Full Length Article|Articles in Press
• PDF [1 MB]PDF [1 MB]
• Top

# Quantitative assessment of automated purification and concentration of E. coli bacteria

Open AccessPublished:February 17, 2023

## Highlights

• Automated Dual-filter method for Applied REcovery, or aDARE, was designed, built, and tested to automate the process of bacteria purification and concentration.
• 42 ± 13-fold enrichment was achieved using the automated system in 5.5 min.
• Using 0.22 µm and 5 µm pore size filters (both 25 mm diameter) with 5 mL sample volume, a 900 µL retentate recovery volume was optimal for bacterial recovery and enrichment.

## Abstract

Automated methods for rapidly purifying and concentrating bacteria from environmental interferents are needed in next-generation applications for anything from water purification to biological weapons detection. Though previous work has been performed by other researchers in this area, there is still a need to create an automated system that can both purify and concentrate target pathogens in a timely manner with readily available and replaceable components that could be easily integrated with a detection mechanism. Thus, the objective of this work was to design, build, and demonstrate the effectiveness of an automated system, the Automated Dual-filter method for Applied Recovery, or aDARE. aDARE uses a custom LABVIEW program that guides the flow of bacterial samples through a pair of size-based separation membranes to capture and elute the target bacteria. Using aDARE, we eliminated 95% of the interfering beads of a 5 mL-sample volume containing 107 CFU/mL of E. coli contaminated with 2 µm and 10 µm polystyrene beads at 106 beads/mL concentration., The target bacteria were concentrated to more than twice the initial concentration in 900 µL of eluent, resulting in an enrichment ratio for the target bacteria of 42 ± 13 in 5.5 min. These results show the feasibility and effectiveness of using size-based filtration membranes to purify and concentrate a target bacterium, in this case E. coli, in an automated system.

## 1. Introduction

From water purification technologies to biological pathogen screening for clinical diagnosis, the ability to detect a target pathogen rapidly and accurately helps us best protect our communities in a variety of different ways. Currently, the target pathogen detection gold-standard is quantitative Polymerase Chain Reaction (qPCR) [

“PCR Test for COVID-19: What It Is, How Its Done, What The Results Mean,” Cleveland Clinic. https://my.clevelandclinic.org/health/diagnostics/21462-covid-19-and-pcr-testing [Accessed 4 November 2021].

]. Though sensitive, this method is time consuming and labor intensive, specifically when various samples are involved, because of the DNA extraction process required prior to the DNA amplification step. Furthermore, for a successful amplification, and thereby successful detection of the target pathogen, PCR requires precision in mixing the expensive reagents required for the steps previously mentioned. A potentially more cost-effective and rapid alternative approach is the use of an antibody-coated biosensor such as a field-effect transistor (FET). FETs decrease labor, have a quicker response time, decrease cost of reagents, and can be mass-produced using semiconductor manufacturing technology. FETs have been used for the detection of organisms for medical diagnostics [,
• Purwidyantri A.
• et al.
Sensing performance of fibronectin-functionalized Au-EGFET on the detection of S. epidermidis biofilm and 16S rRNA of infection-related bacteria in peritoneal dialysis.
,
• Formisano N.
• et al.
Inexpensive and fast pathogenic bacteria screening using field-effect transistors.
], environmental monitoring [,
• Shen F.
• et al.
Integrating silicon nanowire field effect transistor, microfluidics and air sampling techniques for real-time monitoring biological aerosols.
,
• Thakur B.
• et al.
Rapid detection of single E. coli bacteria using a graphene-based field-effect transistor device.
,
• Chang J.
• et al.
Ultrasonic-assisted self-assembly of monolayer graphene oxide for rapid detection of Escherichia coli bacteria.
] and bioresearch [,
• Chang J.
• et al.
Ultrasonic-assisted self-assembly of monolayer graphene oxide for rapid detection of Escherichia coli bacteria.
,
• Tarasov A.
• et al.
A potentiometric biosensor for rapid on-site disease diagnostics.
]. Furthermore, different types of FETs have successfully detected bacteria such as Staphylococcus epidermidis (S. epidermidis) [
• Purwidyantri A.
• et al.
Sensing performance of fibronectin-functionalized Au-EGFET on the detection of S. epidermidis biofilm and 16S rRNA of infection-related bacteria in peritoneal dialysis.
] and Escherichia coli (E. coli) [
• Thakur B.
• et al.
Rapid detection of single E. coli bacteria using a graphene-based field-effect transistor device.
] with a concentration as low as 10 CFU/mL [
• Chang J.
• et al.
Ultrasonic-assisted self-assembly of monolayer graphene oxide for rapid detection of Escherichia coli bacteria.
]. FETs require only microliters of volume of a sample [
• Zhang Y.
• Xu C.
• Guo T.
• Hong L.
An automated bacterial concentration and recovery system for pre-enrichment required in rapid Escherichia coli detection.
,
• Ezenarro J.J.
• Mas J.
• Muñoz-Berbel X.
• Uria N.
Advances in bacterial concentration methods and their integration in portable detection platforms: A review.
], and are amenable to automation due to their small footprint, which allows for faster sample processing and less variation due to human handling.
As with many detection methods, FET sensitivity can be affected by interferents present in real-world samples [

A. A. Fatah, J. A. Barrett, R. D. Arcilesi, K. J. Ewing, C. H. Lattin, and T. F. Moshier, “An introduction to biological agent detection equipment for emergency first responders,” p. 53.

]. Thus, for sensitive FET sensing, it is recommended to purify and concentrate sample upstream, prior to reaching the FET for detection [
• Shen F.
• et al.
Integrating silicon nanowire field effect transistor, microfluidics and air sampling techniques for real-time monitoring biological aerosols.
,
• Thakur B.
• et al.
Rapid detection of single E. coli bacteria using a graphene-based field-effect transistor device.
,
• Walper S.A.
• et al.
Detecting biothreat agents: from current diagnostics to developing sensor technologies.
]. Ideally, this would be done in an automated system capable of handling a range of volumes from milliliters to microliters that would be easily integrated with a FET. Thus, the goal of this work is to introduce such a purification and concentration system.
Traditional ways of purifying and concentrating samples prior to analysis involve chemical, physical, physiochemical, or biological approaches such as centrifugation, and immunomagnetic separation. Centrifugation is a commonly used method that allows for the separation of sample components based on mass or density which can both purify and concentrate to a desired volume by removal of the supernatant and resuspension. Though it allows for processing a range of volumes based on the centrifuge size, this process can be time-consuming, labor intensive, and not amenable to automation [
• Ezenarro J.J.
• Mas J.
• Muñoz-Berbel X.
• Uria N.
Advances in bacterial concentration methods and their integration in portable detection platforms: A review.
,
• Vijayaragavan K.
Virus purification, detection and removal.
]. Similarly, immunomagnetic separation is useful for both purifying and concentrating, but it is limited to small volumes and yields low concentration efficiencies [
• Ezenarro J.J.
• Mas J.
• Muñoz-Berbel X.
• Uria N.
Advances in bacterial concentration methods and their integration in portable detection platforms: A review.
]. On the other hand, filtration is a physical approach that relies on the size of the target pathogen and the filter membrane. It has been used for water sampling [
• Health Organization World
Guidelines for drinking-water quality: fourth edition incorporating first addendum.
], food-borne pathogen detection [
• Ezenarro J.J.
• Mas J.
• Muñoz-Berbel X.
• Uria N.
Advances in bacterial concentration methods and their integration in portable detection platforms: A review.
,], and processing in the dairy industry [,

GEA Process Engineering, “Membrane Filtration in the Dairy Industry.” https://www.lenntech.com/Data-sheets/GE-Osmonics-membrane-filtration-dairy-industry-L.pdf

]. Furthermore, filtration surpasses the other two techniques in that it is simple, low cost, versatile, capable of processing large and small sample volumes, and comes in readily available disposable packaging that can be implemented in an automated system.
Previous work studied the feasibility of filtration, specifically dead-end filtration, via the use of syringe filters, as a means for sample purification and concentration. Zhang et al. [
• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
] tested E. coli purification and concentration using varying pore-size syringe filters and filter membrane materials. With a one-step, manual filtration approach for purification, they were able to recover up to 91% of bacteria. Additionally, they achieved up to a 10-fold concentration of bacteria. Isabel et al. [
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
] also looked at varying syringe filter materials and pore sizes for bacteria (Bacillus spores) purification and concentration from powdery and environmental matrices. They applied a dual-filter approach, which they termed dual-filter method for applied recovery of microbial particles from environmental and powdery samples, or “DARE”, consisting of filtration through a 5-µm pore size syringe filter followed by retention of spores with a 0.45 µm syringe filter for concentration by resuspension of bacteria on the filter surface with a desired volume. This manual procedure processed a sample in 2 min with an average recovery of 51±17% for the bacterial spores. Though a reasonable processing time was reported for this procedure, much like the previous study described [
• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
], there is still a significant amount of labor associated with the manual method, particularly when processing a larger batch of samples. In this work, we investigated the applicability of a dual-filter filtration method, as studied previously [
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
], by further optimizing the flow parameters and automating the filtration and collection stages for minimal labor and handling.
We tested the automated dual-filter filtration system with samples containing E. coli, as our target pathogen and interferents in the form of fluorescent polystyrene beads. E. coli was chosen since it is a bacterium that has been used in many other studies as an airborne pathogen [
• Jing W.
• et al.
Microfluidic device for efficient airborne bacteria capture and enrichment.
] and a pathogen for water-borne and food-borne illnesses [
• Chang J.
• et al.
Ultrasonic-assisted self-assembly of monolayer graphene oxide for rapid detection of Escherichia coli bacteria.
]. The research presented here characterizes the applicability of an automated syringe filter system for the purification and concentration of bacteria to allow for an integrable filtration system that could be combined with a FET for quick, real-time downstream detection.

## 2. Materials and methods

### 2.1 Automated dual-filter method for applied recovery (aDARE)

Fig. 1 illustrates the experimental setup. Filters for aDARE were 25 mm diameter, sterile, hydrophilic polyvinylidene fluoride (PVDF) syringe filters with pore sizes of 0.22 µm (MilliporeSigma, Burlington, MA) and 5 µm (MilliporeSigma, Burlington, MA). Two 3-way rocker valves (Darwin Microfluidics, Paris, FR) controlled the flow direction with three syringe pumps (Harvard Apparatus, Holliston, MA) controlling the flow rate. Ethyl vinyl acetate plastic tubing was used with an inner diameter of 0.02″ and an outer diameter of 0.06″ to minimize the dead volume (McMaster-Carr, Elmhurst, IL). Tubing was connected to the filters using 23-gauge syringe needles (McMaster-Carr, Elmhurst, IL), polypropylene male Luer adapter T-junctions (Cole-Palmer, Vernon Hills, IL), and male-to-male Luer lock adapters (Cole-Palmer, Vernon Hills, IL). Additionally, we used valve fittings for 1/16″ OD tubing (Cole-Palmer, Vernon Hills, IL) to connect the tubing to the valves. To automate the system, all three syringe pumps and valves were controlled using a custom LabVIEW program (National Instruments, Austin, TX) using a 16-bit DAQ (USB 6002, National Instrument, Austin, TX). All pumps were operated for all experiments at 2 mL/min. To avoid cross-contamination between different runs, the valves and tubing were washed with 10 mL of 70% ethanol, followed by 10 mL deionized (DI) water, and then primed with 2 mL of calcium and magnesium free 1 $×$ PBS (Corning; Corning, NY), while the filters were manually replaced for each run.
During the automation, the system undergoes two different flow configurations, as shown in Fig. 1A. The first configuration (red arrows) is the purification and capturing step of the system. Syringe pump 1 forces the sample, of 5 mL volume, containing the target (at 107 CFU/mL for all experiments) and any interferents (at concentrations described later) through the 5 µm filter (to eliminate larger-sized particles) and subsequently through the 0.22 µm filter using valves to route flow as shown, where the bacteria are captured on the second filter membrane's surface (termed retentate). Next, syringe pump 2 forces 1 $×$ PBS buffer, at 4 mL volume, through the same configuration in order to rinse bacteria from tubing, valves, and the 5 µm filter.
For the second configuration (blue arrows), the valves switch the flow direction as shown. Syringe pump 3 (backflow pump) forces varying backflow volume of 1 $×$ PBS through the 0.22 µm pore size filter for retentate recovery into an Eppendorf collection tube for downstream processing. The final retentate volumes in the collection tubes were determined by weighing them with a sensitive mass balance.
A first set of experiments, termed “backflow optimization experiments”, were performed to discover how the backflow volume affects the concentration of bacteria collected from aDARE. From these experiments, we determined the optimal backflow volume to use for subsequent experiments. Syringe pump 3 was filled with 3 mL of 1 $×$ PBS. As syringe pump 3 dispensed its contents, it paused intermittently, in 300 µL increments for collection. The retentate recovery tube was replaced after each 300 µL was eluted, resulting in 10 consecutive equal tubes collected for a total backflow volume of 3 mL. Backflow volume was thus aliquoted into separate retentate recovery tubes which could be analyzed either individually (e.g., tube 1, tube 2, etc., or correspondingly 0.3 mL, 0.6 mL, etc.) or cumulatively (e.g., tube 1, tube 1+2, tube 1+2+3, etc. or again correspondingly 0.3 mL, 0.6 mL, 0.9 mL). The bacterial count and concentration in each tube were determined by qPCR as described below. The exact volumes in the collection tubes were determined by weighing them with a mass balance to account for volume variation during tube swapping.
A second set of experiments were performed to determine how much aDARE can enrich bacteria concentration in the presence of interferents. For these experiments, we filled syringe pump 3 with the optimal backflow volume determined from the previous experiments. As syringe pump 3 dispensed its contents, it was paused once, after a dead volume of 150 µL was expelled from the system (calculated as the total tubing and valve volume between the 0.22 µm filter and collection tube). This dead volume was discarded, to avoid dilution with pathogen-free fluid. Then, the remaining volume from syringe pump 3 was collected into the retentate recovery tube and analyzed.

### 2.2 Bacteria preparation

We used the bacterial pathogen Escherichia coli (E. coli) (ATCC 25922), a cylindrical bacteria with a length of 1-2 µm and a diameter of 0.5-1 µm [
N. R. C. (US) S. G. for the W. on S. L. of V. S. Microorganisms
Correlates of smallest sizes for microorganisms.
,

“bacteria - Diversity of structure of bacteria | Britannica.” https://www.britannica.com/science/bacteria/Diversity-of-structure-of-bacteria [Accessed 30 January 2022].

]. The bacteria were propagated per the vendor's instructions and grown to concentrations of approximately 109 CFU/mL, evaluated using OD600. A cell washing procedure was used to resuspend the cells in 1 $×$ PBS and the cells were stored in a refrigerator (3-4˚C) prior to use.
Prior to experiments, the PBS-washed E. coli was briefly vortexed to mix and break up any pellets and then diluted in 10-fold dilutions using 1 $×$ PBS, plated on BHI agar plates (Hardy Diagnostics, Santa Maria, CA) and incubated at 37 ˚C for 24 hours to grow single colonies [
• Sanders E.R.
Aseptic laboratory techniques: plating methods.
]. Quantification from plating informed the starting concentration of the bacteria prior to the experiment which allowed us to accurately dilute the bacteria down to a desired starting concentration and volume using 1 $×$ PBS.

### 2.3 Quantitative polymerase chain reaction (qPCR)

To quantify the number of copies and concentration of bacteria in the collection tube, we performed qPCR against known copies from a standard curve. Bacteria DNA was extracted and isolated using the E.Z.N.A Bacterial DNA kit (Omega Bio-Tek, Norcross, GA) according to the kit's instructions. Typically, 1 mL per sample underwent extraction, and the DNA was concentrated to a final volume of 150 µL in elution buffer. To prepare the DNA samples for the standard curve of the qPCR runs, DNA from an unknown concentration of bacteria was extracted, and the Denovix dsDNA High Sensitivity Assay (Wilmington, DE) was used, per the manufacturer's instructions.
The PCR analysis for all samples was carried out in the StepOnePlus real-time system (Applied Biosystems) using specific primers and probe for targeting the 16S gene in E. coli (226-kbp genome), as has been done previously [
• Huijsdens X.W.
• Linskens R.K.
• Mak M.
• Meuwissen S.G.M.
• Vandenbroucke-Grauls C.M.J.E.
• Savelkoul P.H.M.
Quantification of bacteria adherent to gastrointestinal mucosa by real-time PCR.
,
• Smati M.
• et al.
Real-Time PCR for quantitative analysis of human commensal escherichia coli populations reveals a high frequency of subdominant phylogroups.
]. The forward primer [5’-CAT GCC GCG TGT ATG AAG AA-3’], reverse primer [5’ -CGG GTA ACG TCA ATG AGC AAA- 3’], and probe [5’- /56-FAM - TA TTA ACT T/ZEN/T ACT CCC TTC CTC CCC GCT GAA /3IABkFQ/ - 3’] were manufactured by Integrated DNA Technologies, Inc. (Coralville, IA).
The PCR mix consisted of 5 µL of 5 $×$ PerfeCta MultiPlex qPCR ToughMix (QuantaBio, Beverly, MA), 0.5 µL of each primer (at 25 µM), 0.5 µL of the probe (10 µM, 5 µL of the DNA template, and 13.5 µL of nuclease-free water to make up a 25 µL total volume per well. The samples were placed in an Applied Biosystem Fast Optical 96-well, 0.1 mL reaction plate (Fisher Scientific, Waltham, MA). All runs contained a negative control of nuclease-free water and a positive control consisting of the E. coli standard DNA templates. The samples then underwent the following thermal conditions: 10 min at 95 ˚C, followed by 45 cycles, with each cycle consisting of 10 sec at 95 ˚C and 30 sec at 60 ˚C. The PCR analysis was performed in technical triplicates per sample, unless stated otherwise.

### 2.4 Interferents

Fluorescent polystyrene beads (Fluoro-Max Dyed Aqueous Fluorescent, Thermo Fisher Scientific, Waltham, MA) of sizes 2 µm and 10 µm with excitation/emission peaks of 412/473 nm and 468/508 nm, respectively, were used to model non-target particles that could be present in real-world samples. The beads were each diluted with 1 $×$ PBS to a final concentration of 106 beads/mL, representative of a high interferent concentration [
• Seinfeld J.H.
• Pandis S.N.
Atmospheric chemistry and physics: from air pollution to climate change.
] when loaded into the sample syringe pump 1. The beads were prepared without surfactants to avoid cell lysis. The lack of surfactant also allowed bead agglomeration as in real-world samples.

### 2.5 Flow cytometry

Fluorescent polystyrene beads were quantified with flow cytometry (Aurora, Cytek; Fremont, CA). Prior to analysis, we added a commercial surfactant; Triton X-100 (Sigma-Aldrich, St. Louis, MO) to our bead solutions with a final concentration of 0.5% to ensure the beads were monodispersed. A 96-well, clear, round bottom plate (Corning; Corning, NY) was used to house 100-150 µL of the sample.

### 2.6 Metrics and analysis

The system performance was evaluated using concentration factor, recovery, and enrichment ratio as follows:
$Concentrationfactor=ConcentrationoutputConcentrationinput$
(1)

$Recovery(%)=NumberoftargetparticlesinoutputNumberoftargetparticlesininput×100%=(Targetconcentration×Targetvolume)output(Targetconcentration×Targetvolume)input×100%$
(2)

$Enrichmentratio=(targetconcentrationnontargetconcentration)output(targetconcentrationnontargetconcentration)input.$
(3)

## 3. Results

We first present the results of a backflow optimization to discover how the backflow volume affects the concentration of bacteria collected from aDARE. We then present the purification abilities of aDARE when input is mixed with interferents.

### 3.1 Backflow optimization

The results from the backflow optimization experiments are shown in Fig. 2. We performed two kinds of collection experiments to study the effectiveness of backflow collection. The first is a series of individual collection retentate tubes of 0.3 mL volume each, where each tube is evaluated for bacterial concentration and recovery (Fig. 2 A, B). Evaluating the individual tubes allows us to see precise changes in backflow effectiveness as fluid flows across the filter. The second experiment also involves a series of individual collection tubes of 0.3 mL, but the tubes are aggregated to allow for cumulative measurement of bacterial concentration and recovery, which simulates how the device will be run in practice (Fig. 2 C, D). The concentration factor and the recovery percentage were evaluated using Eqns. 1 and 2, respectively.
Fig. 2A shows the effect of the backflow volume on the concentration factor for each individually collected retentate tube. It demonstrates a very low concentration factor for the first 2 collected tubes followed by a significant increase in the concentration factor, ranging from 6 to 10 times its initial concentration, for the next 3 tubes (corresponding to a backflow volume of 900 µL up to 1.2 mL). Similarly, the individual recovery (Fig. 2B) showed a recovery percentage below 5% for the initial 600 µL, followed by a recovery percentage that exceeds 70% for the next two tubes. This finding indicates that the overall effect of adding the initial backflow volume to the final collection volume is unfavorable. Thus, it can be eliminated in applications where high concentration factors are desirable. This result can be explained by the existence of a dead volume in the system that is collected in the process prior to recovering any bacteria, in addition to the fact that the bacteria captured on the membrane needs to experience backflow first before dislodging from the membrane.
Fig. 2C shows the cumulative concentration factor resembling the continuous collection of retentate without tubes swapping. The concentration factor exhibits a concave shape where the highest concentration factor of approximately four corresponds to volumes in the range of 900 µL to 1.2 mL. This is followed by a drop in the concentration factor for the larger volumes, thus indicating further dilution of the sample with pathogen-free backflow volume. Fig. 2D shows the cumulative recovery percentage, which appears sigmoidal, showing a low recovery for the initial 600 µL of backflow. This is followed by a rapid increase in the recovery in a nearly linear manner that continues until a volume of 1.5 mL, by which most of the captured bacteria have been recovered from the filter. Beyond the 1.5 mL volume, the recovery exhibits a plateau regardless of the infused backflow volume. We conclude a four-fold concentration factor can be achieved for a backflow volume between 900 µL and 1.5 mL.

### 3.2 Purification

After determining the optimal retentate recovery volume of 900 µL, we repeated the cumulative retentate recovery experiments of Fig. 2 A, B using only that volume, both with and without the presence of interferents (beads). Thus, we loaded 107 CFU/mL of E. coli into the sample syringe pump 1 along with 106 beads/mL of 2 µm beads and 106 beads/mL of 10 µm beads. Fig. 3 shows the resulting concentration factor and recovery percentage. As in Fig. 2 C, D, without beads present, we again were able to achieve a four-fold concentration factor (Fig. 3A). Specifically, a starting concentration of $6.89×103±5.48×102$copies/µL was increased to $3.13×104±1.82×103$ copies/uL (Welch's t-test p < 0.0001), and a recovery percentage (Fig. 3B) of 68 ± 17% was achieved. Upon addition of interferents, we observed a reduction in the concentration factor from 3.7 ± 0.95 to 2.1 ± 0.35 and the recovery percentage from 68 ± 17% to 40.2 ± 7.3% (n = 3).
Using flow cytometry, the interferents’ recovery was evaluated to be 1.1 ± 0.29 % and 0.006 ± 0.001% for the 2-µm and 10-µm beads, respectively, indicating that most of the interferents were successfully removed. Thus, the enrichment ratio of the target relative to the interferents, calculated using Eqn. 3, was 42 ± 13.

## 4. Discussion

Using aDARE, a cumulative concentration factor of approximately 4x and 69% recovery were achieved for backflow volumes in the range of 0.9-1.5 mL. Depending on the specific sensor or application where aDARE is used, it may be more advantageous to collect the cumulative retentate (up to 1.5 mL as described), in order to maximize the number of target bacteria collected without significant dilution. Or one may choose instead to collect only the third 300 µL sample (as described in the Methods section) in order to maximize the concentration factor.
We speculate, but did not test, that the optimal backflow volume is independent of the sample volume but depends on the pathogen, filter material, and diameter in addition to the system's dead volume. Below we discuss the effects of these different materials. We suggest that subsequent researchers calibrate their systems accordingly. For example, in a different study, Zhang et al. filtered the same E. coli strain, diluted in urine, using a 0.22-µm pore size filter but with smaller filter diameter, [
• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
]. They were able to concentrate the bacteria 10-fold over a range of starting concentrations. Although this concentration factor exceeds ours by more than a factor of two, it is important to highlight that they used different starting concentrations, larger sample volumes, and different solvents.
Our primary method for optimization was using the backflow volume. However, we conducted some testing on the effect of filter diameter. A 25 mm diameter filter was chosen to allow for higher flow rates []. It has been hypothesized in other work that smaller membrane diameters enhance bacterial retentate recovery due to less cross-sectional area [
• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
]. Yet, we found there to be no statistically significant difference between the E. coli recovery when using a 13 mm diameter versus a 25 mm diameter. However, in our experiments (not shown) we observed that the 25 mm diameter did allow for a higher overall recovery percentage. Thus, these results further supported our choice of using 25 mm diameter filters.
Filter membrane material is of utmost importance. As shown by Zhang et al., certain membranes, such as cellulose acetate, are not compatible for E. coli filtration [
• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
]. Work by Isabel et al. further confirms the compatibility of PVDF syringe filters for E. coli filtration and recovery [
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
]. Our results further support the use of hydrophilic PVDF membranes. Although Isabel et. al observed an approximate recovery of 60% of their spores using 0.45-µm filters [
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
], our work used 0.22-µm pore size filters to ensure E. coli would get stuck inside the membrane pores. Other work has shown that 0.22-µm filters can achieve up to 66% bacterial recovery [
• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
] which matches the results we obtained.
Upon addition of interferents to the input sample, we observed a reduction in the concentration factor from 3.7 ± 0.95 to 2.1 ± 0.35 and the recovery percentage from 68 ± 17% to 40.2 ± 7.3%, with more than 95% of interferents removed. Thus, aDARE yielded an enrichment ratio increase of 42 ± 13-fold. Isabel et al., manually purified B. atrophaeus subsp. globigii spores under different powdery conditions, recovering 51 ± 17% of the bacterial spores [
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
]. In comparison, we recovered 40.2 ± 7.3%, about 10% less on average. Yet, it is important to recall that the two works involved different interferents, target pathogens and quantification methods. Since target pathogens for both works had comparable diameters and shapes, we suspect factors like the type and amount of interferent used as well as the handling process are responsible for the results. Nevertheless, our work shows that the physical purification and concentration approach using syringe filters can be successfully automated.
The reduced bacteria recovery upon the addition of interferents could be interpreted as the bacteria's interaction with a highly agglomerated interferent matrix, especially since the bead suspension contained no surfactant to keep beads monodisperse. This agglomeration could lead to increase caking or fouling on the filter's membrane. Another cause for reduced and variable recovery is the gram-negativity of our bacterial pathogen of choice, E. coli, which makes the bacteria deformable and potentially able to pass through pores of smaller sizes than its size. More deformable organisms can construct a more compact and higher resistance cake leading to a filtration flux decrease and potentially lower recovery [
• Charcosset C.
3 - Microfiltration.
]. Moreover, cultured bacteria can contain cell-free genomic DNA that can be detected by PCR, but due to its small size, they are lost through the second filter. Thus, this free-flowing DNA is never collected for DNA quantification which can potentially lead us to underestimate the achieved recovery percentages and concentration factor.
Future workers may consider removing possible caking and fouling from the membrane surface via pulsing and adding a small percentage of surfactant, such as 0.002% [
• Charcosset C.
3 - Microfiltration.
], to encourage E. coli detachment from the filter membrane by lowering the interfacial tension between the membrane surface and attached cells. Furthermore, optimization could be performed with respect to the concentration of bacteria and/or interferents. Beads could be substituted with interferents more representative of the real-world interferents such as dust and pollen as well as the bacteria type could be changed. Optimization tests may also be conducted to study the reusability of the syringe filters. The automated system could become further standalone by including reservoirs that will be used for cleaning in between runs.

## 5. Conclusion and future work

We presented aDARE, an automated dual-filter bacterial purification and concentration approach that helps reduce the time-consuming and labor-intensive manual filtration procedures involving syringe filters used in labs. Prior to testing the fully automated system with a 5-mL sample input, we optimized the backflow volume for bacterial retentate recovery and concentration. The use of two syringe filters with different pore sizes was adapted from the work by Isabel et al. [
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
], enhanced, and automated for our E. coli samples. With a sample containing bacteria and fluorescent polystyrene beads (used to mimic interferents likely to be found in real-world samples) at a concentration ratio of 10:1, respectively, we found that aDARE concentrates a 5-mL input sample of bacteria by a factor of two, recovers on average 40% of the bacteria, removes over 95% of the interferents, and yields a post-automation volume of 958 ± 8.5 µL. In 5.5 minutes, at a flow rate of 2 mL/min, the results yielded an enrichment factor of 42 ± 13-fold (n = 3).
Our automated physical size-based approach for the separation, recovery, and concentration of bacteria from environmental interferents for detection has demonstrated promise. In the future, the automated dual filter system could be further miniaturized and integrated with a FET. Until then, aDARE has shown to be a promising option for quickly purifying and concentrating a contaminated sample for a target pathogen to allow for quick, less labor-intensive, real-time detection of target pathogens.

## Disclaimer

The views, opinions and/or findings expressed are those of the authors and should not be interpreted as representing the official views or policies of the Department of Defense or the U.S. Government. Approved for Public Release, Distribution Unlimited.

## Conflicts of interest

There are no conflicts of interest to declare.

## Author's contribution

Nina Sara Fraticelli Guzmán: conceptualization, experimental work, validation, data analysis, investigation, writing and editing manuscript. Mohamed Badawy: conceptualization, experimental work, data analysis, editing. Max A. Stockslager: conceptualization, experimental design, editing. Caitlyn van Zyl: experimental work. Michael L. Farrell: conceptualization, editing. Seth Stewart: conceptualization, editing. David L. Hu: conceptualization, editing. Craig R. Forest: conceptualization, editing.

## Declaration of Competing Interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Mike L. Farrell reports financial support was provided by Defense Advanced Research Projects Agency.

## Acknowledgements

This research was funded by Georgia Tech's Research Institute (GTRI) IRAD program, as well as DARPA grant program BIOAEROSOL COLLECTOR FOR SARS-COV-2 DETECTION, under Agreement No. HR00112190060 (project D9604).
The authors would like to thank Stephanie Ritch, Jared Beyersdorf, and Dr. Phil Santangelo for helpful discussion and insight for this project. We would also like to thank Stephanie Ritch and Dr. Jie Xu from the Food Processing Technology Division at Georgia Tech for their help preparing the bacteria for our experiments.

## References

1. “PCR Test for COVID-19: What It Is, How Its Done, What The Results Mean,” Cleveland Clinic. https://my.clevelandclinic.org/health/diagnostics/21462-covid-19-and-pcr-testing [Accessed 4 November 2021].

2. “Graphene Field Effect Transistors for Biological and Chemical Sensors.” https://www.sigmaaldrich.com/US/en/technical-documents/technical-article/materials-science-and-engineering/organic-electronics/graphene-field-effect-transistors [Accessed 4 November 2021].

• Purwidyantri A.
• et al.
Sensing performance of fibronectin-functionalized Au-EGFET on the detection of S. epidermidis biofilm and 16S rRNA of infection-related bacteria in peritoneal dialysis.
Sens Actuators B. 2015; 217: 92-99https://doi.org/10.1016/j.snb.2014.11.017
• Formisano N.
• et al.
Inexpensive and fast pathogenic bacteria screening using field-effect transistors.
Biosens Bioelectron. 2016; 85: 103-109https://doi.org/10.1016/j.bios.2016.04.063
• Shen F.
• et al.
Integrating silicon nanowire field effect transistor, microfluidics and air sampling techniques for real-time monitoring biological aerosols.
Environ Sci Technol. 2011; 45: 7473-7480https://doi.org/10.1021/es1043547
• Thakur B.
• et al.
Rapid detection of single E. coli bacteria using a graphene-based field-effect transistor device.
Biosens Bioelectron. 2018; 110: 16-22https://doi.org/10.1016/j.bios.2018.03.014
• Chang J.
• et al.
Ultrasonic-assisted self-assembly of monolayer graphene oxide for rapid detection of Escherichia coli bacteria.
Nanoscale. 2013; 5: 3620https://doi.org/10.1039/c3nr00141e
• Tarasov A.
• et al.
A potentiometric biosensor for rapid on-site disease diagnostics.
Biosens Bioelectron. 2016; 79: 669-678https://doi.org/10.1016/j.bios.2015.12.086
• Zhang Y.
• Xu C.
• Guo T.
• Hong L.
An automated bacterial concentration and recovery system for pre-enrichment required in rapid Escherichia coli detection.
Sci Rep. 2018; 8: 17808https://doi.org/10.1038/s41598-018-35970-8
• Ezenarro J.J.
• Mas J.
• Muñoz-Berbel X.
• Uria N.
Advances in bacterial concentration methods and their integration in portable detection platforms: A review.
Anal Chim Acta. 2021; 339079https://doi.org/10.1016/j.aca.2021.339079
3. A. A. Fatah, J. A. Barrett, R. D. Arcilesi, K. J. Ewing, C. H. Lattin, and T. F. Moshier, “An introduction to biological agent detection equipment for emergency first responders,” p. 53.

• Walper S.A.
• et al.
Detecting biothreat agents: from current diagnostics to developing sensor technologies.
ACS Sens. 2018; 3: 1894-2024https://doi.org/10.1021/acssensors.8b00420
• Vijayaragavan K.
Virus purification, detection and removal.
Diss Masters Theses Masters Rep - Open. 2014; https://doi.org/10.37099/mtu.dc.etds/850
• Health Organization World
Guidelines for drinking-water quality: fourth edition incorporating first addendum.
4th ed+1st add. World Health Organization, Geneva2017 ([Accessed 4 November 2021]. [Online]. Available)
4. “Food Pathogen Concentration Techniques.” https://www.rapidmicrobiology.com/test-method/separation-and-concentration-of-microorganisms-from-food-matrices [Accessed 4 November 2021].

5. GEA Process Engineering, “Membrane Filtration in the Dairy Industry.” https://www.lenntech.com/Data-sheets/GE-Osmonics-membrane-filtration-dairy-industry-L.pdf

• Zhang P.
• Kaushik A.M.
• Mach K.E.
• Hsieh K.
• Liao J.C.
• Wang T.-H.
Facile syringe filter-enabled bacteria separation, enrichment, and buffer exchange for clinical isolation-free digital detection and characterization of bacterial pathogens in urine.
Analyst. 2021; 146: 2475-2483https://doi.org/10.1039/D1AN00039J
• Isabel S.
• et al.
Rapid filtration separation-based sample preparation method for bacillus spores in powdery and environmental matrices.
Appl Environ Microbiol. 2012; 78: 1505-1512https://doi.org/10.1128/AEM.06696-11
• Jing W.
• et al.
Microfluidic device for efficient airborne bacteria capture and enrichment.
Anal Chem. 2013; 85: 5255-5262https://doi.org/10.1021/ac400590c
• N. R. C. (US) S. G. for the W. on S. L. of V. S. Microorganisms
Correlates of smallest sizes for microorganisms.
National Academies Press (US), 1999 ([Accessed 4 November 2021]. [Online]. Available)
6. “bacteria - Diversity of structure of bacteria | Britannica.” https://www.britannica.com/science/bacteria/Diversity-of-structure-of-bacteria [Accessed 30 January 2022].

• Sanders E.R.
Aseptic laboratory techniques: plating methods.
J Vis Exp. 2012; : e3064https://doi.org/10.3791/3064
• Huijsdens X.W.
• Linskens R.K.
• Mak M.
• Meuwissen S.G.M.
• Vandenbroucke-Grauls C.M.J.E.
• Savelkoul P.H.M.
Quantification of bacteria adherent to gastrointestinal mucosa by real-time PCR.
J Clin Microbiol. 2002; 40: 4423-4427https://doi.org/10.1128/JCM.40.12.4423-4427.2002
• Smati M.
• et al.
Real-Time PCR for quantitative analysis of human commensal escherichia coli populations reveals a high frequency of subdominant phylogroups.
Appl Environ Microbiol. 2013; 79: 5005-5012https://doi.org/10.1128/AEM.01423-13
• Seinfeld J.H.
• Pandis S.N.
Atmospheric chemistry and physics: from air pollution to climate change.
John Wiley & Sons, 2016
7. Corning, “Corning Filtration Guide.” https://www.corning.com/catalog/cls/documents/selection-guides/t_filterselectionguide.pdf

• Charcosset C.
3 - Microfiltration.
in: Charcosset C. Membrane processes in biotechnology and pharmaceutics. Elsevier, Amsterdam2012: 101-141https://doi.org/10.1016/B978-0-444-56334-7.00003-4