Science of the Total Environment 914 (2024) 169932

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Influence of phosphate on bacterial release from activated carbon

point-of-use filters and on biofilm characteristics
Gemma G. Clark a, *, Dietrich Geisler b, Evan J. Coey a, Lance J. Pollitz a, Farzana R. Zaki c,
Conghui Huang a, Stephen A. Boppart c, d, e, Thanh H. Nguyen a, e, f
a
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, United States of America
b
Department of Computer Science, Cornell University, United States of America
c
Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, United States of America
d
Department of Bioengineering, Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, United States of America
e
Carle Illinois College of Medicine, United States of America
f
Institute of Genomic Biology, University of Illinois at Urbana-Champaign, United States of America

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Phosphate filters released fewer bacteria

than groundwater filters.
• Phosphate biofilms were thicker and
rougher than groundwater biofilms.
• Overall porosity did not differ between
groundwater and phosphate biofilms.
• Phosphate biofilms had fewer pores per
volume of biofilm than groundwater
biofilms.
• Phosphate biofilms had shorter pore-
connecting channels than groundwater
biofilms.

A R T I C L E I N F O A B S T R A C T

Editor: Kyle Bibby Point-of-use (POU) filters certified to remove lead are often composed of activated carbon and have been shown
to release high concentrations of bacteria, including opportunistic pathogens. In this study, we examine the
Keywords: impacts of the common corrosion inhibitor phosphate on biofilm characteristics and the relationship between
Point-of-use filter biofilm structure and bacterial release from POU filters. This knowledge is essential for understanding how best
Biofilm
to use the filters and where these filters fit in a system where other lead contamination prevention measures may
Bacteria
be in place. We measured the bacterial release from activated carbon POU filters fed with groundwater - a
Phosphate
Drinking water common source of drinking water - with and without phosphate. We used optical coherence tomography (OCT)
to quantitatively characterize biofilm growing on activated carbon filter material in which the biofilms were fed
groundwater with and without phosphate. Phosphate filters released significantly less (57–87 %) bacteria than
groundwater filters, and phosphate biofilms (median thickness: 82–331 μm) grew to be significantly thicker than
groundwater biofilms (median thickness: 122–221 μm). The phosphate biofilm roughness ranged from 97 to 142

* Corresponding author.
E-mail address: [email protected] (G.G. Clark).

https://doi.org/10.1016/j.scitotenv.2024.169932
Received 28 August 2023; Received in revised form 29 November 2023; Accepted 3 January 2024
Available online 8 January 2024
0048-9697/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-
nc/4.0/).
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

% of the groundwater biofilm roughness and was significantly greater in most weeks. Phosphate biofilms also
had fewer pores per biofilm volume and shorter channels connecting those pores.

1. Introduction previously been shown to affect the structure of biofilm grown on

polyvinyl chloride (PVC) (Huang et al., 2023). Our results show how
As drinking water infrastructure continues to age and degrade, phosphate affects biofilm characteristics of activated carbon biofilm and
commercially available point-of-use (POU) filters, such as those made by how these differences affect the release of bacteria from activated car­
Brita and PUR, have become increasingly popular as a temporary solu­ bon POU filters.
tion to address drinking water quality problems. Major concerns about
lead and other drinking water contaminants have contributed to 77 % of 2. Material and methods
Americans filtering their drinking water at home (Carollo, 2022). POU
filters certified to remove lead are often composed of activated carbon 2.1. Measuring bacteria release from water bottle filters
and have been shown to increase bacteria in drinking water (Chaidez
and Gerba, 2004; Clark et al., 2022; Nriagu et al., 2018; Tobin et al., We measured the concentrations of bacteria entering and exiting
1981; Wu et al., 2017). While not all bacteria are harmful, previous Brita water bottle filters (Brita Model BB10) composed of porous solid
studies have shown that opportunistic pathogens can break through block activated carbon (Fig. S1a-b), much like those used in larger faucet
POU filters (Geldreich et al., 1985; Molloy et al., 2007; Wu et al., 2021). filters by Brita and other POU filter companies. The total volume of the
In places where lead is a concern, corrosion inhibitors such as ortho­ filter and water bottle cartridge was 750 mL as determined by a tracer
phosphates and polyphosphates are often applied to control the release test. The filter itself was a smaller cylinder (17.5 mm outer diameter, 8
of lead into drinking water (McNeill and Edwards, 2002). It is important mm inner diameter, 80 mm long) and housed in a hollow plastic straw
to understand how the usage of POU filters will pair with other lead within the water bottle. The flow in the water bottle filters was radial
contamination prevention measures concerning microbiological drink­ from the outside of the activated carbon cylinder to the inside, where
ing water quality. filtered water exited through the straw. Sixteen water bottle filters were
Biofilms are collections of microorganisms that accumulate and grow assembled with controllable influent: eight replicates were fed with
on almost any surface (Flemming, 2011; Wingender and Flemming, groundwater (groundwater filters), another eight with groundwater
2011) and are known to harbor and protect opportunistic pathogens containing 2 mg/L as PO4 of phosphate (phosphate filters) to simulate
(Cargill et al., 1992; Cooper and Hanlon, 2010; Falkinham, 2015; the use of a phosphate corrosion inhibitor (McNeill and Edwards, 2002).
Wingender and Flemming, 2011). Previous studies have shown that The groundwater without disinfectant treatment came from a well
increasing phosphorus concentrations can increase the growth of bac­ beneath Newmark Civil Engineering Laboratory (Urbana, IL) after being
teria and biofilms (Chu et al., 2005; Fang et al., 2009; Hozalski et al., passed through a greensand filter to remove iron precipitates. The
2005; Noh et al., 2020). Extracellular polymeric substances (EPS) are groundwater’s pH ranged from 7.4 to 7.7, alkalinity was typically
important for biofilm adhesion (Flemming and Wingender, 2010), and around 280 mg/L as CaCO3, and the water temperature ranged from 12
research has shown that adding phosphorus can decrease the production to 15 ◦ C, which is within the range of typical household cold water (Bors
of EPS (Douterelo et al., 2020; Fang et al., 2009; Noh et al., 2020). This et al., 2017). Additional water chemistry parameters for this ground­
may have concerning implications for the combination of using POU water, including total organic carbon averaging 1–2 mg/L as C, have
filters (which can exacerbate bacteria concentrations in drinking water) been reported in previous studies (Clark et al., 2022; Shen et al., 2017).
and adding phosphate corrosion inhibitors (which may increase bacte­ Sodium phosphate dibasic (Sigma-Aldrich) was mixed with ground­
rial growth and decrease bacterial adhesion to a filter). water at a final concentration of 2 mg/L as PO4 for the phosphate filters.
Numerous studies have examined biofilm growth on activated car­ The water bottle filters were operated at room temperature
bon as biological activated carbon can be intentionally used in (20–23 ◦ C), which falls within typical household heating and cooling
centralized drinking water treatment systems to remove organics (Gibert temperatures of 19–24 ◦ C (Booten et al., 2017). A peristaltic pump was
et al., 2013; Korotta-Gamage and Sathasivan, 2017; LeChevallier et al., used to pump fresh groundwater, or groundwater with 2 mg/L as PO4 of
1992; Stringfellow et al., 1993; Urfer et al., 1997; Weber et al., 1978). phosphate, through the filters at a flow rate of 6.5 mL/s for 2 min 3–5
Biological activated carbon has also been shown to release bacteria times per week. The flow rate corresponded to a Reynolds number of
(Servais et al., 1994; Stringfellow et al., 1993; Zhang et al., 2015), 1200 for water that flows over the inner surface of the activated carbon
especially when filtration velocity increases as high shear forces can cylinder (Text S1). This value is within the range of Reynolds numbers
slough bacteria off of a biofilm (Abe et al., 2012; Paul et al., 2012; Zhang calculated from filter flow rates observed in the field (Clark et al., 2022).
et al., 2015). However, no research has yet examined the characteristics At the beginning of each week, 100-mL samples of water entering and
of biofilms growing on POU filters. Biofilm characteristics, such as exiting each filter were collected before the two-minute filter flushing.
biofilm thickness, roughness, and pore structure, can influence bacterial The filter flushing acted to displace any stagnant water, introduce new
adhesion and detachment, biofilm growth, and nutrient transport within nutrients and bacteria from the groundwater, and simulate infrequent
the biofilm (Ammar et al., 2015; Carrel et al., 2018; Huang et al., 2020; but consistent filter usage with long stagnation periods that one might
Huang et al., 2023; Nerenberg, 2016; Picioreanu et al., 2000; Shen et al., observe in an office drinking fountain or break room sink.
2015; Stewart, 2003; Wu et al., 2012). Understanding the factors gov­ Aliquots of 700 μL from the 100-mL samples were taken to measure
erning bacterial release and biofilm growth on activated carbon is key to the total cell counts (TCC) and intact cell counts (ICC) from each sample
controlling bacterial release from these popular filters and protecting via flow cytometry (Sysmex Partec). TCC were measured by staining the
public health. samples with 1X SYBR Green (Invitrogen), and ICC were measured by
To fill this gap in research and understand how phosphorus may staining samples with 1X SYBR Green and 30 μM propidium iodide
affect biofilm characteristics and bacterial release from filters, we have (Invitrogen) (Berney et al., 2008; Hammes et al., 2008). Samples were
characterized the surface structure and the spatial pore structure of incubated in the dark at 37 ◦ C for 10 min between staining and mea­
biofilm grown on POU filters using optical coherence tomography (OCT) surement (Van Nevel et al., 2013). Concentrations of bacteria were
and compared these results to bacterial release from POU filters over quantified using the flow cytometer’s volumetric counting and the
time. Experiments were performed using groundwater – a common FloMax (Sysmex) software’s electronic gating (Hammes et al., 2008).
drinking water source – with and without phosphate, which has QA/QC calibration beads (Sysmex) were used to check the alignment of

2
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

the instrument’s lasers and the consistency of the instrument settings. collected.
Before sample measurements, the flow cytometer was cleaned with 10 %
bleach, nanopure water, and Sysmex Cleaning Solution. Nanopure water 2.3. Calculating biofilm thickness and roughness from OCT images
was used as a negative control and also used to rinse the instrument in
between samples. The limit of quantification was 20 cell counts/mL and A total of 3000 OCT images were analyzed in this study: 1500 from
the limit of detection was 5 cell counts/mL. groundwater biofilms and 1500 from phosphate biofilms. The 100 cross-
Every two weeks, each of the remaining volumes of the 100-mL sectional OCT images of each biofilm were preprocessed in ImageJ 2.3
samples after flow cytometry processing were filtered through a 0.2- (Fiji) using the following successive steps (Fig. S3):
μm cellulose nitrate filter (mdi Membrane Technologies). We used a
FastDNA SPIN Kit for Soil (MP Biomedicals) to extract DNA from the 1) We removed specular artifacts caused by the light from the super­
filter’s surface and performed qPCR on the DNA extract targeting the luminescent diode reflecting off water droplets at the biofilm surface.
16S rRNA universal bacteria gene (Table S1, Text S2) to complement the 2) Using the segmented line tool, we marked the interface between the
flow cytometry results. Primers and standard sequences were taken from activated carbon surface and the biofilm with pure black (R: 0, G: 0,
Shen et al., 2017. Every qPCR plate contained a standard curve made B: 0) as shown in Fig. S2. Biofilm was differentiated from activated
from triplicate serial dilutions of the 16S gene standard and triplicate carbon by comparing images of samples with biofilm to control im­
negative controls. Triplicates of each sample were measured. Each re­ ages of fresh activated carbon without biofilm.
action well contained 15 μL of solution containing 2 μL of the sample, 3) We used the thresholding function in ImageJ to produce a binary
standard, or negative control; 4.3 μL of molecular grade water, 0.6 μL of black-and-white image that distinguishes the surface of the biofilm
10 μM forward primer, 0.6 μL of 10 μM reverse primer, and 7.5 μL of 2X (white) from the air (black) above.
PowerUp SYBR Green (Applied Biosystems). The limit of quantification
was 5 gene copies (gc) per mL sample, and bacteria were not detected in After each successive step, we exported each slice in the image stack
any of the negative controls. The qPCR efficiency ranged from 92 % to to a single file. A MATLAB code (Derlon et al., 2012; Folmli, 2011) for
97 % (R2 = 0.989–0.999). analyzing OCT images of biofilm was translated into Python and upda­
ted to process our image stacks (Fig. S3). The updated Python code was
2.2. Growing biofilm in reactors to acquire biofilm characteristics verified by processing a subset of images with the MATLAB code and
Python code. The difference between the results of the two codes
In parallel with the water bottle setup, CDC biofilm reactors (Bio­ averaged <10− 12 μm which is well below the resolution of the OCT
Surface Technologies) were used to grow biofilm on slices of the acti­ imaging system.
vated carbon water bottle filters. The biofilm reactors with sliced filter The new Python code read in all images produced in each of the three
material were used so that the biofilm could be characterized at regular successive steps above. Although thresholding was done manually with
time intervals without sacrificing the entire filter and so that the filter ImageJ in the third step, an auto-thresholding function was an option in
apparatus would not have to be deconstructed - and therefore the bio­ the code. The biofilm interface with the air was detected from the binary
film disturbed - in preparation for the characterization methods. Water image. For each pixel column in the image, the biofilm thickness was
bottle filters (identical to those used in the water bottle setup in Section calculated by subtracting the marked biofilm‑carbon interface from the
2.1) were cut into 3–5 mm thick slices and glued to polycarbonate biofilm-air interface. Finally, the average biofilm thickness (Eq. 1),
coupons that were inserted into polypropylene rods (Fig. S1c). The re­ roughness (Eq. 2), and relative roughness (Eq. 3) were calculated for
actors were autoclaved, and then the activated carbon slices were sub­ each image slice using the following equations obtained from previous
merged in groundwater or groundwater containing 2 mg/L as PO4 of studies (Derlon et al., 2012; Janjaroen et al., 2013; Picioreanu et al.,
phosphate. Autoclaving the activated carbon filters may have opened 1998; Shen et al., 2018):
more pores in the filter, as steam does during the activation of producing
1∑n
activated carbon. Although the polymers holding the activated carbon Average biofilm thickness = z = zi (1)
n i=1
block together may have been heat sensitive, the filter did not break or
change in any form detected by optical coherence tomography (OCT). 1∑n
Absolute biofilm roughness = Ra = (|zi − z|) (2)
Fresh groundwater with or without phosphate was pumped into the n i=1
reactors from 10 L tanks that were replenished with fresh influent every ( )
2–3 days. The reactors contained a 45 mm agitator that was set to 40 1∑n |zi − z|
Relative biofilm roughness = Ra′ = (3)
rpm, which corresponds to a Reynolds number of 1200 for water flowing n i=1 z
over the surface of the activated carbon slices (Text S1). The matching
Reynolds number in the biofilm reactors and the water bottle filters was where n is the number of pixel columns (thickness measurements
used to standardize the approximate flow that would pass over the counts), zi is the biofilm thickness for the ith pixel column (μm), and z is
biofilm surface. These reactors were wrapped in aluminum foil to limit the mean biofilm thickness (μm).
algal growth and photosensitive reactions and operated at room tem­ The biofilm thickness produced by this code was the optical biofilm
perature (20–23 ◦ C). thickness. To calculate the physical biofilm thickness, we divided by the
Optical coherence tomography (OCT) is a noninvasive imaging refractive index 1.3, which was previously determined for biofilms
technique that has previously been used to characterize biofilm struc­ (Huang et al., 2020). Both the absolute and relative biofilm roughness
ture and development (Huang et al., 2020; Huang et al., 2023; Janjaroen were calculated from the profile of the biofilm’s surface and were
et al., 2013; Nguyen et al., 2013; Nguyen et al., 2010; Shen et al., 2016; therefore not affected by the refractive index.
Shen et al., 2015; Xi et al., 2006). At various time intervals (2, 4, 5, 6, 7, The Python code produced a set of images which were used to verify
8, 11, and 15 weeks), a slice of activated carbon was removed from each that the biofilm was correctly identified by the code. This included a set
reactor fed with groundwater or groundwater with phosphate. Each of images marking the location of the biofilm surface (biofilm-air
week that slices were removed from the reactors, three sets of 100 cross- interface), and a set of black-and-white images in which the biofilm was
sectional images of the biofilm were taken using OCT (Fig. S2) for each white and everything else was blacked out including the air above and to
biofilm type (groundwater biofilms or phosphate biofilms) at different the sides of the biofilm, and the activated carbon marked below the
locations to account for the heterogeneity of the biofilm. Using previ­ biofilm. The blackout image set was used when constructing a three-
ously described methods (Huang et al., 2020; Huang et al., 2018), OCT dimensional biofilm model.
images with scanning dimensions 3.13 mm × 4.18 mm × 4 mm were

3
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

2.4. Reconstructing the biofilm 3. Results and discussion

Biofilm reconstruction was performed in Avizo (Thermo Fisher Sci­ 3.1. Phosphate-fed filters released less bacteria than groundwater-only
entific) by adapting methods developed in previous studies (Carrel et al., filters
2018; Huang et al., 2020) to fit our biofilm grown on activated carbon. A
flow diagram of the biofilm reconstruction methods is shown in Fig. S3. The cell counts entering both groundwater and phosphate filters in
First, we uploaded the raw OCT image stack and the blackout image Week 0 were significantly higher than those exiting (p < 0.05), indi­
stack from the Python code to Avizo. Using the thresholding function on cating that new filters trap bacteria and thereby remove bacteria from
the blackout images, we defined the biofilm volume (biomass and the filtered effluent (Fig. 1). However, in subsequent weeks, effluent
pores). We examined each slice in the image stack to correct for any bacteria concentrations were significantly higher than influent bacteria
image artifacts produced by the code, and this corrected image made up concentrations for both the groundwater filters (2.6–6.6 times higher)
the biofilm volume. To reduce noise in the raw OCT image stack, we and the phosphate filters (1.4–4.7 times higher) (p < 0.05).
used the Avizo “despeckle” function. From this despeckled image, we The means, standard deviations, medians, and ranges of the absolute
defined the pore spaces using the Avizo “top-hat” function every ~20 bacteria concentrations released from the groundwater and phosphate
image slices in the 100-image stack. Pixels with an intensity of zero were filters are shown in Table 1. Absolute bacteria refer to the raw measured
considered to be pores. Some of the pixels with an intensity of zero bacteria concentrations; normalized bacteria refer to the effluent bac­
include locations outside of the biofilm such as in the air or the activated teria divided by the influent bacteria and are shown in Table S2. All
carbon. Therefore, we removed these points from the pore space with samples measured with flow cytometry exceeded the 20 counts/mL limit
the Avizo “mask” function using the defined biofilm volume. This yiel­ of quantification. In Weeks 4, 6, 8–11, 13–14, the absolute concentra­
ded the biomass space and the pore space. The defined biomass space tions of total bacteria released from groundwater filters were signifi­
and pore space were then used to construct a three-dimensional model of cantly greater than those released from phosphate filters (p < 0.05) with
the biofilm. The porosity of the biofilm was calculated from the volume phosphate filters releasing 57–87 % of the bacteria released by
of pore spaces divided by the volume of the biofilm. From the pore groundwater filters (Table 1). In Weeks 1–3, 5, 7, 12, and 15, there was
space, we analyzed the pore network further by modeling the spatial not a significant difference between the total bacteria released from the
distribution of pores, represented by spheres, and channels connecting groundwater or the phosphate filters (p > 0.05). When normalized by
the pores, represented by cylinders. the concentration of bacteria entering the filters, the total bacteria
released from groundwater filters was greater than that from phosphate
2.5. Statistical analyses filters in Weeks 2, 3, 7, and 12 as well (p < 0.05) (Table S2). However,
after normalizing by the influent bacteria concentrations, bacteria
The data were analyzed using R (version 4.0.4). Independent sam­ released from groundwater filters was less than that from phosphate
ples were compared with two-sample t-tests if they were normally filters in Week 8 (p < 0.05) and not significantly different in Week 11 (p
distributed or the Wilcoxon rank-sum test if either sample was not > 0.05). The trends of the intact bacteria were similar to those of the
normally distributed. Dependent samples were compared with paired- total bacteria when comparing bacteria released from groundwater and
sample t-tests if they were normally distributed or Wilcoxon signed- phosphate filters using both the absolute and normalized data (Table 1).
rank tests if they were not. Shapiro-Wilk tests were used to determine The trends from the qPCR results generally support the findings
the normality of the samples. A significance level of α = 0.05 was used observed in the flow cytometry data (Table 1) with bacteria released
for all statistical tests. from groundwater filters being greater than that released from

Fig. 1. Time series of average bacteria released (solid lines with square markers) from filters and the bacteria in the feedwater entering the filters (dashed lines with
circle markers). Total cell counts are shown in red and blue for the groundwater and phosphate filters, respectively. Intact cell counts are shown in yellow and cyan
for the groundwater and phosphate filters, respectively. “G” refers to the groundwater filters and “P” refers to the phosphate filters. Error bars show the standard
deviation across 8 filters for each week.

4
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

Table 1
Absolute concentrations of bacteria in filter effluent.
Week Absolute TCC Absolute ICC Absolute 16S
± groundwater phosphate groundwater phosphate groundwater phosphate
med (min- (x105 cell (x105 cell (x105 cell (x105 cell (x102 gene (x102 gene
max) counts/mL) counts/mL) counts/mL) counts/mL) copies/mL) copies/mL)
3.1 ± 0.1 4.1 ± 0.2 2.9 ± 0.3 3.5 ± 0.2 6.3 ± 6.9 2.0 ± 2.3
0
3.1 (2.9-3.3) 4.1 (3.7-4.4) 2.7 (2.6-3.6) 3.5 (3.1-3.8) 3.8 (1.4-22.8) 1.0 (0.1-6.3)
3.8 ± 1.2 4.6 ± 1.2 3.5 ± 1.3 4.0 ± 1.0
1
3.2 (2.8-6.5) 4.8 (2.9-6.8) 2.9 (2.5-6.4) 4.2 (2.4-5.7)
6.2 ± 3.4 4.5 ± 2.4 5.4 ± 3.0 3.8 ± 2.1 88.1 ± 111.3 86.5 ± 192.9
2
6.0 (2.3-11.0) 3.5 (2.0-8.1) 5.2 (2.0-10.6) 3.0 (1.4-7.2) 36.9 (1.7-287.0) 22.4 (<LoQ-562.1)
11.7 ± 2.6 10.3 ± 3.3 11.3 ± 2.6 9.5 ± 2.8
3
11.7 (8.3-15.9) 9.8 (6.3-17.2) 11.3 (7.7-15.2) 9.2 (5.7-15.2)
9.9 ± 1.9 7.1 ± 0.9 9.3 ± 1.8 6.5 ± 0.8 15.7 ± 26.7 11.9 ± 6.3
4
9.8 (7.7-12.2) 7.2 (5.8-8.5) 9.4 (7.1-11.6) 6.4 (5.5-7.9) 7.8 (0.2-80.8) 13.4 (0.8-20.0)
12.6 ± 5.8 10.9 ± 3.6 13.3 ± 3.8 9.9 ± 2.6
5
12.4 (1.0-20.0) 9.0 (7.8-16.8) 12.0 (10.0-19.2) 8.7 (7.4-14.2)
10.3 ± 1.6 6.0 ± 2.0 8.8 ± 2.0 5.5 ± 1.9 16.2 ± 17.5 699.5 ± 813.6
6
9.3 (8.9-12.4) 5.3 (3.9-9.3) 8.7 (4.7-10.7) 4.8 (3.4-8.7) 6.7 (1.5-47.4) 224.3 (4.2-1785.6)
11.1 ± 2.4 10.7 ± 2.3 10.2 ± 2.6 10.1 ± 2.2
7
10.4 (7.4-15.6) 10.7 (8.2-15.0) 9.6 (7.0-15.3) 9.9 (7.7-14.3)
11.4 ± 1.8 8.2 ± 1.4 10.8 ± 1.7 7.8 ± 1.4 39.4 ± 55.6 6.5 ± 7.9
8
10.9 (9.6-15.4) 8.5 (5.9-10.1) 10.4 (8.9-14.7) 8.0 (5.4-9.9) 19.5 (0.3-171.3) 2.8 (0.5-20.7)
9.5 ± 1.3 7.3 ± 0.9 8.8 ± 1.2 6.8 ± 1.0
9
9.7 (7.2-10.8) 7.0 (6.4-9.1) 9.0 (6.9-10.1) 6.5 (5.7-9.0)
11.4 ± 1.8 8.6 ± 0.8 10.5 ± 1.7 8.1 ± 0.7 6.8 ± 4.8 1.7 ± 1.1
10
11.3 (9.2-14.4) 8.3 (7.6-10.2) 10.6 (8.3-13.6) 7.9 (7.2-9.5) 5.3 (1.1-14.7) 1.5 (0.6-3.9)
13.5 ± 3.3 10.0 ± 1.8 12.6 ± 3.4 9.1 ± 1.4
11
12.3 (10.4-20.1) 9.8 (7.5-13.2) 11.5 (9.2-19.2) 8.9 (7.2-10.9)
9.8 ± 2.6 7.7 ± 1.2 8.0 ± 2.5 6.9 ± 1.2 10.8 ± 15.2 3.0 ± 4.1
12
9.0 (6.2-14.8) 7.3 (6.5-10.0) 7.3 (4.6-12.8) 6.6 (5.8-9.4) 3.4 (0.3-39.4) 1.0 (0.3-11.1)
10.2 ± 2.1 7.9 ± 1.0 9.0 ± 1.8 7.1 ± 0.8
13
10.1 (6.4-13.8) 7.9 (6.5-9.4) 8.8 (5.8-12.0) 6.8 (6.0-8.3)
9.7 ± 1.5 8.1 ± 1.1 8.3 ± 1.3 6.9 ± 0.9 3.5 ± 2.9 1.1 ± 1.1
14
9.8 (6.5-11.5) 8.6 (6.1-9.0) 8.4 (5.8-9.9) 6.8 (5.4-8.0) 2.6 (0.5-9.5) 0.5 (<LoQ-2.9)
16.0 ± 2.9 14.2 ± 3.6 14.1 ± 3.0 11.8 ± 3.5
15
15.8 (12.2-21.8) 13.4 (10.0-20.6) 14.0 (8.7-19.0) 10.0 (8.8-18.3)

Rows are highlighted in red on weeks in which the concentration of bacteria was significantly greater in groundwater filter effluent and highlighted in
blue on weeks in which phosphate filter effluent was significantly greater. Rows highlighted orange indicate there was no significant difference
between the absolute groundwater and phosphate filter effluent bacteria concentrations that week. TCC and ICC stand for total cell counts and intact
cell counts, respectively. 16S refers to bacteria measured by qPCR. Values are presented as mean ± standard deviation in the first row of each week
and median (minimum-maximum) in the second row of each week. TCC and ICC are in units of 10,000 cell counts/mL and 16S qPCR data are in units
of 100 cell counts/mL. <LoQ indicates qPCR samples were below the 5 gc/mL limit of quantification.

phosphate filters in about half of the weeks sampled. All but two samples Miettinen et al., 1997); however, the influence of phosphate on micro­
in Weeks 2 and 14 from the phosphate filter effluent exceeded the 5 gc/ bial activity can actually depend on the environment (Appenzeller et al.,
mL limit of quantification. The absolute bacteria released from 2001; Fang, 2010; Noh et al., 2020). In carbon-limited systems, phos­
groundwater filters was higher than that from phosphate filters in Weeks phate addition can have limited impact on bacterial density and biofilm
8, 10, and 14 (p < 0.05); lower in Week 6 (p < 0.05); and not signifi­ growth (Batté et al., 2003; Noh et al., 2020). The concentrations of
cantly different in Weeks 2, 4, and 12 (p > 0.05). After normalizing by carbon and phosphorus in this system were 1–2 mg/L as C and 0.6 mg/L
the concentration of bacteria entering the filters (Table S2), the con­ as P (2 mg/L as PO4), respectively, resulting in a 100:60 to 100:30
centrations of bacteria exiting groundwater filters were higher than carbon to phosphorus ratio. Noh et al., 2020 found a carbon to phos­
those exiting phosphate filters in Weeks 2, 6, and 12 (p < 0.05); lower in phorus ratio of 100:20 to be an excess phosphorus condition. Although
Week 8 (p < 0.05); and not significantly different in Weeks 4 and 10 (p having excess phosphorus could have limited the magnitude by which
> 0.05). The concentrations of bacteria determined by qPCR were phosphate increases bacterial growth and release from the filter, the
100–1000 fold lower than those determined by flow cytometry because lower release of bacteria from the phosphate filters is likely more
the DNA extraction process resulted in consistent losses of DNA to attributable to the biofilm structure, which is discussed in Sections 3.2
remove qPCR inhibitors while the flow cytometry method measured and 3.3. The results of this study show that filters increase bacteria
bacteria cells directly. concentrations after two weeks of usage, but the addition of a phosphate
The rapid elevation of bacterial release with filter age is consistent corrosion inhibitor would neither accelerate nor further increase the
with previous findings from both the field and laboratory (Clark et al., release of bacteria from POU filters.
2022; Reasoner et al., 1987; Snyder et al., 1995; Su et al., 2009).
Numerous studies have reported that phosphate can contribute to in­
creases in microbial growth (Chu et al., 2005; Hozalski et al., 2005;

5
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

3.2. Phosphate biofilms were thicker and rougher than groundwater

biofilms

Within two weeks, biofilm had visibly and quantifiably grown on the
activated carbon filter media (Fig. 2a, Fig. 3, Fig. S2) to a median
thickness of 154 μm in groundwater biofilms and 82 μm in phosphate
biofilms (Table S3). Although biofilms were imaged on the outside of the
filter, it is reasonable to assume that the biofilm can grow on the inside
and outside of the filter because bacteria have been shown to break
through the filter (Wu et al., 2021). The median groundwater biofilm
thickness ranged from 122 to 221 μm over 15 weeks and the median
phosphate biofilm thickness ranged from 82 to 331 μm over the same
period (Fig. 2a). In Week 2, the groundwater biofilms were significantly
thicker than the phosphate biofilms (p < 0.05). Every week thereafter,
the phosphate biofilms were significantly thicker than the groundwater
biofilms (p < 0.05). The periodic rise and fall of the biofilm thickness
and the bacterial release were likely due to a combination of natural
variation and the filters cycling through stages of filters trapping bac­
teria and biofilm growth (periods of low bacterial release and thick
biofilm) and filters releasing bacteria and biofilm detachment (periods
of high bacterial release and thin biofilm).
The absolute roughness of the phosphate biofilms was significantly
greater than that of the groundwater biofilms in Weeks 2, 4, 5, 6, and 7
(p < 0.05), not significantly different in Weeks 8 and 11 (p > 0.05), and
significantly lower in Week 15 (p < 0.05, Fig. 2b). The mean and median
absolute roughness of the phosphate biofilms were 97–142 % and 89 %–
145 % that of the groundwater biofilms, respectively. The median ab­
solute roughness ranged from 65 to 79 μm in phosphate biofilms and
from 54 to 73 μm in groundwater biofilms (Table S4). When normalizing
by the average biofilm thickness to calculate the relative roughness,
phosphate biofilms were still significantly rougher than groundwater
biofilms in Weeks 2, 4, and 7 (p < 0.05, Fig. S4).
Three-dimensional models of the biofilms (Fig. 3a-b) and the pore
network (Fig. 3c-d) were developed from the OCT images to further
characterize the biofilm. These models provided a visual representation
of the biofilm structure and allowed us to quantify the pore structure of
the biofilms. Porosity was calculated from the pore volume divided by
the total biofilm volume. Biofilm porosity rose and fell over time
(Fig. 2c) and ranged from 0.28 to 0.58 in groundwater biofilms (median
0.45) and 0.23–0.58 in phosphate biofilms (median 0.49). There was no
significant difference between the porosities of the groundwater and
phosphate biofilms in any week (p > 0.05). Meanwhile, the total biofilm
volume (made up of biomass and pores) of the phosphate biofilms was
significantly greater than that of the groundwater biofilms in every week
except Week 2 (p < 0.05, Fig. 2d, Fig. 3).
Our finding that biofilm thickness and biofilm volume increased with
the addition of phosphate aligns with previous studies showing phos­
phate can increase bacterial activity, biofilm growth, and biofilm
thickness (Chu et al., 2005; Fang et al., 2009; Hozalski et al., 2005). This
increase in biofilm thickness and volume when phosphate was added
was likely due to an increased abundance of phosphorus as a key
nutrient for bacterial growth. The addition of phosphate has also been
shown to decrease the production of EPS (Douterelo et al., 2020; Fang
et al., 2009). EPS has been found to aid biofilm adhesion and bacterial
aggregation (Flemming and Wingender, 2010). While decreases in EPS
production with the addition of phosphate would indicate that bacteria
would be less likely to attach to the fresh activated carbon or biofilm
surface, the lower release of bacteria from phosphate filters can be
explained by the greater roughness of the phosphate biofilms.
Phosphate biofilms were rougher than groundwater biofilms in most
weeks (Fig. 2b, Fig. S4), and phosphate filters released less bacteria than
groundwater filters (Fig. 1, Table 2, Fig. S4). Increases in phosphate Fig. 2. (a) Biofilm thickness, (b) absolute biofilm roughness, (c) biofilm
porosity, and (d) biofilm volume for groundwater and phosphate biofilms after
have been shown to affect species composition in biofilms, favoring
2, 4, 5, 6, 7, 8, 11, and 15 weeks of biofilm growth in the CDC biofilm reactors.
microorganisms that can metabolize phosphate (Douterelo et al., 2020;
Li et al., 2016). The species composition of a biofilm can be decisive in
determining a biofilm’s structure, including the biofilm thickness and

6
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

Fig. 3. 3D reconstruction of (a) groundwater biofilms and (b) phosphate biofilms, and the corresponding pore networks for the (c) groundwater biofilm and (d)
phosphate biofilm for 15-week old biofilm.

roughness (Murga et al., 1995). The higher roughness of the phosphate results in a “texture” or graininess in the OCT images. This texture in
biofilms observed in our study may have resulted in less bacterial release OCT images includes speckle, which is the constructive/destructive
from the phosphate filters because of greater bacterial adhesion to and interference of coherent light waves scattering off sub-resolution parti­
less bacterial detachment from the biofilm. cles or structures within the sample. While we cannot physically resolve
The relationships observed between higher substrate surface or visualize the sub-resolution structures below 5 μm, these structures
roughness and greater bacterial colonization – including greater adhe­ can still be detected based on the difference in the speckle pattern or
sion and less detachment – are in agreement with previous studies texture, as shown in a previous study (Zaki et al., 2023). Given this
(Ahmad et al., 2017; Korotta-Gamage and Sathasivan, 2017; Oder et al., limitation of the OCT system, pores and channels with equivalent radii
2015; Percival et al., 1999; Shen et al., 2015; Tong and Derek, 2022). below 5 μm were included in the counts of pores and channels but not
The increase in bacterial adhesion to rougher biofilms can be attribut­ compared quantitatively.
able to greater interception of the bacteria by the rough edges of the The number of pores and channels normalized by the biofilm volume
biofilm and an increased surface area onto which the bacteria could for both groundwater and phosphate biofilms were steady over time
attach (Saranya et al., 2011; Shen et al., 2015; Webster et al., 1999). (Fig. S6). Although the total number of pores were not significantly
Rougher biofilms can protect bacteria from detaching by reducing col­ different in groundwater and phosphate biofilms (p > 0.05), ground­
lisions with other particles and reducing the liquid shear (Al-Amshawee water biofilms had significantly more pores normalized by biofilm vol­
et al., 2021; Gjaltema et al., 1997; Shen et al., 2015; Wu et al., 2012). A ume than phosphate biofilms (p < 0.05). A distribution of the
schematic illustrating this hypothesized influence of roughness on bac­ normalized pore and channel numbers of the groundwater and phos­
terial release is shown in Fig. S5. Our findings suggest that the addition phate biofilms is shown in Fig. 4a. The number of pores per 106 μm3
of a phosphate corrosion inhibitor creates a rougher biofilm on the biofilm ranged from 8.1 to 11.5 (median of 10.5) and 0.2 to 18 (median
activated carbon POU filter and may help reduce the release of bacteria of 8.7) in groundwater and phosphate biofilms, respectively. The num­
from the filters. ber of channels per 106 μm3 biofilm ranged from 1.0 to 2.4 (median of
The overall porosity did not have a clear impact on the bacterial 1.7) and 0.6 to 4.2 (median of 2.0) in groundwater and phosphate bio­
release from filters as there were no consistent differences in the overall films, respectively.
porosity between groundwater and phosphate biofilms. After analyzing The distributions of pore equivalent radii, channel equivalent radii,
the pore network of the groundwater and phosphate biofilms, the pore and channel equivalent length are shown in Fig. 4b-d. As shown in the
structure of the biofilm could have been influential. split violin plots with short bodies and long necks, the equivalent radii
and lengths were skewed right. Quantitatively, the median equivalent
radii and lengths each week (pore equivalent radii: 6–8 μm, channel
3.3. Phosphate biofilms had fewer pores and shorter pore-connecting
equivalent radii: 20–26 μm, channel equivalent length: 131–166 μm)
channels than groundwater biofilms
were all lower than the mean equivalent radii and lengths in that same
week (pore equivalent radii: 10–26 μm, channel equivalent radii: 22–30
The axial resolution of the OCT imaging system was 5 μm. OCT can
μm, channel equivalent length: 148–191 μm).
both image structures that are within its imaging resolution (or larger)
In Weeks 2, 4, 5, 7, and 15, the channel equivalent length of
and detect object structures that are below the optical resolution
groundwater biofilms was significantly greater than that in phosphate
because of the scattering that comes from sub-resolution structures and

7
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

structure of biofilms grown on other filter materials may be different.

4. Conclusions

As POU filters become increasingly popular to remove lead from and

improve taste in drinking water, it is essential that we understand the
broader impacts of using these filters and how these filters fit in a system
with other methods for addressing elevated lead levels in drinking
water, such as adding a corrosion inhibitor. Numerous studies have
shown that activated carbon POU filters increase concentrations of
bacteria in filtered drinking water (Chaidez and Gerba, 2004; Clark
et al., 2022; Nriagu et al., 2018; Tobin et al., 1981; Wu et al., 2017),
some of which may be opportunistic pathogens (Geldreich et al., 1985;
Molloy et al., 2007; Wu et al., 2021). Our study is the first to characterize
biofilm grown on activated carbon POU filters and assess the impacts of
phosphate on these biofilms. Phosphate filters released significantly less
bacteria than groundwater filters in most weeks. Phosphate biofilms
were thicker and had a greater surface roughness, which may have
resulted in less bacterial release from the phosphate filters. Phosphate
biofilms also had fewer pores per biofilm volume and shorter channels
connecting the pores, which may have contributed to a less stiff biofilm
Fig. 4. (a) Boxplots of the number of pores and channels normalized by the than the groundwater biofilms, thereby causing less bacterial release
total biofilm volume for groundwater and phosphate biofilms. Split violin plots from the phosphate filters. In summary, the presence of a phosphate
of (b) pore equivalent radius, (c) channel equivalent radius, and (d) channel
corrosion inhibitor in systems with commercially available activated
equivalent length for groundwater and phosphate biofilms. Week is abbrevi­
carbon POU filters could reduce the bacterial release from the filters.
ated “Wk.”
This knowledge is important for limiting exposure to and infection from
bacteria in drinking water when using POU filters. This study also has
biofilms (p < 0.05). In other weeks, there was no significant difference
broader implications for installing new POU filters in systems where
between groundwater and phosphate channel lengths (p > 0.05). There
corrosion inhibitors are already in use: that adding a POU filter will not
were some differences in the distributions of pore equivalent radii and
elevate bacteria in drinking water any more than adding a filter would if
channel equivalent radii calculated to be statistically significant when
no corrosion inhibitors were in use.
comparing groundwater biofilms against phosphate biofilms. However,
the median differences were smaller than the resolution of the OCT
Declaration of Generative AI in scientific writing
imaging system and so the results of those statistical comparisons are not
reported. The median pore equivalent radius ranged from 6.7 to 7.0 μm
The authors declare that no generative AI or AI-assisted technologies
in groundwater biofilms and 6.4–8.0 μm in phosphate biofilms over the
were used in the production of this manuscript, writing or otherwise.
15-week growth period. The median channel equivalent radius ranged
from 21 to 25 μm in groundwater biofilms and 20–26 μm in phosphate
CRediT authorship contribution statement
biofilms. The median channel equivalent length ranged from 149 to 161
μm in groundwater biofilms and 131–166 μm in phosphate biofilms. Gemma G. Clark: Conceptualization, Data curation, Formal anal­
In our study, groundwater biofilms had more pores per biofilm vol­
ysis, Investigation, Methodology, Visualization, Writing – original draft,
ume than phosphate biofilms, and groundwater biofilms had longer
Writing – review & editing. Dietrich Geisler: Software. Evan J. Coey:
channels connecting the pores than phosphate biofilms had. The
Data curation, Investigation. Lance J. Pollitz: Data curation, Investi­
porosity and pore structure of a material, including biofilms, can affect
gation. Farzana R. Zaki: Investigation, Resources. Conghui Huang:
the stiffness of the material or its ability to resist deformation (Huang
Conceptualization, Methodology. Stephen A. Boppart: Resources.
et al., 2020; Huang et al., 2023; Laspidou and Aravas, 2007; Savchenko
Thanh H. Nguyen: Conceptualization, Funding acquisition,
et al., 2014; Schaffler and Burr, 1988). More porous biofilms with longer
Supervision.
and larger pore channels can lead to less stiff biofilms (Huang et al.,
2020; Huang et al., 2023; Laspidou and Aravas, 2007). Decreases in
Declaration of competing interest
stiffness can result in more bacteria detaching from the biofilm (Shen
et al., 2018; Tierra et al., 2015). This is because the EPS structure be­
The authors declare that they have no known competing financial
comes more complex with increasing elasticity or decreasing stiffness
interests or personal relationships that could have appeared to influence
and forms biofilm streamers that are easier to detach (Tierra et al.,
the work reported in this paper.
2015). Our findings suggest that the addition of phosphate led to bio­
films grown on activated carbon with fewer pores and shorter con­
Data availability
necting channels, which may have resulted in a stiffer biofilm less likely
to detach bacteria. The structure that a biofilm grows into is likely
Data will be made available on request.
substrate dependent. Previous studies have shown how bacterial
composition dictates biofilm structure (Murga et al., 1995). Our study
Acknowledgements
found that phosphate led to a thicker biofilm with fewer pores and
shorter connecting channels when grown on activated carbon. However,
This project was supported by NSF Grant 1855609 and by the NIH/
a previous study with the same growth conditions found that phosphate
NIBIB Center for Label-free Imaging and Multiscale Biophotonics
led to a thinner and more porous biofilm when grown on PVC (Huang
(CLIMB) Grant P41EB031772. Additionally, the authors would like to
et al., 2023). Our study presents results from Brita activated carbon POU
thank Amaja Craft, Alexandra Loya, Aarya Patel, and Gang (Hubert)
filters. The general structure of commercially available activated carbon
Zheng for their contributions to the laboratory experiments and data
POU filters will be similar; however, the impacts of phosphate on the
analyses.

8
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

Appendix A. Supplementary data Huang, P.C., Chaney, E.J., Shelton, R.L., Boppart, S.A., 2018. Magnetomotive
displacement of the tympanic membrane using magnetic nanoparticles: toward
enhancement of sound perception. IEEE Trans. Biomed. Eng. 65, 2837–2846.
Supplementary data to this article can be found online at https://doi. Huang, C., Sun, P.P., Won, J., Wang, Y., Boppart, S.A., Nguyen, T.H., 2020. Effect of
org/10.1016/j.scitotenv.2024.169932. nonphosphorus corrosion inhibitors on biofilm pore structure and mechanical
properties. Environ. Sci. Technol. 54, 14716–14724.
Huang, C.H., Clark, G.G., Zaki, F.R., Won, J., Ning, R.S., Boppart, S.A., et al., 2023.
References Effects of phosphate and silicate on stiffness and viscoelasticity of mature biofilms
developed with simulated drinking water. Biofouling 39, 36–46.
Abe, Y., Skali-Lami, S., Block, J.C., Francius, G., 2012. Cohesiveness and hydrodynamic Janjaroen, D., Ling, F.Q., Ling, F., Monroy, G., Derlon, N., Morgenroth, E., et al., 2013.
properties of young drinking water biofilms. Water Res. 46, 1155–1166. Roles of ionic strength and biofilm roughness on adhesion kinetics of Escherichia coli
Ahmad, M., Liu, S.T., Mahmood, N., Mahmood, A., Ali, M., Zheng, M.S., et al., 2017. onto groundwater biofilm grown on PVC surfaces. Water Res. 47, 2531–2542.
Effects of porous carrier size on biofilm development, microbial distribution and Korotta-Gamage, S.M., Sathasivan, A., 2017. A review: potential and challenges of
nitrogen removal in microaerobic bioreactors. Bioresour. Technol. 234, 360–369. biologically activated carbon to remove natural organic matter in drinking water
Al-Amshawee, S., Yunus, M.Y.B., Lynam, J.G., Lee, W.H., Dai, F., Dakhil, I.H., 2021. purification process. Chemosphere 167, 120–138.
Roughness and wettability of biofilm carriers: a systematic review. Environmental Laspidou, C., Aravas, N., 2007. Variation in the mechanical properties of a porous multi-
Technology & Innovation 21. phase biofilm under compression due to void closure. Water Sci. Technol. 55,
Ammar, Y., Swailes, D., Bridgens, B., Chen, J.J., 2015. Influence of surface roughness on 447–453.
the initial formation of biofilm. Surf. Coat. Technol. 284, 410–416. LeChevallier, M.W., Becker, W.C., Schorr, P., Lee, R.G., 1992. Evaluating the
Appenzeller, B.M.R., Batte, M., Mathieu, L., Block, J.C., Lahoussine, V., Cavard, J., et al., performance of biologically-active rapid filters. J. Am. Water Works Assoc. 84,
2001. Effect of adding phosphate to drinking water on bacterial growth in slightly 136–146.
and highly corroded pipes. Water Res. 35, 1100–1105. Li, S.S., Wang, C., Qin, H.J., Li, Y.X., Zheng, J.L., Peng, C.R., et al., 2016. Influence of
Batté, M., Koudjonou, B., Laurent, P., Mathieu, L., Coallier, J., Prévost, M., 2003. Biofilm phosphorus availability on the community structure and physiology of cultured
responses to ageing and to a high phosphate load in a bench-scale drinking water biofilms. J. Environ. Sci. 42, 19–31.
system. Water Res. 37, 1351–1361. McNeill, L.S., Edwards, M., 2002. Phosphate inhibitor use at US utilities. J. Am. Water
Berney, M., Vital, M., Hülshoff, I., Weilenmann, H.U., Egli, T., Hammes, F., 2008. Rapid, Works Assoc. 94, 57–63.
cultivation-independent assessment of microbial viability in drinking water. Water Miettinen, I.T., Vartiainen, T., Martikainen, P.J., 1997. Phosphorus and bacterial growth
Res. 42, 4010–4018. in drinking water. Appl. Environ. Microbiol. 63, 3242–3245.
Booten, C., Robertson, J., Christensen, D., Heaney, M., Brown, D., Norton, P., et al., 2017. Molloy, S.L., Ives, R., Hoyt, A., Taylor, R., Rose, J.B., 2007. The use of copper and silver
Residential Indoor Temperature Study. U.S. Department of Energy Office of Science in carbon point-of-use filters for the suppression of Legionella throughput in
and Technical Information. domestic water systems. J. Appl. Microbiol. 104, 998–1007.
Bors, J., O’Brien, K.R., Kenway, S.J., Lant, P.A., 2017. Regional-scale variability of cold Murga, R., Stewart, P.S., Daly, D., 1995. Quantitative-analysis of biofilm thickness
water temperature: implications for household water-related energy demand. variability. Biotechnol. Bioeng. 45, 503–510.
Resour. Conserv. Recycl. 124, 107–115. Nerenberg, R., 2016. The membrane-biofilm reactor (MBfR) as a counter-diffusional
Cargill, K.L., Pyle, B.H., Sauer, R.L., Mcfeters, G.A., 1992. Effects of culture conditions biofilm process. Curr. Opin. Biotechnol. 38, 131–136.
and biofilm formation on the iodine susceptibility of Legionella-Pneumophila. Can. Nguyen, C.T., Tu, H.H., Chaney, E.J., Stewart, C.N., Boppart, S.A., 2010. Non-invasive
J. Microbiol. 38, 423–429. optical interferometry for the assessment of biofilm growth in the middle ear.
Carollo, R., 2022. Survey Reveals why More Americans than Ever Are Filtering their Biomed. Opt. Express 1, 1104–1116.
Water. Aquasana. Nguyen, C.T., Robinson, S.R., Jung, W., Novak, M.A., Boppart, S.A., Allen, J.B., 2013.
Carrel, M., Morales, V.L., Beltran, M.A., Derlon, N., Kaufmann, R., Morgenroth, E., et al., Investigation of bacterial biofilm in the human middle ear using optical coherence
2018. Biofilms in 3D porous media: delineating the influence of the pore network tomography and acoustic measurements. Hear. Res. 301, 193–200.
geometry, flow and mass transfer on biofilm development. Water Res. 134, 280–291. Noh, J.H., Yoo, S.H., Son, H., Fish, K.E., Douterelo, I., Maeng, S.K., 2020. Effects of
Chaidez, C., Gerba, C., 2004. Comparison of the microbiologic quality of point-of-use phosphate and hydrogen peroxide on the performance of a biological activated
(POU)-treated water and tap water. Int. J. Environ. Health Res. 14, 253–260. carbon filter for enhanced biofiltration. J. Hazard. Mater. 388.
Chu, C.W., Lu, C.Y., Lee, C.M., 2005. Effects of inorganic nutrients on the regrowth of Nriagu, J., Xi, C., Siddique, A., Vincent, A., Shomar, B., 2018. Influence of household
heterotrophic bacteria in drinking water distribution systems. J. Environ. Manag. 74, water filters on Bacteria growth and trace metals in tap water of Doha, Qatar. Sci.
255–263. Rep. 8, 8268.
Clark, G.G., Pan, W., Giammar, D.E., Nguyen, T.H., 2022. Influence of point-of-use filters Oder, M., Kompare, B., Bohinc, K., Torkar, K.G., 2015. The impact of material surface
and stagnation on water quality at a preschool and under laboratory conditions. roughness and temperature on the adhesion of Legionella pneumophila to contact
Water Res. 211, 118034. surfaces. Int. J. Environ. Health Res. 25, 469–479.
Cooper, I.R., Hanlon, G.W., 2010. Resistance of Legionella pneumophila serotype 1 Paul, E., Ochoa, J.C., Pechaud, Y., Liu, Y., Line, A., 2012. Effect of shear stress and
biofilms to chlorine-based disinfection. J. Hosp. Infect. 74, 152–159. growth conditions on detachment and physical properties of biofilms. Water Res. 46,
Derlon, N., Peter-Varbanets, M., Scheidegger, A., Pronk, W., Morgenroth, E., 2012. 5499–5508.
Predation influences the structure of biofilm developed on ultrafiltration Percival, S.L., Knapp, J.S., Wales, D.S., Edyvean, R.G.J., 1999. The effect of turbulent
membranes. Water Res. 46, 3323–3333. flow and surface roughness on biofilm formation in drinking water. J. Ind. Microbiol.
Douterelo, I., Dutilh, B.E., Calero, C., Rosales, E., Martin, K., Husband, S., 2020. Impact Biotechnol. 22, 152–159.
of phosphate dosing on the microbial ecology of drinking water distribution systems: Picioreanu, C., van Loosdrecht, M.C.M., Heijnen, J.J., 1998. Mathematical modeling of
fieldwork studies in chlorinated networks. Water Res. 187. biofilm structure with a hybrid differential-discrete cellular automaton approach.
Falkinham, J.O., 2015. Common features of opportunistic premise plumbing pathogens. Biotechnol. Bioeng. 58, 101–116.
Int. J. Environ. Res. Public Health 12, 4533–4545. Picioreanu, C., van Loosdrecht, M.C., Heijnen, J.J., 2000. A theoretical study on the
Fang, W., 2010. Biofilm formation and control in a model drinking water distribution effect of surface roughness on mass transport and transformation in biofilms.
system with phosphorus addition. Department of Civil Engineering. Degree of Biotechnol. Bioeng. 68, 355–369.
Philosophiae Doctor. National University of Singapore, p. 199. Reasoner, D.J., Blannon, J.C., Geldreich, Edwin E., 1987. Microbiological characteristics
Fang, W., Hu, J.Y., Ong, S.L., 2009. Influence of phosphorus on biofilm formation in of third-faucet point-of-use devices. J. Am. Water Works Assoc. 79, 6.
model drinking water distribution systems. J. Appl. Microbiol. 106, 1328–1335. Saranya, N., Saravanan, S., Moorthi, A., Ramyakrishna, B., Selvamurugan, N., 2011.
Flemming, H.C., 2011. Microbial biofouling: unsolved problems, insufficient approaches, Enhanced osteoblast adhesion on polymeric Nano-scaffolds for bone tissue
and possible solutions. Biofilm Highlights 5, 81–109. engineering. J. Biomed. Nanotechnol. 7, 238–244.
Flemming, H.C., Wingender, J., 2010. The biofilm matrix. Nat. Rev. Microbiol. 8, Savchenko, N., Sevostyanova, I., Sablina, T., Gomze, L., Kulkov, S., 2014. The influence
623–633. of porosity on the elasticity and strength of alumina and zirconia ceramics.
Folmli, I., 2011. Quantification of Biofilm Morphology Using Optical Coherence International Conference on Physical Mesomechanics of Multilevel Systems 2014
Tomography and Quantitative Image Analysis. Master of Science. Swiss Federal (1623), 547–550.
Institute of Technology, Zurich. Schaffler, M.B., Burr, D.B., 1988. Stiffness of compact-bone-effects of porosity and
Geldreich, E.E., Taylor, R.H., Blannon, J.C., Reasoner, D.J., 1985. Bacterial colonization density. J. Biomech. 21, 13–16.
of point-of-use water treatment devices. J. Am. Water Works Assoc. 77, 72–80. Servais, P., Billen, G., Bouillot, P., 1994. Biological colonization of granular activated
Gibert, O., Lefevre, B., Fernandez, M., Bernat, X., Paraira, M., Calderer, M., et al., 2013. carbon filters in drinking-water treatment. J. Environ. Eng.-Asce 120, 888–899.
Characterising biofilm development on granular activated carbon used for drinking Shen, Y., Monroy, G.L., Derlon, N., Janjaroen, D., Huang, C., Morgenroth, E., et al., 2015.
water production. Water Res. 47, 1101–1110. Role of biofilm roughness and hydrodynamic conditions in Legionella pneumophila
Gjaltema, A., van der Marel, N., van Loosdrecht, M.C.M., Heijnen, J.J., 1997. Adhesion adhesion to and detachment from simulated drinking water biofilms. Environ. Sci.
and biofilm development on suspended carriers in airlift reactors: hydrodynamic Technol. 49, 4274–4282.
conditions versus surface characteristics. Biotechnol. Bioeng. 55, 880–889. Shen, Y., Huang, C.H., Monroy, G.L., Janjaroen, D., Derlon, N., Lin, J., et al., 2016.
Hammes, F., Berney, M., Wang, Y., Vital, M., Köster, O., Egli, T., 2008. Flow-cytometric Response of simulated drinking water biofilm mechanical and structural properties
total bacterial cell counts as a descriptive microbiological parameter for drinking to long-term disinfectant exposure. Environ. Sci. Technol. 50, 1779–1787.
water treatment processes. Water Res. 42, 269–277. Shen, Y., Huang, C., Lin, J., Wu, W., Ashbolt, N.J., Liu, W.T., et al., 2017. Effect of
Hozalski, R.M., Esbri-Amador, E., Chen, C.F., 2005. Comparison of stannous chloride and disinfectant exposure on Legionella pneumophila associated with simulated drinking
phosphate for lead corrosion control. J. Am. Water Works Assoc. 97, 89–103. water biofilms: release, inactivation, and infectivity. Environ. Sci. Technol. 51,
2087–2095.

9
G.G. Clark et al. Science of the Total Environment 914 (2024) 169932

Shen, Y., Huang, P.C., Huang, C.H., Sun, P., Monroy, G.L., Wu, W.J., et al., 2018. Effect of Weber, W.J., Pirbazari, M., Melson, G.L., 1978. Biological growth on activated carbon-
divalent ions and a polyphosphate on composition, structure, and stiffness of investigation by scanning Electron-microscopy. Environ. Sci. Technol. 12, 817–819.
simulated drinking water biofilms. Npj Biofilms Microb. 4. Webster, T.J., Siegel, R.W., Bizios, R., 1999. Osteoblast adhesion on nanophase ceramics.
Snyder, J.W., Mains, C.N., Anderson, R.E., Bissonnette, G.K., 1995. Effect of point-of-use, Biomaterials 20, 1221–1227.
activated carbon filters on the bacteriological quality of rural groundwater supplies. Wingender, J., Flemming, H.-C., 2011. Biofilms in drinking water and their role as
Appl. Environ. Microbiol. 61, 4291–4295. reservoir for pathogens. Int. J. Hyg. Environ. Health 214, 417–423.
Stewart, P.S., 2003. Diffusion in biofilms. J. Bacteriol. 185, 1485–1491. Wu, M.Y., Sendamangalam, V., Xue, Z., Seo, Y., 2012. The influence of biofilm structure
Stringfellow, W.T., Mallon, K., Digiano, F.A., 1993. Enumerating and disinfecting and total interaction energy on retention by biofilm. Biofouling 28, 1119–1128.
Bacteria associated with particles released from Gac filter-Adsorbers. J. Am. Water Wu, C.-C., Ghosh, S., Martin, K.J., Pinto, A.J., Denef, V.J., Olson, T.M., et al., 2017. The
Works Assoc. 85, 70–80. microbial colonization of activated carbon block point-of-use (PoU) filters with and
Su, F., Luo, M., Zhang, F., Li, P., Lou, K., Xing, X., 2009. Performance of microbiological without chlorinated phenol disinfection by-products. Environ. Sci. Water Res.
control by a point-of-use filter system for drinking water purification. J. Environ. Sci. Technol. 3, 830–843.
(China) 21, 1237–1246. Wu, C.-C., Love, N.G., Olson, T.M., 2021. Bacterial transmission and colonization in
Tierra, G., Pavissich, J.P., Nerenberg, R., Xu, Z.L., Alber, M.S., 2015. Multicomponent activated carbon block (ACB) point-of-use (PoU) filters. Environ. Sci.: Water Res.
model of deformation and detachment of a biofilm under fluid flow. J. R. Soc. Technol. 7, 1114–1124.
Interface 12. Xi, C.W., Marks, D., Schlachter, S., Luo, W., Boppart, S.A., 2006. High-resolution three-
Tobin, R.S., Smith, D.K., Lindsay, J.A., 1981. Effects of activated carbon and dimensional imaging of biofilm development using optical coherence tomography.
bacteriostatic filters on microbiological quality of drinking water. Appl. Environ. J. Biomed. Opt. 11.
Microbiol. 41, 646–651. Zaki, F.R., Monroy, G.L., Shi, J., Sudhir, K., Boppart, S.A., 2023. Texture-based speciation
Tong, C.Y., Derek, C.J.C., 2022. Membrane surface roughness promotes rapid initial cell of otitis media-related bacterial biofilms from optical coherence tomography images
adhesion and long term microalgal biofilm stability. Environ. Res. 206. using supervised classification. Res. Square (under review).
Urfer, D., Huck, P.M., Booth, S.D.J., Coffey, B.M., 1997. Biological filtration for BOM and Zhang, Y.J., Li, P.F., Zhou, L.L., 2015. Study on the release of HPC and particles in
particle removal: a critical review. J. Am. Water Works Assoc. 89, 83–98. ozonation and biological activated carbon processes. Chem. Eng. J. 276, 37–43.
Van Nevel, S., Koetzsch, S., Weilenmann, H.-U., Boon, N., Hammes, F., 2013. Routine
bacterial analysis with automated flow cytometry. J. Microbiol. Methods 94, 73–76.

10