Bacteriostatic effect of sequential hydrodynamic and ultrasound-induced stress



John S. Chapman, Ashland Research Center, 500 Hercules Road, Wilmington, DE 19808, USA. E-mail:



To elucidate the mechanism of action of a nonchemical microbial control technology employing coupled hydrodynamic and ultrasound-induced stress.

Methods & Results

The effects of a laboratory model system using a commercial nonchemical device on Pseudomonas putida revealed growth and respiration were inhibited without a loss of viability from the treated population. Damage to cell membranes was evident using fluorescent microscopy and a reporter strain containing lux genes fused with a membrane damage stress-response promoter. Other reporter strains also indicated the possible involvement of DNA and protein repair systems. A consequence of treatment was a reduced ability to form biofilms.


The nonchemical device caused a biostatic effect on treated cells induced by sublethal damage to several cellular systems, including cell membranes.

Significance and Impact of the Study

The study demonstrates that biostasis can be an effective mechanism for microbial control in some industrial systems and provides insight into understanding and applying this device and other nonchemical microbial control technologies to real-world problems of microbial contamination.


Microbial contamination of cooling water systems and industrial process water with consequent biofilm formation reduces the ability of these systems to remove heat from industrial and environmental cooling systems and degrades water quality in industrial processes. The reduced efficiency imposes costs on plant and building operations through increased energy expenditures, maintenance and reduced manufacturing output. In some cases, biofilms in the system can act as harbours for pathogenic bacteria including Legionella (Atlas 1999). Microbial control in these systems is often achieved through the application of both oxidizing and/or nonoxidizing biocides. However, the use of chemical biocides may, in some cases, have negative effects on the environment, the health and safety of workers or process operating parameters. In these cases, nonchemical microbial control technologies may offer substantial advantages directly by reducing the use of biocides and indirectly by reducing the use of other water control chemicals.

Understanding the mechanism of action of microbial control technologies can lead to more efficient strategies for their application and provide insight into increasing their potency. The most common measure of the effectiveness of a microbial control treatment programme is to measure the viability of the population being treated. However, cellular viability is the outcome of a large number of subcellular processes and is not a signal of sufficient resolution to elucidate antimicrobial mechanisms. The chemical biocides such as formaldehyde, isothiazolones and cationic biocides are bactericidal, but an examination of their mechanisms of action at the cellular level indicates different routes to lethality, such as inactivation of critical enzymes (Rossmore and Sondossi 1988), the intracellular generation of radicals (Chapman and Diehl 1995) and membrane disruption (Denyer 1990). Thus, an in-depth investigation into the effects of this proprietary nonchemical technology on various aspects of bacterial structure and physiology was expected to lead to insights about its application and improvement.

The device used in this study is a proprietary, nonchemical, microbial control technology used in recirculating cooling towers, related cooling applications and certain industrial process water streams (Fig. 1). It is a side-stream device that directs water through a venturi into which air is aspirated and then to a chamber fitted with a number of ultrasound emitters (the number of emitters is dependent on the system volume). A venturi is a constriction in the pipe; as water enters the venturi, its velocity increases; and upon exiting the constriction rapidly decreases, which causes turbulence that is enhanced by the aspirated air. Further turbulence occurs as the water exits the circular venturi into the square (cross-sectional) ultrasound chamber.

Figure 1.

Diagram of the microbial control device. Water is mixed with air (20% v/v); forced through a venturi; and exposed to six low power, high frequency (1·65 MHz) ultrasound emitters. The black arrow shows the direction of the water flow.

In the ultrasound chamber, the water is exposed to low intensity, high frequency ultrasonic energy. The antimicrobial effects of ultrasound are believed to include the generation of turbulence, shear stress and cavitation (Gogate 2007). Cavitation is a process in which microbubbles form, grow and collapse extremely rapidly (Lambert et al. 2010). The collapse of these microbubbles leads to the local release of high energy in the form of heat, turbulence and the release of free radicals due to pyrolysis of the water (Gogate et al. 2003). However, cavitation only occurs in high energy intensity systems. The antimicrobial action of ultrasound at various frequencies and intensities has been well documented and has been shown to be effective against algae (Broekman et al. 2010)) and bacteria (Drakopoulou et al. 2009), including Legionella (Declerck et al. 2010).

In this study, we examine the antimicrobial mechanism of sequential hydrodynamic and ultrasonic-induced stress on a model microbial population using a commercial, nonchemical, microbial control device.

Materials and methods

Organism and culture conditions

The bacterial strains used in this work are listed in Table 1. Pseudomonas putida ATCC 47054 (American Type Culture Collection, Rockville, MD, USA) was used as the test organism in all but the reporter strain experiments. It was routinely cultured in minimal salts glucose plus 0·01% yeast extract medium (M9YG) at 28°C with shaking. The Escherichia coli stress-response reporter strains used in this work were routinely cultured in M9YG and incubated at 37°C. The Ps. putida inoculum was prepared by diluting one litre of overnight culture with one litre of fresh media and incubating at 28°C with shaking for 2 h prior to diluting the inoculum into the sump.

Table 1. List of strains and plasmids used in this work
Bacterial strain or plasmidPhenotype or genotypeReference
Pseudomonas putida ATCC 47054Wild-typeATCC
Escherichia coli RFM443F-galK2 lac-74 rpsL200(Drolet et al. 1995)
E. coli RFM443 pUCD61517·55 kb broad-host range Apr Kanr plasmid with promoterless luxCDABE from Vibrio fischeri(Rogowsky et al. 1987)
E. coli RFM443 pGrpE::luxCDABEE. coli grpE promoter fused to luxCDABE(Van Dyk et al. 1994)
E. coli RFM443 pRecA::luxCDABEE. coli recA promoter fused to luxCDABE(Vollmer et al. 1997)
E. coli RFM443 pKatG::luxCDABEE. coli katG promoter fused to luxCDABE(Belkin et al. 1996))
E. coli RFM443 pFabA::luxCDABEE. coli fabA promoter fused to luxCDABE(Belkin et al. 1997)

Device description and operating conditions

A lab-scale model system incorporating a commercially available device was used in this work. Figure 1 depicts the arrangement of the aspirator, venturi and ultrasound emitters. The unit is a Sonoxide™ B15 model (Ashland Water Technologies, Wilmington, DE, USA) containing six ultrasound emitters that operate at a frequency of 1·65 MHz and a power output of 10 watts cm−2. Total system volume is approximately 17 l, and the flow rate is 52 l min−1, powered by a 1 HP pump. Air is aspirated into the system through the venturi, such that 20% of the volume of the fluid traversing the emitter chamber is air. Cells transit the unit approximately three times every minute, and the unit was in continuous operation throughout the course of the experiment. The water used in the experiments was autoclaved tap water cooled to room temperature. The autoclaved water had a total hardness of 223 mg l−1 (as CaCo3), a pH range of 7·1–7·3 and was free of chlorine as determined by a Hach test. Temperature was maintained at 25°C with a recirculating heater/chiller.

Viable counts

Viable cells were counted using an automated adaptation of the Most Probable Number (MPN) procedure of (Halvorson and Ziegler 1933) on a Biomek 3000 Workstation (Beckman-Coulter, Carlsbad, CA, USA) in 96-well microplates. Samples of 20 µl were diluted 1 : 10 into microplate wells in column 1 containing 180 µl of trypticase soy broth (TSB). The procedure was performed in quadruplicate. The samples were diluted 1 : 10 across wells two through ten. After incubation overnight at 28°C, wells were scored visually for growth/no growth, and the data were entered into an Excel spreadsheet containing an algorithm for translating wells of growth into the Most Probable Number of bacteria in the undiluted samples along with its 95% confidence interval values.

Growth inhibition

Samples removed from the sump or control flask after 30 or 60 min of treatment were diluted 1 : 2 into M9YG media, and 280 µl aliquots were dispensed into the wells of a 96-well microplate in quadruplicate. The plate was placed into a microplate reader, and absorbance readings at 540 nm were taken every 20 min for 24 h. Recovery from lag times was measured as the time at which the absorbance readings reached a value of 0·1.


Respiration of the bacterial populations was performed using OxiTop bottles (Wissenschaftlich-Technische Werkstätten GmbH, Weilheim, Germany) according to the manufacturer's instructions. Samples of 120 ml were removed from the sump or control flask at 30-min intervals for up to 120 min of treatment and diluted 1 : 2 into either M9YG or TSB media and placed into the OxiTop bottles. The bottles were sealed with the manometric measuring head, and oxygen concentration readings were taken every 4 min for 24 h. The recovery period after lag is defined as the time at which the graph of oxygen consumption passed the zero point.

Membrane damage

Membrane-damaged cells were assayed with the LiveDead fluorescent stain (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. The two dyes (Syto9 and propidium iodide) were mixed in equal volumes, and 1 ml of cell suspension was treated with 3 µl of the mixture. A 7·5 µl aliquot was placed on a microscope slide and covered with a cover slip; the sample was examined at 1000× magnification with oil immersion. Green and red cells were enumerated, with a sample consisting of 100–125 cells. Each slide was sampled three times, and the percentages of red cells were calculated.

Biofilm-forming assay

The ability of cells to colonize a surface and form biofilms was measured as follows: Samples were removed from the sump and control flask after 30 min of exposure and 180 µl aliquots placed in the well of a 96-well microplate. Treated and untreated samples contained ~1·6 × 106 cfu ml−1. Sampling was replicated eight times, as were control wells with no cells. The plates were incubated for 30 min to allow for attachment of cells to the well surface. Using the Biomek 3000 Workstation, the wells were washed with physiologically buffered saline (PBS) three times to remove unattached cells using a pump speed of 40 µl s−1 to provide a slow controlled removal and dispensing of media to avoid dislodging attached cells. After the third wash step, 180 µl of fresh M9YG was added to the wells, and the plates were incubated at 28°C for 2 h to amplify the attached biomass. After the incubation period, the attached biomass was measured using a modification of the method described in (Peeters et al. 2008): wells were washed three times with PBS, and then, 0·01% crystal violet solution was added to the wells to stain attached biomass. After 15-min incubation, the wells were washed three times with PBS, and 180 µl of 95% ethanol was added to the plates to solubilize the bound crystal violet from the attached biomass. The amount of crystal violet in the wells was quantified by measuring the absorbance at 540 nm using a Tecan M200 Pro Infinity (Tecan Group Ltd., Mannedorf, Switzerland) microplate reader.

Stress-response reporter strain assay

The nature of the damage inflicted upon the cells was indicated by the response of the E. coli reporter strains listed in Table 1. The regulatory element of an E. coli stress-response gene is fused to the structural genes that produce a functional luciferase, and thus emits light when the stress-response regulatory element is activated by its cognate stress. The fabA construct responds to membrane damage, the recA construct responds to DNA damage, grpE to protein damage and katG responds to the presence of oxygen radicals. The E. coli reporter strains were cultivated overnight in M9YG supplemented with 50 µg ml−1 kanamycin. The overnight cultures were diluted 1 : 2 with fresh M9YG media with 50 µg ml−1 kanamycin 2 h prior to treatment. Fifty millilitre cultures were inoculated into the sump and stirred gently with a pipette to mix thoroughly prior to turning on the device. At intervals, samples were removed from the sump and the control flask, and 280 µl aliquots were dispensed to a white 96-well microplate. Luminescence was measured in a Tecan M200 Pro Infinity plate reader.

Statistical analysis

Data were analysed using Minitab 15 Statistical Software (Minitab, Inc., State College, PA, USA).


Effect of exposure on viability

Pseudomonas putida was diluted into water in the sump of the bench-top model to an initial concentration of 1·6 × 106 ± 6·9 × 105 cfu ml−1. The population was treated for 120 min, with samples being removed every 30 min for viability determination by MPN. Figure 2 shows the results of the experiment. A two-sample t-test demonstrated no statistically significant difference in population levels between the control and treated populations at any time over the 120-min time course.

Figure 2.

Time course of population levels of Pseudomonas putida suspended in buffer with (□) and without treatments (■). Values are the average of six to eleven determinations plus and/or minus the standard deviations.

Impact of treatment on resumption of growth

Suspensions of Ps. putida were treated and then diluted 1 : 2 into either M9YG or TSB. Control aliquots were removed prior to the initiation of treatment, and treated aliquots were removed after 30 and 60 min of treatment. Eight samples of each aliquot were dispensed to microplates, and growth was measured by determining the increase in the absorbance of the suspension at 540 nm using a microplate reader. The length of the lag phase is defined as the time it takes for the cultures to reach an absorbance of 0·1.

Typical responses of control and treated samples are depicted in Fig. 3a (M9YG) and Fig. 3b (TSB) with the values plotted being the average of eight samples from a single treatment. The average responses from multiple experiments are summarized in Table 2. Lag phases of control cultures are longer in M9YG then in TSB. Treatment of the cultures for either 30 or 60 min produced similar (statistically indistinguishable) lengths of lag phases in both media. The lag phases are longer in each treated condition compared with the untreated control but approach statistical significance (two-tailed test, < 0·01) only in M9YG media. However, a one-tailed t-test comparing the lag period differential between control and treated conditions to the hypothetical mean of zero (expected if no difference exists between treated and control) indicates each differential is highly significant and confirms treated populations have longer lag phases than control populations.

Table 2. Length of growth lag periods in hours, with and without treatment, and the lag period differential
MediaExposureControlTreatedDifferential N P, two-tailedP, one-tailed
  1. Values are the average plus and minus the standard deviation of N experiments.

  2. The ‘P, two-tailed’ value is for the comparison of the control and treated values, and the ‘P, one-tailed’ value is for the comparison of the difference in the lag period values with the hypothetical mean of zero.

TSB303·9 ± 1·75·7 ± 2·11·7 ± 0·450·1810·001
603·3 ± 1·05·3 ± 1·82·1 ± 0·940·0910·017
M9YG309·3 ± 2·012·7 ± 1·83·4 ± 0·450·0220·000
608·5 ± 1·212·0 ± 0·53·5 ± 1·040·0020·007
Figure 3.

Resumption of growth by control (■) and treated (□) cultures diluted into TSB (a) and M9YG (b) after a 30-min exposure. Graphs are representative of six trials.

Impact of treatment on resumption of respiration

Samples withdrawn at the same time, and from the same location, as those described in the section above were diluted 1 : 2 into M9YG, and their rates of oxygen consumption measured. The results of these experiments are reported in Table 3. In all cases, treated samples resumed respiration later than untreated control samples. Statistical analysis (two-sample t-test) indicated the delay in recovery of respiration was significant (< 0·01). Testing the significance of the difference in the length of the delay against the hypothesized length of delay amongst the control samples (zero) using a one-tailed t-test indicates statistical significance at the 30- and 60-min exposures but not the 90-min exposure. Failure to demonstrate significance at 90 min is probably due to the large standard deviation for those samples (1·07 h). Nevertheless, the preponderance of data supports the contention that the onset of recovery of respiration is delayed in treated samples.

Table 3. Length of respiration lag periods in hours, with and without treatment, and the lag period differentials for Pseudomonas putida in M9YG
ExposureControlTreatedDifferential N P, two-tailedP, one-tailed
  1. The ‘P, two-tailed’ value is for the comparison of the control and treated values, and the ‘P, one-tailed’ value is for the comparison of the differential in the lag period values with the hypothetical mean of zero.

302·63 ± 0·834·76 ± 0·822·13 ± 0·6540·010·007
602·48 ± 0·685·56 ± 0·423·08 ± 0·4930·0030·008
902·83 ± 0·825·55 ± 0·522·72 ± 1·0730·0080·048

Membrane integrity in treated samples

The fluorescent dyes Syto9 and propidium iodide (PI) were applied to treated and untreated control populations to visualize cells with damaged membranes. Cells with intact membranes exclude propidium iodide and fluoresce green while cells with damaged membranes allow entry of the propidium iodide, which binds to DNA and fluoresces red.

Cells treated in the system for 0–120 min were stained with the dyes, and the number of red and green cells was counted manually. An average of 740 cells per experiment was counted in the control samples, and an average of 603 cells was counted in treated samples.

Analysis of the data indicated that numbers and ratios of red to green cells from 30 min of exposure were identical to those obtained from 120-min exposures; the data from the different time points were pooled and analysed as such. This lack of a difference between 30- and 120-min samples was evident in both control and treated sample pools, indicating control cells suspended in buffer did not undergo membrane stress during the course of the experiment. Control samples had an average of 2·2 ± 1·1% red cells, and treated samples had an average of 16·5 ± 11·1% red cells. The ratio of red cells in treated samples compared with untreated samples was 7·1 ± 3·5. A two-sample t-test comparing the percentage of red cells in untreated vs treated populations yields a P value of 0·002, indicating a significant difference in spite of the high variability in the treated population. Analysing the ratio of red cells in treated samples with a one-sample t-test and using a hypothetical mean ratio of 1·0 (expected if comparing untreated cells to untreated cells) yields a P value of less than 0·000, a highly significant result.

Ability of treated cells to form biofilms

The ability of cells to form a biofilm was measured in a simple assay in which cells were allowed to settle on a surface for a short period, the cells remaining in suspension removed, and the settled cells grown for a short time to amplify the signal of their biomass. Multiple samples (40 replicates) of untreated control and treated populations were compared in three experiments.

Table 4 shows the results of this experiment. After subtracting the value of blank samples, which had no biofilms, the treated cultures produced biofilms with 26% of the absorbance of the untreated control biofilms. A two-sample t-test comparing the biofilms derived from treated and untreated populations yields a P value of 0·000, indicating a highly significant result (the statistical programme will not calculate a P value smaller than 0·000). The high background absorbance values (52% of the treated value) were a cause of concern, but comparison of the two values yields a P value of 0·000, indicating they are indeed different from each other.

Table 4. Biofilm-forming ability of untreated and treated Pseudomonas putida in the microplate assay
 AbsorbanceStandard deviation N
  1. The values are the average and standard deviation calculated from N samples.


Induction of bacterial repair systems in treated populations

Escherichia coli strains in which the luxBCDE (luciferase) genes were fused to the regulatory elements for the fabA, grpE, recA and katG stress-response genes (responsive to membrane damage, protein damage, DNA damage and radical-mediated damage, respectively) were treated for 30 min, and their response was monitored in a microplate luminometer. Unexposed control cultures were also monitored. Figure 4 reports the results of these experiments.

Figure 4.

Response of luxBC reporter strains to treatment showing increase in relative light units (RLU). Stress-response promoters fused with luxBCDE are fabA (♦), katG (▲), grpE (●), and recA (■). The response of the fabA construct (□) to untreated conditions is shown for comparison; none of the other control strains had more than 3% of the treated condition RLU at 90 min. Values are the average of three replicates plus and/or minus the standard deviations.

The most immediate response was observed with the fabA reporter strain; it was also the largest response observed. The fabA control had a moderate induction upon transfer to the media, but this soon declined to baseline levels. The next largest response was observed with the recA reporter strain, and its control culture had a negligible response. The onset of its response was also delayed compared with that of fabA reporter strain. The onset of the grpE reporter strain response was slower still and reached barely one-sixth the level of the fabA reporter strain at its maximum. The katG reporter strain barely responded, although its luminescence values were higher than its controls.

The reporter strain data suggest that membrane damage is a primary effect of exposure to the device, but that elements of DNA and protein damage are eventually evident. However, the grpE response system in particular has overlapping elements with other response systems, and severe enough damage to one cellular component (e.g. membranes) can cause a cascading response through the interwoven pathways of the cell that may eventually manifest itself in activation of stress-response pathways that are mechanistically distant from the primary damage. Direct evidence of protein damage, for instance increased rates of protein turnover, would provide supporting evidence. Free radicals do not seem to be a major element of treatment stress; the literature indicates that radicals are a more prominent mechanism with high energy, low frequency ultrasonic cavitation than with the low energy, high frequency ultrasound employed in this device (Arrojo et al. 2007).


The desire to be environmentally responsible is a significant driver in the decision to utilize nonchemical means of controlling microbial growth in industrial applications. However, balancing this desire with demonstrable efficacy is still a requirement for cost-effective decision-making. In addition to fulfilling this objective, establishing a mechanism of action for any given microbial control technology provides a rational basis for attempts to improve the technology and yields insight into the application of the technology and measurement of its efficacy.

While the mechanisms of action of some classes of biocides are well defined, a similar level of understanding for nonchemical microbial control technologies, apart from ultraviolet light, is not widely available. This work is an attempt to determine the impact of coupled hydrodynamic and ultrasound-induced stress on a defined population at the cellular level.

One of the first observations of treated populations is they do not exhibit a measureable reduction in viability, which is a marked contrast to chemical microbial control technologies. While bacteriostatic mechanisms are familiar in the antibiotic world, the authors are unaware of any other commercial microbial control technology that utilizes a bacteriostatic mechanism. It is possible that some cells are dying in the course of exposure to the device, but this number does not rise to the level of statistical significance or by itself provide evidence of a dose–response trend. Thus, it is likely the main effect of treatment is bacteriostatic rather than bactericidal.

Treated cells demonstrate a requirement for an increased lag time to resume both growth and respiration than untreated cells upon transfer into fresh media. Bacteria transferred from one environment to another typically show a lag in their growth that is interpreted as the time required to sense and respond appropriately to the new nutritional and environmental conditions. Our interpretation is the increase in the lag times beyond those observed in untreated populations represents time needed to repair damage before the cell can expend resources on adjusting to its new environment and resume growth. This increased lag is observed when measuring both growth and respiration. Both processes are the result of the function of a number of subcellular processes, and these experiments cannot resolve which process or processes are damaged.

Direct evidence of a specific type of cell damage is provided by the increase in the number of membrane-damaged cells in treated populations as observed using the pair of fluorescent dyes, Syto 9 and propidium iodide. Cells with damaged membranes are permeable to propidium iodide, and the increase in the percentage of red cells in treated populations indicates membrane damage. These red-fluorescing cells probably represent only those cells with the most severe membrane damage; the actual fraction of cells in the population with damaged membranes is likely higher. The extent of membrane damage required for a red-fluorescing response is unknown, and while the correlation of the extent of membrane damage required for cell death or growth inhibition is of interest, these experiments cannot inform on that topic.

Further evidence of the nature of damage inflicted upon the cells is indicated by the response of the lux reporter strains to treatment. A previous study (Vollmer 1998) with these strains examined their response to high frequency (1 MHz) and high power (500 W cm−2) ultrasound and determined the grpE construct (monitoring heat shock and protein damage) had the greatest response, followed by the fabA reporter strain, and then smaller equivalent responses from the katG and recA reporters. This system, using higher frequency (1·65 MHz) and lower power (10 W cm−2), produced the largest and most immediate response with the fabA construct followed by lower and slower responses from the recA and grpE reporters. A very minimal response was observed from the katG reporter strain. The higher power and lower frequency system is much more likely to produce true cavitation, which is characterized by the generation and collapse of microbubbles producing localized high temperatures and free radicals. The much lower power employed in this device is likely to be insufficient to generate cavitation. The results reported here support the idea that membrane damage caused by turbulence and shear stress is an early and significant event in the antibacterial mechanism of the device. Secondary events may include protein and DNA damage as suggested by the timing of the onset of the response of the grpE and recA reporter strains.

The reduction in the ability of treated cells to form biofilms may be related to inhibition of growth. Cells settle on a surface irrespective of their health because this is thought to be a simple stochastic process. However, making the transition from a cell settling reversibly onto a surface to one that is encased in an extracellular polymeric matrix in a nascent biofilm requires the expenditure of energy on the part of the cell, which is not available if it is engaged in repairing damage. Alternatively, surface structures or processes required to establish biofilms may be directly damaged by treatment with the device.

This microbial control device is currently deployed in cooling towers as a side-stream unit that treats the system volume once every 3–4 h, which coincides well with the duration of the experimentally observed growth and respiration lags in our model system. Thus, it is conceivable that the microbial population in a treated cooling tower is in a static, nongrowing state. If the incoming microbial population is maintained at a concentration below that in the tower bulk fluid, the treated population in the tower will eventually be reduced through blowdown, which is the term applied to water removed from the tower periodically to help control water chemistry. However, the presence of biofilms in a treated system is problematic in that this population is not directly treated by the flow-thru device. Paradoxically, substantial anecdotal evidence indicates high biofilm populations are not found in towers treated with this device. These devices are currently employed in hundreds of cooling towers throughout Europe and North America. This work does not directly address biofilm control, and this remains an unexplored and potentially productive research topic. It is possible to speculate about potential linkages between treated planktonic populations and their impact on biofilms. One theory is the dynamic equilibrium between recruitment of cells into the biofilm, and biofilm dispersion is altered to favour dispersion. This may involve the reduced ability of treated cells to form biofilms (observed in this work), or it may be that treated cells produce molecules that trigger dispersal of biofilms by signalling the presence of stressing conditions, much like the cis-2-decenoic signalling system described by Davies (Davies and Marques 2009).

Speculation aside, though, this work demonstrates that bacteriostasis is an effective microbial control strategy in certain situations, and this can be successfully achieved by nonchemical devices. Detailed examination of other existing and proposed devices is required to determine the extent to which bacteriostasis is involved as a mechanism and to carefully determine the situations in which they are useful.


Thanks are due to Dr Melvin Czechowski and Olaf Pohlmann who provided valuable insight and helpful discussion. We are also grateful to Dr Amy Vollmer for her kind gift of the reporter strains. Rebecca Collins provided able technical assistance. The authors of this work are employees of Ashland Water Technologies, the manufacturer of Sonoxide™, the microbial control device described in this work.