Optical properties
Optical properties consisting of A254, SUVA254, FI, SR and
EEM-PARAFAC for the six samples for SPE-DOM analysis are presented in Table 1. Although large differences in DOC and A254 were observed for the
different samples (DOC range of 5.8 to 28.6 ppm; A254 from 0.202 to
0.844) some of the qualitative optical parameters such as the SR values
(range 0.91 to 0.98) and the FI values (range 1.30 to 1.44) all fell into a
relatively narrow range. In contrast, the SUVA254 values covered a
larger range from 2.72 to 5.11. A linear correlation was observed between
the DOC and the A254 (r2= 0.95).
SR and FI values were quite similar for all sample pairs and among
samples, suggesting that the molecular weight distribution and the
soil/higher plant vs. microbial contributions were quite similar among these
samples, or that the mineralization of wetland DOM leads to similar
compositional features for systems with different organic matter sources.
Detailed molecular characterizations of DOM in headwater streams from
different climatic regions (biomes) have been reported to exhibit remarkably
similar bulk characteristics, although site-specific features were also
identified in each case (Jaffé et al., 2012). However, in the case of
the SUVA254, some clear compositional variations between different
wetland DOM became apparent, where the samples from the more strongly
soil-OM (or peat) influenced, long-hydroperiod sites, featured higher
SUVA254 values compared to those with larger microbial and
emergent/aquatic plant influenced, short-hydroperiod sites. Indeed,
SUVA254 showed higher values for the peat-based FCE-L compared to the
marl-based FCE-S, the Paraguay River sample PAN-L compared to the wetland
channel PAN-S, and in the seasonally flooded Boro River floodplain OKA-L
compared to the occasional floodplain OKA-S. Although the differences in FI
for the PAN samples were not significant, FI values for the other sites were
expectedly inversely correlated to SUVA254 (Jaffé et al., 2008).
The application of the FCE PARAFAC model to the OKA and PAN samples resulted
in an excellent fit leaving no significant residues and was properly
validated. The application of the FCE PARAFAC model to assess fluorescence
characteristics of DOM from other wetlands was previously reported for the
Okavango Delta (Cawley et al., 2012). In addition, the distribution of the
EEM-PARAFAC components was also surprisingly similar among the six samples
with C1 being dominant, followed by C3 > C5 > C4 > C6
and C7 and with C2 and C8 showing the lowest relative
abundance. This trend is consistent with previous reports for the greater
Everglades ecosystem (Yamashita et al., 2010; Chen et al., 2013).
EEM-PARAFAC results for the three wetlands are shown in Fig. 1. The data
for FCE are presented as two separate sub-groups representing Everglades
National Park (ENP) sites and the Water Conservation Area 2 (WCA2), an area
located north of the ENP boundary where water resources are heavily managed
and agricultural runoff is significant (Yamashita et al., 2010). In general
terms, no significant differences were observed in the PARAFAC component
distributions between FCE, OKA and PAN, and the only difference of
significance was the relative abundance of the C2 PARAFAC component which
was higher in the WCA2 compared to all other study regions. In agreement
with the above, comparing the EEM-PARAFAC distributions between and among
the six stations, and amongst the larger data sets of collected surface water
samples (Table 1; Fig. 1), no statistically significant differences were
observed, although the range in values was large. Component C2 has been
suggested to be photo-chemically stable or possibly a photo-degradation
product (Chen et al., 2010; Cawley et al., 2012; Chen and Jaffé, 2014)
and has also been identified as derived from the oxidation of soil OM, being
exported from the Everglades Agricultural Area (EAA; located to the north of
the WCA; Yamashita et al., 2010). As such it is not surprising that the
levels for C2 are enriched in waters from the WCA2, which receives
significant canal inputs from the EAA. In the other wetlands and at
freshwater marshes in more distant regions of the Everglades, C2 is only a
relatively minor component of the DOM fluorescence signal. However, the most
interesting aspect of this comparison is that the FCE-based PARAFAC model
provided a perfect fit for both the OKA and PAN samples, suggesting that the
overall fluorescent properties of the DOM in the three wetlands are quite
similar.
NMR study
1H NMR spectra
High-field (800 MHz) NMR spectra with cryogenic detection performed on six
samples (paired long- and short-hydroperiod sites from each wetland) revealed
an exceptional coverage and chemical description of wetland organic proton
and carbon chemical environments. The 1H NMR spectra of wetland DOM
acquired with solvent suppression showed the prevalence of rather smooth
bulk signal envelopes reflecting intrinsic averaging from massive signal
overlap with a considerable variance in abundance for all major chemical
environments. In addition, rather minor superimposed sharp individual NMR
resonances were indicative of biological signatures and occurred in the
order PAN > OKA > FCE (Figs. 2 and S1). From higher
to lower field (from right to left), abundant (a) aliphatics, (b) “acetate-analogues”, (c) carboxyl-rich alicyclic molecules (CRAM),
(d) “carbohydrate-like” and methoxy, (e) olefinic, and (f) aromatic NMR
resonances showed well-visible and rather broad maxima (letters given
according to Fig. S1).
Superimposed small NMR resonances indicative of comparatively abundant
biological and biogeochemical molecules were most significant in the
aromatic section (f), well noticeable in sections (e) and (a), and of
continual lesser occurrence in the order c > b > d
(Fig. S1). The area-normalized 1H NMR spectra of the six DOM samples
(Fig. 2) showed more variance than their respective 1H NMR section integrals (Table 2), a plausible consequence of intrinsic averaging across
sizable chemical shift windows (Hertkorn et al., 2007). One-dimensional
1H NMR spectra of wetland SPE-DOM revealed clear distinctions according
to sample location, with pronounced congruence between the three pairs of
samples (Fig. S1). Within sample pairs, internal differences mainly referred
to intensity variations (denoting variable abundance) rather than to
alterations of NMR resonances positioning (denoting molecular diversity).
The relative disparity was largest between both FCE-L and FCE-S whereas PAN
and OKA samples were more alike among themselves (Fig. S2a–d). Otherwise, molecular divergence was most obvious in the case of
unsaturated protons (δH > 5 ppm). Subtle
relative changes in composition between pairs of samples were readily
visualized by superposition NMR spectra in which the relative NMR section integrals of each aromatic and aliphatic substructure had been normalized
to 100 % (Fig. S2e–g).
The larger discrimination observed between 1H NMR spectra of DOM from
different wetlands in comparison with the intrinsic variance among DOM
within each wetland already suggested the presence of an individual molecular
signature, characteristic of each particular wetland. Table 2 shows the
respective 1H NMR section integrals for the six samples under study.
Generally, the OCH, XCCH and
CCCH aliphatic chemical environments represented
nearly equal contributions to make up approx. 90 % of the spectrum with
the CCCH units consistently exceeding 30 %.
Carboxyl-rich alicyclic molecules (CRAM) and functionalized and pure
aliphatics followed the order FCE (L > S) > PAN ≈ OKA. Molecular divergence was most noticeable in the chemical
environment of unsaturated protons, where the ratio of aromatic to olefinic
protons declined in the order FCE > PAN > OKA. Here,
Har (δH > 7 ppm)
and C = CH, O2CH (δH : 5.3–7 ppm) contributed less than 5 % each to the overall
spectra. The differences in the NMR spectra (L-S) obtained for FCE, OKA and PAN wetland
DOM were computed from area-normalized NMR spectra (Fig. S2c–d) and
indicated congruent behavior for OKA and PAN SPE-DOM in the purely
aliphatic section (δH < 3 ppm), with moderate increase
of CnCH groups (n > 1; δH < 1.6 ppm).
The alterations in FCE-based aliphatics were governed by
a marked increase of CRAM whereas the abundance of CnCH decreased
(Fig. S2d). Interestingly, a rather concordant decline
of methoxy groups (primarily methyl esters) was observed for both FCE and
PAN (Fig. S2d). Polycarboxylated and PAH-derived aromatics (δH > 8 ppm)
were markedly increased in FCE-L as compared with FCE-S
(see below).
For improved assessment of unsaturated protons, the respective chemical
shift range was divided into several sections, comprising (f1; letters
according to Fig. S1) polycyclic and polycarboxylated aromatics as well as
six-membered nitrogen heterocycles (δH > 8 ppm);
(f2) electron-withdrawing substituents (COX; Perdue et al., 2007;
δH≈ 7.3–8.0 ppm); (f3) electroneutral
substituents (alkyl, H, R; δH≈ 7.0–7.3 ppm);
(f4) electron-donating substituents (OR, OH, phenolics; δH ≈ 6.5–7.0 ppm); (e1) polarized and conjugated olefins;
(δH≈ 5.5–6.5 ppm); (e2) isolated olefins
(δH≈ 5.0–5.5 ppm) – this section features however
contributions from anomeric protons and certain ester groups (see discussion
of 2-D NMR spectra). The relative and absolute abundance of electroneutral
substituted and phenolic aromatic compounds was maximal in OKA, and
declined through PAN to FCE. The ratio of conjugated olefins and aromatics
was similar in FCE and PAN; however, the abundance of these units was lower
by ca. 30 % in FCE. DOM from FCE-L showed higher proportions of isolated
olefins and, possibly, anomeric positions within carbohydrates.
TOCSY NMR spectra (800 MHz, CD3OD) of wetland SPE-DOM.
(a) TOCSY cross-peaks between aliphatic protons
(X–Csp3H–Csp3H–X; X:
C, O) for samples FCE-S. (b–d) TOCSY cross-peaks
between unsaturated protons (X–Csp2H–Csp2H–X; X: C, O) for samples FCE-S,
PAN-L and OKA-S respectively. Section (a): H3C–Cn–X cross-peaks, with n= 1 (δH > 3) and n > 1 (δH < 3);
where X is any heteroatom, likely oxygen; section (b): –C–CH–CH–Cn–X–, intra-aliphatic cross-peaks; section (c): α,β-unsaturated and conjugated double
bonds: HColefin= ColefinH–(C = O)–X; section (d): polarized α,β-unsaturated
double bonds: HColefin= ColefinH–(C = O)–X; section (e): congested fjord region in polycyclic
aromatics; section (f): aromatics HCaromatic–CaromaticH with
ortho- or/and para-oxygenated substituents (classic aromatic substitution of DOM);
section (g): condensed and strongly electron-withdrawing aromatics HCaromatic–CaromaticH (multiply
carboxylated, N-heterocycles); section (h): (more extended) polycyclic
aromatics, polycarboxylated aromatics, N-heterocycles. (d) Sections of
chemical shift for substituted aromatics as proposed by the SPARIA model
(substitution patterns in
aromatic rings by increment analysis): COR, electron-withdrawing substituents; R, electroneutral substituents; OR,
electron-donating substituents (Perdue et al., 2007).
Within this, the FCE samples showed the lowest proportion of unsaturated
protons, and among them, the short-hydroperiod site FCE-S was marginally
depleted in abundance of carboxylated aromatic protons compared to the
longer-hydroperiod site FCE-L, possibly due to higher light exposure at the
short-hydroperiod site. Such differences among samples from PAN and OKA were
not significant. Ratios of aliphatic to aromatic signals (CCCH / Har; see data in Table 2)
were also highest for the FCE samples, suggesting enrichment in
microbial-derived DOC (periphyton sources) compared to the PAN and OKA
samples, but also featuring differences between long- and short-hydroperiod
sites, where preservation of aliphatics at long-hydroperiod sites seemed to
be favored for all wetlands. These differences may at first conflict with
previous reports which found larger periphyton contributions to DOC at FCE-S
compared to FCE-L (Chen et al., 2013) suggested to be related to drying and
re-wetting of periphyton mats during the dry-to-wet transition at FCE-S, and
higher relative contributions of soil-derived DOM in FCE-L compared to
FCE-S. Similarly, in the case of the long- and short-hydroperiod comparison,
the higher CCCH / Har ratios coincided with higher SUVA values for
the DOM-L samples, suggesting a difference in the relative contribution of
microbial vs. higher plant/soil-derived DOM for CDOM compared to bulk DOM.
CDOM, often used as a proxy for DOM, only represents a small fraction of the
bulk DOC and does not include aliphatic molecules as those determined here.
As such, while being a convenient and useful proxy for DOC sources,
CDOM-based measurements might be less sensitive for the evaluation of
compositional differences between similar samples.
Methoxy NMR resonances for FCE-S compared to FCE-L were not only more
abundant, but were also shifted to lower field, indicating increased
fractions of aromatic methylethers and methylesters. FCE-S undergoes
periodic drying and thus exposure of soil OM (SOM) to atmospheric conditions
and intense sunlight exposure of DOM after high evaporation (drying)
conditions. As such, much of the SOM can be aerobically oxidized to CO2
creating marl soils. It is thus plausible that increased aerobic microbial
oxidation and photo-exposure at this short-hydroperiod site might enhance
DOM oxidation compared to the long-hydroperiod site (FCE-L). In addition,
while OKA showed an appreciable shoulder at δH > 3.75 ppm indicative of aromatic methyl esters and ethers at, however, reduced
relative abundance, this distinction was absent in both PAN and FCE (Fig. 2).
1H, 13C HSQC NMR cross-peaks of FCE-S; section of
unsaturated (olefinic and aromatic) protons δH= 4…10.5 ppm. Assignment in analogy to South Atlantic SPE-DOM
FMAX (Hertkorn et al., 2013) with key substructures denoted as follows. Section (a): anomeric CH in carbohydrates (sp3-hybridized); section (b):
isolated olefins; section (c): C-conjugated olefins, certain five membered N-,
O- and S-heterocycles (δH < 6.5 ppm); section (d):
multiply oxygenated aromatics including oxygen heterocycles, lignin
derivatives, syringyl units (S2/6); section (e): phenols, classical oxygenated
DOM aromatics, lignin derivatives, guaiacyl units (G2), certain admixture of
carbonyl derivatives (likely carboxylic units), causing down-field 1H NMR chemical shift (δH > 7.3 ppm); section (f):
classical DOM aromatic, lignin derivatives, guaiacyl units (G5/6),
para-coumarate (C3/5); section (g): classical DOM aromatics with high
proportion of carboxylated units; at δH > 8 ppm:
multiply carboxylated aromatics, classical PAH and certain six-membered
nitrogen heterocycles; sterically uncongested PAH; section (h): α,β-unsaturated double bonds for δC > 140
ppm, including double bonds adjacent to aromatics: C–HColefin= ColefinH–(C = O), Car–X;
section (i): nitrogen heterocycles, heteroatom substituted polycyclic
aromatics; section (j): certain six-membered nitrogen heterocycles, very
likely with more than one nitrogen. The green area highlights the HSQC cross-peak region accessible to single benzene rings substituted by common
electron-withdrawing, neutral and electron-donating common substituents of
natural organic matter; SPARIA: substitution
patterns in aromatic rings by increment analysis (Perdue et al., 2007).
In addition to the characteristics described above, the FCE samples showed
the largest proportion of aromatic compounds substituted with carbonyl
derivatives (most likely carboxylic acids; δH > 7.3 ppm). This pattern is in accordance with the presence of
dissolved black carbon (DBC) at these wetland sites, where the highest
abundance was reported for the FCE samples (Ding et al., 2014a). The
relatively large fraction of protons with very large down-field chemical
shift (δH > 8 ppm) suggested the presence of
six-membered nitrogen heterocycles as well as that of polycyclic aromatic
hydrocarbons (PAH). These units followed the abundance order PAN > OKA > FCE and could be related in part to the
presence of dissolved black nitrogen (DBN; Ding et al., 2014b). However, the
ratio of olefinic protons (δH ∼ 5.2–6.8 ppm)
to aromatic protons (δH > 6.8 ppm; but see HSQC
cross-peaks; Figs. 4 and S4) followed the order FCE > PAN ≈ OKA.
The distribution of aromatic protons in OKA indicated an elevated abundance of
electroneutral (alkyl, H; δH≈ 7.0–7.3 ppm) and
electron-donating substituents (OR, OH; δH < 7.0 ppm)
in contrast to both FCE and PAN SPE-DOM which showed similar distribution of
aromatic protons with larger proportions of electron-withdrawing
substituents (COR; δH > 7.3 ppm) at, however,
different overall abundance (Fig. 2; Table 2). In contrast, the abundance of
aromatics with electroneutral (R) or electron-donating substitution (OR)
with δH ∼ 7.3–6.6 ppm (Perdue et al., 2007)
followed the order OKA > PAN >> FCE (Fig. 2), likely reflecting the enhanced relative contributions of higher plant-derived DOM (in different degrees of oxidation) for the OKA and PAN compared
to the FCE. In conclusion, one-dimensional 1H NMR spectra showed a
considerable molecular divergence of aromatic molecules in the DOM of the
three wetlands, where the compositional features seemed driven by both source
strengths and variations in biogeochemical processing.
Although some methoxy groups can be formed by reaction of hydroxyl groups in
natural DOM and methanol during storage at ambient temperature (as SPE-DOM;
Flerus et al., 2011), the HCO NMR section integral,
which was found to be typically larger by ∼2 % for the
respective short-hydroperiod samples (Table 2), might reflect larger
abundance of native methyl esters at these sites or larger abundance of DOM
methanolysis products.
13C NMR spectra
13C NMR spectra of wetland DOM were not overly conspicuous, with
limited variance of spectra appearance and 13C NMR section integrals
(Fig. S3; Table S1). The abundance of non-functionalized aliphatics followed
the order FCE-L > FCE-S > PAN > OKA,
whereas aromaticity followed a near-reverse order FCE-L ≈ FCE-S < OKA ≈ PAN. DOM from FCE-L showed depletion of
carbohydrates and increase of lipid-like compounds (Table S1). The near-invariant abundance of carbonyl derivatives (most likely carboxylic acids)
for all DOM could imply that a sizable proportion of low-field 1H NMR
resonances with chemical shift δH > 7.3 ppm, which
were more abundant in PAN than in the others (Fig. 2; see also aromatic
TOCSY cross-peaks, Fig. 3), actually represented (substituted) PAH (with
δC < 140 ppm; Hertkorn et al., 2013) rather than
(poly)carboxylic aromatics (with δC ∼ 167–187 ppm; Figs. 2 and S2; Tables 2 and S1). Computed average H / C ratios from a
basic reverse 13C NMR-based mixing model ranged in the order FCE-L > FCE-S > PAN-S ≈ OKA-L (13C NMR spectra
of PAN-L and OKA-S were not acquired) and primarily reflected variable
content of aliphatic structures (δC ∼ 0–47 ppm). The computed O / C ratio was near-equal for the OKA, PAN and FCE-S
samples, whereas that of FCE-L was lower by ∼0.07 units.
Here, a reduced abundance of oxidized aliphatic units (HCalO) was primarily responsible, because phenolic and
carboxylic content followed the order OKA-L ≈ PAN-S > FCE.
2-D NMR spectra
The 2-D NMR spectra provided remarkable richness in detail and refined
preliminary assignment-proposals from the one-dimensional 1H and
13C NMR spectra. TOCSY NMR spectra (Fig. 3) revealed a wide range of
methyl groups (H3C–CH–X; X:
C, O; Fig. 3a, section a); a contiguous, ill resolved cross-peak reflected a
large number of intra-aliphatic correlations (C–CH–CnH–CH–C; n= 0–2; Fig. 3a,
section b), and fewer cross-peaks in between oxygenated aliphatics
(O–CH–CH–O; Fig. 3a, δH > 3.4 ppm). Protons bound to sp2-hybridized carbon
produced better-resolved TOCSY cross-peaks and were part of various α, β-unsaturated olefins (Fig. 3b, section c–d), oxygenated and
carbonyl (COX) derivatives of benzenes with up to three COX substituents
(Fig. 3c, section f–h) as well as six-membered nitrogen heterocycles and
more extended aromatic systems with up to several aromatic rings (Fig. 3b and c, section (e) and Figs. 3d and 4). As mentioned earlier, such compounds might be
related to the presence of combustion-derived compounds such as DBC and DBN
(Ding et al., 2014a, b) and even dissolved black sulfur (DBS) (Hertkorn et al., 2013;
see attendant discussion of FTICR mass spectra). In contrast to common
five-membered heterocycles, (di)benzothiophene derivatives exhibit NMR
resonances ranging from δH: 7.4–8.1 ppm; corresponding HSQC
cross-peaks of DBS would appear in section (g), Fig. 4.
HSQC NMR spectra of PAN and OKA did not show peculiar features which were
not observable in those of both FCE samples and therefore will be not
discussed here. The HSQC NMR spectra of both FCE-S and FCE-L were remarkably
similar and produced near-identical overlay NMR spectra with some
discernible variance in HSQC cross-peak amplitude rather than positioning
(data not shown). This behavior is expected from comparison of the
one-dimensional 1H NMR spectra. These display differences in relative
amplitude rather than positioning of NMR resonances which is indicative of
variance in abundance of certain molecules rather than variance in molecular
diversity (see, however, discussion of CHOS compounds present in FCE DOM as
derived from FTICR mass spectrometry). About 90 % of overall HSQC cross-peak integral resided in a contiguous expansive superimposed assembly of
HSQC cross-peaks originating from protons bound to sp3-hybridized
carbon (Fig. S4).
Methylene (CH2) selective 1H, 13C DEPT –
HSQC NMR spectrum of SPE-DOM FCE-S with assignment of major substructures;
general colors: CH3: red; CH2 green, and CH: gray; section (a):
C–CH3 cross-peaks; section (b): C=C–CH3 and –SCH3
cross-peaks; section (c): acetate H3C–C( = O)–O–C–; section (d):
C2CH2 cross-peaks; section (e):
–C–CH2–COOH cross-peaks; section (f):
C3CH cross-peaks; section (g): only methoxy
(OCH3) cross-peaks are shown here; see
insert: section (g1): H3COH (HD2COD
shows methine carbon); sections (g2) and g3: aliphatic (g2) and
aromatic (g3) methyl esters H3CO–C( = O)–C–; section (g4) and (g5): aromatic (g4) and aliphatic (g5) methyl
ethers H3CO–C–C;
section (f): C3CH cross-peaks; section (h):
oxomethylene (OCH2) cross-peaks, likely
from carbohydrates; section (i): OC2CH cross-peaks;
section (j): methylene bound to esters –C–H2CO–C( = O)–Z– (see main
text).
The resolution of these expansive aliphatic HSQC cross-peaks of FCE-S (Fig. S4) could be remarkably improved by spectral editing according to carbon
multiplicity (Fig. 5). The combination of methyl- and methylene-selective
DEPT-HSQC NMR spectra revealed well-discriminated cross-peaks for all three
types of protonated carbon; i.e., methyl, methylene and methine (Fig. 5). The
chemical diversity of X–CH3 groups as indicated by DEPT HSQC cross-peaks (section (a), Fig. 5) was noteworthy, and the near-Gaussian distribution
of C–CH3 cross-peak amplitude in 1H and 13C NMR frequencies
indicated near-maximum diversity of aliphatic chemical environments
associated with these methyl groups. However, classical methyl groups
terminating extensive, purely aliphatic units (δH < 1.0 ppm;
CCCCH3 units) contributed less than
20 % to the total CCH3 HSQC cross-peak
integral. The large majority of C–CH3
units was sufficiently proximate to carbonyl derivatives (i.e., most likely
carboxylic acids) to let those experience down-field chemical shift
anisotropy from these nearby carbonyl groups, resulting in chemical shifts
ranging from δH ∼ 1.0 to 1.7 ppm
(cross-peak a; Fig. 5). Alicyclic structures (e.g., CRAM;
Hertkorn et al., 2006) facilitate clustering of chemical environments as
shorter paths of chemical bonds between different substituents are realized
in rings rather than in open chains. Another ∼20 % of CCH3 in FCE was bound to olefins (C = C–CH3),
with a possible contribution of S–CH3 groups (section (b); Fig. 5).
The carbon-bound methylene (C–CH2–C)
cross-peak occupied an impressively large area down to δH ∼ 3.5 ppm, well into the proton chemical shift range
commonly attributed to OCH units. The two major
chemical environments discriminated were methylene more distant to COX
(C–CH2–Cn–COX, with n≥ 1, and
δH < 2.1 ppm cross-peak d; Fig. 5), and methylene
groups directly proximate to carboxylic groups (in α-position; i.e.,
C–CH2–COX, with δH > 2.1 ppm cross-peak e; Fig. 5). The former shows a wider range
of remote carbon substitution as indicated by the substantial spread of
respective carbon chemical shifts (ΔδC: 24/16 ppm,
respectively for section (d/e) HSQC cross-peaks; Fig. 5; see also Fig. 8b in
Hertkorn et al., 2013). A wide variety of aliphatic and aromatic
methylesters and methylethers were also found, the latter being virtually
absent in marine SPE-DOM. Here, aliphatic methyl esters were most abundant
(section (g2) in Fig. 5), aromatic methyl esters (section (g3) in Fig. 5) and methyl ethers (section (g4) in Fig. 5) were of similar abundance,
and clearly recognizable aliphatic methyl ethers were also present (section (g5) in Fig. 5). Oxomethylene (OCH2)
occurred in the form of carbohydrate side chains (section (h); Fig. 5), and a
remarkable set of aliphatic oxomethylene (OCH2) HSQC cross-peaks
(δH/C∼ 3.4–4.0/58–72 ppm; section (j) in Fig. 5)
was present in SPE-DOM FCE-S, which does not correspond to common lignin
β-aryl ether units, which resonate in this 1H and 13C NMR
chemical shift range, but commonly comprise Car–CH–O–,
i.e., methine substructures. Analogous oxomethine
substructures are also found in phenylcoumaran, resinol and dibenzodioxocin
units as well, whereas oxomethylene units with δH > 4.5 are rare in common lignins (Ralph et al., 1998; Yelle et
al., 2008; Martinez et al., 2008; Wen et al., 2013; Yuan et al., 2011). This
peculiar HSQC cross-peak was discovered in FCE-S wetland SPE-DOM (section (j)
HSQC cross-peak in Fig. 5) but since then has also been observed (in
retrospect) with lesser distinction in other SPE-DOM including those from
marine sources. The singular positioning of a methylene group in the 1H
and 13C NMR chemical shift space strongly restrains the potential
diversity of its chemical environments: it has to represent a OCH2
group (methylene as defined by the phase in 1H, 13C DEPT HSQC NMR
spectra; single oxygen because of δC: any O–CH2–O
environment would resonate at δC > 90 ppm).
Similarly, common O–CH2–N chemical environments would resonate at
higher field than observed in both δH/C, but cannot be excluded
entirely in case of peculiar remote substitution. The most plausible
substructure is OCH2C; then, δH from 5.3–5.7 ppm
warrants presence of an ester group: this implies a
–C–(C = O)–O–CH2–C
substructure. However, alkylation alone will not produce the necessary low
field δH observed. This leaves –C–(C = O)–O–CH2–C = O as a
plausible group; possibly confined with a carboxylic group such as
–C–(C = O)–O–CH2–COOH or as an ester –C–(C = O)–O–CH2–COOR. Both
these substructures have a decent propensity to form enols with variable
double-bond character –C–(C = O)–O–CH = CH(OH)2. A partial double bond
character, which might be possibly controlled by mutual interactions in the
complex DOM mixture of molecules, would also explain the observed spread of
chemical shift in 1H and 13C NMR frequencies in this HSQC cross-peak even if the methylene group itself in –C–(C = O)–O–CH2–COOH is
four (carbon) or five (proton) bonds away from the most proximate atom
position where substitution may affect its chemical shift.
Several thousands of acid and ester derivatives of acetoacetic acid
– H3C–(C = O)–O–CH2–COOH – are known in the literature. Here, many of
the common esters comprise lipid substructures such as n-alkanes, sterane
and other polyalicyclic hydrocarbons, trimethylammonium salts, among others,
suggesting a natural origin of these compounds also in wetland SPE-DOM.
While substructures with –O–CH2–COOZ (Z: H, R) will produce distinct
“oxomethylene (OCH2C)” cross-peaks in
1H, 13C DEPT HSQC NMR spectra (section (j) cross-peak; Fig. 5), the
derivatives with –O–CHCHn–COOZ substructures (Z: H, R) will contribute
to the 1H NMR down-field section of the expansive “oxomethine
(OCHC2)” 1H, 13C DEPT HSQC cross-peak
(section (i) cross-peak; Fig. 5) and will not be readily discerned owing to a
larger variance in remote substitution. In addition, oxomethylene units
without geminal and vicinal adjacent protons will very likely produce
intense singlet NMR resonances, contributing to the enhanced visibility of
HSQC cross-peaks even at rather limited relative abundance. Further
evaluation of aliphatic spin systems in FCE-L provided evidence for massive
aliphatic branching in CCCH units and of large
chemical diversity of remote carboxylic substitution (Fig. S6).
TOCSY and HSQC NMR spectra demonstrated the presence of olefinic and aromatic
unsaturation in all wetland SPE-DOM (Figs. 3, 4 and S5). The FCE-S
showed the most informative detail of HSQC cross-peaks arising from
unsaturated Csp2H groups (Fig. 4). In comparison with marine SPE-DOM
(see Fig. S5 and attendant discussion), wetland SPE-DOM displayed a more
restricted chemical diversity of conjugated olefins (Fig. 4) whereas all
kinds of oxygenated aromatics, i.e., those substituted with
electron-withdrawing (e.g., COOH) and electron-donating (e.g., OR)
substituents were much more abundant and chemically diverse in (all) wetland
DOM (not all data shown). The latter finding is indicative of polyphenol
input from vascular plants (e.g., lignin derivatives) into wetland DOM
whereas aromatics in marine SPE-DOM mainly reflect marine natural products
(Fig. S5). In general, aromatic unsaturation (as deduced from proton NMR
integrals; Table 2) followed the order PAN > OKA > FCE
(Fig. S1), whereas olefinic unsaturation followed the order OKA ∼ PAN > FCE (Fig. S1). Aliphatic-to-aromatic ratios
changed across the different samples with an order of FCE > PAN > OKA, suggesting higher relative contributions from periphyton
in the FCE, whilst the PAN and OKA were more influenced by higher
plant-derived organic matter including lignins. The olefinic-to-aromatic
ratios (FCE-L: 0.44; FCE-S: 0.39; PAN: 0.38; OKA: 0.41) were computed from
adapted 1H NMR section integrals (δH: 10–6.5 ppm
(aromatics)/6.5–5.0 ppm (olefins); Table 2) owing to
HSQC cross-peak positioning which indicated major contribution of oxygenated
aromatics CarO (at δH: 7.0–6.5 ppm; Fig. 4) and showed lower values than oceanic DOC – see Hertkorn et
al. (2013), who reported olefinic-to-aromatic ratios in the range of 1.2 to
3.0. It is likely that this significant difference is due to the
contributions of higher plant, lignin-rich carbon in the wetlands compared
to marine DOM. The slightly elevated olefin content found in FCE-L may
result from the contribution of periphyton-derived DOM in the Everglades
(Maie et al., 2005; Chen et al., 2013; see also above). In addition, all
three sites are known for frequent and seasonal fires and have been reported
to contain DBC (Ding et al., 2014a) in abundances
close to 10 % of their DOC on a global average (Jaffé et al., 2013).
However, the DBC content (as %DOC) in the FCE was higher (as high as
20 % of DOC) than for the PAN samples (13 and 14 % for PAN-L and
PAN-S respectively), and these were higher than the OKA samples (9.4 and
6.3 % for OKA-L and OKA-S respectively) studied here (Jaffé et al.,
2013). In addition, the presence of six-membered N-containing heterocycles
in these samples might be indicative of the presence of DBN (Fig. 4i and j), which has previously been reported in the FCE (Maie et al.,
2006) and proposed to consist of polyaromatic molecules containing
pyrrolic-N, and multiple carboxylic substituents (Wagner et al., 2015b). With regards to the degree of
oxidation of the aromatic signal, the OKA showed the highest proportion of
electron-donating groups (Figs. 2, S1 and S2), such as phenols and
ethers, possibly related to lignin oxidation products, while the FCE
featured the highest shares of electron-withdrawing substituents (e.g.,
carboxyl groups) possibly associated with DBC. Although all three ecosystems
have climates leading to high light exposure, the high levels of DOC in the
FCE suggest some degree of self-shading, while DOC in the OKA is generally
lower and the system is known for its capacity to photo-degrade DOM (Cawley
et al., 2012). Thus, the degree of photo-exposure of the DOM combined with
combustion by-products such as DBC, may be the driver controlling the
oxidation state of the aromatic fraction. The photo-reactivity of DBC in
marine environments has recently been shown (Stubbins et al., 2012) and may
play a role in the lower DBC levels observed in the OKA samples.
Left panel: negative electrospray 12 T FTICR mass spectra of
Wetlands SPE-DOM (inset in Fig. 6f shows an enlarged mass view of a mass
range of 6.0 Da. Right panel: expansion of the mass segment m/z = 465.00–465.20 (Δm = 0.2 Da; asterisk in inset Fig. 6f), with
assignment according to CHO, CHNO, CHOS and CHNOS molecular series.
(a)
OKA-L; (b) OKA-S; (c) PAN-L; (d) PAN-S; (e) FCE-L; (f) FCE-S.
FTICR mass spectrometry
Ultrahigh-resolution Fourier transform ion cyclotron mass spectra (FTICR/MS)
of DOM may provide several thousands of mass peaks for individual
samples (Koch et al., 2005; Kujawinski et al., 2009), of which many hundreds
were assigned here to extended CHO, CHNO, CHOS and CHNOS molecular series
(Schmitt-Kopplin et al., 2010) based on the technique's excellent mass
accuracy and mass resolution (Fig. 6 and Table 3). Although detailed
FTICR/MS data are derived only from a few paired DOM samples (long and
short hydroperiod) for each wetland, a slightly higher number of mass peaks
(relative difference < 6 %) and of assigned molecular formulas
(relative difference < 1 %) was observed for the FCE-L compared to
the FCE-S, whereas elevated counts of mass peaks and assigned molecular
compositions were found in the case of the PAN-S and OKA-S samples (relative
difference < 2 %; Table 3). Molecular weights ranged in the order
FCE-S > FCE-L ∼ PAN > OKA (Table 3).
This admittedly minor molecular weight difference was not reflected in the
SR values of these samples (Table 1) which were quite similar. However,
SR only represents a molecular weight proxy for CDOM and might not be
sensitive enough to reflect minor differences accurately. In general, while
SPE-DOM of both OKA and PAN showed near 57 ± 2 % CHO, 8 ± 2 % CHOS, 33 ± 2 CHNO and < 1 % CHNOS molecules, the mass
spectra of FCE samples were fundamentally different compared with respect to
both OKA and PAN as well as among themselves (Fig. 6; Table 3; see also Fig. 7). Sample FCE-S appeared most distinct from all other samples both with
respect to total count of ions, overall mass peak distribution and with
respect to molecular diversity within nominal mass ranges (Fig. 6). Here,
FTICR mass spectra of both FCE samples showed the conspicuous doublets of
CHO/CHOS pairs visible at high resolution (Δm (C-3H4S) = 2.4 mDa) indicating a nominal exchange of H4S against C3
(Schmitt-Kopplin et al., 2010), whereas all other samples showed both lower
abundances and diversity of CHOS compounds (Figs. 6, 7 and S7). In
case of the FCE samples, CHOS and CHNOS compounds were markedly enriched at
the expense of CHO and CHNO compounds. While the proportion of CHNO (21 ± 1 %) and CHNOS (9 ± 1 %) molecules were similar for both
FCE samples, the abundance of CHOS molecules in FCE-S was elevated by more
than 10 %, predominantly at the expense of CHO molecules. The overall
abundance of sulfur in the FCE was nearly 4-fold when compared with that
of the OKA and PAN samples (Table 3), leading to a significant difference in
composition between the three sites as indicated by the PCA (Fig. 8a).
Van Krevelen diagrams of six wetlands
SPE-DOM: (a) OKA-L, (b) OKA-S, (c) PAN-L, (d) PAN-S, (e) FCE-L,
(f) FCE-S, obtained from negative electrospray 12 T FTICR mass spectra. Only
molecular assignments bearing combinations of C, –H, –O, –N, and –S atoms are
shown; color coded according to molecular series as follows: CHO, blue;
CHOS, green; CHNO, orange; CHNOS, red. Bubble areas reflect the relative
intensities of respective mass peaks. (f) Labels for CHOS compounds
correspond to key molecules, section (a): saturated sulfolipids; section (b):
unsaturated sulfolipids; section (c): common CHOS compounds in DOM, possibly
sulfonated carboxylic-rich alicyclic compounds (CRAM); d: aromatic black
sulfur.
Comparative analysis of van Krevelen diagrams derived from
negative electrospray 12 T FT-ICR mass spectra of all six wetlands SPE-DOM.
(a) Clustering diagram based on the similarity values between the spectra of
six wetlands SPE-DOM using Pearson correlation coefficient; (b) molecular
compositions common to all six wetlands SPE-DOM, (c) unique molecular
compositions common in FCE samples (FCE-L and FCE-S); (d) unique molecular
compositions with high abundance in both PAN samples; (e) unique molecular
compositions with high abundance in both OKA samples; (f) unique molecular
compositions common in all four PAN and OKA. The aromaticity index AI (Koch
and Dittmar, 2006) provided denotes single aromatic compounds for AI > 0.5 (bright blue triangle).
Comparative analysis of (left) H / C vs. m/z and (right) H / C vs. O / C van Krevelen diagrams derived from negative electrospray 12 T FTICR
mass spectra of the two Florida Coastal Everglades SPE-DOM FCE-S and FCE-L
(see also Fig. 8). (a) Molecular compositions with high abundance in Florida
Coastal Everglades SPE-DOM FCE-S; section (a): oxygen-deficient (poly)aromatic
black sulfur; CHNOS: suite of highly oxygenated CHNOS molecules; section (b):
common CHOS molecules in DOM; section (c): saturated sulfolipids. The
aromaticity index AI (Koch and Dittmar, 2006) provided in the upper right
van Krevelen diagram denotes single aromatic compounds for AI > 0.5 (bright blue triangle)
and polyaromatic compounds for AI > 0.67 (bright purple triangle); (b) Molecular compositions with high
abundance in Florida Coastal Everglades SPE-DOM FCE-L; section (d): a distinct
set of oxygen-rich aromatic CHOS compounds, likely associated with
ether-linked aromatic units; see main text.
CHOS compounds observed in all wetland samples already showed a remarkable
chemical diversity (Fig. 8b). However, the chemical dissimilarity of CHOS
compounds common to both FCE samples remarkably exceeded that found in OKA
and PAN, covering a substantial share of the CHOS chemical space from O / C ratio: 0.3–0.8 and H / C ratio 0.6–1.7, respectively (Fig. 8c). Here, four
groups of CHOS molecules were differentiated based on their positioning in
H / C against O / C van Krevelen diagrams (Fig. 7f): (a) saturated sulfolipids
with H / C ratio > 2 and intermediate O / C ratio, suggesting the
presence of sulfur in elevated oxidation states; (b) unsaturated
sulfolipids with a rather restricted H / C and O / C ratio; (c) a very large and
expansive set of molecularly diverse CHOS molecules with a bandwidth of O / C ratios
similar to CHO compounds but reaching out to higher saturation
(larger H / C ratio) than the latter (Fig. 7e–f); (d) unique to FCE-S
(with traces in OKA-L) was a large set of aromatic and oxygen-deficient
“black sulfur” compounds (DBS; Fig. 7f; section (a) in Fig. 9a) similarly positioned
like CHOS compounds in Atlantic open-ocean abyssopelagic SPE-DOM at 5446 m
depth (Fig. S8 in Hertkorn et al., 2013), but covering a larger mass range
(Fig. S7f). Sections (b) and (c) CHOS compounds were also observed in PAN and
OKA, whereas black sulfur compounds were rare in OKA-L (Fig. S8) and
virtually absent in the other samples except FCE-S. DOM-type CHOS compounds
common to all six wetland samples were on average more saturated and
oxygenated than their respective CHO and CHNO counterparts, suggesting also
here the presence of sulfur in elevated oxidation states (Fig. 8b).
The CHOS compounds of both FCE samples not only differed fundamentally from
those found in OKA and PAN, but were also remarkably diverse in both FCE-L
and FCE-S samples itself. Figure 9 indicates CHOS compounds present with
elevated abundance in either FCE-S (Fig. 9a) or FCE-L (Fig. 9b). The most
peculiar feature of FCE-S was a hydrogen-deficient pool of (poly)aromatic
CHOS compounds (section (a) mass peaks; Fig. 9a) in extended molecular series
with limited degree of oxidation (O / C ratio < 0.22), ranging from
m/z ∼ 300–600. The positioning in both van Krevelen and mass-edited
H / C ratio diagrams (Fig. 9) was in accordance with that of “black sulfur” in abyssopelagic South Atlantic SPE-DOM (Fig. S8 and Fig. 16 in Hertkorn et al., 2013), but its signature was more
conspicuous and showed larger richness of diverse CHOS compounds in FCE-S.
While sulfur can be readily inserted into any C–C and C–H bond, analogous
to oxygen, organic sulfur can also occupy oxidation states ranging from -2
to +6, an option not available to oxygen. Nevertheless, the manifest
oxygen deficiency of the proposed highly unsaturated CHOS molecules (section (d); Fig. 7f) suggests the presence of reduced sulfur in the form of
sulfides. Aromatic CHOS molecules will then most likely occur as
benzothiophene derivatives, a chemical environment of sulfur largely
favored in mineral oils (Purcell et al., 2007; Liu et al., 2010; Muller et
al., 2012). While both black carbon as well as black nitrogen (Wagner et
al., 2015b) have been reported in the FCE (Ding et al., 2014a, b; Maie
et al., 2006), the presence of this “black sulfur” was not previously
observed at FCE. The environmental factors driving the high abundance of
these compounds at FCE-S remain unclear but may be related to the higher
fire frequency at short-hydroperiod sites and possibly soil charring. A
small set of CHOS compounds with more average H / C and O / C ratios
(section (b)
mass peaks; Fig. 9a) was accompanied by a rather minor set of highly
oxygenated CHNOS compounds, with an O / C ratio > 0.75 (Fig. 9a).
In contrast to the FCE-S, the FCE-L sample displayed an oxygenated set (O / C
ratio > 0.6) of a few dozen hydrogen-deficient (H / C ratio < 1.1) CHOS molecules in truncated molecular series and at rather
low mass (m/z < 400; Fig. 9b section (d) mass peaks). These molecules
were most likely composed of oxygenated aromatics connected by (some) ether
bridges, which rather likely originate ultimately from plant and/or algal
polyphenols. Apart from PAH-derived compounds, which are commonly rather
oxygen deficient (O / C ratio < 0.3), these structures represent one
of the most plausible motifs of very hydrogen-deficient molecules found in
DOM (H / C ratio < 1). The large extent of average oxygenation makes
sulfur functional groups in elevated oxidation states, e.g., sulfones,
sulfonates or sulfates, likely candidates for this group of CHOS compounds.
In addition, a rather expansive cloud of abundant CHOS and less common CHNOS
compounds at mass range m/z 200–550, with large and variable extent of
oxygenation (O / C ratio: 0.4–0.95) was prominent in FCE-L and near-absent
in FCE-S. The sizable expansion of this cloud with a huge range of H / C ratios testified to a rather large overall diversity of these unique CHOS
molecules found solely in FCE-L; CHNOS compounds seemed to follow suit but
with a lesser overall diversity: highly oxygenated (O / C ratio > 0.8) and hydrogen-rich CHNOS molecules (H / C > 1.6) were missing
even if every added nitrogen carried one intrinsic hydrogen into analogous
CHO molecular formulas. This higher molecular diversity for the FCE-L site
may be driven by higher soil-derived (peat soils) DOM contributions at this
site compared to FCE-S (marl soils) (Chen et al., 2013) and a higher degree
of DOM preservation at this deeper, less photo-exposed site.
It has to be mentioned that this cloud encircled the common molecular series
of several hundreds of CHOS and CHNOS compounds found in both FCE samples
(Figs. 7e–f and 8c). The significantly higher presence of sulfur-containing
molecular formulas for the FCE samples is likely the result of higher inputs
of sulfate to the Everglades compared to the Pantanal and Okavango.
Firstly, Everglades is a coastal wetland where sea spray may be an important
contributor to sulfate. In addition, it is the most anthropogenically
impacted wetland of the three being compared, where runoff from agricultural
lands within the Everglades watershed is likely the most important
contributing factor to the sulfur load of the system as it is an ingredient
of fertilizer applications. The CHO and CHNO components specific for the OKA
and PAN samples are shown in Fig. 8f and suggest, in agreement with the
NMR data, a higher degree of oxidized, H-deficient materials at these sites
compared to the FCE. This is particularly true for the PAN which show unique
molecular formulas for oxidized, H-deficient CHO and CHNO components (Fig. 8d),
whereas molecular formulas unique for the OKA are relatively few (Fig. 8e).
Comparative analyses of van Krevelen diagrams between the six sites as shown
in Fig. 8a clearly cluster the FCE samples separately from the OKA and PAN.
Cluster analysis showed a clear distinction between the FCE on the one hand and
the OKA and PAN samples on the other hand, with less pronounced but
significant differences between the paired, long- and short-hydroperiod
samples at each site (Fig. 8a). Among pairs of DOM samples, similarity
according to FTICR/MS-based cluster analysis was in the order PAN > OKA > FCE (Fig. 8a),
whereas one-dimensional 1H NMR spectra clustered according to increasing
dissimilarity in the order OKA < PAN < FCE (Fig. S2). This discrepancy is readily explained
by the different recognition of aliphatic groups in FTMS (insensible) and
NMR spectra (quantitative depiction). CHO and CHNO molecules ionized by
negative ESI occupied rather similar expansive regions with near-average H / C and O / C elemental ratios (Fig. 8b). This is a common feature of DOM
molecular distribution as derived from FTICR mass spectra. Here, the largest
number of feasible and chemically reasonable isomeric molecules will project
on single mass peaks at average H / C and O / C elemental ratios, contributing
to larger overall mass peak amplitude – this applying even more
specifically to van Krevelen diagrams, in which different molecular
compositions with identical elemental ratios contribute to the same data
points (Hertkorn et al., 2007; Lechtenfeld et al., 2014). Analogously, the
distribution of CHO, CHOS, CHNO and CHNOS molecular series roughly
coincided, with some displacement of CHOS molecules in both FCE samples,
towards higher H / C ratio (i.e., higher aliphatic character).
At first glance, the H / C vs. O / C (Fig. 7) as well as the H / C vs. m/z (Fig. S7)
plots showed near-uniform fingerprints for OKA and PAN, covering larger
areas in the van Krevelen diagrams in the case of CHO compared with CHNO
compounds (Fig. 8b), suggesting an increased overall chemical diversity of
CHO compounds. In addition, the paired wetland samples clustered separately
for the high and low hydroperiods respectively, suggesting that molecular
compositions differ among sites with different hydrology. The weighted average O / C and H / C values were remarkably similar for PAN and OKA showing
rather marginal variance between different sites or between high and low
hydroperiod (Fig. 7a–d; Table 3). In comparison, FCE-S showed a
considerably decreased O / C ratio. While computed O / C ratios of wetland DOM
exceeded those found in oceanic DOM by about 0.2 units (Table 3; Table 4 in
Hertkorn et al., 2013), the H / C ratio of wetland DOM was approximately 0.15
units higher in comparison. Even if ionization selectivity in negative ESI
FTICR mass spectra applied, the 1H NMR section integrals indicate
analogous trends of relative saturation, or alternatively, hydrogen
deficiency between wetland and marine DOM. In comparison with average
wetland SPE-DOM, average open-ocean SPE-DOM showed lesser abundance of
aromatics (by 2–3 %), lower proportions of OCH
chemical environments (by 8 %) and, especially, higher abundance of pure
aliphatics (i.e., CCCH units; by 12 %). This implies
that marine DOM shows lower abundance of hydrogen-deficient (unsaturated)
and higher abundance of hydrogen-rich (purely aliphatic) molecules than
wetland DOM, in line with the elevated H / C ratio as derived from FTICR mass
spectra. Similarly, the higher abundance of oxygen-rich OCH
chemical environments in wetland DOM as seen by 1H NMR
section integral (8 % relative increase) was in accordance with the
increased O / C ratio found in their FTICR mass spectra (an increase of 0.2
units).
Comparative analysis of van Krevelen diagrams (Fig. S8) obtained solely from
the four PAN and OKA samples confirmed the previously observed higher
similarity between PAN-L and PAN-S compared to OKA-L and OKA-S sample pairs
(Figs. 7 and 8d–f). Molecular compositions with unique high
abundance when derived from all six wetland samples were sparse and
non-significant in the case of OKA (Fig. 8e; see also Fig. S8b), whereas
molecular compositions with unique high abundance in all four OKA and PAN
samples occupied a rather dense, contiguous section of hydrogen-deficient
(H / C ratio < 1) and oxygenated (O / C ratio ∼ 0.3–0.7) CHO and CHNO molecules (Fig. 8f).
In agreement with its high degree of
photo-oxidation, the OKA contained higher proportions of highly oxygenated
CHO (O / C ratio > 0.7) and CHNO (O / C ratio > 0.5)
molecules, and a few rather abundant (and easily ionizable) sulfolipids
(Fig. S8b), whereas PAN SPE-DOM displayed larger proportions of
hydrogen-deficient CHO molecules of considerable chemical diversity and
extent of oxygenation (O / C ratio ∼ 0.2–0.9), and several
dozens of CHNO molecules similarly positioned but with more limited range of
oxygenation and, hence, overall chemical diversity (Fig. 8d).
Remarkably, with the exception of a tiny section of CHO molecules (H / C ∼ 1.1; O / C ∼ 0.4), both CHO and CHNO molecular
series for OKA and PAN nearly perfectly superimpose in the H / C against O / C van Krevelen diagram (Fig. 8f). It is very likely that these CHO molecules
jointly present in PAN + OKA mainly represent oxygenated aromatic
molecules, possibly connected by ether linkages. This is one of the most
comprehensive ways to envision such hydrogen-deficient molecules of
conceivable natural product origin, and in agreement with the NMR data
suggesting a higher degree of oxidized, H-deficient materials at these sites
compared to the FCE. This is particularly true for the PAN which shows
unique molecular formulas for oxidized, H-deficient CHO and CHNO components,
whereas molecular formulas unique for the OKA are relatively few (Figs. 8e–f and S8b–c).