Introduction
Isoprenoid and branched glycerol dialkyl glycerol tetraethers (GDGTs) are
principal constituents of the prokaryotic cell membrane (Pearson and
Ingalls, 2013; Schouten et al., 2013, and references therein). Differences in
the GDGT core structures are crucial for distinguishing between the archaeal and
bacterial origins of these components, with isoprenoid alkyl chains and a
2,3-di-O-alkyl-sn-glycerol stereoconfiguration being specific to archaea and
branched alkyl chains and a 1,2-di-O-alkyl-sn-glycerol stereoconfiguration
to bacteria (Weijers et al., 2006a). Both types of tetraether lipids have a
high potential to be preserved in the sediment record (Schouten et al.,
2013) and have been reported in abundance from terrestrial and marine
environments, e.g. in the water column and sediments of oceans and lakes
(Hopmans et al., 2000, 2004; Schouten et al., 2012; Tierney and Russel,
2009; Zink et al., 2010; Naeher et al., 2014), in ponds (Tierney et al.,
2012; Loomis et al., 2014; Huguet et al., 2015), in hot springs (Pearson et
al., 2004; Reigstad et al., 2008; Pitcher et al., 2009), in geothermally
heated soils (Peterse et al., 2009a), in peat bogs (Sinninghe Damsté et
al., 2000; Weijers et al., 2006a, 2010), in grassland soils (Weijers et al.,
2007, 2010; Naeher et al., 2014), in forest soils (Hopmans et al., 2004;
Weijers et al., 2007, 2010), in permafrost soils (Peterse et al., 2009b;
Bischoff et al., 2014), in loess soils (Huguet et al., 2012), in podzols
(Huguet et al., 2010), in garden and agricultural soils (Leininger et al.,
2006; Weijers et al., 2010; Sinninghe Damsté et al., 2012) as well as in
paddy soils (Bannert et al., 2011; Ayari et al., 2013).
It is well known that archaea are involved in biogeochemically important
processes, including methanogenesis, anaerobic methane oxidation (AMO) and
aerobic ammonia oxidation (Boetius et al., 2000; Leininger et al., 2006;
Thauer et al., 2008; Stahl and de la Torre, 2012; Offre et al., 2013).
Distributions of isoprenoid GDGTs (iGDGTs) were initially used to
characterize archaeal communities in marine environments with two major
groups of archaea being distinguished: Thaumarchaeota (formerly recognized as mesophilic
Crenarchaeota) and Euryarchaeota (see Pearson and Ingalls, 2013, and reference therein).
Ammonia-oxidizing members of the Thaumarchaeota are currently the only known biological
sources of crenarchaeol, a GDGT structure that contains four cyclopentane
ring systems and an additional cyclohexane ring moiety (Sinninghe Damsté
et al., 2002). In addition, Thaumarchaeota contain varying amounts of GDGTs with 0 to 4
cyclopentane rings (Sinninghe Damsté et al., 2012; Schouten et al., 2013;
Pearson and Ingalls, 2013).
GDGT-0 is another common tetraether lipid that is present in a majority of
archaea (Pearson and Ingalls, 2013; Schouten et al., 2013, and references
therein; Villanueva et al., 2014), including for example mesophilic
methanogens (Koga et al., 1998; Koga and Morii, 2005; Villanueva et al.,
2014; Bauersachs et al., 2015). In addition, the presence of high abundances
of GDGT-0 at sites with active AMO suggests a close relationship between
microbial consortia involved in the production and consumption of methane
(Pancost et al., 2001; Blumenberg et al., 2004; Schouten et al., 2013). In
periodically flooded soils (paddy soils), methanogenic lineages, such as
Methanosarcinales, Methanocellales, Methanobacteriales
and Methanomicrobiales, were found (Liesack et al., 2000; Watanabe et al., 2006, 2013) at varying abundances in continuously flooded as well as in
alternatingly flooded and dried paddy fields (Watanabe et al., 2013). The distribution of
methanogens in soils has not yet been extensively studied by using the
GDGT-0 vs. crenarchaeol ratio. However, this ratio in conjunction with
stable isotope analysis has been applied successfully in soils, sediments
and the water column of Lake Rotsee (Naeher et al., 2014) to identify
methanogenic conditions. Likewise, Ayari et al. (2013) have shown that in a
rice field, where samples were collected before and after flooding, the
ratio of GDGT-0 / crenarchaeol increased upon flooding, when methanogenic
conditions had been established.
iGDGTs with multiple cyclopentane rings have been reported from anaerobic
methanotrophic archaea (ANME) of the ANME-1 cluster as well as
Thaumarchaeota and extremophilic Euryarchaeota and
Crenarchaeota (Blumenberg et al., 2004; Pearson and Ingalls, 2013;
Schouten et al., 2013, and references therein). The presence of iGDGTs has
been predominantly investigated in marine, limnic and other aquatic habitats,
but they have also been reported from soils. Here, the specific environmental
conditions controlling their distribution are less well studied (Weijers et
al., 2006b; Leininger et al., 2006; Sinninghe Damsté et al., 2012; Ayari
et al., 2013). An improved knowledge of environmental factors influencing
iGDGT compositions has been gained from cultivation experiments, which
demonstrated that growth temperature, pH and oxygen content affect GDGT
synthesis (Wuchter et al., 2004; Elling et al., 2015; Qin et al., 2015).
Probably the most commonly used archaeal-based proxy in marine systems is the
TEX86 (tetraether index of Thaumarchaeota-derived tetraethers
consisting of 86 carbons), which correlates well with surface water
temperatures (Schouten et al., 2002). Culture experiments revealed the effect
of increasing temperature to raise the number of cyclopentane rings (Schouten
et al., 2013, and references therein). Regional studies on altitudinal
mountain transects confirmed a dependency of the iGDGT cyclization on
temperatures in soil systems (Liu et al., 2013; Coffinet et al., 2014; Yang
et al., 2016), but additional factors, e.g., pH or soil moisture, may
influence the archaeal community and therefore the lipid composition found in
soils as well (Wang et al., 2013; Xie et al., 2015).
High abundances of branched GDGT (brGDGTs) have previously been reported from
soils worldwide (Weijers et al., 2007, 2010; Peterse et al., 2009a; Huguet et
al., 2010, 2012). Information on the biological sources of these components,
however, is still very limited (Hopmans et al., 2004; Weijers et al., 2007,
2010). Molecular investigations in peat bogs demonstrated that brGDGTs
occurred in highest concentrations in the catotelm, the bottom layer of peats
(Weijers et al., 2006a, 2010). This was used to infer anaerobic and
acid-tolerant bacterial species as brGDGT sources, e.g. microbes belonging to
Acidobacteria, the most abundant bacteria in this environment
(Weijers et al., 2006a, 2009, 2010). This is supported by the presence of a
tetra-methylated brGDGT that was recently identified in two cultured
acidobacterial strains (Sinninghe Damsté et al., 2011). In addition,
ether-bound 5-methyl iso-diabolic acid was detected in four
mesophilic species of subdivision 4 of the Acidobacteria as a
potential breakdown product of penta-methylated brGDGT (Sinninghe Damsté
et al., 2014). Soil bacteria producing brGDGTs have been proposed to be
obligate anaerobes following a heterotrophic mode of life (Oppermann et al.,
2010; Weijers et al., 2006a, 2010). The presence of brGDGTs in oxic soils
infers aerobically living bacteria to produce these lipids, but anaerobic
bacteria residing in anoxic microhabitats may be possible sources as well
(Schouten et al., 2013). The distribution of brGDGTs in soils is related to
growth temperature (mean annual air and soil temperature) and soil pH
(Schouten et al., 2002; Weijers et al., 2007, 2009; Peterse et al., 2009a,
2012). Indices which denote the degree of methylation and cyclization of
brGDGTs, the methylation of branched tetraethers (MBTs) ratio and the cyclization ratio of branched tetraethers (CBT) indices,
have previously been employed to reconstruct mean annual air temperatures
(MATs) using a global soil calibration (Weijers et al., 2009). More recently,
Peterse et al. (2012) defined the modified methylation index of branched
tetraether (MBT′), which represents the ratio of tetra-methylated brGDGT
(GDGT-Ia, Ib and Ic) to the seven most abundant brGDGTs (GDGT-Ia, Ib, Ic,
IIa, IIb, IIc and IIIa).
However, factors other than temperature and pH also seem to affect the
distribution of brGDGTs in natural ecosystems. For example, the relatively
broad scatter of calculated MAT in arid soils (Peterse et al., 2012) as well
as values deviating from the trend in the highest elevations of a transect
sampled on Mt. Kilimanjaro (Sinninghe Damsté et al., 2008) have been
interpreted as indicating an influence of water content and vegetation type
on the brGDGT pool. In addition, several authors noted that changes in the
distribution of brGDGT are strongly correlated with MAT on local scales as,
for example, in altitudinal transects of Mt. Rungwe and Mt. Gongga (Peterse
et al., 2009c; Coffinet et al., 2014). In agricultural soils from the same
area, the type of soil management and the vegetation cover can differ,
leading to variable soil water contents and soil temperatures (Liu et al.,
2014; Awe et al., 2015), which affect the local microbial community. Soil
microbes respond to environmental stress induced by, e.g., starvation, oxygen
limitation or acidification (Frostegård et al., 1993; Aanderud et al.,
2015). The last of these results
in the predominance of brGDGTs without cyclopentyl moieties in soils and
explains the dependency of soil pH and CBT (Weijers et al., 2007).
Besides pH, the redox potential (Eh) is an important factor that affects the
diversity and abundance of soil microorganisms. The Eh expresses the activity
of electrons which influence microbial metabolic reactions in soils. As
individual microorganisms are adapted to specific Eh conditions, an increase
in, e.g., soil moisture is accompanied by a decrease in Eh because of the
consumption of oxygen by microbes (Husson, 2013). Further parameters, which
regulate the Eh are temperature, organic matter content, or soil tillage, the
last of these modifying the soil structure and soil aeration (Husson, 2013,
and references therein). Agricultural management therefore may contribute to
controlling redoximorphic conditions. In contrast to upland soil, i.e.
without water flooding and associated crop plants, including corn or maize,
wheat, barley, rape, cassava, sugar cane, cotton, banana and various
vegetables, rice paddy soil management with
repeated puddling of the surface soil as well as frequent flooding and alternating
draining practices leads to a reduced Eh in the surface layer
(Kögel-Knabner et al., 2010; Kölbl et al., 2014). Prevailing anoxic
conditions are assumed to restrict the decomposition rate of organic matter
(Lal, 2002; Sahrawat, 2005), leading to high activities of methanogenic
archaea (Liesack et al., 2000) and, in combination with the application of
mineral fertilizer, to high denitrification rates producing nitrous oxide
(Xiong et al., 2007). In contrast, oxic conditions are associated with high
Eh, as in upland soil and in paddy soil after draining where ammonia
oxidation can occur. The last of these is either performed by
ammonia-oxidizing archaea (AOA) or bacteria (AOB) (Leininger et al., 2006)
depending on the soil pH, with AOA being more active in acidic soils and AOB
in alkaline soils (Jiang et al., 2015).
Rice serves as major staple food for 50 % of the world's population and
paddy rice cropland occupies an area of 157 million ha. This is equivalent
to 18 % of the agricultural land use area of the 10 major crops worldwide
and illustrates the importance of paddy agroecosystem utilization (FAO,
2003). This profound anthropogenic influence on aquatic agroecosystems will
dictate their biogeochemical and geomicrobiological properties and processes,
which were determined from GDGT distribution and warrant further
investigation. Only limited information on microbial assemblages and their
activity in paddy soils is currently available (Bannert et al., 2011; Ayari
et al., 2013). The study of such agroecosystems is of particular interest for
both soil scientists and geochemists for similar reasons, as man-made
environmental constraints can be compared to natural ones. To identify the
anthropogenically induced ecosystem properties, reflected in microbial
community structures, we studied the tetraether lipid composition in soils of
different agricultural management systems, which developed in subtropical
(Italy, SW China) as well as in tropical (Indonesia, Philippines, Vietnam)
climates. In addition to the management type, including differences in
cropping style (upland crop plants vs. wetland rice), the intensity of the
management and the duration of utilization were distinctive criteria in the
investigation of effects on the microbial lipids in rice paddy soil
(periodically flooded), upland (non-flooded) and forest soils. This study
compares non-flooded and flooded agroecosystems of different agricultural
uses with respect to their GDGT composition (including GDGT palaeoproxies) to
widen our knowledge on the sources and properties of GDGTs in terrestrial
agroecosystems on local, regional and global scales.
Discussion
Distribution of isoprenoid GDGTs in soils
iGDGTs constitute between 0.9 and 25.7 % (35 % in Cixi soils) of all
GDGTs (Table 1), indicating substantial contributions of archaeal lipids to
most of the investigated soils. Forest and bushland soils had lowest relative
mean abundances of iGDGTs (5.8 ± 2.6 %), followed by tropical paddy
(9.3 ± 4.0 %) and upland soils (9.8 ± 6.0 %). The
proportion of iGDGTs was highest in Chinese and Italian upland soils
(21.1 ± 8.0 %) compared to their adjacent paddy soils and all other
remaining soils (13.3 ± 5.0 %). The fact that the iGDGT content was
significantly (p < 0.01; Mann–Whitney U test) lower in
tropical soils (including from the Philippines, Vietnam, Indonesia; n= 116) compared to subtropical soils (including from China and Italy; n= 51) suggests that the composition of the microbial consortia varies on
regional to global scales. In addition, the differentiation between upland
and paddy soils with higher amounts of iGDGTs in the former may indicate
management-induced (regulating the water regime, nutrient availability,
oxygen availability and/or redox conditions) variations of GDGT-containing
microorganism. In general, at locations with the same climate and substrate,
different management types best explain significantly different GDGT
distributions (p < 0.05; Mann–Whitney U test). Regardless of
whether paddy, upland or forest management, all soil types differ in their
microbial lipid pattern that may be influenced by differing inputs of plant
organic matter, differing fertilization practices and redox conditions. The
last of these is controlled by
flooding and draining practices on paddy soils, which seem to favour growth
and input of brGDGT-containing bacteria and/or the improved preservation of
fossil brGDGTs compared to the adjacently located aerated upland soils.
iGDGT distribution patterns described from cultured archaea (Koga et al.,
1998; Pancost et al., 2001; Blumenberg et al., 2004; Koga and Morii, 2005)
and their comparison with soils may provide insights into the archaeal
community structure and the biological processes that they mediate (Ayari et
al., 2013; Yang et al., 2016). The most abundant iGDGTs in our sample set are
GDGT-0 and crenarchaeol. The latter is considered a highly specific
biological marker for ammonia-oxidizing Thaumarchaeota (Leininger et
al., 2006; Pitcher et al., 2010; Sinninghe Damsté et al., 2012; Pearson
and Ingalls, 2013), which, in the form of groups 1.1a, 1.1b, 1.1c and 1.3,
have been reported to be present in soils worldwide (Pester et al., 2012;
Oton et al., 2016). Differences in the ammonia-oxidizing archaea community
composition of group 1.1b Thaumarchaeota in soils may be influenced
by climatic conditions, as demonstrated in soils of various geographical
origins (Pester et al., 2012). This dependency was not found for the relative
abundance of crenarchaeol in soils investigated here using the Mann–Whitney
U test. To date, molecular investigations on cultivated
Thaumarchaeota report GDGTs only for groups 1.1a and 1.1b (Pitcher
et al., 2010, 2011; Sinninghe Damsté et al., 2012). Sinninghe Damsté
et al. (2012) showed that group 1.1a Thaumarchaeota (marine and
other environments) and group 1.1b Thaumarchaeota (soils and other
environments) can be separated from each other based on the relative
abundance of the crenarchaeol regioisomer, with a proportion of the
crenarchaeol regioisomer < 5 % being indicative of group 1.1a
and > 10–20 % of group 1.1b Thaumarchaeota
(Sinninghe Damsté et al., 2012). The same authors observed higher
abundances of the crenarchaeol regioisomer in soils rather than in marine or
lacustrine environments (Sinninghe Damsté et al., 2012). Crenarchaeol and
its regioisomer are present in all analysed soil samples, which is in
agreement with a previous study (Weijers et al., 2006b). The amount of
crenarchaeol is generally higher in upland soils (46.4 ± 12.9 %, n= 37) compared to adjacent paddy soils (22.5 ± 14.5 %, n= 119; Fig. 2a), possibly suggesting management-induced differences in the
archaeal community structure. The abundance of the crenarchaeol regioisomer
varies from 3 to 21 % compared to that of crenarchaeol (mean value of
9 ± 4 %, n= 170) and shows no differences between soils and/or
management types (Fig. S2).
Box-plot diagrams of (a) crenarchaeol,
(b) GDGT-0, (c) GDGT-0 / crenarchaeol ratio and (d)
TEX86 in upland (NP, brown), paddy (P, blue), marsh (grey), forest
(For), bamboo cultivated (Bamb, red) and bushland (Bush, violet) soils.
Abbreviations refer to different sampling locations: Italy (IT), China (C),
Philippines (PH), Vietnam (VN), Sumatra (SUM) and Java (JAV). The vertical
line separates subtropical from tropical locations. Numbers in all plots
indicate samples listed in Table S1. The dashed line in (c) marks
the GDGT-0 / crenarchaeol value of 2 that is the boundary to higher
proportions of methanogens, which reveal values > 2. Note the
logarithmic scale for GDGT-0 / crenarchaeol ratios. Note different symbols
(circle or asterisk) for outliers that are more than 1.5 (or 3) box lengths
from one hinge of the box.
Angel et al. (2012) observed that methanogenic archaea are ubiquitous in
soils and are active only in anoxic, highly reducing environments, e.g. under
flooded conditions. One distinct feature of paddy soil management vs.
management of all other soils is the periodic flooding and draining of soils,
which leads to highly variable redox conditions throughout the course of a
year (Kögel-Knabner et al., 2010; Kölbl et al., 2014). Paddy soils
are known for high methanogenic activity and as significant sources of
atmospheric CH4 (Conrad, 2007; Thauer et al., 2008; Serrano-Silva et
al., 2014) with little changes in the methanogenic community structure
between flooding events (Krüger et al., 2005; Watanabe et al., 2006,
2009). In turn, this suggests that the overall lipid pool in paddies does not
change significantly after draining the fields for harvesting.
Despite GDGT-0 being a common component in many archaea, an elevated ratio
of GDGT-0 / crenarchaeol with a threshold > 2 has been used
previously to indicate a dominance of methanogenic archaea in a given
sedimentary environment. This notion primarily applied to for lake sediments,
where the threshold in GDGT-0 / crenarchaeol > 2 has been
attributed to methanogenesis occurring under anoxic and organic-matter-rich
conditions (Blaga et al., 2009; Naeher et al., 2014). Paddy soils are known
to release high amounts of methane during flooding periods (Thauer et al.,
2008). Therefore, Ayari et al. (2013) suggested that the 3- to 6-fold
increase in the GDGT-0 / crenarchaeol ratio, determined on the intact polar
lipid fraction, in paddy soils after flooding is associated with GDGT-0
synthesis by methanogenic Euryarchaeota. We adopted this assumption and
compared different kinds of soil management with respect to their iGDGT
composition. In the investigated soils, the GDGT-0 / crenarchaeol ratio ranged
from 0.1 to 121.6, with highest ratios observed in Philippine and Vietnamese
paddy soils (Fig. 2c, Table 1). In oxic upland and forest soils the mean
GDGT-0 / crenarchaeol ratio was ≤ 1, which indicates that methanogenic
archaea are only a minor component of the microbial community at these
sites. In addition, a few paddy soils (e.g. sites in Cixi and in
Italy) had GDGT-0 / crenarchaeol ratios comparable to those observed in upland
soils, which can be explained by the management form including higher
intensities of crop rotation with upland crops under non-flooded conditions
on these fields. However, if soils from the same region are compared, the
ratio was generally 3–27 times higher in soils which are under paddy
management compared to adjacent upland soils, indicating increased
abundances and activity of methanogens in flooded soils.
TEX86 values from all sites ranged from 0.3 to 0.9 (Fig. 2d, Table 1)
without an apparent geographical trend. However, TEX86 values were on
average 1.3 times higher in upland, bushland and forest soils compared to
the adjacent paddy soils within the same region. For example, the ratios of
upland and paddy soil TEX86 values were highest in the subtropical
locations of Cixi and Italy (∼ 1.5; Table 1). None or only
minor differences in TEX86 values were noted in the Jasinga and Ngawi
upland and paddy soils of Indonesia. Because of the relation between the
TEX86 and temperature, one explanation for this difference could be
that the periodic water layer on paddy soils may protect the soil surface
from excessive heating and therefore results in lower mean annual soil
temperatures (MSTs) in both soil types. Previous studies of altitudinal
mountain transects support this suggestion, as the soil TEX86 was
negatively correlated with elevation and therefore with decreasing
temperatures, e.g. in the Qinghai–Tibetan Plateau (r= -0.81, r2= 0.65,
p < 0.01; Liu et al., 2013) and Tanzania (r= -0.71, r2= 0.50,
p < 0.0001; Coffinet et al., 2014).
In the soils investigated here, the relative proportion of GDGT-3 and the
crenarchaeol regioisomer together with GDGT-1 mainly affected the
TEX86. Low TEX86 values, as observed in paddy soils, are the
result of high relative abundances of GDGT-1 and low proportions of GDGT-3.
This suggests that paddy soil characteristics such as alternating redox
conditions and higher water content control the presence of GDGT-1. High
contents of cyclopentyl moieties in archaeal membrane lipids are known to be
associated with ANME archaea, which synthesize
significant quantities of GDGT-1, GDGT-2 and GDGT-3 (Pancost et al., 2001;
Blumenberg et al., 2004). Interestingly, two divergent trends in the direction of
increased TEX86 values were observed for GDGT-2 (Fig. 3a), with an
increase in the GDGT-2 content to a TEX86 value of 0.70 and a
subsequent decrease if values exceed this threshold (Fig. 3a). This change
may again indicate that the archaeal community differs in dry upland or forest
soils and flooded soils.
Cross plots showing (a) the relative abundance
(% of the sum of GDGT-1, GDGT-2, GDGT-3 and crenarchaeol regioisomer) vs.
TEX86 and (b) the relationship between the most abundant
iGDGTs (GDGT-0 and crenarchaeol) and lower concentrated iGDGTs (GDGT-1, GDGT-2,
GDGT-3, and crenarchaeol regioisomer) as TEX86. GDGT-0 / crenarchaeol
> 2 and TEX86 < 0.6 are diagnostic for methanogens.
Two outliers from the Ifugao site (Philippines) with a GDGT-0 / crenarchaeol
ratio > 69 were excluded from the figure. Note the logarithmic
scale for GDGT-0 / crenarchaeol ratios. The filled circles in (a)
denote paddy soils and the non-filled circles denote upland, marsh, forest,
bamboo and bushland soils.
Fig. 3b shows that there is only a weak relationship between the relative
abundance of GDGT-0 and TEX86 (logarithmic r= -0.67,
r2= 0.45, p < 0.0001). However, both the TEX86
and the GDGT-0 / crenarchaeol ratio show clear differences in soils under
paddy (grey background in Fig. 3b) and upland management for adjacent sites, suggesting that a comparison of both parameters may allow distinguishing
anoxic or oxic conditions in soils. In general, paddy soils plotted within a
field characterized by GDGT-0 / crenarchaeol ratios > 2 and
TEX86 values < 0.6 (Fig. 3b), possibly denoting a
diagnostic area for the abundance of methanogenic archaea. The
GDGT-0 / crenarchaeol ratio also differs between the various paddy soils, with
exceptionally high ratios in the Philippine Ifugao and Vietnamese Lao Cai
soil (Table S1). At these sites, longer flooding periods (> 5
month per year) compared to Chinese and Indonesian soils are the likely
explanation for the high ratios.
Distribution of branched GDGTs in soils
In the soils investigated here, the relative proportion of brGDGTs to the
total GDGT pool was high and varied from 65.0 to 99.1 % (Table 1). Forest
soils generally contained the highest abundances of brGDGTs (> 92 %), while they were significantly lower in upland and paddy soils (Fig. 4a).
Pearson's correlation analysis indicated that the SOC content was not
related to the relative abundance of brGDGT (r= 0.22, r2= 0.05,
p < 0.01).
In general, the tetra-methylated GDGT-Ia was the most abundant brGDGT in
acidic soil and was the only brGDGT to increase in relative abundance with
decreasing pH (r= -0.75, r2= 0.56, p < 0.001; Fig. 5). All
other brGDGTs increased in relative abundance with pH (p < 0.001;
Table S2), with the highest correlations observed for GDGT Ib (r= 0.83,
r2= 0.69), GDGT IIb (r= 0.79, r2= 0.62) and GDGT IIIb
(r= 0.71, r2= 0.50). Our results thus suggest that the
monocyclization, in particular, of brGDGT is strongly controlled by pH (r= 0.86,
r2= 0.74,p < 0.001) with alkaline conditions favouring the synthesis
of brGDGT with one cyclopentane moiety (Fig. 5). Similar observations have
previously been made in a set of globally distributed upland soils (Weijers
et al., 2007; Peterse et al., 2012).
Box-plot diagrams of (a) the relative proportion of
brGDGT in the total GDGT pool and (b) the BIT index in soil. Note
different symbols (circle or asterisk) for outliers that are more than 1.5
(or 3) box lengths from one hinge of the box. Abbreviations and subdivisions
as in Fig. 2.
Relative abundance of brGDGT plotted vs. measured soil
pH. Note logarithmic scale for relative abundance. Dashed lines indicate
neutral soil conditions, which delimitate the interval between 6.6 and 7.3 pH
units.
Weijers et al. (2007) proposed the lower number of cyclopentyl moieties in
brGDGT as a protection mechanism of bacterial cell membranes within acidic
soils. The decrease in the amount of cyclopentyl moieties in brGDGT is
thought to be associated with a decrease in membrane permeability that
regulates the internal pH of bacteria under acidic conditions (Weijers et
al., 2007). In soils investigated here, the CBT ratio varied between -0.04
to 2.13 (Table 1) and showed a negative correlation with increasing soil pH
(r= -0.81, r2= 0.65, p < 0.001; Fig. 6a). In neutral to
alkaline soils (with pH values > 6.5) CBT values stayed rather
constant with an offset observed between paddy soils (mean 0.34) and upland
soils (mean -0.01; Fig. 6a). Wang et al. (2014) also found no apparent
correlation between pH and CBT in alkaline soils in a study of arid and
subhumid Chinese soils. However, a predominant dependency of CBT on soil
water content and the mean annual precipitation (MAP) was observed (Wang et
al., 2014). In our study, varying degrees of soil moisture may be one
possible explanation for the varying CBT values in paddy and upland soil,
especially under alkaline conditions (Fig. 6a).
Plot of (a) the cyclization ratio of branched tetraethers
(CBT) vs. soil pH and of (b) the modified methylation index of
branched tetraethers (MBT′) vs. soil pH. Dashed lines indicate neutral soil
conditions, which delimitate the interval between 6.6 and 7.3 pH units.
Regressions line of all soils is black, the line of upland, marsh, forest,
bamboo and bushland soils is brown and the line for paddy soils is blue.
Abbreviations as in Fig. 2. Red lines in (a) show the offset between
paddy and upland soil, which have > 6.2 pH values.
The degree of methylation of brGDGTs (MBT′) has previously been shown to
correlate with MAT and pH (Weijers et al., 2007; Peterse et al., 2012). Our
results demonstrate that the MBT′ generally shows low values in paddy soils
compared to the adjacently located upland soils, except for the Chinese
soils of Cixi (Table 1). The difference in MBT′ between soils from the same
sampling area denotes a lower influence of MAT on the MBT′ than on the pH,
which was weakly related to the MBT′ (r= -0.55, r2= 0.31, p < 0.001;
Fig. 6b). The MBT′ was mainly controlled by the relative abundance of
GDGT-Ia and GDGT-IIa, both of which were strongly related to MAP (Peterse et
al., 2012). As the latter is largely similar at adjacent sites, we consider
the paddy-soil-specific management techniques, including periodically
flooding the soils, as responsible for the low GDGT-Ia and high GDGT-IIa
content in paddy soils compared to upland soils (Table S1).The temperatures
inferred from brGDGT patterns, i.e. TMC values, were generally lower in
paddy soils compared to the adjacent upland soils (Table 1), suggesting that
TMC reflects mean annual soil temperature rather than air temperature.
Vegetation cover and soil moisture affect soil temperature, in particular in
surface soils (Seneviratne et al., 2010; Liu et al., 2014; Awe et al.,
2015). This led us to hypothesize that soil moisture and/or soil temperature
regulates composition of brGDGTs in adjacent subaquatic and upland soils of
identical air temperature as recognized by their respective TMC.
A recently developed method separates the structural isomers of brGDGTs from their methyl groups located at positions 5 and 6 (De Jonge et al., 2013). De
Jonge et al. (2014) showed that the new CBT5ME, calculated without
6-methyl brGDGTs, correlated more strongly with soil pH than the regular CBT,
which includes both isomers, the 5- and 6-methyl brGDGTs. In addition, these
authors found no correlation between pH and the newly developed
MBT5ME′, which is calculated without the 6-methyl isomer, but a stronger
correlation of this index with MAT. De Jonge et al. (2014) thus demonstrated
that co-elution of GDGTs can affect estimation of pH values. Conventional
methods, such as the one employed in this study, are not suited to fully
separate the different structural isomers of brGDGTs, and hence it is
possible that some scatter observed between our CBT-reconstructed and
measured pH may result from the analytical set-up (Fig. 6a). However, the
overall good covariation of CBT and pH for our sites suggests that the
partial co-elution of brGDGT only had a minor effect on the calculation of
the lipid-based proxies used in this study.
Influence of management systems on GDGT distributions
The BIT index quantifies the relationship between acyclic brGDGTs and
crenarchaeol and has been used previously to determine the input of
terrestrially derived organic matter to marine and lake environments
(Hopmans et al., 2004; Weijers et al., 2007). The interpretation of BIT
values in soil is not that straight forward as all GDGTs are terrestrially
derived. Thus, variations in BIT values must be governed by a microbial input
whose GDGT distribution is currently only incompletely known.
Wang et al. (2013) observed a positive correlation between increasing soil water content
and BIT values in Chinese marsh soils. In our sample set, the BIT index was
slightly higher in paddy soils than in the adjacent upland soils (Fig. 4b).
Furthermore, higher values were observed generally in paddy soils from
tropic (1.02–1.04-fold) compared to subtropic (1.07–1.11-fold) locations. In
contrast to the general trend, we found highest BIT values (1.27-fold) in
the subtropical paddy soils of the Cixi location. In this area, the
BIT values in marsh and upland soils (0.61–0.89) were comparatively low,
indicating that the latter have a mixed lipid composition, with crenarchaeol
originating predominantly from the residual parent substrate (tidal wetland
sediment) and in smaller quantities also from the current microbial soil
community. Similar results were found in a study of plant wax lipids, which
confirm the mixed organic matter composition in these soils
(Mueller-Niggemann and Schwark, 2015). Except for the higher contribution of
crenarchaeol to the marsh soils, our results show that brGDGTs clearly
dominate over iGDGTs originating from Thaumarchaeota in all of the investigated soil types.
Interestingly, based on relations of brGDGTs to crenarchaeol,
Thaumarchaeota seem to be more abundant in upland soils compared to forest and
periodically flooded paddy soils (Fig. 4b). Low redox conditions as assumed
for paddy soils may thus lead to an enrichment of brGDGTs either by higher
production or increased preservation of brGDGTs compared to crenarchaeol in
wetland soils. Our results thus contradict those of Peterse et al. (2015),
who performed a 152-day experimental study, where soils were incubated under
water to simulate the development of an aquatic environment under aerobic
conditions. Contrastingly to our observations, lower BIT values were
measured in flooded soils, potentially due to a higher contribution of
crenarchaeol while brGDGTs remained unchanged until the end of the
experiment.
Principal component analysis (PCA) based on standardized relative
abundances of six iGDGTs in 170 investigated soils. The first principal
component (PC1) accounted for 53.9 % of the total variance and the second
(PC2) for 29.9 %. Panel (a): symbols and colours in upper legend
denote different management forms. Abbreviations as in Fig. 2. Panel
(b): colours of the sample site symbols in lower legend are
indicative of the number of rice cultivation cycles per year (“No. of
rice yr-1”).
PCA was performed to obtain information on the major factors that control the
variability of the distribution of iGDGTs and brGDGTs. Results of this
analysis indicate that crenarchaeol exerts a major control on the iGDGT
composition in upland soils (Fig. 7a). The component loading score of GDGT-0
is opposite to crenarchaeol and has the highest negative score in PC1. In
general, soils can be sorted into two groups on the basis of their scores on
the first component. Paddy soils load negatively and all other soils load
positively on PC1. Paddy soils that plot in the quadrant of upland soils are
characterized by a higher intensity of crop rotation with upland crops on the
fields. The iGDGT composition of periodically flooded paddy soils is mainly
controlled by GDGT-0 and that of non-paddy upland soils by crenarchaeol
derived from Thaumarchaeota. In flooded rice paddy soils, oxygen
availability determines the development of microbial consortia adapted to
more anoxic conditions such as GDGT-0-synthesizing methanogenic archaea (Koga
et al., 1998; Koga and Morii, 2005). The variance in PC2 is mainly associated
with the relative abundance of GDGT-2 and separating forest and bushland
soils from all other soils. The larger scatter of paddy soils on PC2 is
explained by the number of rice cultivation cycles per year, which apparently
influence the GDGT-2 content significantly (Fig. 7b). Methanogenic archaea
were found to be phylogenetically related to ANME archaea (Krüger et al.,
2003; Shima et al., 2012). ANME archaea are a well-known source of iGDGTs
(including GDGT-2) in natural environments (Pancost et al., 2001; Blumenberg
et al., 2004). Both the interaction of methanogenic and methanotrophic
archaea as well as the fact that ANME are an abundant source of GDGT-2 could
explain the relationship between higher numbers of rice cultivation cycles,
which induce increased methanogenesis through abundant redox cycling, and the
presence of GDGT-2. MAT and MAP had no obvious influence on discriminating
between different types of agricultural soil via iGDGT distribution
(Fig. S3).
Principal component analysis (PCA) based on standardized relative
abundances of nine brGDGTs in 170 investigated soils. The first principal
component (PC1) accounts for 69.1 % of the variance and the second (PC2)
for 14.3 %. Panel (a): symbols and colours in upper legend
denote different management forms. Abbreviations as in Fig. 2. Panel
(b): colours of the sample site symbols in lower legend are
indicative of the mean annual precipitation.
Principal component analysis (PCA) based on commonly used indices
and ratios for the 170 investigated soils. The first principal component
(PC1) accounts for 33.5 % of the variance and the second (PC2) for
21.4 %. Panel (a): symbols and colours in upper legend denote different
management forms. Abbreviations as in Fig. 2. Panel (b): colours of
the sample site symbols in lower legend are indicative of the number of rice
cultivation cycles per year (“No. of rice yr-1”).
PCA analysis of the relative abundances of brGDGT shows an opposite relation
of GDGT-Ia to all other brGDGTs, with the highest component loading score on
PC1 for GDGT-Ia (Fig. 8). The cyclopentane-ring-containing GDGT-IIb and
GDGT-IIIb plot negatively on PC1. Higher contents of GDGT-Ia in upland soils
compared to adjacent paddy soils (Table S1) confirm that tetra-methylated
brGDGTs may be useful in separating different agricultural soils. GDGT-IIa
has the lowest loading score on PC1 but the highest on PC2. Upland soils load
separately from paddy soils along the PC2, with a variation of relative
abundance of the cyclic GDGT-Ib and GDGT-Ic playing the most important role.
In contrast, paddy soils are mainly influenced by the abundance of GDGT-IIa
and GDGT-IIIa, which both show only a low correlation with pH (Table S2). We
rather assume that they are dependent on soil moisture, due to the lack of
correlation between the GDGT distribution and soil properties (e.g. pH) as
well as climate factors (e.g. precipitation, air temperature) in adjacently
located paddy and upland soils. The main ecological difference between paddy
and upland soil is the water budget, and thus we interpret this environmental
variable as causing the offset in GDGTs. The first PC, explaining 69.11 %
of the variance, indicates a separation between locations, with a strong
negative score in subtropical Italian and Chinese soils and more positive
scores in soils originating from the tropics (Fig. 8a). The MAP (Fig. 8b) and
MAT (Fig. S4) gradients of sampling locations on PC1 confirm a relation of
climatic parameters to the variation of acyclic brGDGTs.
PCA analysis of environmental parameters as well as of indices of bacterial
and archaeal GDGTs indicated that separation of paddy and upland soil is
mainly controlled by the intensity of methanogenesis (Fig. 9a). The
GDGT-0 / crenarchaeol ratio and the BIT index had the highest positive
loading score on PC2. The SOC and TN loaded in the same quadrant as the BIT
index, suggesting that a positive correlation between the amount of organic
matter and acyclic brGDGT, especially in paddy soils, prevailed. Alternating
anoxic conditions in paddy soils are known to favour the preservation and
therefore the accumulation of organic matter (Lal et al., 2002), which could
lead to an increase in heterotrophic and brGDGT-producing bacteria. In
general, the CBT loading was the opposite of that of the soil pH on PC1,
indicating their negative relation to each other. The internal separation of
paddy soils via the number of rice cultivation cycles is evident by high
loading scores of the CBT and MBT′ (Fig. 9b). Apparently, the increase in
the MBT′ is linked to the number of rice cycles and therefore to the
lowering of penta- and hexa-methylated brGDGT during increasing redox cycles.
Similar loading scores as well as similar directions of climatic parameters,
such as MAP and MAT, and of CBT and MBT′ also indicated that these factors
were linked to each other. In addition to methanogenesis, differences in MAT
and soil water content seemed to be secondary factors controlling the
distribution of brGDGT in soils, which also allowed a separation between
upland and paddy management. It should be considered though that MAT is not
identical to MST as the latter was also affected by, e.g., the albedo and
soil management, which can be different in the adjacent soils (Liu et al.,
2014; Awe et al., 2015, and references therein). The reflection coefficient
of the surface differs in agricultural soils as a consequence of management
practices, which influence the soil bulk density (via tillage), the plant
cover (function of the crop leaf area index) and the soil water content. For
example, Awe et al. (2015) found differences in soil temperature as a
consequence of management practices, with lower temperatures in soils under
chiselling and conventional tillage compared to no tillage.
Time plots of various GDGT ratios and indices in soils of
the Chinese Cixi region: (a) ratio of branched vs. isoprenoid GDGTs,
(b) the TEX86, (c) the CBT and (d) MBT′.
Note logarithmic scale for the cultivation time. Numbers in plot
(c) reflect soil pH values.
Effects of long-term management on GDGT distributions
Changes in GDGT distribution within two Cixi chronosequences with different
cropping systems, one under continuous non-flooded upland and the other under
paddy management, indicated specific adaption processes during the long-term
usage at each site. Marsh soils were the first soils to develop after the
construction of dykes on tidal wetland sediments and therefore represent the
starting point of the subsequent soil development. We observed high BIT
values (∼ 0.77) already in the surface horizon of the marsh soils,
indicating the rapid adaption of the microbial community to more terrestrial
conditions. A plot of the brGDGT / iGDGT ratio over time provides
evidence for a dominance of brGDGT over iGDGT in all soils, with values of
this ratio varying between 2 and 6 in upland soils (Fig. 10a). In contrast to
paddy soils, which had a 4-fold increase in the ratio after 2000 years of
rice cultivation, such an increase suggests an influence of long-term
processes on the proportion of archaeal and bacterial soil microorganisms.
These processes may include desalinization, decalcification through leaching
as shown in changes in soil pH values (Fig. S5a), fertilization activities,
and organic matter input and accumulation (Fig. S5b). Paddy soil management
is known to strongly affect the accumulation of organic matter (Wu, 2011;
Mueller-Niggemann et al., 2012; Kölbl et al., 2014) as the periodically
anaerobic conditions result in a slower degradation of organic matter (Lal et
al., 2002). Kölbl et al. (2014) investigated the response of redox
dynamics to changing water conditions over a 1-year time period in 100-, 700-
and 2000-year-old paddy soils. They noted a change in the redox potential
towards anoxic conditions, already after 5 days of flooding. After
stabilization, the redox potential was in the same range in all soils (-170
to -200 mV), independent of the duration of paddy management. In upland
soils, permanent oxic conditions were persistent throughout the time period
investigated. The results of Kölbl et al. (2014) demonstrate that the
rapid establishment of anoxic conditions and the long-term usage of paddy
soils may lead to an increase in organic carbon concentrations over time.
Time plot of MBT′–CBT-derived temperatures (TMC) in
soils of Cixi. Note logarithmic scale for cultivation time.
Within the upland soil chronosequence, the TEX86 does not change
significantly over the 700-year cultivation time and averages 0.7 (Fig. 10b).
In paddy soils, by contrast, the TEX86 decreased from the initial
marsh soil value of 0.7 to values of 0.3 within only 50 years of paddy
management. Rotation between paddy- and upland-type cultivation resulted
in a comparatively high TEX86 value of 0.5 in the 2000-year-old paddy
soils (Fig. 10b). Our results thus suggest that management systems
significantly affect the microbial soil community. Long-time paddy
management also led to the successive increase in ammonia-oxidizing
Thaumarchaeota based on high relative abundances of crenarchaeol, indicating either a
recovery process of water-stressed soil Thaumarchaeota or the enrichment of fossil
crenarchaeol. The latter is potentially explained by the management type
used in the Cixi area, with one wetland rice season and one dry intercrop
season per year that influence the presence of aerobic and anaerobic
microbes in these paddy soils. In particular, the periodically anaerobic
conditions may result in a slower degradation of organic matter (Lal et al.,
2002). GDGTs may originate from a mixed source of microbial membrane lipids
that were recently deposited (during the oxic as well as in the anoxic
period) additionally to the previously preserved ones. Thus, higher
proportions of crenarchaeol, e.g., as marker for terrestrial ammonia
oxidizers, being active during the oxic intercrop period, were detected but
in lower amounts than commonly observed in upland soils (Table S1). At the
same time, the proportion of methanogenic archaea, which was estimated by
using the GDGT-0 / crenarchaeol ratio, decreased during the long-term paddy
management from 5.0 in the 50-year-old to 2.8 in the 2000-year-old paddy soil.
The pH values ranged between 8.0 in marsh soil and 5.5 in the 2000-year paddy
soil. The paddy management (including flooding practices) thus leads to
enhanced decalcification of soils compared to the non-flooded upland
management. However, most soils have an alkaline or neutral pH with
exception of the 700-year upland soil and the 2000-year paddy soils, which all
had pH values < 6.5 (Fig. S5a). It has previously been demonstrated
that the CBT is negatively correlated with increasing pH values (Weijers et
al., 2007; Peterse et al., 2012). In the soils of the Cixi chronosequences a
negative correlation was also observed, which was higher for paddy soils
(r= -0.94, r2= 0.88, n= 4, p < 0.001) than for upland
soils (r= -0.69, r2= 0.47, n= 5, p < 0.001).
Interestingly, an offset of CBT values between paddy and upland soils with
no apparent changes during cultivation time was noted (Fig. 10c). In
addition, the CBT was higher in the younger of both marsh soils, probably
because of the greater soil water content in the ∼ 10-year-old
compared to the ∼ 35-year-old marsh soil as a result of the
progressive dewatering during marsh soil pedogenesis. The observation
regarding the CBT values supports the idea that soil moisture, in addition to
pH, controls the degree of cyclization of brGDGTs under alkaline conditions, possibly as a reaction to water stress or oxygen deprivation in
microorganisms. The increase in CBT values in acidic soils (Fig. 10c) also
suggests that low soil pH results in the increased synthesis of brGDGTs with
no cyclopentyl moieties.
Except for the youngest paddy soils (50 years), the MBT′ was slightly lower in
Cixi upland soils compared to their corresponding paddy soils with identical
cultivation time (Fig. 10d). This is in contrast to the observations that
paddy soils in general showed a lower MBT′ compared to the adjacent upland
soils (Fig. 6b). This may indicate that soil bacteria living under
contrasting pH regimes adapt the composition of their membrane lipids in a
different fashion, even if the agricultural management is comparable.
The CBT and MBT′ are both considered to be strongly related to MAT (Weijers
et al., 2007; Peterse et al., 2012), which is largely similar for paddy and
upland soils from the same sampling region. However, the calculated
TMC was different in adjacent paddy and upland soils (Table S1) and
gradually increased during long-term management in both chronosequences
(Fig. 11) from 14.4 to 17.8 ∘C in paddy soils and
from 17.1 to 19.3 ∘C in upland soils.
In general, temperatures were approximately 1.4 ∘C higher in
upland soils compared to soils under paddy management with the same
cultivation time. This implies that the management type affects the MST,
which in turn controls the membrane lipid composition of brGDGT-producing
bacteria.