Shear rate sensitizes bacterial pathogens to H₂O₂ stress

Significance

For over 100 years, bacteria have been studied in simplified conditions. During infection, bacteria colonize complex environments containing fluid flow, exemplified by the bloodstream. Here, we use microfluidics to examine how flow impacts bacterial pathogens. We discover that flow triggers an effect like “wind-chill” that makes cells sensitive to chemical concentrations 100 to 1,000 times lower than traditionally used. Using flow intensities and chemical concentrations found in the human bloodstream, we demonstrate that host-relevant flow impacts bacterial pathogens. Demonstrating the importance of our findings, we establish that flow sensitizes both Pseudomonas aeruginosa and Staphylococcus aureus to chemical stress. Overall, our research highlights the need to study flow to capture the full complexity of how bacterial pathogens operate in realistic contexts.

Abstract

Cells regularly experience fluid flow in natural systems. However, most experimental systems rely on batch cell culture and fail to consider the effect of flow-driven dynamics on cell physiology. Using microfluidics and single-cell imaging, we discover that the interplay of physical shear rate (a measure of fluid flow) and chemical stress trigger a transcriptional response in the human pathogen Pseudomonas aeruginosa. In batch cell culture, cells protect themselves by quickly scavenging the ubiquitous chemical stressor hydrogen peroxide (H₂O₂) from the media. In microfluidic conditions, we observe that cell scavenging generates spatial gradients of H₂O₂. High shear rates replenish H₂O₂, abolish gradients, and generate a stress response. Combining mathematical simulations and biophysical experiments, we find that flow triggers an effect like “wind-chill” that sensitizes cells to H₂O₂ concentrations 100 to 1,000 times lower than traditionally studied in batch cell culture. Surprisingly, the shear rate and H₂O₂ concentration required to generate a transcriptional response closely match their respective values in the human bloodstream. Thus, our results explain a long-standing discrepancy between H₂O₂ levels in experimental and host environments. Finally, we demonstrate that the shear rate and H₂O₂ concentration found in the human bloodstream trigger gene expression in the blood-relevant human pathogen Staphylococcus aureus, suggesting that flow sensitizes bacteria to chemical stress in natural environments.

Classically, research on bacterial stress responses has focused on chemical stressors such as nutrient availability (1), pH (2), antibiotics (3), and oxidative stress (4). Bacteria respond to chemical perturbations with well-studied physiological responses to survive and grow in stressful situations (5–8). For simplicity, bacterial stress responses have been largely studied in batch cell culture, which has allowed researchers to identify and characterize many of the important signaling pathways associated with stress. However, simplified experimental systems neglect the dynamic mechanical features of natural and host systems (9).

Recently, a surge of research on bacterial mechanosensing has revealed that fluid flow impacts virulence (10, 11), biofilm formation (12), and gene expression (13). While bacterial transcriptional responses to flow were assumed to require the cellular measurement of forces (10, 12), one report challenged this assumption and revealed that flow can trigger bacterial gene expression in a force-independent manner (13). This report proposed naming flow-sensitive transcriptional responses “rheosensitive” (as rheo- is Greek for flow), due to lack of direct evidence that bacteria respond to flow by measuring forces (13). Thus, it is currently debated how flow generates bacterial transcriptional responses.

Here, we discover that flow generates transcriptional responses in the human pathogens Pseudomonas aeruginosa and Staphylococcus aureus by sensitizing them to H₂O₂ stress. We quantitatively establish that physical shear rate (a measure of fluid flow) triggers bacterial responses by abolishing spatial gradients of H₂O₂ generated by bacterial scavenging. Thus, flow triggers bacterial responses through a biophysical mechanism that it is highly reminiscent of “wind chill.” We also observe that bacteria in flow are sensitive to concentrations of the host-generated antimicrobial H₂O₂ that are 100–1,000 times less than those traditionally used in lab conditions. Notably, the incorporation of flow explains the long-standing mystery of why H₂O₂ concentrations required to generate responses in simplified lab conditions are orders of magnitude higher than those found in hosts.

Results

To understand how flow affects gene expression, we focused on the flow-sensitive transcriptional response in the bacterium P. aeruginosa. P. aeruginosa responds to flow by upregulating a suite of genes (13), including many genes upregulated during human infection (14). As a representative example of the larger flow-sensitive response, we focused our efforts on the fro operon. The fro operon is rapidly and robustly upregulated by flow (13) and required for full infection in multiple animal models (15, 16). For single-cell analysis, we used a fro reporter strain that reports on fro expression with yellow fluorescent protein (YFP) and encodes a constitutively expressed mCherry for normalization. To confirm that flow induces fro expression, cells in a microfluidic device were simultaneously subjected to flow from a syringe pump and imaged with a fluorescence microscope (Fig. 1A). Throughout this study, we represent the intensity of flow using shear rate, which is calculated using flow rate and channel dimensions (13). Consistent with previous results (13), fro is strongly induced after 3 h of exposure to a shear rate of 800 s⁻¹ in an LB medium (Fig. 1B). Thus, the fro reporter represents a valuable tool to dissect how flow generates transcriptional responses in bacteria.

Fig. 1.

Conditioning media suppresses flow-sensitive gene expression in P. aeruginosa. (A) Microfluidic setup used in this study. Microfluidic devices are custom fabricated with polydimethylsiloxane and glass coverslips. Channels are 50 µm tall × 500 µm wide. Bacteria adhere to the channel wall. Microfluidic devices are simultaneously subjected to flow from a syringe pump and imaged 1 cm into the channel by a fluorescence microscope. (B) Fluorescence and phase images of the P. aeruginosa fro reporter strain in flow (at a shear rate of 800 s⁻¹) over 180 min. Cell density in channels remains relatively constant due to the combinatorial effects of cell growth and cell departure. Images are representative of three biological replicates. (Scale bars, 5 µm.) (C) Quantification of fro expression by dividing YFP intensity by mCherry intensity as described in SI Appendix, Fig. S1. Green represents cells exposed to LB media, while gray represents cells exposed to conditioned LB. Shaded regions show SD of three biological replicates.

How does flow generate a transcriptional response? One possibility is that flow affects a rate-dependent biophysical process such as chemical transport. To test this hypothesis, we flowed fresh LB medium and LB medium that had been conditioned by P. aeruginosa into independent channels of the device. A conditioned medium was generated by exposing fresh, sterile LB to a culture of P. aeruginosa for 60 min, followed by filter-sterilizing the media to remove cells. We determined that cells exposed to flow with LB induced fro expression fivefold, but fro was not induced after 3 h of exposure to flow in conditioned LB (Fig. 1 B and C). Thus, our results suggest that chemical transport of a molecule into or out of cells underlies how bacteria respond to flow.

What is the identity of the flow-sensitive molecule? To examine whether the molecule was conserved, we tested the effect of conditioning media with Escherichia coli, S. aureus, and Enterococcus faecalis. Conditioning media with E. coli or S. aureus led to complete loss of fro induction in flow, while conditioning media with E. faecalis resulted in high fro induction (Fig. 2A). P. aeruginosa, E. coli, and S. aureus are catalase-positive (17), while E. faecalis is catalase-negative (18). Catalase specifically degrades H₂O₂ (19), supporting the hypothesis that the flow-sensitive molecule is H₂O₂. Consistent with the literature (17, 18), we confirm that P. aeruginosa, E. coli, and S. aureus can deplete H₂O₂ from media in 30 min, while E. faecalis cannot (SI Appendix, Fig. S2). To test the role of H₂O₂ scavenging enzymes during the conditioning of media, we compared the effect of conditioning media with wild-type E. coli and an E. coli ∆ahpC ∆katG ∆katE ∆ccp mutant known to lack scavenging ability (20). While media conditioned with wild-type E. coli prevented fro induction, media conditioned with the E. coli ∆ahpC ∆katG ∆katE ∆ccp mutant (20) was still capable of inducing fro expression (SI Appendix, Fig. S3). Taken together, our results support the hypothesis that the flow-sensitive molecule is H₂O₂.

Fig. 2.

H₂O₂ is a flow-sensitive molecule. (A) P. aeruginosa fro expression after 180 min in flow (at a shear rate of 800 s⁻¹) with LB media or media conditioned by P. aeruginosa, E. coli, S. aureus, or E. faecalis for 60 min. Quantification shows the average and SD of three biological replicates. Images show YFP and mCherry fluorescence, as well as phase contrast. (B) fro expression after 180 min in flow (at a shear rate of 800 s⁻¹) with LB media, LB treated with catalase, media conditioned by P. aeruginosa for 60 min, and conditioned media resupplied with 9 μM H₂O₂. Quantification shows the average and SD of three biological replicates. Images are taken 1 cm into the channel and show YFP and mCherry fluorescence, as well as phase contrast. (Scale bars, 5 µm.) Channels are 50 µm tall × 500 µm wide.

We reasoned that if H₂O₂ was the flow-sensitive molecule, it must be present in our media. We tested H₂O₂ levels with a peroxidase assay (21) and determined that our laboratory LB stocks contained approximately 9 μM H₂O₂ (SI Appendix, Fig. S4). The presence of H₂O₂ in LB is a reproducible but underappreciated detail (22, 23). The chemical production of H₂O₂ in LB is mediated by a light-dependent reaction involving riboflavin (23). To test if H₂O₂ in our media was required to trigger fro induction, we repeated our flow experiment with M9 minimal media lacking H₂O₂. Minimal media was incapable of inducing fro expression, supporting our hypothesis that H₂O₂ is required for fro induction (SI Appendix, Fig. S5). To directly test the role of catalase, we treated LB with purified catalase, which depleted H₂O₂ concentrations in LB to effectively zero (SI Appendix, Fig. S6). Catalase-treated media did not induce fro expression in flow, further supporting the hypothesis that the flow-sensitive molecule is H₂O₂. (Fig. 2B). Finally, we tested fro expression with conditioned media resupplied with 9 μM H₂O₂. Reintroduction of H₂O₂ restored robust fro induction (Fig. 2B), demonstrating that the flow-sensitive molecule is H₂O₂.

Armed with our understanding that H₂O₂ is the flow-sensitive molecule, we reanalyzed RNA-sequencing data from a previous microfluidic-based transcriptomic experiment (13). In addition to the fro operon, we identified many flow-induced genes implicated in H₂O₂ scavenging, such as ahpCF, ahpB, and katB (Fig. 3A). AhpCF and AhpB are NADH peroxidases that scavenge H₂O₂ (24–26). KatB is a catalase that converts H₂O₂ into water and oxygen (19, 24). AhpCF, AhpB, and KatB are induced by H₂O₂ and regulated by OxyR, a H₂O₂ sensor located in the cytoplasm (24, 27–30). Thus, our results suggest that shear rate leads to the intracellular accumulation of H₂O₂. In support of this model, an additional H₂O₂-regulated gene trxB2, which encodes thioredoxin reductase (27), was also upregulated in response to shear rate (Fig. 3A). Together, our results suggest the interplay of shear rate and H₂O₂ generate cellular stress and trigger a transcriptional response.

Fig. 3.

Fluid flow upregulates multiple enzymes that scavenge H₂O₂. (A) P. aeruginosa RNA sequencing data (13) showing relative transcript levels comparing flow conditions to no flow conditions. fro represents an average of the froABCD operon and ahpCF represents an average of ahpC and ahpF. Quantification shows the average and SD of three biological replicates. (B) H₂O₂ concentration of LB media over time when treated with wild-type P. aeruginosa and ΔkatA ΔkatB ΔahpCF mutant cells at an OD of 0.4 to 0.5. (C) H₂O₂ concentration of LB media over time when treated with wild-type E. coli and ΔkatE ΔkatG ΔahpC Δccp mutant cells at an OD of 0.4 to 0.5. (D) H₂O₂ concentration of LB media over time when treated with wild-type S. aureus and ΔkatA ΔahpCF mutant cells at an OD of 0.4 to 0.5. All H₂O₂ concentrations were measured using a peroxidase assay (21) and quantification shows the average and SD of three biological replicates.

What is the relationship between flow and H₂O₂? We hypothesize that shear rate replenishes H₂O₂ to overcome rapid scavenging by cells. To explore the kinetics of H₂O₂ scavenging, we used a peroxidase assay (21) to measure H₂O₂ concentrations of LB media treated with P. aeruginosa, E. coli, or S. aureus over 60 min. Additionally, we compared wild-type and mutant strains of P. aeruginosa, E. coli, and S. aureus. At a cell density of 0.5 OD, wild-type P. aeruginosa scavenged about 50% of the available H₂O₂ in 30 s, while a ∆katA ∆katB ∆ahpCF mutant was significantly impaired at scavenging H₂O₂ (Fig. 3B). Wild-type E. coli scavenges about 90% of the available H₂O₂ in 30 s, while an ∆ahpC ∆katG ∆katE ∆ccp mutant had significantly diminished scavenging ability (20) (Fig. 3C). Finally, wild-type S. aureus scavenged about 25% of the available H₂O₂ in 30 s and a ∆katA ∆ahpCF mutant was also significantly impaired at H₂O₂ scavenging (Fig. 3D). Together, our batch cell culture experiments reveal that bacteria rapidly scavenge H₂O₂, providing support to our hypothesis that flow triggers a biological response by replenishing H₂O₂.

How sensitive are bacterial cells to H₂O₂ in flow? We measured fro induction at a range of H₂O₂ concentrations, while maintaining a constant shear rate of 800 s⁻¹. We observed that 2 µM H₂O₂ did not induce fro expression. In contrast, 4 µM H₂O₂ led to a threefold induction and 8 µM H₂O₂ led to a fivefold induction in fro expression (Fig. 4A and SI Appendix, Fig. S7A). Thus, the minimum concentration necessary to elicit a flow-sensitive response at a shear rate of 800 s⁻¹ is approximately 4 µM H₂O₂ (Fig. 4A). Surprisingly, 4 µM is 100 to 1,000 times lower than the H₂O₂ concentration traditionally studied in batch cell culture (24, 31–35). Moreover, the H₂O₂ concentration in the human bloodstream is thought to be between 1 and 5 µM (36). Together, these results support the hypothesis that bacteria in flow are highly sensitive to H₂O₂ and suggest that flow generates stress in natural environments.

Fig. 4.

Simulation and experiment determine the minimal shear rate to trigger flow-sensitive gene expression. (A) P. aeruginosa fro expression after 180 min in flow (at a shear rate of 800 s⁻¹) with LB media with varied H₂O₂ concentrations imaged 1 cm into the channel. (B) Relationship between shear rate and the relative H₂O₂ concentrations 1 cm into a channel from our simulations. Quantification shows the average and SD of multiple bins in the middle of the channel. (C) Visual representation of H₂O₂ molecules in simulated microfluidic channels from the start (0 cm) to the end (2 cm) of a 2-cm channel at different shear rates. Green dots represent individual H₂O₂ molecules experiencing diffusion and fluid flow from left to right. To represent bacterial scavenging, 1 of every 100 molecules that hit the lower surface is removed. This assumption is based on calculations that use the scavenging kinetics defined in Fig. 3. (D) fro expression after 180 min in flow (at 8 μM H₂O₂) with varied shear rates imaged 1 cm into the channel. (E) fro expression near the start (1.5 cm) and end (25.5 cm) of a 27-cm long channel after 180 min in flow at 8 μM H₂O₂ with a shear rate of 240 s⁻¹. Quantification in panels A, D, and E shows the average and SD of three biological replicates. Green and gray in panels A, D, and E signify expression levels that are statistically different with P < 0.05.

To understand the mathematical relationship between flow and H₂O_2, we calculated the Péclet number for our experimental system. The Péclet number describes if shear rate or diffusion is dominant in a particular regime and is proportional to the shear rate divided by diffusion. When shear rate dominates diffusion, the Péclet number is greater than one. When diffusion dominates shear rate, the Péclet number is less than one. To solve for the minimal shear rate required to overtake diffusion, we set the Péclet equation equal to one and solved for shear rate (SI Appendix, Fig. S8). Our calculations show that a shear rate of approximately 166 s⁻¹ should dominate diffusion in our experimental system (SI Appendix, Fig. S8).

To generate testable mathematical predictions, we developed a simulation for the advection-diffusion transport of individual H₂O₂ molecules in a channel. The simulation uses experimental values for shear rate and the estimated diffusion coefficient of H₂O₂ (37). To simulate H₂O₂ removal by cells, we included a feature where 1 of every 100 molecules that contacted the channel surface was removed. We used the value of 1 in 100 based on calculations (described in Materials and Methods) that use the scavenging kinetics defined in Fig. 3. Our simulation shows that as the shear rate increases from 8 to 8,000 s⁻¹, the H₂O₂ concentration at the middle of the channel increases (Fig. 4B). For low shear rates, H₂O₂ molecules are rapidly depleted. In contrast, higher shear rates replenish depleted molecules and maintain the concentration of H₂O₂ close to the initial value. Our simulated results highlight how rapid H₂O₂ scavenging generates gradients of H₂O₂, which are abolished by shear rates of at least 240 s⁻¹. Together, our mathematical calculations and simulations predict that a shear rate of 166 s⁻¹ to 240 s⁻¹ will trigger flow-sensitive gene expression.

To test our mathematical predictions, we measured fro induction at a range of shear rates, while maintaining a constant H₂O₂ concentration of 8 μM. We observed that shear rates of 8 s⁻¹ and 80 s⁻¹ did not trigger fro expression. In contrast, 240 s⁻¹ led to a threefold induction and 800 s⁻¹ led to a fivefold induction in fro expression (Fig. 4D and SI Appendix, Fig. S7B). Thus, the minimum shear rate necessary to elicit a flow-sensitive response at a H₂O₂ concentration of 8 μM is approximately 240 s⁻¹ (Fig. 4D). For perspective, human veins have shear rates of approximately 80 s⁻¹ and human arteries have shear rates of approximately 800 s⁻¹ (38).

To test the prediction that cell scavenging generates spatial H₂O₂ gradients, we examined fro expression at the beginning and end of a long microfluidic channel (SI Appendix, Fig. S9). For this experiment, we used a shear rate of 240 s⁻¹ and a H₂O₂ concentration of 8 μM. Cells at the beginning of the channel induced fro expression 3.5-fold, while cells at the end of channel experienced lower induction levels (Fig. 4E). To further examine how flow transports H₂O₂ molecules, we repeated our long channel experiment at a shear rate of 80 s⁻¹. As predicted by our simulation, a shear rate of 80 s⁻¹ results in low fro expression 1.5 cm into the channel and at 25.5 cm into the channel (SI Appendix, Fig. S10A). Repeating our long channel experiment at a shear rate of 800 s⁻¹ resulted in high fro expression at 1.5 cm into the channel and at 25.5 cm, which further validates our model and shows that flow can mediate long-range transport of H₂O₂ (SI Appendix, Fig. S10B). Based on our simulation, we hypothesized that cell density would affect H₂O₂ levels. We tested this hypothesis by performing an experiment where cell density was altered while shear rate was held constant at 80 s⁻¹ and H₂O₂ concentration was held constant at 8 µM H₂O₂. Supporting our hypothesis and further validating our simulation, higher cell density led to lower fro expression and lower cell density led to higher fro expression (SI Appendix, Fig. S11). Together, our biophysical experiments support our mathematical predictions, establish that flow modulates chemical gradients, and demonstrate that physiological levels of flow trigger a transcriptional response in P. aeruginosa.

Does flow generate stress in other bacteria? The biological requirements for a flow-sensitive response are a permeable membrane, H₂O₂ scavenging, and H₂O₂ sensing capacity. As these three features are widespread in bacteria, we hypothesized that the human pathogen S. aureus would also exhibit flow-sensitive gene expression. We focused our efforts on S. aureus as it infects the human bloodstream, which contains high shear rates (38) and low micromolar H₂O₂ concentrations (36). We generated a S. aureus YFP fluorescent reporter to the promoter of ahpCF. In S. aureus, ahpCF expression is induced by H₂O₂ through the function of the H₂O₂-sensing transcriptional regulator PerR (39, 40) (SI Appendix, Fig. S12). When we subjected S. aureus cells to a shear rate of 800 s⁻¹ and a H₂O₂ concentration of 8 μM, ahpCF was induced fourfold compared to no flow conditions (Fig. 5). Thus, flow triggers gene expression in S. aureus, suggesting that the flow-sensitive stress response we have discovered is widely conserved.

Fig. 5.

Shear rate triggers flow-sensitive gene expression in S. aureus. (A) S. aureus ahpCF expression over 180 min in the presence of flow (green) or no flow (gray). Cells contained a P_ahp::YFP transcriptional reporter. Both treatments had 8 µM H₂O₂. Flow treatment was at a shear rate of 800 s⁻¹. Quantification shows the average and SD of at least three biological replicates. (B) YFP fluorescence and phase contrast images taken 1 cm into the channel, representative of at least three biological replicates. (Scale bars, 5 µm.) Channels are 50 µm tall × 500 µm wide.

Discussion

Collectively, our results provide a molecular mechanism of how flow generates stress and triggers a transcriptional response (SI Appendix, Fig. S12). We show that our flow-sensitive transcriptional response hinges on the quantitative relationships between shear rate, H₂O₂, and cell scavenging. First, H₂O₂ from the environment diffuses into cells, where it is scavenged by catalases and NADH peroxidases (Fig. 3). Second, H₂O₂ scavenging results in a zone of depletion and H₂O₂ spatial gradients in the environment (Fig. 4). Third, shear rates above 240 s⁻¹ replenish H₂O₂ in the environment, abolish spatial gradients, and lead to increased accumulation of H₂O₂ in cells (Fig. 4). Fourth, cells sense intracellular H₂O₂ levels and activate a transcriptional response that induces factors to mitigate H₂O₂ damage (Fig. 5). Thus, flow triggers a biological response by countervailing the ability of cells to remove a chemical stressor from the environment.

Reports have classified bacterial responses to flow as mechanosensing (10, 12) or rheosensing (13). Classification of bacterial flow responses as mechanosensing was influenced by a few well-established examples of force-sensing in eukaryotes (41, 42). Based on this classification, the study of bacterial flow responses focused on shear force and largely neglected the effects of flow on chemical transport. In contrast, the field of fluid dynamics has long appreciated that flow is a major driver of chemical transport (43). The attempt to classify bacterial flow responses as rheosensing (13) was intended to provide a more conservative view, as the mechanisms underlying these responses were unknown. Now, with the data presented here, it is clear that flow can trigger a bacterial response by affecting chemical transport of small molecules such as H₂O₂. Our results provide proof-of-principle that flow could amplify or nullify the effect of a wide variety of molecules, such as amino acids, carbon sources, oxygen, and antimicrobials. Thus, in the current manuscript, we refrain from classifying the flow-sensitive response as mechanosensing or rheosensing, as we prefer the more generalizable explanation that flow has a critical role in generating dynamics of chemical transport.

As an analogy to our findings, we note that flow-sensitive gene expression is conceptually like wind chill. In our system, shear rate generates a biological response by abolishing a chemical gradient. Wind chill describes how wind speed generates a biological response by abolishing a temperature gradient. Historically, the concept of wind chill did not exist until 1945 (44). However, it is now widely reported by the National Weather Service due to the critical importance of its effect on human health. By analogy, we propose that in order to understand how cells interact with their environment in natural systems, it is essential to include flow in experimental systems.

Our discovery that bacteria in flow are sensitive to H₂O₂ concentrations that are 100 to 1,000 times lower than traditionally used should provoke a paradigm shift in the way we think about H₂O₂ stress. For simplicity, studies on H₂O₂ stress in bacteria have primarily used batch cell cultures and millimolar amounts of H₂O₂ (24, 31–35). However, natural environments are unlikely to contain such high H₂O₂ concentrations (9). Additionally, it has been noted that H₂O₂ sensors OxyR and PerR are likely sensitive to much lower concentrations (23, 29, 45). Thus, the H₂O₂ concentrations required to generate bacterial responses in laboratory conditions do not reflect the environments where bacteria live. Our observation that bacteria in flow are sensitive to low-micromolar levels of H₂O₂ likely reconciles this discrepancy. Natural bacterial environments, such as the human bloodstream, contain low-micromolar levels of H₂O₂ (36). Without considering the effect of flow, these levels are insufficient to generate bacterial responses. In light of our study, it is apparent that H₂O₂ in flowing blood is sufficient to affect bacteria and potentially has a role in the restriction of bacterial growth.

Materials and Methods

The bacterial strains used in this paper are described in SI Appendix, Table S1. The primers used in this study are listed in SI Appendix, Table S2. The plasmids used are described in SI Appendix, Table S3. Information on the generation of mutant strains, H₂O₂ experiments, microfluidic experiments, microscopy, image quantification, and mathematical simulations are provided in Materials and Methods section of SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix. The raw data and code used to support this study are available in Github (https://github.com/padrong/ImageAnalysis-Adaptations) (46).