential for biofilm formation. Independent component analysis (ICA), a machine learning algorithm, was adopted to comprehensively identify the biofilm genes of Vibrio vulnificus, a fulminating human pathogen, in this study. ICA analyzed the large-scale transcriptome data of V. vulnificus cells under various biofilm and planktonic conditions and then identified a total of 72 sets of independently co-regulated genes, iModulons. Among the three iModulons specifically activated in biofilm cells, BrpT-iModulon mainly consisted of known genes of the regulon of BrpT, a transcriptional regulator controlling biofilm formation of V. vulnificus. Interestingly, the BrpT-iModulon additionally contained two novel genes, VV1_3061 and VV2_1694, designated as cabH and brpN, respectively. cabH and brpN were shared in other Vibrio species and not yet identified by DE analyses. Genetic and biochemical analyses revealed that cabH and brpN are directly up-regulated by BrpT. The deletion of cabH and brpN impaired the robust biofilm and rugose colony formation. CabH, structurally similar to the previously known calcium-binding matrix protein CabA, was essential for attachment to the surface. BrpN, carrying an acyltransferase-3 domain as observed in BrpL, played an important role in exopolysaccharide production. Altogether, ICA identified two novel genes, cabH and brpN, which are regulated by BrpT and essential for the development of robust biofilms and rugose colonies of V. vulnificus.
Author summary
Biofilm formation is essential for pathogenic bacteria to survive under environmental stresses and host immune surveillances. Unlike conventional differential expression (DE) analysis, independent component analysis (ICA), a machine learning algorithm, could comprehensively analyze gene expression profiles under various conditions of biofilm and planktonic cells and identify the genes essential for biofilm formation. In this study, ICA decomposed large-scale transcriptome data of a devastating pathogen Vibrio vulnificus into the iModulons, sets of independently co-regulated genes. Among the iModulons activated in biofilm cells, BrpT-iModulon contained two novel genes not yet identified by DE analyses, cabH and brpN, in addition to the previously known genes of the BrpT regulon. Our results indicated that CabH and BrpN contribute to biofilm and rugose colony formation. CabH, carrying calcium-binding repeats, and BrpN, harboring an acyltransferase-3 domain, were identified to play a crucial role in surface attachment and exopolysaccharide production, respectively. Consequently, ICA newly identified cabH and brpN which are regulated by BrpT and are essential for robust biofilm and rugose colony development of V. vulnificus.
Citation: Lee H, Im H, Hwang S-H, Ko D, Choi SH (2023) Two novel genes identified by large-scale transcriptomic analysis are essential for biofilm and rugose colony development of Vibrio vulnificus. PLoS Pathog 19(1): e1011064. https://doi.org/10.1371/journal.ppat.1011064
Editor: Matthew Parsek, University of Washington, UNITED STATES
Received: September 12, 2022; Accepted: December 13, 2022; Published: January 19, 2023
Copyright: © 2023 Lee et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: This work was supported by the National Research Foundation of Korea (NRF) and funded by the Ministry of Science, Information and Communications Technology, and Future Planning (https://www.msit.go.kr) (2021K1A3A1A20001134, to SHC). This work was also supported by Cooperative Research Program for Agriculture Science and Technology Development (Project No. PJ016298, to HL), Rural Development Administration (https://www.rda.go.kr/), Republic of Korea. The funders had no roles in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Bacteria often exist as a biofilm lifestyle, which is a sessile community attached to a surface [1]. Biofilms provide bacteria with protection from not only environmental stresses such as nutrient limitation but also host immune systems during infection [2]. Thus, biofilm formation is essential for the survival and virulence of a broad range of pathogens [3,4]. Biofilm formation includes sequential developmental stages composed of surface attachment, microcolony formation, and maturation into three-dimensional structures [1,5]. Mature biofilms are specialized and highly differentiated communities of bacteria covered by an extracellular polymeric matrix consisting of exopolysaccharides (EPSs), proteins, nucleic acids, and lipids [6]. Biofilm formation is essential for Vibrio vulnificus, a fulminating foodborne pathogen, to colonize and survive in oysters, the major infection route of the bacteria, and to cause diseases in mice [7–10].
Several studies have been conducted to understand the molecular mechanisms of biofilm formation in V. vulnificus [11–16]. BrpR, a transcriptional regulator governing biofilm formation in V. vulnificus, regulates downstream genes according to the intracellular levels of bis-(3´-5´)-cyclic dimeric guanosine monophosphate (c-di-GMP), which is a universal bacterial second messenger [11]. BrpR directly activates the expression of brpLG and brpT [11]. BrpL and BrpG play an important role in biofilm and rugose colony formation, through participating in EPS production [11]. BrpT is another transcriptional regulator controlling biofilm formation of V. vulnificus and activates the expression of brpS, cabABC, and the brp locus (brpABCDFHIJK) [12,13]. BrpS is also a transcriptional regulator that activates cabABC and represses brpT [13]. CabA is a calcium-binding protein and secreted to the cell exterior through CabB and CabC to form a structure of the biofilm matrix [14]. In concert with BrpL and BrpG, products of the brp locus also participate in the production of EPS [11,15]. In addition, another gene cluster named the rbd locus is involved in the biofilm formation of V. vulnificus [16]. The expression of the rbd locus is independent of c-di-GMP levels, and its activation signal has not yet been identified [16].
Differential expression (DE) analyses of the genes expressed more in a biofilm than in planktonic cells can simply predict a set of genes required for the biofilm formation under a single condition [17–19]. However, diverse sets of genes are preferentially expressed in a biofilm depending on the conditions such as its developmental stages and growth environments [20,21]. Therefore, independent component analysis (ICA), a machine learning-based algorithm decomposing a mixture of signals into independent components [22,23], can be adopted to comprehensively identify the genes essential for biofilm formation under various conditions. ICA analyzes the genes expressed from the cells under various conditions and identifies the sets of independently co-regulated genes, iModulons [24–26]. Since the genes comprising a specific iModulon are co-regulated, the regulatory mechanism for the newly discovered genes in an iModulon can be easily predicted [27,28].
In this study, ICA decomposed the large-scale transcriptome data of V. vulnificus obtained from biofilm and planktonic cells under various conditions into the iModulons. Among the iModulons specifically activated in biofilm cells, BrpT-iModulon mainly consisted of known genes of the BrpT regulon. Interestingly, the BrpT-iModulon additionally included two novel genes, VV1_3061 and VV2_1694, designated as cabH and brpN, respectively. Transcript and molecular biological analyses demonstrated that BrpT positively regulates cabH and brpN by directly binding to their upstream regions. Genetic analyses showed that cabH and brpN contribute to the development of biofilms through enhanced surface attachment and EPS production, respectively. Taken together, ICA comprehensively analyzed the gene expressions under various conditions of the biofilm and planktonic cells of V. vulnificus and newly identified the genes of the BrpT regulon essential for biofilm formation.
Results
ICA identified 72 iModulons of V. vulnificus
Raw read counts of RNA-seq data from V. vulnificus cells under 14 biofilm and 147 planktonic, a total of 161 conditions were gathered to create large-scale transcriptome data (Fig 1A and S1 and S2 Datasets). ICA decomposed the transcriptome data into 83 independent components containing the entire genes with varied gene coefficients (Fig 1B and S3 Dataset). Simultaneously, ICA also calculated the activities of the independent components to account for the expression of the constituting genes under the various conditions (Fig 1B and S4 Dataset). The cumulative explained variance (CEV) of the 83 independent component activities covered 80% of the total explained variance of the large-scale transcriptome data calculated by principal component analysis (PCA) (S1A Fig), verifying that ICA appropriately decomposed the transcriptome data.
(A) Individual gene expression profiles from biofilm and planktonic cells under a total of 161 conditions were combined into the large-scale transcriptome data. Red and blue represent the high and low expression levels of the genes, respectively. (B) ICA decomposed the transcriptome data into 83 independent components and outputted two matrices, ‘Gene coefficient’ and ‘Activity’. The ‘Gene coefficient’ matrix displayed the varied gene coefficients of 4,264 genes constituting the 83 independent components (S3 Dataset). The ‘Activity’ matrix indicated the activities of each independent component along the 161 conditions (S4 Dataset). (C) The independent component-01 was converted into the iModulon-01 by selecting the genes with significant gene coefficients. The same procedure was applied to the remaining 82 independent components.
The varied gene coefficients represent the effect of the changed activities of the independent component on the expression levels of the constituting genes. A positive or negative gene coefficient implies that the increased activities of the independent component activated or repressed the constituting gene expression, respectively. Most of the gene coefficients are distributed around zero (S1B Fig), indicating that the alteration of the independent component activities significantly affects only a few genes. Thereby, the significantly affected genes in each independent component are defined as the element genes of a specific iModulon (Fig 1C, see Materials and Methods for details). After the removal of iModulons containing less than 4 element genes, a total of 72 iModulons were identified from the 83 independent components (S1 Table and S5 Dataset).
The resulting 72 iModulons were annotated with distinct biological functions based on the known or predicted roles of the element genes (S1 Table). Then, the iModulons were grouped into 10 categories, such as ‘Structural components and signal transduction’, ‘Virulence and stress response’, ‘Translation’, ‘Metal homeostasis’, ‘Carbon and nitrogen source utilization’, ‘Amino acid and nucleotide biosynthesis’, and ‘Transport system’, according to their biological functions (Fig 2). The iModulons grouped into the ‘Genomic difference’ category account for the difference in gene expression between strains, such as either V. vulnificus MO6-24/O and CMCP6 (S2A Fig), or wild-type and regulator knockout mutants (S2B Fig). The iModulons with biological functions that do not belong to other categories were grouped into the ‘Miscellaneous functions’ category. The iModulons whose biological functions are unclear were grouped into the ‘Unknown’ category.
72 iModulons identified by ICA were grouped into 10 categories according to their biological functions as illustrated with different colors. The number of genes in each iModulon is indicated by the size of the box.
BrpT-iModulon contained cabH and brpN
To identify the iModulons containing the element genes essential for biofilm formation, the iModulon activities in the biofilm and planktonic cells were compared. The activities of iModulon-40, -57, and -64 significantly increased in the biofilm cells (S3 Fig), suggesting that their element genes are essential for biofilm formation. While most of the element genes comprising iModulon-40 and -64 were uncharacterized (S2 and S3 Tables), iModulon-57 consisted mostly of the genes of the regulon of BrpT, a known transcriptional regulator controlling biofilm formation of V. vulnificus (Fig 3A and Table 1) [11,13]. Also, the activities of iModulon-57 were greatly correlated with the expression levels of brpT (Pearson R = 0.74, P < 10−10) (Fig 3B). These observations led us to designate the iModulon-57 as BrpT-iModulon. Interestingly, two hypothetical element genes, VV1_3061 and VV2_1694, which were not known to be regulated by BrpT, were also found in the BrpT-iModulon (Fig 3A and Table 1). Since the other element genes in the BrpT-iModulon were known to be regulated by BrpT, it was possible that VV1_3061 and VV2_1694 are also the members of the BrpT regulon. To examine the possibility, the correlation between the expression levels of brpT and the VV1_3061 or VV2_1694 genes was examined. As shown in Fig 3C and 3D, the expression of VV1_3061 and VV2_1694 was positively correlated with that of brpT (Pearson R = 0.67, P < 10−10 and Pearson R = 0.46, P < 10−9, respectively), proposing that VV1_3061 and VV2_1694 might be activated by BrpT.
(A) A histogram of the gene coefficients in the BrpT-iModulon. The gray dashed lines show the cutoff values of the gene coefficients. Newly identified members of the BrpT regulon, VV1_3061 and VV2_1694, are presented with red bars. (B to D) Scatter plots of the expression levels of brpT and the activities of the BrpT-iModulon (B), the expression levels of VV1_3061 (C) and VV2_1694 (D). The expression levels are represented as log-TMM, log-transformed read counts after normalization using the trimmed mean of M-values (TMM) method. The Pearson R values between the two variables are denoted in the boxes. Each dot of the plots represents a single biological replicate. The red lines represent each regression line of the plots.
The open reading frames (ORFs) of VV1_3061 and VV2_1694 were predicted to encode a calcium-binding protein and an acyltransferase, respectively (Table 1). In V. vulnificus, cabA and brpL, the essential biofilm genes, also encode a calcium-binding protein and an acyltransferase, respectively [11, 14]. Thus, the structural similarities between the proteins encoded by VV1_3061 and cabA, and VV2_1694 and brpL were further investigated. As shown in Fig 4A, the repeats in the protein encoded by VV1_3061 were comparable to the calcium-binding repeats of CabA. Moreover, the structural superimposition of the two proteins showed the root-mean-square deviation (RMSD) value of 0.550 Å, indicating that these are highly analogous (Fig 4B). The structural similarity of the proteins led us to designate VV1_3061 as cabH. Meanwhile, a domain of the protein encoded by VV2_1694 was homologous to the acyltransferase-3 (AT3) domain of BrpL (Fig 4A). Although the whole structures of the proteins were not highly analogous (Fig 4C), the domain homologous to the AT3 domain of BrpL led us to designate VV2_1694 as brpN.
(A) Protein repeats and domains of CabA, CabH, BrpL, and BrpN predicted by using InterPro. The numerical values represent the amino acid numbers of each protein. IPR001343, RTX calcium-binding nonapeptide repeat; IPR002656, acyltransferase-3 domain. (B, C) Superimpositions of the predicted protein structures of CabA (green) and CabH (magenta) (B), and BrpL (green) and BrpN (magenta) (C). Protein structures were predicted by using the Alphafold2 algorithm and the superimpositions of the proteins were performed by using PyMOL. RMSD, root-mean-square deviation.
cabH and brpN are positively and directly regulated by BrpT
To experimentally verify that cabH and brpN are positively regulated by BrpT, the expression of cabH and brpN was compared in the brpT-deletion mutant (ΔbrpT) and its isogenic parent strain. Since BrpT is involved in c-di-GMP-induced biofilm formation [13,15], the JN111 strain, whose intracellular c-di-GMP levels are elevated by the addition of arabinose [14], was used as the parent strain in this study. The transcript levels of cabH and brpN in ΔbrpT were significantly decreased compared with those in the parent strain (Fig 5A). The decreased transcript levels of cabH and brpN in ΔbrpT were restored by complementation of brpT (Fig 5B). These results confirmed that the expression of cabH and brpN is positively regulated by BrpT. To investigate whether BrpT regulates cabH and brpN directly, BrpT was purified, and its binding to the upstream regions of cabH and brpN was examined by electrophoretic mobility shift assays (EMSAs). The addition of BrpT to the labeled DNA probes resulted in retarded bands in a BrpT concentration-dependent manner (Fig 5C and 5D). The binding of BrpT was specific because the assays were performed in the presence of poly(dI-dC) as a non-specific competitor. Moreover, the same unlabeled DNA fragment, which was used as a self-competitor, showed competition for the BrpT binding in a dose-dependent manner (Fig 5C and 5D), verifying the specific binding of BrpT to the upstream regions of cabH and brpN. Taken together, these results indicated that BrpT regulates the expression of cabH and brpN by directly binding to their upstream regions.
(A, B) Effects of BrpT on the expression of cabH and brpN. Total RNAs were isolated from the V. vulnificus strains grown in LB supplemented with 0.01% (w/v) arabinose (A) or 0.03% arabinose, 100 μg/ml kanamycin, and 100 μg/ml ampicillin for complementation experiments (B). The expression of brpT, cabH, and brpN was determined by qRT-PCR analysis, and the expression of each gene in the parent strain was set as 1 (n = 3). Error bars represent the SD. Statistical significance was determined by the Student’s t test (**, P < 0.005). ND, not detected. Parent and Parent (pJK1113), parent strain; ΔbrpT and ΔbrpT (pJK1113), brpT mutant; ΔbrpT (pJK1113::brpT), complemented strain. (C, D) Direct binding of BrpT to upstream regions of cabH and brpN. The 6-FAM labeled DNA fragments (10 nM) of the upstream regions of cabH (C) and brpN (D) were incubated with increasing amounts of BrpT as indicated. For competition analysis, the same but unlabeled DNA fragments were used as self-competitors. Various amounts of self-competitors were added as indicated to the reaction mixtures before the addition of BrpT (4 μM). Each gel representing the mean result from at least three independent experiments was photographed using ChemiDoc Touch Imaging System (Bio-Rad). B, bound DNA; F, free DNA.
CabH and BrpN are essential for biofilm and rugose colony formation
Since cabH and brpN were included in the BrpT-iModulon that was activated in biofilm cells (Table 1 and S3B Fig), the involvement of cabH and brpN in biofilm formation was further investigated. The biofilm levels of the cabH-deletion mutant (ΔcabH) and brpN-deletion mutant (ΔbrpN) were reduced when compared with those of the parent strain (Fig 6A). The reduced biofilm levels of ΔcabH and ΔbrpN were restored by complementation of cabH and brpN, respectively (Fig 6B). These results indicated that CabH and BrpN are crucial for biofilm formation of V. vulnificus. In addition, the biofilm level of ΔbrpT decreased when compared with that of the parent strain (Fig 6A), further supporting the previous observation that cabH and brpN are activated by BrpT (Fig 5).
(A, B) For quantitative analysis of the biofilm, biofilms of the V. vulnificus strains were grown in VFMG supplemented with 0.01% arabinose (A) or 0.01% arabinose, 100 μg/ml kanamycin, and 100 μg/ml ampicillin for complementation experiments (B) in 96-well polystyrene microtiter plates for 24 h. Then supernatants were removed from the wells and the remaining biofilms were stained with 1% crystal violet. The crystal violet was eluted and its A570 was determined to quantify the biofilms (n = 3). Error bars represent the SD. Statistical significance was determined by the Student’s t test (*, P < 0.05; **, P < 0.005). Parent and Parent (pJK1113), parent strain; ΔcabH and ΔcabH (pJK1113), cabH mutant; ΔbrpN and ΔbrpN (pJK1113), brpN mutant; ΔbrpT, brpT mutant; ΔcabH (pJK1113::cabH) and ΔbrpN (pJK1113::brpN), complemented strains.
The colony morphology of ΔcabH and ΔbrpN was also compared with that of the parent strain. In accordance with the diminished biofilm formation (Fig 6), ΔcabH and ΔbrpN exhibited reduced colony rugosity when compared with the parent strain (Fig 7A). The changed colony morphologies of ΔcabH and ΔbrpN were recovered by complementation of cabH and brpN, respectively (Fig 7B). These results demonstrated that CabH and BrpN are important for the rugose colony development of V. vulnificus. Furthermore, the reduction of colony rugosity caused by the deletion of brpN was more severe than that of cabH (Fig 7A and 7B), suggesting that BrpN plays a more crucial role in rugose colony development. Additionally, ΔbrpT showed a smooth colony (Fig 7A), again supporting the previous observation that cabH and brpN are activated by BrpT (Fig 5).
(A, B) The V. vulnificus strains were spotted onto VFMG agar supplemented with 0.02% arabinose (A) or 0.02% arabinose, 100 μg/ml kanamycin, and 100 μg/ml ampicillin for complementation experiments (B) and incubated for 24 h. Each colony representing the mean rugosity from at least three independent experiments was visualized using a stereomicroscope (Stemi 305, Zeiss). All images are shown at the same scale, and 1-mm scale bars are shown on the images of the parent strain. Parent and Parent (pJK1113), parent strain; ΔcabH and ΔcabH (pJK1113), cabH mutant; ΔbrpN and ΔbrpN (pJK1113), brpN mutant; ΔbrpT, brpT mutant; ΔcabH (pJK1113::cabH) and ΔbrpN (pJK1113::brpN), complemented strains.
CabH is crucial for surface attachment
Biofilms of ΔcabH and ΔbrpN were further visualized in large scales using polystyrene and glass test tubes. As shown in Fig 8, when compared with the parent strain, ΔcabH exhibited reduced biofilm formation, especially defects in the attachment to both polystyrene and glass tubes. These observations suggest that CabH contributes to robust biofilm formation by enhancing surface attachment. The role of CabH on surface attachment was further compared with that of CabA, a calcium-binding protein [14]. As shown in Fig 8, the cabA-deletion mutant (ΔcabA) reduced biofilm formation of V. vulnificus, but the attachment of ΔcabA to the test tubes was comparable with that of the parent strain, confirming the previous observation that CabA is important for biofilm structural development rather than surface attachment [14]. Additionally, one interesting point was that among the mutants, ΔcabH specifically showed more reduced biofilm formation in smooth-surfaced glass tubes than in polystyrene tubes (Fig 8). The result suggests that ΔcabH with impaired surface attachment has more difficulties in attaching to smoother surfaces.
(A, B) Biofilms of the V. vulnificus strains were grown in VFMG supplemented with 0.01% arabinose in polystyrene (A) or glass (B) test tubes for 24 h. Then supernatants were removed from the tubes and the remaining biofilms were stained with 1% crystal violet. The stained biofilms were then washed with vibration in PBS to remove loosely attached cells. To evaluate the attachment to the test tube surface, each test tube representing the mean result from at least three independent experiments was photographed using a mobile camera. Stained, photographed after the removal of the crystal violet solution; Washed, photographed after being washed with PBS. (C, D) After being washed with PBS, biofilms in the polystyrene (C) or glass (D) test tubes were quantified by eluting the crystal violet and measuring its A570 (n = 3). Error bars represent the SD. Statistical significance was determined by the Student’s t test (*, P < 0.05; **, P < 0.005; ns, not significant). Parent, parent strain; ΔcabH, cabH mutant; ΔbrpN, brpN mutant; ΔcabA, cabA mutant.
BrpN is involved in EPS production
Previously, BrpN was predicted as an AT3 domain-containing acyltransferase (Fig 4A) and essential for the biofilm and rugose colony formation (Figs 6, 7, and 8). However, the attachment of ΔbrpN to the test tubes was comparable with that of the parent strain (Fig 8A and 8B), suggesting that BrpN contributes to biofilm and rugose colony formation through a function other than surface attachment. Since the AT3 domain-containing acyltransferase is known to contribute to EPS biosynthesis [29, 30], the EPS production of ΔbrpN was further compared with that of the parent strain using SDS-PAGE. EPS extract from the parent strain showed a high intensity, but that from ΔbrpN exhibited a much lower intensity (Fig 9). This result confirmed that BrpN participates in EPS production, implying that reduced biofilm and rugose colony formation of ΔbrpN (Figs 6, 7, and 8) is due to diminished EPS production. In addition, the role of BrpN in EPS production was further compared with that of another AT3 domain-containing protein BrpL. As shown in Fig 9, the EPS production from ΔbrpN was lower than that from the brpL-deletion mutant (ΔbrpL). Furthermore, the EPS production from the brpN brpL double-deletion mutant (ΔbrpN ΔbrpL) was even lower than that from either ΔbrpN or ΔbrpL (Fig 9). Together, these results suggest that BrpN is more crucial than BrpL in EPS production, and their activities for EPS production are additive.
(A) EPS extracts were prepared from the V. vulnificus strains and then resolved on a 4% polyacrylamide gel by SDS-PAGE. The gel was stained with Stains-All and the gel representing the mean result from at least three independent experiments was photographed using a mobile camera. (B) EPS production of the mutants was quantified from the intensity of each lane, and the production of the parent strain was set as 100% in each experiment (n = 3). Error bars represent the SD. Statistical significance was determined by the Student’s t test (*, P < 0.05; **, P < 0.005). Parent, parent strain; ΔbrpN, brpN mutant; ΔbrpL, brpL mutant; ΔbrpN ΔbrpL, brpN brpL double mutant.
Discussion
In this study, ICA decomposed the large-scale transcriptome data from 14 biofilm and 147 planktonic cells of V. vulnificus into 72 iModulons (Figs 1 and 2, S1 Table, and S1, S2, and S5 Datasets). Among the iModulons, iModulon-40, -57, and -64 showed significantly increased activities in the biofilm cells (S3 Fig). iModulon-40 included the rbdIJ genes, which contribute to biofilm formation (S2 Table and S4A Fig) [16]. The signal for the expression of rbdIJ is not c-di-GMP, a known signal for bacterial biofilm formation, and rather still unknown [16]. Therefore, identification of the actual signal for the expression of rbdIJ could extend our understanding of the mechanisms for V. vulnificus biofilm formation. Meanwhile, most of the characterized element genes in iModulon-64 were involved in central carbon metabolism (S3 Table and S4B Fig). Biofilm formation can be affected by the availability of carbon sources, and thereby is regulated concurrently with central carbon metabolism in several pathogens [31–33]. Further investigation of the regulatory networks of iModulon-64 may reveal the link between biofilm formation and the nutritional availability of V. vulnificus.
iModulon-57, designated as BrpT-iModulon, contained two hypothetical element genes, VV1_3061 and VV2_1694, which are located apart from the previously known BrpT regulon and thus have not yet been recognized as the BrpT regulon (Fig 3 and Table 1). The product of VV1_3061 was named CabH and VV2_1694 was named BrpN, based on the structural similarity with CabA and domain homogeneity with BrpL, respectively (Fig 4). The deletion of brpT significantly decreased the expression levels of cabH and brpN when compared with those in the parent strain (Fig 5A and 5B). In addition, the direct and specific binding of BrpT to the upstream regions of cabH and brpN was demonstrated (Fig 5C and 5D). The combined results indicated that cabH and brpN are the novel members of the BrpT regulon and up-regulated directly by BrpT. Interestingly, cabH and brpN have not yet been identified by comparisons of the biofilm and planktonic transcriptome data of V. vulnificus under a specific condition (S5 Fig). These results suggest that ICA decomposition of the large-scale transcriptome data under various biofilm and planktonic conditions can uncover novel genes involved in bacterial biofilm formation, yet unrevealed by conventional DE analyses.
The cabH gene is a newly identified member of the BrpT regulon and predicted to encode CabH containing calcium-binding repeats (Fig 4A). Homologs of CabH are found in other Vibrio species such as Vibrio sp. 03_296, V. fluvialis, and V. parahaemolyticus, indicating that the protein may be conserved in other Vibrio species. However, no functional roles of CabH have yet been identified. Deletion of cabH reduced biofilm levels along with the decreased attachment of V. vulnificus to the test tubes (Figs 6 and 8), suggesting that CabH contributes to biofilm formation by playing a crucial role in surface attachment. The protein structure of CabH is highly analogous to CabA (Fig 4B), which is secreted to the cell exterior by the CabBC secretion system and a component of a biofilm matrix [14]. In addition, cabH is activated by BrpT along with cabBC (Fig 5) [13]. These findings suggest that CabH is also a biofilm matrix component secreted by CabBC. The reduced colony rugosity of ΔcabH further suggests that CabH is a biofilm matrix constituent secreted out of the cell (Fig 7). Moreover, since CabH possesses calcium-binding repeats similar to those of CabA (Fig 4A), CabH would be functional when the extracellular calcium is present, as observed in CabA [14].
The brpN gene is another newly identified member of the BrpT regulon and predicted to encode an acyltransferase containing an AT3 domain (Fig 4A), which is involved in EPS biosynthesis [29,30]. Homologs of BrpN are found in Vibrio species such as Vibrio sp. 03_296 and V. mytili, indicating that the protein is also conserved in other Vibrio species. However, no functional roles of BrpN have yet been identified. Deletion of brpN reduced biofilm formation and colony rugosity along with decreased EPS production of V. vulnificus, suggesting that BrpN contributes to biofilm and rugose colony development by producing EPS (Figs 6, 7, 8 and 9). Additionally, EPS production was more diminished in ΔbrpN than in ΔbrpL, and further reduced in ΔbrpN ΔbrpL (Fig 9). Consequently, the biofilm level and colony rugosity were more decreased in ΔbrpN than in ΔbrpL, and further diminished in ΔbrpN ΔbrpL (S6 Fig). These results suggest that BrpN is more crucial for EPS production, biofilm formation, and rugose colony development than BrpL, and the two acyltransferases function additively. Interestingly, brpN is in chromosome II of V. vulnificus, distantly located from the other EPS genes involving brpLG and the brp locus. Moreover, the GC content of brpN is significantly different from that of the other EPS genes (S4 Table), indicating that V. vulnificus may have received brpN transferred horizontally to further elaborate its EPS production [34,35].
As depicted in Fig 10, this study expanded the current understanding of V. vulnificus biofilm formation using ICA. ICA decomposed the large-scale transcriptome data from the biofilm and planktonic cells under diverse conditions into iModulons. Further investigation of the BrpT-iModulon, activated in the biofilm cells, discovered cabH and brpN as the new members of the BrpT regulon. BrpT directly activates the expression of cabH and brpN in addition to brpS, the cabABC operon, and the brp locus. BrpS activates cabABC expression and represses brpT expression to constitute a negative feedback loop tuning the brpT expression precisely. CabH, structurally similar to a biofilm matrix protein CabA, has calcium-binding repeats and contributes to the surface attachment. BrpN carries an AT3 domain and participates in the biosynthesis of EPS along with the proteins encoded by the brp locus. Consequently, ICA identified the novel genes of the BrpT regulon, cabH and brpN, which were not discovered by conventional DE analyses, but essential for robust biofilm and rugose colony development of V. vulnificus. Accordingly, ICA can also be adopted to identify the genes required for biofilm formation of other pathogens and to understand their characteristics.
BrpT-iModulon contained cabH and brpN, the newly discovered genes in this study, in addition to brpS, cabABC, and the brp locus. Transcriptional regulator BrpT activates the expression of cabH, brpN, brpS, cabABC, and the brp locus. Transcriptional regulator BrpS activates the expression of cabABC and represses the expression of brpT. CabH is involved in the surface attachment and CabA is crucial for biofilm structural development. BrpN and the proteins expressed from the brp locus participate in the EPS biosynthesis. This sophisticated regulatory network elaborately controls robust biofilm and rugose colony development in V. vulnificus.
Materials and methods
Generation of the input data set of ICA comprising the V. vulnificus transcriptome data
Raw sequencing reads of available RNA-seq data of V. vulnificus were collected from the Sequence Read Archive (SRA) database at NCBI (https://www.ncbi.nlm.nih.gov/sra) (S1 Dataset). All reads were trimmed using Trimmomatic (v0.39) [36] with the following options “LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36”. The trimmed reads were mapped to the reference genome of V. vulnificus (GenBank accession numbers NC_014965.1 for chromosome I and NC_014966.1 for chromosome II) using STAR aligner (v2.7.8a) with the following options “—outSAMtype BAM—alignIntronMax 1” [37]. The mapped reads were counted using HTSeq (v0.13.5) with the following options “-s no -r pos -t gene -f bam -m intersection-strict” [38]. To ensure the quality of the data, the genes with less than 10 counts per million (CPM) were removed before further analysis. The read counts were normalized using the trimmed mean of M-values (TMM) method from the edgeR package and log-transformed by log2(TMM+1), referred to as log-TMM [39,40]. Biological replicates with R2 < 0.9 between log-TMM values were removed to eliminate the technical noise, and the remaining transcriptome data were used as the input data set of ICA (S2 Dataset).
Identification of iModulons using ICA
ICA was performed on the input data set as described previously [25,28]. The FastICA algorithm from the Scikit-learn Python package (v0.22.1) was executed 256 times with random seeds to construct independent components [41,42]. The resulting independent components were clustered using the DBSCAN algorithm from the Scikit-learn Python package to identify the reproducible components [43]. To select the co-regulated genes of each independent component, the gene with the largest absolute gene coefficient value was removed repeatedly, and D’Agostino’s K2 test statistic was calculated for each remaining distribution after the removal [44]. Once the test statistic dropped below a cutoff, the removed genes were designated as the co-regulated genes, the element genes of the iModulons.
To decide the K2 test statistic cutoff, RegulonDB (v10.9) was used to perform a two-sided Fisher’s exact test (FDR < 10−5) between the top 30 element genes with the highest absolute gene coefficients of each independent component and the previously known genes of the regulon of every regulator [45,46]. Among the regulators, a regulator with the lowest P value was linked to each independent component. Then, the F1-scores were calculated between the regulon of the independent component-linked regulators and the element genes in the independent components selected on the basis of the K2 test statistic cutoff varying from 200 to 1,000 with an increment of 50. Since the average of calculated F1-scores was highest at the K2 test statistic cutoff of 600 (S1C Fig), this cutoff was used to determine the iModulons [26]. The iModulons containing less than 4 genes were discarded and thus, 72 iModulons were identified eventually.
Calculation of CEV of PCA and ICA
PCA of the input data set was conducted with the Scikit-learn Python package (v0.22.1) [42]. The CEV of the PCA results was calculated by adding the explained variance ratios of the principal components in the order of numbers using the NumPy Python package [47]. The CEV of the ICA results was calculated by adding the explained variance ratios of the independent components as described above. The Matplotlib Python package was used to visualize the CEV of the PCA and ICA results [48].
Correlation analysis of the expression levels of the genes or the activities of the iModulons
The expression levels of the genes and activities of the iModulons for each RNA-seq condition were obtained from S2 and S4 Datasets, respectively, by using Pandas and NumPy Python packages [47,49]. The Pearson correlation analyses between the expression levels of a gene and the activities of an iModulon, or between the expression levels of two genes were conducted with the SciPy Python package [50]. The Matplotlib Python package was used to visualize the plots of the correlation analysis [48].
Prediction, modeling, and visualization of the protein structures
The repeats and domains of the proteins were predicted by using InterPro (v88.0) [51]. The structural predictions of the full-length proteins were obtained from the Alphafold2 algorithm [52]. Visualization of the predicted protein structures and superimposition of two proteins was conducted with PyMOL (v2.5.2) [53].
Strains, plasmids, and culture conditions
The strains and plasmids used in this study are listed in S5 Table. Unless otherwise noted, the V. vulnificus strains were grown aerobically in Luria-Bertani (LB) medium supplemented with 2% (w/v) NaCl (LBS) at 30°C. The Vibrio fischeri minimal medium containing glycerol (50 mM Tris-HCl, pH 7.2, 50 mM MgSO4, 300 mM NaCl, 10 mM KCl, 0.33 mM K2HPO4, 18.5 mM NH4Cl, 10 mM CaCl2, and 32.6 mM glycerol) (VFMG) was used for biofilm formation [11]. When required, antibiotics were added to the media at the following concentrations: kanamycin, 100 μg/ml and ampicillin, 100 μg/ml. To manipulate the intracellular c-di-GMP levels, V. vulnificus JN111, which carries dcpA encoding a diguanylate cyclase [54] on the chromosome under the control of the arabinose-inducible promoter PBAD [55], was constructed from the wild-type strain V. vulnificus CMCP6 previously [14]. JN111 was used as a parent strain in this study (S5 Table), and intracellular c-di-GMP levels of the V. vulnificus strains were manipulated by adding different concentrations of arabinose to the growth media.
Generation and complementation of the deletion mutants
For the construction of the isogenic deletion mutants, target genes were inactivated in vitro by deletion of each ORF using the polymerase chain reaction (PCR)-mediated linker-scanning mutation as described previously [56]. Briefly, the deleted ORF fragment was amplified by PCR with appropriate primer pairs (S6 Table), and the amplified fragment was cloned into SphI-SpeI-digested pDM4 [57]. E. coli S17–1 λpir, containing pDM4 with the desired insert, was used as a conjugal donor to the parent strain to generate the deletion mutant (S5 Table). The conjugation and isolation of the transconjugants were conducted using a method described previously [58]. The brpN brpL double mutant was generated identically from the brpL mutant by additional deletion of brpN. For complementation of the deletions, respective ORF was amplified by PCR with appropriate primer pairs (S6 Table), and the amplified fragment was cloned into NcoI-SphI-digested pJK1113 [59] under an arabinose-inducible promoter PBAD. The plasmids were transferred into the appropriate mutants by conjugation as described previously [60].
RNA purification and qRT-PCR
Relative transcript levels in the total RNA isolated from the V. vulnificus strains grown to the stationary phase (an A600 of 2.0) in LB (added arabinose and antibiotics when required) were determined by quantitative reverse transcription-PCR (qRT-PCR). Total RNA from the V. vulnificus cells was isolated by using an RNeasy mini kit (Qiagen, Valencia, CA, United States) and quantified using a NanoDrop One spectrophotometer (Thermo Scientific, Waltham, MA, United States). cDNA was synthesized from 400 ng of the total RNA using a PrimeScript RT Master Mix reagent kit (Takara, Osaka, Japan) in a final volume of 20 μl. Real-time PCR amplification of the cDNA was performed with the CFX96 real-time PCR detection system (Bio-Rad, Hercules, CA, United States) with appropriate primer pairs (S6 Table) as described previously [61]. Relative expression levels of the target gene were calculated by using the 16S ribosomal RNA, rrsH, as an internal reference for normalization.
Protein purification and EMSA
To overexpress BrpT, pSH1819 carrying the brpT gene on pET-28a(+) (Novagen, Madison, WI, United States) was used as described previously [13]. The His6-tagged BrpT was expressed in E. coli BL21 (DE3) and purified by affinity chromatography using Ni-NTA agarose (Qiagen). For EMSA, the 279-bp cabH upstream region was amplified by PCR using unlabeled CABHUP_F and 6-carboxyfluorescein (6-FAM)-labeled CABHUP_R as primers (S6 Table). Similarly, the 300-bp brpN upstream region was amplified by PCR using unlabeled BRPNUP_F and 6-FAM-labeled BRPNUP_R as primers (S6 Table). The 6-FAM-labeled DNA (10 nM) was incubated with different amounts of purified BrpT for 45 min at 25°C in a 20-μl reaction mixture containing 1 × BrpT-binding buffer (10 mM Tris-HCl, pH 7.5, 50 mM KCl, 75 mM NaCl, 10 mM MgCl2, 1 mM DTT, 0.1 mM EDTA, 1 μg bovine serum albumin, 5% (w/v) glycerol), and 0.1 μg of poly(dI-dC) as a non-specific competitor. For the competition analysis, various concentrations of unlabeled DNA fragments were added as a self-competitor to the reaction mixture before incubation. Electrophoretic analysis for the DNA-protein complexes was performed as described previously [62].
Quantitative analysis and visualization of the biofilms
To quantify the biofilms of the V. vulnificus strains, each well of the 96-well polystyrene microtiter plates (Nunc, Roskilde, Denmark) was inoculated with 200 μl of culture diluted to an A600 of 0.05 in VFMG supplemented with 0.01% (w/v) arabinose (added antibiotics when required). After static incubation for 24 h at 30°C, supernatants were removed from the wells and the remaining biofilms were stained with 1% (w/v) crystal violet solution for 15 min. The biofilms were quantified by elution of stained biofilms with ethanol and measurement of absorbance at 570 nm (A570) as described previously [63]. To visualize the biofilms, biofilms of the strains were formed and stained as described above but in a larger scale (1 ml) using polystyrene (SPL, Seoul, Republic of Korea) or glass round-bottom test tubes. To remove loosely attached cells, the remaining biofilms were washed with a vibration of 1,200 rpm for 20 sec in 1 ml of phosphate-buffered saline (PBS). The stained biofilms were photographed by a mobile camera after the removal of the crystal violet solution or PBS. Additionally, after being washed with PBS, the biofilms were quantified by elution of stained biofilms with ethanol and measurement of A570 as described above.
Colony morphology assay
For the analysis of the colony morphology, 2 μl of cultures grown to an A600 of 0.8 were spotted onto VFMG agar supplemented with 0.02% (w/v) arabinose (added antibiotics when required). The resulting colonies grown at 30°C for 24 h were visualized using a Stemi 305 stereomicroscope (Zeiss, Oberkochen, Germany) equipped with an Axiocam 105 color camera (Zeiss).
EPS analysis
Exopolysaccharide was prepared following the procedures previously described [64]. Briefly, each culture grown on an LBS agar plate containing 0.02% (w/v) arabinose was suspended in PBS and diluted to an A600 of 1.0. The suspensions were vigorously shaken to elute the EPS from the cells. The cells and debris were removed by centrifugation, and the supernatant was treated with RNase A (50 μl/ml), DNase I (50 μg/ml with 10 mM MgCl2), and proteinase K (200 μg/ml). Subsequently, the remaining polysaccharide fraction was extracted with phenol-chloroform, precipitated with 2.5 × volumes of 100% ethanol, and resuspended in distilled water. The EPS resuspensions were resolved on a 4% polyacrylamide gel by SDS-PAGE and stained with Stains-All (Sigma-Aldrich, St. Louis, MO, United States). The gel was subsequently destained as described previously [65] and photographed by a mobile camera. The intensity of stained EPS in each lane was determined using ImageJ software (NIH, Bethesda, MD, United States), and the ratio of the intensity of mutants to that of the parent strain was calculated.
Data analysis
Average and standard deviation (SD) values were calculated from at least three independent experiments. The experimental data were analyzed by Student’s t tests using GraphPad Prism 7.0 (GraphPad Software, San Diego, CA, United States). The significance of the differences between experimental groups was accepted at a P value of < 0.05.
Supporting information
S1 Fig. Summary of data processing.
(A) Cumulative explained variance (CEV) for the transcriptome data of V. vulnificus was calculated by using principal component analysis (PCA, blue line) and independent component analysis (ICA, orange line). Using the 83 independent components, ICA reconstructed 80% of the total explained variance of the transcriptome data calculated by PCA. (B) A histogram of the gene coefficients in the entire independent components. Most of the gene coefficients were distributed around zero, and the gene coefficients in any independent component display a similar distribution. (C) Average F1-scores calculated under the varied D’Agostino K2 statistic cutoff ranging from 200 to 1,000 with an increment of 50. The optimal cutoff value was identified as 600 where the highest average F1-score was observed (shown in a red dot).
https://doi.org/10.1371/journal.ppat.1011064.s001
(TIF)
S2 Fig. iModulons identified by genomic differences between strains.
(A, B) Box plots of the iModulon activities under various conditions of the V. vulnificus cells. Box plot whiskers, box bounds, and the center line represent extrema, upper and lower quartiles, and the median value, respectively. MO6-24/O; CMCP6; ΔnsrR; ΔsmcR; ΔhlyU; others, transcriptome data from V. vulnificus MO6-24/O; V. vulnificus CMCP6; the nsrR mutant; the smcR mutant; the hlyU mutant; all other strains except the corresponding mutants, respectively.
https://doi.org/10.1371/journal.ppat.1011064.s002
(TIF)
S3 Fig. Identification of the iModulons specifically activated in biofilm cells.
(A to C) Box plots of the iModulon activities under various conditions of biofilm and planktonic cells. Box plot whiskers, box bounds, and the center line represent extrema, upper and lower quartiles, and the median value, respectively. The gray dashed line represents zero. Early and mature biofilm cells indicate the biofilm cells incubated in VFMG for 1.5 h and 6~13 h, respectively. Among the various conditions of planktonic cells in the transcriptome data, conditions using M9G, LBS, and MHB as the growth medium were selected as the representatives. M9G, M9 minimal medium supplemented with 0.4% (w/v) glucose; LBS, Luria-Bertani (LB) medium supplemented with 2% (w/v) NaCl; MHB, Mueller-Hinton broth.
https://doi.org/10.1371/journal.ppat.1011064.s003
(TIF)
S4 Fig. Validation of iModulon-40 and iModulon-64.
(A, B) Ordered correlation matrix. Colors indicate the Pearson R values between the activities of iModulon-40 (A) or iModulon-64 (B) and the expression levels of the element genes, respectively. Red and blue represent the strongest positive (+1) and negative (-1) correlation, respectively.
https://doi.org/10.1371/journal.ppat.1011064.s004
(TIF)
S5 Fig. Differentially expressed genes in the V. vulnificus biofilm cells.
(A to C) Volcano plots from differential expression analysis between biofilm and planktonic cells. Total RNAs were isolated from the biofilm and planktonic V. vulnificus cells after incubation for 1.5 h (A), 6 h (B), and 10 h (C) in VFMG. Transcriptome analysis plotted the genes down-regulated or up-regulated in the biofilm cells (see S1 Methods for details). The black dots represent differentially expressed genes and the red dots represent cabH and brpN as indicated. The gray dashed lines indicate the cutoffs for differential expression of fold change > 2 and P value < 0.05.
https://doi.org/10.1371/journal.ppat.1011064.s005
(TIF)
S6 Fig. Biofilm formation and colony morphology of the brpN, brpL, and brpN brpL double mutants.
(A) Effects of brpN and brpL deletions on biofilm formation. For quantitative analysis of the biofilm, biofilms of the V. vulnificus strains were grown in VFMG supplemented with 0.01% arabinose in 96-well polystyrene microtiter plates for 24 h. Then supernatants were removed from the wells and the remaining biofilms were stained with 1% crystal violet. The crystal violet was eluted and its A570 was determined to quantify the biofilms (n = 3). Error bars represent the SD. Statistical significance was determined by the Student’s t test (*, P < 0.05; **, P < 0.005). (B) Effects of brpN and brpL deletions on colony morphology. The parent strain and mutants were spotted onto VFMG agar supplemented with 0.02% arabinose and incubated for 24 h. Each colony representing the mean rugosity from at least three independent experiments was visualized using a stereomicroscope (Stemi 305, Zeiss). All images are shown at the same scale, and a 1-mm scale bar is shown on the image of the parent strain. Parent, parent strain; ΔbrpN, brpN mutant; ΔbrpL, brpL mutant; ΔbrpN ΔbrpL, brpN brpL double mutant.
https://doi.org/10.1371/journal.ppat.1011064.s006
(TIF)
S1 Table. Overview of the iModulons identified by ICA.
https://doi.org/10.1371/journal.ppat.1011064.s007
(DOCX)
S2 Table. The element genes included in the iModulon-40.
https://doi.org/10.1371/journal.ppat.1011064.s008
(DOCX)
S3 Table. The element genes included in the iModulon-64.
https://doi.org/10.1371/journal.ppat.1011064.s009
(DOCX)
S4 Table. The GC contents of brpN, brpLG, the brp locus, and chromosome II.
https://doi.org/10.1371/journal.ppat.1011064.s010
(DOCX)
S5 Table. Bacterial strains and plasmids used in this study.
https://doi.org/10.1371/journal.ppat.1011064.s011
(DOCX)
S6 Table. Oligonucleotides used in this study.
https://doi.org/10.1371/journal.ppat.1011064.s012
(DOCX)
S1 Dataset. Information of raw sequencing reads of V. vulnificus transcriptome data retrieved from the SRA database at NCBI.
https://doi.org/10.1371/journal.ppat.1011064.s013
(XLSX)
S2 Dataset. Input data set of ICA comprising 4,264 genes and 161 conditions.
https://doi.org/10.1371/journal.ppat.1011064.s014
(XLSX)
S3 Dataset. Gene coefficients of 4,264 genes constituting the 83 independent components.
https://doi.org/10.1371/journal.ppat.1011064.s015
(XLSX)
S4 Dataset. Activities of the 83 independent components along the 161 conditions.
https://doi.org/10.1371/journal.ppat.1011064.s016
(XLSX)
S5 Dataset. Identified iModulons and their element genes.
https://doi.org/10.1371/journal.ppat.1011064.s017
(XLSX)
S1 Data. Underlying numerical values of Figs 5A, 5B, 6A, 6B, 8C, 8D, and 9B and S1A, S1C, S5A, S5B, S5C, and S6A Figs.
https://doi.org/10.1371/journal.ppat.1011064.s018
(XLSX)
S1 Methods. Differential expression analysis of the genes in biofilm and planktonic cells.
https://doi.org/10.1371/journal.ppat.1011064.s019
(DOCX)
References
- 1. O’Toole G, Kaplan HB, Kolter R. Biofilm formation as microbial development. Annu Rev Microbiol. 2000;54:49–79. pmid:11018124
- 2. Hall-Stoodley L, Stoodley P. Biofilm formation and dispersal and the transmission of human pathogens. Trends Microbiol. 2005;13(1):7–10. pmid:15639625
- 3. Koo H, Allan RN, Howlin RP, Stoodley P, Hall-Stoodley L. Targeting microbial biofilms: current and prospective therapeutic strategies. Nat Rev Microbiol. 2017;15(12):740–55. pmid:28944770
- 4. Costerton JW. Cystic fibrosis pathogenesis and the role of biofilms in persistent infection. Trends Microbiol. 2001;9(2):50–2. pmid:11173226
- 5. Watnick P, Kolter R. Biofilm, city of microbes. J Bacteriol. 2000;182(10):2675–9. pmid:10781532
- 6. Flemming HC, Wingender J. The biofilm matrix. Nat Rev Microbiol. 2010;8(9):623–33. pmid:20676145
- 7. Choi G, Choi SH. Complex regulatory networks of virulence factors in Vibrio vulnificus. Trends Microbiol. 2022;30(12):1205–16. pmid:35753865
- 8. Kim SM, Park JH, Lee HS, Kim WB, Ryu JM, Han HJ, et al. LuxR Homologue SmcR Is Essential for Vibrio vulnificus Pathogenesis and Biofilm Detachment, and Its Expression is Induced by Host Cells. Infect Immun. 2013;81(10):3721–30. pmid:23897607
- 9. Froelich B, Oliver JD. The Interactions of Vibrio vulnificus and the Oyster Crassostrea virginica. Microb Ecol. 2013;65(4):807–16. pmid:23280497
- 10. Pu M, Duriez P, Arazi M, Rowe-Magnus DA. A conserved tad pilus promotes Vibrio vulnificus oyster colonization. Environ Microbiol. 2018;20(2):828–41. pmid:29235706
- 11. Hwang SH, Im H, Choi SH. A Master Regulator BrpR Coordinates the Expression of Multiple Loci for Robust Biofilm and Rugose Colony Development in Vibrio vulnificus. Front Microbiol. 2021;12. pmid:34248894
- 12. Chodur DM, Rowe-Magnus DA. Complex Control of a Genomic Island Governing Biofilm and Rugose Colony Development in Vibrio vulnificus. J Bacteriol. 2018;200(16). pmid:29760209
- 13. Hwang SH, Park JH, Lee B, Choi SH. A Regulatory Network Controls cabABC Expression Leading to Biofilm and Rugose Colony Development in Vibrio vulnificus. Front Microbiol. 2020;10. pmid:32010109
- 14. Park JH, Jo Y, Jang SY, Kwon H, Irie Y, Parsek MR, et al. The cabABC Operon Essential for Biofilm and Rugose Colony Development in Vibrio vulnificus. Plos Pathog. 2015;11(10). pmid:26406498
- 15. Guo YZ, Rowe-Magnus DA. Identification of a c-di-GMP-Regulated Polysaccharide Locus Governing Stress Resistance and Biofilm and Rugose Colony Formation in Vibrio vulnificus. Infect Immun. 2010;78(3):1390–402. pmid:20065022
- 16. Guo YZ, Rowe-Magnus DA. Overlapping and unique contributions of two conserved polysaccharide loci in governing distinct survival phenotypes in Vibrio vulnificus. Environ Microbiol. 2011;13(11):2888–900. pmid:21895917
- 17. Whiteley M, Bangera MG, Bumgarner RE, Parsek MR, Teitzel GM, Lory S, et al. Gene expression in Pseudomonas aeruginosa biofilms. Nature. 2001;413(6858):860–4. pmid:11677611
- 18. Resch A, Rosenstein R, Nerz C, Gotz F. Differential gene expression profiling of Staphylococcus aureus cultivated under biofilm and planktonic conditions. Appl Environ Microb. 2005;71(5):2663–76. pmid:15870358
- 19. Shemesh M, Tam A, Steinberg D. Differential gene expression profiling of Streptococcus mutans cultured under biofilm and planktonic conditions. Microbiol-Sgm. 2007;153:1307–17. pmid:17464045
- 20. An DD, Parsek MR. The promise and peril of transcriptional profiling in biofilm communities. Curr Opin Microbiol. 2007;10(3):292–6. pmid:17573234
- 21. Stewart PS, Franklin MJ. Physiological heterogeneity in biofilms. Nat Rev Microbiol. 2008;6(3):199–210. pmid:18264116
- 22. Hyvarinen A, Oja E. Independent component analysis: algorithms and applications. Neural Networks. 2000;13(4–5):411–30. pmid:10946390
- 23. James CJ, Hesse CW. Independent component analysis for biomedical signals. Physiol Meas. 2005;26(1):R15–R39. pmid:15742873
- 24. Sastry AV, Hu A, Heckmann D, Poudel S, Kavvas E, Palsson BO. Independent component analysis recovers consistent regulatory signals from disparate datasets. Plos Comput Biol. 2021;17(2). pmid:33529205
- 25. Sastry AV, Gao Y, Szubin R, Hefner Y, Xu SB, Kim D, et al. The Escherichia coli transcriptome mostly consists of independently regulated modules. Nat Commun. 2019;10. pmid:31797920
- 26. Rychel K, Decker K, Sastry AV, Phaneuf PV, Poudel S, Palsson BO. iModulonDB: a knowledgebase of microbial transcriptional regulation derived from machine learning. Nucleic Acids Res. 2021;49(D1):D112–D20. pmid:33045728
- 27. Rajput A, Tsunemoto H, Sastry AV, Szubin R, Rychel K, Sugie J, et al. Machine learning from Pseudomonas aeruginosa transcriptomes identifies independently modulated sets of genes associated with known transcriptional regulators. Nucleic Acids Res. 2022;50(7):3658–72. pmid:35357493
- 28. Im H, Lee JH, Choi SH. Independent Component Analysis Identifies the Modulons Expanding the Transcriptional Regulatory Networks of Enterohemorrhagic Escherichia coli. Front Microbiol. 2022;13:953404. pmid:35814713
- 29. Pearson C, Tindall S, Potts JR, Thomas GH, van der Woude MW. Diverse functions for acyltransferase-3 proteins in the modification of bacterial cell surfaces. Microbiology. 2022;168(3). pmid:35253642
- 30. Pearson CR, Tindall SN, Herman R, Jenkins HT, Bateman A, Thomas GH, et al. Acetylation of Surface Carbohydrates in Bacterial Pathogens Requires Coordinated Action of a Two-Domain Membrane-Bound Acyltransferase. Mbio. 2020;11(4). pmid:32843546
- 31. Bitoun JP, Liao SM, Yao X, Xie GG, Wen ZZT. The Redox-Sensing Regulator Rex Modulates Central Carbon Metabolism, Stress Tolerance Response and Biofilm Formation by Streptococcus mutans. Plos One. 2012;7(9). pmid:23028612
- 32. Sadykov MR, Hartmann T, Mattes TA, Hiatt M, Jann NJ, Zhu YF, et al. CcpA coordinates central metabolism and biofilm formation in Staphylococcus epidermidis. Microbiol-Sgm. 2011;157:3458–68. pmid:21964732
- 33. O’Toole GA, Gibbs KA, Hager PW, Phibbs PV, Kolter R. The global carbon metabolism regulator Crc is a component of a signal transduction pathway required for biofilm development by Pseudomonas aeruginosa. J Bacteriol. 2000;182(2):425–31. pmid:10629189
- 34. Syvanen M. Horizontal Gene-Transfer: Evidence and Possible Consequences. Annu Rev Genet. 1994;28:237–61. pmid:7893125
- 35. Ravenhall M, Skunca N, Lassalle F, Dessimoz C. Inferring Horizontal Gene Transfer. Plos Comput Biol. 2015;11(5). pmid:26020646
- 36. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30(15):2114–20. pmid:24695404
- 37. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29(1):15–21. pmid:23104886
- 38. Anders S, Pyl PT, Huber W. HTSeq—a Python framework to work with high-throughput sequencing data. Bioinformatics. 2015;31(2):166–9. pmid:25260700
- 39. Robinson MD, Oshlack A. A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biol. 2010;11(3). pmid:20196867
- 40. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2010;26(1):139–40. pmid:19910308
- 41. Hyvarinen A. Fast and robust fixed-point algorithms for independent component analysis. Ieee T Neural Networ. 1999;10(3):626–34. pmid:18252563
- 42. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011;12:2825–30.
- 43. Ester M, Kriegel H-P, Sander J, Xu X. A density-based algorithm for discovering clusters in large spatial databases with noise. Kdd. 1996;96:226–231.
- 44. Dagostino RB, Belanger A, Dagostino RB. A Suggestion for Using Powerful and Informative Tests of Normality. Am Stat. 1990;44(4):316–21.
- 45. Salgado H, Santos A, Garza-Ramos U, van Helden J, Diaz E, Collado-Vides J. RegulonDB (version 2.0): a database on transcriptional regulation in Escherichia coli. Nucleic Acids Res. 1999;27(1):59–60. pmid:9847141
- 46. Santos-Zavaleta A, Salgado H, Gama-Castro S, Sanchez-Perez M, Gomez-Romero L, Ledezma-Tejeida D, et al. RegulonDB v 10.5: tackling challenges to unify classic and high throughput knowledge of gene regulation in E. coli K-12. Nucleic Acids Res. 2019;47(D1):D212–D20. pmid:30395280
- 47. Harris CR, Millman KJ, van der Walt SJ, Gommers R, Virtanen P, Cournapeau D, et al. Array programming with NumPy. Nature. 2020;585(7825):357–62. pmid:32939066
- 48. Hunter JD. Matplotlib: A 2D graphics environment. Comput Sci Eng. 2007;9(3):90–5.
- 49. McKinney W. Data structures for statistical computing in python. Proceedings of the 9th Python in Science Conference. 2010;445:51–6.
- 50. Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, Cournapeau D, et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat Methods. 2020;17(3):261–72. pmid:32015543
- 51. Hunter S, Apweiler R, Attwood TK, Bairoch A, Bateman A, Binns D, et al. InterPro: the integrative protein signature database. Nucleic Acids Res. 2009;37:D211–D5. pmid:18940856
- 52. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–9. pmid:34265844
- 53. DeLano WL. Pymol: An open-source molecular graphics tool. CCP4 Newsl Protein Crystallogr. 2002;40(1):82–92.
- 54. Nakhamchik A, Wilde C, Rowe-Magnus DA. Cyclic-di-GMP regulates extracellular polysaccharide production, biofilm formation, and rugose colony development by Vibrio vulnificus. Appl Environ Microb. 2008;74(13):4199–209. pmid:18487410
- 55. Guzman LM, Belin D, Carson MJ, Beckwith J. Tight Regulation, Modulation, and High-Level Expression by Vectors Containing the Arabinose PBAD Promoter. J Bacteriol. 1995;177(14):4121–30. pmid:7608087
- 56. Jang KK, Gil SY, Lim JG, Choi SH. Regulatory Characteristics of Vibrio vulnificus gbpA Gene Encoding a Mucin-binding Protein Essential for Pathogenesis. J Biol Chem. 2016;291(11):5774–87. pmid:26755724
- 57. Milton DL, OToole R, Horstedt P, WolfWatz H. Flagellin A is essential for the virulence of Vibrio anguillarum. J Bacteriol. 1996;178(5):1310–9. pmid:8631707
- 58. Jang KK, Lee ZW, Kim B, Jung YH, Han HJ, Kim MH, et al. Identification and characterization of Vibrio vulnificus plpA encoding a phospholipase A2 essential for pathogenesis. J Biol Chem. 2017;292(41):17129–43. pmid:28855258
- 59. Lim JG, Bang YJ, Choi SH. Characterization of the Vibrio vulnificus 1-Cys Peroxiredoxin Prx3 and Regulation of Its Expression by the Fe-S Cluster Regulator IscR in Response to Oxidative Stress and Iron Starvation. J Biol Chem. 2014;289(52):36263–74. pmid:25398878
- 60. Choi G, Kim D, Im H, Choi SH. A Nitric Oxide-Responsive Transcriptional Regulator NsrR Cooperates With Lrp and CRP to Tightly Control the hmpA Gene in Vibrio vulnificus. Front Microbiol. 2021;12. pmid:34093504
- 61. Choi G, Jang KK, Lim JG, Lee ZW, Im H, Choi SH. The transcriptional regulator IscR integrates host-derived nitrosative stress and iron starvation in activation of the vvhBA operon in Vibrio vulnificus. J Biol Chem. 2020;295(16):5350–61. pmid:32169898
- 62. Lee ZW, Hwang SH, Choi G, Jang KK, Lee TH, Chung KM, et al. A MARTX Toxin rtxA Gene Is Controlled by Host Environmental Signals through a CRP-Coordinated Regulatory Network in Vibrio vulnificus. Mbio. 2020;11(4). pmid:32723914
- 63. Ko D, Choi SH. Comparative genomics reveals an SNP potentially leading to phenotypic diversity of Salmonella enterica serovar Enteritidis. Microb Genom. 2021;7(5). pmid:33952386
- 64. Kim HS, Lee MA, Chun SJ, Park SJ, Lee KH. Role of NtrC in biofilm formation via controlling expression of the gene encoding an ADP-glycero-manno-heptose-6-epimerase in the pathogenic bacterium, Vibrio vulnificus. Mol Microbiol. 2007;63(2):559–74. pmid:17241201
- 65. Kelley JT, Parker CD. Identification and Preliminary Characterization of Vibrio Cholerae Outer Membrane Proteins. J Bacteriol. 1981;145(2):1018–24. pmid:7462157