Instructions for use

Title Assessment of the efficacy of membrane filtration processes to remove human enteric viruses and the suitability ofbacteriophages and a plant virus as surrogates for those viruses

Author(s) Shirasaki, N.; Matsushita, T.; Matsui, Y.; Murai, K.

Citation Water research, 115, 29-39https://doi.org/10.1016/j.watres.2017.02.054

Issue Date 2017-05-15

Doc URL http://hdl.handle.net/2115/74004

Rights © 2017, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 Internationalhttp://creativecommons.org/licenses/by-nc-nd/4.0/

Rights(URL) http://creativecommons.org/licenses/by-nc-nd/4.0/

Type article (author version)

Additional Information There are other files related to this item in HUSCAP. Check the above URL.

File Information shirasaki_clean revised manuscript1 (includes Tables- and Figures).pdf

Hokkaido University Collection of Scholarly and Academic Papers : HUSCAP

1

Assessment of the efficacy of membrane filtration processes to remove human enteric viruses 1

and the suitability of bacteriophages and a plant virus as surrogates for those viruses 2

3

4

N. Shirasaki*, T. Matsushita, Y. Matsui, K. Murai 5

6

Division of Environmental Engineering, Faculty of Engineering, Hokkaido University, N13W8, 7

Sapporo 060-8628 Japan 8

9

*Corresponding author. Tel.: +81-11-706-7282; fax: +81-11-706-7282; e-mail address: 10

nobutaka@eng.hokudai.ac.jp 11

12

13

Abstract 14

Here, we evaluated the efficacy of direct microfiltration (MF) and ultrafiltration (UF) to remove 15

three representative human enteric viruses (i.e., adenovirus [AdV] type 40, coxsackievirus [CV] B5, 16

and hepatitis A virus [HAV] IB), and one surrogate of human caliciviruses (i.e., murine norovirus 17

[MNV] type 1). Eight different MF membranes and three different UF membranes were used. We 18

also examined the ability of coagulation pretreatment with high-basicity polyaluminum chloride 19

2

(PACl) to enhance virus removal by MF. The removal ratios of two bacteriophages (MS2 and 20

φX174) and a plant virus (pepper mild mottle virus; PMMoV) were compared with the removal 21

ratios of the human enteric viruses to assess the suitability of these viruses to be used as surrogates 22

for human enteric viruses. The virus removal ratios obtained with direct MF with membranes with 23

nominal pore sizes of 0.1 to 0.22 µm differed, depending on the membrane used; adsorptive 24

interactions, particularly hydrophobic interactions between virus particles and the membrane 25

surface, were dominant factors for virus removal. In contrast, direct UF with membranes with 26

nominal molecular weight cutoffs of 1 to 100 kDa effectively removed viruses through size 27

exclusion, and >4-log10 removal was achieved when a membrane with a nominal molecular weight 28

cutoff of 1 kDa was used. At pH 7 and 8, in-line coagulation–MF with nonsulfated high-basicity 29

PACls containing Al30 species had generally a better virus removal (i.e., >4-log10 virus removal) 30

than the other aluminum-based coagulants, except for φX174. For all of the filtration processes, the 31

removal ratios of AdV, CV, HAV, and MNV were comparable and strongly correlated with each 32

other. The removal ratios of MS2 and PMMoV were comparable or smaller than those of the three 33

human enteric viruses and MNV, and were strongly correlated with those of the three human enteric 34

viruses and MNV. The removal ratios obtained with coagulation–MF for φX174 were markedly 35

smaller than those obtained for the three human enteric viruses and MNV. However, because MS2 36

was inactivated after contact with PACl during coagulation pretreatment, unlike AdV, CV, MNV, 37

and PMMoV, the removal ratios of infectious MS2 were probably an overestimation of the ability of 38

3

coagulation–MF to remove infectious AdV, CV, and caliciviruses. Thus, PMMoV appears to be a 39

suitable surrogate for human enteric viruses, whereas MS2 and φX174 do not, for the assessment of 40

the efficacy of membrane filtration processes to remove viruses. 41

42

43

Keywords: Electrostatic interaction, Hydrophobic interaction, Microfiltration, Nonsulfated 44

high-basicity PACl, Pepper mild mottle virus, Ultrafiltration 45

46

47

1. Introduction 48

The provision of safe drinking water is essential for ensuring public health. However, increases in 49

the global populations of humans and domestic animals have resulted in increased demand for safe 50

drinking water, which has led to the use of alternative water sources, sometimes of compromised 51

quality, in many parts of the world (Bosch, 2007; Ferrer et al., 2015). These population increases 52

have also led to increased fecal contamination of water sources, and because large numbers of 53

pathogenic microorganisms are shed in the feces of infected people and animals, water sources 54

receiving sewage discharge are often contaminated with those microorganisms (Rose et al., 1991; 55

Albinana-Gimenez et al., 2006; Bosch, 2007). Therefore, water safety plans built on the principles 56

of multiple barriers, hazard analyses, and critical control points are crucial not only for preventing 57

4

the contamination of source water via human and animal waste but also for providing adequately 58

treated and disinfected drinking water (WHO, 2011). 59

Membrane filtration, particularly low-pressure membrane filtration, is commonly used as an 60

absolute barrier to microorganism contamination in the production of drinking water. However, 61

although complete removal of bacteria and protozoa has been reported for the low-pressure 62

membrane filtration processes of direct microfiltration (MF) and ultrafiltration (UF) (Jacangelo et 63

al., 1995; Howe 2006), varying virus removal ratios have been reported for these processes 64

(Jacangelo et al., 1995; Madaeni et al., 1995; Urase et al., 1996; van Voorthuizen et al., 2001; 65

Langlet et al., 2009; Boudaud et al., 2012; ElHadidy et al., 2013; Matsushita et al., 2013; Ferrer et 66

al., 2015). The mechanisms underlying the removal of viruses by low-pressure membrane filtration 67

include size exclusion and adsorptive interactions (i.e., hydrophobic and electrostatic interactions) 68

between the virus particles and the membrane surface. The relative contribution of these 69

mechanisms to the removal of viruses depends on the characteristics of the virus particles and 70

membrane such as the relative size of the virus particles to the size of the membrane pores and the 71

hydrophobicity and surface charge of the virus particles and membrane surface (Urase et al., 1996; 72

van Voorthuizen et al., 2001; ElHadidy et al., 2013). Indeed, since the pore sizes of membrane 73

filters are generally larger than the size of virus particles, there is almost no removal of viruses via 74

direct MF when the relative contribution of absorptive interactions to virus removal is negligible 75

(van Voorthuizen et al., 2001; Matsushita et al., 2013). 76

5

Coagulation pretreatment prior to membrane filtration, particularly in MF processes, is widely 77

used in the water industry to improve the quality of drinking water, and the combination of 78

coagulation pretreatment and MF processes (coagulation‒MF) also shows promise as an effective 79

means of removing virus particles. Indeed, under appropriate coagulation conditions, the virus 80

removal performances of coagulation‒MF with membrane with a nominal pore size larger than the 81

size of the virus particles are similar or greater than the performance of direct UF with membrane 82

with a nominal molecular weight cutoff (MWCO) smaller than the size of the virus particles 83

(Matsui et al., 2003; Matsushita et al., 2013). Aluminum-based coagulants such as polyaluminum 84

chloride (PACl) and alum are common coagulants in coagulation‒MF processes, and coagulation 85

conditions such as coagulation pH, coagulant dosage, and coagulation time have been shown to 86

affect virus removal performance (Matsushita et al., 2005; Shirasaki et al., 2009; Tanneru et al., 87

2013). In addition, coagulant properties, such as the distribution of the aluminum hydrolyte species 88

in the coagulant, which might be partly controlled by basicity ([OH−]/[Al3+]), have also been shown 89

to affect membrane permeability and the quality of water processed by means of coagulation‒MF 90

(Zhao et al., 2010; Kimura et al., 2015). Kimura et al. (2015) reported that increasing the basicity of 91

PACl coagulants from the typical basicity of 1.5 to a high basicity of 2.1 not only increased the 92

removal of dissolved organic carbon and reduced the concentration of residual aluminum in the 93

filtrate but also mitigated irreversible membrane fouling, which is a major cause of reduced 94

membrane permeability that increases operating costs, in coagulation‒MF processes. Thus, the 95

6

integration of pretreatment with high-basicity PACls into MF processes may increase the removal of 96

virus particles. 97

The virus removal performances of direct MF, direct UF, and coagulation‒MF processes are often 98

evaluated by using bacteriophages such as MS2, Qβ, and φX174 as surrogates of human enteric 99

viruses because these viruses do not infect humans and are easier to cultivate than human enteric 100

viruses (Jacangelo et al., 1995; Urase et al., 1996; van Voorthuizen et al., 2001; Matsui et al., 2003; 101

Matsushita et al., 2005; Langlet et al., 2009; Shirasaki et al., 2009; Boudaud et al., 2012; Tanneru 102

and Chellam 2012; ElHadidy et al., 2013; Matsushita et al., 2013; Tanneru et al., 2013; Ferrer et al., 103

2015). However, data on the removal of human enteric viruses via membrane filtration is limited, 104

particularly regarding fundamental principles such as the contributions of the size exclusion and 105

adsorptive interaction mechanisms to virus removal via direct MF or UF and the effects of 106

coagulation conditions and coagulant properties on virus removal via coagulation‒MF, partly due to 107

the need for bio-containment facilities for the safe handling of viruses and the fact that virus 108

cultivation is labor intensive and time consuming (Ryu et al., 2010). Thus, whether these 109

bacteriophages are adequate surrogates for human enteric viruses in membrane filtration processes 110

remains unknown. 111

Recently, a metagenomic analysis revealed that pepper mild mottle virus (PMMoV; genus 112

Tobamovirus, family Virgaviridae), which infects pepper species, was the most abundant viral RNA 113

in human feces (concentrations up to 109 virus particles/g of feces; Zhang et al., 2006). In addition, 114

7

because PMMoV is more frequently detected at higher concentrations in environmental waters, 115

including drinking water sources, than are human enteric viruses (Hamza et al., 2011; Haramoto et 116

al., 2013), PMMoV may be a potential indicator of fecal contamination of surface water. Thus, if 117

the removal efficiencies of PMMoV and human enteric viruses are comparable, PMMoV could be a 118

useful surrogate for evaluating the efficacy of membrane filtration processes to remove human 119

enteric viruses. 120

In the present study, we used eight types of MF membranes and three types of UF membranes to 121

investigate the efficacy of direct membrane filtration processes to remove the representative human 122

enteric viruses (i.e., adenovirus [AdV], coxsackievirus [CV], and hepatitis A virus [HAV]) included 123

in the Fourth Drinking Water Contaminant Candidate List (CCL4) published by the US 124

Environmental Protection Agency (2016), and murine norovirus (MNV) as a surrogate of human 125

caliciviruses. In addition, we examined the effects of membrane pore size and membrane material, 126

which are related to the hydrophobic and electrostatic interactions between virus particles and the 127

membrane surface, on virus removal. Next, we examined the potential of pretreatment with a 128

high-basicity PACl to enhance virus removal via membrane filtration by comparing virus removal 129

efficiencies obtained with several PACls and alum under various coagulation conditions. Finally, we 130

examined removal efficiencies of the bacteriophages MS2 and φX174, and of the plant virus 131

PMMoV, to assess the suitability of these viruses as surrogates for human enteric viruses. 132

133

8

134

2. Materials and methods 135

2.1. Source water, coagulants, and membranes 136

On 30 September 2015, river water was sampled from the Edo River (Tokyo, Japan; see Table 1 for 137

water quality), which is the source water for the Kanamachi Water Purification Plant (Tokyo, Japan). 138

The source water samples were stored at 4 °C until use and brought to 20 °C immediately prior to 139

use. 140

To investigate the effect of coagulant type (i.e., effects of coagulant basicity and sulfate content) 141

on virus removal via in-line coagulation pretreatment and MF, we used five aluminum-based 142

coagulants (provided by Taki Chemical Co., Kakogawa, Japan). Specifications of the coagulants are 143

shown in Table 2 and are described in Supplementary Information. 144

To investigate the effects of membrane pore size and membrane material on virus removal via 145

membrane filtration, we used eight commercially available MF membranes and three commercially 146

available UF membranes (provided by Millipore Corp., Billerica, MA, USA, and 3M Corp., 147

Maplewood, MN, USA). Specifications of the membranes are shown in Table 3. 148

149

2.2. Characterization of coagulants and membranes 150

The aluminum hydrolyte species in the coagulants were analyzed by using liquid 27Al nuclear 151

magnetic resonance (NMR) spectroscopy after the coagulants were diluted with Milli-Q water to a 152

9

concentration of 54 g-Al/L (i.e., 2.0 M-Al). The details of the liquid 27Al NMR analysis are 153

described in our previous report (Shirasaki et al., 2014). 154

The hydrophobicities of the membranes were determined by means of water contact angle 155

measurement. The surface charges (i.e., zeta potential) of the membranes were determined with a 156

Zetasizer Nano ZS (50-mW, 532-nm green laser; Malvern Instruments, Malvern, Worcestershire, 157

UK) equipped with a surface zeta potential cell kit (ZEN1020; Malvern Instruments). Details of the 158

hydrophobicity and zeta potential measurements are provided in Supplementary Information. 159

160

2.3. Human enteric viruses, MNV, bacteriophages, and PMMoV 161

AdV type 40 Dugan strain (ATCC VR-931), CV B5 Faulkner strain (ATCC VR-185), HAV IB 162

HM175/18f strain (ATCC VR-1402), and MNV type 1 CW1 strain (ATCC PTA-5935) were 163

obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and propagated 164

in human lung carcinoma epithelial cells (A549 cells; ATCC CCL-185, obtained from ATCC), 165

buffalo green monkey kidney epithelial cells (BGM cells; kindly supplied by Dr. Daisuke Sano, 166

Hokkaido University, Sapporo, Japan), fetal rhesus monkey kidney epithelial cells (FRhK-4 cells; 167

ATCC CRL-1688, obtained from ATCC), and murine macrophage cells (RAW264.7 cells; ATCC 168

TIB-71, obtained from ATCC), respectively. The details of the propagation and purification of AdV, 169

CV, HAV, and MNV are described in our previous reports (Shirasaki et al., 2016a; Shirasaki et al., 170

2017). The concentrations of AdV, CV, HAV, and MNV in the purified solutions were 171

10

approximately 105–6, 107, 105–6, and 106 plaque-forming units (PFU)/mL, respectively, as evaluated 172

by means of plaque assays (Shirasaki et al., 2016a; Shirasaki et al., 2017). 173

F-specific RNA bacteriophage MS2 (NBRC 102619) and somatic DNA bacteriophage φX174 174

(NBRC 103405) were obtained from the National Institute of Technology and Evaluation 175

Biological Research Center (Kisarazu, Japan), as were the Escherichia coli bacterial hosts in which 176

the bacteriophages were propagated (NBRC 13965 for MS2, NBRC 13898 for φX174). The details 177

of the propagation and purification of the bacteriophages are described in our previous report 178

(Shirasaki et al., 2016a). The concentrations of MS2 and φX174 in the purified solutions were 179

approximately 1010 and 107–8 PFU/mL, respectively, as evaluated by means of plaque assays 180

(Shirasaki et al., 2016a). 181

PMMoV pepIwateHachiman1 strain (MAFF 104099) was obtained from the National Institute of 182

Agrobiological Sciences Genebank (Tsukuba, Japan) and propagated in Nicotiana benthamiana 183

(seeds kindly supplied by Dr. Kenji Nakahara, Hokkaido University). The details of the propagation 184

of PMMoV are provided in Supplementary Information. The concentration of PMMoV in the stock 185

solution was approximately 107 lesions/mL, as evaluated by using a local lesion count assay with N. 186

tabacum cv. Xanthi-nc (seeds also kindly supplied by Dr. Kenji Nakahara). 187

188

2.4. Direct membrane filtration experiments and coagulation‒MF experiments 189

Direct membrane filtration experiments and coagulation‒MF experiments were conducted with 190

11

virus-spiked river water at 20 °C. The details of the experimental procedures are provided in 191

Supplementary Information. 192

193

2.5. Virus assay 194

The real-time polymerase chain reaction (real-time PCR), which can detect all viruses regardless of 195

their infectivity or the existence of aggregates, was used to quantify viral DNA and RNA. 196

Specifically, viral DNA of AdV or φX174 was quantified by means of real-time PCR, and viral 197

RNA of CV, HAV, MNV, MS2, or PMMoV was quantified by means of real-time 198

reverse-transcription PCR (real-time RT-PCR). The details of the real-time PCR and real-time 199

RT-PCR methods are provided in Supplementary Information. 200

201

2.6. Characterization of PMMoV 202

The electrophoretic mobility of PMMoV was measured in prepared Milli-Q water. The 203

hydrophobicity of PMMoV was estimated by using the bacterial adherence to hydrocarbon method 204

(Rosenberg et al., 1980), with some modifications. Stock solution of PMMoV was suspended in 205

specially prepared Milli-Q water (for the measurement of the electrophoretic mobility) or 206

phosphate-buffered saline (for the measurement of the hydrophobicity) at approximately 104–5 207

lesions/mL. The details of the electrophoretic mobility and hydrophobicity measurements are 208

described in our previous reports (Shirasaki et al., 2016a; Shirasaki et al., 2017). 209

12

210

211

3. Results and discussion 212

3.1. Virus removal by means of direct MF or UF 213

Figure 1 shows the removal ratios (log10[C0/Cf]) of the viruses via direct MF with eight different 214

membrane filters. The surface is typically considered hydrophobic (low wettability) if the water 215

contact angle is greater than 90° and hydrophilic (high wettability) if the water contact angle is less 216

than 90° (Wang et al., 2016). Thus, the surfaces of PVDF-0.1-HP and PVDF-0.22-HP were deemed 217

to be hydrophobic, the surface of PTFE-0.1 was deemed to be neutral, and those of the remaining 218

membranes were deemed to be hydrophilic, as assessed by means of water contact angle 219

measurement (Table 3). The surface of ZP-0.1 was positively charged whereas the surfaces of the 220

other membranes were negatively charged in Milli-Q water at pH 7 (Table 3). The virions of AdV, 221

CV, HAV, MNV, MS2, and φX174 are spherical with diameters of 70‒90 nm, 22‒30 nm, 22‒30 nm, 222

27‒40 nm, 26 nm, and 28 nm, respectively, and the virions of PMMoV are elongated, rigid 223

cylinders with a diameter of 18 nm and a length of 300‒310 nm (Fauquet et al., 2005). The virus 224

particles were negatively charged and stably monodispersed by electrical repulsion at pH 7; the 225

isoelectric points of AdV, CV, HAV, MNV, MS2, and φX174 ranged from 1.6 to 3.8 (Shirasaki et al., 226

2016a; Shirasaki et al., 2017) and that of PMMoV was 3.2 (Figure S1). Thus, if the adsorptive 227

interactions between the virus particles and membrane surface were negligible, the viruses were 228

13

expected to pass through the MF membranes because the diameters of the viruses were smaller than 229

the nominal pore size of the MF membranes (0.1‒0.22 µm). Indeed, limited removal ratios 230

(<0.5-log10) were obtained for AdV, CV, HAV, and MNV when the hydrophilic, negatively charged 231

MF membranes (i.e., PVDF-0.1-LP, PVDF-0.22-LP, PTFE-0.1, and PC-0.1) were used at pH 7, 232

although MCE-0.1 was an exception (range, 0.5‒2.1-log10). In contrast, the hydrophobic or 233

positively charged MF membranes effectively removed the viruses from the water; the removal 234

ratios of AdV, CV, HAV, and MNV for PVDF-0.1-HP, PVDF-0.22-HP, and ZP-0.1 were 3.6- to 235

4.6-log10, 2.8- to >4.2-log10, and 1.8- to 2.7-log10, respectively. Similar results were obtained for 236

MS2, φX174, and PMMoV. These results suggest that hydrophobic and electrostatic interactions are 237

the main mechanisms by which viruses are removed during MF and that the contribution of size 238

exclusion was negligible. Madaeni et al. (1995) reported 1- to 4-log10 removal of MS2 was 239

achieved with PVDF-0.22-HP at pH 7, whereas almost no removal was observed with 240

PVDF-0.22-LP, which is in agreement with our results. In addition, because the removal ratios of 241

AdV, CV, HAV, and MNV obtained with the hydrophobic and negatively charged membranes 242

PVDF-0.1-HP and PVDF-0.22-HP were higher than those obtained with the hydrophilic and 243

positively charged membrane ZP-0.1, the contribution of hydrophobic interactions was likely 244

greater than that of electrostatic interactions to the removal of the viruses at pH 7. 245

Figure 2 shows the removal ratios of the viruses via direct UF with three different filters. Because 246

the relative molecular masses of AdV, CV, HAV, MNV, MS2, φX174, and PMMoV (150‒180 × 106, 247

14

8‒9 × 106, 8‒9 × 106, 15 × 106, 4 × 106, 6 × 106, and 40 × 106 Da, respectively; Sinsheimer 1959; 248

Fauquet et al., 2005) were much larger than the nominal MWCOs of the three UF membranes 249

(1‒100 kDa), effective virus removal by all of the filters was predicted for all of the viruses. As 250

expected, 3- to >5-log10 removal of AdV, CV, HAV, and MNV were observed when the UF 251

membrane with a nominal MWCO of 100 kDa (i.e., RC-100k) was used at pH 7. In addition, the 252

virus removal ratio increased as the nominal MWCO of the UF membranes was decreased from 100 253

kDa to 1kDa; >4-log10 removal of all the viruses used in the present study was achieved with the 254

UF membrane with a nominal MWCO of 1 kDa (i.e., RC-1k). These results indicated that direct UF 255

is effective for the removal of different types of virus, including three types of human enteric 256

viruses. Other groups have also reported the usefulness of direct UF for the removal of 257

bacteriophages, and have suggested that the bacteriophage removal ratio is strongly dependent on 258

the physicochemical characteristics of the membrane (Jacangelo et al., 1995; Langlet et al., 2009; 259

Boudaud et al., 2012). Because the hydrophobicities and surface charges of the three UF 260

membranes were comparable (Table 3) and the roles of hydrophobic and electrostatic interactions in 261

virus removal were likely negligible, size exclusion must have been the main mechanism by which 262

the viruses were removed. Thus, the difference in the virus removal ratios of RC-1k, RC-10k, and 263

RC-100k was considered to be due to the differences in the nominal MWCOs of these UF 264

membranes. 265

266

15

3.2. Effect of adsorptive interactions on virus removal 267

As described in section 3.1, adsorptive interactions between the virus particles and membrane 268

surface were likely the dominant factors for virus removal via direct MF at pH 7. To confirm this 269

finding, we conducted additional filtration experiments with PVDF-0.1-LP, PVDF-0.1-HP, 270

MCE-0.1, and ZP-0.1 where the filtration volume of the virus-spiked river water was increased 271

from 50 mL up to 250 mL (Figure 3a–g) or 1000 mL (Figure 3h–n). The virus removal ratios of all 272

of the viruses obtained with PVDF-0.1-LP were very low and remained constant up to the 273

maximum filtration volume of 1000 mL. Jacangelo et al. (1995) reported that the removal ratio of 274

MS2 obtained with a polysulfone MF membrane with a pore size of 0.2 µm increased with 275

increasing mass of kaolinite deposited on the membrane surface at pH 7. Although the source water 276

used in the present study contained suspended solids (turbidity = 2.8 NTU; Table 1), the removal 277

ratios obtained did not increase with filtration volume, possibly due to differences in the mass of 278

suspended solids deposited on the membrane surface between our study and that of Jacangelo et al. 279

(1995). 280

The virus removal ratios obtained with PVDF-0.1-HP after 50 mL of filtration were similar to 281

those obtained in the direct MF experiments (Figure 1); however, the removal ratios gradually 282

decreased as the filtration volume was increased from 50 to 250 mL and from 200 to 600 mL 283

(Figure 3). These results indicate that the most accessible hydrophobic adsorption sites for virus 284

adsorption on the PVDF-0.1-HP membrane were gradually saturated and were almost exhausted at 285

16

600 mL of filtration. Thus, hydrophobic adsorptive interactions between the virus particles and 286

hydrophobic membrane surface were the dominant factors for virus removal via direct MF until the 287

adsorption sites were saturated with hydrophobic substances present in the source water. In addition, 288

because the virus removal ratios obtained with ZP-0.1 also gradually decreased with increasing 289

filtration volume, except in the case of HAV, electrostatic adsorptive interactions between the virus 290

particles and the positively charged membrane surface also likely contributed to the virus removal 291

ratios obtained. In contrast, the change in the virus removal ratios obtained with MCE-0.1 and 292

increasing filtration volume differed, depending on the type of virus in the spiked water; the virus 293

removal ratios of AdV, HAV, MNV, and PMMoV decreased with increasing filtration volume, 294

whereas those of CV, MS2, and φX174 remained constant during 1000 mL of filtration. Because the 295

contact angle and surface charge of MCE-0.1 were similar to those of PTFE-0.1 and PC-0.1, 296

respectively, and the virus removal ratios of PTFE-0.1 and PC-0.1 were <0.5-log10, as described in 297

section 3.1, the virus removal ratios observed for MCE-0.1 could not be explained simply by the 298

contact angle and surface charge of the membranes. Further investigation, such as elucidation of the 299

pore size distribution of the membranes, is needed to determine why the hydrophilic, negatively 300

charged membrane MCE-0.1 more effectively removed viruses compared with the other hydrophilic, 301

negatively charged MF membranes. 302

Natural organic substances present in surface water are known to adsorb onto the membrane 303

during filtration and cause a decline in flux (i.e., membrane fouling). Indeed, the fouling rate for 304

17

PVDF-0.22-HP is considerably larger than that for PVDF-0.22-LP when filtration is performed with 305

water containing natural organic substances (Fan et al., 2001). We further investigated the 306

competitive adsorption between viruses and natural organic substances in the source water by using 307

PVDF-0.1-HP (Figure S2). The virus removal ratios obtained after prefiltration with 200 or 400 mL 308

of the unspiked river water were smaller than those obtained without prefiltration, indicating that 309

natural organic substances in the source water used in the present study competed with the viruses 310

for the hydrophobic adsorption sites on PVDF-0.1-HP. Therefore, the reduction of virus removal 311

performance by adsorption competition between virus and natural organic substances should be 312

considered when hydrophobic interactions are the dominant removal mechanism during membrane 313

filtration. 314

315

3.3. Virus removal by means of coagulation‒MF 316

Figure 4 shows the removal ratios obtained with in-line coagulation‒MF. As described in section 317

3.1, limited virus removal (<0.2-log10) was obtained with PVDF-0.1-LP without coagulation 318

pretreatment at pH 7 (Figure 1). However, the removal ratios of AdV, CV, HAV, and MNV were 319

improved when the water was subjected to coagulation pretreatment with 0.54 mg-Al/L of 320

coagulant prior to MF (Figure 4a), suggesting that during coagulation pretreatment the virus 321

particles were incorporated into the aluminum floc, which was larger than the nominal pore size of 322

the MF membrane, and the virus-containing floc was then effectively removed by size exclusion. 323

18

The virus removal ratios obtained were dependent on coagulant type; virus removal ratios of 0.5- to 324

2-log10 were obtained with in-line coagulation‒MF with alum or PACl-1.5s as the coagulant, 325

whereas virus removal ratios of 1- to >4-log10 were obtained with the high-basicity PACls. In 326

particular, with the nonsulfated, high-basicity PACls (i.e., PACl-2.1ns and PACl-2.5ns), virus 327

removal ratios of >4.3-log10 for AdV, >3.5-log10 for HAV, and 3.3‒3.6-log10 for MNV were obtained, 328

and CV, MS2, and PMMoV were removed more efficiently than with the other coagulants. However, 329

the virus removal ratios of φX174 were comparable, regardless of the coagulant used. 330

The virus removal ratios increased as coagulant dosage was increased from 0.54 to 1.08 mg-Al/L 331

at pH 7. For AdV, HAV, and MNV, virus removal ratios of >3.9-log10 were obtained with in-line 332

coagulation‒MF, irrespective of the type of coagulant. In addition, PACl-2.1ns and PACl-2.5ns had 333

the highest virus removal ratios, irrespective of the type of virus, except in the case of φX174 (range, 334

>4.0- to >5.7-log10). 335

The virus removal ratios obtained with pretreatment with alum, PACl-1.5s, or PACl-2.1s (dose 336

1.08 mg-Al/L) markedly decreased when the pH of the treated water was increased from 7 to 8 337

(Figure 4b) and roughly corresponded to those obtained with a coagulant dose of 0.54 mg-Al/L at 338

pH 7 (Figure 4a). Even when the coagulant dosage was increased from 1.08 to 2.16 mg-Al/L at pH 339

8, almost no or only a slight improvement in the virus removal ratios was observed for alum, 340

PACl-1.5s, or PACl-2.1s. In contrast, the high virus removal ratios obtained with PACl-2.1ns or 341

PACl-2.5ns were retained (range, >4.1- to >6.0-log10) even at pH 8, although the case of φX174 was 342

19

an exception. These results indicate that the type of coagulant strongly affected the virus removal 343

performance of in-line coagulation‒MF, and that nonsulfated, high-basicity PACls were the most 344

effective for removing viruses not only at neutral pH but also under weakly alkaline pH. In addition, 345

the coagulation conditions (i.e., coagulation pH and coagulant dosage) also strongly affected the 346

virus removal performance of in-line coagulation‒MF, although the magnitude of the effects of the 347

coagulation conditions on virus removal performance was dependent on the type of coagulant; the 348

virus removal performances of alum, PACl-1.5s, and PACl-2.1s were more sensitive to changes in 349

the coagulation pH and coagulant dosage than were those of PACl-2.1ns and PACl-2.5ns. 350

We previously reported that PACl-2.1ns has a higher colloid charge density than do alum, 351

PACl-1.5s, and PACl-2.1s, probably due to its higher colloidal aluminum content and absence of 352

sulfate (Shirasaki et al., 2014). In addition, we previously reported that the Al30 species 353

[Al30O8(OH)56(H2O)24]18+, which is classified as a colloidal aluminum species, as assessed by the 354

ferron method (Chen et al., 2007), probably plays a major role in the removal not only of 355

bacteriophages (Shirasaki et al., 2014) but also of human enteroviruses (Shirasaki et al., 2016b) 356

during coagualtion. To confirm whether Al30 species were present in the nonsulfated, high-basicity 357

PACls used in the present study, we analyzed the coagulants by using 27Al NMR (Figure 5). In the 358

27Al NMR spectra, the signals at 0, 4, 10 to 12 (broad peak), 63, 70 (broad peak), and 80 ppm were 359

attributed to the Al monomer species, the Al dimer and trimer species, the octahedral Al of the 360

external shells in Al13 species [AlO4Al12(OH)24(H2O)12]7+ and Al30 species, the central tetrahedral Al 361

20

in Al13 species, the central tetrahedral Al in Al30 species, and the internal standard, respectively 362

(Allouche et al., 2000; Chen et al., 2007). Whereas no signal for Al30 species was found in the 363

spectra of alum and PACl-1.5s, the spectra for PACl-2.1s contained a clear broad peak for the 364

octahedral Al of external shells in Al13 and Al30 species at 10 to 12 ppm and a very weak signal at 365

70 ppm for Al30 species (Figure 5). These results indicated that PACl-2.1s, but not alum or 366

PACl-1.5s, contained Al30 species. Therefore, the presence of the Al30 species in PACl-2.1s probably 367

led to the higher virus removal performances observed compared with alum and PACl-1.5s. The 368

spectra of PACl-2.1ns and PACl-2.5ns showed clear broad peak at 70 ppm, which indicate the 369

presence of Al30 species, whereas that of PACl-2.1s did not (Figure 5). These results indicate that 370

PACl-2.1ns and PACl-2.5ns a greater amount of Al30 species than did the other coagulants examined. 371

Because Al30 species are the highest-charged polycations ever characterized among PACls, and they 372

confer a stronger floc formation capacity over a broader pH range and wider coagulant dosage 373

compared with other Al species in PACls (Chen et al., 2006; Zhang et al., 2008), the presence of the 374

Al30 species in PACl-2.1ns and PACl-2.5ns likely resulted in the virus removal performances being 375

the highest obtained under the coagulation conditions examined in the present study (Figure 4). The 376

high virus removal performances obtained with PACl-2.1ns and PACl-2.5ns at a dose of 1.08 377

mg-Al/L were comparable with those achieved with the UF membrane with a nominal MWCO of 1 378

kDa (Figure 2). Therefore, in-line coagulation‒MF with a nonsulfated, high-basicity PACl is a 379

potential alternative to direct UF with membranes with a relatively low MWCO for the removal of 380

21

human enteric viruses. In addition, because the cake layer of aluminum floc particles generated by 381

the coagulation of the river water with PACl-2.5ns was more effective for the removal of human 382

enteric viruses than was that formed with PACl-1.5s (Figure S3; see Supplementary Information for 383

details of the effect of the cake layer on virus removal), further increases in virus removal by the 384

formation and growth of the cake layer is predicted when the filtration volume during in-line 385

coagulation‒MF is increased, particularly when a nonsulfated, high-basicity PACl is used. 386

Prevost et al., (2016) reported that most human enteric viruses, including AdV and NV, detected 387

in river water had an intact capsid and were still infectious, as assessed by means of an intercalating 388

dye pretreatment coupled with PCR method. In addition, because some serotypes of viruses are 389

highly resistant to treatments such as UV disinfection (e.g., AdV type 40) and free-chlorine 390

disinfection (CV B5) (Nwachuku et al., 2005; Cromeans et al., 2010), multiple-barrier approaches 391

that both remove and inactivate viruses are important for preventing infection by human enteric 392

viruses in drinking water (Shannon et al., 2008). Thus, in-line coagulation‒MF with a nonsulfated, 393

high-basicity PACl could be an effective physical barrier for human enteric viruses such as AdV 394

type 40 and CV B5. 395

396

3.4. Relationship between enteric virus removal and bacteriophage or PMMoV removal 397

To investigate whether the bacteriophages MS2 and φX174, and the plant virus PMMoV, are 398

suitable surrogates for AdV, CV, HAV, and caliciviruses, Pearson’s correlation coefficients (r) for 399

22

the removal ratios of viruses obtained with direct filtration (MF and UF), or in-line coagulation‒MF 400

were calculated by using the Excel Toukei BellCurve software (Social Survey Research Information 401

Co., Tokyo, Japan; Tables 4 and 5). Virus concentrations that were below the quantification limit of 402

the PCR method were assigned the quantification limit when calculating the correlation coefficients. 403

The removal ratios obtained with AdV, CV, HAV, and MNV were strongly correlated with each 404

other in the direct MF and UF (r = 0.91–0.96) and the in-line coagulation‒MF experiments (r = 405

0.76–0.95). Although these viruses have similar isoelectric points (3.6‒3.8), they have different 406

surface hydrophobicities; HAV and MNV are more hydrophobic than are AdV and CV, as estimated 407

by means of the bacterial adherence to hydrocarbon method (Shirasaki et al., 2016a; Shirasaki et al., 408

2017). Therefore, it was likely that the similarities in the surface charge properties of the viruses 409

resulted in the comparable removal ratios obtained for AdV, CV, HAV, and MNV, and the 410

differences in the surface hydrophobicities of the viruses did not affect the virus removal 411

performances of membrane filtration processes. 412

The removal ratios of MS2 were strongly correlated with those obtained for the three human 413

enteric viruses and MNV in the direct MF and UF (r = 0.85–0.97) and the in-line coagulation‒MF 414

experiments (r = 0.81–0.95); the removal ratios of MS2 and of the three human enteric viruses and 415

MNV were comparable (Figure S4a,d). These results suggest that MS2 is a potential surrogate for 416

AdV, CV, HAV, and caliciviruses in membrane filtration processes when the virus removal ratios are 417

evaluated by means of PCR. However, some groups, including ours, have reported that MS2 is 418

23

inactivated after contact with PACl during coagulation (Matsushita et al., 2011; Kreissel et al., 419

2014), whereas AdV, CV, and MNV were not (Shirasaki et al., 2016a; Shirasaki et al., 2017), as 420

assessed by means of a combination of plaque assay and PCR method. These results imply that 421

MS2 is also inactivated during coagulation–MF with PACl, unlike AdV, CV, and MNV, and 422

therefore that the removal ratios of MS2, as determined by means of plaque assay, which can detect 423

only infectious viruses, probably resulted in overestimation of the ability of coagulation–MF to 424

remove infectious AdV, CV, and caliciviruses. Therefore, if the virus removal performances of 425

membrane filtration processes, including coagulation–MF, are evaluated by using MS2 as a 426

surrogate of human enteric viruses, the PCR method is more appropriate than the plaque assay for 427

assessing virus removal efficacy. 428

The removal ratios obtained for φX174 were also strongly correlated with those of the three 429

human enteric viruses and MNV in the direct MF and UF experiments (r = 0.89–0.97), and these 430

removal ratios had an approximately 1:1 correlation (Figure S4b). However, the removal ratios of 431

φX174 were clearly smaller than those of the three human enteric viruses and MNV in the 432

coagulation–MF experiment and the range was limited (0.7–1.9-log10), unlike that for the other 433

viruses (Figure S4e), although moderate or strong correlations between were found among the 434

removal ratios of φX174 and those of the three human enteric viruses and MNV (r = 0.51–0.83). 435

Therefore, φX174 appears to be an appropriate surrogate for AdV, CV, HAV, and caliciviruses 436

during direct MF or UF, but it does not appear to be an appropriate surrogate for those viruses 437

24

during coagulation–MF. 438

The removal ratios of PMMoV were strongly correlated with those of the three human enteric 439

viruses and MNV in the direct MF and UF (r = 0.92–0.98) and the in-line coagulation‒MF 440

experiments (r = 0.73–0.96), and the removal ratios of PMMoV were comparable with or somewhat 441

smaller than those of the three human enteric viruses and MNV (Figure S4c,f). PMMoV has an 442

isoelectric point of 3.2 (Figure S1), which is similar to that of AdV, CV, HAV, and MNV (i.e., 443

3.6‒3.8; Shirasaki et al., 2016a; Shirasaki et al., 2017). The degree of surface hydrophobicity of 444

PMMoV, as estimated by means of the bacterial adherence to hydrocarbon method, is probably 445

intermediate between the viruses with a relatively low hydrophobic surface (i.e., AdV and CV) and 446

the viruses with a relatively high hydrophobic surface (i.e., HAV and MNV) because PMMoV was 447

transferred to the solvent phase only when p-xylene was used as the solvent (Figure S5); AdV and 448

CV remained completely in the water phase after mixing with the test solvents, whereas HAV and 449

MNV were transferred to the solvent phase when n-octane or p-xylene were used as the solvent 450

(Shirasaki et al., 2016a; Shirasaki et al., 2017). Therefore, given that there were no marked 451

differences between the surface hydrophobicity of PMMoV and those of AdV, CV, HAV, and MNV, 452

it must have been the similar surface charges among the viruses that led to the comparable removal 453

ratios obtained. In addition, we confirmed that, unlike MS2, PMMoV was not inactivated by contact 454

with PACl during coagulation (data not shown), as evaluated by means of a combination of a local 455

lesion count assay and a PCR method. These results suggest that PMMoV is potentially a suitable 456

25

surrogate for AdV, CV, HAV, and caliciviruses during membrane filtration. Moreover, because 457

PMMoV has been detected at higher concentrations in drinking water sources than have human 458

enteric viruses (Haramoto et al., 2013), PMMoV could be a useful target virus for evaluating the 459

virus removal performances of drinking water treatment plants that use membrane filtration 460

processes. Evaluating virus removal efficiencies in this way may be a useful means of determining 461

the efficacy of membrane filtration processes for the removal of human enteric viruses as part of the 462

management of the spread of waterborne diseases caused by exposure to human enteric viruses in 463

drinking water. 464

465

466

4. Conclusions 467

(1) Different membranes had different virus removal performances during direct MF, even though 468

the nominal pore sizes of the membranes (0.1–0.22 µm) were larger than the diameters of the 469

virus particles used in the present study. Adsorptive interactions, in particular hydrophobic 470

interactions between the virus particles and the membrane surface, were the dominant factor for 471

virus removal until the adsorption sites were saturated with viruses and natural organic 472

substances. 473

(2) Direct UF with membranes with nominal MWCOs of 1 to 100 kDa effectively removed viruses 474

via size exclusion because the nominal MWCOs of the UF membranes were smaller than the 475

26

relative molecular masses of the viruses used in the present study; >4-log10 removal was 476

achieved for all of the viruses examined when a 1-kDa membrane was used. 477

(3) In-line coagulation–MF with nonsulfated, high-basicity PACls containing Al30 species removed 478

viruses more efficiently than did in-line coagulation–MF with the other aluminum-based 479

coagulants; high virus-removal performances (i.e., >4-log10 at pH 7 and pH 8) were obtained. 480

(4) The removal ratios of AdV, CV, HAV, and MNV were strongly correlated with each other, and 481

comparable removal ratios were observed for those viruses with membrane filtration. 482

(5) Although comparable removal ratios were obtained with direct MF and UF for the 483

bacteriophages MS2 and φX174 and the three human enteric viruses and MNV, MS2, unlike 484

AdV, CV, MNV, and PMMoV, was inactivated after contact with PACl during coagulation. In 485

addition, the removal ratios obtained with coagulation–MF for φX174 were clearly smaller than 486

those of the three human enteric viruses and MNV. Therefore, MS2 and φX174 do not appear 487

to be appropriate surrogates for human enteric viruses for the assessment of the efficacy of 488

coagulation–MF processes. 489

(6) The removal ratios of PMMoV obtained not only with direct MF or UF but also with 490

coagulation–MF were strongly correlated with those of the three human enteric viruses and 491

MNV, and were similar to or somewhat smaller than those of the three human enteric viruses 492

and MNV. Thus, PMMoV is a potential surrogate for human enteric viruses for the assessment 493

of the efficacy of membrane filtration processes. 494

27

495

496

Acknowledgements 497

We thank Dr. Kenji Nakahara (Division of Applied Bioscience, Research Faculty of Agriculture, 498

Hokkaido University, Japan) for providing us with seeds of Nicotiana benthamiana and Nicotiana 499

tabacum cv. Xanthi-nc and teaching us how to cultivate these plants. We also thank Dr. Daisuke 500

Sano (Division of Environmental Engineering, Faculty of Engineering, Hokkaido University, Japan) 501

for providing us with BGM cells. 502

This work was supported by the Japan Society for the Promotion of Science [grant numbers 503

25709044, 2013; 16H06103, 2016; 16H06362, 2016; 15H04064, 2015]; the Ministry of Health, 504

Labour and Welfare of Japan; and the Bureau of Water Works, Tokyo Metropolitan Government, 505

Japan. 506

507

508

References 509

Albinana-Gimenez, N., Clemente-Casares, P., Bofill-Mas, S., Hundesa, A., Ribas, F. and Girones, R. 510

(2006) Distribution of human polyomaviruses, adenoviruses, and hepatitis E virus in the 511

environment and in a drinking-water treatment plant. Environmental Science and Technology 512

40(23), 7416-7422. 513

28

514

Allouche, L., Gerardin, C., Loiseau, T., Ferey, G. and Taulelle, F. (2000) Al30: a giant aluminum 515

polycation. Angewandte Chemie International Edition 39(3), 511-514. 516

517

Bosch, A. (2007) Human Viruses in Water. Elsevier Academic Press, Amsterdam, The Netherlands. 518

519

Boudaud, N., Machinal, C., David, F., Bourdonnec, A.F.L., Jossent, J., Bakanga, F., Arnal, C., 520

Jaffrezic, M.P., Oberti, S. and Gantzer, C. (2012) Removal of MS2, Qβ and GA bacteriophages 521

during drinking water treatment at pilot scale. Water Research 46(8), 2651-2664. 522

523

Chen, Z.Y., Fan, B., Peng, X.J., Zhang, Z.G., Fan, J.H. and Luan, Z.K. (2006) Evaluation of Al30 524

polynuclear species in polyaluminum solutions as coagulant for water treatment. Chemosphere 525

64(6), 912-918. 526

527

Chen, Z.Y., Luan, Z.K., Fan, J.H., Zhang, Z.G., Peng, X.J. and Fan, B. (2007) Effect of thermal 528

treatment on the formation and transformation of Keggin Al13 and Al30 species in hydrolytic 529

polymeric aluminum solutions. Colloids and Surfaces A: Physicochemical and Engineering Aspects 530

292(2-3), 110-118. 531

532

29

Cromeans, T.L., Kahler, A.M. and Hill, V.R. (2010) Inactivation of adenoviruses, enteroviruses, and 533

murine norovirus in water by free chlorine and monochloramine. Applied and Environmental 534

Microbiology 76(4), 1028-1033. 535

536

ElHadidy, A.M., Peldszus, S. and Van Dyke, M.I. (2013) An evaluation of virus removal 537

mechanisms by ultrafiltration membranes using MS2 and φX174 bacteriophage. Separation and 538

Purification Technology 120, 215-223. 539

540

Fan, L.H., Harris, J.L., Roddick, F.A. and Booker, N.A. (2001) Influence of the characteristics of 541

natural organic matter on the fouling of microfiltration membranes. Water Research 35(18), 542

4455-4463. 543

544

Fauquet, C. M., Mayo, M. A., Maniloff, J., Desselberger, U. and Ball, L. A., Ed. (2005) Virus 545

Taxonomy: Eighth report of the International Committee on Taxonomy of Viruses. Elsevier 546

Academic Press, London, UK. 547

548

Ferrer, O., Casas, S., Galvan, C., Lucena, F., Bosch, A., Galofre, B., Mesa, J., Jofre, J. and Bernat, X. 549

(2015) Direct ultrafiltration performance and membrane integrity monitoring by microbiological 550

analysis. Water Research 83, 121-131. 551

30

552

Hamza, I.A., Jurzik, L., Uberla, K. and Wilhelm, M. (2011) Evaluation of pepper mild mottle virus, 553

human picobirnavirus and Torque teno virus as indicators of fecal contamination in river water. 554

Water Research 45(3), 1358-1368. 555

556

Haramoto, E., Kitajima, M., Kishida, N., Konno, Y., Katayama, H., Asami, M. and Akiba, M. 557

(2013) Occurrence of pepper mild mottle virus in drinking water sources in Japan. Applied and 558

Environmental Microbiology 79(23), 7413-7418. 559

560

Howe, K. J., Ed. (2006) Membrane Treatment for Drinking Water and Reuse Applications: A 561

Compendium of Peer-Reviewed Papers. American Water Works Association, Denver, CO, USA. 562

563

Jacangelo, J.G., Adham, S.S. and Laine, J.M. (1995) Mechanism of Cryptosporidium, Giardia, and 564

MS2 virus removal by MF and UF. Journal American Water Works Association 87(9), 107-121. 565

566

Kimura, M., Matsui, Y., Saito, S., Takahashi, T., Nakagawa, M., Shirasaki, N. and Matsushita, T. 567

(2015) Hydraulically irreversible membrane fouling during coagulation-microfiltration and its 568

control by using high-basicity polyaluminum chloride. Journal of Membrane Science 477, 115-122. 569

570

31

Kreissel, K., Bosl, M., Hugler, M., Lipp, P., Franzreb, M. and Hambsch, B. (2014) Inactivation of 571

F-specific bacteriophages during flocculation with polyaluminum chloride–a mechanistic study. 572

Water Research 51, 144-151. 573

574

Langlet, J., Ogorzaly, L., Schrotter, J.C., Machinal, C., Gaboriaud, F., Duval, J.F.L. and Gantzer, C. 575

(2009) Efficiency of MS2 phage and Qβ phage removal by membrane filtration in water treatment: 576

applicability of real-time RT-PCR method. Journal of Membrane Science 326(1), 111-116. 577

578

Madaeni, S.S., Fane, A.G. and Grohmann, G.S. (1995) Virus removal from water and wastewater 579

using membranes. Journal of Membrane Science 102, 65-75. 580

581

Matsui, Y., Matsushita, T., Inoue, T., Yamamoto, M., Hayashi, Y., Yonekawa, H. and Tsutsumi, Y. 582

(2003) Virus removal by ceramic membrane microfiltration with coagulation pretreatment. Water 583

Science and Technology: Water Supply 3(5-6), 93-99. 584

585

Matsushita, T., Matsui, Y., Shirasaki, N. and Kato, Y. (2005) Effect of membrane pore size, 586

coagulation time, and coagulant dose on virus removal by a coagulation-ceramic microfiltration 587

hybrid system. Desalination 178(1-3), 21-26. 588

589

32

Matsushita, T., Shirasaki, N., Matsui, Y. and Ohno, K. (2011) Virus inactivation during coagulation 590

with aluminum coagulants. Chemosphere 85(4), 571-576. 591

592

Matsushita, T., Shirasaki, N., Tatsuki, Y. and Matsui, Y. (2013) Investigating norovirus removal by 593

microfiltration, ultrafiltration, and precoagulation-microfiltration processes using recombinant 594

norovirus virus-like particles and real-time immuno-PCR. Water Research 47(15), 5819-5827. 595

596

Nwachuku, N., Gerba, C.P., Oswald, A. and Mashadi, F.D. (2005) Comparative inactivation of 597

adenovirus serotypes by UV light disinfection. Applied and Environmental Microbiology 71(9), 598

5633-5636. 599

600

Prevost, B., Goulet, M., Lucas, F.S., Joyeux, M., Moulin, L. and Wurtzer, S. (2016) Viral 601

persistence in surface and drinking water: suitability of PCR pre-treatment with intercalating dyes. 602

Water Research 91, 68-76. 603

604

Rose, J.B., Gerba, C.P. and Jakubowski, W. (1991) Survey of potable water supplies for 605

Cryptosporidium and Giardia. Environmental Science and Technology 25(8), 1393-1400. 606

607

Rosenberg, M., Gutnick, D. and Rosenberg, E. (1980) Adherence of bacteria to hydrocarbons - a 608

33

simple method for measuring cell-surface hydrophobicity. FEMS Microbiology Letters 9(1), 29-33. 609

610

Ryu, H., Mayer, B. and Abbaszadegan, M. (2010) Applicability of quantitative PCR for 611

determination of removal efficacy of enteric viruses and Cryptosporidium by water treatment 612

processes. Journal of Water and Health 8(1), 101-108. 613

614

Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G., Marinas, B.J. and Mayes, A.M. 615

(2008) Science and technology for water purification in the coming decades. Nature 452(7185), 616

301-310. 617

618

Shirasaki, N., Matsushita, I., Matsui, Y., Kobuke, M. and Ohno, K. (2009) Comparison of removal 619

performance of two surrogates for pathogenic waterborne viruses, bacteriophage Qβ and MS2, in a 620

coagulation-ceramic microfiltration system. Journal of Membrane Science 326(2), 564-571. 621

622

Shirasaki, N., Matsushita, T., Matsui, Y., Marubayashi, T. and Murai, K. (2016a) Investigation of 623

enteric adenovirus and poliovirus removal by coagulation processes and suitability of 624

bacteriophages MS2 and φX174 as surrogates for those viruses. Science of the Total Environment 625

563-564, 29-39. 626

627

34

Shirasaki, N., Matsushita, T., Matsui, Y. and Marubayashi, T. (2016b) Effect of aluminum hydrolyte 628

species on human enterovirus removal from water during the coagulation process. Chemical 629

Engineering Journal 284, 786-793. 630

631

Shirasaki, N., Matsushita, T., Matsui, Y., Murai, K. and Aochi, A. (2017) Elimination of 632

representative contaminant candidate list viruses, coxsackievirus, echovirus, hepatitis A virus, and 633

norovirus, from water by coagulation processes. Journal of Hazardous Materials 326, 110-119. 634

635

Shirasaki, N., Matsushita, T., Matsui, Y., Oshiba, A., Marubayashi, T. and Sato, S. (2014) Improved 636

virus removal by high-basicity polyaluminum coagulants compared to commercially available 637

aluminum-based coagulants. Water Research 48, 375-386. 638

639

Sinsheimer, R.L. (1959) A single-stranded deoxyribonucleic acid from bacteriophage φX174. 640

Journal of Molecular Biology 1(1), 43-53. 641

642

Tanneru, C.T. and Chellam, S. (2012) Mechanisms of virus control during iron 643

electrocoagulation–Microfiltration of surface water. Water Research 46(7), 2111-2120. 644

645

Tanneru, C.T., Rimer, J.D. and Chellam, S. (2013) Sweep flocculation and adsorption of viruses on 646

35

aluminum flocs during electrochemical treatment prior to surface water microfiltration. 647

Environmental Science and Technology 47(9), 4612-4618. 648

649

Urase, T., Yamamoto, K. and Ohgaki, S. (1996) Effect of pore structure of membranes and module 650

configuration on virus retention. Journal of Membrane Science 115(1), 21-29. 651

652

USEPA (U.S. Environmental Protection Agency). (2016) Drinking Water Contaminant Candidate 653

List 4. EPA-HQ-OW-2012-0217, Office of Water, U.S. Environmental Protection Agency, 654

Washington, DC, USA. 655

656

van Voorthuizen, E.M., Ashbolt, N.J. and Schafer, A.I. (2001) Role of hydrophobic and electrostatic 657

interactions for initial enteric virus retention by MF membranes. Journal of Membrane Science 658

194(1), 69-79. 659

660

Wang, Z.X., Elimelech, M. and Lin, S.H. (2016) Environmental applications of interfacial materials 661

with special wettability. Environmental Science and Technology 50(5), 2132-2150. 662

663

WHO (World Health Organization). (2011) Guidelines for Drinking-water Quality, fourth ed.. 664

World Health Organization, Geneva, The Swiss Confederation. 665

36

666

Zhang, P.Y., Wu, Z., Zhang, G.M., Zeng, G.M., Zhang, H.Y., Li, J., Song, X.G. and Dong, J.H. 667

(2008) Coagulation characteristics of polyaluminum chlorides PAC-Al30 on humic acid removal 668

from water. Separation and Purification Technology 63(3), 642-647. 669

670

Zhang, T., Breitbart, M., Lee, W.H., Run, J.Q., Wei, C.L., Soh, S.W.L., Hibberd, M.L., Liu, E.T., 671

Rohwer, F. and Ruan, Y.J. (2006) RNA viral community in human feces: prevalence of plant 672

pathogenic viruses. PLoS Biology 4(1), 108-118. 673

674

Zhao, B.Q., Wang, D.S., Li, T., Chow, C.W.K. and Huang, C. (2010) Influence of floc structure on 675

coagulation-microfiltration performance: effect of Al speciation characteristics of PACls. Separation 676

and Purification Technology 72(1), 22-27. 677

678

679

37

Table 1. Edo River water quality. 680

681

682

683

684

685

pH 7.5

Turbidity (NTU) 2.8

DOC (mg/L) 1.0

UV260 (cm-1) 0.020

Alkalinity (mg-CaCO3/L) 36.0

38

Table 2. Specifications of the aluminum-based coagulants used in the present study. 686

687

688

689

690

691

Aluminum concentration Sulfate concentration(% [w/w] as Al2O3) (% [w/w])

Alum Aluminum sulfate 0.0 8.1 22.6 1.3

PACl-1.5s PACl 250A 1.5 10.1 2.9 1.2

PACl-2.1s PACl 700A 2.1 10.1 2.0 1.2

PACl-2.1ns ‒ 2.1 10.4 0.0 1.2

PACl-2.5ns Takibain #1500 2.5 23.2 0.0 1.3

Taki Chemical Co.

Coagulants Basicity ManufacturerProduct name Relative density at 20ºC

39

Table 3. Specifications of the filtration membranes used in the present study.a 692

693

694

695

696

697

Nominal pore size or Contact angle Zeta potentialnominal molecular weight cutoff (°) (mV)

PVDF-0.1-LP Durapore VVLP MF 0.1 µm Hydrophilic polyvinylidene fluoride 20.2 ± 2.0 -39.1 ± 0.7 Millipore Corp.

PVDF-0.22-LP Durapore GVWP MF 0.22 µm Hydrophilic polyvinylidene fluoride ND ND Millipore Corp.

PVDF-0.1-HP Durapore VVHP MF 0.1 µm Hydrophobic polyvinylidene fluoride 136.5 ± 4.6 -33.6 ± 4.2 Millipore Corp.

PVDF-0.22-HP Durapore GVHP MF 0.22 µm Hydrophobic polyvinylidene fluoride ND ND Millipore Corp.

PTFE-0.1 Omnipore JVWP MF 0.1 µm Hydrophilic polytetrafluoroethylene 87.9 ± 3.1 -50.2 ± 1.2 Millipore Corp.

MCE-0.1 MF-Millipore VCWP MF 0.1 µm Mixed cellulose esters 55.9 ± 2.0 -50.1 ± 4.0 Millipore Corp.

PC-0.1 Isopore VCTP MF 0.1 µm Polycarbonate 44.6 ± 1.5 -33.0 ± 0.7 Millipore Corp.

ZP-0.1 Zeta Plus 90S MF 0.1‒0.2 µm Cellulose, etc. Very low 12.0 ± 7.1 3M Co.

RC-1k Ultracel PLAC UF 1 kDa Regenerated cellulose 14.6 ± 1.7 -25.9 ± 0.4 Millipore Corp.

RC-10k Ultracel PLGC UF 10 kDa Regenerated cellulose 12.2 ± 1.4 -12.7 ± 7.0 Millipore Corp.

RC-100k Ultracel PLHK UF 100 kDa Regenerated cellulose 15.8 ± 0.8 -28.7 ± 4.7 Millipore Corp.a ND, not determined. Contact angle and zeta potential of PVDF-0.22-LP and PVDF-0.22-HP are almost the same as those of PVDF-0.1-LP and PVDF-0.1-HP, respectively, according to the manufacturer’s information.Because ZP-0.1 has very high wettability, we could not accurately measure the contact angle of this membrane.

MaterialMembranes Product name MF or UF Manufacturer

40

Table 4. Pearson’s correlation coefficient matrix for the removal ratios obtained with direct MF and UF (n = 11).a 698

699

700

701

702

703

AdV CV HAV MNV MS2 φX174 PMMoV

AdV 1.00

CV 0.91** 1.00

HAV 0.93** 0.96** 1.00

MNV 0.93** 0.96** 0.91** 1.00

MS2 0.85** 0.97** 0.90** 0.91** 1.00

φX174 0.97** 0.92** 0.89** 0.95** 0.85** 1.00

PMMoV 0.98** 0.92** 0.92** 0.94** 0.84** 0.98** 1.00a Asterisks indicate a statistically significant correlation (**P < 0.01).

41

Table 5. Pearson’s correlation coefficient matrix for the removal ratios obtained with in-line coagulation‒MF (n = 18).a 704

705

706

707

708

709

AdV CV HAV MNV MS2 φX174 PMMoV

AdV 1.00

CV 0.76** 1.00

HAV 0.94** 0.85** 1.00

MNV 0.88** 0.92** 0.95** 1.00

MS2 0.81** 0.93** 0.87** 0.95** 1.00

φX174 0.51* 0.83** 0.63** 0.71** 0.77** 1.00

PMMoV 0.73** 0.96** 0.84** 0.93** 0.97** 0.83** 1.00a Asterisks indicate a statistically significant correlation (*P < 0.05, **P < 0.01).

42

710

711

712

Figure 1. Removal of virus particles by means of direct MF with different types of membranes 713

at pH 7. Values are the means of three experiments, and error bars indicate standard deviations. 714

Arrows indicate that the virus concentrations were below the quantification limit. 715

716

717

PVDF-0.1-LP

Log

arith

mic

vir

us r

emov

al(L

og10

[C0/C

f])

0

1

2

3

4

5

6

7

PVDF-0.22-LP

PVDF-0.1-HP

PVDF-0.22-HP

PTFE-0.1 MCE-0.1 PC-0.1 ZP-0.1

AdVCVHAVMNVMS2φX174PMMoV

43

718

719

720

Figure 2. Removal of virus particles by means of direct UF with membranes with different 721

nominal MWCOs at pH 7. Values are the means of three experiments, and error bars indicate 722

standard deviations. Arrows indicate that the virus concentrations were below the quantification 723

limit. 724

725

726

Log

arith

mic

vir

us r

emov

al(L

og10

[C0/C

f])

AdVCVHAVMNVMS2φX174PMMoV

RC-1k RC-10k RC-100k0

1

2

3

4

5

6

7

44

727

728

729

730

Log

arith

mic

AdV

rem

oval

(Log

10[C

0/Cf])

Log

arith

mic

AdV

rem

oval

(Log

10[C

0/Cf])

Filtration volume (mL) Filtration volume (mL)

-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

PVDF-0.1-LP PVDF-0.1-HPMCE-0.1ZP-0.1

(h) AdV(a) AdV

-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

Filtration volume (mL) Filtration volume (mL)

Log

arith

mic

CV

rem

oval

(Log

10[C

0/Cf])

Log

arith

mic

CV

rem

oval

(Log

10[C

0/Cf])

(i) CV(b) CV

-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000

Filtration volume (mL) Filtration volume (mL)

-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

Log

arith

mic

HAV

rem

oval

(Log

10[C

0/Cf])

Log

arith

mic

HAV

rem

oval

(Log

10[C

0/Cf])

(j) HAV(c) HAV

Filtration volume (mL) Filtration volume (mL)

Log

arith

mic

MN

V r

emov

al(L

og10

[C0/C

f])

Log

arith

mic

MN

V r

emov

al(L

og10

[C0/C

f])

-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000

(k) MNV(d) MNV

45

731

732

733

734

Figure 3. Change in virus removal ratio with filtration volume during direct MF at pH 7. 735

Filtration volumes were 50–250 mL (a–g) and 200–1000 mL (h–n). Arrows indicate that the virus 736

concentrations were below the quantification limit. 737

738

739

-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

Filtration volume (mL) Filtration volume (mL)

Log

arith

mic

MS2

rem

oval

(Log

10[C

0/Cf])

Log

arith

mic

MS2

rem

oval

(Log

10[C

0/Cf])

(l) MS2(e) MS2

-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000

Filtration volume (mL) Filtration volume (mL)

Log

arith

mic

φX

174

rem

oval

(Log

10[C

0/Cf])

Log

arith

mic

φX

174

rem

oval

(Log

10[C

0/Cf])

-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

(m) φX174(f) φX174

-1

0

1

2

3

4

5

6

7

0 200 400 600 800 1000

Filtration volume (mL) Filtration volume (mL)

Log

arith

mic

PM

MoV

rem

oval

(Log

10[C

0/Cf])

Log

arith

mic

PM

MoV

rem

oval

(Log

10[C

0/Cf])

-1

0

1

2

3

4

5

6

7

0 50 100 150 200 250

(n) PMMoV(g) PMMoV

46

740

741

742

743

Figure 4. Removal of virus particles by means of in-line coagulation–MF with different 744

aluminum-based coagulants at pH 7 (a) and pH 8 (b). The PVDF-0.1-LP membrane was used. 745

Values are the means of three experiments, and error bars indicate standard deviations. Arrows 746

indicate that the virus concentrations were below the quantification limit. 747

748

749

Log

arith

mic

vir

us r

emov

al(L

og10

[C0/C

f])

alum PACl-1.5s

PACl-2.1s

PACl-2.1ns

PACl-2.5ns

alum PACl-1.5s

PACl-2.1s

PACl-2.1ns

PACl-2.5ns

0

1

2

3

4

5

6

7 (a) pH 7

AdVCVHAVMNVMS2φX174PMMoV

0.54 mg-Al/L 1.08 mg-Al/L

Log

arith

mic

vir

us r

emov

al(L

og10

[C0/C

f])

alum PACl-1.5s

PACl-2.1s

PACl-2.1ns

PACl-2.5ns

alum PACl-1.5s

PACl-2.1s

0

1

2

3

4

5

6

7 (b) pH 8 1.08 mg-Al/L 2.16 mg-Al/L

47

750

751

752

Figure 5. 27Al NMR spectra of the coagulants used in the present study. 753

PACl-1.5s

PACl-2.1s

alum

PACl-2.1ns

PACl-2.5ns

10‒12

70

0

80

10‒1263

63

63

63

70

80

80

80

10‒1280

000

0

10‒1270

4

4

4