PLoS ONE plos plosone PLOS ONE 1932-6203 Public Library of Science San Francisco, CA USA 10.1371/journal.pone.0247910 PONE-D-20-04971 Research Article Physical sciencesChemistryChemical compoundsCarbon dioxide Earth sciencesHydrologySurface water Ecology and environmental sciencesPollutionWater pollution Earth sciencesMarine and aquatic sciencesBodies of waterRivers Ecology and environmental sciencesAquatic environmentsFreshwater environmentsRivers Earth sciencesMarine and aquatic sciencesAquatic environmentsFreshwater environmentsRivers Ecology and environmental sciencesWater quality Ecology and environmental sciencesNatural resourcesWater resources Physical sciencesPhysicsClassical mechanicsPressurePartial pressure Biology and life sciencesOrganismsEukaryotaAnimalsVertebratesAmniotesBirdsCanaries Biology and life sciencesZoologyAnimalsVertebratesAmniotesBirdsCanaries Canary in the coliform mine: Exploring the industrial application limits of a microbial respiration alarm system Industrial application limits of a microbial respiration alarm system https://orcid.org/0000-0001-9126-6362 Stone Wendy Data curation Formal analysis Investigation Methodology Project administration Resources Validation Visualization Writing – original draft 1 * https://orcid.org/0000-0002-4166-0799 Louw Tobi M. Formal analysis Methodology Validation Writing – review & editing 2 https://orcid.org/0000-0003-2817-2069 Booysen Marthinus J. Methodology Resources Software Writing – review & editing 3 Wolfaardt Gideon M. Conceptualization Funding acquisition Supervision Writing – review & editing 1 4 Water Institute and Department of Microbiology, Stellenbosch University, Stellenbosch, South Africa Department of Process Engineering, Stellenbosch University, Stellenbosch, South Africa Department of E&E Engineering, Stellenbosch University, Stellenbosch, South Africa Department of Chemistry and Biology, Ryerson University, Toronto, Canada Zhang Dawei Editor Chinese Academy of Sciences, CHINA

The authors have declared that no competing interests exist.

 * E-mail: wstone@sun.ac.za 4 3 2021 2021 16 3 e0247910 20 2 2020 16 2 2021 2021 Stone et al This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Fundamental ecological principles of ecosystem-level respiration are extensively applied in greenhouse gas and elemental cycle studies. A laboratory system termed CEMS (Carbon Dioxide Evolution Measurement System), developed to explore microbial biofilm growth and metabolic responses, was evaluated as an early-warning system for microbial disturbances in industrial settings: in (a) potable water system contamination, and (b) bioreactor inhibition. Respiration was detected as CO2 production, rather than O2 consumption, including aerobic and anaerobic metabolism. Design, thresholds, and benefits of the remote CO2 monitoring technology were described. Headspace CO2 correlated with contamination levels, as well as chemical (R2 > 0.83–0.96) and microbiological water quality indicators (R2 > 0.78–0.88). Detection thresholds were limiting factors in monitoring drinking water to national and international standards (0 CFU/100 mL fecal coliforms) in both open- (>1500 CFU/mL) and closed-loop CO2 measuring regimes (>100 CFU/100 mL). However, closed-loop detection thresholds allow for the detection of significant contamination events, and monitoring less stringent systems such as irrigation water (<100 CFU/mL). Whole-system respiration was effectively harnessed as an early-warning system in bioreactor performance monitoring. Models were used to deconvolute biological CO2 fluctuations from chemical CO2 dynamics, to optimize this real-time, sustainable, low-waste technology, facilitating timeous responses to biological disturbances in bioreactors.

 http://dx.doi.org/10.13039/501100007601 Horizon 2020 689925 Wolfaardt Gideon M. GMW and WS were funded by the European Union’s Horizon 2020 research and innovation program (https://ec.europa.eu/programmes/horizon2020/en), grant agreement No 689925. The study reflects only the authors’ views. The EU is not responsible for any use that may be made of the information it contains. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of manuscript. Data Availability All relevant data are publicly accessible via Kaggle (https://www.kaggle.com/wendystone/a-microbial-respiration-alarm-system). Introduction

Carbon dioxide production is a universal biological indicator of respiration, and is a parameter indicative of life, often harnessed as an indicator of ecosystem health [1, 2]. Pursuing a similar bird’s-eye-view, Kroukamp and Wolfaardt [3] developed the Carbon Dioxide Evolution Measurement System (CEMS), which harnesses microbial CO2 production to study whole-biofilm metabolic profiles. Unlike standard respirometry [4], CO2 rather than O2 is monitored as the by-product of glycolysis and the Krebs cycle. This metabolic pathway is common to aerobic and anaerobic metabolism, and the measurement of CO2 fluctuation is also indicative of photosynthetic metabolism. The system traps biofilm-produced CO2, passing it over an analyzer on CO2-free sweeper gas, and logging the data. Online platforms allow for remote, real-time monitoring of disturbances. It is used primarily to study pure culture laboratory biofilms, and has generated information on the relationship between biofilm inoculum, nutrients and metabolism [5, 6]; metabolic responses to antibiotic treatments [7, 8] and to track cellulolytic activity [9]. This design has also been adapted to a closed-loop CO2 accumulation system to monitor low-level microbial metabolic patterns during desiccation [10].

Understanding microbial activity at community and species levels does require the genetic, proteomic and metabolic profiling of the specific active and dormant organisms within an ecosystem. However, the traditional narrow focus on pathogenic species in water quality monitoring may well contribute to the many failures in detecting outbreaks, as indigenous communities could mask pathogens in our efforts to detect them. There are practical benefits and unique perspectives facilitated by the immediate and high-level availability of whole-ecosystem metabolic footprints. This is particularly true when coupled with modelling, which can predict many of the physico-chemical CO2 fluctuations, resolving the biological data from total CO2 information.

We evaluate the potential of using CEMS as an alarm in commercial and industrial settings, analogous to the miner’s canary. The technology is assessed as an indicator in two similar but converse undesirable industrial scenarios: (1) the contamination of potable water systems, and (2) the inhibition of bioreactors for the treatment of waste- or industrial service water. The second scenario is especially relevant, since failure of wastewater treatment systems can result in serious environmental pollution, due to unexpected inflow of chemicals that inhibit growth, or careless incorporation of industrial waste streams without taking biodegradability into account. For instance, wastewater loading impact is commonly calculated based on COD rather than contaminant toxicity.

Any fluctuations of CO2 production above a pre-determined minimum respiration rate can act as a high-level indicator of microbial contamination events in potable water systems, providing maintenance staff a warning to investigate more fully. Whether water is stored in a tank or distributed in pipes in systems such as dental chairs, microbial contamination of potable water is often the cause of disease outbreaks [11, 12]. With growing water scarcity challenges in drought-ridden areas, and the associated popularity of decentralized water systems gaining traction particularly in developing countries, remote monitoring for microbial contamination is paramount. Moreover, the microbiological monitoring of drinking water is limited by time (culturing), cost and expertise (quantitative real-time PCR, enzyme assays), which hampers the public provision of decentralized point-of-use water treatment systems. Most of these techniques involve the mass use of disposable reagents and equipment, a critical consideration in the current environmental waste crisis. CEMS has the benefit of real-time and remote monitoring of microbial activity, with equipment that is long-lasting and energetically sustainable using solar power, generally liberally available in drought-ridden areas.

In the converse scenario, in the biological treatment of sewage and industrial wastewater, microbial activity is responsible for converting dissolved pollutants into the gas phase, solid phase for removal as biomass, or into more innocuous chemical forms. Their metabolic activity can simultaneously act as a reliable indicator of changes in the waste treatment process, including changes in chemical composition (nutrient spikes, toxins, pH), temperature and failure of mixing (oxygenation vs anaerobiosis). Optimal functionality should be associated with a quantifiable CO2 steady-state, and predictable fluctuation thresholds. Companies that utilize waste treatment bioreactors can be fined extensively if the reactors fail and toxic waste is released into water sources, recorded in governmental non-compliance records [13–15]. These fines are proportional to waste volume, and an immediate, durable indicator of bioreactor failure due to the disruption of microbial metabolism is both environmentally and financially valuable.

In this study, the thresholds of this CO2 monitoring technology were shown to limit use as an early-warning system for potable water, but were sufficient to meet agricultural water standards. The system was an effective alarm for bioreactor disturbances. Modelling was employed to tease out physico-chemical CO2 fluctuations from the metabolic response of the microbial aggregate, which is the active role player in waste remediation.

 Materials and methods Sampling: River water and wastewater

The CEMS performance was evaluated in two different contexts: for the monitoring of clean water, which should ideally contain very low levels of microbial contamination, and the monitoring of an active bioreactor. To assess the former, river water samples were collected from the polluted Plankenbrug River, Stellenbosch, South Africa (-33.933983; 18.85095) [16]. This polluted river water was diluted with pure water to mimic a range of pollution levels, from low to high. For the latter, Return Activated Sludge (RAS) samples were collected from a municipal wastewater treatment works in Cape Town, South Africa.

 Canary CEMS for potable water storage systems

Sterile 5 L Erlenmeyer flasks were filled with river water, mixed continuously to ensure aeration, and no additional nutrients added. The flasks were sealed and the CEMS tube for headspace gaseous analysis was placed directly above the liquid surface. A small air inlet port prevented negative pressure. The headspace gas was transferred via a peristaltic pump (flow rate = 15 mL/min; carrier gas = ambient air) through a LiCor CO2 analyzer (Campbell Scientific, Stellenbosch, South Africa), with continuous data logging per second (Fig 1 –open loop). Thus, CO2 measurements were investigated as a difference between atmospheric and reactor CO2 levels, rather than absolute CO2 production, as measured in CO2-free air in previously described CEMS studies [5]. Atmospheric baselines were measured repeatedly on different days (10 days, minimum of 5 hrs per baseline), to assess variation. Two sensors were used throughout the experiment, calibrated every 3–7 days with a CO2 gas standard (AFROX, Cape Town, South Africa), and randomly exchanged between control and experimental reactors throughout the study. The random sensor exchange was to monitor potential sensor drift and for normalizing, preventing the constant exposure of each sensor to either high or low CO2 concentrations. Averages and standard deviations were calculated in Microsoft Excel for Mac (v 15.25), throughout the study.

10.1371/journal.pone.0247910.g001 Schematic of the laboratory-scale canary CEMS alarm system.

To assess minimum thresholds for whole-reactor CO2 monitoring, as an indicator of (1) water storage contamination events and (2) metabolic disturbances of bioreactors.

 Headspace effect

The effect of headspace volume on CEMS sensitivity was evaluated at different volumes: (1) 0.9 L headspace (5.0 L liquid), (2) 0.55 L headspace (5.35 L liquid) and (3) 0.25 L headspace (5.65 L liquid), as described in Table A1 in the S1 Appendix. In order to assess lower detection thresholds, both undiluted and 25% samples (v river water/v sterile tap water) were evaluated. The average total microbial CO2 production [ppm CO2 –ppm CO2 (ambient control)] was plotted against headspace volume. Liquid surface area was recorded per treatment. The control was an identical reactor containing sterile (autoclaved, 121 °C, 15 psi, 30 min) river water. All further measurements were assessed at 0.9 L headspace (5.0 L liquid). The differences between means were compared individually at decreasing headspaces, between (1) & (2), and (2) & (3), for diluted and undiluted reactors.

 Minimum detection threshold

In order to correlate CO2 accumulation and SANS241 parameters [17], headspace accumulation of CO2 in the river water reactors was measured (0.9 L headspace), along with the chemical and microbiological profiles (described below). To determine minimum thresholds, river water samples were subsequently diluted with sterile tap water (0%, 1%, 10%, 25%, 50%, 100% (v river water/v sterile tap water)), representing an increasingly contaminated potable water system, and evaluated for the same metabolic, chemical and microbiological profiles. Measurements were done for separate, triplicate experiments. These were assessed for linear correlations in Microsoft Excel for Mac (v 15.25) goodness-of-fit reported as R2, assessed for significance at a 95% fit. These were plotted alongside the South African water quality guideline’s chemical and microbiological limits for application in monitoring potable water storage [17]. National guidelines are based on the World Health Organization guidelines [18], and comparable to those of leading countries like Canada [19].

 Chemical profile of river water samples

Chemical Oxygen Demand (Spectroquant COD Cell Test, 1.14541.0001), Nitrogen (Spectroquant Nitrogen (total N) Cell Test, 1.00613.0001), Ammonium (Spectroquant Ammonium NH4+ Test, 1.14752.0001), Sulphate (Spectroquant Sulfate SO42- Cell Test, 1.14548.0001) and Phosphate (Spectroquant Phosphate Test, 1.14842.0001) were measured colorimetrically using a Merck Spectroquant Pharo300 photometer, according to the manufacturer’s instructions. Samples were heated in a Spectroquant TR320 Thermocycler, thoroughly mixed by inversion before analyzing and filtered with cellulose nitrate, 1.2 μm pore size filters (Sartorius Stedim Biotech, South Africa). Spectroquant kits and apparatus were sourced from Merck (Modderfontein, South Africa). Correlations between chemical profiles and headspace detection of CO2 were determined in triplicate against linear regression models in Microsoft Excel for Mac (v 15.25), and goodness-of-fit reported as R2, assessed for significance at a 95% fit.

 Microbiological profile of river water samples

Standard Heterotrophic Plate Counts were conducted (APHA Method 9215) on Standard Plate Count media (Yeast Extract, 2.5 g/L; Pancreatic Digest of Casein, 5.0 g; Glucose, 1.0 g/L; Agar, 15 g/L; pH 7.0; 26°C). Total coliform counts were assessed on EndoAgar (Merck, APHA Filtration Method 9222; 37°C). Gram negative enteric bacteria were assessed on MacConkey Agar (APHA Filtration Method 9222; 37°C). In addition to filtration, where bacterial loads were higher, reactor samples were diluted in Physiological Saline Solution (9 g/L NaCl), and enumerated on Standard Plate Count, Endo and MacConkey Agar. Chemicals were purchased from Sigma-Aldrich (Johannesburg, South Africa). Correlations between microbiological profiles and headspace detection of CO2 were determined in triplicate against linear regression models in Microsoft Excel for Mac (v 15.25), and goodness-of-fit reported as R2, assessed for significance at a 95% fit.

 ATP measurements

ATP [Relative Light Units (RLU)/50 μL] was measured as a confirmation of metabolic profiles. River water was added (50 μL) to Hygenia Ultrasnap ATP surface swabs (Fischer Scientific, Ottawa, Canada), shaken for 15 s and measured in a Hygenia EnSURE luminometer according to manufacturer’s instructions. These are used to measure microbial surface contamination in the food and hospital industry [20], but were harnessed here as a simple tool for confirming metabolic activity in a river water dilution series. Dispersion of flocs was critical for unbiased readings. Samples (1.5 mL) were vortexed (2 minutes) prior to analysis. Correlations between metabolic profiles and headspace detection of CO2 were determined in triplicate against linear regression models in Microsoft Excel for Mac (v 15.25), and goodness-of-fit reported as R2, assessed for significance at a 95% fit.

 Closed-loop design

To investigate lower detection thresholds, the system was redesigned to accumulate CO2 over time (Fig 1 –closed loop), by circulating the reactor headspace continuously over the CO2 analyzer. An overnight culture of E. coli (3 g/L Tryptic Soy Broth, TSB, 26 °C, rotary shaker) was diluted (sterile tap water) and inoculated into a 5 L Erlenmeyer flask containing autoclaved tap water, at final concentrations of 101, 102 and 103 CFU/100 mL, with continuous stirring. Overnight E. coli concentrations were pre-determined with growth curves at room temperature rather than 37 °C in TSB, to prevent temperature variation upon transfer. Final cell concentrations were checked by extraction from the sampling port, and counted on Tryptic Soy Agar using filtration (TSA, 3 g/L TSB, 15 g/L agar), as described above for total heterotrophic analysis. The accumulation of CO2 over time was measured at each cell concentration, before the cell concentration was increased via the injection port.

 Canary CEMS for active bioreactor disturbances

For RAS bioreactors, the flasks were filled (4.5 L) with synthetic wastewater and sterilized by autoclaving. The reactors were inoculated with 0.5 L activated sludge, corresponding to a liquid volume of 5L and a headspace volume of 0.9 L, and allowed to equilibrate to steady state CO2 production with constant stirring. One of the two reactors was sterilized in the autoclave after the addition of RAS and maintained under identical conditions as an abiotic control. The reactors were treated with identical environmental stressors. These chemical and physical disturbances included (1) chlorine [ChlorGuard commercial bleach; final reactor concentration 15% (v/v)], (2) acidification [(a) pH 6.6 dropped to pH 5.0, ~200 μL concentrated (38%) HCl and (b) pH 6.6 dropped to pH 3.3, ~600 μL concentrated HCl)] and (3) temperature fluctuations [23°C dropped to 16°C; 5% (v/v) ice, in a 4 °C refrigerator]. The chlorine treatment was repeated, once with reactor mixing (aerated) and once without (sedimentary). For the pH disturbances, soluble CO2 was measured pre- and post-acidification with a MerckMillipore Carbon Dioxide Titrimetric MCarbon Test (Modderfontein, South Africa), according to manufacturer’s instructions. A Jenway 3510 pH Meter, calibrated using the 3 buffer method, was used to monitor pH. Titrations were performed and data collected using a HI902 Auto-titrator (Hanna Instruments®) equipped with a HI1131 pH probe (Hanna Instruments®). The instrument was pH calibrated beforehand using standards pH 4, pH 7 and pH 10 (Merck Millipore). Up titrations were performed to endpoint pH 11 using standardized 0.1 N NaOH after which subsequent down titrations were performed with standardized 0.1 N HCl to endpoint pH 5.

For all treatments, cell concentrations were measured pre- and post-disturbance. Standard plate counts were assessed with dilutions on Tryptic Soy Agar (Sigma Aldrich, Johannesburg, South Africa).

Mathematical modelling of physico-chemical and microbiological CO<sub>2</sub> release

A dynamic mathematical model (the details of which can be found in S1 Appendix) was used to discriminate between physico-chemical- and microbiological CO2 release from the liquid. A fundamental assumption of this model was that the microbiological CO2 production was constant. While this assumption may appear limiting, it has little effect on the results interpretation, as demonstrated shortly.

The model accounted for microbiologically produced CO2, speciation into bicarbonate (HCO3- and carbonate CO32-), and CO2 mass transfer between the liquid- and gas phases. The model was implemented using MATLAB R2018a (The MathWorks, Inc., Natick, MA).

All model parameters are provided in Table A1 in S1 Appendix. Two parameters were unknown: the microbiological CO2 production rate (r˙CO2) and the liquid-air mass transfer coefficient (KLA). Both were determined by regression of experimentally measured gaseous CO2 concentrations to model predictions.

 Threshold alarm software

Remote alarm functionality was achieved by developing an external software application in the Python programming language to intercept the data packets on the LiCor LI-820 CO2 analyzer’s serial port. This allowed users to specify an upper and lower threshold, and to configure an email message, sent when serial port readings surpassed any of the pre-determined thresholds.

 Results

The measurement of microbial CO2 production for the remote monitoring of water treatment system performance was assessed. Two experimental scenarios were evaluated, on the polar ends of a metabolic spectrum: (a) for the remote monitoring of microbial contamination in potable water systems (where the ideal microbial levels are zero), and (b) for the remote monitoring of steady state microbial activity in bioreactors for waste treatment (where steady-state CO2 production is optimal within a pre-determined window).

Headspace volume and detection sensitivity

Respiration was monitored in reactor headspaces, which were open to the atmosphere via a defined inlet port. Baselines were predictable at 385 ppm CO2 ± 45 ppm for the open-loop system, with variation attributed to human activity in the vicinity increasing CO2 levels during the day. Closed-loop baselines had a similar standard deviation if measured at separate instances over different days, but a variation of less than 5 ppm over days if sealed. Headspace CO2 was allowed to accumulate before open-loop measurement. In river water reactor systems, a smaller headspace decreased the sensitivity of the whole-system CO2 production (Fig 2, Student’s t-Test, p<0.05). This was counter-intuitive, as lower headspace volumes are correlated with greater river water volumes, increasing the total biomass and thus presumably resulting in increased CO2 production. These unexpected results are likely due to two reasons: (1) the system is open to ambient air and a smaller headspace is more rapidly replaced by ambient air drawn into the system, and (2) the conical flasks in the laboratory set-up meant a larger headspace was equivalent to a larger surface area for gas-liquid CO2 exchange. The tension between the influences of these design parameters on the data indicates that case-by-case optimization is necessary for effective industrial CEMS application as an early-warning alarm system for microbial disturbances in water storage systems.

10.1371/journal.pone.0247910.g002 The effect of reactor volume and headspace on (a) the whole-system CO<sub>2</sub> measurement.

The first reactor (1) contained undiluted river water at (A) 5 L with 0.9 L headspace, (B) 5.35 L with 0.55 L headspace, and (C) 5.65 L with 0.25 L headspace. The second reactor (2) contained diluted Plankenbrug river water (25% v/v sterile tap water), at the same volume:headspace ratios. Controls were a separate reactor with undiluted autoclaved Plankenbrug river water. Respiration was compared by subtracting the average CO2 production of the sterile control reactor from the average CO2 production of the reactor over a given time period. Error bars represent standard deviation of measurements per second over 1 hour.

The dynamic model results compare well to the experimentally measured CO2 partial pressure (Fig 3a), but the rate of CO2 production (as determined by regression) was so low in the river water that the regression was insensitive to orders of magnitude variations in r˙CO2 in the open-loop configuration, which is to be expected with only a few cells responsible for the metabolic footprint. Fig 3b shows the variation in the coefficient of determination (1-TSS/RSS, where TSS is the total sum of square errors from the mean, and RSS is the residual sum of squares) as a function of the regressed parameters KLA and r˙CO2. Clearly, the values for r˙CO2 can vary significantly without influencing the fraction of explained variance. This implies that the CEMS device is not suitable for such low CO2 production rates in an open-loop design. However, the combination of modelling and experimentation as performed in this study can be used to indicate the range at which the system can be used, guiding further modification of the system into a closed-loop design as described below.

10.1371/journal.pone.0247910.g003 Modelling of CO<sub>2</sub> limits of detection.

(a) Predicted vs measured values of CO2 partial pressure in the headspace above undiluted river water for varying headspace volumes. CO2 was allowed to accumulate in the headspace prior to turning on the CEMS. (b) Residual Sum of Squares as a function of the regressed parameters r˙CO2 and KLA. At low contamination levels, the rate of CO2 production can vary by multiple orders of magnitude without significantly affecting the measured response, thereby indicating the CEMS limits of detection.

The open-loop CEMS data proved insufficient to accurately predict the rate of CO2 production in systems with low levels of microbial contamination. However, the steady state CO2 concentrations in such systems were still significantly higher than in their abiotic counterparts. This implies that the steady state CO2 concentration as detected by the CEMS can be used as an empirical predictor of water quality. Given the enhanced sensitivity at larger headspace volumes, correlating to larger surface area, a liquid volume of 5 L (corresponding to a headspace volume of 0.90 L) was used for all subsequent experiments.

 Monitoring microbial contamination: Potable water systems

Elevated CO2 concentrations in the reactor headspace was measurable for undiluted polluted river water in an open system. In order to determine the sensitivity thresholds of the system, the river water was diluted (50%, 25%, 10%, 1%, 0% v/v) with sterile tap water, representing an increasingly contaminated water storage system, and the profiles were assessed at each dilution.

chemical (COD, Total-N, NH4+-N, SO4-),

microbiological (standard heterotrophic bacteria, total coliforms, enteric bacteria), and

metabolic (CO2, ATP)

The dilution profile of polluted river water produced a measurable CO2 production trend, which was correlated to a suite of water quality parameters, described in the national water quality guidelines [17]. Adenosine triphosphate (ATP) was included as an independent measure of metabolic activity, detected with fluorometry on swabs typically utilized in surface metabolic monitoring. This technique confirmed the metabolic profile with tight correlation to CO2 data (R2 = 0.832; Fig 4b), but generated high variation at higher bacterial loads, likely due to challenges in microbial floc dispersion. All of the chemical and microbiological parameters also correlated with CO2 production (Fig 4b and 4c), and the point at which the detection system’s correlation with the chemical parameters (total N, NH4-N and sulphate) was useful fell above the minimum thresholds according to South African Water Quality Guidelines [17]. In contrast, the limits prescribed by these guidelines for COD and the microbial loads were well below the lowest threshold of the system. The correlation (R2 = 0.88; Fig 4) between enteric bacterial load and CO2 production indicated promising potential for detection of larger contamination events. However, the target water quality range for total heterotrophs, according to SANS241 [17], is 0–100 CFU/mL, which is below the lowest threshold of the CO2 correlation in open-loop configuration, and the target water quality range for coliforms is drastically more stringent at 0–5 CFU/100 mL.

10.1371/journal.pone.0247910.g004 <p>(a) Pollutant effects on CO<sub>2</sub> measurements. CO<sub>2</sub> profiles for increasing river water concentrations (pollution levels) were plotted for 2 independent sampling events. (b) Pollutant effects on CO<sub>2</sub> measurements. River water chemistry was correlated against CO<sub>2</sub> at the above-mentioned dilutions, including COD (mg/L), sulphate (mg-SO4-/L), total nitrogen (mg-N/L) and ammonia (mg-NH4+/L). The results of triplicate experiments are plotted. Minimum thresholds are indicated with dashed lines, as determined by the South African Guidelines for Drinking Water Standards [<xref ref-type="bibr" rid="pone.0247910.ref017">17</xref>]. (c) River water energetic (ATP) and microbiological profiles were correlated against CO<sub>2</sub> at the above-mentioned dilutions, including fecal coliforms (FC), enteric bacteria (BC) and heterotrophic bacteria (HB). The results of triplicate experiments are plotted. Minimum thresholds are indicated in red, as determined by the South African Guidelines for Drinking Water Standards [<xref ref-type="bibr" rid="pone.0247910.ref017">17</xref>].</p> </caption> <graphic mimetype="image" position="float" xlink:href="info:doi/10.1371/journal.pone.0247910.g004" xlink:type="simple"/> </fig> <p>Based on these threshold limitations of the open-loop CEMS alarm system, measuring CO<sub>2</sub> production on a continuously replaced carrier gas, the design was adapted to measure closed-loop CO<sub>2</sub> accumulation, in an attempt to decrease the lower threshold (<xref ref-type="fig" rid="pone.0247910.g001">Fig 1</xref>). It was only assessed against the microbial load, which is the limiting parameter (<xref ref-type="fig" rid="pone.0247910.g004">Fig 4</xref>). In this case, the rate of CO<sub>2</sub> accumulation, calculated from the slopes, was compared (ppm/h) at 10 CFU/100 mL, 10<sup>2</sup> CFU/100 mL and 10<sup>3</sup> CFU/100 mL (total coliforms). Here, the CO<sub>2</sub> accumulation rates were statistically distinguishable as low as 10 CFU/100mL according to linear best fit models, however the tool is truly useful over 100 cfu/100 mL, as the slopes at 10 and 100 cfu/mL only differ by 5 ppm, and control variation is 1.8 ppm (<xref ref-type="fig" rid="pone.0247910.g005">Fig 5</xref>). An automated flushing mechanism to facilitate closed-loop daily monitoring has been designed in-house, to allow for a pre-determined window of CO<sub>2</sub> accumulation before flushing, circumventing perpetual CO<sub>2</sub> accumulation.</p> <fig id="pone.0247910.g005" position="float"> <object-id pub-id-type="doi">10.1371/journal.pone.0247910.g005</object-id> <label>Fig 5</label> <caption> <title>Headspace accumulation of CO2 in in a closed-loop CEMS design.

A control (C) reactor containing sterile water was compared to E. coli inoculated (I) reactors. The inoculation concentration was increased twice: from an initial concentration of 10 CFU/mL over time range (A), it was increased to 100 CFU/mL for time (B) and finally (C) 1000 CFU/mL. The CO2 generation rates m(C) and m(I) in the control and inoculated reactors, respectively, are indicated in the figure, clearly showing a detectable difference in each case.

It is clear from these thresholds that the open-loop Canary CEMS is an indicator of significant, sudden contamination events, but is not suited for the monitoring of drinking water to potable national and international standards. The lack of suitability is based largely on sensitivity of the system for measuring microbial activity at levels as low as guidelines demand, which is effectively zero. The alarm effectively detects microbial loads higher than 102−103 CFU/mL, above the maximum thresholds for potable water. In attempting to compare the detection range to other widely promoted measurements techniques, such as qPCR, the definition of Limits of Quantification (LoQ’s) are informative, including precision and accuracy, only under stated experimental conditions [21]. Forootan et al. [22] explore the challenges of defining thresholds of detection and quantification of qPCR. In their context, the LoD (Limit of Detection) was 2 molecules, and the LoQ was 16 molecules. This is tenfold lower than the CEMS alarm, but still demands time, laboratory facilities and expertise.

Contamination events that have warranted reporting due to associated public illness include ranges as wide as 20 CFU/mL [23] to 10300 CFU/mL [24]. This work demonstrates the utility of the technology as an early warning system for such significant contamination events, as well as for applications in aquaculture, irrigation (water quality limited to 100 CFU/mL, according to South African National Guidelines [25]) and service water for washing equipment.

 Monitoring bioreactor disturbances

The alarm system was proven effective as an indicator of microbial inhibition and metabolic disturbances in active return activated sludge (RAS) reactors. In the case of a disinfectant (chlorine, Fig 6a and 6b), a pH disturbance (pH drop from 6.6 to 5.0 and 3.3, HCl, Fig 6c & 6d respectively) and a temperature disturbance (temperature decrease: 22°C to 16°C, Fig 6e), the CO2 immediately fluctuated well outside of pre-determined windows defined by steady-state CO2 measurements.

10.1371/journal.pone.0247910.g006 CO<sub>2</sub> fluctuations in response to physico-chemical disturbances.

The CO2 profiles of (a) a mixed and sedimentary RAS reactor with the addition of chlorine compared to (b) an autoclave sterilized RAS control reactor; as well as a mixed RAS reactor acidified from pH 6.6 to (c) 5.0 and (d) 3.3, with concentrated HCl, and (e) a sudden cooling event (ice; ~5% v/v) in a mixed RAS reactor connected to the system at 1 h. Measurement intervals are indicated with dashed lines.

The first disinfectant treatment triggered an approximately 5-fold increase in the metabolic activity of a mixed reactor (Fig 6a). The CO2 measurements subsequently steadily decreased over days, remaining almost 2-fold higher than the pre-treated whole-reactor CO2 production rate. The control reactor showed a slight increase in CO2 production upon addition of chlorine, probably due to chemical oxidation of organic material in the water, although the metabolic measurements were negligible in comparison to the non-sterile reactor confirming the biological metabolic response (Fig 6b). The sedimentary reactor demonstrated a similar, although less dramatic, response upon disinfectant treatment, suggesting a higher sedimentary chlorine demand. A second treatment with the same disinfectant concentration dropped the measurable CO2 production rate to zero for both the mixed and sedimentary reactors, and cell survival was also approximately zero after the second chlorine treatment. It is well established that microbes respond to stressors such as antibiotic challenges with a spike in CO2 production, likely due to the metabolic shifts for adaptation [7, 8], particularly with continued exposure to the chemical inhibitor. Higher CO2 production is potentially ascribed to the metabolic cost associated with the upregulation of efflux pump activity in the cell [7]. This work indicates a similar metabolic response to an oxidizing agent (Fig 6a).

Similarly, a pH decrease from 6.6 caused the microbial consortium to increase its metabolic rate (Fig 6c and 6d). The measurable metabolic rate was approximately 10-fold higher at pH 5.0 than at pH 3.3. Acidification to pH 3.3 caused a drop in cell concentration from 109 CFU/mL to 107 CFU/mL in 2 h and to 103 CFU/mL in 24 h, whereas cell concentrations remained constant at 109 CFU/mL with a milder pH drop to 5.5. The 10-fold difference in CO2 production is due to an interaction between this difference in biological contribution, as well as physico-chemical CO2 speciation.

Temperature and pH affect the liquid phase speciation of CO2 (see S1 Appendix) and will affect the liquid-to-headspace mass transfer rate. A decrease in pH shifts the carbonate equilibria from ionic bicarbonate towards dissolved CO2, which would then escape to the atmosphere and subsequently increase the headspace CO2 concentration. Assuming the liquid- and gas phases are in equilibrium and that the acidification process is rapid enough such that the total (combined liquid phase and headspace) molar amount of CO2 in the reactor remains constant, it is shown that the proportional increase in CO2 partial pressure pCO2 is given by Eq 1 (subscripts 0 and f indicate reactor conditions before and after acidification, see S1 Appendix): pCO2,fpCO2,0=VGRT0+VLKH(T0)(1+K1(T0)10-pH0+K1(T0)K2(T0)10-2pH0)VGRTf+VLKH(Tf)(1+K1(Tf)10-pHf+K1(Tf)K2(Tf)10-2pHf) Where VG and VL are the gas- and liquid-phase volumes, R is the universal gas constant, and KH, K1 and K2 are the Henry’s constant and dissociation constants of dissolved CO2. The equilibrium constants are all functions of temperature.

Within this study, the ratio of steady state CO2 partial pressures (before and after acidification) due to purely chemical effects, calculated using Eq 1, is compared to the ratio of steady state CO2 partial pressures as measured using the CEMS (Table 1). When the pH was dropped to a value of 5.0 from pH 6.6, the measured 24-fold increase in CO2 partial pressure far exceeded the values solely attributed to a disturbance in the chemical equilibrium, thus indicating that acidification triggered a metabolic response. In the case of acidification to pH 3.3, the measured 2.3-fold increase in CO2 partial pressure was slightly lower compared to that calculated for a purely chemical response. This indicates a complete cessation of metabolic activity. The lower than expected increase in CO2 partial pressure can be attributed to the continuous removal of CO2 from the reactor headspace by the CEMS.

10.1371/journal.pone.0247910.t001 Ratio of CO<sub>2</sub> partial pressures before and after acidification due to purely chemical effects (calculated) compared to combined chemical and metabolic effects (measured).
Initial pH Final pH pCO2,fpCO2,i(calculated) pCO2,fpCO2,i(measured)
6.6 5.0 2.5 23.9
6.6 3.3 2.7 2.3

In contrast, temperature fluctuations did not lead to a spike in CO2 production rates, but rather a drop in respiration to approximately 65% upon cooling (Fig 6e). There was no associated decrease in microbial load with the temperature fluctuation. The phenomenon of decreased respiration with decreasing temperature is well-demonstrated in soil microbial communities [26].

The temperature decrease, from 22 °C to 16 °C, would affect the chemical equilibrium mathematically represented by KH, K1 and K2. In this case, the CO2 partial pressure due to chemical effects alone was calculated to be 10%, underestimating the measured decrease of 40%, clearly indicating a decrease in metabolic activity.

The LiCor CO2 monitoring system software was also adapted to send an email in response to CO2 fluctuations outside pre-determined thresholds, alerting the system manager of metabolic disturbances, remotely and immediately. This adaptation eliminates the need for expertise to assess microbial activity, and the consequent performance of the bioreactor. It makes it accessible to technicians, which is critical in implementing this technology, particularly in rural, under-developed areas. Design adjustments might involve the circulation of a representative reactor sample through an adjacent compartment controlled for volume and headspace, rather than direct measurement of the reactor headspace, as well as the measurement of biofilms in pipes.

There is no silver bullet for the monitoring of water storage contamination events, be it potable, irrigation or water recirculation for cleaning—a common practice in drought-ridden areas. The high frequency of water quality system collapse demands creative additions to the current suite of water-monitoring techniques. There is a tension between the resolution of microbial water quality information, the cost and the speed at which information is accessible.

Thus, the efficiency of this technology can be compared to competing technology [27]. Such competing technologies, based on fluorescence assays and enzyme detection, still involve grab samples, and demand technical know-how and complex, energy-expensive technology. However, the true benefit of this work lies not in competition, but rather in the amplified benefits of multipronged approaches to water quality monitoring. The cumulative effect of both high-level, remote, online alarm systems and more detailed pathogen identification, will provide more effective protection to communities susceptible to outbreaks. In addition, the particular benefit of this technology is the accessibility to under-developed areas without geographical and economic access to standard laboratory facilities: often the most vulnerable communities.

 Conclusions

The Carbon Dioxide Evolution Monitoring System was shown to only be indicative of relatively high levels of microbial contamination in water systems. The lower thresholds are above the maximum coliform and heterotrophic loads permitted for potable drinking water according to WHO and national standards, although significantly improved by the redesign to a closed-loop system. In contrast to potable water standards, it was proven effective for significant contamination events and for monitoring water for broader applications, for example to national and international irrigation standards. The canary CEMS was also effective as an immediate, remote alarm for metabolic disturbances in a RAS reactor treated with a disinfectant, as well as pH and temperature fluctuations. The aim of the system is to act as an immediate indicator of the metabolic responses of the microbial aggregate, which was resolved from the physico-chemical CO2 equilibration by mathematical modelling. If thresholds and minimum limits are understood and optimized for individual application per system, the universal principle of whole microbiome respiration has notable potential as an early-warning technology for microbial disturbances in industrial settings.

 Supporting information Mathematical model of CO<sub>2</sub> release.

(DOCX)

The authors are grateful to Jaco van Rooyen (Department of Process Engineering, Stellenbosch University) for his assistance with pH auto-titrations.

 References Pan Y, Birdsey RA, Fang J, Houghton R, Kauppi PE, Kurz WA, et al. A large and persistent carbon sink in the world’s forests. Science. 2011;333(6045): 988–993. doi: 10.1126/science.1201609 21764754 Ballantyne AP, Alden CB, Miller JB, Tans PP, White JWC. Increase in observed net carbon dioxide uptake by land and oceans during the past 50 years. Nature. 2012;487(7409): 70. doi: 10.1038/nature11299 22859203 Kroukamp O, Wolfaardt GM. CO2 production as an indicator of biofilm metabolism. Appl Environ Microbiol. 2009;75(13): 4391–4397. doi: 10.1128/AEM.01567-08 19346353 Spanjers H, Vanrolleghem P, Olsson G, Dold P. Respirometry in control of the activated sludge process. Water Sci Technol. 1996;34(3–4): 117–26. Kroukamp O, Dumitrache RG, Wolfaardt GM. Pronounced effect of the nature of the inoculum on biofilm development in flow systems. Appl Environ Microbiol. 2010;76(18): 6025–6031. doi: 10.1128/AEM.00070-10 20639376 Bester E, Kroukamp O, Hausner M, Edwards EA, Wolfaardt GM. Biofilm form and function: carbon availability affects biofilm architecture, metabolic activity and planktonic cell yield. J Appl Microbiol. 2011;110(2): 387–398. doi: 10.1111/j.1365-2672.2010.04894.x 21122038 Jackson LM, Kroukamp O, Wolfaardt GM. Effect of carbon on whole-biofilm metabolic response to high doses of streptomycin. Front Microbiol. 2015;6. doi: 10.3389/fmicb.2015.00953 26441887 Ronan E, Edjiu N, Kroukamp O, Wolfaardt G, Karshafian R. USMB-induced synergistic enhancement of aminoglycoside antibiotics in biofilms. Ultrasonics. 2016;69: 182–190. doi: 10.1016/j.ultras.2016.03.017 27111871 Dumitrache A, Lynd LR, Liss SN, Wolfaardt GM. Tracking the cellulolytic activity of Clostridium thermocellum biofilms. Biotechnol Biofuels. 2013;6: 175. doi: 10.1186/1754-6834-6-175 24286524 Stone W, Kroukamp O, McKelvie J, Korber DR, Wolfaardt GM. Microbial metabolism in bentonite clay: saturation, desiccation and relative humidity. Appl Clay Sci. 2016;129: 54–64. Blackburn BG, Craun GF, Yoder JS, Hill V, Calderon RL, Chen N, et al. Surveillance for waterborne-disease outbreaks associated with drinking water—United States, 2001–2002. MMWR Surveill Summ. 2004;53(8): 23–45. 15499307 Coleman DC, O’Donnell MJ, Shore AC, Russell RJ. Biofilm problems in dental unit water systems and its practical control. J Appl Microbiol. 2009;106(5): 1424–1437. doi: 10.1111/j.1365-2672.2008.04100.x 19187140 Companies pay out more than £1.5m for breaking environment laws. The Guardian. 2017 Jan 30 [Cited 2019 June 25]. https://www.theguardian.com/environment/2017/jan/30/companies-pay-out-more-than-15m-for-breaking-environment-laws 360 Environmental [Cited 2019 June 25]. https://www.360environmental.co.uk/legislation/other/waste_prosecutions/ Alberta Government Website [Cited 2019 June 25]. https://www.alberta.ca/environmental-compliance-prosecutions-closed-files.aspx Paulse AN, Jackson VA, Khan W. Comparison of microbial contamination at various sites along the Plankenbrug: and Diep Rivers, Western Cape, South Africa. Water SA. 2009;35(4): 469–478. South African Water Quality Guidelines. Volume 1: Domestic Water Use. Department of Water Affairs and Forestry, 1996A; 2nd Edition. Guidelines for Drinking Water Quality, 2nd Edition, Vol 1: Recommendations. World Health Organization, Geneva, Switzerland. 1993. Guidelines for Canadian Drinking Water Quality—Summary Table. Water and Air Quality Bureau, Healthy Environments and Consumer Safety Branch, Health Canada, Ottawa, Ontario. 2019. Boyce JM, Havill NL, Havill HL, Mangione E, Dumigan DG, Moore BA. Comparison of fluorescent marker systems with 2 quantitative methods of assessing terminal cleaning practices. Infect Cont Hosp Ep. 2011;32(12): 1187–1193. doi: 10.1086/662626 22080657 Protocols for Determination of Limits of Detection and Limits of Quantitation; Approved Guideline. EPA-17, Clinical Laboratory Standards Institute. 2004;24(34): 1–39. Forootan A, Sjöback R, Björkman J, Sjögreen B, Linz L, Kubista M. Methods to determine limit of detection and limit of quantification in quantitative real-time PCR (qPCR). Biomol Detect Quant. 2017;12: 1–6. Hänninen ML, Haajanen H, Pummi T, Wermundsen K, Katila ML, Sarkkinen H, et al. Detection and typing of Campylobacter jejuni and Campylobacter coli and analysis of indicator organisms in three waterborne outbreaks in Finland. Appl Environ Microbiol. 2003;69(3): 1391–1396. doi: 10.1128/aem.69.3.1391-1396.2003 12620821 Boccia D, Tozzi AE, Cotter B, Rizzo C, Russo T, Buttinelli G, et al. Waterborne outbreak of Norwalk-like virus gastroenteritis at a tourist resort, Italy. Emerg Infect Dis. 2002;8(6): 563. doi: 10.3201/eid0806.010371 12023910 South African Water Quality Guidelines. Volume 4: Agricultural Water Use: Irrigation. Department of Water Affairs and Forestry, 1996B, 2nd Edition. Lloyd J, Taylor JA. On the temperature dependence of soil respiration. Funct Ecol. 1994; 315–323. Burnet JB, Dinh QT, Imbeault S, Servais P, Dorner S, Prévost M. Autonomous online measurement of β-D-glucuronidase activity in surface water: is it suitable for rapid E. coli monitoring? Water Res. 2019;152: 241–250. doi: 10.1016/j.watres.2018.12.060 30677635 10.1371/journal.pone.0247910.r001 Decision Letter 0 Zhang Dawei Academic Editor 2021 Dawei Zhang This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 0

19 Oct 2020

PONE-D-20-04971

Canary in the coliform mine: Exploring the industrial application limits of a microbial respiration alarm system

PLOS ONE

Dear Dr. Stone,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Dec 03 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Dawei Zhang, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Additional Editor Comments (if provided):

This manuscript describes an early-warning system for microbial disturbances in industrial settings by exploring microbial biofilm growth and metabolic responses. It is an interesting paper. However, the experiment design and statistical analysis need to be improved. I would like to receive the revised paper if you can address the reviewer's comments.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is an interesting paper and for the most part is carefully performed and well written. However, some of the statistical analyses and some figures don’t look correct to me and I am requesting a revision primarily to address this.

L122-125 It is stated that river water was used to assess “very low levels of microbial contamination”. However, the river is from a highly polluted source as stated earlier in the manuscript and as is supported by the chemical data that show very high COD. It needs to be explained here that the river water will be added in very low concentrations to pure water in order to mimic clean water.

L131-141 , 149-161

Its not stated explicitly how many sensors were used so I assume it was just one, and the sensor was used in sequential experiments to measure differences. Standard deviations were calculated from the STD of measurements obtained as a time series during a single sensor run.

The problem with this is that any temporal drift in the sensor output will be interpreted as a difference between treatments. It also violates the assumptions of a t-test which requires independent measurements when in this case, all data for the same treatment were autocorrelated. It is presumably this that accounts for the supposed significant differences between some treatments in Fig 2 despite near identical means. Instead, the figure should show the average and STD derived from repeated (independent) experiments with the treatments run in randomized order to counteract systematic biases caused by drift. If the experiment wasn’t repeated then variance cant be assessed.

L380 “ Headspace CO2 concentration had a linear correlation with the 380 dilution of the river water,” between 25 and 100% strength (Fig 4; R2 = 0.986) and all other data with R2 in Fig 4.

AND

L287 “Thistechnique confirmed the metabolic profile with tight correlation to CO2 data (R2=0.99; Fig 4),

The R2 values reported are much higher than would be expected from the scatter seen in the data in the figures. The only explanation I can think of is that the data were averaged before calculating R2, which is the difference between the two plots shown here: https://i.imgur.com/GDZmrjv.png . This artificially deflates variance by performing stats on already processed data.

L414 Fig 5 caption doesn’t match the figure. I don t understand what Fig 5 is showing.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

 10.1371/journal.pone.0247910.r002 Author response to Decision Letter 0 Submission Version 1

29 Nov 2020

Editor Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf

and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Authors:

Thank you for the guidance. We have gone carefully through formatting, and have met the stipulated requirements. If further editing is necessary, we are happy to do so.

As primary and corresponding author, I made the mistake of not using track changes as I edited.

However, the only edits we made were those recommended by the reviewers and described below (including Line numbers of edits). In addition, I formatted according to the links provided (headings, figure and table captions, and references). All figures were corrected using PACE.

I highlighted the changes in yellow, to account for the mistake of not using track changes before saving the new, edited document.

Additional Editor Comments (if provided):

This manuscript describes an early-warning system for microbial disturbances in industrial settings by exploring microbial biofilm growth and metabolic responses. It is an interesting paper. However, the experiment design and statistical analysis need to be improved. I would like to receive the revised paper if you can address the reviewer's comments.

Authors:

We thank the editor for the opportunity to respond to the reviewer’s comments and resubmit. We hope we have addressed all comments to your satisfaction, and are available for follow up if further issues need clarification.

The reviewer’s comments were astute, and we are grateful for the improvement of the manuscript through the editorial process.

Reviewer:

L122-125 It is stated that river water was used to assess “very low levels of microbial contamination”. However, the river is from a highly polluted source as stated earlier in the manuscript and as is supported by the chemical data that show very high COD. It needs to be explained here that the river water will be added in very low concentrations to pure water in order to mimic clean water.

Authors:

We thank the reviewer for this clarification, and have adjusted the sentence as follows:

“This polluted river water was diluted with pure water to mimic a range of pollution levels, from low to high.” (Line 118-119)

Reviewer:

L131-141 , 149-161 Its not stated explicitly how many sensors were used so I assume it was just one, and the sensor was used in sequential experiments to measure differences. Standard deviations were calculated from the STD of measurements obtained as a time series during a single sensor run.

The problem with this is that any temporal drift in the sensor output will be interpreted as a difference between treatments. It also violates the assumptions of a t-test which requires independent measurements when in this case, all data for the same treatment were autocorrelated. It is presumably this that accounts for the supposed significant differences between some treatments in Fig 2 despite near identical means. Instead, the figure should show the average and STD derived from repeated (independent) experiments with the treatments run in randomized order to counteract systematic biases caused by drift. If the experiment wasn’t repeated then variance cant be assessed.

Authors:

Two sensors were used, rather than one, but the reviewer is correct in this statistical observation. The sensors were regularly interchanged throughout all the runs, between control and experimental reactors, and between high and low CO2 accumulation, in order to assess drift. We can confirm that sensor drift is negligible. Both this random, regular exchange between sensors during experimental work confirms this, and, as stated earlier in the manuscript “Respiration was monitored in reactor headspaces, which were open to the atmosphere via a defined inlet port. Baselines were predictable at 385 ppm CO2 ± 45 ppm for the open-loop system, with variation attributed to human activity in the vicinity increasing CO2 levels during the day. Closed-loop baselines had a similar standard deviation if measured at separate instances over different days, but a variation of less than 5 ppm over days if sealed.” The low variance under similar controlled conditions on different days allows us to safely conclude that any sensor drift is negligible. We thank the viewer for this meticulous caution, and have added a sentence about using two sensors with random interchange during experimentation (Line 132-137), for replicability of the work.

We agree that employing a t-test was inappropriate on our part, due to the use of correlated data. Since the assumptions of a t-test were violated, we have removed the significance indicators on the graph and have removed that statistical description in ‘Material and Methods’. However, based on the established low variance in sensor measurements and the noticeable difference between the measured respiration for the Undiluted River Water sample in the 0.9 L headspace experiment and all five other experiments, we can still confidently conclude that microbial activity is detectable and a larger headspace volume enhances measurement sensitivity, thus supporting the methodology developed subsequently.

Reviewer:

L380 “Headspace CO2 concentration had a linear correlation with the 380 dilution of the river water,” between 25 and 100% strength (Fig 4; R2 = 0.986) and all other data with R2 in Fig 4.

AND

L287 “This technique confirmed the metabolic profile with tight correlation to CO2 data (R2=0.99; Fig 4).

The R2 values reported are much higher than would be expected from the scatter seen in the data in the figures. The only explanation I can think of is that the data were averaged before calculating R2, which is the difference between the two plots shown here: https://i.imgur.com/GDZmrjv.png . This artificially deflates variance by performing stats on already processed data.

Authors:

The reviewer is correct, and the statistical values were calculated based on averages. In an attempt to fit all of the graphs into an aesthetically streamlined format, presenting multiple assessments per quadrant, the statistical analyses were incorrectly collapsed into a format that, as the reviewer points out, deflates variance.

We have expanded them all as recommended by the reviewer and avoided passing the correlations through zero. We kept them on separate graphs for ease of audience interpretation, as single images become too busy. (Lines 350-363). We apologize for the conflation of the statistical values, and are grateful to the reviewer for encouraging the more rigorous analysis and stringent interpretation.

Reviewer:

L414 Fig 5 caption doesn’t match the figure. I don t understand what Fig 5 is showing.

Authors:

Thank you for bringing the lack of articulation to our attention. We’ve expanded the figure caption to provide a clearer description of the figure. We hope this brings clarity: if it is still unclear, we are happy to edit further. (Line 396-401).

Submitted filename: Response to Reviewers.docx

 10.1371/journal.pone.0247910.r003 Decision Letter 1 Zhang Dawei Academic Editor 2021 Dawei Zhang This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 1

20 Jan 2021

PONE-D-20-04971R1

Canary in the coliform mine: Exploring the industrial application limits of a microbial respiration alarm system

PLOS ONE

Dear Dr. Stone,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Mar 06 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Dawei Zhang, Ph.D.

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: (No Response)

Reviewer #3: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Partly

Reviewer #3: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: No

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: The authors present an interesting method for monitoring respiration by quantifying CO2. The device is tested in proof of concept experiments focusing on detection of changes to baseline respiration rates in each system – river water and activate sludge. Overall the data interpretation and modeling need to be improved.

Data and approach – It is unclear why the lumped mass transfer coefficient was fit. This could be evaluated using a control experiment. The potential issue in simultaneously fitting the lumped mass transfer coefficient and the CO2 production rate is that these parameters are likely to be correlated in ways that suggest potential for non-unique optimization/fit. The manuscript describes using R2 for assessing the two parameter fit (r and KLA). It would be helpful to establish that a one parameter model is not capable of describing the data, and then using an information criteria metric (e.g., Akaike) to establish the additional fitting parameter is warranted. A superior approach would be to independently determine KLA (fit to a control test) so as to only fit the rate parameter (i.e., what the device aims to measure) to the experimental data.

What is the implication of the variability in ambient CO2? The manuscript (Line 284) suggests that the baseline is 385 +/-45 ppm. What is the +/- here? The manuscript implies on the next line that it may be a standard deviation, but it is unclear. Since the monitoring device is based on differencing with this ambient condition (Line 127-128), what is the implication on the sensitivity of the motorizing? Later in the manuscript (Line 380-390) there is discussion about detection limit in terms of CFUs. How does the CFU detection limit depend on the substrate and nutrient condition of the water? That is, is using a detection limit in terms of CFUs system dependent? It stands to reason this type of conversion of the detection limit would depend on the rate at which the microbial community can oxidize the carbon. Seemingly this would depend on the form of the carbon present, the type of microbial community, and any limitations on carbon utilization (e.g., carbon, nutrient, and inhibitor/competitor concentrations).

Application – What is unclear from the work is the extent to which the CO2 monitoring system can help in application. The data show that the microbial community respond to perturbations established by adjusting chlorine concentration and pH, but what about monitoring a real plant? How much variability is present in the CO2 concentrations, and what is the time-scale of this variability compared to the time-scale of an upset in the community? Would upsets be detected earlier using this type of CO2 monitor? Also, how do the authors envision going from a beaker type test to a large, open top aeration tank? Is this a sample and test method, or is there potential for real time monitoring of the process? If the latter, is there a need for an array of sensors, or some sort of gas collection device? I raise these questions because the manuscript implies that the method can help in industrial settings without actually testing industrial settings.

River water vs activated sludge – the manuscript is often redundant in terms of the comparison between the two waters. The rationale for investigating these two end members of microbial activity are clear and can be stated once – ideally in the methods section. As I understand the manuscript, the results using the river water were not meaningful (Lines 290-298 suggest confounding factors limited the sensor or its interpretation). Unfortunately, this makes inclusion of the river water concept and data substantially less important. I recommend that the river water aspect be removed from the manuscript, or repeated using an experimental approach which resolves the issues related to the apparatus.

Minor items

Figure 3a – the use of predicted here for the modeled CO2 is incorrect. The manuscript states that the modeling results shown in figure 3a are a fit. Predictions have no fitted parameters (i.e., parameters are determined independent of the experiment being modeled). The modeling results for the 0.25 L case should be smooth though there appears to be a small increase then decrease in modeled CO2 around 2 hr. If this feature is real (i.e., the plot is small and the data line may be misleading the eye), then what in the model can account for the presence of this type of behavior? See above concern about the two parameter fit.

Reviewer #3: This manuscript provides a methodology to detect aberrations in microbial growth in an industrial setting. The advantage of CEMS is its utilization of sustainable, existing technology to function as a first-pass ‘canary’. Overall the manuscript is technically sound after the first revision. However, we would like to see more details on the points that we made below.

1. For figure 3a & b:

For Figure 3a, there is a discontinuation of the undiluted 0.90 L headspace experimental data between 3 - 6 hr point. We would like to see more explanation on how the measurement was done in specific.

2. For the description of figure 3 in the main manuscript, refer to the mathematical appendix for the equations and parameters to more clearly explain the regression.

3. L 416-421: The authors stated that this work demonstrates an early warning system for such significant contamination events (20 CFU/ml to 10300 CFU/ml). It would make a stronger argument if the authors explain what other methods are there for this kind of detection and give a comparison why this system would be better than the others.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

 10.1371/journal.pone.0247910.r004 Author response to Decision Letter 1 Submission Version 2

28 Jan 2021

Response to the generous editorial and reviewer comments have been provided as a separate PDF document (Response to Reviewers) during the submission process, as recommended throughout the editorial process. It allows for the formatting necessary for a coherent response, especially in the modelling sections.

Submitted filename: Response to Reviewers.pdf

 10.1371/journal.pone.0247910.r005 Decision Letter 2 Zhang Dawei Academic Editor 2021 Dawei Zhang This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Submission Version 2

17 Feb 2021

Canary in the coliform mine: Exploring the industrial application limits of a microbial respiration alarm system

PONE-D-20-04971R2

Dear Dr. Stone,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Dawei Zhang, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The paper has been extensively modified from its first submision and in my opinion the reviewer comments have been adequately addressed. I asked for a minor revision so that the authors can ensure that the journals data policy has been completely fulfilled (it may be necessary to share raw data via a repository).

Reviewer #3: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #3: No

 10.1371/journal.pone.0247910.r006 Acceptance letter Zhang Dawei Academic Editor 2021 Dawei Zhang This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

23 Feb 2021

PONE-D-20-04971R2

Canary in the coliform mine: Exploring the industrial application limits of a microbial respiration alarm system

Dear Dr. Stone:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Dawei Zhang

Academic Editor

PLOS ONE