The AMI005 strain of N. fowleri (EDF internal collection, LNHE, Chatou, France) was isolated from the cooling water of a power station (unpublished data). N. fowleri was grown (for 3–5 days) at 43°C on non-nutrient agar (NNA, Indicia Biotechnology, Oullins, France) previously overlaid with an Escherichia coli suspension and identified by an enzyme-linked immunosorbent assay (Indicia Biotechnology) using monoclonal antibody 5D12 (Pougnard et al., 2002). The maintenance of the strain was carried out by transplanting every 3 months. Subculturing consists of cutting a cube of agar and placing it with the contaminated side on a new medium. Incubation was then carried out between 2 and 5 days at 43°C. Following this maintenance procedure, the strain was kept in the dark at 20–25°C in the culture box (sealed tightly with parafilm). N. fowleri trophozoites were harvested under microscope examination after a 2-day culture on E. coli mats against a 5-day culture for cysts. Suspensions of N. fowleri trophozoites were prepared by gently scraping the amoebic migration front of 10 plates and poured them into 5 ml of phosphate buffer saline solution at a pH value of 7.4 (PBS) for further use.

The biofilm on the coupons was first removed by scraping with a sterile swab into 150 ml of bacteria-free PBS. Consequently, the swab containing the extracted biofilm and the coupons were treated with ultrasounds for 10 min (ultrasonic bath, 140 W, 50/60 Hz; Fisher Scientific). As recently reviewed by Azeredo et al. (2017), the methodologies used for biofilm detachment are of great importance; therefore, we have preliminary verified and optimized the power and duration of the ultrasound treatment with respect to the highest cell detachment and lowest mortality of amoebae (Goudot et al., 2012; Perrin et al., 2018). Bacterial and amoebic analyses were conducted on the same dispersed biofilm from the same coupon.

The number of bacterial cells in the biofilm extracts was determined by direct count using epifluorescence microscopy. The cells were stained with the DNA fluorochrome SybrGreen I (S7567, InVitrogen) and recovered on a black nucleopore filter with a membrane of 0.2 μm pore size (Goudot et al., 2012).

Dry weight and copper analysis have been performed on biofilm extracted from some campaigns. Copper was analyzed by anodic stripping voltammetry and the standard addition method after mineralization. The detection limit was 0.1 μg mg⁻¹ dry weight.

Thermophilic FLA, including N. fowleri, were counted using the most probable number (MPN) approach described by Pougnard et al. (2002). The MPN approach determines the concentration of viable N. fowleri—both trophozoites and cysts—without allowing the two forms to be discriminated. Briefly, five replicates of 1 ml of samples (with an appropriate dilution) were spread onto NNA plates previously inoculated with E. coli. The plates were incubated at 43°C, and the presence of an amoebic migration front was assessed daily for 5 days by microscopic examination. Amoebae fronts were further analyzed to determine the presence of Naegleria species using a flagellation test by incubating vegetative forms in demineralized water at 37°C for 4 h (Behets et al., 2003). Among the flagellated amoebae, N. fowleri were identified using an enzyme-linked immunosorbent assay (Indicia Biotechnology) with monoclonal antibody 5D12 as previously described by Reveiller et al. (2003). All other detected amoebae that do not meet flagellation and ELISA tests were classified as thermophilic FLAs other than N. fowleri.

A flat-plate open-channel reactor previously described by Goudot et al. (2012) was operated in continuous flow mode. The inlet flow and the recycle flow rates were maintained at 1.9 and 810 ml min⁻¹, respectively. The hydraulic retention time was 24 h. The flow presented a laminar velocity profile in the length direction, characterized by a shear rate of 17 s⁻¹. The bacteria present in the river water adhered spontaneously to the different surfaces of the reactor. The biofilm grew at 42°C on slide coupons measuring ~22 cm² (8 × 2.5 × 0.1 cm) at the bottom of the reactor (Goudot et al., 2012) made of different test materials (polyvinyl chloride, stainless steel, brass, and titanium). From 24 h on, it is expected that the biofilm will present ~10⁶ bacteria cm⁻². A schematic view of this experimental setup is presented in Figure 1.

Test campaigns (named C1–C6) were conducted for a period of 45 days each. Each test campaign was conducted with the reactor fed by the same fresh water. For C1–C4, two runs (named R1 and R2) were consecutively performed; for C5 and C6, a single run was performed (R1). The assignments for each independent campaign are shown in Table 1. Between each run, the reactor was acid-cleaned (HCl 1 M, 4 h), disinfected (100 mg Cl₂ L⁻¹, 4 h), and finally rinsed with deionized water (five times the working volume).

The glass slide coupons were used as a reference material to evaluate the variability between campaigns and assays. Subsequently, in addition to the tested material (polyvinyl chloride, stainless steel, brass, and titanium), several coupons made of glass were installed in the reactor.

The reactor was fed with Loire River freshwater (Dampierre-en-Burly, France), collected 1–3 weeks before each experiment. The water was stored in darkness in a agitated and refrigerated (4°C) tank for the duration of the experiments. Microbial and physicochemical characteristics of the inlet water are shown in Table 2. Apart from the parameters of dried suspended solids and nitrate, it can be considered that no drastic variation in water quality was observed.

A suspension of N. fowleri in the trophozoite form (10⁸-10⁹ amoebic trophozoites L⁻¹) was prepared extemporaneously and was inoculated into the reactor in a single injection 24 h after its start-up to reach 10⁵ trophozoites L⁻¹ (final concentration). A previous study had established that these conditions allowed viable N. fowleri to colonize biofilms (Goudot et al., 2012).

For each assay, seven to nine samples of three coupons were randomly collected for analysis over the course of the 45 days of the experiments. The coupons were gently washed with 10 ml of bacteria-free PBS to remove cells and deposits not strongly attached to the support material (i.e., not considered part of the biofilm).

The apparent specific growth rate (μ_expo) for amoebae and the bacteria/amoebae ratio (BAR) were calculated as previously described by Goudot et al. (2012) when data permitted. BARs were calculated for each experiment according to the Equation (1):

where “bacteria density” and “amoebae density” are initial cell densities (cells × cm⁻²), i.e., densities measured on the substratum at the time of the injection of N. fowleri. Therefore, BAR is assumed to represent the amount of prey available per amoeba in the biofilm at the time of the amoeba injection into the reactor.

Comparisons between samples (assays) were assessed using ANOVA, where statistical significance was considered (p < 0.05). Statistical analyses were performed using XLSTAT Version 2010.1.01 (Addinsoft, Paris, France).

Error values for the fitting curves were given by Kaleidagraph software v.5.0.

The variation in bacterial cell density in the biofilms developed on the tested materials (polyvinyl chloride, stainless steel, brass, and titanium) was monitored over time (Figure 2). The first two campaigns, C1 and C2, were performed with polyvinyl chloride and stainless steel (Figures 2A, C), and campaigns C3 and C4 were performed with titanium and brass (Figures 2E, G). Two additional campaigns, C5 and C6, were performed on stainless steel and titanium (Figure 2I). In addition, experiments involving the biofilms growing on glass coupons, which shared the same reactor as tested materials serving as reference material, were conducted (Figures 2B, D, F, H, J).

The results show that all substrata, including glass coupons, were colonized by bacteria (Figure 2). This colonization is assumed to be a result of both the attachment and growth of the bacteria. A temporal analysis of the bacterial colonization kinetics reveals two major phases: a first phase from 0 to 3 days characterized by a fast increase in bacterial densities, followed by a second phase from 3 to 45 days characterized by a slowdown or slowup in biofilm development. A more pronounced decrease, than in the other tests, occurred with the glass coupons of the C4 campaign, where brass or titanium was tested (runs R1 and R2) between 2 and 16 days (Figure 2H). However, on a statistical basis, the results did not allow for the establishment of a ranking list of the tested materials according to their ability to promote or not the bacterial colonization, since no significant difference was noted between the assays. This aspect suggests that the biofilm implantation conditions of the autochthonous amoebae (FLA) and N. fowleri we conducted were similar between the materials.

In addition to bacteria, thermophilic FLA and N. fowleri densities were monitored over the course of the experiments (Figure 3). In all cases, no thermophilic FLA was detected on the substrata (<0.3 amoeba cm⁻²) before the spike of N. fowleri and in the fresh water inlet (Table 2).

For campaigns C1, C2, and C5 all done on PVC and SS, a significant increase of N. fowleri within a range of 30–60 N. fowleri cm⁻² was shown on SS and PVC coupons within the first weeks after the injection (Figures 3A, C, I). These results clearly demonstrate that N. fowleri has colonized the biofilm of the coupons. After reaching 100–250 cells × cm⁻², a slight decrease in the amoebae density was observed, followed by a relative stability or a small decrease, indicating that the amoebae population was maintained on these materials during the 40 days of the experiment (Figures 3A, C, I). Thus, whatever the two materials investigated (PVC and SS), the colonization of the coupons by amoebae was not significantly different. Actually, the colonization was mainly due to N. fowleri, which remains the main detectable thermophilic FLA (compare FLA with N. fowleri). Similar trends were obtained with glass sharing the same reactor (Figures 3B, D, J), although a relative stronger decline in amoebae was observed with glass for the two runs of the C1 campaign (Figure 3B). The amoebal surface colonization was also coupled with the persistence of autochthonous amoebae as well as the experimentally injected N. fowleri (Supplementary Figure S1) in water over the 6 weeks of assays.

With the C3, C4, and C6 campaigns all done on brass and titanium, the amoeba colonization profile was different. On brass, the colonization of the biofilm was limited to 15 or 2 FLA cells cm⁻², and N. fowleri was not able to colonize the biofilm after the spike (<0.3 cells cm⁻² after the 7th day, Figures 3E, G). The glass coupons sharing the same reactor with brass coupons (C3-R1) indicated a similar trend, namely, a transient installation of FLA with no true colonization by N. fowleri, which was no longer detectable after 2 days (Figures 3F, H). In contrast, runs with titanium showed a significant increase in FLA during the first days, with a maximum of 40–300 amoebae cm⁻², depending on the campaigns (Figures 3E, G, I). Thereafter, the FLA densities were relatively stabilized ~10–30 amoebae cm⁻². Moreover, on titanium, the FLA presented equal or higher densities than those of N. fowleri (campaigns C4-R2 and C6-R1), suggesting the emergence of autochthonous amoebae at the expense of N. fowleri (Figures 3G, I). Based on morphological characteristics (not shown), such autochthonous amoebae were assumed to belong to the genus Hartmannella. Compared to other assays in the present study, this appears to be the fastest and highest implantation of indigenous amoebae. A decrease to a range of 1–10 amoebae cm⁻² of this FLA population was observed from the third day. Similar colonization of FLA and N. fowleri was noted on the reference material (glass) (Figures 3F, H, J). These differences in the surface colonization of amoebae between brass and titanium were also observed with the same trends in water (Supplementary Figure S1) during the 6 weeks of testing.

Finally, these results suggest that the implantation and growth of amoebae, FLA, and N. fowleri were favored in the freshwater biofilms developed on polyvinyl chloride, stainless steel, or, to a lesser extent, titanium, contrary to brass, which led to the disappearance of amoebae and N. fowleri in particular.

For runs where N. fowleri growth was observed, apparent specific growth rates (μ_expo) were calculated and plotted against the bacteria/amoeba ratio (Equation 1; Figure 4). The runs C3-R1-B, C4-R1-B, and C4-R2-T could not be considered due to the lack of growth of N. fowleri.

The increase of μ_expo with BAR was adjusted by the Monod equation (Equation 2):

In such an equation, we assume that BAR is equivalent to the growth-limiting substrate concentration (Goudot et al., 2012). The maximal specific growth rate (μ_max) derived from the Equation (2) is equal to 0.18 ± 0.07 h⁻¹ for all tested materials (PVC, SS, and T) and 0.14 ± 0.02 h⁻¹ for glass. The half saturation constants (K_S) were 1.2 × 10⁶ ± 0.9 × 10⁶ bacteria × amoeba⁻¹ and 1.1 × 10⁵ ± 0.5 × 10⁵ bacteria × amoeba⁻¹ and the generation time (G) was 5 and 4 h for glass and all other materials, respectively (Figure 4). These data suggest that the variability of N. fowleri observed on glass, PVC, SS, and Ti was explained by the variation in the available substrates (i.e., biofilm cell density per amoeba, or BAR). Since N. fowleri colonization was not observed on any of the brass coupons and the apparent density of biofilm bacteria was not significantly different from that of the other substrata (Figures 2E, G), one can consider that BAR was not the limiting growth factor on brass and on glass sharing the same reactor as brass coupons.

In our experiments, the measurement of the dissolved copper in the water for all the assays showed that in the presence of brass coupons (campaigns C3 and C4), the concentration of dissolved copper (30–110 μg L⁻¹) was 10–30 times greater than the copper concentration in water in contact with other materials (under the detection limit of 3 μg L⁻¹) (not shown). To check the presence of copper within biofilms, analyses were conducted on biofilm extracts from campaigns C3 and C4. All extracts other than those from brass showed very low or values below the detection limit (<0.1 μg mg⁻¹ dry weight, DW). On the contrary, biofilms developed onto brass coupons presented an average copper concentration of 9 ± 2 or 17 ± 5 μg mg⁻¹ DW (Supplementary Table S1).

No significant colonization by the amoebae was observed on brass, suggesting that brass does not promote an effective colonization. It is then logical to consider a direct toxic effect of copper. Indeed, while small amounts of copper are an essential trace element in most pro- and eukaryotic organisms, it can become toxic with higher concentrations. This fact has already been reported that electrolytically generated copper and silver concentrations at a ratio up to 80:800 μg L⁻¹, respectively, and a dissolved copper concentration of 500 μg L⁻¹ caused no significant inactivation of N. fowleri after 72 h exposure (Cassells et al., 1995). More recently, Grechnikova et al. (2020) found a maximum semi-inhibitory concentration (IC₅₀) for N. fowleri equal to 1.6 mM (100 mg L⁻¹). This fact suggests that this amoeba a relatively copper-resistant eukaryotic microorganism and copper concentrations in the water of our reactor (1,000 × lower than the IC₅₀ mentioned above) are unlikely to directly affect N. fowleri. However, the copper level extracted from the biofilm on brass coupons (9–17 μg mg⁻¹ DW) suggests that the copper concentration present in the biofilm could locally reach toxic values for amoebae: 9–17 μg mg⁻¹ DW corresponds to ~8 μg cm⁻² (Supplementary Table S1, taking into account the value of DW for three coupons of 22 cm² each), and by considering a biofilm thick of 20 μm (i.e., a final volume of biofilm of ~2 mm³ cm⁻²⁾, it would mean amoebae eat biofilms having ~4 mg Cu cm⁻³ (i.e., an immense 4 g L⁻¹). Therefore, we cannot exclude a direct toxic effect of copper on amoebae. However, the concentrations for which such an effect is expected should also influence the bacteria present in the biofilms. However, our results do not show anything of the sort. To ensure this aspect, further research into the levels of copper actually absorbed by amoebae living on copper substratum should be conducted.

Competition between amoebae is another parameter that can control biofilm colonization by N. fowleri, as has already been discussed in other studies (Stahl and Olson, 2021). In the present study, the runs showing a higher density of FLA than that of N. fowleri suggest that native amoebae took over the established N. fowleri. However, tests with a similar or slightly lower N. fowleri density than FLA indicated that N. fowleri had successfully implanted itself. In the first cases, this could result from competition favorable to the autochthonous; in the second case, the competition would have been favorable to N. fowleri. We can therefore consider that small heterogeneities in the densities of native amoebae present in the water of the rivers supplying the reactors could have contributed to the differences obtained in our study, rather than a different behavior of the amoebae with regard to the nature of the substrate. To draw a conclusive statement on this point, the ideal would have been to carry out all the materials to be tested simultaneously with the same river water.

Finally, after examination of these hypotheses, the direct or indirect inhibition of N. fowleri by the direct toxic effect of copper or by trophic relationship with bacteria (prey-predator) remains realistic. It would then be interesting to observe the variation in diversity of the bacterial communities when they are exposed to brass because this could help us identify, if any, possible bacterial factors governing the abundance of N. fowleri. However, this possibility remains to be confirmed. It is conceivable that in direct contact with brass, the toxicity threshold is reached or even exceeded (i.e., more than a factor of a thousand, as our estimation of the copper level suggests). Therefore, the copper level within the biofilm, would be much greater than that found in the water bulk and would be becomes sufficient to inhibit the growth of N. fowleri.