D. acidovorans (ATCC 15668; American Type Culture Collection, Manassas, VA), S. paucimobilis (CDC SP-01), M. mucogenicum (CDC MM-01), and Methylobacterium sp. (CDC ME-01) were grown from frozen stocks on R2A agar (Becton Dickinson Diagnostic Systems, Sparks, MD), incubated at ambient room temperature for 7 days, and passaged no more than two times before inoculation into biofilm. The CDC isolates were identified by multiple methods, including morphological characteristics on R2A agar. M. mucogenicum CDC MM-01, an environmental isolate from hospital water associated with an outbreak of central line associated blood stream infections, was identified by fluorescence high-performance liquid chromatography analysis of mycolic acids and hsp65, 16S rRNA, and rpoB gene sequencing (Cooksey et al. 2008). S. paucimobilis CDC SP-01 is an environmental isolate from a 10 μg ml⁻¹ Fentanyl Citrate in 0.9% Sodium Chloride IV Bag involved in a separate hospital outbreak. Methylobacterium sp. CDC ME-01 is an environmental isolate from water in an outpatient dialysis facility. Identification of both S. paucimobilis CDC SP-01 and Methylobacterium sp. CDC ME-01 was confirmed by 16S rRNA gene sequencing.

Autoclaved DeKalb County (Georgia) municipal drinking water (treated surface water containing free chlorine as the secondary disinfectant; DeKalb County Public Works Department; EPA ID: GA0890001), collected over the 7 month period of the study, was used for all experiments. The mean water quality parameters measured in water leaving the treatment plant include temperature, 188C (standard deviation, SD = 2.5); total organic carbon (TOC), 0.79 mg l⁻¹ (SD = 0.11); dissolved O₂, 14 mg l⁻¹ (SD = 0.38); alkalinity, 21 mg l⁻¹ (SD = 1.7); hardness (total), 13 mg l⁻¹; nitrate, 0.49 mg l⁻¹ (SD = 0.09); free chlorine, 1.40 mg l⁻¹ (SD = 0.14); and chloramines, 0.03 mg l⁻¹ (SD = 0.01). pH equaled 8.85 (SD = 0.49), as measured in autoclaved water in the laboratory. Free chlorine in the autoclaved drinking water was measured using the FAS-DPD method (APHA 2005) to confirm that no residual levels were present prior to experimentation.

For all assays and model experiments, single species inocula were prepared in either autoclaved drinking water (for the paired growth assays) or Butterfield Buffer (Becton, Dickinson and Company, Sparks, MD; Donlan et al. 2005) to a concentration equivalent to 0.10 OD₆₀₀ (Hach DR/4000U spectrophotometer, Hach Company, Loveland, CO). For the paired growth assays, the single species suspensions were stored overnight in an environmental chamber (Caron 6030, CARON Products & Services, Inc, Marietta, OH) at 25°C, to allow the bacteria to acclimate to the water. The absorbance at 600nm was measured and adjusted to 0.10 OD₆₀₀ if necessary the next day, before use in the assay. An aliquot of each inoculum was diluted and plated onto R2A, to determine the concentration of each species. The mean cell density of each species suspension was 8.66 log₁₀ CFU ml⁻¹ (SD = 0.16), 8.81 (SD = 0.16), 8.60 (SD = 0.15) and 8.22 (SD = 0.15) for D. acidovorans, S. paucimobilis, M. mucogenicum, and Methylobacterium sp., respectively (n = 13).

Preliminary paired-growth biofilm assays were performed in 96-well polystyrene plates by adding overnight suspensions of each possible pairing of the four model species at a 1:1 proportion, for a final concentration of 10⁵ CFU ml⁻¹ per species. Controls of single species in autoclaved drinking water were also run. The plates were incubated in an environmental chamber at 25°C and 46% humidity for 72 h. To sample the biofilm, plates were rinsed three times in 250 μl well⁻¹ of phosphate buffered saline (PBS, pH 7.2) before extracting the biofilm by pipetting vigorously for 10 s to resuspend the cells in 250 μl well⁻¹ of PBS with 0.02% Tween^® 80 (PBST). Biofilm organisms were enumerated by 1:10 serially diluting suspensions and spread-plating onto R2A agar.

The CBR (Biosurface Technologies, Bozeman, MT, USA) was run by following published protocols (Donlan et al. 2004; Goeres et al. 2005). The CBR contains eight coupon holders with space for three disk coupons per holder. These were filled with PVC coupons (13 mm diameter, 4 mm thickness; Biosurface Technologies, Bozeman, MT, USA) that had been washed in laboratory detergent (Versaclean, Fisher Scientific, Pittsburgh, PA), rinsed 5x in deionized water, rinsed a final time in 70% ethanol, and air-dried. The assembled CBR was sterilized with ethylene oxide, the outflow line was clamped, and the reactor was filled with 350 ml of autoclaved drinking water. Suspensions of the four model bacteria were prepared as described above and inoculated into the CBR at a final concentration of ~10⁵ CFU ml⁻¹ per species. The CBR was run in batch mode for 24 h, with the inner paddle rotating at 100 rpm, in an environmental chamber at 25°C. After 24 h, the reactor was continuously supplied with autoclaved drinking water contained in a 20-l carboy, at 2.5 ml min⁻¹ (140 min residence time), for at least 13 additional days.

PVC coupons were harvested by aseptically removing a sampling rod from the reactor and transferring them to conical 50 ml tubes containing 10 ml of PBST. CBR water was sampled by inserting a pipet through the empty sample rod slot before closing the hole with a sterile replacement rod. Coupons were subjected to three alternating 30 s cycles of sonication (frequency of 42 kHz [±6%], Branson Ultrasonics Corp., Danbury, CT) and vortexing. Previously determined recovery efficiency of NTM biofilm from plastic and stainless steel coupon surfaces ranged between 90 and 99% (Williams et al. 2009). Biofilm suspensions were serially diluted and spread-plated on R2A agar. Plates were incubated at room temperature for up to 2 weeks. Colonies of each organism were differentiated and enumerated based on colony morphology and time of appearance on R2A (Table 1).

NH₂Cl was generated by adding ammonium chloride (NH₄Cl) and sodium hypochlorite (NaOCl) in a 3:1 molar ratio to 0.0035M phosphate buffer at pH ≥ 8.5. The NH₂Cl concentration was determined with a Monochlor F kit (Hach Company, Loveland, CO), and read in a Hach DR 4000 spectrophotometer at 655 nm. The batch disinfection assays were performed on 14-day old biofilm grown on PVC coupons in the CBR, placed in chlorine demand-free conical glass tubes containing 45 ml of NH₂Cl solution at the target concentration for disinfection (1 or 2 mg l⁻¹). At each sampling time point, coupons were removed from the disinfectant solution and placed in 10 ml of 0.01% sodium thiosulfate to inactivate the NH₂Cl before processing. The NH₂Cl residual concentration and plate count following neutralization with sodium thiosulfate were measured on the remaining 45 ml of disinfectant solution. The log CFU cm⁻² of each species for each time point or NH₂Cl concentration, triplicate coupons from at least two replicate batch disinfection experiments, were used to calculate the log₁₀ reduction for biofilm and water samples. Two direct viability measurements were performed on a subset of biofilm samples exposed to 0 or 2 mg l⁻¹ of NH₂Cl for 5 or 24 h (see Methods subsection ‘Direct viability detection of suspended PWDS biofilm’).

Continuous flow disinfection assays were performed by filling the CBR and supply carboy with autoclaved drinking water containing the target NH₂Cl concentration of 1 mg l⁻¹. To achieve the target concentration of NH₂Cl in the reactor immediately at the beginning of the continuous flow disinfection, the entire top of the reactor (including the sampling rods) was removed aseptically and the volume of NH₂Cl stock needed to achieve the target concentration of NH₂Cl was added directly into the reactor. The solution in the CBR was mixed for 5 s before the top was replaced and flow of treated drinking water was initiated at a rate of 2.5 ml min⁻¹. At each time point, a rod was removed as previously described (Donlan et al. 2004; Goeres et al. 2005), neutralized in 30 ml of 0.01% sodium thiosulfate, and then processed to recover the biofilms, as described above. Water samples were also neutralized by adding 20 μl of 0.1 N sodium thiosulfate into 8 ml of reactor water, before diluting and plating. The concentration of NH₂Cl in the reactor was monitored over time, and kept at 1.0 ± 0.2 mg l⁻¹ by adding a calculated amount of the stock solution directly into the reactor, when needed. The log₁₀ CFU cm⁻² of each species was calculated as an average of between 9 and 12 biofilm samples, for each time point, from at least three replicate continuous flow disinfection experiments. The log₁₀ reduction was calculated for biofilm and water samples using pre-disinfection Day 14 samples as the control.

Biofilm on PVC coupon surfaces was directly observed using epifluorescence microscopy after staining with Sybr Green I (Invitrogen, Carlsbad, CA), as previously described (Williams et al. 2009). Briefly, coupons containing 14-day biofilm were fixed in 3.7% formaldehyde and stored at 4°C until needed. Coupons were rinsed by dipping into PBS (phosphate buffered saline, Fisher Scientific, Pittsburgh, PA) before immersing in 2 ml of a 5x solution of Sybr Green I (Invitrogen) in room temperature TE buffer at pH 8 (10 mM Tris, 1 mM EDTA; Mediatech, Inc., Manassus, VA). Coupons were stained in the dark at room temperature for 10 min, rinsed in sterile deionized water for 1 min, and air-dried in the dark. Coupons were attached to microscope slides and coverslips were mounted with ProLong Gold antifade solution (Invitrogen). Biofilm was viewed through a Zeiss Axioplan epifluorescence microscope (Carl Zeiss, Thornwood, NY) equipped with a fluorescein filter set (480/40 nm excitation, 505 nm long-pass dichroic mirror, 535/50 nm emission) and an ApoTome. Image z-stacks were obtained through a 63x oil immersion objective.

Biofilm viability after batch disinfection with 2 mg l⁻¹ of NH₂Cl was determined in treated and control suspended biofilm samples by two direct methods, viz. detection of esterase activity, measured with ChemChrome V6 and a solid-phase cytometer (ScanRDI; Chemunex, Ivy-Sur-Seine, France), and membrane permeability, measured by BacLight^™ viability staining (Molecular Probes, Life Technologies, Grand Island, NY). Aliquots of suspended biofilm from triplicate coupons were processed for esterase activity as previously described (Baudart et al. 2002; Shams et al. 2007). Second aliquots were processed to detect membrane permeability with BacLight by following the manufacturer’s protocol. Stained suspensions were filtered through black polycarbonate membrane filters, 0.2 μm pore size (Whatman, Inc. Piscataway, NJ), and air-dried on microscope slides in the dark. Coverslips were secured on filters with BacLight^™ mounting fluid. Samples were viewed through a Zeiss Axioplan epifluorescence microscope described above. Green and red bacteria were enumerated in 10 microscope fields per coupon, totaling at least 500 bacteria.

All statistical analyses were performed on log₁₀ transformed colony enumeration data using SPSS PASW Statistics 18, Release Version 18.0.0 (SPSS, Inc., 2009, Chicago, IL). One-way ANOVA was performed on Day 14 base model data from multiple experiments to determine the source of variation among mean total log₁₀ cell densities. The repeatability SD of the base model was calculated as previously described for the CBR (Goeres et al. 2005), comparing the mean log₁₀ total CFU cm⁻² for all four species combined, per coupon harvested on Day 14, between three replicate experiments. Log₁₀ reductions of viable cells from batch and continuous flow disinfection experiments were calculated using different controls. For the batch disinfections, the control coupons were manipulated in the same way as the experimental coupons, however they were not exposed to NH₂Cl. For the continuous flow disinfections, control coupons were harvested on Day 14, immediately prior to the addition of NH₂Cl to the CBR. Significant differences in log₁₀ values between disinfected and untreated samples were determined by t-test (p < 0.05).

All four species of the base model grew in the CBR and remained in the biofilm during the entire 2 week period, over three experiments (Figure 1a). Of the four species, α-Proteobacteria, typically Methylobacterium sp., were the dominant bacteria in the biofilm. Total culturable biofilm (colony enumeration of all four species combined) on Day 14 for replicate experiments ranged from 6.16 to 6.65 log₁₀ CFU cm⁻². The repeatability SD was 0.24, indicating that the model system provides repeatable total viable biofilm quantities on PVC surfaces. The estimated within-experiment variance was 0.0724 (6 df) and the estimated between-experiment variance was 0.0316 (2 df). There was no significant difference in the mean total log₁₀ biofilm cell densities for the three experiments (ANOVA, p = 0.34), with 70% of the variance coming from within the experiments. Direct observation of biofilm collected on Day 14 demonstrated that the four species formed a patchy biofilm on the PVC surface, the majority of which occurred in a scattered monolayer, with some microcolonies up to 11 mm thick (Figure 2).

At the end of the 24 h batch mode the total viable unattached or suspended cell density had increased slightly in the CBR liquid (Figure 1b). The unattached cell density decreased when the reactor was switched from batch mode to continuous flow mode at 24 h, and stabilized at 5.1 log₁₀ CFU ml⁻¹ (SD = 0.3) from Day 4 through the end of the reactor run on Day 14. Unattached M. mucogenicum was not detected in CBR water samples after 24 h (n_exp = 5), but was present in the biofilm (3.9 log₁₀ CFU cm⁻²; SD = 0.4, n = 3). Low log density of unattached M. mucogenicum was detected sporadically in water samples collected over the remaining incubation days, with maximum log density observed on Day 11, 1.3 log₁₀ CFU ml⁻¹, SD = 1.7.

As preliminary work to confirm compatibility of the four species in a multi-species biofilm, each of the model system organisms was tested for its ability to enhance or inhibit biofilm formation of the other species using a paired-growth biofilm assay. M. mucogenicum produced significantly more biofilm when grown in the presence of D. acidovorans (p = 0.043), S. paucimobilis (p = 0.0052), and Methylobacterium sp. (p = 0.0042) than when grown alone. Biofilm growth of D. acidovorans, S. paucimobilis, or Methylobacterium sp. was neither enhanced nor inhibited by the presence of the other model organisms.

Achieving concentrations of 1 or 2 mg l⁻¹ of NH₂Cl initially during batch disinfection resulted in actual concentrations of 0.8 mg l⁻¹ (SD = 0.1, n_exp = 3) or 1.6 mg l⁻¹ (SD = 0.3, n_exp = 5) of NH₂Cl, respectively, over 24 h, calculated as weighted averages. Total culturable biofilm experienced one log reduction (LR) after exposure to 2 mg l⁻¹ of NH₂Cl (Figure 3, p < 0.01). Exposure for 5 or 24 h resulted in essentially the same LR, 1.05 (SD = 0.51, n_exp = 5) and 1.03 (SD 0.6, n_exp = 5), respectively. D. acidovorans and S. paucimobilis were the most susceptible to disinfection. However, the log density of M. mucogenicum remained the same or increased slightly during disinfection while the log density of the three Proteobacteria species decreased, resulting in a population shift to favor M. mucogenicum as the largest component of the biofilm (Figure 3a, b). One mg l⁻¹ of NH₂Cl had little effect on the biofilm in 5 h, with 0.37 LR (SD = 0.32, n_exp = 4, p = 0.06). After 24 h, however, LR was slightly higher: 0.6 (SD = 0.15, n_exp = 2, p < 0.05).

Two direct viability measurements of biofilm bacteria after batch disinfection with 2 mg l⁻¹ of NH₂Cl provided similar results that contrasted with culturable bacterial enumeration. Esterase activity and membrane integrity were not reduced significantly in biofilm exposed to 2 mg l⁻¹ of NH₂Cl in comparison to untreated control biofilm (Table 2, n = 3 coupons). Similar biofilm log densities were measured in untreated control biofilm by culture, esterase activity, and viable bacteria with intact membranes. The log density of total biofilm bacteria detected with the BacLight^™ kit was only slightly higher than the log density of viable bacteria, since the majority of bacteria were viable (green).

During the batch disinfection experiments, cells of each PWDS species detached from biofilm on PVC coupons and remained in suspension in vessels containing NH₂Cl and control vessels with no NH₂Cl. Control vessels contained 5.48 (SD 0.48) log density of detached culturable bacteria (range, 4.45–6.71 log₁₀ CFU ml⁻¹). Counting all four species together, detached bacteria experienced a 2.01 (SD = 0.66, n_exp = 5) or 2.41 (SD = 0.88, n_exp = 5) LR after exposure for 5 or 24 h to 2 mg l⁻¹ NH₂Cl, respectively (Figure 3c, d, p < 0.01), demonstrating greater susceptibility to disinfection than biofilm bacteria. Exposure to 1 mg l⁻¹ of NH₂Cl resulted in 1.38 (SD = 0.8, n_exp = 4) or 1.41 (SD = 0.33, n_exp = 2) LR of detached bacteria after 5 or 24 h, respectively. Within the overall enumeration, however, the quantity of detached M. mucogenicum increased significantly to 3 log₁₀ CFU ml⁻¹ after exposure for 5 h to both a target of 1 and 2 mg l⁻¹ NH₂Cl (Figure 3c). Also, although detached M. mucogenicum was only recovered erratically from water with no NH₂Cl, it was consistently cultured from water at both time points and NH₂Cl levels during batch disinfection.

Studies were performed to assess the repeatability of the response of the PWDS biofilms to disinfection, with 1 mg l⁻¹ of NH₂Cl under continuous flow. Targeting 1 mg l⁻¹ of NH₂Cl in the CBR for the continuous flow disinfection experiments resulted in a mean concentration of 0.9 mg l⁻¹ NH₂Cl over 28 h (SD = 0.2, n_exp = 3). The response of each species to continuous flow and batch disinfection with 1 mg l⁻¹ of NH₂Cl were similar. There was a statistically significant decrease in the total amount of bacteria recovered from the biofilm at all time points sampled during the continuous flow disinfection (range 0.06–1.30 log₁₀ CFU cm⁻², p < 0.05), although the overall decrease was less than that which occurred in the batch disinfection. The numbers of D. acidovorans and S. paucimobilis cultured from the biofilm in the continuous flow disinfection was significantly reduced at all time points (p < 0.01, data not shown). The quantity of M. mucogenicum recovered from the biofilm increased during exposure for 28 h to 1 mg l⁻¹ of NH₂Cl. However, this increase was not statistically significant, probably due to the large variability of M. mucogenicum log density found in 0 h samples (Figure 4a).

The recovery of detached D. acidovorans and S. paucimobilis in the CBR decreased significantly during continuous flow disinfection (Figure 4b). Detached M. mucogenicum was recovered from the majority of water samples containing NH₂Cl from batch and continuous flow disinfection experiments, with higher log density than found in untreated water (Table 3).