Effect of Quorum Sensing on the Ability of Desulfovibrio vulgaris To Form Biofilms and To Biocorrode Carbon Steel in Saline Conditions

Sulfate-reducing bacteria (SRB) are considered key contributors to biocorrosion, particularly in saline environments. Biocorrosion imposes tremendous economic costs, and common approaches to mitigate this problem involve the use of toxic and hazardous chemicals (e.g., chlorine), which raise health and environmental safety concerns. Quorum-sensing inhibitors (QSIs) can be used as an alternative approach to inhibit biofilm formation and biocorrosion. However, this approach would only be effective if SRB rely on QS for the pathways associated with biocorrosion. These pathways would include biofilm formation, electron transfer, and metabolism. This study demonstrates the role of QS in Desulfovibrio vulgaris on the above-mentioned pathways through both phenotypic measurements and transcriptomic approach. The results of this study suggest that QSIs can be used to mitigate SRB-induced corrosion problems in ecologically sensitive areas.

ABSTRACT

Sulfate-reducing bacteria (SRB) are key contributors to microbe-induced corrosion (MIC), which can lead to serious economic and environmental impact. The presence of a biofilm significantly increases the MIC rate. Inhibition of the quorum-sensing (QS) system is a promising alternative approach to prevent biofilm formation in various industrial settings, especially considering the significant ecological impact of conventional chemical-based mitigation strategies. In this study, the effect of the QS stimulation and inhibition on Desulfovibrio vulgaris is described in terms of anaerobic respiration, cell activity, biofilm formation, and biocorrosion of carbon steel. All these traits were repressed when bacteria were in contact with QS inhibitors but enhanced upon exposure to QS signal molecules compared to the control. The difference in the treatments was confirmed by transcriptomic analysis performed at different time points after treatment application. Genes related to lactate and pyruvate metabolism, sulfate reduction, electron transfer, and biofilm formation were downregulated upon QS inhibition. In contrast, QS stimulation led to an upregulation of the above-mentioned genes compared to the control. In summary, these results reveal the impact of QS on the activity of D. vulgaris, paving the way toward the prevention of corrosive SRB biofilm formation via QS inhibition.

IMPORTANCE Sulfate-reducing bacteria (SRB) are considered key contributors to biocorrosion, particularly in saline environments. Biocorrosion imposes tremendous economic costs, and common approaches to mitigate this problem involve the use of toxic and hazardous chemicals (e.g., chlorine), which raise health and environmental safety concerns. Quorum-sensing inhibitors (QSIs) can be used as an alternative approach to inhibit biofilm formation and biocorrosion. However, this approach would only be effective if SRB rely on QS for the pathways associated with biocorrosion. These pathways would include biofilm formation, electron transfer, and metabolism. This study demonstrates the role of QS in Desulfovibrio vulgaris on the above-mentioned pathways through both phenotypic measurements and transcriptomic approach. The results of this study suggest that QSIs can be used to mitigate SRB-induced corrosion problems in ecologically sensitive areas.

INTRODUCTION

Sulfate-reducing bacteria (SRB) are anaerobic microorganisms that utilize hydrogen and a range of organic compounds, such as lactate, acetate, pyruvate, and malate, to reduce sulfate and produce hydrogen sulfide (H₂S) (1). Sulfate reducers are generally thought to play an important role in the corrosion of metal surfaces exposed to seawater (2, 3). Although corrosion is mainly a chemical process involving metal oxidation and dissolution, it was found that SRB utilize hydrogen during sulfate reduction, which in turn affects the chemical dissolution of metal surfaces (4). Corrosive H₂S produced by SRB further compromises the structural integrity of metals (i.e., chemical microbe-induced corrosion [CMIC]). Furthermore, members of SRB, such as Desulfovibrionaceae and Desulfobulbaceae, can directly uptake electrons from the metal through pili, nanowires, or outer membrane proteins and cause corrosion (i.e., electrical microbe-induced corrosion [EMIC]) (4, 5).

For both CMIC and EMIC, the presence of a biofilm matrix establishes anoxic niches in which SRB proliferate resulting in localized corrosion and pits (6). In addition, biofilm matrix facilitates direct contact between the metal surface and bacterial outer membrane proteins (e.g., cytochromes and hydrogenases) or with electroconductive nanowires (7, 8). In Desulfovibrio vulgaris, biofilm formation is dependent on filament and flagellar biosynthesis (9, 10). Coincidentally, in Pseudomonas aeruginosa and Vibrio cholerae, some of their genes related to biofilm formation (e.g., rhamnolipids [11] and extracellular polymeric substance [12] production) are controlled by the quorum-sensing (QS) system. This led us to wonder whether biofilm formation and subsequently biocorrosion could also be associated with QS in D. vulgaris.

QS is a communication system based on the exchange of small molecules called autoinducers. When the density of cells reaches a threshold, autoinducer binds to receptors to initiate a cascade of reactions that lead to either expression or repression of certain genes (13). QS systems of Vibrio fischeri (14), Vibrio harveyi (15), and Pseudomonas aeruginosa (16) are relatively well studied compared to QS systems in SRB, where the exact mechanism of the QS, as well as its linkage to biocorrosion, remains largely unknown.

Data mining in the NCBI database revealed the presence of QS protein homologs in many SRB (17). In particular, homologs of proteins involved in the QS-controlled phosphorylation cascade (LuxR and LuxO) and in the acyl homoserine lactone (AHL) synthesis were found in D. vulgaris (17, 18). This is verified by the detection of different carbon chains of AHLs, including C₈-AHL, C₁₀-AHL, and C₁₂-AHL, in pure cultures of D. vulgaris (19, 20). Furthermore, it was found that exposing SRB to QS inhibitors in saline conditions diminished biofilm formation, sulfate reduction, and AHL production (20). This finding suggests a possible connection between QS and the above-mentioned bacterial functions. Nevertheless, a complete understanding of the QS effect at the gene level is still missing. The role of QS in SRB-induced biocorrosion is also not well elucidated.

This study aims to understand the role of the QS system in SRB and to identify which genes correlate with the QS system using D. vulgaris as a model marine bacterium. Specifically, we sought to assess the influence of the QS system on biofilm formation and biocorrosion in a seawater environment. To achieve these objectives, the QS system of D. vulgaris was perturbed using either a stimulating cocktail of AHLs or an inhibitory cocktail of QS inhibitors (QSIs). The bacterial response to the two treatments was analyzed in terms of cell activity, lactate and sulfate consumption, biofilm formation, and biocorrosion capacity. These bacterial traits were also investigated through gene expression pattern upon treatment application. The results indicate a correlation between the expression or repression of genes (e.g., cell activity, biofilm formation, sulfate reduction, electron transfer, and lactate metabolism) with the stimulation or repression of the QS system (through AHL or QSI application). This study provides an improved understanding on the role of QS in D. vulgaris in initiating biocorrosion, and it suggests the effectiveness of the QS inhibition approach to control SRB-associated biocorrosion.

RESULTS

Effect of QS modulation on planktonic bacteria.

The ATP concentration per cell increased at the same rate prior to AHL or QSI treatment (Fig. 1A). At 72 h, AHL or QSI treatment was applied. This treatment time was selected based on the previous finding that, in D. vulgaris, a possible connection between sulfate reduction and the alteration of the QS system is most significant during the early exponential phase (20). After QSI treatment, ATP concentration per cell was significantly lower compared to the control (P < 0.05), with a difference that ranged from 0.12 to 0.30 pM/cell. In contrast, AHL treatment resulted in an ATP concentration in average of 0.16 pM/cell higher than that of control (Fig. 1A). Despite the difference in ATP concentration per cell, the different treatments did not result in a significant change (P > 0.05) in cell number among control, AHL, and QSI treatments (see Fig. S1 in the supplemental material).

FIG 1.

Concentrations of ATP per cell (A) and of lactate (B), sulfate (C), and acetate (D) in the planktonic phase during the experimental time in control, QSI [100 μM (Z-)-4-bromo-5-(bromomethylene)-3-butyl-2(5H)-furanone, 100 μM 3-oxo-C₁₂-(2-aminocyclohexanone), and 5 mM γ-aminobutyric acid], and AHL (N-octanoyl-homoserine lactone [C₈-HSL], N-decanoyl-homoserine lactone [C₁₀-HSL], and N-dodecanoyl-homoserine lactone [C₁₂-HSL], 5 μM each) treatments and under abiotic control conditions. The dashed red line indicates the moment of the treatment injection (72 h). Error bars show the standard deviations from three biological replicates. Statistical differences were evaluated by one-way ANOVA, with a confidence level of 95% (P < 0.05).

Lactate was not utilized in the abiotic control but was consumed at similar rates in all the conditions prior to treatment (Fig. 1B and Table S1). After AHL application, lactate consumption rate was 0.45 ± 0.05 mM/h and was significantly higher compared to the other conditions (P < 0.05) (see Table S1 in the supplemental material). The total lactate utilization at the end of the experiment shows that QSI treatment resulted in the lowest total consumption compared to the control and AHL treatments (P < 0.05) (Table S1). Sulfate consumption (Fig. 1C) followed a similar trend of lactate utilization, with a significantly lower sulfate reduction rate upon QS inhibition (P < 0.05) (Table S1).

Due to increased lactate utilization rates upon AHL treatment, acetate was consequentially produced at a higher rate (0.71 ± 0.10 mM/h) compared to both control and QS inhibition condition (P < 0.05) (Table S1), and the acetate concentration stabilized at ca. 48 mM (Fig. 1D).

Biofilm and surface characterization.

Coupon biofilm analysis showed that a significantly higher number of cells and ATP per surface unit were obtained with AHL treatment, whereas QSI addition significantly decreased both parameters compared to the control (P < 0.05) (Fig. 2A).

FIG 2.

Coupon biofilm ATP concentration and cell density (A), surface roughness index and maximum pit depth (B), corrosion rate calculated through linear polarization resistance (C), and total corrosion rate measured by weight loss analysis (D) in the control, QSI, and AHL treatments (as described in the legend to Fig. 1) and under abiotic control conditions. Error bars show the standard deviations from three biological replicates. Statistical differences were evaluated through one-way ANOVA, with a confidence level of 95% (P < 0.05).

QSI and AHL treatments altered the impact of biocorrosion on the carbon steel coupon surface. Surface roughness and maximum pit depth (Fig. 2B) were both higher when the coupons were in AHL-treated cultures compared to the control (P < 0.05), whereas these surface topography values were significantly lower when QSI was applied (P < 0.05). The extent of surface biocorrosion can be observed in Fig. S2A through S2C, where scanning electron microscopy (SEM) images showed higher density of pits was obtained when bacteria were in contact with the AHL cocktail. The medium alone caused slight or no signs of biocorrosion (Fig. S2D). Atomic force microscopy (AFM) photos of coupon surfaces reiterated the same observation (Fig. S3). No apparent surface pits were observed in abiotic controls that contained only the AHL and QSI molecules (Fig. S4).

Corrosion evaluation through electrochemical and weight loss analysis.

Linear polarization resistance and weight loss analyses were conducted to quantitatively evaluate the respective daily and total corrosion rates arising from the different treatments. An increase in the corrosion rate was observed at day 4, after 1 day from treatment exposure, for all of the biotic conditions. AHL-treated samples showed a rapid increase in the corrosion rate that reached 1.1 mm/year, while QSI treatment resulted in a lower increase in the corrosion rates compared to control and AHL treatment (Fig. 2C). Under all conditions, corrosion rates started to decrease after day 5. Weight loss analysis further confirmed that the highest total corrosion rate was achieved upon AHL treatment (0.41 mm/year, P < 0.05), while the lowest was obtained in the presence of QSI (ca. 0.08 mm/year, P < 0.05) (Fig. 2D).

Overview of gene expression.

Genes that correlate with the QS system perturbation were selected based on the scoring system presented in the supplemental material. A total of 162 genes, accounting for 4.6% of the total genes present in D. vulgaris, were found to correlate with QS stimulation or inhibition using a score of ≥2. Particularly, 45 of these genes (1.3% of the total genes) were highly correlated with the QS system (a score of ≥3). Figure 3 shows the summarized numbers of differentially expressed genes, while Table S2 lists the gene fold changes compared to the control for each of the treatments at each time point. In general, at 6 h of AHL or QSI treatment, the number of downregulated genes was comparable between the two treatments, although there were more genes showing upregulation in the presence of AHL, in particular the ones related to biofilm formation, electron transfer, and the regulatory genes (Fig. 3). Subsequently, stimulation of the QS system through AHL treatment resulted in an upregulation of most of the genes at 24 and 48 h. An opposite trend was observed upon QSI treatment, where most of the genes were downregulated (Fig. 3).

FIG 3.

Numbers of differently expressed genes grouped in functions in AHL and QSI treatments at three time points compared to the relative control. Upregulated genes are shown in red; downregulated genes are shown in blue. The size of the circle represents the number of differently expressed genes (DEGs).

Regulatory genes.

Both QS stimulation (by AHL treatment) and inhibition (by QSI treatment) led to expression of genes encoding regulatory proteins such as transcriptional and response regulators and histidine kinases (Table S2.1). Transcriptional regulator and histidine kinase are part of the gene machinery that allows the QS system to regulate different bacterial functions. A histidine kinase and two transcriptional regulators belonging to the LuxR family were strongly correlated with QS stimulation or inhibition, indicating that the assigned scoring system is able to pinpoint QS-associated genes. Furthermore, some of these genes that correlated with the QS stimulation or inhibition were found to be grouped in the same operons and therefore likely to work in the same pathway or to interact with each other (Table S2.1).

Cell activity and membrane transport.

The two treatments showed an important influence on cell activity. ATP synthase genes were upregulated 24 and 48 h from AHL addition and downregulated after QSI treatment compared to the relative control (Fig. 3 and 4, in yellow), suggesting a higher cellular activity in AHL-treated samples compared to the other conditions. This observation is in agreement with the higher ATP concentration measured in AHL-treated samples (Fig. 2A).

FIG 4.

Simplified representation of the D. vulgaris cell, including proteins encoded in genes related to lactate and pyruvate metabolism (in green), sulfate reduction (in blue), electron transfer (in red), and ATP synthases (in yellow). Proteins shown in a solid color are encoded in genes that are controlled by the QS system (protein complexes are shown in a solid color when one of their components was controlled by QS). A dark or light color represents a higher or a lower grade of QS influence, respectively. A net pattern indicates proteins encoded by genes that were not impacted by the QS modulation (CM, cytoplasmic membrane, OM, outer membrane). The figure was adapted from Strittmatter et al. (44).

Genes encoding the large and small subunits of ribosomal proteins were upregulated by AHL treatment, indicating higher protein production and cell activity (Fig. 3). Conversely, QSI treatment had a slightly lower impact in the downregulation of these genes compared to the control. From this group, four large and four small ribosomal subunits were strongly correlated with the QS system modulation (Table S2.3) and were located in a total of three operons (Table S2.3).

Similarly, the expression of genes related to membrane transfer, particularly those encoding ATP-binding cassette (ABC) transporters, also correlated with the AHL and QSI treatments (Table S2.4). Some of the ABC transporter genes were involved in the biogenesis of cytochromes and thus facilitate electron transfer.

Metabolism and anaerobic respiration.

D. vulgaris is able to use various electron donors, such as lactate, pyruvate, and hydrogen, to reduce sulfate and generate energy (21). Figure 4 illustrates genes related to lactate and pyruvate metabolism, sulfate reduction, and electron transfer that were affected by QS system perturbation in our study.

Genes related to lactate oxidation, such as lactate dehydrogenase (ldh), pyruvate-flavodoxin oxidoreductase (por) and acetate kinase (ack), were upregulated in AHL-treated samples at 24 and 48 h from exposure compared to the control (Fig. 4, in green). In QSI-treated samples, these genes were instead downregulated at 24 h from exposure, while at 48 h, QSI treatment affected only a smaller number of genes (Table S2.5). Also, pfl encoding pyruvate formate-lyase correlates with QS treatment.

Genes involved in the sulfate reduction pathway (Fig. 4, in blue), including sulfate adenylyltransferase (sat), adenylsulfate reductase (apsAB), and dissimilatory sulfite reductase (dsrABC), were affected by the inhibition or the stimulation of the QS system. Among them, three genes related to the dissimilatory sulfite reductase were grouped in the same operon and most likely transcribed in the same pathway. Furthermore, a similar trend was observed for genes encoding sulfur carrier and sulfotransferase (Table S2.6). This is with the exception of the ppa gene, which encodes proteins involved in the hydrolysis of pyrophosphate.

Electron transfer.

Electron carriers are required for sulfate reduction and to establish proton motive force. Figure 4 (in red) and Table S2.7 showed that genes encoding ferroxidin (Fd), cytochromes (Cyt), hydrogenases (Hase), and formate dehydrogenase (Fdh) were strongly correlated with the addition of either AHL or QSI molecules. Some of the cytochromes were transcribed and regulated together in the same operons. Genes associated with membrane complexes such as DsrMKJOP, QmoABC, Coo, and Rnf were also affected by the two different treatments (Table S2.7). In particular, genes related to QmoABC were found in the same operon, together with the ones related to ApsAB (Table S2.6 and S2.7), even if they participate in two different bacterial functions, namely, electron transfer and sulfate reduction. However, not all electron transfer genes are affected by QS. For example, Tmc, Hmc, Qrc, Nuo, and Och, most of which are associated with membrane complexes, were not influenced by any of the treatments compared to the control (Fig. 4, net pattern).

Biofilm formation.

Genes associated with biofilm formation were related to flagellar biosynthesis, pilus assembly, and extracellular polymeric substance (EPS) synthesis and transport. Seven genes associated with pilus assembly (rcpC, tadB, and tadD) and EPS (DVU_0670 and glycotransferases) were highly correlated with the addition of both treatments (Table S2.8). Some of these genes were transcribed together in the same operon (Table S2.8).

CRISPR genes.

Almost all the genes encoding clustered regularly interspaced short palindrome (CRISPR)-associated proteins were upregulated compared to the control after AHL treatment (Fig. 3). However, unlike the other genes in which QSI treatment would lead to a corresponding downregulation of the same genes, QSI inhibition did not have any impact on these CRISPR genes.

Hypothetical proteins.

Seven hypothetical proteins with unknown functions were also affected by both AHL and QSI treatments (Table S2.10). Among them, two genes were strongly correlated with the QS alteration. One of these genes is located in a cluster of other genes with unknown functions, while the second one is surrounded by genes encoding bacteriophage-related proteins (Fig. S5).

Baseline gene expression change in time in control samples.

Given that D. vulgaris naturally produces AHL, a separate analysis was made for the gene expressions in the control without any additional AHL or QSI application. This is done to determine whether differential gene expressions described in earlier subsections when comparing AHL application versus control at each individual time point were also observed when AHL is produced at baseline levels. Over time, some of the genes in Table S2 were differentially expressed in the control samples. In general, a downregulation trend was observed when comparing the gene expression of the control samples at 24 and 48 h with the one at 6 h (Table S3). All of these genes were even more downregulated compared to the same control samples after QSI addition. However, genes involved in biofilm formation and in the regulatory functions were mostly upregulated at time 24 and 48 h, coinciding with an increase in cell density and likely AHL amount produced (Table S3). In addition, by 48 h, some genes related to membrane transport and electron transfer were also upregulated, although the majority of the genes in these categories remained downregulated compared to 6 h. However, genes that were upregulated in control samples were expressed at an even higher level when AHL was further applied. For example, the two LuxR family genes that were found to be upregulated in the 48-h control samples were even more upregulated when we compared AHL samples to the relative control at the same time point (Table S2.1). This confirms the ability of our AHL treament to increase the detection sensitivity of gene expressions, while not deviating too much from the actual response of the bacterial culture.

Confirmation analysis of LuxR family motif sequences.

The expression of genes is regulated by the presence of transcription factors that control the transcription of the gene. To confirm that the different genetic expression was effectively influenced by QS system alteration, we searched for LuxR family boxes by sequencing from 400 bp upstream to 50 bp downstream of each gene in Table S2. A total of 238 hits with a P value of <0.05 were found (Table S4). The number of hits was greater than the 162 genes detailed in Table S2 because some of these genes were characterized by more than one motif sequence. Particularly, for 116 genes over 162 genes influenced by the QS system, we found at least one transcription factor sequence related to the LuxR or the las-rhl family. This indicates that most of the genes that were affected by the addition of the treatments were characterized by transcription factor related to QS. In addition, by performing the Benjamini-Hochberg procedure to correct the initial motif prediction significance P values for the multiple hypothesis tests, a more stringent q value was obtained. Based on this, we found that 19% of the hits had a q value of <0.05 (Table S4 in boldface) and were related to genes of various bacterial functions connected to the LuxR family transcription regulator vqsR.

DISCUSSION

Understanding the QS functionality in SRB represents a crucial step to develop a QS inhibition approach and to effectively reduce biofilm formation and biocorrosion. Earlier studies revealed the presence of proteins that are homologous to the transcriptional regulator LuxR in D. vulgaris Hildenborough (17, 18). In the present study, two genes encoding proteins of the LuxR family were upregulated naturally in the control temporal baseline (Table S2.1). D. vulgaris, in fact, produces AHL molecules itself, with increasing cell density (20). Moreover, these two genes were strongly correlated with the QS system alteration (Table S2.1), showing an even higher magnitude of up- or downregulation in AHL and QSI samples, respectively. In addition, upregulation or downregulation of other transcriptional regulator and histidine kinase genes followed QS stimulation and inhibition, respectively (Fig. 3). Proteins encoded by these genes are fundamental because they are part of the cellular machinery that allows the QS system to regulate different bacterial functions (13). Interestingly, two of these genes encode QS family proteins (YebC and LysR) that were found to control biofilm formation and AHL production in P. aeruginosa (22, 23). These transcriptional regulators can possibly be part of the QS system in D. vulgaris and regulate functions similar to those in P. aeruginosa. More detailed studies are required to determine the individual role of the proteins encoded by these genes in D. vulgaris.

These observations suggest the ability of the applied treatments and analysis conducted in this study to elucidate pathways, including those already known, that are involved in QS signal response. For instance, it was observed that energy production and protein transcription in D. vulgaris correlated with AHL and QSI treatments. Applying AHL or QS inhibitors did not impact the total cell number (Fig. S1), confirming that QS does not affect bacterial growth, but it affected cell activity, the capacity to generate energy and to produce proteins and enzymes. In fact, the amount of ATP per cell produced after AHL treatment was higher compared to the other conditions (Fig. 1A). This greater cell activity was confirmed by the transcriptomic analysis. Upregulation and downregulation of genes related to ATP synthase (Fig. 4, in yellow) and protein transcription (ribosomal subunit genes) correlated with AHL and QSI treatment, respectively (Fig. 3).

Besides affecting cellular activity and the production of proteins and enzymes, QS alteration appeared to also correlate with the expression of genes related to transport of materials through the membrane. For example, ABC transporters are generally involved in the transport of nutrients and amino acids into the cell. Many genes encoding these proteins showed upregulation or downregulation compared to the control depending on the type of QS perturbation (Fig. 3). In addition, the same was observed for the expression of genes encoding membrane efflux pumps (Table S2.4), which facilitate the transfer of protons across the periplasmic membrane and are required to sustain a higher ATP synthesis. Furthermore, some of the genes in this class encoded transporter protein involved in the biosynthesis of cytochromes. Upregulation of membrane transport-related genes was also observed when a QS signal generation-deficient mutant of P. aeruginosa was provided with AHL to stimulate the QS system (24).

The higher activities associated with transporters and efflux pumps could have indirectly resulted in the higher cell activity and protein synthesis discussed earlier for D. vulgaris. However, the expression of genes related to other membrane proteins involved in the electron transfer to the sulfate reduction pathway (Tmc, Hmc, and Qrc) was not affected by QS (Fig. 4, in net pattern). Unlike DsrMKJOP and QmoABC, which are considered essential for sulfate reduction and showed differential gene expressions upon AHL and QSI applications, Tmc, Hmc, and Qrc are not always present in all SRB, and they play a role only in the presence of certain organic substrates (25). In contrast, along with heterodisulfide (hdr) reductase, dsrMKJOP, and qmoABC, other genes encoding periplasmic hydrogenases, cytochromes, and ferredoxins also correlated with QS. These results are in agreement with an earlier study on P. aeruginosa, in which the stimulation of the QS system led to the upregulation of genes related mainly to cytochromes (11).

As expected, considering the regulation of electron transfer genes by QS modulation, sulfate reduction was also observed to likewise be affected by both AHL and QSI treatments (Fig. 1B and C). For instance, several genes involved in sulfate reduction and lactate metabolism were differentially expressed in the presence of AHL and QSI treatments compared to the relative control (Fig. 3). D. vulgaris oxidizes lactate to pyruvate and then, with the addition of acetyl coenzyme A, to acetate. Pyruvate can also be converted to formate through the pyruvate-formate lyase (pfl). Formate is then oxidized to CO₂ by formate dehydrogenase (fdh), an enzyme involved in the periplasmic electron movement. Electrons from lactate are available for sulfate reduction. When sulfate is internalized, it is converted to adenylylsulfate (APS) by the sulfate adenylyltransferase (sat). APS is reduced to sulfite by the APS reductase (aprAB) and then converted to hydrogen sulfide by the dissimilatory sulfite reductase (dsrABC) (26). Particularly, apsB and qmoABC were found in the same operon, indicating that they were transcribed and regulated together, even if they are associated with different bacterial functions. This suggests the interconnection of electron transfer and sulfate reduction pathways and the synergistic QS effect on both of them. All of the cited essential genes involved in lactate oxidation and sulfate reduction pathways showed a correlation with QS modulation in our study (Fig. 4, in green and in blue, respectively, and Table S2.5 and S2.6). A similar impact on metabolism-related genes was also described for Burkholderia glumae, in which phosphate metabolism and the biosynthesis of various amino acids were controlled by QS (27). Furthermore, it was found that AHL addition led to the upregulation of a gene encoding a sulfite reductase in P. aeruginosa (11). P. aeruginosa is a thiosulfate reducer possessing part of the genetic machinery that is found in D. vulgaris. A previous study also showed that D. vulgaris exposed to each individual compound of the QSI cocktail used in our study showed a decline in sulfate reduction (20), suggesting a relationship between sulfate reduction and QS. This finding, together with our results, supports the idea that sulfate reduction and subsequently biocorrosion might be indeed linked to QS. Although QSI treatment was correlated with a general downregulation of the above-cited bacterial functions after 6 h of treatment, the AHL treatment effect was not as apparent after 6 h of treatment (Fig. 3). This could be due to a longer bacterial acclimatization to the AHL treatment or to a slower molecular assimilation of AHL compared to QSI.

In this study, the amount of biofilm on the carbon steel surface formed under the different conditions (Fig. 2A) correlated with the expression of genes related to EPS production and flagellum and pilus biosynthesis (Tables S2.7 and S2.8). The expression of these genes was affected in an opposite way depending on whether AHL or QSI was applied (Fig. 3). The same genes were also upregulated over the temporal baseline of the control samples, suggesting that an actual gene response was captured in our analysis despite artificially inflating the amount of AHL concentrations for increasing detection sensitivity. These genes were previously found to be crucial for biofilm formation in D. vulgaris (9, 10). By impacting biofilm formation, QS could indirectly affect biocorrosion (Fig. 2B to D). Surface contact facilitated by the biofilm structure could enhance the EMIC process both directly, through nanowires (28) and cytochromes (8), or indirectly via electron carriers (29). Furthermore, the higher sulfate consumption and the upregulation of genes related to sulfate reduction in AHL-treated samples might indicate higher production of H₂S in the environment. This gas is highly corrosive, and it reacts with the metal surface, stimulating the corrosion process.

This study also unveiled a less described role of QS in bacterial protection. The CRISPR adaptive immune system provides resistance to bacteriophages (30). Recent studies have shown that QS can modulate the activity of the CRISPR system by activating the expression of cas genes (27). It was proposed that maintaining the CRISPR system under QS control allows an efficient response when the risk of phage infection is high (31). Likewise, we found that AHL addition was correlated with an upregulation of different CRISPR genes (Fig. 3 and Table S2.9). In addition, from the analysis of the position in the D. vulgaris genome of QS-controlled genes with unknown function, it was observed that the hypothetical protein 205 (Hp205) is placed next to genes related to phage eliminase and phage tail (Fig. S5). However, CRISPR genes were not downregulated when QSI inhibitors were applied, suggesting a different, non-QS-based pathway of silencing these genes. Further studies are required to clarify the effect of QS on these functions in D. vulgaris.

In summary, QS alteration is correlated with the modulation of different key functions, including ATP production, protein transcription, membrane transport, lactate oxidation, sulfate reduction, biofilm formation, and biocorrosion in D. vulgaris. QS-associated effects were observed at both phenotypic and molecular levels. At the molecular level, a lot of the genes that are differentially expressed were characterized by the transcription regulator vqsR. However, further and more in-depth studies are required to determine the exact mechanism or causative effect of AHL and QSI on the QS system of D. vulgaris. Nevertheless, our findings provide important indications about the role of QS in SRB, and open up the possibility to exploit a QS inhibition approach to reduce biofilm formation and biocorrosion related to SRB.

MATERIALS AND METHODS

Bacterial cultivation conditions.

Desulfovibrio vulgaris Hildenborough (21) was propagated in marine Desulfovibrio Postgate medium (DSMZ) with concentrations of 50 mM lactate and 25 mM sulfate. The medium had the following composition: 25.0 g/liter NaCl, 0.5 g/liter K₂HPO₄, 1.0 g/liter NH₄Cl, 2.4 g/liter Na₂SO₄, 0.1 g/liter CaCl₂⋅2H₂O, 0.97 g/liter MgSO₄, 0.005 g/liter FeSO₄⋅7H₂O, 1.0 g/liter yeast extract, and 0.1% (wt/vol) sodium resazurin as an O₂ indicator. After heat sterilization, the medium was supplemented with filter-sterilized (0.22-μm pore size) sodium lactate (50 mM) and a solution composed of 0.1 g/liter sodium thioglycolate and 0.1 g/liter ascorbic acid. The pH of the medium was adjusted to 7.5 and aseptically transferred to sterile tubes or vials sealed with butyl rubber stoppers. The medium was purged with N₂ for 15 min and placed in the anaerobic chamber (Coy Laboratory Products, Inc., Grass Lake, MI), where l-cysteine (100 mg/liter) was added as an oxygen scavenger. For all of the experiments, a 5% (vol/vol) bacterial inoculum was maintained, and the cultures were incubated at 30°C without shaking.

Anaerobic reactor setup.

Four serum bottles were equipped with five carbon steel coupons (1030; ChemWorld, Taylor, MI), each with a surface area of 1.4 cm². Coupons were wet polished with sandpaper up to 600-grit finish, cleaned with ethanol, and dried with nitrogen. After UV sterilization, the coupons were immersed in 800 ml of the marine DSMZ. The medium in each serum bottle was purged with N₂ for 1 h before bacterial inoculation in the anaerobic chamber. One serum bottle was not inoculated and served as an abiotic control. All the serum bottles were incubated at 30°C without shaking for 7 days. After 72 h of incubation, one serum bottle was injected with a cocktail of N-octanoyl-homoserine lactone (C₈-HSL), N-decanoyl-homoserine lactone (C₁₀-HSL), and N-dodecanoyl-homoserine lactone (C₁₂-HSL) (Cayman Chemical, Ann Arbor, MI), each at a concentration of 5 μM (AHL treatment). The concentrations of the three AHLs used in our study were not physiologically relevant since they are higher than the baseline concentration produced by D. vulgaris. However, it was necessary to excite the bacterial QS system and to improve detection sensitivity when performing subsequent transcriptome sequencing (RNA-seq) analysis, as shown in an earlier study (11). At the same time, one serum bottle was injected with a cocktail of QSIs used in a previous study (20). The cocktail was composed of (Z-)-4-bromo-5-(bromomethylene)-3-butyl-2(5H)-furanone at 100 μM, 3-oxo-C₁₂-(2-aminocyclohexanone) at 100 μM, and γ-aminobutyric acid at 5 mM (QSI treatment). The treatments were spiked after 72 h from incubation based on the indication of our earlier study, which showed that the alteration of the QS system has a more pronounced effect in the early exponential phase (20). The remaining inoculated serum bottle was not injected, serving as biotic control. The experiment was repeated to obtain a total of three biological replicates. In addition, the possible impact of the two treatments on the coupon surface was assessed. An abiotic experiment with the addition of QSI and AHL cocktail was performed in triplicates.

Planktonic cell analysis.

Bacterial culture aliquot (5 ml) was sampled every 12 h from the serum bottles in an aseptic manner. The sampling was conducted in the anaerobic chamber to avoid any oxygen contamination.

To assess cell activity during the incubation period, the level of ATP per cell was calculated. For cell enumeration, 100-μl portions of bacterial suspension were diluted 10³ times, stained with SYBR green (Thermo Fisher Scientific, Waltham, MA), and counted by BD Accuri C6 flow cytometer (BD Bioscience, Franklin Lakes, NJ). The ATP concentration was quantified using a Celsis amplified ATP reagent kit and an Advance luminometer (Celsis, Westminster, United Kingdom). Theoretically, 2 mol of lactate consumed and 1 mol of sulfate reduced led to the production of 2 mol of acetate (32), as shown in the following reaction:

$$2{\text{C}}_3{\text{H}}_5{\text{O}}_3{}^{\text{–}}{\text{+ SO}}_4{}^{\text{2–}}\text{=}2{\text{C}}_2{\text{H}}_3{\text{O}}_2{}^{\text{–}}{\text{+ 2HCO}}_3{}^{\text{–}}{\text{+ HS}}^{\text{–}}{\text{+ H}}^{+}$$

Lactate and acetate concentrations were quantified by using a high-performance liquid chromatograph equipped with an HPX-87 H ion exchange column (300 by 7.8 mm; Bio-Rad, CA) and a UV detector. Sulfuric acid (5 mM) was used as a mobile phase at a flow rate of 0.6 ml/min. The sulfate concentration in the planktonic phase was quantified using an ICS-1600 ion chromatograph (Dionex Corporation, Sunnyvale, CA) with KOH as an eluent. Data from the chromatography analysis were processed using Chromeleon 7.0 software. The lactate, acetate, and sulfate consumption rates were calculated before and after the treatments.

Coupon biofilm harvesting and surface analysis.

After 7 days of incubation, the coupons from the serum bottles were harvested aseptically, and then biofilm and surface characterization was performed to assess the effect of the different treatments. All of the coupons from each serum bottle were washed, placed separately in 1 ml of 1× phosphate-buffered saline (PBS), and sonicated for 5 min to detach the biofilm from the metal surface. The cell suspension was assessed for the number of attached cells and the ATP concentration using the protocols described above. The other three coupons were removed from the PBS and placed in Clark solution for 30 s to remove any recalcitrant biofilm and corrosion products (33).

One coupon was observed under an FEI Teneo SEM (Thermo Fisher Scientific, Hillsboro, OR) to examine the degree of surface corrosion and the presence of pits caused by the formed biofilm. The second coupon was analyzed through AFM to determine the surface roughness. For each coupon, two areas of 50 by 50 μm each were analyzed using Bruker Dimension ICON equipment (Santa Barbara, CA) in soft tap mode at a constant spring of 40 N/m and a resonant frequency of 300 kHz. The AFM images were analyzed on Pico image software (Keysight Technologies, Inc., Santa Rosa, CA). The last coupon was used to determine the maximum depth of the pits using a Dektak profilometer (Bruker, Billerica, MA). For each coupon, two sections 500 μm in length were analyzed. In addition, SEM, AFM, and surface profile analyses were performed to assess the abiotic effects of both AHL and QSI molecules alone on the coupon surface.

Electrochemical and corrosion analysis.

Quantitative evaluation of biocorrosion was obtained through electrochemical measurements and weight loss analysis. Four Communicable Disease Centre bioreactors (BioSurface Technologies, Bozeman, MT) equipped with five carbon steel coupons (1030; BioSurface Technologies, Bozeman, MT) were operated at 30°C with continuous stirring in batch mode for 7 days with continuous N₂ purging. The reactors were filled with 600 ml of marine DSMZ, and three of them were inoculated (5%) with D. vulgaris in the exponential phase. One reactor was not inoculated (i.e., abiotic control). After 72 h, one reactor was injected with the QSI cocktail and one with the AHL cocktail. The abiotic control was inoculated with both cocktails, while the remaining inoculated reactor was used as control. Coupons in the reactors (surface area, 1.3 cm²) were previously electrocoated with a protective layer of Powercron 6000CX (PPG Industrial Coatings), and one surface (exposed surface) was polished up to 600 grit finish. Three coupons for each reactor were weighed before to initiate the experiment for the weight loss analysis.

Two coupons were soldered with a copper wire, and they served as working electrode for the electrochemical measurements. A platinum-coated mesh was used as counterelectrode, and a double-junction Ag/AgCl electrode (3.5 M) was used as reference electrode. The reference electrode was held in a Luggin capillary filled with agar (1.5% [wt/vol]) containing 3% KCl (wt/vol). All electrochemical measurements were conducted using a Gamry-600 potentiostat connected to an electrochemical multiplexer ECM8 (Gamry Instruments, Warminster, PA). The linear polarization resistance from −0.1 to 0.1 V, with a scan rate of 1 mV/s, was measured daily to obtain the corrosion rate using the Gamry software. At the end of the incubation exposure, three coupons were weighed to measure the weight loss after biofilm and corrosion product removal using Clark solution. The total corrosion rate related to all the incubation times expressed as millimeters per year was calculated according to standard methods (33). The experiment was repeated twice more for a total of three replicates.

RNA extraction and sequencing.

Samples from the inoculated serum bottles (control, AHL, and QSI treatments) were taken after 6, 24, and 48 h from the treatment application. Biomass preservation and RNA extraction were performed as described previously (34). Briefly, 15 ml of suspended culture was centrifuged at 6,500 × g for 30 min, and the pellet was resuspended in 2 ml of 1× PBS. Thereafter, 4 ml of RNAprotect cell reagent (Qiagen, Hilden, Germany) was added to avoid RNA degradation, and the solution was incubated at room temperature for 5 min. After incubation, the mixture was centrifuged at 6,500 × g, and the pellet was stored at –80°C until RNA extraction. RNA was extracted using an RNeasy Midikit (Qiagen), including a DNase treatment, and the RNA concentration was measured using an Invitrogen RNA HS Qubit 2.0 assay kit (Thermo Fisher Scientific).

The RNA quality was assessed with a 2200 Tapestation bioanalyzer (Agilent Technologies, Santa Clara, CA). Afterward, the samples were enriched in mRNA by rRNA removal using a Ribo Zero rRNA removal kit (Illumina, San Diego, CA). Finally, RNA-seq libraries were prepared and submitted to the KAUST genomic Core Lab for RNA-seq on an Illumina HiSeq 4000 platform.

RNA transcriptomic analysis.

D. vulgaris DNA was extracted and the whole genome was sequenced using the PacBio platform. The genome sequence was assembled as described elsewhere (35–37). The structural annotation of the final genome assembly was performed with Rast (38). Genome assembly revealed the presence of a circular chromosome of 3,526,512 bp and a plasmid of 201,796 bp, accounting for a total of 3,538 genes. The in-house assembled genome shared the most similarity with the reference genome of Desulfovibrio vulgaris Hildenborough (21), indicating no contamination of the SRB culture.

Transcriptomic analysis was performed with CLC Genomic Workbench 8.0 (CLC Bio, Cambridge, MA), as described elsewhere (39). The in-house assembled and annotated D. vulgaris genome described above was used as reference genome. RNA-seq reads were first mapped to the whole bacterial genome to assess the quality of the samples and the absence of contamination. The rest of the analysis was performed mapping the reads only to the coding sequence of the annotated genome. Reads were mapped to the genome only if the fraction aligned sequence was >0.9 and if the number of nucleotides matching other genome regions was <10. The percentage of mapped reads to both coding sequences only and to the whole genome can be found in Table S5.

After read mapping, biological replicates were assigned to the same category, and a mean expression value was calculated. In addition, a scaling correction was applied to normalize each expression value with the total number of reads. Normalized gene expression value was defined as reads per kilobase per million (RPKM). The different effects of each of the two treatments were compared to the control at the same time point. A Baggerly proportion-based test was used for statistical comparison (40). A fold change of 2 and a P value of <0.05 were selected as threshold parameters in the selection of differently expressed genes. Based on their upregulation (fold change, >2) or downregulation (fold change, <−2) in the two treatments, a scoring system was applied to each gene. A minimum total score of 2 was used to consider a gene to be likely correlated with the QS system. A gene with a score of ≥3 was considered strongly correlated with the QS system. Genes that did not fit these parameters (fold change, P value, and scoring system) were not considered for further analysis. Genes that showed a strong correlation with the QS were analyzed to evaluate whether other genes in their same operon were also affected by QS modulation. The operon structure was found in the OperonDB database. Operon grouping for this database was performed through an algorithm that infers the probability that two adjacent genes belong or not to the same operon (41). In addition, to further evaluate the baseline temporal trend for the control samples, the gene expression of control samples after 24 and 48 h from the moment of the treatments in AHL and QSI samples was compared to the one after 6 h. More details are provided in the supplemental material.

Transcription factor binding site analysis.

The analysis of LuxR family box sequencing was performed to confirm that the different genetic expression was effectively influenced by QS system alteration. Fimo v5.0.5 (42) was used to predict transcription factor binding site motifs in a sequence window reaching from 400 bp upstream to 50 bp downstream of each gene transcriptional start site. The results were filtered, retaining only hits with a P value of <0.05. In addition, to correct the initial motif prediction significance P value for the multiple hypothesis test, we performed the Benjamini-Hochberg procedure. This test provides the minimal false discovery rate (FDR), a more stringent threshold at which the corresponding P value is considered significant. Hits with an FDR q value of <0.05 were further investigated. Fimo was provided with binding site motifs of 13 transcription factors of the LuxR family, obtained from the collectTF database (43). The following transcription factors were included (transcription factor symbol followed by the UniProt identifier in parentheses): hapR (A0A0H3Q915), lasR (P25084), lasR (P54292), luxR (A7MXJ7), luxR (B5EV73), luxR (P35327), opaR (Q79YV4), rsaL (G3XD78), smcR (Q7ME71), traR (P33905), vpsT (Q9KKZ8), vqsR (Q9I0P6), and vsrD (Q8XVU0). In addition, the las-rhl motif defined by Schuster et al. (11) was also included.

Statistical test.

To evaluate statistical differences, one-way analysis of variance (ANOVA) was performed, with a confidence level set at 95% (P < 0.05).

Data availability.

The raw read data are available on the European Nucleotide Accession Short Reads Archive repository under accession number PRJEB33204. The assembled genome is available from the ENA SRA repository under accession number ERS3567812.