Characterization of Disinfection By-Products from ... - ACS Publications

Article pubs.acs.org/est

Characterization of Disinfection By-Products from Chromatographically Isolated NOM through High-Resolution Mass Spectrometry Bradley D. Harris,† Taylor A. Brown,† Jimmie L. McGehee,† Dominika Houserova,† Benjamin A. Jackson,† Brandon C. Buchel,† Logan C. Krajewski,§ Andrew J. Whelton,‡,∥ and Alexandra C. Stenson*,† †

Department of Chemistry and ‡Department of Civil Engineering, University of South Alabama, Mobile, Alabama 36688, United States § Department of Chemistry and Biochemistry, Florida State University, Tallahassee, Florida 32306, United States S Supporting Information *

ABSTRACT: As levels of natural organic matter (NOM) in surface water rise, the minimization of potentially harmful disinfection by-products (DBPs) becomes increasingly critical. Here, we introduce the advantage that chromatographic prefractionation brings to investigating compositional changes to NOM caused by chlorination. Fractionation reduces complexity, making it easier to observe changes and attribute them to specific components. Under the conditions tested (0.1−0.4 g of Cl to g of C without further additives), the differences between highly and less oxidized NOM were striking. Highly oxidized NOM formed more diverse Clcontaining DPB, had a higher propensity to react with multiple Cl, and tended to transform so drastically as to no longer be amenable to electrospray-ionization mass spectral detection. Less-oxidized material tended to incorporate one Cl and retain its humiclike composition. N-containing, lipidlike, and condensed aromatic structure (CAS)-like NOM were selectively enriched in mass spectra, suggesting that such components do not react as extensively with NaOCl as their counterparts. Carbohydrate-like NOM, conversely, was selectively removed from spectra by chlorination.

■

INTRODUCTION Chlorination is among the most popular methods of combating waterborne diseases. Unfortunately, reactions between chlorinating agents and natural organic matter (NOM) create potentially harmful disinfection by-products (DBPs).1 Correlation between chronic exposure to DBPs and bladder cancer2−4 and between DBP formation and early-warning toxicity markers5 is suspected. As NOM concentrations increase in surface waters,6,7 improved removal is desirable, especially to meet drinking water maximum contaminant levels of trihalomethanes (THMs) and haloacetic acids (HAAs) set by the Environmental Protection Agency.8 Meanwhile, Hrudey and others make strong arguments that THMs and HAAs are not potent enough at typical drinking-water concentrations to be solely responsible for health impacts associated with DBPs.1,9 Correspondingly, because bioassay experiments (e.g., Neale et al.5) do not optimally capture volatile DBPs such as THMs and HAAs, it is likely that toxicity observed in such experiments is derived, in part, from unknown DBPs. It must, therefore, be examined which NOM components are most reactive toward chlorinating agents and form the most potent DBPs. Because each water-treatment plant (WTP) is different in terms of both raw-water composition and treatment procedure, it is important to independently evaluate all variables (dosage, pH, flocculation, and ration) and NOM components (e.g., ligninlike, © XXXX American Chemical Society

tanninlike, hydrocarbons, and lipids). Here, the most polar components of NOM (fulvic acids and their subfractions) were analyzed because these are the components that likely remain in solution after flocculation, the step after which many WTPs chlorinate. Fulvic acid is expected to resemble the NOM remaining postflocculation because it is established that flocculation of NOM is selective toward large, hydrophobic molecules such as humic acids, leaving hydrophilic NOM behind.10−12 It is acknowledged, however, that much is still uncertain about the effect of flocculation, as results differ between fractionated NOM and whole water,11 fractionation schemes13 and analyses.10−12,14−16 Electrospray-ionization Fourier transform ion cyclotron mass spectrometry (ESIFTICR-MS) results, for example, suggest that flocculation preferentially removes NOM components of low H/C ratios and middle to high O/C ratios.14−16 The high O/C selectivity is surprising because it suggests relatively polar material. In short, much is still unknown about the effect of flocculation, but fulvic acid as a model for NOM remaining postflocculation is a good starting point. Suwannee River fulvic acid (SRFA) was Received: July 16, 2015 Revised: October 29, 2015 Accepted: November 10, 2015

A

DOI: 10.1021/acs.est.5b03466 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology chosen because it is arguably the best-characterized fulvic acid and readily available for parallel studies. Similarly, for chlorination, size-based separation, fluorescence, NMR, and UV work have long suggested that highmolecular-weight, hydrophobic, aromatic NOM has the highest THM-formation potential.17 However, Uyguner et al. examined UV−vis, fluorescence, and NMR techniques and found that none of these measures provided universally reliable THMformation potential predictions across NOM types.18 Zhang et al.,14,19 Lavonen et al.,15 and Gonsior et al.16 pioneered the use of ESI-FTICR-MS to characterize nonvolatile DBPs. Zhang et al. concluded that NOM molecules with low O/C ratios react preferentially, whereas Lavonen et al. and Gonsior et al. concluded the opposite. Given the complexity of NOM, the number of possible reaction pathways14,15,20 and the different ways in which “most reactive” can be assessed, it is not surprising that different NOM components can appear most reactive. Notably, no prior team could test their hypotheses on NOM fractionated into high- versus low-O/C ratio components. Additionally, the complexity of the starting NOM impedes data interpretation, as some components are converted into DBPs that take up the same van Krevelen space or the same NMR shift region as other, pre-existing, NOM components. Novel here is the fact that the complexity of the starting material is significantly reduced through chromatographic fractionation so that transformations to NOM are clearly visible and can more readily be attributed to specific components. Furthermore, the chromatographic step adds additional information (retention time and polarity), which constrains data interpretation and allows for a more direct correlation between the synthetic NOM-like standards and the actual NOM components, as is demonstrated on a select number of synthetic standards. Finally, the combination of a universally available NOM standard with offline fractionation allows for the independent assessment of all relevant parameters through varied analyses across research teams. Here, we demonstrate this potential by analyzing SRFA subfractions at two different Cl dosages for pH and residual Cl.

Table 1. Concentration, Doses, and Elution Times sample ID

TOC (mg/L)

Cltot (mg/L)

SRFA FRC 1 FRC 7a

16.8 ± 0.2 23.9 ± 0.5 17.62 ± 0.02 114.0 ± 0.5 10.9 ± 0.1

3.44 4.20 1.24

8.03 ± 0.07 N/A

FRC 30b FRC 50c FRC 65 DI water

Cltot/ TOC

refractionated (Y/N)

elution time (min)

0.2 0.2 0.1

N N Y

N/A 1.78−1.97 2.69−3.25

8.59

0.1

Y

6.67−7.21

4.10

0.4

Y

10.4−11.2

0.76

0.1

N

13.3−13.6

8.60

N/A

N

N/A

a

Fractions 6−8 combined. bFractions 28−30 combined. cFractions 49−52 combined. Cltot refers to total free Cl concentration.

available, samples were refractionated (i.e., reinjected and purified with the same HPLC method (Table 1)). Selectivity. Predictable fractionation makes more NOM components accessible to analysis,22−25 FTMS provides unprecedented compositional detail. However, isolation of SRFA from NOM, HPLC separation, and desalting with ZipTip pipet tips all incur sample loss. Moreover, ESI of complex mixtures is highly selective (e.g., ref 26), preventing accurate mass balance tracking in a meaningful way. Thus, ESIFTICR-MS of select fractions provides only a snapshot of some representative transformations. One advantage of the method described here, however, is that simple, robust, and predictable HPLC separation, combined with the use of standard reference material, readily allows future analyses to expand upon what is possible in a single study and eventually provide as complete a picture of NOM transformations as possible. Chlorination. SRFA and five chromatographic subfractions (FRC) (two early-eluting (FRC 1 and 7), one mid-eluting (FRC 30), and two late-eluting (FRC 50 and 65)) were chlorinated at 0.2 ± 0.1 g/g total Cl to total organic C (Cltot/ TOC), similar to doses employed in WTPs (e.g., Lavonen et al.15). Specific doses are listed in Table 1. The poor precision in dose measurements is the result of sample limitation, which affected the reliability of mass and TOC measurements. Please see the Concentration, Dose, and Stoichiometry and Sample Limitation sections for a more detailed discussion. Importantly, the apparent variation in dose did not affect the results, given that the fraction that received the highest apparent dose (FRC 50) showed the least transformation, whereas the fraction that received the lowest apparent dose (FRC 7) showed the greatest transformation. Overall, no correlation exists between the degree of transformation and DBPs formed and the apparent dose. Conversely, a direct correlation exists between the elution order and the transformation and DBPs formed. A 5.65−6.00% NaOCl stock (laboratory grade) was used to chlorinate. Accurate free Cl concentration was determined with a Pocket Colorimeter II and DPD Free Chlorine Reagent Powder (Hach). Aliquots for prechlorination samples were dried immediately as detailed above. During chlorination, all samples sat in a dark cabinet at room temperature (∼22 °C) for 3 days. Afterward, aliquots for postchlorination analyses were dried. At the Cl to TOC doses employed here (Table 1), residual free Cl concentration (Cltot) was 0 after 3 days for all samples. Samples had essentially the same pH after chlorination (average pH: 4.8 ± 0.9). As expected, early fractions and SRFA were

■

EXPERIMENTAL SECTION Materials. Unless stated otherwise, all chemicals were purchased from Sigma-Aldrich or Fisher Scientific and used asis. DI water was obtained from a Synergy Ultrapure 18.3 Millipore system. SRFA Standard II (2S101F) was purchased from the International Humic Substances Society (IHSS). HPLC Fractionation. SRFA was fractionated and pooled as described previously.21 Briefly, SRFA was separated using a Water’s Corporation X-Bridge phenyl column (3.5 mm, 4.6 mm × 150 mm). The mobile phase consisted of DI water (18.2 MΩ) and 0.1% formic acid (Optima LC/MS grade) with a step gradient of increasing acetonitrile (Optima LC/MS) and 0.1% formic acid concentration. Fractions were dried with a Labconco Centrivap Cold Trap at 30 °C overnight. In addition to pooling fractions from replicate separations, adjacent fractions were combined, where necessary, to obtain sufficient quantities for analysis (Table 1). Refractionation. In a control experiment, nitrile gloves, markers and other common laboratory items were identified as the significant sources of chemical-noise ions (e.g., CHO3S class and fatty acids). These were reduced by avoiding gloves, most plastics, and markers and by carefully cleaning the work area before opening samples. Where enough material was B


Article

Environmental Science & Technology slightly more acidic than the two late fractions, but all measured pHs were below 6. Thus, reactions may have produced mainly kinetically favored products and proceeded under slightly acidic conditions. Total Organic Carbon. TOC was measured at the Dauphin Island Sea Laboratory (DISL). Aliquots were reconstituted in house with 9 mL of DI water before transport at ∼4 °C. TOC analysis was conducted on a Shimadzu TOC-L CPH/CPN in accordance with EPA method 415.1. Calibration standards of 0, 2, 4, and 5 mg of C/L were employed. More detail is available in the Supporting Information. Desalting. ESI-FTICR-MS aliquots, including the DI water control, were desalted with a microscale solid-phase extraction method on a 10 μL C4 or C18 ZipTip (Millipore) modified from Kido et al.27 To avoid noncovalent adducts of Cl− forming in the ESI source, we used formic acid instead of hydrochloric acid as in Gonsior et al.16 Briefly, aliquots were reconstituted in 2% formic acid under sonication (2 min) and then centrifuged (2 min). Elution occurred with a decreasing step gradient of methanol (100:0, 70:30, 50:50, and 30:70 MeOH/H2O by volume). All visible material was removed with the initial 100% methanol elution. Once ZipTipped, samples were immediately dried as above. Early-eluting fractions (FRC 1 and 7) were ZipTipped with a C4 ZipTip (all others with a C18 ZipTip). Potential contamination from the NaOCl stock, ZipTips, and formic acid was addressed in two ways. (1) A water control was chlorinated and ZipTipped (equivalent to the samples). Peaks that were prominent in the water control were disregarded, as were peaks that could have arisen from carry-over (i.e., were present in the control or blanks and were at less than threshold abundance in the sample spectrum). More detail is available in the Supporting Information. (2) An unchlorinated fraction (FRC 30 pre) was analyzed with and without ZipTipping (Figures 1 and 2) to ascertain how the specific ZipTipping procedure employed here affected the spectra. ESI-FTICR-MS. ESI-FTICR-MS analyses were performed at the National High Magnetic Field Laboratory (Tallahassee, FL) on a custom-built, passively shielded 9.4 T ESI Fourier transform ion cyclotron resonance mass spectrometer in negative-ion mode.28 Samples were dissolved in 70:30 H2O/

Figure 2. Effect of ZipTipping: Fraction 30 pre analyzed with and without ZipTipping (same solution split in two). Top row: Kendrick plots. (A) overlay without abundance information. (B) FRC 30 without ZipTiping. (C) FRC 30 after ZipTipping. Bottom row: van Krevelen plots. (D) Overlay without abundance information. (E) FRC 30 without ZipTipping. (F) FRC 30 after ZipTipping. WZ, without ZipTipping. AZ, after ZipTipping. Ovals encircle CHON with very high DBEs (21−24); these formulas are of suspect assignment but were retained in the plot to allow for extremely conservative assessment of all possible changes induced by the C18 ZipTip.

MeOH by volume and direct-injected. A dual-stage ion funnel (front of funnel 1: −120 V, back of funnel 1: −70 V, front of funnel 2: −40 V, back of funnel 2: −10 V) and quadrupole mass-preselection of ions greater than m/z 300 were employed to increase sensitivity to NOM. The scan range was m/z 200− 1500. A blank was injected before and after each sample. The number of scans collected varied from 100 to 371 (Table S-1 and Figure S-4 caption) for a maximum potential variation in S/N of a factor of 1.9. Samples collected for only 100 scans were those that significantly exceeded their chlorinated counterparts in S/N. Notably, Kido et al.27 report that reproducibility (peaks shared between spectra) of ESIFTICR-MS results for NOM does not significantly improve between 150 and 500 scans. Mass Calibration and Data Analysis. Mass spectra were analyzed with Composer 7 (Sierra Analytics). Because of sample limitation, no internal standard (which might have swamped the NOM signal) had been added. Thus, spectra were calibrated on previously identified SRFA and characteristic chemical noise ions in several steps. Reliable calibration did not extend significantly past m/z 550 because S/N, mass accuracy, and resolving power all decrease with increasing m/z, whereas the number of isobars and possible elemental combinations increase. Residual errors for calibrants are listed in Table S-1 and were below 0.6 ppm. Elements considered were C, H, N, O, S, and Cl. Up to three Cl were considered for all spectra. The inclusion of up to three N and S, one P, and the combination of N and Cl were spot-checked. No credible assignments were found for N > 2, S > 1, P > 0 or the combination of N and Cl. Where possible, the presence of 37Cl isotopologues at credible relative abundances was spot-checked. In most instances, the S/N ratio is so low for CHOCls that the isotopologue assignment for CHOCls with a single Cl is not generally reliable. In prechlorination samples, we erred on the side of retaining formula assignments even if they did not appear strictly credible (on the basis of location in van Krevelen plots, DBE relative to other ions within the same class, and relative abundance). In postchlorination spectra, we did the opposite. In combination, we erred on the side of under-

Figure 1. Average oxidation number of C versus fraction number. Solid squares: CHOpre. Diamonds: CHOpost. Stars: CHOCl1post. Green: SRFA whole. Orange (FRC 30) empty square: before ZipTipping; empty triangle: after ZipTipping. Black line: CHOpre. Light gray line: CHOCl1post. Dark gray line: CHOpost. See also Figure S-3. C


Article

Environmental Science & Technology

Differences in Size. As observed previously, spectra of later eluting fractions are more likely to contain multiply charged ions.23 On the basis of m/z spacing between monoisotopic ions and their 13C isotopologues, we determined that up to three charges were observed for FRC 65. FRC 50 and 30 contain doubly and singly charged ions. FRC 1 and 7 contain only singly charged ions (Table S-1). Therefore, later-eluting fractions contain molecules large enough to allow multiple charging.23 Unfortunately, multiply charged ions could not be assigned by Composer, primarily because their accurate monoisotopic masses overlap exactly with those of singly charged ions (i.e., in formula, they are the equivalent of multimers of singly charged NOM). Also, the presence of multiply charged ions in the later-eluting fractions was closely tied to overall S/N ratio (Table S-1), not to chlorination, and thus revealed nothing about changes in size caused by chlorination. Within the limited range for which molecular formulas were obtained (m/z 300 − ∼550), Mn, Mw, Mz, and Kendrick plots do not reveal differences in size between unchlorinated fractions. However, for all chlorinated samples (except FRC 65 and SRFA), the highest observable mass decreased postchlorination. For all samples (except FRC 50), Mn, Mw, and Mz also decreased slightly postchlorination. Together, this might indicate oxidative fragmentation but requires further examination. Spectral Transformations after Chlorination. Overall, signal abundance differed by a factor of ∼50 between the strongest spectrum (FRC 7 pre) and the weakest (FRC 1 post). Figure S-4 shows total abundance for each spectrum. For earlyeluting fractions, the signal was lower after chlorination, as expected. For late-eluting fractions, the opposite trend was observed. LCMS data confirm the trend (Figure S-2B). Given that this trend occurs at widely different experimental conditions, the data suggest that it corresponds to actual differences between reaction pathways of late- versus earlyeluting samples. Conversely, the effect of the weak base (NaOCl) on the ionization potential of the less-acidic, lateeluting fractions should be insignificant here because (a) solution pH of FTMS samples was acidic, (b) residual NaOCl was removed from the organic material, either through ZipTipping (FTMS) or HPLC (LCMS), and (c) concentration of 0.1% FA in the mobile phase (LCMS) should have compensated for any residual effect of the weak base prior to injection. Enhancement in signal post chlorination for late-eluting fractions may indicate reaction pathways, such as the addition of ClOH to double bonds, as proposed as the major pathway by Lavonen et al.15 Such an addition would produce C−OH next to an electron-withdrawing C−Cl group and likely improve gasphase acidity. The opposite trend for early-eluting fractions (i.e., decrease in signal) suggests significant conversion of NOM to DBPs not amenable to FTMS. Total abundance for FRC 1 and FRC 7 drop by a factor of 2 and 33, respectively. NOM signal relative to the chemical noise (S/Nchem), which can serve as a quasiinternal standard here, drops even more substantially (by a factor of ∼50 for FRC 1 and ∼4000 for FRC 7). Bulk SRFA was transformed similarly to early fractions (S/Nchem decreased by a factor of ∼400, Table S-1). In contrast, for FRC 65, S/ Nchem increased by a factor of ∼4; for FRC 50, it stayed essentially the same (increase by a factor of ∼1.4).

estimating the number of newly formed DBPs. Finally, the background-subtract and exclusion-list features in Composer 7 were used to subtract carry-over and chemical noise that was also present in the chlorinated DI water control, as best possible. More detail on advantages and limitations of this approach is available in the Supporting Information. Cltot, UV−Vis, LCMS, and Chlorination of Standards. A total of eleven subfractions from five pooled fractionations were analyzed for TOC, and then diluted and dosed with NaOCl at a ratio of 1.6 Cltot/TOC, mass-to-mass. This dose was chosen because preliminary experiments (e.g., Figure S-1) indicated that at Cltot/TOC ratios between 1.6 and 3.2 showed the largest differentiation in chlorine reaction potential. Samples were chlorinated for 3 days as above. Residual Cl was measured with an eight-point calibration curve (r2 > 0.98) on a Shimadzu UV-2101 PC Spectrometer (Shimadzu, Japan) at 528 nm. A total of three replicate measurements were collected in each assay. At least two assays were examined for each fraction. Additional data were collected with a Hach colorimeter (as above). Absorbance at 254 nm was measured for specific UV absorbance (SUVA). LCMS analysis of unchlorinated controls and chlorinated samples was conducted with the same HPLC method described above on an LTQ Velos ion-trap mass spectrometer (Thermo Scientific) in negative-ion mode, scanning from m/z 250−900. Chlorination and LCMS analysis were repeated for 15 standards with elution times resembling the fractions’ (Table S-2). Limitations and Shortcomings. Limitations of ESI-MS in terms of reproducibility, selectivity, adduct -formation, and peak splitting are detailed in the literature (e.g., refs 15, 26, 27, and 29−33). Descriptions of limitations relevant here are provided in the Supporting Information.

■

RESULTS AND DISCUSSION

Differences between Unchlorinated Fractions. As reported previously21,23 and evident from Figure 3, SRFA fractions transition gradually from high-O/C and low-H/C (early-eluting) to the opposite (late-eluting). Average oxidation numbers of C (Cos values, Figure 1) mirror this decrease in oxidation from early to late fractions. FRC 1 and 7 and FRC 50 show no overlap in van Krevelen space. In short, early-eluting HPLC fractions of SRFA correspond most closely to the material predicted by Lavonen et al.15 and Gonsior et al.16 to form the most diverse CHOCls, whereas late-eluting SRFA corresponds to the material predicted by Zhang et al.14,19 to be preferentially used up in the chlorination reaction. Surprisingly, the latest-eluting fraction (FRC 65), not previously analyzed, does show some compositional overlap with early eluting fractions (Figures 3−5, best visible in Figure 5). This overlap likely derives from relatively polar molecules that are well retained on the phenyl stationary phase because they contain aromatic moieties spatially separated from polar functional groups. Examples are ibuprofen and fenoprofen (Table S-2), which elute at a similar retention times to FRC 65. Differences in Polarity. LCMS data of reinjected, unchlorinated fractions (Figure S-2A) furthermore illustrate that each fraction elutes as an individual peak. Although early and mideluting fractions contain compounds with identical formulas (as do mid and late-eluting fractions), LCMS peaks for equivalent fractions (FRC 6, 29, and 49) show no chromatographic overlap. Thus, isomers in those fractions are chemically distinct. D


Article


Figure 3. Van Krevelen plots of all formulas assigned for fractions 1−65 (right, post; left, pre). FRC 30* pre is an identical sample to the sample of FRC 30 post that is not chlorinated or ZipTipped and is analyzed three months later on the same instrument but without the ion funnel.

FRC 1, the CHOCls in prechlorination spectra appear in an entirely different region of the van Krevelen plot from any other NOM or DBPs and do not persist after chlorination. All CHOCls detected in FRCs 1 and 30 post are actually newly formed. Differences in Heteroatom Distribution between Unchlorinated and Chlorinated Fractions. N-containing compounds were only detected in early fractions (FRC 1 and 7), consistent with the literature,22,34 and grew in number and relative abundance after chlorination. This trend is confirmed by an increase in N/C after chlorination for all fractions (data not shown). CHOs are likely preferentially removed through the formation of nonamenable DBPs, the desalting step, or the drying process. Additionally, fragmentation or oxidation may make certain CHONs amenable that were previously suppressed in the ESI source because of low gas-phase acidity compared to other NOM components or the need for highcharge states to fall within the effective m/z range of the instrument (e.g., proteins, large heterocycles). Both transitions would lead to an apparent enrichment in N over C. Similar apparent enrichment will be discussed below for CHOs and referred to as “selective enrichment”. Differences in C Oxidation. The average oxidation number of C (Cos) was calculated equivalent to Lavonen et al.15 Figure 1 shows a plot of Cos versus fraction number. What is striking is that inclusion of Cl does not significantly change the oxidation in polar fractions (Figure 1: black versus light gray line). This may indicate that reaction pathways, such as electrophilic substitution, are favored in polar SRFA. For FRC 65, however, CHOCls are significantly more oxidized than the starting material (CHOs pre). This trend is in agreement with reaction pathways such as Cl−OH addition (which increases O/C and DBE/C ratios). Under the conditions employed for synthetic standards, which may not have matched fractions perfectly in pH or dose, most analytes favored substitution regardless of retention time. Molecules that do not incorporate Cl (CHOS post) differ in overall oxidation from the starting material (CHOs pre) according to elution order (Figure 1: dark gray line versus black line). For late-eluting fractions (FRC 50 and 65), average oxidation for CHOs remains essentially unaffected by chlorination. In early fractions, however, the average oxidation

Synthetic standards with retention times equivalent to each fraction (Table S-2) similarly showed that early-eluting material tend to lose parent signal and form a number of products that could not readily be identified with low-resolution LCMS. Lateeluting standards, meanwhile, tend to gain parent signal and form expected CHOCls (replacement of H with Cl or addition of Cl plus OH). Inferences from Formulas. CHOCl Diversity. Differences in diversity of CHOCls formed and their relative contribution to the total spectrum also show clear differences in the extent to which chlorination transformed each fraction. In FRC 7 post, 40% of all assignable formulas contained Cl. All other fractions averaged ∼9% (FRC 30−11%, SRFA − 10%, FRC 50−8%, FRC 65−8%, FRC 1−7%). In general, this percentage is lowest and approximately equal for late fractions, midrange for midfractions and bulk SRFA, and highest for early fractions. FRC 1 is an outlier in this trend, either because its total abundance was so low that a significant portion of its CHOCls was below the detection limit or because carbohydrates make up a significant portion of this fraction. As discussed below, carbohydrates tend to turn into DBPs nonamenable to ESIFTICR-MS. Nonamenable DBPs here include volatiles like THMs and HAAs, aprotics, and molecules that do not readily fall within the 300−1000 m/z range in negative-ion mode. FRC 7 was the only sample in which incorporation of more than 1 Cl was observed (up to 3), although overall signal abundance in FRC 7 postchlorination was the second lowest. The number of CHOCls formed was also greatest in FRC 7 (134-Cl1, 115-Cl2, and 55-Cl3) followed by bulk SRFA (74Cl1), late-eluting fractions (FRC 65:63-Cl1, FRC 50:60-Cl1), mideluting fraction (FRC 30:54-Cl1) and FRC 1 (24-Cl1). This diversity pairs with a near-Gaussian distribution in total abundance (low abundance for FRC 7 and SRFA, medium for FRC 65, high for FRC 50, medium for FRC 30 and low for FRC 1). Thus, high diversity of CHOCls cannot be attributed to better detectability of DBPs at high S/N but appears related to the chemical makeup of each fraction. Pre-existing CHOCls. FRCs 1 and 30 were the only fraction for which CHOCls might have been present prior to chlorination (Table S-4). Because we erred on the side of retaining suspect CHOCls in prechlorination samples, formula misassignments are possible. Meanwhile, as Figure 4 shows for E


Article


Figure 4. Van Krevelen plots by class for fractions 1−65 (right, post; left, pre). FRC 30* pre is an identical sample to the sample of FRC 30 post that is not chlorinated or ZipTipped and is analyzed 3 months later on the same instrument but without the ion funnel. Blue, CHOs. Green, CHOCls. Red, CHONs.

of CHOs decreases with chlorination. This finding is nonintuitive at first because chlorination is an oxidation, yet CHOs appear to get reduced. The results can be explained through selective enrichment. Highly oxidized CHOs disappear from the spectrum as they react with NaOCl. Both the overall loss of signal described above and the formation of CHOCls with high oxidation numbers (Figure 1) support this hypothesis. Meanwhile, less-oxidized molecules remain and become more detectable through reduced competition in the ESI source. The combined removal of the most oxidized material and the improved detectability of less-oxidized material result in a lowering of the overall average Cos. Figure 1 also contains data for FRC 30 pre analyzed both with and without desalting through a ZipTip. The effect of ZipTipping on Cos values is minimal (Figure 1), although ZipTipping does enrich NOM in material of high MW, low number of O, and high DBE (Figures 2 and S-5), as expected for a hydrophobic membrane and consistent with the literature.22,34,35 However, both the solid-phase extraction (SPE) results in the literature (e.g., refs 15, 22, 27, and 35) and overlays of data from the same sample with and without ZipTipping (Figures 2 and S-5) show that C18 SPE does not induce the marked compositional changes observed here between pre and post chlorination samples. For instance, here, an increase in N/C is observed postchlorination and Ziptipping, whereas the literature shows that C18 membranes are selective against N-containing NOM.34,35 Finally, if desalting had induced the changes in oxidation (Figure 1), selective enrichment of specific van Krevelen regions (Figures 3−5), and the appearance of CHOCls, all C18 extracts (FRC 30, 50, 65, and SRFA) should exhibit equivalent transformations. Instead, differences track with elution order: early fractions differ from mid and bulk, which differ from lateeluting fractions. Unfortunately, the effect of the C4 membrane could not be independently investigated here because of severe sample limitation for early-eluting fractions. However, C18 membranes function exactly as expected, suggesting that C4 likely do as well. C4 SPE membranes should have the same selectivity toward hydrophobic material as C18 but to a lesser extent.

Figure 5. Van Krevelen plots by treatment and compound category. Oval: ions with 2O (possibly fatty acids) that were surprisingly detected in both early and late fractions here.

DBP Formation: Van Krevelen Plots. Figures 3−5 illustrate that HPLC fractionation prior to chlorination reveals notable differences between pre- and postchlorination spectra and avails detail as to which components are relatively inert, and thus selectively enriched, and which are highly reactive, and thus substantially converted. The same level of detail is not accessible from bulk measurements of whole water.14−16 Van Krevelen plots of all classes (Figure 3) indicate that FRC 1−30 and bulk SRFA experienced far more significant conversion through chlorination than late fractions. In chlorinated early fractions, FRC 1 and 7 post, ions with low O/C and high H/C ratios (H/C > 1.2, O/C < 0.4) dominate in the region of the van Krevelen space, where proteins and lipids transition into lignin.36 In Figure 4, van Krevelen plots coded by class reveal that these newly visible low O/C ratio ions contain no Cl; therefore, they are likely the origin for the decrease in Cos of CHOs discussed above. Kendrick plots (Figure S-6) show that these lipidlike CHOs are also of low molecular weight. Notably, Liu et al.22 also noted the presence of low-MW fatty acids in their most hydrophilic fractions of NOM. The possible origin of low-MW material in the lipid- or peptidelike regions of van Krevelen space in early F


Article


underscores the selective enrichment of N-containing SRFA in the early fractions discussed above. Superimposed van Krevelen plots for all samples of CHOs pre, CHOs post and CHOCls (Figure 5) re-emphasize that prechlorination (Figure 5A) early fractions (FRC 1 and 7) and FRC 50 show no overlap. FRC 65 pre surprisingly encompasses almost the entire van Krevelen space for SRFA. For material not containing Cl, postchlorination, the most notable transformation is evident in early fractions. Early fractions originally did not contain any ions below O/C ratios of 0.5, but after chlorination, they extend over all four quadrants. CHOs newly appearing in quadrant I postchlorination (e.g., for early fractions) are unlikely to have formed from CHOs observed prechlorination in quadrants II−IV because that would involve chemical reduction (either through loss in O, gain of H, or both). Selective enrichment is the more likely explanation. As observed in Figures 3 and 4, late-eluting fractions do not undergo as notable a transformation, perhaps because their main abundance is in quadrant I, which appears the least transformed during chlorination. For CHOCls, the most notable difference between the starting material (CHOs pre) and the product (CHOCls post) is again in the early fractions. A significant portion of their abundance pre chlorination is in quadrant II (carbohydratelike region of the van Krevelen diagram36), which is entirely unrepresented in CHOCls and much diminished in CHOs postchlorination. H/C and O/C ratios alone do not infallibly identify compound classes, but experiments with synthetic standards (Table S-2) show that simple sugars elute at equivalent retention times to FRC 1. Also, these sugars were similarly “removed” from the spectrum rather than transformed by expected addition or substitution. Furthermore, Navalon et al.37 noted that in some waters, THMs (which are not amenable to ESI-FTICR-MS) derived mainly from oligosaccharides and oligopeptides. Finally, Woods et al.43 also found a positive correlation between the polarity and hydrophilicity (as for early eluting fractions here) and the carbohydrate content of NOM. CHOCls that can be observed in ESI-FTICR-MS stand out in terms of how well they overlap in van Krevelen space for nearly all samples (except FRC 50). CHOCls are similar to each other and similar in Cos to the starting material (CHOs pre). Most unchlorinated material remaining (CHOs post) also exists in the same quadrant as chlorinated DBPs formed in the same fraction. In short, CHOCls amenable to ESI-FTICR-MS appear similar in composition to the humic starting material. Early fractions are the notable exceptions. Interestingly, of the very few CHOCls detected in FRC 50, all were also detected in raw water by Lavonen et al.15 prior to treatment (Figure S-7). Presence in prechlorination samples suggests that these types of CHOCls form readily (e.g., with Cl− or HCCl3 (g) in the ESI source) or accumulate in the environment from chlorinated water returning to the water cycle. Easy formation or persistence of these particular CHOCls may account for their detection even in the least transformed fractions. Finally, all but three of the CHOCls detected in FRC 50 and most in FRC 65 were also detected in chlorinated river water by Zhang et al. (Figure S-7).14 FRC 65 had the second highest overlap with both literature sources. Consistency in formulas assigned between teams using different algorithms goes to strengthening confidence in these assignments. Here, the data provided additional indication that when it comes to

fractions could be fragments from overall polar macromolecules that are of low original concentration or do not compete well for charge in negative-ion mode without the selective removal of the bulk material (e.g., lipoglycans, lipopetides, and glycopeptides). The notable signal in the carbohydratelike region of the van Krevelen space prior to chlorination (best visible in Figure 5) may point to macromolecules including carbohydrate moieties that break up in the ESI source. Consistent with FRC 30 making up a large portion (by mass and UV signal) of bulk SRFA, SRFA and FRC 30 show similar transitions. Most noticeably, ions with low O/C and H/C ratios (H/C ∼0.4−∼0.9, O/C ∼0.1−∼0.4) dominate chlorinated FRC 30 and SRFA in the region of the van Krevelen space, where condensed aromatic structures (CAS) transition into lignin.36 The possible origin of CAS-like signal in postchlorination spectra of mid-fractions and bulk SRFA (FRC 30 and SRFA) could be the SPE extraction (Figure 2), but the low-abundance signal in the CAS-like region36 of the van Krevelen plot encircled in Figure 2 shares no overlap with ions in that region postchlorination. Prechlorination, the signal in the CAS-like region (selectively enhanced through C18 SPE) consists of 11 N-containing ions with extremely high DBEs (21−24). Ions that are selectively enriched in the CAS-like region through chlorination do not contain N, are far more numerous (121), and have far lower DBEs (12−16). Thus, the more likely origin is selective enrichment of pre-existing CAS-like SRFA. CAS-like structures do appear to chlorinate; 34 of the 121 ions in this region contain Cl (ClO3 − ClO6; DBE: 12−16) but remain ESI-FTICR-MS-amenable in the transformation, far more so than the other components (degraded tannins and lignins), which appear to be selectively removed. The selective enrichment theory is strengthened by close correlation between new regions of the van Krevelen plot appearing postchlorination in early and mideluting fractions and reduction in overall signal (Figure S-4) for those fractions post chlorination. As noted above, late-eluting fractions, which are centered at lower O/C values, show little transformation in Figure 3 other than a general broadening of the most abundant signal into slightly higher O/C and DBE space of the van Krevelen plot. Broadening signal is consistent with the increase in total signal abundance and S/Nchem discussed above. Differences between FRC 30 and SRFA and FRC 50 and 65 postchlorination cannot be attributed to desalting (all these fractions were desalted on a C18 ZipTip). Overall, Figure 3 suggests that early and mideluting SRFA primarily form DBPs that are no longer humiclike or no longer amenable to ESI-FTICR-MS, whereas late-eluting SRFA retain much of their humiclike characteristics. When van Krevelen is plotted by class (Figure 4), it is possible to distinguish between Cl-containing ions (CHOCls) and CHOs. It is notable that CHOCls in early fractions, generally group at lower H/C ratios (higher DBEs) than CHOs, in agreement with carbon oxidation trends discussed above (higher DBE infers higher Cos). Inclusion of Cl into molecules of low H/C (or high DBE), likely highly aromatic molecules, is in agreement with the established correlation between SUVA values and TTHM formation potentials in the literature,17 as well as SUVA values and total Cl consumptions here (see below). It is also in agreement with findings of both Zhang et al.14 and Lavonen et al.15 Finally, Figure 4 also G


Article


chromatograms indicate that sample absorbs much more strongly at 254 nm than at 340 nm (Figure S-10). High SUVA values for mid-eluting fractions are consistent with results by Woods et al.,38 who determined that NOM of midpolarity exhibits the highest aromatic content. Thus, it appears that early fractions, which so clearly transformed more noticeably in ESI-FTICR-MS, are not more reactive, per se. This may partially be the result of concentrations being measured as TOC. Low-MW samples contain more molecules at the same TOC concentration than high-MW samples and thus receive lower doses of Cl per molecule. Also, high Cl/ TOC ratio doses affect pH (Table S-3), which was close to neutral after 3 days for all fractions (except FRC 31) here but acidic above. Reaction pH, and whether Cl is in excess throughout the reaction period, also affect which reaction pathways predominate.20 Because of the complexity and variability of NOM concentrations in drinking water (e.g., references 5, 11, 15, and 39), it is impossible for one study to provide universal insights. At best, each experiment leads to further testable hypotheses. Future work involves the independent variation of Cl dose and pH and the analysis of additional types of NOM (humic acid and whole NOM). Another important aspect of future work is the need to scale up to reduce uncertainty in mass and TOC measurements and to make more parallel studies feasible. Finally, the correlation between DBPs formed by different components, toxicity, and ease of removal of different DBPs from drinking water will be examined.

composition and formulas, differentiation between humic samples may not lie in the humiclike CHOCls formed but rather in the material selectively concentrated (CHOs post). This observation belies the probable differences in structure, functionality, and toxicity between CHOCls formed, which cannot be evaluated with broadband MS alone. DBP Formation: DBE versus C Plots. Correlation between DBE and number of C (Figure S-8) confirms that late eluting fractions show less differentiation postchlorination. CHOCls appear in a similar area of the DBE versus C space for nearly all samples, confirming relative unity in the CHOCls formed as noted above. The most transformed fractions, FRC 1 and 7, appear enriched in CHOs with the same or lower DBE values compared to the starting material but form CHOCls with higher DBE values (in agreement with Figure 5B versus C). The complementary nature of these changes in CHOs versus CHOCls is expected on the basis of selective enrichment; as Cl is incorporated into NOM of high DBE, the corresponding CHOs disappear, giving the semblance of CHOs with low DBE increasing. Class Distribution. Consistent with previous findings, Figure S-9B illustrates that early and late fractions produce CHOCls centered at similar oxygen classes. Figure S-9A furthermore confirms that early- and mideluting fractions (especially FRC 7 and 30) are transformed more significantly. The relative abundance of CHOCls is greatest in these spectra, whereas the abundance of CHOs remaining is lowest. Inferences from FTMS Experiments. Because all fractions analyzed above used up all available Cl and the ESI source is selective, no fraction can be deemed most reactive. However, NOM components react quite differently. (A) The leastoxidized material, (Figure 5, quadrant I) appears most refractory and retains the most CHO signal, especially ligninlike36 components. (B) FTMS-amenable CHOCls tend not to favor quadrant II (a high-O and high-H region), especially the carbohydratelike space of that quadrant. (C) Ions in quadrant II (highly oxygenated but largely aliphatic material like carbohydrates) appear to be selectively removed, similar to carbohydrate synthetic standards. (D) Early fractions become enriched in quadrant I, the inert, low-O and high-H region, especially in the lipid- and proteinlike space of that quadrant. (E) Bulk and middle fractions become enriched in quadrant III (low-O and low-H region), especially in the CAS-like space of that quadrant. (F) SRFA of high DBE are selectively transformed to CHOCls in agreement with the literature.14,15,17 (G) Early fractions, which are highly polar and highly oxidized, undergo the most drastic transformations, lose the most signal, and form most diverse CHOCls, consistent with Lavonen et al.15 (H) CHOs selectively enriched after chlorination differ more notably from the starting material than CHOCls do. In other words, CHOCls formed that are amenable to ESIFTICR-MS largely retain their humic-like character. SUVA, Residual Chlorine, and ΔpH. To obtain quantitative data, we measured residual free Cl for samples dosed at a higher Cl to NOM ratio. Consistent with expectation, 17 Figure S-10 illustrates that the chlorine consumed, and the ΔpH associated with chlorination, track with SUVA values. Somewhat surprisingly, all fractions have similar SUVA values and reactivities except for a narrow band in the mid-elution range that exhibits extraordinarily high SUVA values (FRC 31) followed by extraordinarily low ones (FRC 36). Accordingly, where SUVA values peak, the original

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.5b03466. Additional details on the experimental procedures, challenges, and limitations. Figures showing the relationship between Chlorine consumed and NOM concentration, LCMS results of SRFA Fractions, average oxidation number of C, total signal abundance, effect of ZipTipping, Kendrick plots, overlap in CHOCl formulas with literature, DBE versus number of carbon relationships in SRFA subfractions, oxygen class abundance, and quantitative measurements. (PDF) Tables showing summaries of select FTMS results and settings and LCMS results on chlorinated synthetic standards; retention time, pH and SUVA fraction data; and a list of formulas. (XLSX)

■

AUTHOR INFORMATION

Corresponding Author

*Phone: (251) 460-7432; fax: (251) 460-7359; e-mail: [email protected]. Present Address ∥

Division of Environmental and Ecological Engineering and Lyles School of Civil Engineering, Purdue University, West Lafayette, IA 47907, United States. Author Contributions

The manuscript was written through contributions by all authors. All authors have given approval to the final version of the manuscript. All undergraduate authors contributed equally and substantially. H


Article


(13) Iriarte-Velasco, U.; Á lvarez-Uriarte, J. I.; González-Velasco, J. R. Removal and structural changes in natural organic matter in a Spanish water treatment plant using nascent chlorine. Sep. Purif. Technol. 2007, 57 (1), 152−160. (14) Zhang, H.; Zhang, Y.; Shi, Q.; Hu, J.; Chu, M.; Yu, J.; Yang, M. Study on Transformation of Natural Organic Matter in Source Water during Chlorination and Its Chlorinated Products using Ultrahigh Resolution Mass Spectrometry. Environ. Sci. Technol. 2012, 46 (8), 4396−4402. (15) Lavonen, E. E.; Gonsior, M.; Tranvik, L. J.; Schmitt-Kopplin, P.; Köhler, S. J. Selective Chlorination of Natural Organic Matter: Identification of Previously Unknown Disinfection Byproducts. Environ. Sci. Technol. 2013, 47 (5), 2264−2271. (16) Gonsior, M.; Schmitt-Kopplin, P.; Stavklint, H.; Richardson, S. D.; Hertkorn, N.; Bastviken, D. Changes in dissolved organic matter during the treatment processes of a drinking water plant in sweden and formation of previously unknown disinfection byproducts. Environ. Sci. Technol. 2014, 48 (21), 12714−12722. (17) Matilainen, A.; Gjessing, E. T.; Lahtinen, T.; Hed, L.; Bhatnagar, A.; Sillanpäa,̈ M. An overview of the methods used in the characterisation of natural organic matter (NOM) in relation to drinking water treatment. Chemosphere 2011, 83 (11), 1431−1442. (18) Uyguner, C. S.; Hellriegel, C.; Otto, W.; Larive, C. K. Characterization of humic substances: implications for trihalomethane formation. Anal. Bioanal. Chem. 2004, 378 (6), 1579−1586. (19) Zhang, H.; Zhang, Y.; Shi, Q.; Ren, S.; Yu, J.; Ji, F.; Luo, W.; Yang, M. Characterization of low molecular weight dissolved natural organic matter along the treatment trait of a waterworks using Fourier transform ion cyclotron resonance mass spectrometry. Water Res. 2012, 46 (16), 5197−5204. (20) Rook, J. J. Chlorination reactions of fulvic acids in natural waters. Environ. Sci. Technol. 1977, 11 (5), 478−482. (21) Stenson, A. C. Reversed-Phase Chromatography Fractionation Tailored to Mass Spectral Characterization of Humic Substances. Environ. Sci. Technol. 2008, 42 (6), 2060−2065. (22) Liu, Z.; Sleighter, R. L.; Zhong, J.; Hatcher, P. G. The chemical changes of DOM from black waters to coastal marine waters by HPLC combined with ultrahigh resolution mass spectrometry. Estuarine, Coastal Shelf Sci. 2011, 92 (2), 205−216. (23) Capley, E. N.; Tipton, J. D.; Marshall, A. G.; Stenson, A. C. Chromatographic Reduction of Isobaric and Isomeric Complexity of Fulvic Acids To Enable Multistage Tandem Mass Spectral Characterization. Anal. Chem. 2010, 82 (19), 8194−8202. (24) Koch, B. P.; Ludwichowski, K.-U.; Kattner, G.; Dittmar, T.; Witt, M. Advanced characterization of marine dissolved organic matter by combining reversed-phase liquid chromatography and FT-ICR-MS. Mar. Chem. 2008, 111 (3−4), 233−241. (25) Gaspar, A.; Harir, M.; Hertkorn, N.; Schmitt-Kopplin, P. Preparative free-flow electrophoretic offline ESI-Fourier transform ion cyclotron resonance/MS analysis of Suwannee River fulvic acid. Electrophoresis 2010, 31 (12), 2070−2079. (26) Novotny, N. R.; Capley, E. N.; Stenson, A. C. Fact or artifact: the representativeness of ESI-MS for complex natural organic mixtures. J. Mass Spectrom. 2014, 49 (4), 316−326. (27) Kido Soule, M. C.; Longnecker, K.; Giovannoni, S. J.; Kujawinski, E. B. Impact of instrument and experiment parameters on reproducibility of ultrahigh resolution ESI FT-ICR mass spectra of natural organic matter. Org. Geochem. 2010, 41 (8), 725−733. (28) Kaiser, N. K.; Quinn, J. P.; Blakney, G. T.; Hendrickson, C. L.; Marshall, A. G. A novel 9.4 T FTICR mass spectrometer with improved sensitivity, mass resolution, and mass range. J. Am. Soc. Mass Spectrom. 2011, 22 (8), 1343−1351. (29) Kostiainen, R.; Kauppila, T. J. Effect of eluent on the ionization process in liquid chromatography−mass spectrometry. J. Chromatogr. A 2009, 1216 (4), 685−699. (30) Flerus, R.; Koch, B. P.; Schmitt-Kopplin, P.; Witt, M.; Kattner, G. Molecular level investigation of reactions between dissolved organic matter and extraction solvents using FT-ICR MS. Mar. Chem. 2011, 124 (1−4), 100−107.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors acknowledge Drs. Alan Marshall and Amy McKenna for institutional support and access to the 9.4T FT-ICR mass spectrometer and Mrs. Laura Linn for TOC analysis and MAWWS. We thank Divya Goel and Charlie Weeks (students at St. Paul’s Episcopal School, Mobile, AL, at the time of preparation of this manuscript) for laboratory assistance and manuscript editing. Financial support from NSF DMR 1157490, NSF CHE 1039944, the Center for Environmental Resiliency (University of South Alabama), the State of Florida, and the University of South Alabama is also gratefully acknowledged. The authors thank Dr. David Stranz and Sierra Analytics for tailoring features in the Composer software to characterize CHOCl and perform threshold-based background subtraction.

■

REFERENCES

(1) Hrudey, S. E. Chlorination disinfection by-products, public health risk tradeoffs and me. Water Res. 2009, 43 (8), 2057−2092. (2) Villanueva, C. M.; Cantor, K. P.; Grimalt, J. O.; Malats, N.; Silverman, D.; Tardon, A.; Garcia-Closas, R.; Serra, C.; Carrato, A.; Castaño-Vinyals, G.; et al. Bladder Cancer and Exposure to Water Disinfection By-Products through Ingestion, Bathing, Showering, and Swimming in Pools. Am. J. Epidemiol. 2007, 165 (2), 148−156. (3) Michaud, D. S.; Kogevinas, M.; Cantor, K. P.; Villanueva, C. M.; Garcia-Closas, M.; Rothman, N.; Malats, N.; Real, F. X.; Serra, C.; Garcia-Closas, R.; et al. Total fluid and water consumption and the joint effect of exposure to disinfection by-products on risk of bladder cancer. Environ. Health Perspect. 2007, 115 (11), 1569−1572. (4) Cantor, K. P.; Villanueva, C. M.; Silverman, D. T.; Figueroa, J. D.; Real, F. X.; Garcia-Closas, M.; Malats, N.; Chanock, S.; Yeager, M.; Tardon, A.; et al. Polymorphisms in GSTT1, GSTZ1, and CYP2E1, disinfection by-products, and risk of bladder cancer in Spain. Environ. Health Perspect. 2010, 118 (11), 1545−1550. (5) Neale, P. A.; Antony, A.; Bartkow, M. E.; Farré, M. J.; Heitz, A.; Kristiana, I.; Tang, J. Y. M.; Escher, B. I. Bioanalytical Assessment of the Formation of Disinfection Byproducts in a Drinking Water Treatment Plant. Environ. Sci. Technol. 2012, 46 (18), 10317−10325. (6) Eikebrokk, B.; V, R. D. NOM increase in Northern European source waters: Discussion of possible causes and impacts on coagulation/contact filtration processes. Water Sci. Technol. Water Supply 2004, 4 (4), 47−54. (7) Driscoll, C. T.; Driscoll, K. M.; Roy, K. M.; Mitchell, M. J. Chemical Response of Lakes in the Adirondack Region of New York to Declines in Acidic Deposition. Environ. Sci. Technol. 2003, 37 (10), 2036−2042. (8) USEPA. Rule Fact Sheet: Stage 2 Disinfectants and Disinfection Byproducts Rule. http://water.epa.gov/lawsregs/rulesregs/sdwa/ stage2/regs_factsheet.cfm (accessed Sep 15, 2014). (9) Bull, R. J.; Reckhow, D. A.; Li, X.; Humpage, A. R.; Joll, C.; Hrudey, S. E. Potential carcinogenic hazards of non-regulated disinfection by-products: Haloquinones, halo-cyclopentene and cyclohexene derivatives, N-halamines, halonitriles, and heterocyclic amines. Toxicology 2011, 286, 1−19. (10) Matilainen, A.; Vepsäläinen, M.; Sillanpäa,̈ M. Natural organic matter removal by coagulation during drinking water treatment: A review. Adv. Colloid Interface Sci. 2010, 159 (2), 189−197. (11) Sharp, E. L.; Jarvis, P.; Parsons, S. A.; Jefferson, B. Impact of fractional character on the coagulation of NOM. Colloids Surf., A 2006, 286 (1−3), 104−111. (12) Zhang, X.; Minear, R. A.; Barrett, S. E. Characterization of High Molecular Weight Disinfection Byproducts from Chlorination of Humic Substances with/without Coagulation Pretreatment Using UF−SEC−ESI-MS/MS. Environ. Sci. Technol. 2005, 39 (4), 963−972. I


Article

Environmental Science & Technology (31) Mopper, K.; Stubbins, A.; Ritchie, J. D.; Bialk, H. M.; Hatcher, P. G. Advanced Instrumental Approaches for Characterization of Marine Dissolved Organic Matter: Extraction Techniques, Mass Spectrometry, and Nuclear Magnetic Resonance Spectroscopy. Chem. Rev. 2007, 107 (2), 419−442. (32) Marshall, A. G.; Hendrickson, C. L. Fourier transform ion cyclotron resonance detection: principles and experimental configurations. Int. J. Mass Spectrom. 2002, 215 (1−3), 59−75. (33) Marshall, A. G.; Hendrickson, C. L.; Jackson, G. S. Fourier transform ion cyclotron resonance mass spectrometry: a primer. Mass Spectrom. Rev. 1998, 17 (1), 1−35. (34) Reemtsma, T.; These, A.; Linscheid, M.; Leenheer, J.; Spitzy, A. Molecular and Structural Characterization of Dissolved Organic Matter from the Deep Ocean by FTICR-MS, Including Hydrophilic Nitrogenous Organic Molecules. Environ. Sci. Technol. 2008, 42 (5), 1430−1437. (35) Sleighter, R. L.; Hatcher, P. G. Molecular characterization of dissolved organic matter (DOM) along a river to ocean transect of the lower Chesapeake Bay by ultrahigh resolution electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry. Mar. Chem. 2008, 110 (3−4), 140−152. (36) Hockaday, W. C.; Purcell, J. M.; Marshall, A. G.; Hatcher, P. G.; Baldock, J. A. Electrospray and photoionization mass spectrometry for the characterization of organic matter in natural waters: a qualitative assessment. Limnol. Oceanogr.: Methods 2009, 7, 81−95. (37) Navalon, S.; Alvaro, M.; Alcaina, I.; Garcia, H. Multi-method characterization of DOM from the Turia river (Spain). Appl. Geochem. 2010, 25 (11), 1632−1643. (38) Woods, G. C.; Simpson, M. J.; Koerner, P. J.; Napoli, A.; Simpson, A. J. HILIC-NMR: Toward the Identification of Individual Molecular Components in Dissolved Organic Matter. Environ. Sci. Technol. 2011, 45 (9), 3880−3886. (39) Tsai, M.-L.; Yen, C.-C.; Lu, F.-J.; Ting, H.-C.; Chang, H.-R. Environmentally relevant concentration of arsenic trioxide and humic acid promoted tumor progression of human cervical cancer cells: In vivo and in vitro studies. Environ. Toxicol. 2015, DOI: 10.1002/ tox.22121.

J


Characterization of Disinfection By-Products from ... - ACS Publications

Recommend Documents