Introduction

Chemical disinfection of water with strong oxidizing agents is a widely extended practice used to produce potable water in most developed countries. However, this process has the unintended consequence of forming disinfection byproducts (DBPs). Since the discovery of the first carcinogenic DBP (i.e., chloroform) in drinking water in 1974 [1, 2], concerns over DBP exposure have increased through the years because of epidemiologic studies showing associations between consumption of chlorinated water and bladder and colorectal cancer [3–6], and adverse reproductive outcomes, including spontaneous abortion, increased risk of low birth weight at term, still birth, and preterm delivery [7–16].

The formation of DBPs is mainly related to the type of the disinfection treatment applied, and the nature of the water source in terms of its natural organic matter (NOM) and bromide and iodide content. More than 600 DBPs have been identified in drinking water to date [17–20]. However, despite intense identification efforts, about 50 % of the halogenated DBPs formed during water chlorination still remains unidentified, as assessed through total organic halogen (TOX) measurements [17]. Most iodine containing DBPs (iodo-DBPs) belong to the aforementioned unknown fraction, and these chemicals are becoming a new focus because of their higher toxicity compared with the brominated and chlorinated analogues [21–27]. In addition, one of these iodo-DBPs, iodoacetic acid, was recently shown to be tumorigenic in mice [28]. Furthermore, iodide levels in waters intended for drinking are likely to increase in the near future in coastal areas affected by climate change (where water scarcity will lead to an increased use of groundwater resources, moving seawater intrusion inland, and desalinated water) and areas impacted by hydraulic fracturing activities that produce highly saline water with high iodine content (as recently reported by Parker et al. [29]). Based on this, further research on iodo-DBPs, in terms of compound detection and confident chemical characterization, in addition to toxicity assessment, is required. This is critical in order to evaluate the potential risk of exposure to these compounds and the possible consequences on human health.

Characterization of iodo-DBPs in disinfected water is a challenging task because of the combined challenge of their occurrence at low concentrations and the complexity of the analytical matrix. This has been usually carried out using high resolution mass spectrometry (HRMS) techniques. Magnetic sector mass spectrometry coupled to gas chromatography has been the technique of choice to discover semivolatile iodo-DBPs in DBP mixtures [18, 30]. Operation of this type of analyzer in an untargeted manner (i.e., full scan mode) on one hand requires a high degree of expertise and on the other may not provide the required sensitivity and scan speed at a high mass resolving power. The recent development of Orbitrap-based GC-MS benchtop instruments enables a unique opportunity to operate using full scan acquisition with very high mass resolving power and mass accuracy, whilst at the same time offering the ability to detect very low concentrations. Thus, the present work aimed at evaluating the application of Orbitrap GC-MS for the characterization of iodinated DBPs in chlorinated and chloraminated DBP mixtures. In order to accomplish this, DBP mixtures were generated in lab-scale disinfection reactions using waters with various NOM, iodide, and bromide content, and were analyzed. Automatic peak deconvolution software was used to detect compounds, and putative identification of these compounds was performed using mass spectral matching against commercially available libraries combined with a comprehensive accurate mass-based rationalization of the entire mass spectrum.

Materials and methods

Chemicals and reagents

All reagents were purchased from Sigma-Aldrich (Barcelona, Spain) unless otherwise specified. Amberlite XAD-2 and Supelite DAX-8 resins were used for water extraction. Anhydrous Na2SO4 was used to dry the extracts. The list of solvents used includes: Chromasolv grade methanol (≥99.9 %, MeOH), and ethyl acetate for pesticide residue analysis (≥99.9 %, EtAc). Potassium phosphate dibasic trihydrate (K2HPO4·H2O) and potassium phosphate monobasic (KH2PO4) (≥98 %) were used to buffer at pH 7.5 the disinfection reactions. Sulfuric acid (95–97 %, H2SO4), hydrochloric acid (≥37 %, HCl), and sodium hydroxide pellets (≥98 %, NaOH) were ACS grade.

Reverse osmosis isolated NOM from Nordic Lake (NL) (Skarnes, Norway) was purchased from the International Humic Substances Society (IHSS) (St. Paul, MN, USA). Purified water (18 MΩ/cm) from an Aurum ultrapure water system (Sartorius, Madrid, Spain) was used to prepare all reagent solutions and to perform disinfection reactions with NL NOM.

Reagents used to measure free chlorine in the water samples and in the chlorine (HOCl/OCl-) and monochloramine (NH2Cl) solutions by means of the N,N-diethyl-p-phenylenediamine (DPD) ferrous titrimetric method [31] were: barium diphenylamine-4 sulfonate for redox titration, potassium dichromate (>99 %, Cr2K2O7), ethylenediaminetetraacetic acid disodium salt dehydrate (99–101 %, EDTA), DPD sulfate salt (>98 %), ferrous ammonium sulfate hexahydrate (99 %, FAS), orthophosphoric acid (85 %, H3PO4), and sodium phosphate dibasic (99 %, Na2HPO4).

Sodium hypochlorite (NaOCl) solution (10 %, w/v, reagent grade) was purchased from Panreac (Barcelona, Spain). Ammonium chloride (>99.99 %, NH4Cl) was used to produce NH2Cl.

Disinfection reactions

All disinfection reactions were performed in a Pyrex glass reaction vessel filled up to the top (headspace-free; total volume of 17.1 L). Reactions were conducted over 72 h at room temperature (22–26 °C) in the dark, under continuous stirring using a magnetic stir plate and a polytetrafluoroethylene (PTFE)-coated stir bar. In all experiments, water was buffered to pH 7.5 with phosphate buffer (10 mM) prior to oxidant addition. Either H2SO4 or NaOH (1 N) was used to adjust the solution pH.

NOM solutions were prepared at a concentration of 5 mg/L by weighing the corresponding amount of NL NOM. Five hundred ppb of bromide (as KBr) and 50 ppb of iodide (as KI) were also added to this matrix in order to promote formation of iodinated and brominated DBPs during disinfection reactions. Similar bromide and iodide levels have been reported to occur in source waters used for drinking water production at different locations [18, 22, 32, 33].

In the case of the Llobregat River (LLOB) experiments, water was collected directly from the midpoint of the river at a time at one single location. The sampling point was located close to the intake of a drinking water treatment plant that gives service to part of Barcelona and its metropolitan area. Due to the high natural levels of iodide and bromide of this surface water, disinfection reactions were performed without the extra addition of iodide and bromide ions. Characteristics of the water matrices used for the experiments are summarized in Table S1 [see Electronic Supplementary Material (ESM)].

The chlorine dose was selected based on the specific chlorine demand of the NOM of each water matrix that resulted in approximately 0.5 mg/L of residual chlorine at the end of the disinfection reaction. The disinfectant doses used to treat NL NOM solutions and LLOB water were 4 mg/L and 7.5 mg/L of Cl2, respectively.

Monochloramine reactions were carried out with freshly prepared preformed NH2Cl. This solution was prepared at a 0.7 Cl/N molar ratio by dropwise addition of HOCl to a NH4Cl solution at pH 8.5. The reaction was performed under continuous stirring with the reaction flask immersed in an ice bath for 30 min.

Two different procedural blanks were also analyzed: (a) Purified water containing NL NOM, buffered but not treated (no disinfectant), and (b) LLOB water buffered but not treated. Both blanks were concentrated in the same manner as the treated samples, and they were used to investigate whether the compounds detected and identified were generated during disinfection treatments (i.e., DBPs) or were contaminants already present in the source waters or were artifacts generated during the sample preparation treatments.

Generation of DBP mixture concentrates

Up to 16 L of the treated water was concentrated using XAD resins. Details about the preparation and cleaning of these resins can be found elsewhere [34] and are provided in ESM. After acidification, the water samples (pH < 1) were passed through a resin column that contained a combination of DAX-8 resin over XAD-2 resin at a ratio of 770:1 (v/v) of water to resin to maximize adsorption of organic compounds and to minimize breakthrough. The combination of these two resins is effective for absorbing a wide range of organic compounds. Retained analytes were eluted with EtAc at a ratio of 9.5:1 (v/v) of solvent to resin. The EtAc extract was dried with the aid of a separator funnel and Na2SO4, rotary-evaporated to approximately 5 mL, and further concentrated under a gentle stream of N2 to 0.8 mL (see detailed extraction procedure in ESM). This extract was directly injected into the GC-Orbitrap MS for analysis of iodo-DBPs.

GC-Q/Orbitrap MS analyses

A total of six water concentrates were analyzed by means of GC Orbitrap based MS. The Q Exactive GC was coupled to a Trace 1300 GC (Thermo Scientific). A volume of 1 μL of the DBP mixture extract was injected in splitless mode using a TriPlus RSH injector (Thermo Scientific) with a splitless time of 1 min. The injector temperature was set at 280 °C.

In order to increase the resolution of chromatographic peaks, the GC separation of extracted components was carried out on a Thermo Scientific TR-5MS column, 60 m × 0.25 mm (i.d.) × 0.25 μm (film thickness) using helium as carrier gas at a constant flow of 1.2 mL/min and the following GC temperature program: initial temperature of 40 °C held for 1 min, increased at a rate of 15 °C/min to 325 °C and held for 10 min. The GC was interfaced with the Q Exactive GC instrument via a transfer line heated at 280 °C. The source temperature was set at 250 °C. MS analyses were performed using electron ionization (EI) at 70 eV in full-scan mode using a scan range of m/z 50–650 and a resolution of 60,000 (full width at half-maximum [FWHM] at m/z 200).

Data analysis

Data were both manually queried and automatically processed with TraceFinder software. TraceFinder provided automatic peak detection, spectral deconvolution, and compound identification based on spectral library matching and accurate mass information. In all cases, a mass window of ±2 ppm was used to enable generation of highly selective extracted ion chromatograms with reduced chemical interferences from the matrix background.

Results and discussion

Compound identification workflow

Figure 1 summarizes the workflow applied for the detection and molecular structure characterization of iodo-DBPs. Data acquired in EI was sent to the TraceFinder that uses unique software workflows and algorithms to automate the discovery and identification of compounds. First, the software performs peak detection, spectral deconvolution, spectral library searching (e.g., NIST, Wiley, or custom made) and validates candidate library hits using high resolution filtering (HRF) based on candidate elemental formulae [35].

Fig. 1
figure 1

Workflow applied for identification of iodo-DBPs and confidence levels of the iodo-DBP molecular structures proposed

Full size image

Data processing with TraceFinder was simultaneously performed for all DBP mixtures generated with the same matrix (i.e., NL NOM, chlorinated NL NOM, and chloraminated NL NOM) to visualize fold changes with respect to a particular analyte of interest. This process provides a heat map of components that facilitates the identification of the most relevant compounds generated during each disinfection process. Due to the large number of compounds detected following deconvolution (>2500 peaks were found in the chloraminated NL NOM extract using a total ion current (TIC) intensity threshold of 500,000 and a signal-to-noise (S/N) threshold of 10:1), a mass filter was employed to limit the chemicals of interest only to those containing iodine (m/z 126.90392). This resulted in fewer compounds detected (less than 15 major peaks in the chloraminated NL NOM extract), but with all of them containing iodine in their mass spectra.

Moreover, despite the fact that iodine does not produce a distinctive mass spectral isotopic pattern (it has only one stable isotope), iodo-DBPs may also include chlorine or bromine atoms in their structures, which give distinctive isotopic patterns, due to their two stable isotopes each (m/z 35 and 37 for chlorine and m/z 79 and 81 for bromine). Consequently, this feature was also used for elucidation of iodo-DBP chemical structures.

Following the discussion started by Schymanski et al. [36] on identification confidence levels for small molecules, three different confidence levels for iodo-DBP identification were considered (Fig. 1). If no analytical standards were available, but enough evidence was obtained through fragment rationalization (assisted by the Mass Frontier software) and the presence of isotopic patterns in the molecular ion, an identification confidence level 3 for the chemical structure proposed was assigned. If there was also evidence provided by a mass spectral library match (such as NIST), the confidence level for the chemical structure proposed was risen from level 3 to level 2. The confidence level was further risen to 1 if the chemical structure proposed could be confirmed with the analysis of a reference analytical standard by a match of retention times and mass spectra.

Iodinated DBPs in disinfected water extracts

According to the literature [18, 30, 37–39], iodo-DBP formation (as well as subsequent cytotoxicity and genotoxicity of treated water [24]) is enhanced during chloramination reactions when iodide is present in source waters prior to disinfection. This was also observed in the experiments performed here (normalized extracted ion chromatograms [XICs] at m/z 126.90392 for the different analyzed extracts are provided in Figures S1 and S2, ESM). Interestingly, this was more noticeable in NL NOM extracts compared with the LLOB water extracts. Profiles of formed iodo-DBPs in NL NOM and LLOB water samples were also different, with a higher number of peaks containing iodine in NL NOM than in LLOB extracts, which could be attributed to the difference in NOM characteristics of the waters tested. According to specific UV absorbance (SUVA) measurements (see ESM Table S1), which indicate the aromatic content per unit concentration of carbon [40], the organic matter of the LLOB water has a relatively high content of hydrophobic, aromatic, and high molecular weight fractions, whereas NL NOM contains mainly non-humic, hydrophilic, and low molecular weight materials. Thus, different reactivity with disinfecting agents is expected. The increased formation of iodo-DBPs in NL NOM extracts is consistent with three other studies showing increased formation of iodo-trihalomethanes (I-THMs) from hydrophilic and low molecular weight precursors [39, 41, 42]. However, LLOB water extracts are expected to contain overall higher amounts of total organic iodine (TOI) (parameter not measured in this study) and high-molecular weight iodo-DBPs (less amenable to GC/MS analysis than low and medium molecular weight compounds) than NL NOM extracts, since hydrophobic and high MW precursors (related to high SUVA values) were observed to be more reactive with iodine in the formation of TOI and unknown TOI (not attributed to I-THMs) [42]. Furthermore, different levels of bromide and iodide in the tested waters (Br–/I– concentration ratios of 8 and 46 in NL NOM solutions and LLOB water, respectively, see Table S1 in ESM) may also rule iodine and bromine incorporation into NOM, and consequently, iodo-DBP formation. In this respect, Jones et al. reported an increased iodo-DBP formation (I-THMs) in waters with low Br–/I– ratio compared with waters with high Br–/I– ratio (2 versus 10). Additionally, I–/dissolved organic carbon (DOC) ratio has also been pointed as an important factor in iodo-DBP (I-THMs) formation (e.g., increased formation of I-THMs in waters presenting high I–/DOC ratios) [41].

A total of 11 different iodo-DBPs, most of them iodine-containing halomethanes, were detected and identified in the extracts analyzed. Chemical structures were proposed for all compounds with a minimum confidence level of 2 after applying the workflow described in the previous section, with the exception of ethyl iodoacetate and iodoethene (confidence level 3). Experimental and theoretical masses of molecular and fragment ions, the mass difference (in ppm) between them, the assigned elemental compositions for each diagnostic ion, and the proposed chemical structures for the identified DBPs are shown in Tables 1 and 2.

Table 1 Iodo-DBPs characterized in disinfected NL NOM waters
Full size table
Table 2 Iodo-DBPs characterized in disinfected LLOB waters
Full size table

For some DBPs, the molecular ion was not observed because of extensive fragmentation under EI, as shown in Fig. 2 for dichloroiodomethane. To determine the elemental composition, the presence of the molecular ion is essential for correct mass spectrum interpretation. In this respect, the use of softer ionization sources, e.g., negative/positive chemical ionization (NCI or PCI) or atmospheric pressure chemical ionization (APCI) for GC may help to identify and confirm additional iodo-DBPs in these extracts.

Fig. 2
figure 2

Mass spectrum of dichloroiodomethane (tR = 6.33 min) (a) in an analytical standard solution, and (b) in the chloraminated NL NOM extract

Full size image

During this study, the scan range was not extended below m/z 50 and, as such, some lower mass fragments of these compounds were not measured. This was the case for the peaks appearing at 8.03 and 8.14 min, as shown in Fig. 3. The elemental composition of the molecular ion and the fragments observed suggests these peaks to be ethyl iodoacetate and iodoethene, respectively. As it can be observed in the mass spectral library records for these compounds, m/z 29 (C2H5 +) and m/z 42 (COCH2 +) in the case of peak at 8.03 min, and m/z 27 (CH2 = CH+) in the case of peak at 8.14 min, are additional fragments of these compounds that contribute to increase the confidence level in the molecular structure proposed. The remaining fragments observed fit with the proposed structures; however, because the fragments’ relative abundance is not in complete agreement with that provided by a NIST mass spectral library search (see Fig. 3), the molecular structure assigned to the elemental composition of C4H7O2I and C2H3I cannot be explained with a confidence level higher than 3. Differences in fragments’ relative abundance could be attributed to differences in the type of mass analyzers used to construct the spectral library used. Ethyl iodoacetate was present in all investigated extracts (including the blanks), and therefore it may be formed during the experimental procedure because of the use of ethyl acetate as extraction solvent. However, the intensity of this peak in the chloraminated NL NOM extract was 10 times higher [60 × 106 counts per second (cps)] than in the other samples (6 × 106 cps) and, therefore, it was considered to be also formed during chloramination of the NL NOM solutions. A similar situation was observed for the peak appearing at retention time 3.7 min, tentatively identified as iodomethane (confidence level 2) (see ESM Fig. S3). Iodomethane was present in all NL NOM extracts, but its concentration noticeably increased during chlorination and chloramination of NL NOM solutions.

Fig. 3
figure 3

Fragment rationalization for mass spectra of peaks appearing at (a) tR = 8.03 min, and (b) tR = 8.14 min in the chloraminated NL NOM extract and comparison with mass spectra of the tentatively assigned molecular structures provided by the MS library (NIST chemistry webbook)

Full size image

Ion mass spectra for additional iodo-DBPs identified are provided as supplementary material (see ESM Figs. S4–S11). All iodo-DBPs identified in NL NOM and LLOB disinfected waters were generated during both chlorination and chloramination treatments, except chlorodiiodomethane and bromodiiodomethane, which were exclusively formed during chloramination. Similar levels of iodo-DBPs were found in chloraminated and chlorinated LLOB waters, whereas iodo-DBP formation was enhanced during chloramination of NL NOM solutions compared with chlorination (i.e., 8- to 66-fold higher base peak areas at XIC 126.90392 in the chloraminated extract compared with the chlorinated extract, up to 145 in the case of diiodomethane).

Despite the fact that most of the characterized iodo-DBPs belong to the well-known iodo-DBP class of I-THMs, this work reveals the formation of chloriodomethane and ethyl iodoacetate in chloraminated waters for the first time, and points to iodoethene as a novel iodo-DBP that may generate during disinfection of waters containing iodine.

Conclusions

This work has shown the successful application of GC-Orbitrap MS for the characterization of iodinated DBPs in disinfected water extracts. By using this novel GC-MS technology, further scientific insights into the formation of volatile and semivolatile DBPs generated during disinfection treatments are possible.

To the authors’ knowledge, this work would represent the first evidence on the formation of the tentatively identified iodo-DBPs, chloroiodomethane and ethyl iodoacetate, in chloraminated water, despite the fact that their presence in chlorinated waters is well known [30]. In the case of iodoethene, no previous records on its formation during water disinfection processes were found in the peer-reviewed literature.

The consistent sub-ppm mass accuracy even for compounds present at low concentrations, and the power of high resolution, allows for clear mass spectrum interpretation and ultimately for compound identification and chemical structure elucidation. Further investigations using softer ionization techniques, such as chemical ionization, may help to identify and confirm the presence of additional iodo-DBPs in these extracts.

Ultimately, the workflow described here can also be applied to discover “unknown” chlorine- and bromine-containing DBPs in DBP mixtures, as the identification of all major components of DBP mixtures generated during disinfection treatments is essential to further minimize the potential risks of exposure.