Developments in Molecular Level Characterization of Naphthenic Acid Fraction Compounds Degradation in a Constructed Wetland Treatment System

: The reclamation of oil sands process-a ﬀ ected water (OSPW) is a matter of environmental importance because of the aquatic toxicity to biota. This study describes reﬁnements in advanced analytical methods to assess the performance of biological treatment systems for OSPW, such as constructed wetland treatment systems (CWTSs). Assessment of treatment e ﬃ ciency by measurement of the degradation of naphthenic acid fraction compounds (NAFCs) in OSPW is challenging in CWTS due to potentially interfering constituents such as humic acids, organic acids, salts, and hydrocarbons. Here we have applied a previous weak anion exchange (WAX) solid-phase extraction (SPE) method and high-resolution Orbitrap-mass spectrometry (MS) to remove major interferences from the NAFC analysis. The reﬁnements in data processing employing principal component analysis (PCA) indicates that the relative abundance of NAFCs decreased with time in the treated OSPW relative to the untreated OSPW. The most saturated NAFCs with higher carbon numbers were relatively more degraded as compared to unsaturated NAFCs. The use of Kendrick plots and van Krevelen plots for assessment of the performance of the CWTS is shown to be well-suited to detailed monitoring of the complex composition of NAFCs as a function of degradation. The developments and application of analytical methods such as the WAX SPE method and high-resolution Orbitrap-MS are demonstrated as tools enabling the advancement of CWTS design and optimization, enabling passive or semi-passive water treatment systems to be a viable opportunity for OSPW treatment.


Introduction
The mining of oil sands in northeastern Alberta, Canada, generates large volumes of petroleum-contaminated water known as oil sands process-affected water (OSPW). The OSPW is stored numbers (−2 to −4) followed first-order kinetics with half-lives between 19 and 28 weeks. Nevertheless, the more complex NAs (−6 to −12) with more carbon numbers (17)(18)(19)(20) were more recalcitrant with half-lives between 37 and 52 weeks. Pilot-scale CWTS treatment of OSPW was found to decrease the concentration of NAs, hydrocarbons, and metals, as well as toxicity to Ceriodaphnia dubia [20].
A fundamental concern in the analysis of organic compounds in petroleum-derived waters such as OSPW is the complicated nature of the matrix, which requires extraction/fractionation and chemical clean-up [26]. To address this analytical challenge and improve the evaluation of the performance of remediation techniques, there is a need to improve both the extraction methods and high-resolution mass spectrometry analysis of NAFCs [16]. There are several recent breakthroughs for monitoring the fate of NAFCs in water treatment based on methods employing Fourier transform mass spectrometry [27]. The implementation of analytical methods using such instruments have vastly enhanced the determination of composition and abundance of NAFCs in complex samples according to carbon number and DBE. The changes in the distribution of DBE and carbon numbers are used to monitor the efficiency, progress of remediation, water treatment process designs, and development of environmental discharge guidelines.
The most important step in the analysis of COPCs prior to instrumental analysis is the isolation, clean-up, and concentration processes of complex environmental samples. Sample extraction offers the possibility of resolving OSPW composition by eliminating potentially confounding sample matrices (e.g., salinity, suspended solids) to enable the characterization and quantification of the organic compounds [16,26,28]. The sample preparation can facilitate the evaluation of the fate of NAFCs in CWTSs, where measurement of the biodegradation and toxicity reduction are crucial for optimization and scale up [29]. The use of the solid phase extraction (SPE) methods continues to grow as the selectivity of sorbent materials and availability for specific applications, such as OSPW extraction and concentration, improve [30,31]. Several SPE sorbents materials are applicable to assessment of organic compounds in OSPW including ENV+, Oasis Hydrophilic-Lipophilic-Balanced (HLB), C18 and ion exchange [29,[32][33][34]. For the current research, weak anion exchange (WAX) SPE proved to be ideal for extraction of NAFCs to evaluate CWTS-facilitated treatment processes. In an earlier study, we reported the selectivity of WAX SPE compared with previously used ENV+ for NA extraction from OSPW in the presence of wetland plants [35]. The use of the WAX SPE method for analysis of organic acids in OSPW was essential to facilitate their detection and fate in constructed wetland. This study provides further evidence that the WAX SPE method combined with high-resolution Orbitrap-MS is viable and robust for measurement of the biodegradation of OSPW-NAs in CWTSs. Additionally, in our previous study, Ajaero et al. [36], we discussed and explored the composition and transformation of NAFCs in a CWTS according to carbon number and DBE and the electrospray ionization (ESI). Here, the NAFCs distributions in CWTSs are revisited [36]. We present new visualizations of the results including PCA, Kendrick plots and van Krevelen plots for the first time for improved understanding of the performance of the treatment wetland for OSPW. Emphasis is on refinement of analytical methods for data processing to provide improved insights on wetland performance. A rigorous assessment of the optimization of CWTs based on differences in wetland design; hydraulic retention time; horizontal flow versus vertical flow; wetland plant type; and positive controls to evaluate the kinetics of degradation; are not presented here. The latter are topics of another communication with focus on wetland performance [37].

Oil Sand Process-Affected Waters
The OSPW used in this study was obtained in July 2015 from Muskeg River Mine external tailings facility (MRM-ETF). Water was collected 1-2 m below the water surface of a tailings pond. The OSPW was transported in a 1000 L plastic tote to the Contango Strategies test facility in Saskatoon, SK, Canada where it was stored for about 6 months for the constructed wetland experiments. Assessment of Environments 2020, 7, 89 4 of 16 the influent water chemistry showed that it was stable under the storage conditions and within the lower ranges of OSPW COPCs measured at other locations in the Athabasca oil sands area as reported in [20,38].

Constructed Wetland Treatment Systems
The study was carried out in the pilot-testing facilities in Saskatoon, Canada, where the pilot-scale CWTS was set up to assess the degradation of OSPW-NAFCs. The CWTS consists of four plastic Rubbermaid ® bins in a series, each with a diameter of 42.5 cm filled with sand (porosity 0.23) up to 30 cm with an overlying water depth of 20 cm. The bins were planted with Carex aquatilis (Sedges) and aquatic bryophytes (mosses) and operated as a horizontal surface flow system. The CWTS was operated as a closed loop system for 27 days and water from the outflow was redirected back into the first cell of the series, to promote detectable changes in the degradation of NAFCs and to conserve the limited volumes of OSPW available for the experiments. This specific CWTS was not operated as a stand-alone system but was part of an integrated system of other CWTS and controls; full details of which are given elsewhere [37]. The flow rates were regulated by FMI ® QG 400 pumps (Fluid metering Inc., Syosset, NY, USA) at 20 mL/min ± 3 mL/min. Samples from the outflow were collected in 500-mL plastic containers from the last cell of the wetland system at multiple time points and stored in a refrigerator at 4 • C before extraction and analysis within 24 h. The temperature at the greenhouse was maintained at 22 • C during daytime with a 4 • C night setback for a 12 h daylight time. All CWTSs are subject to evapotranspiration (ET), which raises local concentrations of the constituents and HRT, which in turn may lead to a perception of inadequate treatment efficiency due to measurement of concentrations rather than loads [39]. The measured concentrations of NAFCs were determined as reported in [36], and also adjusted for ET to assess load removed (and remaining) to enable the evaluation of treatment performance.

Sample Preparation and Extraction
A detailed procedure of the SPE sample preparation using WAX sorbent is described elsewhere [35]. Briefly, WAX SPE cartridges (200 mg, 33 µm polymeric; Phenomenex (Torrance, CA, USA) were conditioned with 5 mL of methanol followed by 5 mL of Milli-QH 2 O. Then 10 mL aqueous sample from the treatment wetlands (pH~7.0) was loaded onto the cartridges. The NAFCs adsorbed to the cartridges were eluted using 5 mL methanol + 5% ammonium hydroxide. The samples were dried under a steady stream of nitrogen and reconstituted in 1 mL 50/50 acetonitrile/Milli-QH 2 O + 0.1% ammonium hydroxide for high-resolution Orbitrap-MS analysis.

Analysis of NAFCs
The analysis of NAFCs by high-resolution negative-ion electrospray ionization (ESI) Orbitrap-MS was performed at Environment and Climate Change Canada (Saskatoon, SK, Canada). A Surveyor MS pump (Thermo Fisher Scientific Inc., San Jose, CA, USA) was used for loop injection of 5 µL of OSPW extract into the mass spectrometer with 50:50 acetonitrile/water containing 0.1% ammonium hydroxide as the eluent at 200 µL/min. A dual pressure linear ion trap-Orbitrap-MS (LTQ Orbitrap Elite Thermo Fisher Scientific) was used for mass spectrometric analysis. Mass spectra data were collected in the range of 100-600 m/z in full scan mode with a resolution of 240,000 at m/z 400. Mass calibration was performed using n-butyl benzenesulfonamide (212.07507 m/z scan-to-scan mass calibration correction). The ionization conditions of the ESI source are described elsewhere [36,40]. Data acquisition and quantitative analysis were controlled using Xcalibur version 2.2 software (Thermo Fisher Scientific). Data processing was done with Composer version 1.5.3 (Sierra Analytics, Inc. Modesto, CA, USA) by assignment of chemical formulas according to heteroatom class and DBE of classical NAs and NAFCs. Molecular assignments were based on mass accuracies of <2 ppm. The relative abundance of individual compound classes was obtained by dividing a given peak by the sum of the intensities of all peaks observed by the negative-ion ESI Orbitrap. The estimation of the concentration of NAFCs was performed with a five-point external standard calibration of NAFCs at known concentrations, as reported in past studies [8] within the working range of the calibration curve (10-100 mg/L). The determination of NAFC concentration in this investigation, as well as in other studies, is based on semi-quantitative measurements. The semi-quantification is because of the lack of authentic standards for unambiguous quantification and non-existence of standard procedure for isolation and determination of individual NAFCs [36,40]. Accuracy and precision were measured for replicates (n = 3) of a Sigma-Aldrich (Oakville, ON, Canada) commercial NA. Additionally, as a quality assurance step, replicate analysis (n = 3) of a laboratory blank (Milli-Q water) was conducted following a similar procedure for the commercial NAs.

PCA, Kendrick Plot, and van Krevelen Plot Procedure
Principal component analysis (PCA) was conducted by exporting spectral data from Xcalibur to Composer version 1.5.6 software (Sierra Analytics, Modesto, CA, USA), to R 3.6.2 (R Core Team, 2020) where data was transformed, analyzed, and plotted. Exact-mass data from XCalibur was assigned to molecular formulae by Composer, and formula data was exported to R. Lists of identifiable molecular formulae, which were collected from each sample, and compiled into a master list of formulae present across all samples (339 common formulae) in R, along with molecular characteristics including heteroatom counts, ratios, Kendrick masses and Kendrick mass defects. Relative abundances of formulae in each wetland were assigned to formulae found on this master list by comparing the counts of each individual formula to the most abundant peak detected in each respective sample. Principal component analysis (PCA) was conducted by comparing differences in relative abundance of compounds (as defined above) and plotted using the "ggplot 2" [41] and "ggbiplot" [42] packages in R 3.6.2 [43]. Kendrick and van Krevelen plots were prepared to show the differences between day 0 (i.e., raw OSPW) and the final sample taken (day 27) from the wetland by comparing the relative abundance of all common molecular formulae. Both Kendrick and van Krevelen plots were plotted using the "ggplot 2" [41] package in R 3.6.2 [43].

Principal Component Analysis
The data were initially evaluated by principal component analysis (PCA) ( Figure 1). Nearly 83% of the dataset's variability is captured along both principal components (PC1 = 73.4% explained variability, PC2 = 9.4% explained variability). This analysis focused on separation across PC1 because it accounts for approximately 8x greater variability than PC2. A few factors may introduce sample data separation across PC2, such as slight temporal differences in the microbiological communities in the treatment setup [44], which was not evaluated in this experiment. In addition, the apparent sample variability across PC2 could be due to a slight arch effect, a well-known consequence of data ordination using PCA [45]. The most prominent differences between samples are also apparent across PC1; therefore, this analysis focuses on separation between samples along PC1.
Samples separate steadily from the far left of PC1 to the far right over time, suggesting consistent and gradual changes lead to differences in NAFC profiles. Variability between samples reaches a maximum between the untreated OSPW and samples on day 27; these samples are therefore subsequently compared to highlight major compositional differences in NAFC molecular characteristics observed across the experiment. Additionally, the distribution along PC1 increases in distance over time, which may indicate further dynamics of the biotransformation where at least two mechanisms for degradation occur. The first one from day zero to day 21; and the second after day 21. To evaluate how formula characteristics, contribute to sample separation along PC1, Kendrick [46] and van Krevelen plots of all formulae detected across the experiment were prepared by colourcoding formulae according to their factor loadings along PC1. The Kendrick plot ( Figure 2) highlights that the highest m/z saturated components (i.e., m/z range 200-275, Kendrick mass defect 0.05-0.1) are the most strongly associated with a low PC1 sample score, and that more unsaturated (i.e., higher Kendrick mass defect) components are associated with a high PC1 sample score. This suggests that on Day 27, NAFCs in the treatment wetland had a higher proportion of highly-unsaturated components than did the Raw OSPW. The van Krevelen of the formulae (Figure 3) offers complementary information; only high H/C ratio components (i.e., saturated) are associated with a low PC1 score. In addition, there are a few low-O/C components (approximately 1.2-1.5 H/C ratio and 0.1 O/C ratio) that are associated with a low PC1 score. These two plots together predict that most sample variability is likely attributable to differences in relative abundance of saturated components; higher relative abundance of saturated components is expected from the Raw OSPW, whereas more unsaturated components persist to Day 27.  To evaluate how formula characteristics, contribute to sample separation along PC1, Kendrick [46] and van Krevelen plots of all formulae detected across the experiment were prepared by colour-coding formulae according to their factor loadings along PC1. The Kendrick plot ( Figure 2) highlights that the highest m/z saturated components (i.e., m/z range 200-275, Kendrick mass defect 0.05-0.1) are the most strongly associated with a low PC1 sample score, and that more unsaturated (i.e., higher Kendrick mass defect) components are associated with a high PC1 sample score. This suggests that on Day 27, NAFCs in the treatment wetland had a higher proportion of highly-unsaturated components than did the Raw OSPW. The van Krevelen of the formulae (Figure 3) offers complementary information; only high H/C ratio components (i.e., saturated) are associated with a low PC1 score. In addition, there are a few low-O/C components (approximately 1.2-1.5 H/C ratio and 0.1 O/C ratio) that are associated with a low PC1 score. These two plots together predict that most sample variability is likely attributable to differences in relative abundance of saturated components; higher relative abundance of saturated components is expected from the Raw OSPW, whereas more unsaturated components persist to Day 27. To evaluate how formula characteristics, contribute to sample separation along PC1, Kendrick [46] and van Krevelen plots of all formulae detected across the experiment were prepared by colourcoding formulae according to their factor loadings along PC1. The Kendrick plot ( Figure 2) highlights that the highest m/z saturated components (i.e., m/z range 200-275, Kendrick mass defect 0.05-0.1) are the most strongly associated with a low PC1 sample score, and that more unsaturated (i.e., higher Kendrick mass defect) components are associated with a high PC1 sample score. This suggests that on Day 27, NAFCs in the treatment wetland had a higher proportion of highly-unsaturated components than did the Raw OSPW. The van Krevelen of the formulae (Figure 3) offers complementary information; only high H/C ratio components (i.e., saturated) are associated with a low PC1 score. In addition, there are a few low-O/C components (approximately 1.2-1.5 H/C ratio and 0.1 O/C ratio) that are associated with a low PC1 score. These two plots together predict that most sample variability is likely attributable to differences in relative abundance of saturated components; higher relative abundance of saturated components is expected from the Raw OSPW, whereas more unsaturated components persist to Day 27.

Interpretation of Kendrick and van Krevelen Plots
To visualize the predictions discernible from the PC1-coded Kendrick plot, sample data were plotted in a Kendrick plot [46] showing the difference in relative abundance of various NAFCs across the full spectrum ( Figure 4). As illustrated in Figure 4, the most substantial decreases in relative abundance were observed from compounds with an approximate Kendrick Mass of 220-280 with Kendrick mass defects (KMD) ≤ 0.1. There were also some modest decreases in the relative abundance of compounds with approximate Kendrick Masses of 240-325 with an approximate KMD of 0.165, and a smaller series with approximate Kendrick Masses of 280-315 with an approximate KMD of 0.175. On the other hand, most compounds from Kendrick Mass 200-300 and KMDs from 0.1-0.15 simultaneously increased in relative abundance. In addition, there are some formulae across the Kendrick plot that were not present in the untreated OSPW that increased slightly in relative abundance. These trends suggest that the most saturated compounds (i.e., low mass defects) from m/z 220-300 are the most susceptible to changes and/or dissipation in this system. Unsaturated compounds across a similar mass range therefore appear to be resistant to degradation and/or dissipation in the CWTS, leading to an increase in relative proportion of those unsaturated compounds.
Complementary observations were made by plotting data in a van Krevelen plot showing the net-change in the relative abundance of major Ox (i.e. O2, O3, O4, O5, O6, and O7) formulae to simultaneously examine the change in oxygen content and unsaturation of NAFCs across the sampling period ( Figure 5). As shown in Figure 5, some O2 compounds with relatively high H/C ratios (i.e., H/C ratio > 1.6) decrease the most in relative abundance. There are also some compounds from H/C 1.2-1.5 and low O/C ratio (i.e., ~0.1) that also decrease in relative proportion; since these are O2-NAFCs, these are necessarily the longest (i.e., highest carbon number) compounds among the O2-NAFCs detected. In comparison, O3-O7 compounds with lower H/C ratios (i.e., H/C ratio ≈ 1.5) increase in relative proportion. In addition, many formulae observed in the final sample were not initially observed in the untreated OSPW, increasing slightly in relative abundance. Increases are most prominent at a higher O/C ratio (i.e., O/C ratio < 0.2), which in combination with the trends observed in the previous Kendrick plot (Figure 4), suggest that changes in NAFCs in the CWTS are due, at least in part, to oxidative reactions. Smaller unsaturated compounds are the least susceptible

Interpretation of Kendrick and van Krevelen Plots
To visualize the predictions discernible from the PC1-coded Kendrick plot, sample data were plotted in a Kendrick plot [46] showing the difference in relative abundance of various NAFCs across the full spectrum ( Figure 4). As illustrated in Figure 4, the most substantial decreases in relative abundance were observed from compounds with an approximate Kendrick Mass of 220-280 with Kendrick mass defects (KMD) ≤ 0.1. There were also some modest decreases in the relative abundance of compounds with approximate Kendrick Masses of 240-325 with an approximate KMD of 0.165, and a smaller series with approximate Kendrick Masses of 280-315 with an approximate KMD of 0.175. On the other hand, most compounds from Kendrick Mass 200-300 and KMDs from 0.1-0.15 simultaneously increased in relative abundance. In addition, there are some formulae across the Kendrick plot that were not present in the untreated OSPW that increased slightly in relative abundance. These trends suggest that the most saturated compounds (i.e., low mass defects) from m/z 220-300 are the most susceptible to changes and/or dissipation in this system. Unsaturated compounds across a similar mass range therefore appear to be resistant to degradation and/or dissipation in the CWTS, leading to an increase in relative proportion of those unsaturated compounds.
Complementary observations were made by plotting data in a van Krevelen plot showing the net-change in the relative abundance of major O x (i.e., O 2 , O 3 , O 4 , O 5 , O 6 , and O 7 ) formulae to simultaneously examine the change in oxygen content and unsaturation of NAFCs across the sampling period ( Figure 5). As shown in Figure 5, some O 2 compounds with relatively high H/C ratios (i.e., H/C ratio > 1.6) decrease the most in relative abundance. There are also some compounds from H/C 1.2-1.5 and low O/C ratio (i.e.,~0.1) that also decrease in relative proportion; since these are O 2 -NAFCs, these are necessarily the longest (i.e., highest carbon number) compounds among the O 2 -NAFCs detected. In comparison, O 3 -O 7 compounds with lower H/C ratios (i.e., H/C ratio ≈ 1.5) increase in relative proportion. In addition, many formulae observed in the final sample were not initially observed in the untreated OSPW, increasing slightly in relative abundance. Increases are most prominent at a higher O/C ratio (i.e., O/C ratio < 0.2), which in combination with the trends observed in the previous Kendrick plot (Figure 4), suggest that changes in NAFCs in the CWTS are due, at least in part, to oxidative reactions. Smaller unsaturated compounds are the least susceptible to oxidation, increasing in relative to oxidation, increasing in relative abundance; conversely, increasing molecular size and saturation were associated with decreases in relative abundance.  As a note of caution, this method of comparing base-peak normalized abundance between unlike samples may not capture some changes, especially if various mixture components to oxidation, increasing in relative abundance; conversely, increasing molecular size and saturation were associated with decreases in relative abundance.  As a note of caution, this method of comparing base-peak normalized abundance between unlike samples may not capture some changes, especially if various mixture components  As a note of caution, this method of comparing base-peak normalized abundance between unlike samples may not capture some changes, especially if various mixture components simultaneously increase in concentration while the dominance of the base peak decreases. Such disparities between samples would generate a general increase in relative abundance, wherein many components may simultaneously increase in relative abundance in a comparison between two samples, preventing the observation of actual decreases in concentrations of sample components. Similarly, if the base peak were to increase in concentration while other sample components remained consistent, this would lead to the opposite phenomenon, decreasing relative abundance of numerous sample components simultaneously. However, this does not negate differences that are observed, and only serves to provide a note of caution about interpretation of "real" differences in sample component concentrations.

Heteroatom Class Distribution
The relative abundances of heteroatom classes in the untreated and wetland treated samples are displayed in Figure 6. For all assigned peaks, the relative abundance of each heteroatom class was determined by normalizing the peak abundance to the total abundance. From our previous work, the contribution of the O x compounds (x = 2-6) to the total abundance was the greatest, as compared to other compound classes in the untreated OSPW [36], and is in line with results from another study [27]. The O 2 -NAs accounted for 56.1% of the total NAFCs detected in the OSPW, followed by O 3 -NAs and O 4 -NAs. The abundance of the oxygen-containing compounds follows the order O 2 -NAs ≥ O 3 -NAs≈O 4 -NAs ≥ O 5 -NAs ≥ O 6 -NAs, which is characteristic of OSPW composition [16,47]. The evaluation of the transformation of NAFCs in the wetland system indicates that abundance of O x -NAs (3 ≤ x ≤ 6) increased while the O 2 -NAs species decreased (Figure 4). The O 2 -NAs decreased and total O x -NAs increased in the wetland system compared to the untreated OSPW ( Table 1). The increased abundance of the O x -NAs in the wetland systems is due to the oxidation of O 2 -NAs to higher O x -NAs. The oxidized NAs (O x -NAs) are deemed as intermediate products of the biodegradation of O 2 -NAs [48]. Also, past literature found that O 2 -NAs were transformed to O x -NAs during the biodegradation of commercial NAs [49]. It is worth mentioning that in earlier findings, high proportions of the O x -NAs (x = 3-6) were used as marker of OSPW sources for groundwater and surface water samples [50].
With respect to sulfur containing NAFC species, SO 3 , SO 4 , and SO 5 were the most abundant species, with proportions in the untreated OSPW as shown in Table 1. As reported in our previous study [36], the relative abundance of SO 4 and SO 5 increased while that of the SO 3 species decreased after treatment in the CWTS. Among the NAFCs, the nitrogen containing compounds (NO x ) showed the lowest abundance in both the untreated and CWTS treated OSPW samples detected by high-resolution negative-ion ESI-Orbitrap-MS (Table 1) [36]. A previous study reported a low abundance of nitrogen-containing species in the acid extracted organic fractions of OSPW characterized by ESI (−) [11]. Nitrogen-containing compounds have been detected in high abundance by ESI (+) and atmospheric pressure photoionization [34,51]. Additionally, low abundances of sulfur-containing and nitrogen-containing compounds 3.9% and 0.5%, respectively, were detected in the acid extractable organics of OSPW with ESI-FT-ICR-MS [11].
can penetrate the lipid bilayer and disrupt membrane integrity, resulting in cell death [36]. The O2-NA species implicated in toxicity are C ≥ 17 NAs [58], C15-18 and DBE = 4, and C14-17 and DBE = 3 [59]. The CWTS transformation of the classical NAs is expected to reduce OSPW toxicity.  [36] The feasibility of any treatment process involves both analytical evaluation and toxicity tests for any wastewater to be deemed safe for return to the aquatic environment. Currently, the critical requirements for the evaluation and careful choice of the best treatment options for OSPW are based on the analytical and toxicological assessment of the OSPW [38]. However, to date, there is no standard analytical method for OSPW-NA characterization and quantification, and therefore continuous development of robust analytical methods for the identification and measurement of NAs is warranted. The anticipated environmental discharge of treated OSPW will require a good analytical approach for the evaluation of the degradation of the most potent toxic components in the wastewater. The refinements in the data processing of the WAX SPE Orbitrap MS method provided improved insights into the fate of NAFCs in the CWTS. The refinements of the analytical tools may thus be useful for monitoring the progress of environmental remediation.

Degradation of Individual NA Species
The carbon number and DBE are structural characteristics of classical NAs, which provide evidence on their structural affinity for degradation. These are useful parameters for assessing the efficiency of degradation of classical NAs and the selectivity of treatment. In complex systems such as CWTSs, several processes such as sorption, photolysis, hydrolysis, speciation/ionization, oxidation, reduction, and biotransformation contribute to the removal of COPCs (e.g., NAFCs) [20]. However, this study did not investigate the major removal mechanism of OSPW-NAFCs in the CWTS and the elucidation of the major removal process requires further research. The high-resolution Orbitrap-MS data provided information on the molecular distribution of classical NA species. The plots of O 2 -NA concentration versus carbon number and DBE value before and after wetland treatment are given in Figure 6A,B. A similar NA profile was observed in both the untreated and treated OSPW; however, the classical NA concentration in the wetland system was lower. The total concentrations of O 2 -NAs in untreated OSPW decreased from 40.3 mg/L to 12.7 mg/L (after correction for measured ET, transformation efficiency of 69.5%), indicating that the CWTS has the efficiency to degrade O 2 -NAs. NA subgroups with higher initial concentrations (carbon numbers 12-17 and 13-16 with DBE 3 and DBE 4 respectively) were degraded significantly [36].
To further illustrate the changes in NA species that occurred in the treatment wetland, the ratios of NA Σno /NA Σni and NA ΣDBEo /NA ΣDBEi for carbon number and DBE respectively were calculated. NA Σno and NA Σni is the total concentration of O 2 -NAs with the equivalent carbon number in the outflow on day 27 of the treatment and the inflow samples on day zero respectively and used to indicate the extent of degradation of NAs with each carbon number. The NA ΣDBEo /NA ΣDBEi ratio is the total concentration of NAs with the equivalent DBE values in the outflow and inflow samples, respectively, and is used to indicate the extent of degradation of NAs with each DBE. A lower ratio (<1.0) means a better degradation in the treatment system [52,53]. To assess the influence of carbon number and DBE on degradation of classical NAs, a plot was generated showing changes in the most prevalent carbon numbers (11)(12)(13)(14)(15)(16)(17)(18)(19) and DBE values (2-8) of total NA and O 2 -NA species. Figure 6C show the degradation of O 2 -NAs according to DBE values and carbon number value in the wetland system after treatment for 27 days. The NA ΣDBEo /NA ΣDBEi ratio showed no distinctive pattern for the degradation of DBE compounds.
The ratio of NA Σno /NA Σni clearly indicates that the extent of degradation is directly proportional to the increase in carbon number. It is of interest that the CWTS showed high efficiency for the degradation of the more recalcitrant NAs. An earlier study observed that classical NAs with higher carbon numbers in crude oil were preferentially degraded under anaerobic conditions [54]. The increase in carbon number is correlated to a more hydrophobic character of NAs [55]. This may result in their accumulation in the wetland systems. The local concentration of NAs in a system could be increased because of hydrophobic interactions between NAs-NAs and NA/colloidal particles, making them more susceptible to biodegradation [52]. Bacterial cells that have unique hydrophobic surfaces are capable of directly adsorbing accumulated hydrophobic pollutants [52]. Therefore, more hydrophobic organic contaminants (or higher lipophilicity) are more available for adsorption onto or permeation through the cell membrane which in turn improves efficiency of biodegradation under certain conditions [56]. However, very high logKow may affect the bioavailability of NAFCs for biotransformation because these compounds could be strongly adsorbed to sediment or plants. Furthermore, the pH and the characteristics of the media can influence permeation into membranes.
As discussed above, there was an increase in the rate of transformation of O 2 -NA in the CWTS, as the concentration of O x -NAs increased. The O x -NAs have more -OH groups than O 2 -NAs making them less hydrophobic which may lead to reduced toxicity [57], because more hydrophobic compounds can penetrate the lipid bilayer and disrupt membrane integrity, resulting in cell death [36]. The O 2 -NA species implicated in toxicity are C ≥ 17 NAs [58], C15-18 and DBE = 4, and C14-17 and DBE = 3 [59]. The CWTS transformation of the classical NAs is expected to reduce OSPW toxicity.
The feasibility of any treatment process involves both analytical evaluation and toxicity tests for any wastewater to be deemed safe for return to the aquatic environment. Currently, the critical requirements for the evaluation and careful choice of the best treatment options for OSPW are based on the analytical and toxicological assessment of the OSPW [38]. However, to date, there is no standard analytical method for OSPW-NA characterization and quantification, and therefore continuous development of robust analytical methods for the identification and measurement of NAs is warranted. The anticipated environmental discharge of treated OSPW will require a good analytical approach for the evaluation of the degradation of the most potent toxic components in the wastewater. The refinements in the data processing of the WAX SPE Orbitrap MS method provided improved insights into the fate of NAFCs in the CWTS. The refinements of the analytical tools may thus be useful for monitoring the progress of environmental remediation.

Conclusions
The present investigation provided improved insights and confirmed earlier reports on extraction techniques used to monitor the degradation of NAFCs in CWTS. The result highlighted the importance of the sample preparation method on the identification of OSPW-NAFCs. The refinements in the data processing (PCA, Kendrick plots and van Krevelen plots) and combined use of the WAX SPE method and high-resolution Orbitrap-MS revealed new insights on wetland performance. The horizontal flow wetland system demonstrated high efficiency for the degradation of more hydrophobic NAs (with higher carbon numbers) compared to the less hydrophobic NAs (with lower carbon numbers). The finding is contrary to previous studies which showed that the carbon numbers may have either an inverse correlation or little influence on NA biodegradation [49,60]. These collective findings demonstrate the suitability of the WAX SPE and high-resolution Orbitrap-MS techniques for monitoring the degradation of NAFCs in CWTS. Moreover, in view of the lack of methods to measure the NAFCs directly in the wetland plant tissue, it is difficult to differentiate between possible adsorption and biodegradation mechanism in the CWTS. While the CWTS evidently degraded and oxidized the NAFCs, it is beyond the scope of the evidence presented here to conclude that biodegradation was the sole mechanism of the treatment.