The relationship between MX [3-Chloro-4-

(dichloromethyl)-5-hydroxy-2(5H)-furanone],

routinely monitored trihalomethanes, and other

characteristics in drinking water in a long-term

survey.

AUTHOR NAMES

Rachel B. Smith†, James E. Bennett†, Panu Rantakokko‡, David Martinez§, Mark J.

Nieuwenhuijsen†§, Mireille B. Toledano†*

AUTHOR ADDRESS

†MRC-PHE Centre for Environment and Health, Department of Epidemiology and Biostatistics,

School of Public Health, Imperial College London, St Mary’s Campus, Norfolk Place, London,

W2 1PG, UK. ‡National Institute for Health and Welfare, Chemicals and Health Unit, Kuopio,

Finland. §Centre for Research in Environmental Epidemiology, (CREAL), Barcelona, Spain.

§Universitat Pompeu Fabra (UPF), Barcelona, Spain. §CIBER Epidemiología y Salud Pública

(CIBERESP), Madrid, Spain.

1

1

2

3

4

5

6

7

8

9

11

12

13

14

15

16

17

18

AUTHOR INFORMATION

Corresponding Author

* Dr Mireille B. Toledano, MRC-PHE Centre for Environment and Health, Department of

Epidemiology & Biostatistics, Imperial College London, St Mary’s Campus, Norfolk Place,

London W2 1PG, UK. Tel:+44 20 7594 3298 Fax:+44 20 7594 3196 Email:

m.toledano@imperial.ac.uk

Author Contributions

JB conducted statistical analysis, including modeling of trihalomethane data, and wrote initial

drafts of the paper. RBS was responsible for design of the water sampling survey in Bradford,

water sample collection, collection of additional data from water company, and writing final

draft of paper. PR conducted laboratory analyses of MX on water samples, and QA/QC of MX

data and analytical methods. DM conducted statistical analysis. MBT was responsible for

funding and oversight of UK aspects of HiWATE, and study design. MJN was responsible for

funding, oversight, overall study design of HiWATE. All authors contributed to interpretation of

data and analysis, and the manuscript was written through contributions of all authors. All

authors have given approval to the final version of the manuscript.

KEYWORDS

Disinfection by-products (DBPs); drinking water; Mutagen X (MX); seasonality;

trihalomethanes

2

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

ABSTRACT

MX (3-Chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone) is a drinking water disinfection

by-product (DBP). It is a potent mutagen and is of concern to public health. Data on MX levels

in drinking water, especially in the UK, are limited.

Our aim was to investigate factors associated with variability of MX concentrations at the tap,

and to evaluate if routinely measured trihalomethanes (THMs) are an appropriate proxy measure

for MX.

We conducted quarterly water sampling at consumers’ taps in eight water supply zones in and

around Bradford, UK, between 2007 and 2010. We collected 79 samples which were analysed

for MX using GC-HRMS. Other parameters such as pH, temperature, UV-absorbance and free

chlorine were measured concurrently, and total THMs were modeled from regulatory monitoring

data.

To our knowledge this is the longest MX measurement survey undertaken to date.

Concentrations of MX varied between 8.9 and 45.5 ng/L with a median of 21.3 ng/L. MX

demonstrated clear seasonality with concentrations peaking in late summer/early fall.

Multivariate regression showed that MX levels were associated with total trihalomethanes, UV-

absorbance and pH. However, the relationship between TTHM and MX may not be sufficiently

consistent across time and location, for TTHM to be used as a proxy measure for MX in

exposure assessment.

3

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

TOC/ABSTRACT ART

4

62

63

64

INTRODUCTION

The formation of MX (3-Chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone) during the

chlorination of water was first discovered as a result of the observation that bleaching of sewage

from wood pulp caused strong mutagenic activity of the waste water.1 The compound was later

identified as a disinfection by-product (DBP) in chlorinated drinking water.2, 3 MX is a minor

component of the highly complex mixture of DBPs by concentration, but it can give rise to a

considerable proportion of the total mutagenicity. Estimates of mutagenicity attributable to MX

in various chlorinated drinking water samples from the US, Finland, Australia, China, Japan,

Russia, Spain and the UK have ranged from 2% to 67%.4

Exposure to drinking water mutagenicity has been associated with increased risks of bladder and

kidney cancer in humans,5 and reduced birth weight and increased risk of small-for-gestational

age,6 which could reflect exposure to MX or other mutagenic DBPs, or a mixture of these. MX

is a potent direct-acting genotoxicant, and has been shown to have promoter activity, to induce

oxidative stress, and to be the most potent carcinogen of all the DBPs in animal studies.7 MX

has also been classified as possibly carcinogenic to humans.8 MX induces point mutations in

human cells in vitro, and is capable of inactivating tumour-suppressor genes.9 MX and its

analogs have been ranked as DBPs of priority concern .10, 11 The contribution of MX to toxicity

in drinking water may vary according to the specific DBP mixture in a drinking water system,

but given its genotoxic and carcinogenic potential, MX is of interest from a public health

perspective and it is important to understand how it relates to more prevalent DBPs, such as

trihalomethanes (THMs).

5

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

Previous studies have examined the factors influencing MX formation in simulated laboratory

chlorination.12-14 The aim of this paper was to investigate the factors associated with variability

of MX concentrations in public drinking water supplies at the tap, in order to inform and

improve exposure assessment in epidemiological studies of DBPs, and to understand if routinely

measured trihalomethanes (THMs) are an appropriate surrogate for epidemiological inference

regarding possible health risks of MX. This paper presents occurrence data for MX in samples

collected in the North of England between 2007 and 2010. It addresses a data gap, as there is

little information on MX concentrations in drinking water supplies in the UK. To our knowledge,

it is the longest MX measurement survey undertaken worldwide to date, thus providing a unique

opportunity to evaluate long-term temporal variation of MX.

MATERIALS AND METHODS

Sampling

MX sampling at the tap was undertaken as part of a larger water sampling campaign covering a

suite of DBPs.15 The study area supplied by Yorkshire Water company covered 8 Water Supply

Zones (WSZs, supplying <50,000 people) in and around Bradford in the North of England. Two

WSZs were supplied solely by water treatment plant A, four WSZs are supplied solely by

treatment plant B, one WSZ is supplied by a mixture of treated water from A and B and one

WSZ supplied by a mixture of A and up to 3 further plants. Disinfection was via dosing of either

chlorine or sodium hypochlorite. The raw water is drawn from a mixture of upland surface water

sources.

6

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

MX samples were collected quarterly from Quarter 3 of 2007 to Quarter 1 of 2010. In total, 79

samples were collected. The location for sample collection in each WSZ was chosen by a

random address generator from the water company’s customer database, in accordance with

standard regulatory monitoring. Sampling locations thus differ for each quarter and are customer

addresses (residential and business). All samples are finished drinking water at the point of use.

Together with MX concentrations, a large range of water parameters such as pH, temperature,

UV-absorbance, free and total chlorine, conductivity, bromide, colour and turbidity were

measured concurrently.

Analytical Methods

GC-HRMS determination of MX

The first version of the MX method was published more than 10 years ago 16 and a short version

of improvements more recently.17 Full details of sample pretreatment, instrumental analysis,

quantification and quality control are given in the Supporting Information and only a brief

description here.

MX samples were collected in a 250 ml polyethylene bottle to which ammonium sulphate was

added prior to sampling. Samples were then acidified to pH 2-3 with HCl and frozen at -20°C

until analysis. Thawed samples were pumped with tubing pump through a train of 0.45 µm

syringe filter, Waters Sep-Pak Plus tC18 cartridge (tC18) and Waters Oasis HLB Plus cartridge.

The HLB cartridge was dried with air and MX was eluted from it with acetone that was

evaporated to dryness. Internal standard (13C labelled 2,4,5-trichlorophenol), 4% H2SO4-

7

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

isopropanol, nonane and hexane were added to extracts. MX and internal standard were

isopropylated at 85°C for 1 hour. Then ultrapure water and more hexane were added, and mixed.

Hexane phase was separated, washed with ultrapure water, dried with sodium sulfate, evaporated

to a volume of about 500 µl, and transferred to autosampler vials. Recovery standard PCB-30

was added, and solvent was concentrated to 50 µl in nonane.

Instrumental analysis was performed with gas chromatograph (Hewlett Packard 6890) coupled to

high resolution mass spectrometer (Waters Autospec Ultima). The column used was Agilent

DB-5MS (20 m, i.d. 0.18 mm, film 0.18 µm).

Final results were calculated by standard addition. Limit of quantification for MX was 0.5 ng/L.

For quality control one ultrapure water reagent blank and one spiked Kuopio tap water control

sample (10ng/l of MX added) were analysed in each batch of samples. Blank levels were

negligible. Average spike recovery in 15 batches of samples was 98%, and relative standard

deviation of recoveries was 18%. Uncertainty of measurement was 51%.

Free and total chlorine were measured in situ by DPD visual comparator test kit. pH and

temperature were recorded in situ using a pH meter (Hanna HI-98128 waterproof pHep pH/c

meter). UV-absorbance was measured by UV at 254nm, TOC was measured by chemical

oxidation/infra-red spectrometry, bromide by ion chromatography, colour and turbidity by

spectrometry, and conductivity by potentiometry.

8

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

Modelled Total Trihalomethanes

Routine regulatory monitoring data for THMs from samples collected from customers’ taps in

the distribution system were provided by Yorkshire Water for the 8 WSZs in the study area.

Thus, all data used in this study reflects measurements from taps in the distribution system.

Regulatory drinking water sampling, analytical testing and data quality is subject to independent

audit by the United Kingdom Drinking Water Inspectorate (DWI)..

This routine THM monitoring dataset contained 358 sample points across the years 2006 to

2010. In order to estimate THM concentrations for those months in which no data were collected

and in order to provide robust estimates of the monthly concentrations in each WSZ, a predictive

model was used. Any bromoform or DBCM values below the limit of quantitation (LOQ) were

treated as zero concentration. No sample points were below LOQ for chloroform or BDCM.

Total trihalomethane (TTHM) for each sample point was calculated as the sum of chloroform,

bromodichloromethane (BDCM), dibromochloromethane (DBCM) and bromoform. TTHM

levels ranged from 14.2 to 95.6 µg/L. The data points were approximately evenly distributed

across the WSZs, months and years. Log-transformed TTHM was modelled using linear

regression (in the statistical package R), with a spline in month, a factor for year and a factor for

WSZ included in order to provide monthly WSZ-specific concentrations. The predictive model

was validated using 10-fold cross-validation20 (i.e. splitting the data into 10 parts and using 9

parts to predict the 10th, repeated 10 times) for the R2. The predictive model gave an R2 value

(calculated using 10-fold cross-validation) for observed versus predicted values of 0.76,

indicating a close fit to the routine THM monitoring data. This R2 reflects the predictive

performance within this routine monitoring dataset; we would expect performance when

9

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

predicting the unobserved TTHM concentrations in the actual MX tap samples to be slightly

lower. Other water characteristics were excluded from the predictive models in order to avoid

confounding any relationship between MX and the predicted TTHM in the planned analysis. The

model generated month and WSZ specific estimates of TTHM concentrations which were then

linked to corresponding MX samples.

Statistical Analysis

All statistical analyses were conducted using R 2.15.2.21 In descriptive analyses, continuous

variables (predicted TTHM, pH, free, total and combined chlorine, UV-absorbance, and

temperature) were categorised into tertiles, and due to the skewness of the MX data, geometric

means are presented. Analysis of variance (ANOVA) tests were undertaken for each variable

independently and the percentage of log(MX) variability explained and the overall p-values were

obtained. The percentage of MX variability explained by a particular variable was calculated

from ANOVA as the sum of squares for the variable of interest divided by the total sum of

squares.

Scatter plots of MX against continuous variables were produced with a Lowess smooth (a

locally weighted regression) to aid interpretation of the relationship between variables. In

boxplots of MX concentration by categorical variables, the width of box is proportional to the

square root of the number of observations contributing to that category.

Univariate and multivariate regression were undertaken after log transforming the MX values to

examine predictors of MX variation. Year, quarter, total chlorine, free chlorine, combined

10

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

chlorine, pH, temperature, UV-absorbance, predicted TTHM, colour, and conductivity were

considered as potential covariates in multivariate models. Multivariate models were selected via

a stepwise model selection procedure using the Akaike Information Criteria (AIC). Leave-one-

out cross validation using the current dataset was undertaken in order to obtain a valid value of

R2, i.e. validate the percentage of variability explained by the multivariate models.

RESULTS AND DISCUSSION

MX occurrence. MX was detected well above the limit of quantification in all 79 samples.

Concentrations of MX ranged from 8.9 to 45.5 ng/L with a median of 21.3 ng/L. These levels

are consistent with, and within maximum range of, distribution system samples in Finland (range

15-67 ng/L),22 the US (mean 28 ng/L, range 4 - 80 ng/L),23 Spain (median 16.7 ng/L, range 0.8-

54.1 ng/L),24 Canada (range not detected-141 ng/L)25 but well below the extremely high levels

(mean 180 ng/L) reported in Russia,26 and a US sampling survey (up to 850 ng/L) targeting

plants/distribution systems fed by TOC rich waters.27, 28 The only available data on levels of MX

occurring in the UK drinking water supply are from a 1990 survey of 11 treatment plant samples,

in which concentrations ranged from 1-48 ng/L.29 This range is similar to the present study,

tentatively suggesting that over the long-term there has been little change in MX levels in UK

drinking waters.

Temporal variation. There is clear evidence of a seasonal trend with MX concentrations peaking

in Quarter 3, i.e. late summer/early fall, in our study area (Figure 1, part A). This is similar to a

general pattern of higher MX concentrations in summer compared to winter observed in Canada

in 2007 (means 32 ng/L and 12 ng/L for summer and winter respectively).25 In contrast, Wright

11

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

et al.23 reported higher MX concentrations in spring compared to fall in Massachusetts, US

(mean 31 ng/L and 22 ng/L respectively). Patterns of seasonal variation may vary according to

the pathway by which the specific humic organic matter associated with MX formation enters the

raw water, or variation in average seasonal temperatures, or variation in chlorine dosing, and

these may differ by geographical location.

MX shows very similar temporal variation to TTHM, in the study area (Figure 2). Month and

year explain 59% of variability in MX concentrations (p<0.01) (Table 1). If the year and month

elements are separated out, a smaller percentage of the variability is explained by year of

sampling but results suggest an overall increase in MX concentrations in 2009 compared to 2008

(Figure 1, part B). Comparisons with annual averages for 2007 and 2010 are not valid, because

samples were not collected for all quarters in 2007 and 2010.

To our knowledge, the present study is the longest MX measurement survey undertaken. With

regular quarterly sampling over a 2.5 year timeframe, it has provided a unique opportunity to

evaluate long-term temporal variation. As far as the authors are aware, only Wright et al. 23 have

collected a comparable number of samples (n=88) but, with a shorter timeframe and fewer

sampling windows (Spring 1997, Spring 1998, Fall 1998), their data have limited ability to

assess temporal variation within and between years.

Other DBP species. MX was positively correlated (r=0.60, p<0.001) with modelled TTHM

estimates (Figure 3, part A), which explained 40% of MX variability (p<0.01) (Table 1). In our

study area, chloroform comprised the majority (mean fraction 80.75%) of TTHM, with BDCM,

12

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

DBCM and bromoform contributing 16.17%, 3.00% and 0.08% respectively. The correlation we

observe is similar to a correlation of 0.65 between chloroform and MX reported by Villanueva et

al.24, and a correlation of 0.7 between chloroform and MXsum (MX + ZMX + EMX) reported by

Onstad et al.28 in plant effluent samples, but higher than a correlation of 0.37 between TTHM and

MX reported by Wright et al.23, which suggests that the relationship between TTHM and MX

may vary by geographical location. However, whilst the correlation found here was similar to

some previous work, it must be emphasised that the TTHMs used here were modelled values

representing a WSZ-level and month average, therefore within-WSZ variation between taps and

within-month variability for THMs mean that the correlation (and regression coefficients) can

only be viewed as an approximation of the underlying relationship.

pH. Figure 3 (part B) suggests a possible U-shaped relationship between MX and pH, with

minimum MX values around pH 7.9-8.0. The tertile groupings in Table 1 are not sufficiently

aligned with this minimum to explain a significant percentage of variability, but using a linear

spline to check the relationship we observed that pH explains 10% of the variability in MX. This

could be an artefact of the data, however we note a similar U-shaped pattern in Wright et al.23

with mean MX concentrations of 35.89 ng/L, 21.9 ng/L, 28.9 ng/L in the pH categories 5.66-

7.39, 7.4-8.16 and 8.17-9.90 respectively. Our data do not include pH values lower than 7.4,

whereas in Wright et al. 23 the pH values go as low as 5.66, therefore if the U-shaped relationship

is true and not artefactual we would predict to see a ‘flatter’ U-shaped relationship in our data

compared to the US, which is indeed what we see. A U-shaped relationship, although slightly

differently located, was observed in the early MX-stability studies over a wide pH range. MX

has a local stability minimum around pH 6 -7, and is more stable around pH 8. At and above pH

13

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

10 hydrolytic degradation increases drastically. However, temperature is also very important for

MX stability, at low temperatures degradation can be slow even at pH 9.2, 30

A strong influence of pH on MX formation during laboratory simulated chlorination has

previously been observed, with MX formation usually favoured by chlorination at acidic

conditions, but with no MX detected when chlorination was performed at neutral or alkaline

conditions.13, 14 Previous studies find that MX is relatively stable at acidic pH (≤4), but is

degraded at neutral and alkaline conditions, although not monotonically.2, 30-32 Speciation of MX

is also dependent on pH with ring forms of MX favoured at pH < 7, whereas open forms are

favoured at pH > 7.28 In short, pH appears to be influential to both MX formation and

degradation, and therefore also concentration.

Free, total and combined chlorine.

MX concentrations decrease slightly across increasing total and free chlorine tertiles (Table 1).

However, scatter plots suggest that overall the relationship is weak for both total chlorine and

free chlorine (Figure 3, part C and D), with neither explaining much variability in MX (Table 1).

A scatter plot of combined chlorine (the difference between total and free chlorine) versus MX

revealed no relationship (Figure S1, part A, Supporting Information). Total, free and combined

chlorine were allowed as possible explanatory variables in the multivariate model selection

process. Previous studies observe MX concentration to be associated with chlorine dose,13, 23 but

not with chlorine residual in the distribution system.23

14

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

Organic content. UV-absorbance, colour and total organic carbon (TOC) are all indicators of

organic content, and were highly intercorrelated in this dataset (correlation coefficients ≥ 0.73).

There was a positive linear relationship between MX and UV-absorbance, which explained 30%

variability in MX concentrations (p<0.01) (Table 1, Figure 3 part E). Colour and TOC showed

similar associations with MX (Figure S1, part B and C, Supporting Information). Consistent with

our data, positive relationships between TOC and MX have previously been observed.13, 23 We

cannot comment on the influence of specific types of organic matter in our study, but Xu et al. 12

have previously identified aquatic humic substances and in particular fulvic acids in the organic

matter being the strongest contributors to MX formation in chlorinated natural waters.

Temperature. Temperature explained 14% (p<0.01) of the variability in MX concentrations

(Table 1). MX appeared to increase non-linearly with increasing temperature (Figure 3, part F),

which is consistent with laboratory findings that MX formation increases with temperature (up to

45 oC after which MX stability decreases).13 In contrast, Wright et al.23 found the highest levels

of MX corresponded to the category with the lowest temperature (1.1-6.7 oC), however few of

our samples were in this temperature range.

Water Supply Zone and Treatment Plant. Little or no variability was explained by either WSZ

(0%, p-value 0.947) or Treatment Plant supply (0.2%, p-value 0.675) (Figure S2, part A and B,

Supporting Information). This is as expected since the treatment processes were broadly the

same and the raw water used was from similar surface water sources in a relatively compact

geographical area. A difference between levels seen in two WSZs which were supplied by the

15

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

same treatment plant may have been due to differences in the positions of the WSZs in the

distribution network but there were insufficient data to evaluate this.

Other water parameters. Bromide levels were low with only 6 out of 79 samples exceeding the

limit of detection, with a maximum value of 0.035 mg/L, and thus bromide has not been used

further due to insufficient variability and also due to low potential for the formation of

brominated THMs and brominated analogues of MX.33, 34 Conductivity and turbidity showed no

discernable relationship with MX (Figure S1, parts D and E, Supporting Information).

Univariate and Multivariate Regression Models

We identified factors associated with MX variation via univariate and multivariate regression

models. There were several competing multivariate models which explained the variability in

MX to a similar degree. This is a common finding in the modeling of DBPs where many of the

possible explanatory variables are often highly correlated. Coefficients (and SEs) from two

multivariate models, together with univariate regression coefficients for all variables considered

for final models, are presented in Table 2.

Our observation of a possible U-shaped relationship influenced how we incorporated pH into

multivariate models. We use a linear spline of pH, which gives 2 parameters to estimate the

relationship between MX and pH: one to represent the relationship with pH below 8, and another

to represent the relationship with pH above 8. The combination of both pH terms gives the

relationship at pH value 8.

16

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

Model 1 includes the variables spline of pH, UV-absorbance and predicted TTHM; Model 2

includes the variables Year, spline of pH, UV-absorbance and predicted TTHM. The two models

respectively explain 47% and 52% of the variability in log(MX) concentrations (R2 from leave-

one-out cross-validation, to avoid overestimating the variability explained by these models). The

variables conductivity, colour, temperature, chlorine (total, free or combined) and quarter, were

not selected in any of the possible models.

Models 1 and 2 perform similarly, but Model 1 may be more useful since it is not reliant on the

unmeasurable Year factor, particularly given that years 2007 and 2010 do not represent a full

calendar year of sampling. There may be some differences in MX concentrations between the

years which are not explained by the other available measured explanatory variables, leading to

Year being selected for inclusion in the model. In contrast, the significant quarterly temporal

variation is largely explained by other variables, such as predicted TTHM and UV absorbance,

and a factor for Quarter is not necessary in the multivariate models.

Our models clearly outline the combination of parameters which are important in the prediction

of MX concentrations at the tap, and which should be measured in future studies evaluating MX,

or when conducting MX exposure assessment.

However, although indicative of the relationships we would expect to see in other areas of the

UK using similar treatment processes and water sources, the multivariate models presented here

should only be used to determine zone estimates for this area and within the timeframe of the

study. Extrapolation of the models to other regions or years carries with it the risk that the

17

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

interrelationships between the variables in the models, which give rise to the exact relative sizes

of the model parameter estimates, may vary.

The main influences on MX concentrations at the tap are temporal, together with recognised

determinants of disinfection by-product formation such as UV-absorbance (an indicator of

organic matter) and pH. Our findings are consistent with Wright et al.23 who also find TOC, pH

and temporal factors (amongst other variables) to be significant predictors of MX at the tap in

multivariate regression. We observed a possible U-shaped relationship between MX and pH,

that is consistent with results presented in Wright et al23 for real drinking water samples,

although it was not discussed by these authors. As mentioned above, stability tests confirmed a

U-shaped relationship in the late 1980s, but the minimum occurred at slightly lower pH of 6-7.30

Due to the similar nature of the treatment plants supplying the area no insight could be gained

regarding the influence of the physical and chemical treatment of the raw water on MX levels.

However, these influences have been addressed in previous literature (e.g. Wright et al.23 and

Onstad et al.28). The present study has instead focused on water characteristics in the distribution

system and how these, together with the month and year of sampling, influence the variability in

MX levels at the tap.

One limitation of this work is that the THM concentrations were not sampled concurrently with

the MX, with inference relying instead on predicted levels of TTHM obtained by modelling

routine sampling data. The TTHM predictions were made for each WSZ and month and

therefore do not replicate spatial variability within WSZs, nor day to day variation within each

18

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

month. Also, we have only certain input variables to use, and we will not be able to capture or

explain all variation in multivariate models.

Whilst TTHM is clearly associated with MX, that association varied by year from a correlation

of 0.70 (n=10) for 2007, 0.84 (n=38) for 2008, 0.33 (n=31) for 2009 to -0.02 (n=8) for 2010 data.

There is absolutely no evidence for a lack of predicted TTHM model fit to the routine TTHM

data in 2009 and therefore we must conclude that the weaker correlation between 2009 levels of

MX and predicted TTHMs is real and not an artefact of the use of predicted TTHMs. Nor do the

available water parameter data fully explain the levels of MX in 2009. The coefficient for 2009

in the multivariate model presented is significant. We must conclude that there are further

unexplained sources of variability in MX not accounted for in these data and that these are

particularly strongly present in 2009. These factors could take the form of weather patterns,

operational changes in processing or distribution, or water characteristics not measured in this

study. Wright et al.23 also note considerable variability in the relationship between TTHM and

MX in different sampling time periods.

We have shown that TTHM explains 40% of MX variability, and correlates quite well with MX

in our study area. However, we observe some variation in this relationship by year, and what

limited literature is available also suggests that the relationship between TTHM and MX can

vary by location and time period. We conclude therefore, that TTHM by itself may not be a

sufficiently consistent surrogate for MX for reliable epidemiological inference. Epidemiological

studies investigating potential health effects of MX should undertake sufficient longitudinal MX

19

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

sampling to enable the fitting of multivariate predictive MX models for use in robust exposure

assessment.

20

405

406

Table1: MX concentrations (ng/L) in relation to categorised variables, 2007-2010, Bradford, UK.

Data Range n MinimumMX

(ng/L)

MaximumMX

(ng/L)

GeometricMean MX

(ng/L)

IQR a

MX(ng/L)

All samples 79 8.9 45.5 21.0 13.1Year/ Month 2007 Sept 5 16.1 28.6 22.2 5.5

Nov 5 26.8 45.5 33.9 5.42008 Mar 8 8.9 24.0 14.0 3.0 May 7 13.1 22.4 16.3 4.2 Aug 8 22.5 44.2 30.0 2.8 Nov 7 15.3 23.1 18.9 2.62009 Mar 8 11.0 36.0 18.5 7.2 May 8 14.0 44.2 26.6 18.2 Sept 7 19.7 37.0 28.0 7.8 Nov 8 21.1 28.7 24.3 3.22010 Mar 8 9.6 20.0 13.2 5.1%Variability, p-value 59% p < 0.01

Predicted TTHM

26.2 – 37.1 µg/L 26 8.9 36.0 15.1 6.837.2 – 51.9 µg/L 26 13.5 45.5 22.2 11.252 – 81.5 µg/L 27 14.7 44.2 27.5 8.2%Variability, p-value 40% p < 0.01

pH 7.4 – 7.6 32 9.6 45.5 19.9 12.17.7 – 7.9 22 10.3 40.8 20.8 16.68.0 – 8.7 22 8.9 44.2 23.9 9.0%Variability, p-value 3% p = 0.11

Total Chlorine

<0.05-0.15 mg/L 21 11.0 45.5 23.8 10.20.16-0.25 mg/L 23 10.3 37.0 21.2 13.50.26-0.8 mg/L 34 8.9 40.8 19.4 14%Variability, p-value 4.5% p = 0.06

Free Chlorine

<0.10 mg/L 24 11.0 44.2 22.4 13.20.10 – 0.15 mg/L 24 10.3 45.5 21.2 11.30.20 – 0.75 mg/L 30 8.9 40.8 19.9 15.0%Variability, p-value 2% p = 0.26

UV-absorbance

0.018 – 0.025 m-1 26 8.9 36.0 16.6 7.40.026 – 0.036 m-1 26 10.3 40.8 19.9 9.50.037 – 0.057 m-1 27 16.6 45.5 27.9 8.2%Variability, p-value 30% p < 0.01

Temperature 4.7 – 9.2 oC 22 8.9 36.5 16.5 10.19.3 – 13.9 oC 23 13.1 45.5 23.7 7.914.0 – 19.7 oC 23 13.1 44.2 23.7 10.6%Variability, p-value 14% p < 0.01

21

407408

a IQR, interquartile range

22

409410

Table 2. Univariate and Multivariate Coefficients for linear regression model of log(MX)

Data Range Univariate RegressionSlope (SE)

MultivariateModel 1d

Slope (SE)

MultivariateModel 2 e

Slope (SE)Year 2007 a -0.349 (0.131) * 0.304 (0.102) **Year 2009 -0.040 (0.129) 0.159 (0.069) *Year 2010 -0.438 (0.145) -0.048 (0.132)Quarter 2 b Apr – Jun 0.256 (0.138)Quarter 3 Jul – Sept 0.507 (0.131) ***Quarter 4 Oct - Dec 0.525 (0.142) ***Total Cl (mg/L) <0.05 - 0.80 -0.015 (0.273)Free Cl (mg/L) <0.05 - 0.75 0.048 (0.304)Combined Cl c (mg/L) 0.00 – 0.47 -0.360 (0.703)pH spline values < 8 7.4 – 8.7 -0.322 (0.277) -0.320 (0.213) -0.276 (0.202)pH spline values > 8 0 – 0.7 1.208 (0.549) * 0.938 (0.411) * 0.681 (0.395) .Temperature (oC) 7.3 – 19.7 0.012 (0.014)UV absorbance (1/m) 0.018–0.057 22.10 (4.098) *** 10.433 (5.677) . 10.483 (5.697) .Pred TTHM (µg/L) 30.0 – 69.1 0.013 (0.003) *** 0.577 (0.175) ** 0.508 (0.2) *Colour (Hazen) 0.35-3.70 0.215 (0.060) ***Conductivity (µS/cm) 122-331 0.0006 (0.001)

DoF=71, R2=0.51 (R2

cv1=0.47)DoF=68, R2=0.56

(R2cv1=0.52)

“.” p <0.10, * p<0.05, ** p<0.01, *** p<0.001. aYear coefficients are relative to 2008. b Quarter Coefficients are relative to 1st Quarter (Jan-Mar). c Combined Cl calculated as Total Cl minus Free Cl.R2

cv1: leave-one-out cross validation.d The equation for Model 1 is:

log ( MX )=10.433× UV +0.577× TTHM−0.320 × pH +0.938 × PHc[pH >8]

23

411412413

414415416417418419

Where pHc[pH>8] pHc[ pH>8]=¿{ 0 , pH ≤8

pH−8 , pH >8¿

e The equation for Model 2 is:

log ( MX )=0.304 ×Year2007+0.159 × Year2008−0.048 ×Year2009+10.483 ×UV +0.508× TTHM−0.276 × pH +0.681× PHc[pH> 8]

Where pHc[pH>8] pHc[ pH>8]=¿{ 0 , pH ≤8

pH−8 , pH >8¿

24

420

421

422423424425

426427

428

Figure 1: MX in relation to A) Quarter and B) Year, 2007-2010, Bradford, UK.

B)A)

25

429430431432433434

Figure 2:MX and predicted TTHM in relation to time, 2007-2010, Bradford, UK.

26

435436

437438

Figure 3: MX in relation to A) predicted TTHM and B-F) other water characteristics, 2007-2010, Bradford, UK.

Footnotes for Figure 3: Super-imposed red line is a Lowess smooth, a locally weighted regression line.

F)

A) B)

D)C)

E)

27

439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484

ACKNOWLEDGEMENTS

This research was funded by HiWATE (Health Impacts of Long-Term Exposure to

Disinfection By-products in Drinking Water in Europe) [EU 6th Framework Programme

Contract no. Food-CT-2006-036224], the Joint Environment & Human Health

Programme [NERC grant NE/E008844/1], and an ESRC studentship [PTA-031-2006-

00544]. The MRC-PHE Centre for Environment and Health is funded by the UK Medical

Research Council and Public Health England. We thank Yorkshire Water, particularly

Cameron Hamilton, for assisting this study with provision of routine sampling data,

allowing us to piggy-back additional sampling onto their routine regulatory sampling

programme, and for making their knowledge of the study area available. We thank Teija

Korhonen for her valuable assistance in the laboratory analysis of MX. We thank Nina

Iszatt and Susan Edwards for their valuable assistance with sample and data collection.

SUPPORTING INFORMATION AVAILABLE

Full details of MX analytical methods and quality control are given in Supporting

Information. Plots of MX in relation to combined chlorine, colour, TOC, conductivity,

turbidity, water supply zone, and treatment plant supply are given in Figures S1 and S2 in

Supporting Information. This information is available free of charge via the Internet at

http://pubs.acs.org .

28

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

REFERENCES

1. Holmbom, B.; Voss, R. H.; Mortimer, R. D.; Wong, A., Fractionation, Isolation,

and Characterization of Ames Mutagenic Compounds in Kraft Chlorination Effluents.

Environmental Science & Technology 1984, 18, (5), 333-337.

2. Meier, J. R.; Knohl, R. B.; Coleman, W. E.; Ringhand, H. P.; Munch, J. W.;

Kaylor, W. H.; Streicher, R. P.; Kopfler, F. C., Studies on the Potent Bacterial Mutagen,

3-Chloro-4-(Dichloromethyl)-5-Hydroxy-2(5h)-Furanone - Aqueous Stability, Xad

Recovery and Analytical Determination in Drinking-Water and in Chlorinated Humic-

Acid Solutions. Mutation Research 1987, 189, (4), 363-373.

3. Hemming, J.; Holmbom, B.; Reunanen, M.; Kronberg, L., Determination of the

Strong Mutagen 3-Chloro-4-(Dichloromethyl)-5-Hydroxy-2(5h)-Furanone in Chlorinated

Drinking and Humic Waters. Chemosphere 1986, 15, (5), 549-556.

4. McDonald, T. A.; Komulainen, H., Carcinogenicity of the Chlorination

Disinfection By-Product MX. Journal of Environmental Science and Health, Part C

2005, 23, (2), 163-214.

5. Koivusalo, M.; Jaakkola, J. J. K.; Vartiainen, T.; Hakulinen, T.; Karjalainen, S.;

Pukkala, E.; Tuomisto, J., Drinking-Water Mutagenicity and Gastrointestinal and

Urinary-Tract Cancers - an Ecological Study in Finland. American Journal of Public

Health 1994, 84, (8), 1223-1228.

29

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

6. Wright, J. M.; Schwartz, J.; Dockery, D. W., The effect of disinfection by-

products and mutagenic activity on birth weight and gestational duration. Environ Health

Perspect 2004, 112, (8), 920-5.

7. Richardson, S. D.; Plewa, M. J.; Wagner, E. D.; Schoeny, R.; DeMarini, D. M.,

Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection by-

products in drinking water: a review and roadmap for research. Mutat Res 2007, 636, (1-

3), 178-242.

8. IARC, I. A. f. R. o. C., Some Drinking-water Disinfectants and Contaminants,

including Arsenic. IARC Monographs on the Evaluation of Carcinogenic Risks to

Humans 2004, 84.

9. Hakulinen, P.; Yamamoto, A.; Koyama, N.; Kumita, W.; Yasui, M.; Honma, M.,

Induction of TK mutations in human lymphoblastoid TK6 cells by the rat carcinogen 3-

chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone (MX). Mutat Res 2011, 725, (1-2),

43-9.

10. Woo, Y. T.; Lai, D.; McLain, J. L.; Manibusan, M. K.; Dellarco, V., Use of

mechanism-based structure-activity relationships analysis in carcinogenic potential

ranking for drinking water disinfection by-products. Environ Health Perspect 2002, 110

Suppl 1, 75-87.

11. Hebert, A.; Forestier, D.; Lenes, D.; Benanou, D.; Jacob, S.; Arfi, C.; Lambolez,

L.; Levi, Y., Innovative method for prioritizing emerging disinfection by-products

30

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

(DBPs) in drinking water on the basis of their potential impact on public health. Water

Res 2010, 44, (10), 3147-65.

12. Xu, X.; Lin, L.; Huixian, Z.; Yongbin, L.; Liansheng, W.; Jinqi, Z., Studies on the

precursors of strong mutagen [3-chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-

furanone]MX by chlorination of fractions from different waters. Chemosphere 1997, 35,

(8), 1709-1716.

13. Zhuo, C.; Chengyong, Y.; Junhe, L.; Huixian, Z.; Jinqi, Z., Factors on the

formation of disinfection by-products MX, DCA and TCA by chlorination of fulvic acid

from lake sediments. Chemosphere 2001, 45, (3), 379-385.

14. Backlund, P.; Wondergem, E.; Voogd, K.; Dejong, A., Influence of Chlorination

Ph and Chlorine Dose on the Formation of Mutagenic Activity and the Strong Bacterial

Mutagen 3-Chloro-4-(Dichloromethyl)-5-Hydroxy-5(2h)-Furanone (Mx) in Water.

Chemosphere 1989, 18, (9-10), 1903-1911.

15. Nieuwenhuijsen, M. J.; Smith, R.; Golfinopoulos, S.; Best, N.; Bennett, J.;

Aggazzotti, G.; Righi, E.; Fantuzzi, G.; Bucchini, L.; Cordier, S.; Villanueva, C. M.;

Moreno, V.; La, V. C.; Bosetti, C.; Vartiainen, T.; Rautiu, R.; Toledano, M.; Iszatt, N.;

Grazuleviciene, R.; Kogevinas, M., Health impacts of long-term exposure to disinfection

by-products in drinking water in Europe: HIWATE. J Water Health 2009, 7, (2), 185-

207.

16. Rantakokko, P.; Yritys, M.; Vartiainen, T., Matrix effects in the gas

chromatographic-mass spectrometric determination of brominated analogues of 3-chloro-

31

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

4-(dichloromethyl)-5-hydroxy-2(5H)-furanone. Journal of Chromatography A 2004,

1028, (2), 179-188.

17. Carrasco-Turigas, G.; Villanueva, C. M.; Goni, F.; Rantakokko, P.;

Nieuwenhuijsen, M. J., The effect of different boiling and filtering devices on the

concentration of disinfection by-products in tap water. Journal of environmental and

public health 2013, 2013, 959480.

18. Carrasco-Turigas, G.; Villanueva, C. M.; Goni, F.; Rantakokko, P.;

Nieuwenhuijsen, M. J., The effect of different boiling and filtering devices on the

concentration of disinfection by-products in tap water. Journal of environmental and

public health 2013, 2013, 959480.

19. Malliarou, E.; Collins, C.; Graham, N.; Nieuwenhuijsen, M. J., Haloacetic acids in

drinking water in the United Kingdom. Water Res 2005, 39, (12), 2722-2730.

20. Lewis, P. D., R for Medicine and Biology. Jones and Bartlett Publishers: Sudbury,

MA, 2010.

21. R Development Core Team R: A language and environment for statistical

computing. R Foundation for Statistical Computing., Vienna, Austria. URL

http://www.R-project.org/. 2012.

22. Kronberg, L.; Holmbom, B.; Reunanen, M.; Tikkanen, L., Identification and

Quantification of the Ames Mutagenic Compound 3-Chloro-4-(Dichloromethyl)-5-

Hydroxy-2(5h)-Furanone and of Its Geometric Isomer (E)-2-Chloro-3-(Dichloromethyl)-

32

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

4-Oxobutenoic Acid in Chlorine-Treated Humic Water and Drinking-Water Extracts.

Environmental Science & Technology 1988, 22, (9), 1097-1103.

23. Wright, J. M.; Schwartz, J.; Vartiainen, T.; Maki-Paakkanen, J.; Altshul, L.;

Harrington, J. J.; Dockery, D. W., 3-Chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-

furanone (MX) and mutagenic activity in Massachusetts drinking water. Environ Health

Perspect 2002, 110, (2), 157-64.

24. Villanueva, C. M.; Castano-Vinyals, G.; Moreno, V.; Carrasco-Turigas, G.;

Aragones, N.; Boldo, E.; Ardanaz, E.; Toledo, E.; Altzibar, J. M.; Zaldua, I.; Azpiroz, L.;

Goni, F.; Tardon, A.; Molina, A. J.; Martin, V.; Lopez-Rojo, C.; Jimenez-Moleon, J. J.;

Capelo, R.; Gomez-Acebo, I.; Peiro, R.; Ripoll, M.; Gracia-Lavedan, E.; Nieuwenhujsen,

M. J.; Rantakokko, P.; Goslan, E. H.; Pollan, M.; Kogevinas, M., Concentrations and

correlations of disinfection by-products in municipal drinking water from an exposure

assessment perspective. Environ Res 2012, 114, 1-11.

25. Kubwabo, C.; Stewart, B.; Gauthier, S. A.; Gauthier, B. R., Improved

derivatization technique for gas chromatography-mass spectrometry determination of 3-

chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone in drinking water. Anal Chim Acta

2009, 649, (2), 222-229.

26. Egorov, A. I.; Tereschenko, A. A.; Altshul, L. M.; Vartiainen, T.; Samsonov, D.;

LaBrecque, B.; Maki-Paakkanen, J.; Drizhd, N. L.; Ford, T. E., Exposures to drinking

water chlorination by-products in a Russian city. Int J Hyg.Environ Health 2003, 206,

(6), 539-551.

33

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

27. Krasner, S. W.; Weinberg, H. S.; Richardson, S. D.; Pastor, S. J.; Chinn, R.;

Sclimenti, M. J.; Onstad, G. D.; Thruston, A. D., Occurrence of a new generation of

disinfection byproducts. Environmental Science & Technology 2006, 40, (23), 7175-

7185.

28. Onstad, G. D.; Weinberg, H. S.; Krasner, S. W., Occurrence of halogenated

furanones in US drinking waters. Environmental Science & Technology 2008, 42, (9),

3341-3348.

29. Horth, H.; Fielding, M.; James, C. P.; James, H. A.; Gwilliam, R. D. Identification

of Mutagens in Drinking Water (EC9105) Final report to the Department of the

Environment DoE 2489-M/1; 1990.

30. Holmbom, B.; Kronberg, L.; Smeds, A., Chemical-Stability of the Mutagens 3-

Chloro-4-(Dichloromethyl)-5-Hydroxy-2(5h)-Furanone (Mx) and E-2-Chloro-3-

(Dichloromethyl)-4-Oxo-Butenoic Acid (E-Mx). Chemosphere 1989, 18, (11-12), 2237-

2245.

31. Vartiainen, T.; Heiskanen, K.; Lotjonen, S., Analysis and Some Chemical-

Properties of Mx (3-Chloro-4-(Dichloromethyl)-5-Hydroxy-2(5h)-Furanone), the Potent

Drinking-Water Mutagen. Fresen J Anal Chem 1991, 340, (4), 230-233.

32. Miettinen, I.; Martikainen, P.; Vartiainen, T.; Lotjonen, S., Biochemical and

Chemical Degradation of 3-Chloro-4-(Dichloromethyl)-5-Hydroxy-2(5h)-Furanone (Mx)

in Surface and Drinking-Water. Chemosphere 1993, 27, (9), 1707-1718.

34

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

625

33. Richardson, S. D.; Thruston, A. D.; Caughran, T. V.; Chen, P. H.; Collette, T. W.;

Floyd, T. L., Identification of new drinking water disinfection byproducts formed in the

presence of bromide. Environmental Science & Technology 1999, 33, (19), 3378-3383.

34. Suzuki, N.; Nakanishi, J., Brominated Analogs of Mx (3-Chloro-4-

(Dichloromethyl)-5-Hydroxy-2(5h)-Furanone) in Chlorinated Drinking-Water.

Chemosphere 1995, 30, (8), 1557-1564.

35

626

627

628

629

630

631

632