Nontarget Screening with High Resolution Mass ... - ACS Publications

Subscriber access provided by - Access paid by the | UCSB Libraries

Feature

Non-target screening with high resolution mass spectrometry in the environment: Ready to go? Juliane Hollender, Emma Louise Schymanski, Heinz Singer, and P. Lee Ferguson Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.7b02184 • Publication Date (Web): 06 Sep 2017 Downloaded from http://pubs.acs.org on September 7, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 26

Environmental Science & Technology

1

Non-target screening with high resolution mass spectrometry in the

2

environment: Ready to go?

3

Juliane Hollender1,2*, Emma L. Schymanski1, Heinz P. Singer1, P. Lee Ferguson3.

4

1

5

2

6

3

7

Durham, NC, USA 27708

8

* Corresponding Author

9

Überlandstrasse 133, 8600 Dübendorf, Switzerland

10

Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland Institute of Biogeochemistry and Pollutant Dynamics, ETH Zürich, 8092 Zürich, Switzerland Duke University, Dept. of Civil & Environmental Engineering, Box 90287

Phone: +41 58 765 5493; e-mail: [email protected]

11 12

Abstract

13

The vast, diverse universe of organic pollutants is a formidable challenge for environmental

14

sciences, engineering, and regulation. Non-target screening (NTS) based on high resolution

15

mass spectrometry (HRMS) has enormous potential to help characterize this universe – but

16

is it ready to go for real world applications? In this Feature article we argue that development

17

of mass spectrometers with increasingly high resolution and novel couplings to both liquid

18

and gas chromatography, combined with the integration of high performance computing,

19

have significantly widened our analytical window and have enabled increasingly

20

sophisticated data processing strategies, indicating a bright future for NTS. NTS has great

21

potential for treatment assessment and pollutant prioritization within regulatory applications,

22

as highlighted here by the case of real-time pollutant monitoring on the River Rhine. We

23

discuss challenges for the future, including the transition from research towards solution-

24

centered and robust, harmonized applications.

25 26

ACS Paragon Plus Environment


27

Introduction

28

Since the advent of large-scale commercial production of organic chemicals for use in

29

industry and commerce, the release of anthropogenic chemicals into the aquatic environment

30

has resulted in contamination with complex chemical mixtures that are potentially harmful to

31

aquatic and human life. The publication of Silent Spring1 in 1962 was a transformative

32

landmark that raised public awareness of the impact of chemical pollution on the

33

environment and human health. Concurrently, new developments in analytical chemistry

34

(and especially mass spectrometry) gave scientists powerful means to assess the identity

35

and occurrence of these pollutants, and the pioneers of environmental chemistry quickly took

36

advantage of this. Gas chromatography coupled with mass spectrometry with electron

37

ionization (GC-EI-MS) rapidly became the most powerful detection technique for measuring

38

environmental pollutants. By the early 1970s, the seeds of workflows for identification of

39

unknown compounds, which we now call “non-target screening” (NTS), were sowed in

40

studies reporting impressive structural elucidation of a myriad of heretofore unknown

41

pollutants in rivers, sediments, and waste streams

42

considered “legacy pollutants”: e.g., polycyclic aromatic hydrocarbons, dioxins, chlorinated

43

pesticides, flame retardants, alkylphenols, surfactants, and volatile aromatic hydrocarbons.

44

The reproducible and robust fragmentation afforded by GC-EI-MS, suitable for compilation of

45

standard spectra in libraries, made that technique the most common “magnifying glass” of

46

environmental chemists into the early 2000s. Unknown compound identification became

47

easier over time with the release of large mass spectral databases such as those from NIST

48

and Wiley in the late 1990s, which now contain spectra of several hundreds of thousands of

49

compounds.4 However, the chemical coverage of GC-MS is generally limited to volatile

50

compounds unless derivatization of non-volatiles is performed, while spectral interpretation

51

beyond library searches remains largely the domain of expert analysts. The commonly

52

observed absence or low intensity of molecular ions in GC-EI-MS spectra made molecular

53

formula determination (and subsequent elucidation efforts) of the “unknown” compounds

54

challenging.

2,3

. Many of those compounds are now


Page 2 of 26

Page 3 of 26


55

Three very different but critically important technological developments in analytical

56

chemistry and computing that ultimately revolutionized the way structure elucidation could be

57

performed in complex environmental samples occurred towards the end of the 20th century.

58

These were the advent of (1) softer ionisation techniques such as electrospray (ESI) and

59

atmospheric pressure chemical ionization (APCI), which enabled facile coupling of both gas

60

and liquid chromatography with mass spectrometry while limiting fragmentation and

61

maintaining high sensitivity; (2) robust and sensitive high resolution mass spectrometry

62

(HRMS) instruments, allowing resolving of peaks with much smaller mass differences and (3)

63

the internet, which has opened up completely new possibilities for researchers to exchange

64

and process data, far beyond a stand-alone computer connected to a single instrument.

65

Once John Fenn developed electrospray ionization,5 the tools were available to look beyond

66

volatile substances into more polar, water-soluble, and larger organic molecules such as

67

pesticides, pharmaceuticals, food additives, natural toxins, and drugs of abuse.

68

chromatography (LC) coupled to MS and MS/MS (to provide additional fragment information)

69

took their place beside GC-MS in the analyst’s toolkit, and a vital step was taken towards

70

comprehensive organic pollutant analysis in the environment. HRMS instruments suitable for

71

routine analysis are now capable of near simultaneous, sensitive untargeted detection of

72

thousands of substances within the short time frames necessary for chromatographic

73

separation. These instruments have high mass accuracy (± 0.001 Da), high mass resolution

74

(ratio of mass to mass difference ≥ 20,000) and wide mass range (simultaneous acquisition

75

of ions (full scan) up to 2000 Da).

76

The 2000s also saw the advent of openly accessible online chemical compound databases

77

such as ChemSpider and PubChem containing structures and properties of millions of

78

natural and synthetic organic chemicals, while the 2010s have yielded an explosion of online

79

mass spectral libraries (e.g. MassBank, METLIN, mzCloud) and software packages aimed at

80

processing the mountains of data generated by these HR-MS/MS instruments.

81

convergence of these technological developments has led to a fortuitous situation indeed:


Liquid

The


82

the analytical capabilities available to the environmental analytical chemist today are finally

83

ready to tackle the complexity of environmental samples.

84

When we ask the question of which anthropogenic organic compounds contaminate the

85

environment, we must first define the boundary condition: can we capture the universe of

86

anthropogenic organic chemicals? The Chemical Abstract Services (CAS) now contains

87

over 100 million entries. Reporting under the Registration, Evaluation, Authorisation and

88

Restriction of Chemicals (REACH) legislation in the European Union indicates that Europe

89

produces or imports around 140,000 substances, while the analogous Toxic Substances

90

Control Act (TSCA) in the United States contains around 85,000 chemicals. Estimates

91

indicate that between 30,000 and 70,000 compounds such as pharmaceuticals, biocides and

92

surfactants are used in households alone.6 Modelling approaches, such as those pioneered

93

by Howard and Muir,7 have attempted to estimate the quantities and fate of these myriad

94

chemicals in our environment. However, incomplete or confidential production/use

95

information as well as the generation of transformation products through environmental biotic

96

and abiotic processes complicate such predictions. Despite the immense number of

97

chemicals in production and use, regulatory monitoring is still restricted to only a small

98

number of well-known contaminants, such as 76 priority substances, 17 “watch list”

99

candidates and selected “river basin specific pollutants” for European wide monitoring within

100

the Water Framework Directive. Meanwhile, the United States Clean Water Act regulates

101

126 priority pollutants. While several thousand substances have been detected to date in the

102

environment,8,9 the number of chemicals in production and use suggests that this number is

103

only the tip of the iceberg. The adoption of NTS with HRMS is increasing rapidly at institutes

104

and commercial laboratories.10-12 However, a single measurement of a complex

105

environmental sample typically contains many thousands of signals,13,14 so that even with the

106

most sophisticated instruments and data analysis workflows, it is currently not feasible to

107

identify all the chemical structures present in such samples.

108

In this Feature we focus on the recent application of NTS with HRMS coupled to

109

chromatography and show that it is ready to help solve real world problems, opening up ACS Paragon Plus Environment

Page 4 of 26

Page 5 of 26


110

many opportunities for characterization of processes and identification of heretofore unknown

111

pollutants. The utility of NTS for the aquatic environment is highlighted with a case study of

112

real-time pollutant monitoring in the River Rhine. Aspects of NTS requiring further refinement

113

and improvement for broader, successful application in environmental science and

114

technology are covered, including future needs and opportunities within regulatory

115

frameworks.

116 117

Generic Non-target Screening Workflow

118

Figure 1 outlines a general scheme for NTS of typical environmental samples. As with every

119

investigation, NTS starts with appropriate sampling to answer the study question. Sampling,

120

enrichment and analysis should be planned considering the volatility and polarity of the

121

substance classes of interest. Water soluble, semi-volatile or non-volatile organic pollutants

122

in water are commonly analyzed with solid-phase extraction followed by reversed phase

123

liquid chromatography (LC) combined with electrospray and HRMS (e.g. Time-of-flight or

124

Orbitrap). Typically, HRMS analysis involves acquisition of full scan MS data, containing

125

mostly protonated or deprotonated molecular ions (or other adducts), plus MS/MS (or MSn)

126

data where collision-induced fragmentation of the molecules yields additional structural

127

information. A key advantage to NTS workflows compared to target analysis with low

128

resolution MS is that in addition to storing the physical environmental samples for later

129

analysis (“sample archive”, Fig. 1), data files from current-state full scan HRMS analyses can

130

also be archived and exploited retrospectively, if new questions or new knowledge arise

131

(“digital archive”, Fig. 1).



132 133

Fig. 1: Workflow for non-target screening of environmental samples

134 135

Following data acquisition, pre-processing is extremely important to reduce data quantity and

136

complexity. This typically includes peak detection, annotation or subtraction of compounds

137

present in blanks as well as “componentization” via grouping of isotopes, adducts, multi-

138

charged ions and in-source fragments to define components (i.e., grouping all signals that

139

likely belong to one unique molecular structure). Instrumental noise can be filtered out using

140

replicate measurements.15 As typical environmental samples contain hundreds of homologue

141

series (related substances with varying chain length) such as surfactants or polymers, these

142

can be linked through constant mass and retention time shifts using algorithms such as

143

envihomolog.13,16

144

Alignment of components into profiles across several samples along gradients of time, space

145

or treatment enables prioritization of specific profiles for further evaluation. This includes

146

statistical methods such as principal component analysis, clustering and regression

147

analysis.17 These data reduction routines are critical for prioritizing the most relevant and

148

interesting components in sample sets, and form the basis for hypothesis generation or

149

testing within the context of, for example, treatment technology assessment. Full


Page 6 of 26

Page 7 of 26


150

identification might not even be critical in some cases. For example, to assess treatment

151

options, bulk characterisation parameters such as peak numbers, overall reduction in mass,

152

retention times, or functional groups may provide sufficient valuable information to give

153

insight into treatment effectiveness.

154

Finally, identification of prioritized components involves all information available from MS and

155

MS/MS (molecular ion, isotope pattern, and fragments), spectra and compound databases

156

as well as meta information such as the environmental context (e.g., water or soil) and the

157

emission source or context (e.g., agricultural, household, industrial). Although NTS enables

158

tentative identification without requiring reference standards in advance, such standards are

159

needed for unequivocal confirmation and quantification and to ensure the newly-discovered

160

knowledge is verified for use (potentially as target substances) in subsequent investigations.

161 162

Emerging Analytical Technologies for NTS

163

Although GC-MS and other techniques have been used for NTS for many years, as

164

mentioned above, one of the most important new technologies for NTS of environmental

165

organic pollutants in complex matrices is HRMS.

166

resolution, and accuracy of mass spectrometers provide continuous improvements in the

167

assignment of tentative identities to components in complex environmental media. A critical

168

checkpoint in NTS for identification of an individual component is molecular formula

169

assignment. Molecular formula assignment is facilitated by recent improvements to the mass

170

accuracy and resolution of bench-top instruments such as ultra-high field Orbitrap mass

171

spectrometers (resolution > 450,000 at m/z 200, mass accuracies < 1 ppm), plus the

172

associated fine isotopic structure (e.g. observation of C,N,O, and S isotopomers directly),18

173

and (where possible) fragment information.

Refinements in the speed, sensitivity,

174 175

While LC-ESI-HRMS/MS is a common choice for NTS, many compounds are not observed

176

due to inefficient ionization or incomplete separation. Thus, alternative and complementary

177

separation and ionization methods widen the “analytical window” for NTS screening and give ACS Paragon Plus Environment


Page 8 of 26

178

additional, confirmatory information. Two-dimensional GC (GCxGC) methods coupled with

179

HRMS have been used with great success for NTS of nonpolar, bioaccumulative compounds

180

in the environment.19 Coupling of GC with HRMS through “softer” ionization methods such

181

as atmospheric pressure chemical ionization

182

application of HRMS identification workflows, providing complementary information to the

183

established GC-EI-MS workflows described in the introduction. Because EI-MS databases

184

are much larger and the spectra are more reproducible than MS/MS spectra, spectral match

185

searching with EI-MS, especially coupled with retention index information, often yields good

186

tentative identifications. While MS/MS libraries still suffer from a lack of broad coverage

187

(several thousand compounds, versus hundreds of thousands in EI-MS libraries), these are

188

growing rapidly.4

189

In the case of LC separations, limitations arise with the typical reverse phase

190

chromatography due to excessive polarity of analytes or (in complex samples) extensive

191

component co-elution and consequent matrix effects.23

192

addressed using alternative liquid separation approaches such as HILIC chromatography24,

193

ion chromatography or capillary electrophoresis25 coupled to ESI-HRMS. Multidimensional

194

liquid chromatography (LCxLC) using either two different reverse-phase columns26 or

195

comprehensive separation with size exclusion coupled to reverse-phase columns can be

196

used to address analyte co-elution prior to HRMS analysis.

197

chromatographic methods improve separation power, major challenges remain for integrating

198

the resulting data into conventional NTS data processing workflows, due to the high

199

dimensionality of the data produced by these techniques.

200

Methods beyond MS also hold promise for NTS. Nuclear Magnetic Resonance Spectroscopy

201

coupled to LC is used in natural product and metabolomics applications, but has not been

202

widely applied in environmental applications due to low sensitivity.11 High-resolution ion

203

mobility spectrometers are capable of separating geometric isomers of molecules, which is

204

not possible with co-eluting isobars in MS regardless of resolution. Commercial integration

20,21

or electron capture ionization

22

enables

The former problem has been


While two-dimensional

Page 9 of 26


205

of high resolution ion mobility spectrometers with HR-MS/MS has recently enabled the

206

separation and analysis of isobaric pollutants in wastewater.27

207 208

Challenges in Prioritization

209

The enormous effort for a true unknown identification, which can easily last several months,

210

requires rigorous prioritization approaches to focus on the most relevant sample components

211

(e.g. those that are toxic, persistent, or transformed). Table 1 compiles common prioritization

212

strategies for monitoring studies, assessment of treatment processes and complementary

213

laboratory experiments. Simple approaches include ranking of signal intensity, frequency of

214

occurrence in a dataset, as well as “suspect screening” - searching for masses of

215

compounds that are expected in the sample without the use of reference standards.

216

Fragmentation information helps find structurally related compounds, often applied for

217

identification of transformation products. In silico pathway prediction systems such as

218

enviPath (formerly UM-PPS) predict potential transformation products and these masses

219

can then be searched in NTS data using “suspect screening” approaches. Microbial,

220

oxidative or electrochemical laboratory experiments generate potential transformation

221

products with sufficient concentration to record high quality mass spectra to enable

222

subsequent discovery in environmental samples.

223

Increasingly, statistical approaches are used to prioritize components across related

224

samples, from spatial and time trends through to “before and after” comparison of treatment

225

technologies. This has expanded from time trends in laboratory-based biotransformation

226

experiments28 with increasing recognition that computational methods are essential to

227

support analytical efforts. For example, recently, NTS using LC-HRMS was essential during

228

evaluation of a pilot scale advanced oxidation process (AOP) reactor to treat wastewater to

229

fill gaps remaining after target analysis.29 For a full scale plant, overall component

230

characteristics such as retention time and mass changes were indicative of elimination

231

processes of organic compounds during activated sludge treatment, without full

232

identification.30 ACS Paragon Plus Environment


Page 10 of 26

233

The combination of controlled laboratory experiments with real world monitoring often

234

facilitates prioritization by adding new information.31,32 For example, in lab experiments

235

Kolkmann et al.31 traced mutagenic nitrogenous disinfection byproducts (N-DBPs) formed

236

with reactive N species during UV drinking water treatment by adding

237

Comparison of the labeled and unlabeled samples revealed 84 N-DBPs amongst the

238

thousands of signals, one quarter of which were later detected in samples from actual water

239

treatment facilities and prioritized for identification efforts.

240

One current limitation in such prioritization efforts is analytical matrix effects. Comparison of

241

samples with varying matrices (e.g. wastewater influent and effluent) is hampered by

242

suppression of signals in matrix-rich samples. Currently, isotopic labelled internal standards

243

are often used for correction, but robust methods to correct comprehensively for these

244

influences for unknown compounds with various functional groups and thus varying

245

ionization efficiencies are a definite future need.

15

N-labeled nitrate.

246 247

Table 1. Summary of prioritization approaches used for NTS, with selected examples.

Data-driven Frequency, signal intensity of masses 14,33

Experiment-driven Persistence34, elimination/formation29 process

over

Component with characteristic isotope pattern Reaction-based search of transformation (C, Cl, Br, N, O, S)13,19,35 products to link masses before and after treatment30 Part of homologue series (mass difference, Biological28, electrochemical38, 13,36,37 Kendrick mass defect) transformation product formation

oxidative32

Suspect screening (looking for “known” or Reaction with isotopically-labelled reagents31 predicted chemicals without standard) 39,40 Specific functional groups derivatisation, neutral loss) 41,42

(MS/MS, Effect-directed selection of masses in toxic fractions 43,44

Temporal or spatial profile over several samples 24,33

248 249

A second limitation is accounting for potential toxicity during prioritization. Approaches such

250

as effect-directed analysis use biological effect tests to prioritize chromatographic fractions

251

with unknown components associated with specific toxic effects for identification. However, ACS Paragon Plus Environment

Page 11 of 26


252

the fractionation process can be very time consuming and linking effects to unknown

253

compounds remains difficult. Success stories up to now have been the identification of

254

unknown compounds exhibiting estrogenic and glucocorticoid activity.45,46

255

chromatography47 and/or high-throughput multidimensional micro- and nanofractionation43,48

256

speed up effect-directed analysis immensely and facilitate NTS in the fractions. Virtual effect-

257

directed-analysis attempts to prioritize pollutants for identification with a given effect via

258

statistically correlating chemicals and effects data over a samples set instead of using

259

extensive fractionation.49

260 261

Identification

262

The identification of components in NTS requires gathering evidence from many different

263

sources. “Suspect screening” (see Table 1) is now a common way to expedite NTS.50 The

264

Suspect Exchange from the NORMAN Network of reference laboratories, and research

265

centers for monitoring of emerging environmental substances now contains many different

266

suspect lists. Large compound databases such as PubChem and ChemSpider or the US

267

EPA CompTox Chemistry Dashboard contain many more potential candidates than those on

268

suspect lists. They also contain useful additional data to support identification, such as

269

literature references, patent data, functional uses and toxicological/bioassay data. In general

270

for suspect screening, an exact mass match is not sufficient for identification alone.51 MS/MS

271

libraries are constantly growing

272

are now in the scientific domain. Open digital repositories with continuous screening of

273

spectral libraries such as GNPS52 will likely play a vital role in the future of retrospective

274

screening of HR-MS/MS data (digital archive, Figure 1), but this is not yet used widely in

275

environmental contexts.

276

Beyond spectral libraries, in silico fragmentation techniques assist in the discovery of “known

277

unknowns” (i.e. those chemicals in large compound databases) with NTS. The most

278

sophisticated in silico fragmentation approaches ranked the correct candidate in first place

279

one-third of the time (with an exact mass search of ChemSpider).53 Adding environmental

4,52

Thin layer

and over 1 million MS/MS spectra of >20,000 chemicals



280

context or “meta data” improved this immensely, to over 70%, greatly increasing the success

281

rate for environmental investigations.53,54 True unknowns that have not yet been documented

282

anywhere (e.g. novel transformation products) will require structure generation of candidates.

283

While this is not yet ready for routine application, some successful examples exist for very

284

small molecules where sufficient structural information is available.55

285

A number of challenges remain in candidate selection for non-target identification. Prediction

286

of ionization properties (i.e. which substances ionize under what conditions), has been

287

generally limited to rules based on a certain subset of functional groups,14,39 while broader

288

prediction methods are greatly needed. Retention time indices, while established for gas

289

chromatography and peptides in proteomics research, are not yet established for LC-based

290

techniques and are undergoing collaborative assessment (e.g. within the NORMAN

291

Network). While effect prediction has been incorporated into candidate selection in certain

292

cases, further prediction improvements are needed in terms of modes of action, applicability

293

domains and calculation speed before this is suitable for routine application in toxicant

294

identification.

295 296

A NTS Case-Study: Real-time monitoring of the Rhine River

297

While NTS is complex, it is suitable for routine monitoring applications if given time and cost

298

constraints allow. Processing dozens to hundreds of field study samples within a reasonable

299

period is resource intensive, especially the data evaluation. However, proper workflow

300

optimization and careful data prioritization can enable successful NTS under real world

301

conditions. This is showcased by the international Rhine monitoring station close to Basel on

302

the border of Switzerland and Germany, which is one of stations along the river organized

303

within the International Commission for the Protection of the River Rhine.

304

The Rhine is one of the most important rivers in central Europe, as it is a source of drinking

305

water for 20 million people. The Rhine station at Basel is responsible for daily monitoring of

306

river water quality (including long-term trend monitoring and the detection of accidental spills


Page 12 of 26

Page 13 of 26


307

originating from industry, municipalities, or agriculture) in order to protect downstream

308

populations. The monitoring strategy encompasses daily LC-HRMS and GC-MS screening

309

analysis (Figure 2); so that downstream drinking water suppliers can be warned of spills

310

within the same day and shut down their production before spills reach the drinking water

311

extraction wells.

312

The LC-HRMS monitoring concept entails: (i) automated (semi-)quantitative screening of 320

313

target compounds (with standards) for long-term trend analysis; (ii) screening of 1500

314

suspected compounds (based on usage in the Rhine catchment) to identify peak events and

315

continuous emission patterns; and (iii) NTS to detect accidental spills of previously

316

undetected or unanticipated compounds. Sample measurement is performed within a few

317

hours33 and the subsequent target, suspect and non-target data processing with time trend

318

analysis uses the streamlined software pipeline enviMass to provide final results by the end

319

of each day.

320

In 2014, 90 of 320 targets (mainly pharmaceuticals, pesticides and their transformation

321

products) were detected regularly above their limit of quantification (

Nontarget Screening with High Resolution Mass ... - ACS Publications

Recommend Documents