3.1 Peaks detected in waters and plant extracts
For both water and aqueous plant eluates, the transformation of LC-HRMS output data resulted in a massive dataset (peaks list). After noise, background contaminant and blank correction, 13000 to 29000 and 50000 to 70000 peaks (defined by m/z, retention time and intensity) were considered to be positive detects in river waters and plant extracts, respectively. The positive detects represented contaminants from all possible sources in the environment – both anthropogenic and natural.
In a first step, we identified peaks common to vegetation and adjacent river water, which ranged from 4900 to 18500 peaks for the individual pairs (Figure 2).
For illustration, an aqueous extract of Galanthus nivalis and river water from an adjacent location are discussed here. As displayed in Figure 3, a larger number of common peaks (red spots) were obtained between Galanthus nivalis extracts and rain event water samples (Figure 3 – right) than for water samples under dry weather conditions (Figure 3 – left). A similar trend could be observed for all analysed plant-river water pairs. The majority of peaks in plant extracts (green spots) exhibit a higher retention time and thus hydrophobicity than those in water (blue spots). The peaks in common (red spots) are predominantly in the retention time region of moderately hydrophobic compounds.
The agreement of m/z and retention times still allows for different isobaric compounds detected at the same retention time and thus requires further steps to narrow down to common structures.
3.2 Peak prioritization and structural identification of metabolites
3.2.1 Prioritization of overlapping peaks
Peaks were prioritized for identification using a stepwise filtering approach demonstrated on the basis of Galanthus nivalis and a rain event river water sample from an adjacent location (Figure 4). After limiting positive detects in both samples to common peaks only (8574), the overlapping peaks were ranked based on intensity in plant extracts and corresponding water samples considering two general assumptions. (1) Peaks with low intensity in plant extracts (selected threshold 106) have low probability to enter to river water in a sufficient quantity to be detected. (2) Peaks appearing at higher intensity in river water than in plant extracts are unlikely to origin from the plants. Both criteria were used to exclude peaks of low priority. In our example, this prioritization step reduced the number of peaks to be considered to 1406 which is 8% of the initial peak list. Broad peaks with low intensity and not well-defined retention time were manually eliminated by inspecting the peak shape. In a next step, the elemental composition of each peak was evaluated based on accurate mass (with an error range given in 5 ppm for exact mass) considering the elements CHNOPS – commonly occurring in natural products (Bobeldijk Pastorova et al. 2001; Wolf et al. 2010; Pluskal et al. 2012). Finally, the isotopic fit analysis resulted in 261 (1.5% of initial peaks) tentatively identified candidate peaks.
3.2.2 Identification of unknown SPMs
All 216 peaks selected as candidates were subjected to further identification efforts combining a set of software tools for retrieving possible chemical structure with selection criteria based on database (and software) search and MS/MS fragment consideration as exemplified for two structures below. For a river water sample with high abundance of Galanthus nivalis in the catchment, we perceived 54 out of 216 candidate peaks plausible chemical structure using spectral database (MassBank and MZcloud) search and in-silico fragmentars (Metfrag, CSI Finger ID, CFM-ID). Analysing MS/MS fragment match with reference standards, we were able to identify nine of the metabolites (Figure 4) to confidence level 1 (Schymanski et al. 2014) while three more metabolites obtained in the remaining water samples resulting in a total of twelve identified SPMs and other metabolites. The stepwise identification of unknown SPMs will be demonstrated for two examples.
For one of the candidates, the accurate m/z of the unknown protonated molecule at a retention time of 0.8 min was determined to be 136.0619 mu. The PubChem search for the elemental composition resulted in five molecular formulas within 5 ppm mass accuracy. The isotopic pattern analysis confirmed the presence of N in the unknown molecule, thus compounds not containing N were excluded, which left C5H5N5 to be the only potential candidate with 284 registered chemical structures. Further, the data dependent MS/MS fragment ion masses of the unknown molecule were matched with fragmentation pattern of the suggested molecules in the library. Adenine as the compound with the highest spectral match was selected as potential candidate and confirmed with a reference standard based on retention time and MS/MS fragment (see Figure S2 and S3 in SI).
The second accurate mass, chosen for illustration, is 287.0549 mu eluting with a retention time of 10.4 min. Within the set limit, evaluation of the elemental composition using QualBrower of XCalibur resulted in 22 formulas applying a mass error window of 5 ppm. Formulas containing N and S in addition to C, H and O were discarded since the isotopic pattern analysis of full scan (MS1) spectra (Pluskal et al. 2012) did not provide any evidence on the presence of N and S in the candidate molecule. Consequently, the only remaining molecular formula C15H10O6 (Δ = -0.085 ppm) was taken as potential candidate, for which 302 candidate structures were proposed by the database (PubChem). For the determination of the chemical structure, the data dependent MS/MS fragment ion spectrum was submitted to MetFrag, CFM-ID and CSI:finger ID to compare those with in-silico predicted spectra for candidate structures retrieved from databases such as PubChem, KNApSAcK, natural product and KEGG. Among the structures suggested, the one with highest score and also with highest spectral similarity, namely kaempferol, was selected as plausible candidate structure. This compound could be confirmed in turn with a commercial reference standard based on retention time and MS/MS fragment match (see Figure S4 and S5 in SI). Thus, from the above analysis the suspected unknown molecule was confirmed to be kaempferol.
Following a similar approach, the presence of nicotiflorin, hyperoside, cynaroside (luteolin 7-O-beta-D-glucoside), trifolin (kaempferol-3-O-galactoside), alpinetin, isofraxidin, apiin, guanosine, quercetin and kaempferitrin was confirmed in river waters. The chromatogram and MS/MS spectra for the identified compounds are given in SI (see Figure S6 – S25 in SI). All the detected metabolites were also obtained in plant extracts, except alpinetin and kaempferitrin, with common peaks detected in water and plant samples but confirmed only in water with isobaric but not identical compounds in the plant extracts. Among the detected plant metabolites, 10 are SPMs while the nucleic bases adenine and guanosine are components of DNA and RNA and thus no SPMs in a strict sense but subsumed under the same abbreviation. The chemical structures for the identified metabolites are displayed in Figure 5. See Table S6 in SI for full information on the identified metabolites in both river water and plant extracts.
3.3 Distribution of the identified metabolites in river waters
SPMs of different classes, flavonoids (and their glucosides), coumarins and purine nucleobases were identified and quantified (Figure 6). In total, the presence of twelve SPMs in river waters from both catchments was confirmed with flavonoids being the predominant class detected. In general, most of the identified metabolites contain one or more phenolic groups representing a class of compounds found most abundantly in vegetation (Puri et al. 1998). The identified SPMs were detected in individual water samples at concentrations up to about 5 µg/L (Figure 6, and Table S6 in SI). The highest number and concentrations of identified SPMs have been found in two samples (ELP2 and ELP21) from the ELP catchment collected during heavy rain, while in none of the control (dry weather) samples the identified metabolites were detected (data not shown). This finding supports the hypothesis that rain events drive the leaching of SPMs to surface water.
Most SPMs were detected in water samples from both catchments, with the exception of alipinetin, hyperoside, kaempferitrin and quercetin which were detected in the ELP catchment only. Among the detected SPMs, adenine and isofraxidin were obtained at high frequency in both water samples and plant extracts. This has been followed by cynaroside in water samples and trifolin in plant extracts (Table 1 and Figure S26 in SI). In river waters, SPMs were detected in an overall concentration range of 0.02 to 5.1 µg/L (Figure 6, Table 1).
Table 1: Concentration of the detected metabolites in both river waters and plant extracts.
Metabolites
|
Formula
|
CAS No
|
Precursor ion (m/z)
|
Retention time (min)
|
MDL (µg/L)
|
River water
|
Plant extracts
|
Detection frequency
|
Concentration range (min-max, µg/L)
|
Detection frequency
|
Aqueous extractable concentration range (min-max, µg/g)
|
Adenine
|
C5H5N5
|
73-24-5
|
136.0619
|
0.8
|
0.2
|
7
|
0.4 - 2.6
|
5
|
35.0 - 59.3
|
Alpinetin
|
C16H14O4
|
36052-37-6
|
271.0962
|
10.3
|
0.004
|
2
|
0.023 – 0.050
|
0
|
ND
|
Apiin
|
C26H28O14
|
26544-34-3
|
565.1547
|
9.1
|
0.5
|
4
|
1.2 - 5.1
|
1
|
21.7
|
Cynaroside
|
C21H20O11
|
5373-11-5
|
449.1073
|
8.6
|
0.050
|
5
|
0.2 -2.1
|
3
|
11.1 - 50.6
|
Guanosine
|
C10H13N5O5
|
118-00-3
|
284.0984
|
1.0
|
0.2
|
4
|
1.1 - 4.0
|
5
|
42.8 - 189.5
|
Hyperoside
|
C21H20O12
|
482-36-0
|
465.1017
|
8.6
|
0.3
|
2
|
3.8 - 4.0
|
2
|
18.9 - 22.6
|
Isofraxidin
|
C11H10O5
|
486-21-5
|
223.0599
|
7.4
|
0.014
|
7
|
0.020 – 0.300
|
5
|
0.01 - 16.8
|
Kaempferitrin
|
C27H30O14
|
482-38-2
|
579.171
|
9.3
|
0.050
|
1
|
0.9
|
0
|
ND
|
Kaempferol
|
C15H10O6
|
520-18-3
|
287.0548
|
10.6
|
-
|
3
|
NQ
|
3
|
NQ
|
Nicotiflorin
|
C27H30O15
|
17650-84-9
|
595.165
|
9.2
|
0.2
|
3
|
1.9 - 2.2
|
1
|
88.0
|
Quercetin
|
C15H10O7
|
117-39-5
|
303.0496
|
8.6
|
0.6
|
2
|
1.9 - 2.5
|
2
|
54.6 - 74.7
|
Trifolin
|
C21H20O11
|
23627-87-4
|
449.1073
|
9.1
|
0.2
|
3
|
0.3 - 2.9
|
4
|
25.0 - 36.0
|
NQ – Not quantified, ND – Not detected
The purine bases adenine and guanosine were detected at concentration range of 0.4 – 4.0 µg/L and 35 – 59 µg/L in water samples and plant extracts, respectively (Table 1). Adenine is an aromatic base found in both DNA and RNA of living organisms. The compounds were previously isolated from a variety of plants (e.g., maize, tea and coffee plants) (Ashihara et al. 2008, Suzuki et al. 2007). Guanosine was reported to have neurotrophic and neuroprotective effects, evidenced from rodent and cell models study (Lanznaster et al. 2016).
Flavonoids, a class of natural compounds widely distributed in plants including kaempferol and quercetin were detected in several water samples and plant extracts from ELP and one from Bode catchment. Quercetin was obtained at an average concentration of 2 µg/L. Besides their potential positive effects such as antiproliferative, chemopreventive, and anti-inflammatory activities (Kumar and Pandey 2013), kaempferol and quercetin inhibit the acetylcholinesterase (AChE) activity in vitro (Xie et al. 2014; Orhan et al. 2007; Nugroho et al. 2018; Murray et al. 2013). Quercetin demonstrated toxic and carcinogenic effect in the kidney of male rats (Dabeek and Marra 2019; Dunnick and Hailey 1992).
The flavanone alpinetin and the glycosyloxyflavone kaempferitrin (a 3,7-dirhamnoside of kaempferol) were obtained in river waters from ELP, but not in the investigated plant extracts (despite overlapping peaks by isobaric compounds). However, the metabolites were previously reported from a variety of other plants in the environment – alpinetin from genus Alpinia (flowering plants) and kaempferitrin from Lathyrus (a genus in the legume family Fabaceae) (Hymavathi et al. 2009; Afendi et al. 2012; Conde et al. 1998; Wang et al. 2001; Lee et al. 2008)) no evidence for the presence of such plants along the investigated rivers. The measured concentration of kaempferitrin was 0.9 µg/L while alpinetin was present in concentrations of 23 and 50 ng/L. Besides its antibacterial and anti-inflammatory activities, alpinetin exhibited vasorelaxant effects on rat (Wang et al. 2001). It also showed potential effects in downregulating the immune system in mice (Guan et al. 2014). A study by Zhang et al showed that kaempferitrin competitively inhibited human liver microsomal Cytochrome P450 1A2 activity (Zhang et al. 2019).
The glycosyloxyflavone apiin was measured at high concentration (5 µg/L) in a water sample from the Bode catchment but was also obtained in two water samples from ELP at an average concentration of 2.9 µg/L. Another flavonoid glucoside, namely nicotiflorin (kaempferol 3-O-rutinoside) was obtained in rivers from both catchments – two from ELP and one from Bode catchment – at an average concentration of 2 µg/L. However, both metabolites were detected only in one plant extract each – apiin in Digitalis purpurea and nicotiflorin in Fraxinus excelsior from Bode catchment, though, Fraxinus excelsior is a characteristic plant in the ELP floodplain forest, too. The detection of apiin in ELP water samples indicates leaching also from other frequently occurring plant species (not considered in this work) including Apiaceae (Afendi et al. 2012) and stinging nettle (Urtica dioica) (Orčić et al. 2014). In vitro, apiin displayed anti-inflammatory activity (Mencherini et al. 2007). Nicotiflorin has many interesting pharmacological activities, such as decreasing arterial blood pressure and heart rate and hepatoprotective effects (Harborne and Baxter 1999). It was found to protect against memory dysfunction and oxidative stress in multi-infarct dementia model rats (Huang et al. 2007; Harborne and Baxter 1999).
In only two water samples from ELP, an average concentration of 3.9 µg/L was registered for hyperoside (a quercetin-3-O-D-galactoside). It was also detected in substantial concentrations in plant extracts (Fraxinus excelsior and Galanthus) from close vicinity, from which it could be emitted (Table S2 in SI). It may have potential as a therapeutic agent for the treatment of liver fibrosis (Wang et al. 2016). It improves cardiac function and prevents the development of cardiac hypertrophy via AKT signalling (Wang et al. 2018). Hyperoside was found to present a depressor effect on the central nervous system as well as an antidepressant-like effect in rodents which is, at least in part, mediated by the dopaminergic system (Haas et al. 2011). The water-extractable hypersoside from Hypericum species demonstrated an acetylcholinesterase inhibition effect (Hernandez et al. 2010).
Cynaroside and trifolin glycosyloxyflavones in water samples occurred at concentrations ranging from 0.2 – 2.1 and 0.3 – 2.9 µg/L, respectively (Table 1 and Figure 6). The former was identified in five water samples – four from ELP and one from the Bode catchment while the later was in three samples – two from Leipzig and one from Bode catchment. Both metabolites were also detected in plant extracts from both catchments. Cynaroside shown to cause a prominent anti-oxidant effect, inhibiting lipid and protein oxidation. It also displayed inhibitory effects of human liver cytochrome P450 (CYP) enzymes (Wang et al. 2019). Trifolin (kaempferol-3-O-galactoside), which is a galactose-conjugated flavonol, exhibits antifungal and anticancer effects (Li et al. 2005).
The coumarin, isofraxidin was obtained at an average concentration of 0.03 µg/L in two water samples from each location. In the rest of the water samples, except one from Bode catchment, it was found at an average concentration of 0.2 µg/L. The SPM was quantified in all the plant extracts – the highest being in Fraxinus excelsior, a characteristic tree along the rivers in both catchments. Apart from its numerous pharmacological activity such as antioxidant and anti-inflammatory, isofraxidin inhibited human liver cytochrome P450 (CYP) enzymes (Song et al. 2019).
3.4 Toxic risk estimation
The SPMs have been detected in water samples not as individual compounds but in mixtures of at least three SPMs co-occurring at all sites while at two samples even nine metabolites were detected (13 % sites) (Figure 7(a)). Thus, a preliminary mixture RQ based on a TTC of 0.1 µg/L exceeded 5 (and thus also 1 at all the sites), while at 7 sites a value of 10 and at 3 sites even a value of 50 was exceeded (Figure 7(b)). Individual concentrations of the detected SPMs, except isofraxidin (in three water samples) and alpinetin, were also above the TTC. Thus, toxic risks by individual SPMs and mixtures thereof and a contribution to overall toxicity of surface waters cannot be excluded and demand for additional efforts in hazard characterization.