Disturbing the Spatial Organization of Biofilm Communities Affects Expression of agr-Regulated Virulence Factors in Staphylococcus aureus

ABSTRACT

Staphylococcus aureus uses quorum sensing and nutrient availability to control the expression of agr-regulated virulence factors. Quorum sensing is mediated by autoinducing peptide (AIP), which at a high concentration reduces expression of surface attachment proteins (coa, fnbpA) and increases expression of exotoxins (lukS) and proteases (splA). Nutrient availability can be sensed through the saeS/saeR system. Low nutrients increase expression of saeR, which augments expression of coa and fnbpA, distinct from the activity of AIP. The formation of spatial structure, such as biofilms, can alter quorum sensing and nutrient acquisition. In natural environments, biofilms encounter forces that may alter their spatial structure. These forces may impact quorum sensing and/or nutrient acquisition and thus affect the expression of agr-regulated virulence factors. However, this has not been studied. We show that periodically disturbing biofilms composed of S. aureus using a physical force affected the expression of agr-regulated virulence factors. In nutrient-poor environments, disturbance increased the expression of coa, fnbpA, lukS, and splA. Disturbance in a nutrient-rich environment at low or high disturbance amplitudes moderately reduced expression of coa and fnbpA but increased expression of lukS and splA. Interestingly, at an intermediate amplitude, the overall expression of agr-regulated virulence factors was the lowest; expression of lukS and splA remained unchanged relative to an undisturbed biofilm, while expression of coa and fnbpA significantly decreased. We hypothesize that these changes are a result of disturbance-driven changes in access to AIP and nutrients. Our results may allow the identification of environments where virulence is enhanced, or reduced, owing to a disturbance.

IMPORTANCE Bacteria, such as Staphylococcus aureus, integrate signals from the environment to regulate genes encoding virulence factors. These signals include those produced by quorum-sensing systems and nutrient availability. We show that disturbing the spatial organization of S. aureus populations can lead to changes in the expression of virulence factors, likely by altering the ways in which S. aureus detects these signals. Our work may allow us to identify environments that increase or reduce the expression of virulence factors in S. aureus.

INTRODUCTION

Quorum sensing is an important bacterial behavior that allows bacteria to act as a collective. The process of quorum sensing involves the production and exchange of small diffusible molecules called autoinducers, which allow bacteria to alter gene expression based on the density of the population (1). Since each bacterium produces and secretes autoinducer, the concentration of autoinducer in an environment increases with increasing cell density. Once a sufficiently high concentration of autoinducer, and thus cell density, is reached, gene expression in the bacterial population can be altered. It is thought that quorum sensing evolved to regulate processes that would only be beneficial to the bacterial population when at high density (2). The behaviors regulated by quorum sensing are diverse. On the one hand, these behaviors can be beneficial to organisms other than the bacteria themselves (e.g., bioluminescence [3]). On the other, quorum-sensing-regulated behaviors can have negative outcomes for other organisms. Chief among these is the production of virulence factors that are critical for bacterial pathogenesis (4). Given the ubiquity of quorum sensing and the diversity of behaviors that it regulates, there is a need to identify the conditions under which quorum-sensing behaviors are enhanced or reduced (5).

One bacterial species that requires quorum sensing for pathogenesis is Staphylococcus aureus (6, 7), which uses the agr (accessory gene regulator) system for the regulation of many well-described virulence factors (6). The agr operon consists of four genes (agrA, agrB, agrC, and agrD) that work together to produce and sense a diffusible autoinducing peptide (AIP) (Fig. 1A). Production and accumulation of AIP drives the additional expression of all four agr genes and alters the expression of downstream virulence factors through rnaIII. At low bacterial density, and thus low concentrations of AIP, the expression of rnaIII is inhibited and/or expressed at low levels, allowing for the expression of genes encoding surface attachment proteins, including protein A (spa), fibronectin-binding protein A (fnbpA), and coagulase (coa), which promote surface colonization (6). At high bacterial density, and thus high concentrations of AIP, expression of rnaIII increases. This results in the repression of genes encoding surface attachment proteins and increases expression of genes that encode exotoxins, including α-hemolysin (hla) (8) and leukocidin S (lukS) (9, 10), and proteases, including the serine protease-like protein A (splA) (11). These exotoxins and proteases allow bacteria to lyse host cells, which facilitates access to additional nutrients (6).

FIG 1.

An approach to disturbing the spatial structure of biofilms composed of S. aureus. (A) A minimal representation of quorum sensing and metabolism/nutrient-sensing networks in S. aureus that regulate expression of genes that code for virulence factors through the agr operon. Red arrow, repression; green arrow, activation. AIP (bottom) is produced by the agr operon and is shared with all members of the bacterial population. (B) The Innovotech MBEC biofilm inoculator was used to grow biofilms. Bacteria in the biofilm state adhere to the peg and the sides of the well, while bacteria in the planktonic state grow in the medium surrounding the biofilms. Disturbances to the spatial structure of the population were accomplished by using the linear shaking function of a microplate reader. (C) Schematic of our experimental approach to periodically disturb spatial organization and measure its effect on the expression of select virulence factors. (D) Density of bacteria in the biofilm and planktonic states after 24 h of growth. Biofilm density was measured using a crystal violet assay and OD₅₅₅. Density of planktonic bacteria was measured using OD₆₀₀. Both OD values were converted to CFU per milliliter by using standard curves (see Fig. S1 in the supplemental material). Standard deviations are from 12 biological replicates (P < 0.001, two-tailed t test). (E) The density of bacteria in the planktonic state after biofilms were shaken once in the microplate reader at the amplitude indicated. There was a significant increase in the density of bacteria in the planktonic state at an amplitude of 0.3 mm relative to that of the undisturbed control (P = 0.0495, Wilcoxon [Shapiro-Wilk, P = 0.0150]). Standard deviations from 4 biological replicates are shown.

The expression of agr-regulated virulence factors can also be regulated through changes in metabolism. One global transcriptional regulator that is conserved in multiple bacterial species and that integrates changes in metabolism is CodY (12). Access to a large amount of nutrients, which coincides with high metabolism, allows the activation of CodY (13). As nutrients become limited and metabolism is reduced, CodY becomes deactivated (12, 14). In S. aureus, one system that CodY regulates is the saeS/saeR two-component system (15). The saePQRS operon contains two promoters, P1 and P3. P3 is a weak but constitutive promoter that produces basal amounts of SaeR and SaeS. P1 is a stronger promoter whose activity is regulated by CodY. When nutrients are abundant, CodY represses expression from the P1 promoter. However, as nutrients are depleted, repression from CodY is alleviated. This increases the expression of elements in the saePQRS operon, including saeR, which itself can bind to the P1 promoter, driving further expression from the operon (15). These molecular interactions are congruent with increased expression of the saePQRS operon after the exponential phase of growth when nutrients become limiting (16). The saeS/saeR system regulates the expression of agr targets distinct from AIP (Fig. 1A). Phosphorylation of large amounts of SaeR increases the expression of agr-regulated virulence factors, including coa and fnbpA (15, 17). Thus, expression of virulence factors from the agr operon can be increased through saeR when nutrients are limiting and metabolism is low.

Density-dependent behaviors, including quorum sensing and nutrient acquisition, are relevant in highly dense environments such as biofilms (18). Biofilms are complex, high-density structures that encase bacteria in a matrix of self-produced extracellular polymeric substances (19). The formation of biofilms allows communities of bacteria to become affixed to surfaces and can lead to novel behaviors, including enhanced antibiotic resistance (20), horizontal gene transfer (21), and immune system avoidance (22). The specific architecture of a biofilm is dynamic and can depend on both bacterial species and the growth environment (23). Biofilms can allow the accumulation of important nutrients or other diffusible molecules, including autoinducers such as AIP. Components of the matrix, such as amyloid fibers, have been shown to bind to quorum-sensing molecules (24); the concentration of such molecules can be approximately 1,000-fold greater in the biofilms than in the surrounding environment (25). Moreover, the formation of pores and channels can allow the movement of liquid through the biofilm, carrying with it nutrients and autoinducers (19). Thus, bacteria in biofilms can sense a higher effective concentration of autoinducers relative to bacteria that lack spatial structure (26). Biofilm structure can also impact the metabolism of bacterial residents. In environments that are nutrient poor (i.e., oligotrophic environments), rapid utilization of oxygen and nutrients by bacteria located toward the surface of the biofilm can limit the availability of these compounds to bacteria located deeper within the biofilm (19). These bacteria can experience lower growth rates (27), reductions in metabolism (28), and cell death (25). Thus, while biofilm structure can increase the availability of autoinducers, it may lead to reductions in metabolism; together, this would impact the expression of agr-regulated virulence factors.

Most studies have used continuously mixed (e.g., culture tubes) or completely stationary (e.g., agar plates) conditions to investigate quorum-sensing-regulated expression of virulence factors. However, in a natural setting, bacteria can experience a wide range of physical forces (29, 30), including changes in fluid flow (31–33), vibrations (34, 35), and additional mechanical stressors (36, 37). These physical forces have been shown to alter spatial structure (38, 39) and often fluctuate or occur periodically. For example, vibrational forces (34, 35) and shear forces (40) in aquatic environments fluctuate. If the force is sufficiently strong, the structural organization of bacterial populations can be altered. This has been previously shown to impact the expression of genes regulated by quorum sensing (27) and the production of virulence factors (41). However, we do not understand how these physical forces affect the expression of virulence factors that are regulated by quorum sensing and metabolism. This is important to address; as physical forces can disrupt spatial organization, they can affect both quorum sensing and nutrient availability, thus impacting the expression of virulence factors. In this study, we use the agr and saeS/saeR systems to ask how does disturbance to spatial structure impact the expression of agr-regulated virulence factors? An understanding of this may allow us to gain insight into how such systems regulate virulence in a more natural setting and may establish novel approaches to attenuate the expression of virulence factors.

RESULTS

An experimental setup to perturb spatial structures of biofilms.

We first established a method to grow and perturb biofilms composed of S. aureus. We grew biofilms using the MBEC inoculator device, which is a growth platform consisting of a 96-well plate with a lid containing 96 polystyrene pegs. The pegs are submerged in liquid medium containing bacteria. Biofilms form on the pegs and the walls of the surrounding well, while a population of bacteria in the planktonic state exist in the remaining liquid medium (Fig. 1B). After growing biofilms, and to rationally manipulate metabolism, fresh nutrients can be provided to the biofilms by moving the plate lid to a new 96-well plate containing fresh growth medium (Fig. 1C). This would increase metabolism by providing nutrients (a condition henceforth referred to as nutrient rich). It would also remove accumulated AIP and the majority of bacteria in the planktonic state, the latter of which would reduce the total population density below carrying capacity (i.e., the maximum population size that can be sustained by the growth medium). Moreover, the addition of nutrients would allow growth of bacteria in both states. Thus, in the nutrient-rich condition, disturbance occurs in fresh medium. Alternatively, if nutrients are not replenished prior to disturbance, metabolism could consequently be reduced (a condition henceforth referred to as nutrient poor) but would allow accumulated AIP, bacteria, and any waste products to remain in the planktonic state (Fig. 1C). Thus, in the nutrient-poor condition, disturbance occurs in the same medium that was used to grow the biofilm. Overall, our use of nutrient-rich and -poor conditions allowed us to rationally perturb metabolism.

To confirm the formation of biofilms on the pegs, we placed 190 μL of growth medium containing a 1:100 dilution of an overnight culture of S. aureus in the well surrounding the peg. After 24 h of growth, we measured biofilm density by staining the biofilms in crystal violet (Fig. 1D). We also determined the density of bacteria in the planktonic state by measuring the optical density at 600 nm (OD₆₀₀). We detected the presence of bacteria in both the biofilm and planktonic states, indicating that 24 h was sufficient time to allow the formation of a biofilm and surrounding planktonic population.

In line with previous work (27, 41), we used the linear shaking function of a microplate reader to disturb the spatial structure of the bacterial community (Fig. 1B). This served to periodically transfer bacteria from the biofilm state to the planktonic state. To confirm that the linear shaking function of a plate reader could perturb the spatial structure of biofilms composed of S. aureus, we measured the number of CFU in the planktonic state both before and after a single shaking event. A single shaking event using a 0.1-mm amplitude was insufficient to increase the density of bacteria in the planktonic state compared to an undisturbed control (Fig. 1E). When the shaking amplitude was increased to 0.3 mm, we observed a significant increase in the density of bacteria in the planktonic state (Fig. 1E; P values are provided in the figure legend). Increasing the shaking amplitude further resulted in an increase in the density of bacteria in the planktonic state. However, this increase was insignificant, owing to large variability in the data.

Frequency of disturbance alters the distribution of bacteria and metabolism.

To examine the effects of periodic disturbance on the spatial structure of S. aureus grown in the MBEC device, we disturbed biofilms using an amplitude of 0.3 mm and at three different shaking frequencies: 3 shaking events per hour (3/h), 9/h, and 15/h. After 24 h of disturbance, we then measured the density of bacteria in the biofilm and planktonic states (Fig. 2A). Under the nutrient-poor condition, we observed that shaking significantly reduced the density of bacteria in the planktonic state compared to the undisturbed control (Fig. 2A, left panel). Conversely, total biofilm density, which we determined by summing the density of bacteria in the biofilm state on the peg and on the walls of the well, was on average greater than that of the undisturbed control. Disturbance at 3/h and 9/h, but not 15/h, significantly increased the density of the biofilm in the well. Significant differences in the density of the biofilm on the peg were only observed at 9/h and 15/h (Fig. 2A, right panel). Under the nutrient-rich condition, the density of bacteria in the planktonic state was significantly reduced at all shaking frequencies measured relative to that in the undisturbed control (Fig. 2B, left panel). Total biofilm density was reduced at 9/h but increased at 3/h and 15/h relative to the undisturbed control. At 3/h and 9/h, but not 15/h, there was a significant decrease in the density of biofilm on the peg (Fig. 2B, right panel). At 3/h and 15/h, but not 9/h, there was a significant increase in the density of the biofilm on the walls of the well. Under both nutrient conditions, total biofilm density was ~10⁷ CFU/mL, which comprised ~6% of the total population, since the density of bacteria in the planktonic state was on average ~3 × 10⁸ CFU/mL. Importantly, under both nutrient-rich and nutrient-poor conditions, the distribution of bacteria after 24 h of disturbance was not identical to that of the undisturbed control. This indicated that periodically shaking the MBEC plate could disrupt the spatial distribution of bacteria within the biofilm.

FIG 2.

Periodic disturbance of biofilms alters the spatial distribution of bacteria. (A) Density of bacteria in the biofilm and planktonic states following 24 h of disturbance at the frequency indicated on the x axis (0.3 mm amplitude) in the nutrient-poor condition. (Left) Density of bacteria in the planktonic and biofilm states (summed between biofilms on the peg and on the walls of the well). (Right) Density of bacteria in the biofilm state on either the peg or the walls of the well. *, significant difference from the undisturbed control (P < 0.005, Wilcoxon [P < 0.0001, Shapiro-Wilk]). Standard deviations from a minimum of seven biological replicates are shown. (B) Density of bacteria in the biofilm and planktonic states following 24 h of disturbance at the frequency indicated on the x axis (0.3 mm amplitude) in the nutrient-rich condition. (Left) Density of bacteria in the planktonic and biofilm states (summed between biofilms on the peg and on the walls of the well). (Right) Density of bacteria in the biofilm state on either the peg or the walls of the well. *, significant difference from the undisturbed control (P < 0.036, Wilcoxon [P < 0.0001, Shapiro-Wilk]). Standard deviations from a minimum of six biological replicates were determined.

To determine how the nutrient-poor and nutrient-rich conditions altered metabolism, we quantified the expression of saeR using reverse transcription-quantitative PCR (RT-qPCR). Previous work had shown that saeR expression increases upon nutrient limitation concomitant with a reduction in metabolism (15, 17). We confirmed that increasing nutrients decreased expression of saeR (see Fig. S3A in the supplemental material). Similarly, we found that decreasing nutrients increased expression of saeR (Fig. S3B). To quantify gene expression using RT-qPCR, we extracted RNA from bacteria in the planktonic state, which as noted above comprised ~94% of the total population. When biofilms were disturbed in nutrient-poor conditions, we observed a significant increase in the expression of saeR relative to the undisturbed control (0/h) at shaking frequencies of 9/h and 15/h, but not 3/h (Fig. 3A, left panel). Conversely, when biofilms were disturbed in nutrient-rich conditions, there was a significant decrease in the expression of saeR at all shaking frequencies measured (Fig. 3A, right panel). To provide additional support to the reduction in metabolism, we also we measured the expression of ilvD, a member of the branched-chain amino acids synthesis pathways whose expression has been previously shown to be inhibited by CodY when the concentrations of amino acids and GTP are high (42, 43). We observed significantly higher expression of ilvD in bacteria in the nutrient-poor condition (0/h) compared to the nutrient-rich condition (0/h) (Fig. 3B). These findings were also consistent when we measured the concentration of ATP, which can be used as a surrogate measure of metabolism (44). We found a significantly higher concentration of ATP in the nutrient-rich condition than in the nutrient-poor condition (Fig. 3C). Finally, we found that bacteria would not grow in the nutrient-poor condition, but could readily grow in the nutrient-rich condition (Fig. S4).

FIG 3.

Periodic disturbance of biofilms in nutrient-poor and nutrient-rich conditions alters metabolism, and expression of agrC and rnaIII. (A) Effect of increasing shaking frequency of the expression of saeR in the nutrient-poor (left) and nutrient-rich (right) conditions. *, significantly different from undisturbed control (P ≤ 0.0277, two-tailed t test). Standard deviation from a minimum of three biological replicates were determined. All ΔC_T values here are shown in increased detail in Fig. S5 in the supplemental material. For panels A, D, and E, fold change was determined relative to corresponding 0/h condition. (B) Expression of ilvD in nutrient-poor and nutrient-rich conditions. Transcript abundance was measured in the undisturbed condition (0/h). When transcript abundances were compared to each other, an asterisk indicates P  = 0.0005 (two-tailed t test). Standard deviations from four biological replicates were determined. (C) ATP concentration in bacteria that were grown in nutrient-rich or nutrient-poor conditions. The ATP concentration was determined using the standard curve in Fig. S2. For ATP concentration (measured under both conditions) comparisons, an asterisk indicates P = 0.003 (two-tailed t test). Standard deviations from eight biological replicates were calculated. (D) Effect of increasing shaking frequency on the expression of agrC in the nutrient-poor (left) and nutrient-rich (right) conditions. *, significantly different from undisturbed control (P ≤ 0.0016, two-tailed t test). Standard deviations from a minimum of four biological replicates were determined. (E) Effect of increasing shaking frequency on the expression of rnaIII in the nutrient-poor (left) and nutrient-rich (right) conditions. *, significantly different from undisturbed control (P ≤ 0.0007, two-tailed t test). Standard deviations from a minimum of three biological replicates are shown.

Next, we measured how disturbance affected agr activity by quantifying the expression of agrC and rnaIII. Under nutrient-poor conditions, the expression of agrC significantly increased relative to that at 0/h at shaking frequencies of 9/h and 15/h (Fig. 3D, left panel). Under nutrient-rich conditions, expression of agrC significantly increased at all shaking frequencies measured (Fig. 3D, right panel). Expression of rnaIII was significantly increased at all shaking frequencies in the nutrient-poor condition (Fig. 3E, left panel). However, expression of rnaIII was not different than that at 0/h in nutrient-rich condition (Fig. 3E, right panel). This was likely because of the large amount of rnaIII expression in the 0/h condition (Fig. S5). agr activity was similar between bacteria in the biofilm and planktonic states when measured in the undisturbed condition (P = 0.061) (Fig. S4). Thus, measuring gene expression in bacteria isolated from the planktonic state serves as a surrogate of cells released from the biofilm. Overall, our results showed that disturbing biofilms under either nutrient-poor or nutrient-rich conditions can alter the expression of genes involved in metabolism and in agr activity.

Disturbance under nutrient-poor but not nutrient-rich conditions increases expression of agr-regulated virulence factors.

Using RT-qPCR, we measured the effects of periodic disturbance on the expression of select agr-regulated virulence factors. To accomplish this, we grew biofilms of S. aureus for 24 h, followed by an additional 24 h of disturbance (Fig. 1C). In the nutrient-poor condition, we observed that the expression of coa, fnbpA, splA, and lukS significantly increased at 9/h and 15/h, but not 3/h, relative to 0/h (Fig. 4A and B). This trend matched the previously described increase in the expression of saeR and agrC at 9/h and 15/h but not 3/h (Fig. 3A and D). Changes in the expression of these genes were unlikely a result of the accumulation of waste products in the nutrient-poor condition. Specifically, we found that disturbing bacteria in fresh tryptic soy broth (TSB) medium that had been diluted 10-fold resulted in the same qualitative trends in gene expression as when bacteria were disturbed in the nutrient-poor condition (Fig. S7).

FIG 4.

Periodic disturbance can alter the expression of agr-regulated virulence factors. (A) Effects of increasing shaking frequency on the expression of genes encoding select surface attachment genes (coa and fnbpA) in the nutrient poor condition. *, significantly different from undisturbed control (P ≤ 0.0052, two-tailed t test). Standard deviations are from a minimum of four biological replicates. ΔC_T values can be found in Fig. S6. In all panels, fold change was determined relative to corresponding 0/h condition Panels A and B show data from nutrient poor condition. (B) Effects of increasing shaking frequency on the expression of splA and lukS in the nutrient poor condition. *, significantly different from undisturbed control (P ≤ 0.0212, two-tailed t test). Standard deviations are from a minimum of four biological replicates. (C) Effects of increasing shaking frequency on the expression of coa and fnbpA in the nutrient rich condition. *, significantly different from undisturbed control (P ≤ 0.0443, two-tailed t test). Standard deviations are from a minimum of four biological replicates. (D) Effects of increasing shaking frequency on the expression of splA and lukS in the nutrient rich condition. Standard deviations are from a minimum of four biological replicates. None of the transcript abundances were significantly different from the 0/h control (P ≥ 0.1009, two-tailed t test).

In the nutrient-rich condition and for all shaking frequencies measured, we observed a significant decrease in the expression of coa and fnbpA compared to results with 0/h (Fig. 4C). Conversely, there was no significant difference in the expression of lukS and splA relative to that at 0/h (Fig. 4D). Interestingly, while disturbance into the nutrient-rich condition increased the expression of agrC (Fig. 3D), it did not increase expression of rnaIII (Fig. 3E). The lack of change in the amount of rnaIII, splA, or lukS transcripts was consistent with the regulation of the agr network, as increasing rnaIII increased expression of splA and lukS (Fig. 1A). Overall, disturbance into a nutrient-poor environment tended to increase the expression of agr-regulated virulence factors. Conversely, disturbance into the nutrient-rich condition decreased the expression of surface attachment proteins coa and fnbpA while leaving the expression of splA and lukS unchanged relative to levels in an undisturbed control.

An intermediate amplitude of disturbance leads to the greatest reduction in virulence factor expression.

Previous work showed that the amplitude of a physical stressor can impact the mechanical properties of biofilms and alter the growth of biofilm residents (45, 46). However, an understanding of how changes in the amplitude of a physical stressor and their impact on the expression of agr regulated virulence factors has yet to be demonstrated. To address this gap in knowledge, we disturbed bacteria grown in the MBEC system for 24 h at increasing amplitude and at a shaking frequency of 15/h in the nutrient-rich condition. We used a frequency of 15/h because, when coupled with increasing amplitude, it would provide a wide range of shear forces over which we could study the effects of disturbance on the expression of virulence factors.

We first measured the effect of disturbance on the distribution of bacteria. We observed significant changes in the density of bacteria in the planktonic state at amplitudes of 0.3 mm, 0.7 mm, and 1.1 mm compared to the undisturbed control (0/h, 0 mm) (Fig. 5A). Specifically, the density of this population increased at amplitudes of 0.7 mm and 1.1 mm but decreased at an amplitude of 0.3 mm. We also found that, on average, the density of bacteria in the biofilm state increased at all amplitudes tested relative to 0/h. Significant differences in the density of the biofilm on the pegs were observed at amplitudes of 0.1 mm, 0.7 mm, and 1.1 mm (Fig. 5B). There was a reduction in the density of the biofilm on the peg at an amplitude of 0.1 mm but an increase at amplitudes of 0.7 mm and 1.1 mm. The density of the biofilm attached to the walls of the well was significantly higher at amplitudes of 0.1 mm, 0.3 mm, and 0.7 mm but not 1.1 mm relative to results at 0/h. Overall, changing the amplitude of shaking altered the distribution of bacteria such that after 24 h, the distribution of bacteria differed from that at 0/h.

FIG 5.

Amplitude of disturbance affects expression of agr virulence factors. (A) Density of bacteria in the biofilm and planktonic states following 24 h of disturbance at the amplitudes indicated along the x axis and at 15/h. Density of bacteria in the biofilm state was summed between biofilms on the peg and on the walls of the well. *, significant difference from the undisturbed control (P < 0.036, Wilcoxon [P < 0.0001, Shapiro-Wilk]). Standard deviations are from a minimum of eight biological replicates. (B) Density of bacteria in the biofilm state on either the peg or the walls of the well. *, significant difference from the undisturbed control (P < 0.041, Wilcoxon [P < 0.0001, Shapiro-Wilk]). Standard deviations are from a minimum of eight biological replicates. (C) Effects of increasing shaking amplitude on the expression of saeR. *, significantly different from undisturbed control (P ≤ 0.0066, two-tailed t test). Standard deviations are from a minimum of three biological replicates. ΔC_T values can be found in Fig. S8. In panels C to F, fold change was determined relative to corresponding 0/h condition. (D) Effects of increasing shaking amplitude on the expression of agrC. *, significantly different from undisturbed control (P ≤ 0.0006 two-tailed t test). (Inset) Effects of increasing amplitude on the expression of rnaIII. *, significantly different from undisturbed control (P < 0.0007, two-tailed t test). Standard deviations are from a minimum of three biological replicates. (E, top) Effects of increasing amplitude (shaking frequency, 15/h) on the expression of coa and fnbpA. *, significantly different from undisturbed control (P ≤ 0.0166 two-tailed t test). (Bottom) Average fold change of genes encoding surface attachment proteins. Averages are from data in top panel. Standard deviations are from a minimum of three biological replicates. (F, top) Effects of increasing disturbance amplitude (shaking frequency, 15/h) on the expression of lukS and splA. *, significantly different from undisturbed control (P ≤ 0.0375, two-tailed t test). (Bottom) Average fold change of genes encoding exotoxins and proteases. Averages are from data in top panel. Standard deviations are from a minimum of three biological replicates.

We then used RT-qPCR to measure the expression of saeR, agrC, and rnaIII. We observed a significant decrease in saeR expression at all amplitudes measured compared to that at 0/h (Fig. 5C). Expression of agrC was significantly increased across all amplitudes measured compared to 0/h (Fig. 5D). agrC expression was greatest at an amplitude of 0.1 mm (0/h, P < 0.0001; 0.1 mm to 0.3 mm, P = 0.0373, two-tailed t test). Finally, we observed that there was a significant increase in the expression of rnaIII at shaking amplitudes of 0.1 mm and 1.1 mm, but not 0.3 mm or 0.7 mm (Fig. 5D, inset).

We then measured the expression of genes encoding virulence factors. We observed that the expression levels of coa and fnbpA were generally lower than at 0/h at all amplitudes tested (Fig. 5E, top). However, only at a shaking amplitude of 0.3 mm was the expression of both coa and fnbpA significantly reduced compared to 0/h. Moreover, the average fold reduction in expression of both coa and fnbpA was the greatest at this amplitude (Fig. 5E, bottom). The expression of lukS and splA was the lowest at a shaking amplitude of 0.3 mm (Fig. 5F, top) but was not significantly different than at 0/h. Otherwise, the expression of lukS and splA significantly increased at all other amplitudes measured relative to that at 0/h (Fig. 5F, top). Thus, the lowest average fold change in expression of both lukS and splA was found at an amplitude of 0.3 mm (Fig. 5F, bottom). The changes in gene expression observed in this study were unlikely due to changes in pH. There were no significant differences in pH in the nutrient-poor and nutrient-rich conditions when bacteria were disturbed at 9/h, 15/h (amplitude of 1.1 mm), or 0/h (Fig. S9). Overall, we observed that periodically disturbing S. aureus biofilms at an intermediate amplitude (0.3 mm) could reduce the overall expression of virulence factors. Otherwise, at lower or higher amplitudes, expression of agr-regulated virulence factors was higher than, or no greater than, that of the undisturbed control.

DISCUSSION

We have shown that periodically disturbing a biofilm using a physical force can affect virulence factor production in S. aureus. Disturbance into a nutrient-poor environment revealed that agr-regulated genes encoding surface attachment proteins (coa, fnbpA), exotoxins (lukS), and proteases (splA) were significantly upregulated compared to biofilms that remained undisturbed (Fig. 4). Disturbance into a nutrient-rich environment showed that expression of genes encoding surface attachment proteins were reduced, while those encoding exotoxins and proteases remained unchanged in comparison to the undisturbed condition (Fig. 4). However, this finding was dependent upon the amplitude of the disturbance; increasing or decreasing the amplitude away from 0.3 mm increased the overall expression of agr-regulated virulence factors (Fig. 5). Our results add to our previous work demonstrating that disturbance can impact the expression of genes regulated by quorum sensing (27) and the production of pyoverdine in Pseudomonas aeruginosa (41).

We hypothesize that disturbance affected the concentration of AIP sensed by bacteria in our experiments. In a nutrient-poor environment, disturbance increases the amount of AIP sensed by the bacteria. This was evidenced by the increased expression of agrC and rnaIII relative to the undisturbed control (Fig. 3). Expression of these genes has been previously shown to increase in response to increasing concentrations of AIP (47). Increasing the concentration of AIP sensed by cells would increase expression of exotoxins and proteases. While disturbance also likely allowed more bacteria at the base of the biofilm to acquire any remaining nutrients in the medium that would be otherwise limited by diffusion in the undisturbed condition, these nutrients were quickly utilized. Thus, no additional measurable growth occurred in this medium (Fig. S4). The lack of nutrients reduced metabolism (as measured using ilvD expression and ATP) (Fig. 3) and increased the expression of agr-regulated virulence factors through the saeS/saeR two-component system (15). As a result, expression of surface attachment proteins increased. Importantly, we cannot rule out additional factors that may influence agr expression, such as cell death owing to extended time in nutrient-poor medium or starvation-induced changes in pathways outside the agr network.

In the nutrient-rich environment, the effect of disturbance on determining the amount of AIP sensed appeared to be dependent on shaking frequency and amplitude, but this is not fully understood. While an increase in the expression of agrC was observed at all shaking frequencies and amplitudes tested, expression of rnaIII was not always altered relative to the undisturbed control. At shaking frequencies of 3/h, 9/h, and 15/h (amplitude 0.3 mm), an increase in expression of agrC, but not rnaIII, was observed (Fig. 3). This resulted in a disconnect between repression of surface attachment proteins, whose expression was repressed at high AIP concentrations, and both exotoxins and proteases, whose expression was activated at high AIP concentrations. Specifically, we observed a decrease in expression of surface attachment proteins, while expression of exotoxins and proteases remained unchanged relative to the undisturbed control. Thus, while disturbance under these conditions appears to increase the concentration of AIP sensed (increased expression of agrC), this increase was not “transmitted” through rnaIII to increase expression of exotoxins and proteases. This may be owing to additional enzymes involved in the regulation of elements downstream of rnaIII (i.e., rot) or additional growth phase-dependent networks that interact with the agr locus (e.g., sarA-sarR and sarX-mgrA [48]). Consistency in the expression of agr-regulated virulence factors returned at lower (0.1 mm) or higher (0.7 and 1.1 mm) amplitudes and with a frequency of 15/h. Here, disturbance increased expression of agrC and rnaIII (Fig. 5). This reduced expression of surface attachment proteins and increased expression of exotoxins and proteases. Disturbance in the nutrient-rich condition appeared to increase access to nutrients, as evidenced by the reduction in saeR expression. However, this did not have an observable influence on the expression of agr-regulated virulence factors.

While our results showed that disturbance affected the distribution of bacteria, it was unclear as to how the relative amount of bacteria in the planktonic and biofilm states impacted AIP access, or nutrient access. This certainly warrants future work. Moreover, it is unclear as to how our results would translate to other biofilm growing systems, including those that occur in microfluidic devices, where fluid flow across the biofilm is nearly constant. Fluid flow could remove accumulated AIP and infuse new nutrients; together, this could influence the effects of disturbance. Nevertheless, we surmise that disturbance affects the ability of bacteria to sense AIP, leading to changes in expression of virulence factors that are regulated by agr.

It is possible that S. aureus has mechanisms to sense changes in physical force, which may lead to changes in the expression of virulence factors. For example, environments with strong physical forces may promote the expression of surface attachment proteins; intuitively, augmented expression of these proteins could help S. aureus colonize such an area. Interestingly, regulatory networks and proteins that sense and respond to forces have been described in bacteria; proteins that sense forces play integral roles in the operation of flagella (49), twitching (50), adhesion (51), and biofilm formation (52). In the case of biofilm formation, mechanical forces are known to affect the expression of genes involved in the production of extracellular polymeric substances and extracellular DNA in S. aureus (53). Changes in mechanical force are also known to affect virulence (54), including in Escherichia coli (55) and P. aeruginosa (56). To our knowledge, a system that integrates changes in force and alters expression of virulence factors is yet to be described in S. aureus. While we cannot rule out that an unidentified protein or network that senses changes in force is influencing our results, identification of such a protein or network could pave the way for new mechanisms to rationally perturb biofilm formation, which could alter the expression of virulence factors. Previous work has found that expression of saeR-regulated virulence genes is heterogenous and stochastic, whereas expression of agr-regulated virulence factors is homogenous and deterministic during biofilm development (57). Stochastic expression of saeR may be influencing expression of virulence factors examined in this study. For example, if saeR was expressed in only a subset of cells, then perhaps only those cells would show increased expression of surface attachment proteins in the nutrient-poor condition. However, our experiments lack the resolution to catch stochastic gene expression, as we report population averages.

Interestingly, disturbing the spatial structure of biofilms may have perturbed the balance between bacterial cooperation and competition. Quorum sensing in bacteria has often been considered to be a cooperative behavior (58). Increases in the expression of agrC suggest an increase in cooperation. Due to the close proximity of bacteria in a biofilm, or nutrient availability in the environment, local resources such as nutrients or space become limiting, which increases competition among conspecifics (58). Increased competition can result in reduced metabolism owing to a decrease in available nutrients (59). In our system, an increase in the expression of saeR would suggest an increase in competition. In both nutrient-rich and -poor environments, we hypothesize that disturbance enhances cooperation, as evidenced by the general increase in agrC expression (Fig. 3 and 5). Competition, however, was only increased in the nutrient-poor condition, as noted by the increase in expression of saeR (Fig. 3). In this light, disturbance into a nutrient-rich environment increases cooperation but not competition. This served to reduce the expression of surface attachment proteins, thus reducing overall virulence. However, in a nutrient-poor environment, disturbance increased both cooperation and competition. This served to increase overall virulence, as the expression of both exotoxins and surface attachment proteins increased. Interestingly, while competition can often be regarded as an antagonist to cooperation (60), in this scenario, the combined effect of enhanced competition and cooperation could be quite beneficial to the bacteria during an infection, owing to enhanced virulence.

There is growing interest in developing tools to attenuate the expression of virulence factors in the clinical setting (61). Our results highlight the ability of disturbance to affect quorum-sensing-regulated behaviors and thus may represent a strategy to reduce pathogenesis and infection severity. As many bacteria use quorum sensing to regulate the expression of virulence factors, this approach may be able to reduce pathogenicity of other bacterial species. Interestingly, disturbance may affect multiple bacterial behaviors simultaneously, such as access to nutrients, ability to form biofilm and location of a biofilm, and access to AIP. As there are multiple areas of bacterial physiology that are affected by disturbance, this approach may have the potential to spread out any selective pressure over several targets in the bacteria simultaneously (62). By spreading out the selective pressure, multiple additional peaks in the evolutionary landscape might emerge. This differs from a landscape in which selective pressure is placed on a single target, which serves to limit the number of evolutionary peaks (63). In the latter scenario, where only a few peaks are observed, there is a greater possibility of climbing an evolutionary trajectory that reduces the effectiveness of treatment. Alternatively, with multiple peaks, there is the chance that bacteria will evolve toward a peak that, although it can counteract the effects of disturbance, does not enhance expression of virulence factors. While this notion requires further testing, periodically disturbing the spatial structure of biofilms may represent a viable mechanism to attenuate the expression of virulence factors required for pathogenesis.

MATERIALS AND METHODS

Bacteria, media, and growth conditions.

S. aureus strain USA300 was used throughout this study and was obtained through BEI Resources (NIAID, NIH). Overnight cultures were initiated from single colonies isolated from a tryptic soy agar plate (Becton, Dickinson and Company, Sparks, MD). Colonies were inoculated into 3 mL of TSB and shaken in 15 mL culture tubes (Genesee Scientific, Morrisville, NC) for 24 h at 37°C and at 250 rpm.

Biofilms were grown in an MBEC biofilm inoculator device (Innovotech, Edmonton, AB, Canada). Overnight cultures were washed twice in TSB plus 1% glucose (VWR International, Radnor, PA). Pellets were then resuspended in TSB plus 1% glucose and diluted 100-fold into fresh TSB plus 1% glucose medium. A 190 μL volume of the diluted culture was placed into wells of the MBEC microplate. To reduce evaporation, the surrounding wells were filled with 200 μL of distilled water (dH₂O), and the lid was sealed with three layers of parafilm. The MBEC device was then shaken at 110 rpm at 37°C for 24 h in a shaker incubator that contained a half-filled beaker of dH₂O.

Crystal violet assay.

Biofilms were grown as described above or disturbed as described below for the disturbance assays. To measure the density of biofilms on the peg, the lid of the MBEC device was removed, and the pegs were washed twice at 5 s intervals using 200 μL of 1× phosphate buffered saline (PBS) solution (Fisher Scientific, Fair Lawn, NJ). Bacteria were fixed with 200 μL of 100% ethanol (Fisher Scientific) for 1 min and were stained for 2 min with 200 μL of 0.41% crystal violet (Acros Organics, Fisher Scientific) in 12% ethanol (Fisher Scientific). The pegs were washed four times with 200 μL of 1× PBS and destained in 200 μL of 100% ethanol for 10 min. To measure the incidence of biofilms adhered to the walls of the microplate wells, bacterial cultures were carefully removed from each well. Each well was washed three times with 200 μL of 1× PBS, stained with 200 μL of 0.41% crystal violet for 15 min, and washed three more times with 1× PBS. After the plate was dried for 20 min at room temperature, 200 μL of 95% ethanol was added to each well and the plate was incubated for 20 min. The amount of crystal violet was measured using the OD₅₅₅ in a Victor X4 plate reader (Perkin Elmer, Waltham, MA). OD₅₅₅ values were blanked using an ethanol only control.

Standard curves to determine cell density.

Standard curves of CFU per milliliter as a function of either crystal violet-stained bacteria (measured using OD₅₅₅) or OD₆₀₀ were produced as follows. Biofilms were grown in an MBEC biofilm inoculator device using overnight cultures of S. aureus grown in TSB. The overnight cultures were washed twice in TSB plus 1% glucose. A serial dilution was performed in the MBEC plate in increments of 1:25 with a volume of 190 μL and sealed with three layers of parafilm. The MBEC device was then shaken at 110 rpm at 37°C for 24 h. After 24 h, the lid of the MBEC device was removed and the medium surrounding the pegs was aspirated and placed into a 96-well plate. To generate a plot of OD₆₀₀ versus CFU per milliliter, a serial dilution was performed on the samples and the cell density was measured based on the OD₆₀₀ in a Victor X4 plate reader. Ten microliters of each sample was removed, an additional serial dilution was performed, and CFU were counted on LB agar after 24 h of growth at 37°C. To generate a plot of OD₅₅₅ versus CFU per milliliter, the samples were transferred to a 1.7 mL microcentrifuge tube where they were washed twice in 1× PBS. Pelleted cells were fixed with 200 μL of 100% ethanol for 1 min and were stained for 2 min in 200 μL of 0.41% crystal violet in 12% ethanol. The cells were then washed four times with 200 μL of 1× PBS and destained for 10 min in 200 μL of 100% ethanol. The amount of remaining crystal violet was measured as the OD₅₅₅. CFU per milliliter was plotted as a function of either the OD₆₀₀ or OD₅₅₅, and a linear line placed through zero was used to extract the equation to convert OD into CFU per milliliter.

Effect of a single shake on the distribution of bacteria.

After 24 h of growth, biofilms were washed by placing the lid over the top of a new plate where the wells contained 180 μL of fresh 1× PBS. The microplate was subsequently placed into a Victor X4 plate reader, where biofilms were shaken at specified amplitudes for 10 s using the linear shaking function. Shaking was performed using the fast setting, which shakes the plate linearly at a frequency of 4,800 mm/min. After a single shake, a serial dilution was performed in 1× PBS, and 10 μL aliquots of each dilution were plated on LB agar contained in a 24-well plate, which was subsequently incubated at 37°C overnight. CFU were then quantified using a dilution series.

Disturbance assays.

For nutrient-rich conditions, after 24 h of growth, biofilms were removed from the shaking incubator. The biofilms on the pegs were washed with 190 μL of TSB plus 1% glucose. Biofilms were subsequently placed in 190 μL of fresh TSB plus 1% glucose. To prevent evaporation, the wells surrounding the medium were filled with 200 μL of deionized water and the lid was sealed with three layers of parafilm. For nutrient-poor conditions, biofilms were grown for 24 h and subsequently placed into the plate reader. Importantly, we did not remove the lid nor replace any of the medium for the nutrient-poor condition. Upon placing the plate into a Victor X4 plate reader, biofilms were shaken at specified frequencies and amplitudes for 10 s using the linear shaking function for 24 h. Shaking was performed using the fast setting (frequency of 4,800 mm/min).

ATP assays.

Biofilms were grown in nutrient-poor and nutrient-rich conditions. After 24 h in the undisturbed condition, 100 μL of medium surrounding the peg was transferred into an opaque-walled 96-well plate. The concentration of ATP was measured using the BacterTiter-Glo assay (Promega, Madison, WI) according to the manufacturer’s recommended approach using 100 μL of reagent per well. ATP and cell density were quantified using luminescence and OD₆₀₀, respectively, in a Victor X4 microplate reader. To create a standard curve from which we could determine the concentration of ATP, we used a dilution series of pure ATP (Sigma-Aldrich) and quantified luminescence as described above. Luminescence values were normalized to the OD₆₀₀ and converted to ATP concentration by using the standard curve.

Growth rate.

Growth rate was determined by growing overnight cultures of S. aureus in 3 mL of TSB at 37°C and at 250 rpm for 24 h. For nutrient-rich conditions, bacteria were diluted 1:200 in fresh TSB plus 1% glucose. A 200 μL volume of diluted cells was placed into a 96-well plate. For nutrient-poor conditions, medium from an overnight culture was filtered through a 0.45 μm syringe filter. Bacteria were diluted 1:200 into the filtered medium. Volumes of 200 μL of diluted cells were placed into a 96-well plate. For both nutrient-rich and -poor conditions, the medium was then overlaid with 70 μL of mineral oil, and the plate was placed into a microplate reader that was preheated to 37°C. Every 10 min, OD₆₀₀ values were measured for a total of 24 h. To extract growth rate (nutrient-rich condition only), we followed a previously published approach (64, 65). Growth curves were log transformed and normalized to the initial minimum density. The logistic equation was used to fit the growth curve and to extract the maximum growth rate.

$$y=\frac{A}{\left\{1\hspace{0.17em}+\hspace{0.17em}\text{exp}\left(\frac{4{\text{μ}}_m}{A}\left(\lambda -t\right)+2\right)\right\}}$$

where A is the maximum density, μ_m is the maximal growth rate, and λ is the lag time. For all parameters (A, μ_m, and λ), the lower bound was set to zero. The upper bounds for A, μ_m, and λ were 2, 1, and 10, respectively. Fitting was performed in MATLAB 2022a (MathWorks Inc., Natick, MA) using lsqcurvefit.

RNA extraction.

Total RNA was extracted using the Qiagen RNeasy minikit following the RNAprotect bacteria reagent handbook with modifications. Briefly, S. aureus biofilms were grown as described above. To extract RNA from the planktonic cells, ~180 μL of medium was removed from each well of the MBEC plate, placed into a 1.5 mL centrifuge tube, and centrifuged for 1 min at 12,000 rpm. Pelleted cells were resuspended in 200 μL of RNase-free water. Each tube was then centrifuged once again for 1 min at 12,000 rpm, and any residual supernatant was discarded. To extract RNA from S. aureus biofilms, pegs containing biofilms were individually removed using sterilized forceps. Pegs were snapped off at the base of the lid to avoid any structural damage to formed biofilms. Each peg was placed in 1.7 mL centrifuge tubes containing 400 μL of RNase-free water. Tubes containing the pegs were placed in a sonicator (CPX1800 ultrasonic bath; Fisher Scientific) for 10 min to remove biofilms attached to the pegs. The following was then performed when extracting RNA from either the biofilm or bacteria in the planktonic state. Each tube was then centrifuged for 1 min at 12,000 rpm, and any residual supernatant was discarded. Thirty microliters of a 1 mg/mL solution of lysostaphin from Staphylococcus staphylolyticus (Sigma-Aldrich) in nuclease-free water was then added to each centrifuge tube and incubated at room temperature for 10 min while vortexing every 2 min. Total RNA was then extracted according to the manufacturer’s recommended protocol, including the optional in-column DNase digestion using the RNase-free DNase set. Following RNA extraction, we used the Qiagen DNase Max kit following the Quick Start protocol provided by the manufacturer.

cDNA synthesis.

cDNA synthesis was performed using the Bio-Rad reverse transcription supermix for RT-qPCR kit according to the manufacturer’s recommendations. Reaction mixtures were placed into a Bio-Rad C1000 Touch thermal cycler, and cDNA synthesis was performed with the following program: priming for 5 min at 25°C, reverse transcription for 20 min at 46°C, and inactivation for 1 min at 95°C.

RT-qPCR.

Quantitative RT-qPCR was performed using the Bio-Rad iTaq Universal SYBR green Supermix kit according to the manufacturer’s instructions. RT-qPCR was performed using a Bio-Rad CFX96 Touch real-time PCR detection system. The parameters for each RT-qPCR cycle used were 95°C for 3 min (initial denaturation), 95°C for 10 s, and 60°C for 30 s. A total of 40 cycles followed by melt curve analysis were performed. Primers for RT-qPCR analysis are listed in Table 1. Amplicon specificity and size were confirmed using agarose gel electrophoresis. The average threshold cycle (C_q) was normalized using the comparative threshold cycle (C_T) method (66). All ΔC_T values were normalized using gyrB within each shaking condition (67). To calculate fold changes in gene expression, ΔC_T values from the experimental and control conditions were averaged. We confirmed that the expression of gyrB did not change as a function of shaking frequency (P > 0.17, Kruskal-Wallis test).

TABLE 1.

List of primers used in RT-qPCR

Target gene	Direction	Primer sequence (5′–3′)
gyrB	Forward	CAAATGATCACAGCATTTGGTACAG
	Reverse	CGGCATCAGTCATAATGACGAT
coa	Forward	GAGACCAAGATTCAATAAGCCATCAG
	Reverse	GGGCGAGCGCCATATGATAC
rnaIII	Forward	GTGATGGAAAATAGTTGATGAGTTGTTT
	Reverse	GAATTTGTTCACTGTGTCGATAATCC
fnbpA	Forward	TGGTACTGATGAAGTTGATTTTAGAACACA
	Reverse	AACCATTATCCCAAGTTAAGGTATATCCTCT
lukS	Forward	TGCTACTTCGTTTCATGAATCTAAAGCT
	Reverse	CCCCACTTATCGCTACTTGTATCTT
agrC	Forward	CTTGATAATGCAATTGAGGCATCAAC
	Reverse	GATAGACCTAAACCACGACCTTCAC
saeR	Forward	CAACTGTCGTTTGATGAATTAACACTTATTAACT
	Reverse	CCACAATAACTCAAATTCCTTAATACGCAT
splA	Forward	ACCAGTACCACCCACAAATGCTAC
	Reverse	TCATTCAATTGCCAAAGCAGAAAAG

Statistical analysis.

All statistical tests were performed in JMP Pro 15 (SAS Institute Inc., Cary, NC) as indicated throughout the text and figure legends. A Shapiro-Wilk test was used to assess normality. The P value from the Shapiro-Wilk test is only reported for nonnormally distributed data sets. For normally distributed data sets, unpaired Student’s two-tailed t tests with unequal variance were performed using the “each pair, Student’s t” function. For nonnormally distributed data sets, a Mann-Whitney test (2-sample test, normal approximation) was used. For RT-qPCR data, all statistical tests were performed using ΔC_T values normalized to gyrB. P values were then used to report on significant differences as they related to gene fold change in transcript. All fold change values were plotted relative to the undisturbed control (0/h) with the exception of data presented in Fig. S3.

Data availability.

All raw and supporting data are either presented in the supplemental material or can be located in the Dryad digital repository using the following link: https://datadryad.org/stash/share/EPDIFzrktMLDY_20zlD9zHKcNA5u5gfBJCRb1oiYjmw.