Removal of model viruses, E. coli and Cryptosporidium oocysts from surface water by zirconium and chitosan coagulants

All source waters can potentially be contaminated with human pathogens, originating from animal and human excreta. In order to achieve an acceptable drinking water quality, hygienic barriers are implemented as a part of drinking water treatment. A hygienic barrier is a natural or artificial physical or chemical barrier for removal or inactivation of pathogens. Norwegian Drinking Water Regulations require at least two independent hygienic barriers in the water supply system (Norwegian Food Safety Authority 2001). In order to be considered as a hygienic barrier, a water treatment method should be able to reduce bacteria and virus concentrations by a minimum of 99.9% (3 log₁₀) and parasites by 99% (2 log₁₀) (Norwegian Food Safety Authority 2001).

The combination of coagulation, flocculation and sedimentation followed by rapid granular filtration (conventional treatment) is a commonly employed hygienic barrier. The effectiveness of conventional treatment has been evaluated by numerous studies, for which reported removals range from less than 1 log₁₀ (Harrington et al. 2003) up to 7 log₁₀ (Abbaszadegan et al. 2007), depending on the operational conditions and microbial agents to be removed. Some studies have attempted to assess the relationship between coagulant type and microbial log-removals (Rao et al. 1988; Brown & Emelko 2009). The work by Brown & Emelko (2009) indicated that alternative coagulants can provide similar or higher removal effects on Cryptosporidium parvum, as conventional aluminium (Al) and ferric (Fe). Previous studies have reported high coagulation efficiencies for zirconium (Zr) and chitosan alternative coagulants in terms of colour and turbidity reductions (Christensen et al. 2016); however, little is known about their effects on various waterborne microorganisms. Thus, broader characterization of these alternative agents in terms of their hygienic effects could provide information of relevance.

Removal efficiencies of microorganisms under physicochemical treatment processes are potentially influenced by their morphological, physical, and biochemical characteristics. In the present study, bacteriophages MS2 and Salmonella Typhimurium 28B, Escherichia coli and Cryptosporidium parvum oocysts were selected to cover the size range for waterborne pathogens (Table 2). Data on the isoelectric points (pI) of the selected species are also provided. As pI characteristics for the 28B phage were not available, the present study employed electrophoretic mobility measurements for this microorganism.

The overall aim of the present study was to evaluate the impacts of Zr and chitosan coagulants on the removal of model microorganisms and waterborne pathogens from surface water, relative to a well-characterized and effective polyaluminium chloride (PACl) agent. The present work also assesses the ability of all three coagulants to fulfil the hygienic barrier requirements.

All experiments were performed on water collected from the Glomma River, at Nedre Romerike Water Treatment Plant in Strømmen, Norway in May 2015. The water was stored in a stainless steel tank for 7 months, during which period the water temperature ranged between 5 and 15 °C. The tank was equipped with a recirculating pump, working continuously to resuspend settled particles. The water quality was assessed prior to each run (Table 1).

Turbidity was measured using a 2100 N IS turbidimeter (Hach Company, USA). Colour was measured by DR3900 Hach spectrophotometer (Hach, USA), following ISO 7887:2011 (λ = 410 nm) procedure. Prior to colour measurements, the samples were filtered through a 0.45 μm syringe polypropylene membrane (514-0065, VWR, USA), in order to avoid the influence of turbidity. pH was measured with SenTix^® 41 pH-sensor (WTW, Germany). Total organic carbon (TOC), residual Al and Zr were analysed by commercial laboratories (ALS Laboratory Group Norway AS or Noranalyse AS), following ISO 8245:1999, ISO 17294-1:2004 and ISO 17294-2:2003 procedures, respectively. Inlet and outlet water were also monitored on-line for pH (SensoLyt 700 IQ, WTW, Germany) and turbidity (VisoTurb 700 IQ, WTW, Germany).

Microorganisms included in the present study, their main characteristics and the rationale for their inclusion in the study are provided in Table 2.

The bacteriophage MS2 was propagated using Salmonella Typhimurium WG49 (NCTC 12484) as host (ISO 10705-1:1995). Infectious MS2 was enumerated by a PFU (plaque forming unit) assay, using Escherichia coli Famp as bacterial host, as previously described by Debartolomeist & Cabelli (1991), with minor modifications.

Initial testing revealed that MS2 plaque counts were reduced in the presence of coagulant in both influent and effluent samples. In order to decrease this effect, treatment with an alkaline beef extract (BE) solution was included (Shirasaki et al. 2009). Prior to the MS2 phage enumeration, samples (1 mL) were mixed 1:10 with BE and stirred at 1,500 rpm at +4 °C for a minimum of 5 hours. Virus dilutions were also prepared with the BE solution. Thirteen per cent BE (211520, Becton-Dickinson and Company, USA) was prepared in sterile water, followed by adjustment to pH 9.5–10.0 with 5 N NaOH. The solution was stored at +4°C and used within three days.

Propagation and enumeration of the Salmonella Typhimurium 28B phage was conducted according to ISO 10705-1 (1995), with some modifications. Salmonella Typhimurium type 5 was used as a host strain. The sample (0.5 mL) was mixed with 0.5 mL of host culture and 4 mL of a soft agar. Bottom base agar contained 8.0 g of nutrient broth (105443, Merck, Germany), 0.5 g of yeast extract (111926, Merck, Germany), and 15.0 g of agar in 1 L of distilled water. Soft agar was prepared similarly, but contained 6.5 g of agar only. Unlike MS2, enumeration of infectious 28B phage was influenced only by Zr coagulant; however, BE treatment was not used for that virus.

Escherichia coli (CCUG 17620) was cultivated in brain–heart infusion broth (237500, Becton-Dickinson and Company, USA). The overnight culture (16–18 hours at 37°C) was centrifuged at 1,500 g for 20 min and washed twice with peptone saline diluent (CM0733, Oxoid Ltd, UK), and stored at +4 °C for no longer than 5 days. Enumeration was performed using Colilert-18 with Quanti-Tray/2000 (IDEXX Laboratories, USA) according to the manufacturer's instructions.

Cryptosporidium parvum oocysts were purchased from Moredun Scientific Limited (Scotland, UK). For detection of oocysts in the influent, samples were vortexed for 60 s, pipetted directly on a slide (100 μL) and enumerated by IFAT (immunofluorescent antibody test). Oocysts in the effluent samples (10 L) were concentrated by membrane filtration and centrifugation, followed by separation by IMS (immunomagnetic separation) prior to IFAT. Each slide was counted twice. Concentration and enumeration procedures were based on standard methods (ISO 15553 (2006) and US EPA Method 1623 (2005)). Prior to analysis, two quality control samples, prepared with 10 L tap water and oocysts EasySeed™ spike (TCS Biosciences, UK), were analysed to assess recovery efficiency, which was shown to be between 60 and 70%.

Infectious bacteriophages were enumerated in technical duplicates, whereas E. coli and C. parvum oocysts were quantified from a single measurement.

In an additional attempt to decrease the impact of the coagulants, enumeration of MS2 was also performed by quantitative reverse transcription polymerase chain reaction (RT-qPCR). Samples (140 μL) were added to lysis buffer (560 μL) immediately after sampling and stored at −80 °C. RNA was extracted with the QIAamp^® Viral RNA Mini kit and QIAcube automated system (Qiagen, Germany) according to the manufacturer's instructions. Carrier-RNA (3.1 μg per sample) was spiked after thawing and prior to RNA-extraction. RT-qPCR was performed in a Stratagene AriaMx Real-Time PCR System (Agilent Technologies, Inc., USA) using the RNA UltraSense™ One-Step Quantitative RT-PCR System kit (Invitrogen, USA). Three μL RNA was used in a total volume of 20 μL, using primers, probe and RT-qPCR conditions listed in Table 3 (Dreier et al. 2005). ROX was used as passive reference, and positive and negative controls were included in each run. Each sample was run in technical duplicates and the results were analysed using Agilent AriaMx 1.1 Software and Microsoft Excel.

An initial experiment was run in order to test whether the presence of coagulant would affect the RT-qPCR-assay. Distilled water was spiked with MS2, followed by coagulants (PACl, Zr or chitosan), and compared with a control sample without coagulants. Viral RNA was extracted and enumerated as described earlier. No inhibitory effect was observed for coagulant concentrations up to 10 mg/L.

The main seed suspension consisted of MS2, 28B and E. coli prepared in 15 L of distilled water. Additionally, 1 L of this suspension was spiked with C. parvum. The main suspension was continuously seeded into inlet water during the first 13–15 h of the pilot plant operation. Thereafter, the suspension with C. parvum was seeded for the next 2 h. Two influent and two effluent samples were collected within this period. The first effluent sample was collected after 1 h of C. parvum seeding, corresponding to five pore volumes. Finally, the suspension without C. parvum was seeded again. C. parvum was seeded only during Runs 2 and 3. The seed suspension was cooled on ice and stirred continuously during each run. Influent titres were 6–7 log₁₀ PFU/mL for each of the two phages, 6 log₁₀/100 mL for E. coli and ∼ 3 log₁₀ oocysts/mL for C. parvum.

Microorganism seed suspensions were not purified before use and could therefore elevate TOC concentrations in the raw water, and, consequently, impact the coagulation performances. However, the crude microorganism stocks were diluted at least by 2,500-fold; first, during the preparation of the seed suspensions and then after mixing with raw water. Hence, the uptake of the residual component of the culture medium was considered as minimal. The initial tests also confirmed that the surplus of TOC in raw water did not affect the outlet water quality.

A commercial PACl product, PAX-18, was obtained from Kemira Chemicals (Norway). The product had basicity of 42% and specific gravity of 1.37 g/mL.

Zirconium (IV) oxychloride octahydrate powder was obtained from Teta Vannrensing Ltd (Norway). A working solution of 15% (w/w) was prepared in distilled water.

Commercial chitosan product KitoFlokk™ (low molecular weight [MW, 100 kDa] and deacetylation degree [DD] close to 0.8) was obtained from Teta Vannrensing Ltd, Norway. The concentration of the working solution was 0.5% (w/v) in 0.1 M HCl. All working solutions were stored at room temperature and the chitosan solution was prepared fresh prior to each run.

A schematic and detailed experimental procedure for the filtration system has been described previously (Christensen et al. 2016). Briefly, the pilot-scale system combined coagulation with filtration in a dual-media contact filter column. The column, 10 cm in diameter and 2.5 m high, was packed with support gravel (0.1 m), followed by 0.5 m-layer of Rådasand^® sand (Rådasand AB, Sweden, d₁₀ = 0.4 mm) and 0.8 m of Filtralite^®NC 0.8–1.6 mm material (Weber Leca Raelingen, Norway, d₁₀ = 0.95 mm, 500 kg/m³ dry bulk density). Raw water was pumped at a constant filtration rate of 5.9 m/h. The microorganism suspension, HCl (1 M), and coagulants were fed sequentially to the inlet pipe at constant flow rates. The amount of HCl was set depending on the used coagulant and desired process pH.

As in water treatment practice, the coagulation conditions were optimized in terms of particulate and organic matter removals. A single combination of coagulant dose and pH was used for each selected coagulant. Optimal coagulation conditions were defined, based on residual particles (turbidity) and natural organic matter (NOM, colour) in the filtrate samples. Each coagulant was screened under stable pH conditions, which corresponded to the lower limit of the optimal pH ranges, reported in the literature: 5.7–5.8 for PACl and 4.0–4.2 for Zr and chitosan (Ødegaard et al. 2010; Jarvis et al. 2012; Christensen et al. 2016). A range of low doses was further tested in pilot scale to detect the minimal effective doses that provided turbidity <0.2 NTU, colour <5 mg Pt/L and residual Al and Zr <0.15 mg/L (European Commission 1998, European Directive 98/83) in the outlet water. The following treatments were selected: Run (1) 1.5 mg Al/L at pH 5.8; Run (2) 2.4 mg Zr/L at pH 4.2; Run (3) 1 mg/L of chitosan at pH 4.2; Run (4) 1 mg/L of chitosan mixed with 0.3 mg Zr/L at pH 6.0. Originally, lower coagulation pH was chosen for the last run; however, a slight change in raw water quality during storage affected the system, and the pH was unintentionally raised to 6.0.

Raw water samples were collected as grab samples from the raw water feed tank. Influent and effluent samples were collected from two ports available for manual sampling. The ports were tapped for a couple of minutes prior to sampling. Samples for virus and E. coli enumeration (50–150 mL) were processed within 6 h after sampling, C. parvum oocysts samples were concentrated and enumerated within 2 weeks after sampling. All samples were stored at 4 °C prior to processing.

The procedures for phage purification and electrophoretic mobility measurements were similar to those described previously by Langlet et al. (2008). Forty mL of phage suspension (propagated according to ISO 10705-1) were centrifuged at 4,000 g for 10 min at 4 °C. The supernatant was filtered through a 0.22 μm membrane (SLGP033RB, Millipore, Germany) and concentrated to 6 mL by ultrafiltration (100 kDa, UFC910008, Merck, Germany) at 3,500–4,000 g for 20 min at 20 °C. Caesium chloride (3.65 g, Carl Roth Gmbh, Germany) was dissolved in 5 mL of the phage suspension and centrifuged at 100,000 g for 15 h at 15 °C. The fraction with purified phage was dialyzed (100 kDa MWCO, 131420, Spectrum Laboratories, Inc., USA), against deionized water for 6 h, and against NaNO₃ (1 mM, pH 7.0) for 15 h. The phage was stored at 4 °C prior to measurement of electrophoretic mobility and size distribution using a Zetasizer Nano ZS (Malvern Instruments Ltd, UK). The measurements were performed over a wide range of ionic strengths (1–100 mM NaNO₃) and pH conditions (3.0–6.6; adjusted with either HCl or NaOH). Prior to measurement, solutions with the selected ionic strength and pH were filtered through a 0.1 μm syringe filter (16553, Stedium Biotech GmbH, Germany) and spiked with the purified phage to a final concentration of approximately 10¹¹ PFU/mL. Each measurement was performed in biological triplicates.

Effluent turbidity changed dynamically throughout filter operation. During ripening, the turbidity was compromised, but as the filter became loaded, the quality of the effluent improved gradually and remained stable for the next 15–22 h. Filter operation was terminated by turbidity breakthrough upon reaching maximum filter loading capacity. The dynamics of microorganism retention reflected that of the effluent turbidity to some extent. While a satisfactory turbidity (<0.2 NTU) in the effluent was achieved after 30 min of ripening, concentrations of virus and bacteria were still high. Reduction of microorganisms gradually increased as the cycle progressed. Zirconium coagulant achieved a 3–4 log₁₀ removal of infectious MS2 and 28B phages and 5–6 log₁₀ removal of E. coli and C. parvum (Figure 1). When PACl was used, infectious phages and bacteria were removed by 2.5–3.0 log₁₀ and 4.5 log₁₀, respectively. Bacteria removal by chitosan was similar to those determined for the metal salts, 4.5–5.0 log₁₀, while its effect on infectious phage was slightly lower (∼2.5 log₁₀). Relatively poor C. parvum removal (∼3 log₁₀) was also observed during the run with chitosan. Addition of Zr to the coagulation mixture did not improve chitosan performance for the infectious phage (2.0–2.5 log₁₀ removal) or bacteria (4.4–5.0 log₁₀ removal).

MS2 analysis was strengthened by simultaneous enumeration of total (PCRU) and infectious (PFU) virus particles. The titres obtained for total virus were 0.1–0.2 log₁₀ higher than for infectious virus; however, the removal patterns were similar (Figure 1). For all coagulants, the reduction of total MS2 was close to 2.2–2.6 log_10.

Turbidity breakthrough had a dramatic impact on the number of microorganisms in the effluent water. For PACl and Zr coagulants, turbidity breakthrough occurred after 15 and 22 h of filter operation, respectively. However, for both coagulants, the microbiological water quality started to decline a few hours ahead of turbidity. The impairment was especially dramatic for E. coli. For chitosan, deterioration of the effluent turbidity started after 16–17 h of filter run; however, neither phages nor bacteria were affected and the number of microorganisms continued to decline. Due to the limited number of analyses, the effect of breakthrough on C. parvum was not evaluated. No signs of turbidity or microbiological breakthrough could be identified for the run with chitosan and Zr after 15 h of operation.

The present study assessed the removal capacity of two non-conventional coagulants – Zr and chitosan – for physicochemical and microbiological parameters, and compared their efficiencies with those of PACl. Log-removals for all coagulants were somewhat higher than previously published, despite the application of minimal effective doses. According to a review by Hijnen & Medema (2010), contact filtration systems with traditional Al and Fe coagulants may provide, on average, 0.9 log₁₀ (0.1–1.5 log₁₀ range) removal for viruses, 1.4 log₁₀ (0.8–2.1 log₁₀ range) for bacteria and 3.0 log₁₀ (0.1–5.4 log₁₀ range) for C. parvum. Although the low reductions could be a result of the weighting system used in this review, there is also a concern that contact filtration treatment might be considered inefficient for removal of pathogenic virus and bacteria (Hijnen & Medema 2010). In the present study, the selected coagulants fulfilled the requirements for a hygienic barrier, apart from chitosan, which gave insufficient virus reduction. Addition of Zr to the coagulation mixture did not improve chitosan performance for the removal of both MS2 and E. coli. Poor performance by the mixture is probably due to the relatively high coagulation pH, as acidic pH would likely be more favourable for both coagulants.

Zirconium gave 1.5 log₁₀ fewer E. coli in the effluent than the other coagulants and retained more than 6 log₁₀ of C. parvum oocysts. For MS2 removal, almost no difference in the efficiency of Zr and Al could be detected.

High microbial removal by Zr can be attributed to several factors. Zirconium effectiveness is usually explained in terms of its valency (Jarvis et al. 2012; Hussain et al. 2014), which presumably provides higher charge neutralization power, which is necessary for destabilization of microbial and other colloids. This also explains the higher Zr affinity to organic matter shown in the present and previous studies (Jarvis et al. 2012).

Chitosan is regarded as a potential substitute for the traditionally used Al and Fe coagulants, and may even be preferable due to its properties of biodegradability and non-toxicity, along with the ability to produce less sludge or water with no metal residue (Renault et al. 2009). Brown & Emelko (2009) assessed chitosan reduction properties for C. parvum oocysts in a pilot filter. Removal was shown to be dose-dependent, and the highest tested dose (3 mg chitosan/L) gave 4.2 log₁₀ removal of the pathogen. Treatment with water-soluble chitosan (5 mg/L) and a small ceramic filter removed MS2 and E. coli by 3 and 6 log₁₀, respectively (Abebe et al. 2016). The results achieved by chitosan for similar microorganisms in the present study were poorer. This could be due to setup differences.

It is likely that production of insoluble hydroxide-species could assist the metal-based coagulants in enhanced coagulant–microorganism interactions. Turbidity measurements revealed high levels of hydroxide species in the influent for Al and Zr (Christensen et al. 2016). For chitosan, which does not produce hydroxides and was dosed at low concentrations, such interactions seem to be limited. This may explain the reduced retention of virus-sized biocolloids using chitosan. Higher chitosan dosing could, potentially, enhance microorganism removal. However, high doses were avoided because of the formation of a cake layer on top of the filter (Christensen et al. 2016), which would affect the filtration process, and therefore the barrier properties (Logsdon et al. 2002). Future studies on chitosan may still be required to verify its hygienic properties. Another interesting aspect to address is inactivation of pathogens in the sludge produced with chitosan. Antimicrobial activity of chitosan towards bacteria, viruses, and fungi is well documented (Rabea et al. 2003; Su et al. 2009). A metal-free sludge with reduced levels of infective pathogens would have fewer disposal problems and may be suitable for use in agriculture.

The treatment conditions defined on the basis of turbidity and colour were generally sufficient to meet the criteria for a hygienic barrier for the selected microorganisms. Moreover, a linear relationship could be established between microbial reductions and residual TOC. However, as the number of runs was limited, additional tests are necessary to confirm whether effluent TOC could be used to estimate coagulant efficiency for removal of microorganisms. It is noteworthy that a similarity between the mechanisms for removal of organic matter and microbes has been postulated previously for viruses (Abbaszadegan et al. 2007) and C. parvum oocysts (Xagoraraki & Harrington 2004).

The turbidity parameter is extensively used to assess the efficiency of the physicochemical treatment step for microbial removal (Xagoraraki & Harrington 2004). In the present study, turbidity was of limited reliability as a surrogate. First of all, evidence of a relationship between hygienic effects and effluent turbidity was lacking. Furthermore, the microbiological water quality was shown to be highly compromised for part of the filter cycle, but this was not reflected by online turbidity. These two observations indicate that destabilization and removal processes for inorganic particles and microorganisms differ. Alternatively, the distinction could be explained by turbidimeter failure to register fine particles, detected by microbial analysis.

The discrepancy between effluent turbidity and microbial counts was especially large at ripening and breakthrough stages, and, therefore, longer filter-to-waste periods and early cycle termination were apparently necessary to avoid water quality impairment. Filter ripening could be potentially shortened using elevated coagulant dosing. Furthermore, the cycle length is normally regulated by a fixed period of operation, which is ideally a few hours shorter than the potential turbidity breakthrough. The results indicate that this practice is advantageous in preventing microorganism leakage to the effluent.

Another interesting observation in the present study was the size-related removal pattern for the microorganisms. According to the classical colloid filtration theory, retention of microorganisms in porous media is related to their size, and appears to be more challenging for bacteria-sized particles (Yao et al. 1971) than for viruses and protozoa. However, introduction of a coagulation step prior to filtration changes the effective size of the microorganisms upon their attachment on flocs (Yao et al. 1971). In a study by Nilsen et al. (manuscript in preparation) microorganism removal rates within the filter column were more consistent with a particle size of 20 μm rather than with the size of a single microorganism particle. Nevertheless, the removal efficiencies, in the present study, correlated roughly with the microorganisms' size; that is, E. coli (0.5–1 μm) and C. parvum oocysts (3–5 μm) were usually retained more efficiently than viruses (29–60 nm). Although the number of replicates was insufficient for statistical analysis, it was apparent that the removal of the faecal indicator E. coli did not always reflect that of C. parvum, indicating that other factors, in addition to size, affected microorganism retention. Furthermore, breakthrough occurred earlier for bacteria than for viruses during the termination stage of filtration.

At a characteristic pH, defined as pI, the ability of microorganism colloids to aggregate or attach to filter media and flocs is enhanced (Gerba 1984). The pI of the microorganisms used in the present study were acidic (Table 2). That would imply an advantage of Zr and chitosan coagulants over Al, as they were applied at lower pH conditions. However, the data reported here support this hypothesis just partially. Furthermore, the pI value of both phages used in the present work appeared to be similar, along with their adsorptive and retention behaviour. It was, therefore, concluded that the impact of pI on microorganism removal by physicochemical processes may be important; however, additional factors are likely to be involved as well.

Three different coagulants were applied to evaluate the hygienic effects on viruses, bacteria and protozoan oocysts by the dual-media pilot contact filter. The coagulants were the conventional PACl, little characterized Zr, and eco-friendly chitosan. In general, all tested coagulants demonstrated similar or higher removal efficiencies than previously published for the traditional Al or Fe and contact filtration systems. Each coagulant fulfilled the hygienic requirement, apart from chitosan. Nevertheless, removal of microbes provided by Zr and chitosan were adequate, and comparable to those of the reference PACl coagulant.

The work was supported by the Research Council of Norway (grant no. 226750/O30) and Norconsult consultancy firm. The authors are grateful to the staff at Nedre Romerike Vannverk for provided water samples. The authors also thank Else Marie Aasen, Rannei Tjåland, Elisabeth Furuseth Hansen, Anne Willumsen, Torbjørn Friborg, Øyvin Østensvik, Marte Monshaugen and Olga Anna Osinska from Norwegian University of Life Sciences for the analysis assistance. The surface property measurements of phage 28B were performed using the environmental molecular biology platform from LCPME (Nancy-France) with assistance from Romain Rivet. Kemira Chemicals AS and Teta Vannrensing Ltd are acknowledged for provision of the coagulant samples.