Edited by: Mohiuddin Md. Taimur Khan, Washington State University Tri-Cities, United States
Reviewed by: M. Jahangir Alam, University of Houston, United States; John P. Brooks, United States Department of Agriculture, United States
This article was submitted to Environmental Health, a section of the journal Frontiers in Public Health
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Performing microbiological assays on environmental samples in field settings poses logistical challenges with respect to the availability of suitable equipment or the ability to get samples to the laboratory in a timely fashion. For example, the viability of some bacteria can decrease greatly between sampling and arrival to the laboratory for processing. We developed and constructed rugged, reliable, and cost-effective portable incubators that were used by 10 independent field teams to perform microbiological assays on surface water samples from lakes across Canada. Rigorous testing and validation of our incubators ensured that incubation conditions were consistent within and across all 10 field teams and 2 sampling years. Samples from all sites were processed in duplicate and bacterial counts were highly repeatable within and across sampling teams. Bacterial counts were also found to be statistically equivalent to counts obtained with standard laboratory techniques using a conventional incubator. Using this method, thermotolerant coliforms (TTCs) and
Surveillance and monitoring of bacterial contamination in aquatic environments is critical for public health for water recreation, consumption, and sanitation (
Many monitoring programs for recreational areas sample on a daily or weekly basis (
To minimize temporal degradation, environmental samples would ideally be processed as soon as they are collected. Several studies have investigated the use of microbiological testing during field assessments and in low-resource settings (
A humanitarian outreach program by Engineers Without Borders (Austin, Texas) developed a low-cost, battery-operated incubator, the Armadillo, for use with Petrifilms™, with education about water quality in remote environments via the visually striking results being its primary use (
As part of a national lake assessment (
The custom-built incubator featured a double-chamber design composed of two snap-lid leak-proof food storage containers inside of a 16-quart insulated hard-shell cooler (
Incubator schematics.
The outer heating chamber, a 3.3-L snap-lid leak-proof Tritan™ plastic food storage container, had three small holes drilled into it for wiring the two temperature sensors, two resistance heating pads, and a small axial fan (
Inoculated Petrifilms™ were stacked and placed on top of a piece of ½″ (1.27 cm) plywood that provided a level surface in the incubation chamber. A second piece of plywood with a 2 ¼″ (5.71 cm) hole, being slightly larger than the circular medium area of the Petrifilms™ (2″, 5.08 cm), was placed on top of the Petrifilms™. This was followed by another layer of plywood and a layer of compressible foam, securing the Petrifilms™ in place in the chamber (
Several prototypes of the incubator were made throughout the design process, with the final version outlined here. In total, five incubators were constructed to these specifications and were used by the ten field teams, five teams per year, for the sampling campaigns of 2018 and 2019. While sampling in 2018, one team that occasionally experienced ambient temperatures below 15°C noticed that the incubator struggled to maintain 37°C. Prior to the field campaign of 2019, which extended up to 68° N in the Yukon and Northwest Territories, a layer of reflective bubble foil insulation was added to all inner sides of the incubators for added insulation. A full list of parts, materials, and tools used for construction of the incubators can be found in the
Field teams had constant access to 120-volt AC power, supplied through a power converter from either a vehicle's electrical system, a portable generator, or a deep cycle battery bank, or through direct connection to standard electrical services. The power supply/adapter of the incubator converted this input power to 12-volt DC which was required for its electrical system. As incubators were supplied with continuous power, they were kept on for the entire duration of the 9 week field campaign. A schematic outlining the wiring of the incubator is shown in
As the incubators were custom-built and used independently by different field teams, it was important to ensure temperature consistency within and between incubators. To evaluate this, a thermocouple was placed inside the outer heating chamber of each incubator, with all five thermocouples connected to a Campbell Scientific CR300 Datalogger. Following these tests, the set-point of the thermostat control module of the incubator was adjusted if necessary, to ensure that all incubators held the same internal temperature. The homogeneity of the temperature inside each incubator was evaluated by placing one thermocouple inside the inner incubation chamber, and the other four thermocouples in each corner of the outer heating chamber of one incubator. The time for the incubator to reach 37°C and the temperature stability over time were measured, with the change in temperature when disconnected from power also evaluated in 2019. In 2019, after adding a layer of reflective bubble foil insulation to the incubators, these tests were repeated at ambient temperatures of 4°C and at 24°C. A simulation of adding/removing Petrifilms™ was also performed in 2019, where the incubator was opened and the outer heating chamber and inner incubation chamber was disassembled for 1 min, to determine the time for the temperature to recover to 37°C. The power consumption of the incubator was determined by using a multimeter to measure the amperage when the heating pads were on and off. A logger recorded the time heating pads were on over a 12 h period, which allowed calculation of the mean amperage.
The LakePulse survey involved five independent field teams each year. Each team was identified by a color (blue, green, purple, red, or yellow) and typically sampled one lake per day, in their respective regions, for 8 to 9 weeks during the summers of 2018 and 2019. Each lake was sampled once, for a total of 45 to 49 lakes per team per year.
Single sub-surface grab samples were aseptically collected in sterile 50-mL conical tubes (Fisher Scientific, Product # 05-539-13) at two sites per lake, one in the littoral zone and one above the deepest part of the lake, below referred to as the littoral and index sites, respectively. These were collected from 217 lakes in 2018 and 215 lakes in 2019. Samples were stored on ice or at 4°C for a maximum of 2 h before being processed.
After incubation, Petrifilms™ were removed from the incubator and backlit photographs were taken by the 8-megapixel camera of a 2017 iPad Pro (Apple, Toronto, Canada) while holding the Petrifilms™ up to the sky. Interpretation of the photographic results was later performed manually by a single individual as per the manufacturer's instructions, with the exception that all colonies were counted even if the number was above the recommended maximum countable range for the product. Photos were scored in a randomized order after assignment of a computer-generated randomized name to each JPEG image. After scoring, a random 10% of the images were re-scored to ensure consistency of results.
The data were subjected to several levels of quality control. Before analysis was performed, all samples that were not processed the same day of collection, were incubated beyond 24 ± 2 h, or contained algae or debris such as sediment that impaired scoring of results were removed. For any samples where colony counts differed by more than 10 colonies and/or >30% between duplicates, the photographs and/or the original incubated Petrifilms™, which were stored at 4°C after incubation, were consulted to verify the accuracy of any outliers and they were removed if necessary. Additionally, notes made by the individuals that processed the samples while in the field were consulted to flag and remove samples that may have been problematic. Prior to analysis, samples that had colony counts above the recommended maximum countable range of the product, >150 colony forming units per mL (CFU/mL) (
A comparison to standard laboratory methods was performed to ensure that the data generated from this method were equivalent to those using a conventional laboratory incubator. Three sub-surface water grab samples were collected from a local lake and processed using the same protocol as outlined above. Samples were processed in triplicate and incubated in either a conventional laboratory incubator (Model No. 2005, VWR) or a custom-built incubator at 37°C for 24 ± 2 h. Scoring of results was performed and subjected to the same interpretation and quality control as outlined above.
Data from the 2018 field season were used for method validation and preliminary analyses. R v3.5.0 (
For the first and traditional approach (
A precision criterion should be determined for each individual lab for each type of sample examined to determine what is acceptable reliability for that laboratory (
For the comparison between the custom-built incubator and a conventional laboratory incubator, an F-ratio test was performed using the log-transformed total Petrifilm bacterial loads.
A drawback of the traditional approach using precision criteria is that bacterial colony counts of zero are eliminated due to the need for log transformation, therefore a second approach using more modern statistical methods was also used. As the data were based on bacterial colony counts and each sample was processed in duplicate, we used Generalized Linear Mixed Models (GLMMs) with negative binomial distributions and a log link function using the package lme4 v.1.1-23 (
Temperature was homogeneous throughout the outer heating and inner incubation chambers for each of the five incubators and the temperature record for one of the incubators is shown in
Temperature recordings from five thermocouples placed inside one incubator. Measurements were taken every 30 s for ~4 h at an ambient temperature of 26°C.
When tested at ambient temperatures of 26°C in 2018 and both 24 and 4°C in 2019, all incubators maintained a relatively constant temperature near 37°C during the testing period (
Consistency of internal temperature profiles and recovery times after opening the incubator lid and disassembling the outer and inner chambers for 1 min.
26°C |
Average temperature (°C) | 36.5 | 37.4 | 36.7 | 37.2 | 37.4 | 37.0 |
Temperature range (°C) | 36.2–36.9 | 37.0–37.9 | 36.4–37.1 | 37.0–37.5 | 37.1–37.9 | 36.2–37.9 | |
24°C |
Average temperature (°C) | 36.6 | 37.1 | 36.8 | 36.6 | 37.2 | 36.8 |
Temperature range (°C) | 36.3–37.1 | 36.8–37.8 | 36.4–37.2 | 36.3–36.9 | 36.8–37.5 | 36.3–37.8 | |
Time to recover (minutes) | 11 | 18 | 13 | 11 | 18 | 14.2 | |
4°C |
Average temperature (°C) | 36.7 | 36.4 | 36.5 | 36.4 | 36.6 | 36.5 |
Temperature range (°C) | 36.1–37.6 | 35.7–37.3 | 35.7–37.4 | 35.5–37.4 | 35.8–37.5 | 35.5–37.6 | |
Time to recover (minutes) | 19 | 25 | 23 | 25 | 26 | 23.6 |
When the heating pads were on, the incubator drew approximately 1 A, and when the heating pads were off, the incubator drew ~0.1 A. For a period of 12 h at an ambient temperature of 24°C, the incubators consumed on average 2.9 W. When tested for the same period at an ambient temperature of 4°C, the incubators consumed on average 8.5 W.
To evaluate precision of replicates at the individual level, precision criterion-based analysis was used for a subset of the data, specifically Blue team. Due to a change-over in personnel part way through the 2018 field season, sample processing was performed by two different people over the course of the field season. After removing replicate pairs when either replicate had a count of zero, person A had a precision criterion of 0.357 (
To evaluate precision of replicates at the individual level, the GLMM and ICC approach was used for Blue team's complete dataset (which retains zero values). Person A had an ICC of 0.99 (
The relationship of bacterial loads and three lake metrics, ecozone, size, and sampling site, are shown in
Boxplots of total Petrifilm, thermotolerant coliform, and
Here we provide a framework for future researchers to build their own rugged, reliable, and cost-effective incubators that have been rigorously tested and whose design has been proven to withstand the challenges of field work while maintaining good microbiological standards.
For a study of this scope and scale, rigorous testing and yearly validation of the custom-built incubators was of the highest importance. The temperature is homogeneous throughout the incubator due to the fan and double-chambered design. The fan helps to circulate heat in the outer heating chamber, minimizing hot spots nearest the heating pads and ensuring even heating on all sides of the inner incubation chamber. Temperature fluctuation is minimal, varying by <2°C from the setpoint. The temperature of the incubator recovers in a relatively short period of time after being opened for manipulations, as evidenced by monitoring during the simulated adding/removing of Petrifilms™, even at an ambient temperature of 4°C. Furthermore, the incubation chamber temperature decreases slowly when the incubator is briefly unplugged, further demonstrating its buffering capacity and ability to handle disruptions such as switching power sources (
The temperature performance of all five incubators was validated before each field season, ensuring consistent temperature conditions across all incubators and both sampling years. Testing in advance of the 2019 field campaign at 4°C, after adding the layer of reflective bubble foil insultation, showed that the incubators would be able to maintain temperature even below temperatures expected for Northern Canada in late August. Monitoring power consumption and periods of heating indicated that the incubators were adequately insulated for even a low-temperature environment. While in the field in 2018, several teams occasionally noticed that the internal temperature would rise above 37°C when the incubator was in direct sunlight on hot summer days. When incubators were moved out of direct sunlight, internal temperature would quickly return to 37°C. The level of insulation and the lack of cooling components were more than adequate for this Canadian field study and would similarly be for other studies in temperate climates. To the best of our knowledge, this is the first field-based study to perform both testing and validation of their incubator design.
We performed statistical analyses using two different approaches for validation of our incubator and showed that data generated at the individual and team levels were comparable and reliable, despite one team (Purple) showing more variability. Purple team sampled lakes in the Semi-Arid Plateaux and Prairies ecozones where crop and livestock farming are common. Bacterial loads from their sampled lakes were frequently either below (<0.5 CFU/mL) or above (>150 CFU/mL) the detection thresholds. The approach where counts of zero were removed resulted in a smaller sample size for this team than for the other teams and this helps to explain why their precision criterion was much higher. This also inflated the overall precision criterion using data for all teams. Greater variability for this team was also shown using the second analytical approach, although it is much less pronounced. Model testing indicated that data from all other teams closely follow a Poisson distribution while Purple team's data were overdispersed, leading us to use negative binomial models for our analysis. These models include residual effects that are not included in Poisson models, and therefore ICCs for the other teams were more conservative than if a Poisson distribution was used, although they still show a very high level of reliability across all five teams.
Using the traditional approach, where zeros are removed and data are in log-transformed space, very low counts that differ by only a few colonies are often flagged as problematic (e.g., 4 and 10), although they are biologically acceptable. On the other hand, when counts are higher (e.g., 28 and 64), data are often shown to be acceptable using this approach, even though counts may differ by several fold and are not biologically acceptable. This statistical approach is often the standardized way in which the precision of duplicate analyses is determined by analytical laboratories as part of their quality assurance guidelines (
The more contemporary second approach takes into account the underlying distribution of the data and further highlights the limitations of using log transformation for count data. These contemporary analyses showed that each team had a high and comparable degree of reliability, something not as evident through the traditional log transformation approach. These findings not only support validation of our incubators but also show that the data are highly reliable and there are minimal artifacts in the dataset due to different processing individuals or teams/incubators. This statistical approach allows the data to be more useable for downstream analysis as counts of zero are retained and relationships of various factors with bacterial loads can be investigated directly (
The direct plating method with a sample volume of 1 mL required when using Petrifilms™ is less representative than using a sample volume of 100 mL as frequently employed for methods such as membrane filtration and most probable number (MPN). Concentration of samples by membrane filtration was not feasible for this study as it would have resulted in the total bacterial loads being above the maximum countable range, and the results would also have been impaired by the presence of more algae and sediment on the filters. Although compact MPN methods exist that could be used for field applications, such as the IDEXX Quanti-Tray™, their size, cost, and the requirements for each team to have a tray sealer on site made this an unrealistic option for this study.
Although some of the sampled lakes are in urban and highly impacted environments, many of the lakes are minimally impacted by humans. Therefore, it is not surprising that TTC and
As the lower detection threshold for
This field-based method allows bacterial loads to be accurately determined, including for very remote lakes, far away from traditional testing laboratories. By performing testing on-site within hours of collection, concerns about changes in the bacterial community between the time of sampling and the time of processing are minimized. This method can be used for a variety of future field studies when timely access to a microbiology laboratory is not possible and/or is cost-prohibitive. The incubator described here can be constructed in a matter of a few hours with minimal tools and expertise for a total materials cost of approximately $150 (CDN). Its components are readily available and can be easily purchased from hardware stores and online retailers. At a cost of approximately $2 (CDN) per Petrifilm™, bacterial loads of lake surface waters were determined for a total consumable cost of <$10 per lake. The low initial investment and consumable costs make this system highly cost-effective, and this approach can be used in the future by other research groups operating with modest budgets. The low-cost and flexibility for power supply options may also provide a solution for low- and middle-income countries with unreliable electricity to better perform microbiological testing and medical diagnostics in resource-limited settings (
The methods used here can easily be adapted to suit the needs of future studies and are easily modified to work with many other microbiological testing products, including other products in the 3M™ Petrifilm™ suite. The incubator is inexpensive, easy to assemble, reliable, requires no specialized equipment, and is easily transported due to its ruggedness and low weight. It is easy to ship and travel with by air as there are no chemicals or batteries involved and no components are regulated as dangerous goods. For future research programs that do not have continuous access to electricity, the incubator design is easily modified to run on battery power. Modification to use an 8 Ah 12 V battery, which fits inside of the incubator, or a 100 Ah 12 V deep-cycle battery would power the incubator set at 37°C in an ambient temperature of 24°C for 33 h and 17 days, respectively, making it suitable for use in many different field scenarios.
The design and validation of our incubator answers the call for the development of robust methods for performing bacteriological testing while in the field. Detecting
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
YH and AL secured funding. JW, M-PV, YH, and AL conceived the study. JW and M-PV developed the methods and administrated the project. JW collected, processed, analyzed, and visualized the data. JW and GR carried out the statistical analysis. JW wrote the manuscript with contributions from M-PV. All authors read and revised the manuscript and gave final approval for publication.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank all the individuals, especially Patrick Cliche, who helped brainstorm ideas for the development of our field incubator. We would also like to extend sincere thanks to all those individuals who helped in construction of the incubators and to those who processed samples during the 2018 and 2019 field seasons, without whose efforts this project would have not been possible.
The Supplementary Material for this article can be found online at: