Introduction
Permafrost, including constantly frozen sediments of the Arctic, is a unique
subsurface complex environment where microorganisms retain viability over a
long period of time, from thousands to millions of years (Gilichinsky and
Rivkina, 2011). The impact of climate change on permafrost stability has
recently been discussed widely by the scientific community (Zimov et al.,
2006; Walter et al., 2007; Anthony Walter et al., 2014). The permafrost
deposits of northeastern Siberia, which did not thaw during the Holocene
climatic optimum, have attracted particular interest, especially the late
Pleistocene Ice Complex deposits (Yedoma Suite) that are widespread on the
eastern Siberian coastal plains (Lena–Anabar, Yana–Indigirka, and Kolyma
lowlands) (Schirrmeister et al., 2011). It was found earlier that the
epigenetically (consecutively) frozen sediments of both lake and marine
origin (independent of age) contain biogenic methane in concentration up to
1.3 mmol kg-1 (∼ 30 mL kg-1), whereas methane was either
absent or present at trace concentrations
(< 0.0002 µmol kg-1) in
samples from the sincryogenic (simultaneously frozen) late Pleistocene Ice
Complex located in the Yana–Indigirka and Kolyma lowlands (Rivkina et al.,
2007; Rivkina and Kraev, 2008). However, Yedoma deposits found on the
Tyungylyn Terrace of the Lena River (central Yakutia) in a zone of boreal
forest contained methane in concentrations up to 6000 ppmv in pore space of
permafrost (Brouchkov and Fukuda, 2002). Studies conducted on Kurungnakh
Island, which is situated in the southern part of the Lena River delta,
showed the presence of biogenic methane in Holocene sediments (up to
3.5 mmol kg-1), while methane concentrations in the majority of samples
from late Pleistocene Ice Complex were at a detection level and only some
samples from Ice Complex sediments approximately 28 000 years old showed the
presence of methane in a concentration of 1.0 mmol kg-1 and methane
concentration increased up to 2.0 mmol kg-1 in sediments either
younger than 16 000 years or older than 40 000 years (Bischoff et al., 2013). The
presence of biogenic methane up to 0.4 mmol kg-1 in the Ice Complex
deposits on the Gydan Peninsula located on the Siberian coast of the Kara Sea
between the estuaries of the Ob and Yenisei rivers was also shown during our
recent research (unpublished data). These data indicate that different types
of Yedoma deposits are present in different parts of the Siberian Arctic.
Schirrmeister et al. (2013) pointed out that “the typical Ice Complex formation
consisted of several concurrent cryogenic processes, including sediment
accumulation and freezing, ice segregation, syngenetic ice-wedge growth,
sediment reworking, peat aggradation, cryosol formation, and cryoturbation”. Different ways of Yedoma formation are
presumably reflected in the possibility of the methane formation and its
preservation in Yedoma sediments. Studies reviewed above showed an absence of
methane in the late Pleistocene Ice Complex on the territory of the
Yana–Indigirka and Kolyma lowlands.
Anaerobic microcosm incubation of thawed permafrost samples in a carbon
dioxide – and hydrogen – enriched atmosphere showed methanogenic activity
in epicryogenic sediments only, while this process was not observed in
samples from the sincryogenic late Pleistocene Ice Complex (Rivkina and
Kraev, 2008). Similar results were obtained during experiments with
radioactively labeled substrates (Rivkina et al., 2006, 2004, 2007), showing an absence of methanogenic activity
in the sincryogenic late Pleistocene Ice Complex samples with this process
evidently taking place in epicryogenic sediments of lake or lake-alluvial
origin. Moreover, only from the latter sediments (1.3 mmol
CH4 kg-1) methanogenic archaea were isolated in pure culture
(Rivkina et al., 2007; Krivushin et al., 2010; Shcherbakova et al., 2011).
Viable methanogens were also isolated from the permafrost-affected soils
located in the Lena River delta region (Morozova et al., 2007; Wagner et al.,
2013; Liebner et al., 2007).
The specific distribution of methane and methanogenic activity in permafrost
raises a number of questions relating to the composition of microbial communities and
particulary to abundance and composition of methane-producing microorganisms.
Certainly, the application of new methodologies such as metagenomic analyses
is required to estimate diversity and complexity of permafrost microbial
communities, as well as gaining insight into their metabolic capabilities (Graham et
al., 2012; Jansson and Tas, 2014). Until recently, the determination of
microbial diversity in low-biomass environments, including permafrost, was
problematic. The microbial cell abundances in the ancient permafrost is
10–100 times lower than that in the active layer samples, thereby resulting
in low yields of the total community genomic DNA (gDNA) (Yergeau et al.,
2010). However, using appropriate DNA extraction kits (Vishnivetskaya et al.,
2014) and the whole-community genome amplification technique (Yergeau et al.,
2010), a sufficient amount of gDNA can be obtained for next-generation
sequencing technologies, producing sequences on an unprecedented scale.
Indeed, the first metagenomic analyses of permafrost samples became available
recently (Mackelprang et al., 2011; Jansson and Tas, 2014). Specifically, the
analyses of the metagenomes from active layer soil and 2 m deep
permafrost samples collected in the Canadian High Arctic and Alaska
identified signature genes responsible for hydrogenotrophic and acetoclastic
methanogenesis, methylotrophic methane oxidation, nitrification, and
carbohydrate degradation (Yergeau et al., 2010; Mackelprang et al., 2011).
Here we report results of the comparative metagenomic analyses of the two
ancient permafrost samples similar in age (ca. 30 000 years old) but
of different origins (epigenetically frozen lake sediments vs. sincryogenic
sediments from the late Pleistocene Ice Complex). The aim of the study was to
analyze the abundance and composition of microbial communities from the late
Pleistocene permafrost of different origin and elucidate why the sediments of
the late Pleistocene Ice Complex in the Kolyma–Indigirka lowland do not
contain methane or methanogenic activity. The second objective of the study
was to compare putative metabolic capabilities of the communities and use
these findings for explanation of a possible scenario of the development of
low methane Yedoma sediments.
Materials and methods
Sample collection and description
Samples were collected within the Kolyma–Indigirka lowland in northeastern
Siberia (152–162∘ E, 68–72∘ N) during the summer field
season of 2007 (Fig. 1a). Permafrost sediments were sampled using drilling
equipment that operates without fluids and prevents down-hole contamination.
The sampling technique was tested and described previously (Shi et al.,
1997). Briefly, the surfaces of the 20 to 30 cm long cores were cleaned
immediately by shaving melted layers out with an ethyl alcohol-sterilized
knife and then the frozen internal part of the core was split into 5 cm long
segments; these were placed into sterile aluminum containers and kept frozen
during storage in field and transportation to the Institute of
Physicochemical and Biological Problems in Soil Science, Pushchino. The
physicochemical characteristics of the samples are presented in Table 1. Gas
samples were collected by degassing 50 g of frozen cores in a 150 mL
syringe under nitrogen atmosphere as described previously (Rivkina et al.,
2007; Rivkina and Kraev, 2008). CH4 concentration was measured by
headspace equilibration using a KhPM-4 gas chromatograph (Moscow Chromatograph, Russia) with flame ionization detector and hydrogen as a
carrier gas. The δ13C of methane was
analyzed on a Thermo Finnigan GC Combustion III interface and Deltaplus XL mass
spectrometer (Thermo Electron Corporation, Germany).
Location of sampling sites on the Kolyma lowland (a). The
position of samples IC4 and IC8 in drilling holes and methane content from
different depth of these boreholes (b): 1 – sandy loam; 2 – sand
with gravel; 3 – ice wedges; 4 – methane content.
![]()
Physicochemical characteristics of two permafrost soils.
Samples
Grain-size distribution
Ice
Eh
Ctot
Water extract (1:5)
content
Sand
Silt
Clay
Dry residue
pH
HCO3-
Cl-
SO42-
Ca2+
Mg2+
K+
Na+
%
%
mV
%
%
Meq/100 g dry soil
IC4
79.44
15.84
4.72
17.5
45
1.1
0.130
7.22
0.12
1.40
0.14
0.32
0.42
0.06
1.33
IC8
40.72
52.72
6.56
27.6
167
1.2
0.135
8.16
0.57
0.32
0.48
0.75
0.35
0.02
0.56
Sample IC4 corresponded to the permafrost sediment of lake origin
from the floodplain of the Ambolikha River, borehole DH-4/07, depth
22.5 m (Fig. 1b). Total carbon concentration was ∼ 1.1 % (w/w).
Methane content of this sample was 1.2 mmol kg-1, δ13C = -85 ‰, indicative of biogenic origin. The radiocarbon
age of this sample was 30 696 ± 394 years (J–5829) (Kraev et al.,
2013).
Sample IC8 represented a permafrost soil from the late Pleistocene
Ice Complex (Omolon River), borehole DH-2/07, depth 16 m. Total carbon
concentration was ∼ 1.1 % (w/w). Methane levels in all samples
tested from this borehole were < 0.0002 mmol kg-1 (Fig. 1b).
The age of this sample was estimated to be ∼ 32 000 years, based on
the age determination for the same outcrop which was described recently
(Legendre et al., 2014).
DNA extraction and sequencing
In the laboratory, material from the inner part of the permafrost cores was
subsampled aseptically for DNA isolation. The gDNA was extracted from eight
replicates of ∼ 0.5 g each randomly taken from ∼ 50 g of
permafrost core collected at corresponding depth (Fig. 1) using the
PowerSoil® DNA Extraction Kit (MO BIO
Laboratories Inc., USA). Due to low yield (3–6 ng µL-1),
gDNAs from eight replicates were combined, then purified and concentrated
using Genomic DNA Clean and Concentrator®
Kit (Zymo Research Corporation, USA).
The gDNA sequencing libraries were prepared using
NEBNext® reagents (New England BioLabs Inc.,
USA), according to protocol recommended by the manufacturer, having an
estimated peak insert size of 150 nt. Metagenome sequencing was performed at
the CRG Genomics Core Facility (Centre for Genomic Regulation, Barcelona,
Spain) on an Illumina HiSeq 2000™ machine
using Flow Cell v3 with TruSeq SBS v3 reagents and a 2 × 100 cycle
sequencing protocol.
MG-RAST analysis
Raw sequencing data (i.e., 19.8 Gb, representing 143.7 M sequences with an
average length of 138 bp for IC4, and 19.7 Gb, representing 131.7 M
sequences with an average length of 150 bp for IC8) were uploaded to the
MG-RAST server (Meyer et al., 2008) for gene calling and annotation under ID
4 606 864.3 for IC4 and 4 606 865.3 for IC8. A total of 6.6 % (IC4)
and 3.4 % (IC8) of the sequences failed to pass the quality control (QC)
pipeline, whereas 0.3 % of the total sequences in both data sets were
assigned to ribosomal RNA genes.
For functional assignment, protein sequences of putative ORF were searched
against the M5NR non-redundant protein database (Wilke et al., 2012) with an
e-value threshold of 1e-5, minimum percentage identity of 60 %, and
minimum alignment length of 15 aa. The taxonomic assignments of Illumina
reads were performed against M5NR and M5RNA databases at default
parameters. The best-hit classification method was used in both cases for
match assessment.
Comparative metagenomic data analyses
Statistical analyses of the metagenomes were performed in order to compare
community composition and functional profiles of the permafrost samples.
Statistical significance was calculated using the two-sided Fisher's exact test.
The differences between proportions were analyzed in STAMP (Parks and Beiko,
2010) using the MG-RAST taxonomic profile and the Newcombe–Wilson method
(Newcombe, 1998) at a 95 % confidence interval and with Storey's false discovery rate
(FDR) correction (Storey and Tibshirani, 2003; Storey et al., 2004). Original data
sets were deposited at the NCBI Sequence Read Archive (SRA) under the
accession numbers SRX763249 and SRX751044 (Krivushin et al., 2015).
Results and discussion
Community description
The gDNA yield was higher in IC8 sample, with an average of
0.5 µg g-1 of wet sediment, in comparison to
0.37 µg g-1 in the IC4 sample. For gDNA isolation, the
PowerSoil® DNA Extraction Kit (MO BIO
Laboratories Inc., USA) was selected based on its performance and cost per
sample. Previously it has been shown that the gDNA extraction efficiency
(based on the measurement of mCherry plasmid copies) for this kit was
98 %, whereas efficiency of gDNA extraction from permafrost was estimated
at 16 ± 8 % (Vishnivetskaya et al., 2014). The theoretical levels
of the prokaryotic cell populations for studied samples were calculated from
the total gDNA recovered, an average genome length of 4.7 Mb for the soil
bacterial/archaeal population estimated from the metagenomics data (Angly et
al., 2009; Raes et al., 2007), and a weight of 4.05 fg (Ellenbroek and
Cappenberg, 1991) for a genome of this size. Accounting for an average
extraction efficiency and including a reduction of the total cell population
by the eukaryotic component equal to 25 % (Raes et al., 2007), the
estimated prokaryotic cell populations were 4.3 × 108
cells g-1 for IC4 and 5.9 × 108 cells g-1 for IC8.
For gDNA extractions from permafrost, the
PowerSoil® DNA Extraction Kit (MO BIO
Laboratories Inc., USA) was shown to be reliable, providing similar microbial
community structure results with minimal variations among triplicates (P>0.05) (Vishnivetskaya et al., 2014). Analyses of metagenomes of
the two permafrost samples showed that bacterial genes were dominant and 96.4
and 97.7 % of sequences were assigned to the domain Bacteria in the IC4
and IC8 samples, respectively. Archaea were the second dominant domain
followed by Eukaryotes, while viruses comprised only 0.06 % in IC4 and
0.03 % in IC8 samples. Archaea and Eukaryotes were more abundant in the
IC4 sample compared to the IC8 sample, i.e., 2.4 % vs. 1.3 % and 1.0 %
vs. 0.9 %, respectively. Representation of Bacteria in the metagenomic
data was significantly higher than Eukaryotes. A similar pattern has been
observed in metagenomic data sets obtained from other soils including cold
deserts and tundra (Fierer et al., 2012). Bailly et al. (2007) showed that, in spite of the fact that eukaryotic organisms represented
a large portion of the microbial biomass in soil, detection of eukaryotes in
metagenomes is low because they are physically excluded (by filtration or
centrifugation on density gradients) from the biomass before DNA extraction.
Nevertheless, during the current study it was observed that a number of detected
unparalleled genes of eukaryotic organisms, including plants, algae, fungi,
and protists, were 3.4 and 2.6 times higher in comparison to a number of
Bacterial unparalleled genes detected in the IC4 and IC8 samples,
respectively. Eukaryotic organisms including fungi have a genome size several orders of magnitude higher than the genome size of a bacterium.
Eukaryotic genome sizes vary from, for example, 13.8 Mb for the yeast
Schizosaccharomyces pombe (with 4800 protein coding genes) to 69 Mb
for the ciliate Paramecium tetraurelia (39 600 gene models) (Bailly
et al., 2007). Both of these free-living unicellular eukaryotes were detected
in IC4, while only members of the genus Schizosaccharomyces were
detected in IC8 at lower abundance. It is likely that a metagenomic library
based on gDNA can capture a significant functional diversity in spite of low
representation of a eukaryotic microbial community.
The most abundant microorganisms were comprised of nine bacterial and one
archaeal phyla, with Proteobacteria being the most plentiful, followed by
Actinobacteria, Firmicutes, Bacteroidetes, Chloroflexi, Planctomycetes,
Euryarchaeota, Acidobacteria, Cyanobacteria, and Verrucomicrobia in that
order (Fig. 2). The quantities of all other phyla were less than 1 %.
Sequences affiliated with Actinobacteria (P< 1e-15),
Firmicutes (P< 1e-15), Chloroflexi (P< 1e-15), Cyanobacteria (P< 1e-15),
Acidobacteria (P< 1e-15), Verrucomicrobia (P< 1e-15) were dominant within IC8, whereas α-, β-, γ-, and δ-Proteobacteria (P< 1e-15),
Bacteroidetes (P< 1e-15), Planctomycetes (P< 1e-15), and Euryarchaeota (P< 1e-15)
dominated within IC4. The microorganisms that are involved in methane
formation and methane metabolism, for example methanogenic archaea (phylum
Euryarchaeota) or methanotrophic and methylotrophic bacteria (phylum
Proteobacteria), were twice more abundant in the more reduced IC4 sample
(Eh = 45 mV) than in the less reduced IC8 (Eh = 167 mV). Methanogenic
and methanotrophic reactions operate at opposite redox conditions
of -250 and +250 mV, respectively. Estimated value of the redox
potential in the permafrost samples does not support prevalence of either
methanogenesis or methane oxidation. The observation that the late
Pleistocene samples with higher proportion of sands contain low methane
concentrations (Bischoff et al., 2013) was not supported during the current study
(Table 1). Microorganisms without strong association to methane metabolism
(e.g. phyla Actinobacteria and Chloroflexi) dominated in IC8.
Community analyses of the IC4 and IC8 metagenomes at phylum level.
The phyla present in metagenomes at a > 1 % level are shown.
![]()
Dependence of the microbial community structure on permafrost
characteristics as detected from genus taxonomical level
Methanogenic community
As was shown in previous research (Kraev et al., 2013; Legendre et al.,
2014), methane was not detected in the IC8 sample, but its level in the IC4
sample reached 1–2 mmol kg-1; hence, one would expect to find a
greater abundance of methanogenic species in the IC4 sample. Indeed, the
phylum Euryarchaeota, which contains methanogens, was twice as prevalent in
the IC4 sample, and methanogenic Archaea constituted as much as 1.5 % of the
microbial communities in IC4 (compared to 0.5 % in IC8). The most
abundant methane-producing species was Methanosarcina (0.32 % in
IC4 vs. 0.14 % in IC8), followed by Methanoregula (0.28 %
vs. 0.03 %), Methanoculleus (0.1 % vs. 0.05 %),
Methanosphaerula (0.1 % vs. 0.03 %),
Methanospirillum (0.1 % vs. 0.03 %), and Methanosaeta
(0.1 % vs. 0.03 %). Twelve other methanogenic species were detected
at < 0.1 % in IC4 and at an even lower percentage in IC8.
Methanotrophic bacteria
In the vicinity of the environments where methane is produced,
methane-oxidizing (methanotrophic) bacteria can be found (Khmelenina et al.,
2002). Thus, we analyzed the presence of methanotrophic DNA in our data. To
date, two types of methane-oxidizing bacteria are recognized, type I
methanotrophs belonging to γ-Proteobacteria and type II methanotrophs
from α-Proteobacteria. Indeed, the phylum Proteobacteria clearly
dominated in the IC4 sample (50.0 % vs. 26.5 % in IC8) with the
α-Proteobacteria being more abundant in IC4 (32.7 %) in
comparison to IC8 (12.3 %). On a more refined taxonomical scale the most
abundant order of α-Proteobacteria was Rhizobiales (24.2 % in IC4
vs. 7.7 % in IC8), which contains both methanotrophic and nitrogen-fixing
bacteria. The α-Proteobacteria type II methanotrophs (3.32 % vs.
1.06 %) were represented by the genera Methylocella (0.53 %
vs. 0.14 %), Methylosinus (0.24 % vs. 0.07 %),
Methylocystis (0.22 % vs. 0.06 %), and
Methylobacterium (2.32 % vs. 0.8 %) in IC4 and IC8. The last
genus (Methylobacterium) is a facultative methylotroph, however,
some species are capable of growth on methane. The class γ-Proteobacteria was the most diverse Proteobacteria class (184 species);
nonetheless, it was six-fold less plentiful in comparison to α-Proteobacteria. The γ-Proteobacteria type I methanotrophs, such as
Methylococcus (0.13 % vs. 0.11 %), Methylobacter
(0.12 % vs. 0.05 %), and Methylophaga (0.02 % vs.
0.01 %) were again more abundant in the IC4 than in the IC8 sample. In
this Siberian permafrost, type II methanotrophs dominated over type I
methanotrophs, which is similar to methanotrophic bacteria abundance and
diversity in Canadian High Arctic permafrost (Lau et al., 2015).
Methylotrophs, as well as the subset methanotrophs, play an essential role in
the carbon cycle. Interestingly, obligate methylotrophic bacteria belonging
to the β-Proteobacteria, such as Methylibium (0.17 % vs.
0.16 %), Methylobacillus (0.07 % vs. 0.08 %),
Methylovorus (0.04 % vs. 0.04 %), and Methylotenera
(0.06 % vs. 0.05 %), were detected in both the IC4 and IC8 samples at
equally low levels.
Bacteria of nitrogen cycle
The nitrogen cycle includes several microbial processes such as N2
fixation, ammonification, nitrification, and denitrification. Nitrogen-fixing
bacteria more abundant in the IC4 were represented by the genera
Bradyrhizobium (1.85 % vs. 0.5 %), Sinorhizobium
(1.3 % vs. 0.6 %), Rhizobium (0.82 % vs. 0.4 %),
Rhodospirillum (0.63 % vs. 0.3 %), Afipia
(0.4 % vs. 0.12 %), and the γ-Proteobacteria nitrogen-fixing
species. Other species involved in the nitrogen cycle from the
Hyphomicrobium (2.5 % vs. 0.23 %, capable of
denitrification with methanol), Nitrobacter (1.9 % vs.
0.6 %, capable of oxidizing nitrite into nitrate), and
Rhodopseudomonas (3.8 % vs. 1.2 %, capable of carbon
dioxide and nitrogen fixation) genera were also more abundant in the IC4
sample. However, ammonia-oxidizing and nitrifying bacteria of the class
β-Proteobacteria, such as Nitrosomonas
(∼ 0.1 %) and Nitrosospira (∼ 0.17 %), were
detected in both samples at a similar level. Other bacteria involved in the
nitrogen cycle are members of the phylum Planctomycetes, many of which
conduct anaerobic ammonium oxidation or so-called “anammox” metabolism, a
process of ammonia oxidation by nitrite involvement to yield nitrogen gas.
Four planctomycete genera were more abundant in IC4 compared to the IC8
sample, with Planctomyces (0.63 % vs. 0.34 %) being the most
abundant, followed by Pirellula (0.61 % vs. 0.28 %),
Blastopirellula (0.57 % vs. 0.24 %), and Isosphaera
(0.16 % vs. 0.13 %). Some planctomycetes, e.g., Pirellula,
are able to live in environments with high inorganic sulfate concentrations
(Glockner et al., 2003). Nitrogen-fixing cyanobacteria slightly dominated in
IC8 (0.82 %) in comparison to IC4 (0.71 %); however, the proportion
of nitrogen fixing to the total amount of cyanobacteria was higher in the IC4
(55.5 %) than in the IC8 (48.6 %) sample. Another nitrogen-fixing
bacterium dominating in IC8 (2.6 %) in comparison to IC4 (0.8 %) was
the actinobacterium Frankia, which is characterized by the ability
to engage in a symbiotic relationship with plants, producing nitrogen-fixing
root nodules.
Sulfate-reducing bacteria
The δ-Proteobacteria were also more abundant in IC4 (5.8 %) than
in IC8 (4.5 %). Sulfate-reducing bacteria, namely Desulfovibrio
(0.49 % vs. 0.41 %), Desulfatibacillum (0.22 % vs.
0.08 %), Desulfococcus (0.18 % vs. 0.07 %),
Desulfobacterium (0.14 % vs. 0.04 %),
Desulfomicrobium (0.11 % vs. 0.06 %), and metal-reducing
bacteria of the genus Geobacter (1.13 % vs. 0.82 %), were
more plentiful in the IC4 sample. Two species from the order
Syntrophobacterales were more abundant in IC4; these included the strictly
anaerobic, sulfate-reducing, propionate-degrading bacterium
Syntrophobacter (0.33 % vs. 0.18 % in IC8) and the
benzoate-degrading bacterium Syntrophus (0.62 % vs. 0.13 %).
During growth on certain compounds, both of these organisms are known to form
syntrophic associations with methanogens, e.g., Methanospirillum hungateii, facilitating methane production (Harmsen et al., 1998; Jackson et
al., 1999). The sulfate-reducing bacterium Desulfotomaculum from the
phylum Firmicutes was found at a similar concentration in both samples (i.e.,
0.32–0.33 %). However, another strictly anaerobic bacterium,
Desulfitobacterium, which is capable of using a wide variety of
electron acceptors, such as nitrate, sulfite, metals, humic acids, and
halogenated organic compounds, can use H2 as an electron donor to
facilitate sulfate reduction and methanogenesis (Villemur et al., 2006) was
twice as abundant in the IC4 (0.2 %) than in the IC8 sample (0.1 %).
Sulfate-reducing Thermodesulfovibrio species from the Nitrospirae
division were found at 0.09 and 0.06 % in the IC4 and IC8 samples,
respectively.
Similarities and dissimilarities in the microbial communities based on
functional annotation
In contrast to the taxonomical assignment, the functional annotation of the
metagenomes exhibited a similar structure for the microbial communities.
Essential portions of sequencing data ranging from 51.9 % in IC4 to
49.1 % in IC8 were assigned to unknown proteins, while 37.6 % (IC4)
and 44.1 % (IC8) were annotated proteins. A comparison of the metagenomes
at the function level using the SEED – the database for bioinformatic
research (Overbeek et al., 2005) – genome annotations and profile scatter
plots showed that the IC4 and IC8 metagenomes possess > 83 %
similarity at the function level (Supplement Fig. S1). Functional analysis of
the IC4 and IC8 metagenomes demonstrated that, among the annotated protein
sequences, the most abundant groups represented housekeeping functions,
including carbohydrate metabolism (16.6 and 17.7 %), amino acid
biosynthesis (13.8 and 14.8 %), and vitamin and protein metabolism (12.1
and 13.2 %). Approximately 4.1 and 3.7 % of annotated reads were
related to stress response, including a high proportion of oxidative-stress-
(52 and 48 %) and osmotic-stress- (13 and 12 %) related sequences
(Fig. 3). The functional potential documented through the metagenome provides
the information on how many different genes may exist in a sample, but it
does not give any indication of whether these genes are expressed in
permafrost. If slow metabolic activity at subfreezing temperatures in
permafrost exists as shown in the laboratory studies (Rivkina et al., 2000)
then such a community may develop adaptations directed by sediments'
physicochemical properties. Correlations between methane concentrations and
both methanogenic community structure and functional profile derived from
metagenomics shown in the current study suggest that the genomic data could
be used as biomarkers for methanogenesis; however, additional approaches are
needed to provide evidence for metabolic activity in situ. One such method
is a phospholipid analysis. Phospholipid fatty acids (PLFAs) and phospholipid
ether lipids (PLELs) are suitable biomarkers for estimation of either viable
Bacteria and Eukarya or living Archaea, respectively. For example, Bischoff
et al. (2013) correlated PLEL amounts with methane concentrations in Holocene
sediments, suggesting living archaea in those permafrost samples. Though
their study did not find any correlation between PLFA/PLEL markers and
methane concentration in the late Pleistocene sediments, the possible
presence of the inactive methanogenic community or storage of methane
previously produced and trapped within the sediments cannot be ruled out.
Percentage of gene sequences associated with different functions in
the annotated protein sequences within IC4 and IC8 metagenomes.
![]()
Methane metabolism
Analyses using the KEGG (Kyoto Encyclopedia of Genes and Genomes) database
(Kanehisa and Goto, 2000) (Table 2) demonstrated that several functional
genes directly related to methanogenesis are abundant in the IC4 metagenome
(Fig. 4), including genes coding for F420-dependent methylene-H4 MPT
reductase (Enzyme Commission number, EC 1.5.99.11), formylmethanofuran
dehydrogenases (fmd) (EC 1.2.99.5), CoB-CoM heterodisulfide
reductases (EC 1.8.98.1), F420-reducing hydrogenases (EC 1.12.98.1), and
methylenetetrahydromethanopterin dehydrogenases (mer) (EC 1.5.99.9).
All of these genes, with the exception of F420-dependent methylene-H4 MPT
reductase, were also present in the IC8 metagenome, though in lower amounts.
For example, the gene coding for the enzyme catalyzing the last step of
methanogenesis, methyl-coenzyme M reductase (MCR) (EC 2.8.4.1), was
represented by 1333 hits in IC4 but only by nine hits in the IC8 metagenome
(Table 2).
KEGG abundance of the selected functional gene sequences found
within IC4 and IC8 metagenomes. Genes found in low abundance were not
included in the figure. Genes from methane, nitrogen, and sulfur metabolic
pathways are underlined. The enzyme name that corresponds to each KEGG number
is given in Table 2.
![]()
Number of sequences showing homologies to genes associated with
KEGG pathways within the IC4 and IC8 metagenomes. Data were generated by
searching in hierarchical classification table based upon
SubSystems annotation source.
Enzyme nomenclature
Function and enzyme profile
IC4
IC8
Methane metabolism
EC 1.5.99.11
F420-dependent methylene-H4 MPT reductase
3649
0
EC 1.12.98.1
Coenzyme F420 hydrogenase
1441
370
EC 1.8.98.1
CoB-CoM heterodisulfide reductase
24 601
3982
EC 1.2.99.5
Formylmethanofuran dehydrogenase
10 757
3012
EC 1.5.99.9
Methylenetetrahydromethanopterin dehydrogenases
370
3
EC 2.8.4.1
Methyl-coenzyme M reductase
1333
9
EC 2.3.1.101
Formylmethanofuran–tetrahydromethanopterin N-formyltransferase
3056
1642
EC 2.1.1.86
Tetrahydromethanopterin S-methyltransferase
1997
9
Nitrogen metabolism
EC 1.18.6.1
Nitrogenase (molybdenum–iron) alpha chain nifH
2519
42
EC 1.7.2.1
Copper-containing nitrite reductase
1824
11 405
EC 1.7.2.5*
Nitric oxide reductase
3440
1838
EC 1.7.1.4
Nitrite reductase [NAD(P)H] small subunit
20 713
13 041
EC 1.7.99.4
Respiratory nitrate reductase alpha chain narG
36 476
28 169
EC
Ammonia monooxygenase
509
324
EC 1.7.3.4
Hydroxylamine oxidase
517
103
EC 1.7.99.6
Nitrous oxide reductase
61
46
Sulfur metabolism
EC 1.8.4.8
Phosphoadenylyl-sulfate reductase [thioredoxin]
7521
11 165
EC 1.8.7.1
Ferredoxin-sulfite reductase
2587
8920
EC 2.7.7.4
Sulfate adenylyltransferase subunit 2
15 449
23 435
EC 2.7.1.25
Adenylylsulfate kinase
22 898
22 418
EC 1.8.3.1
Sulfite oxidase
14 353
13 968
Stress response
EC 1.11.1.6
Catalase
74 622
41 693
EC 1.15.1.1
Manganese superoxide dismutase
8922
13 302
EC 1.11.1.7
Peroxidase
35 552
19 698
EC 1.15.1.1
Superoxide dismutase [Cu-Zn] precursor
2855
1327
EC 1.15.1.1
Superoxide dismutase [Fe]
7221
3109
EC 1.1.99.1
Choline dehydrogenase
25 244
18 018
EC 1.2.1.8
Betaine-aldehyde dehydrogenase
10 022
12 782
NR
Chaperone protein DnaK
78 365
89 944
NR
Interacting protein DnaJ
32 897
45 343
* The current KEGG number has been created for enzyme
EC 1.7.99.7.
Nitrogen metabolism
The abundance of genes associated with nitrogen metabolism was higher in the
IC4 metagenome than in that of IC8 (2.2 % vs. 1.7 %). The percentage
of genes related to nitrogen metabolism detected in permafrost was slightly
higher than the relative abundance (1.3 %) of genes responsible for
nitrogen metabolism reported in metagenomes examined from glacier ice of the
Northern Schneeferner (Simon et al., 2009). Notably, in the IC8 metagenome
the gene nifH coding for a nitrogenase (EC 1.18.6.1), which
performs nitrogen fixation, was represented by only seven sequences. These
genes were mainly associated with Proteobacteria (43.7 %) represented by
Bradyrhizobium sp., Azoarcus sp., Mesorhizobium loti, and Actinobacteria (Frankia sp., 41 %). In the IC4
metagenome, 90 features corresponding to nifH gene were detected, and
these nitrogenase sequences were predominantly linked with Proteobacteria
(33.5 %). Nitrogenase genes related to Firmicutes (primarily
Desulfitobacterium hafniense and “Alkaliphilus metalliredigens”) and Actinobacteria (Frankia species) constituted
17.6 and 15 %, respectively. Interestingly, nifH genes of
cyanobacterial origin (primarily linked to Nostoc species) were
detected exclusively within the IC4 metagenome (12.9 %).
Genes connected with the denitrification processes (Table 2), such as nitrate
reductase (EC 1.7.99.4) and nitrite reductase (EC 1.7.2.1), were found in both
metagenomes (Fig. 4). The nitrate reductase (narG, EC 1.7.99.4)
sequences predominating within the IC4 metagenome came from more diverse
phylogenetic groups in contrast to narG gene from IC8. By contrast,
even though nitrite reductase (nirS) genes were significantly
overrepresented in IC8, their presence was detected in similar phylogenetic
groups in both metagenomes (Fig. 5). The sequences related to both nitrite
reductase (EC 1.7.1.4) and nitric oxide reductase (EC 1.7.2.5) were found in
similar phylogenetic groups in both metagenomes with prevalence in IC4
(Fig. 5).
Phylogenetic distribution of the sequences related to nitrogen
metabolism within the IC4 and IC8 metagenomes.
![]()
The ammonium oxidation pathway was represented by a few sequences related to
hydroxylamine oxidase (EC 1.7.3.6) genes in known nitrifying bacteria such as
Nitrosomonas eutropha, Nitrosococcus oceani,
Nitrosospira multiformis, and some others. Gene coding for ammonia
monooxygenase (EC 1.14.99.39) were not detected in either metagenome by a
search with its EC number. However, a search for ammonia monooxygenase using
functional hierarchies such as KEGG orthologs yielded 10 hits in IC4 and
4 hits in IC8, while SubSystems showed presence of 509 hits in IC4 and 324
hits in IC8. The ammonia monooxygenase sequences were annotated as methane
monooxygenase (EC 1.14.13.25). It should be noted that the particulate
methane monooxygenase and ammonia monooxygenase are related and occur in both
methanotrophs and ammonia oxidizers (Holmes et al., 1995). These enzymes have
wide substrate specificity catalyzing the oxidation of various substrates
including ammonia, methane, halogenated hydrocarbons, and aromatic molecules
(Arp et al., 2002). Overall, the low abundance of ammonia oxidizers
presumably represents the rare utilization of this pathway by permafrost
bacteria in studied environments. Similar results were also reported for
metagenomes from other cold environments, e.g., high-Arctic hypersaline
subzero spring (Lay et al., 2013) and Arctic snow packs (Larose et al.,
2013).
Sulfur metabolism
Sequences associated with sulfur metabolism were present in both IC4 and IC8
metagenomes and related to both reduction and oxidation (Table 2 and Fig. 6).
Overall, genes coding for sulfate reduction were more abundant in the IC8
metagenome, including genes for sulfate adenylyltransferase (EC 2.7.7.4),
phosphoadenylyl-sulfate reductase (EC 1.8.4.8), and ferredoxin-sulfite
reductase (EC 1.8.7.1). Higher abundance of enzymes associated with sulfate
reduction (Fig. 4) and higher concentrations of SO42- (Table 1), as
well as an absence of methane in the IC8 sample, may suggest that sulfate reduction
probably dominated methanogenesis or took part in the
anaerobic oxidation of methane at the time the sediments were formed. Taxonomic
distribution of associated species was similar in both metagenomes, with the
exception of sulfate adenylyltransferase, which was represented in IC4 by
sequences related mainly to Proteobacteria and Actinobacteria. In the IC8
sample this gene was of more diverse phylogenetic origin (Fig. 6). A few
sulfur oxidation genes detected were associated with Renibacterium salmoninarum and Gordonia bronchialis in IC8 and
Mycobacterium species and Sinorhizobium meliloti in IC4.
Phylogenetic distribution of the sequences related to sulfur
metabolism (sulfur reduction) within the IC4 and IC8 metagenomes.
![]()
Stress response
Genes associated with stress response were detected in both of the
metagenomes (33 683 hits in IC4 metagenome and 28 557 in IC8) (Table 2).
The three most abundant groups present corresponded to oxidative stress, heat
shock, and osmotic stress response genes. Sequences related to oxidative
stress originated principally from Proteobacteria, Actinobacteria, and
Firmicutes (5148 features in IC4 and 3832 features in IC8) and included genes
for catalase (EC 1.11.1.6), peroxidase (EC 1.11.1.7), and different
superoxide dismutases (EC 1.15.1.1). Their occurrence is presumably
explained by increased oxygen solubility at low temperatures and associated
increase in reactive oxygen species concentration (Chattopadhyay, 2006).
Sequences related to osmotic stress were represented by the genes involved in
the synthesis and uptake of compatible solutes including choline, betaine,
periplasmic glucan, and ectoine. Genes for osmoprotectant ABC transporters
were also detected. Choline dehydrogenase (EC 1.1.99.1) (222 features in IC4
and 213 features in IC8 mainly from Proteobacteria and Actinobacteria) and
betaine-aldehyde dehydrogenase (EC 1.2.1.8) (166 features in IC4 and 186
features in IC8 from Proteobacteria, Actinobacteria, and Firmicutes) were the
most abundant enzymes of this class. This emphasizes the importance of
betaine osmolyte for the osmoprotection of members in microbial communities
from subfreezing environments. The genes that encoded the heat shock proteins were
mainly represented by the chaperone protein DnaK (816 hits in IC4 and 54 in
IC8) and its interacting protein DnaJ (759 hits in IC4 and 67 in IC8). These
proteins are among the most plentiful chaperons in the bacterial cell and
often prevalent in microorganisms from cold environments (D'Amico et al.,
2006).
Implication of metagenomic sequencing data for the analysis of the
samples origin and evolution
The application of biological markers for paleo-reconstructions in various
environmental sites has been used occasionally: for example, utilization of
lipid analyses for petroleum reservoir formation and maturation (Seifert and
Moldowan, 1981), analysis of fossil chironomid assemblages in the Holocene
lake-sediment cores (south-central Alaska) for evaluating anthropogenic
climatic changes and quantitative paleo-temperature reconstructions (Clegg et
al., 2010), and analyses of fossil ostracodal assemblages from the Arctic
seas for reconstruction of coastline and interpretation of environmental
differences in Arctic areas (Stepanova et al., 2010). To track the occurrence
and distribution of microorganisms in the environment, the gDNA and DNA
fragments amplified with PCR can be employed as biological markers.
Therefore, we anticipated that integration of the next-generation sequencing
capabilities with phospholipid analysis and approaches in microbial ecology
(such as linking microbial community composition and environmental processes
involved in carbon, nitrogen, and sulfur cycling) could be used for
paleo-reconstructions.
The late Pleistocene Ice Complex or Yedoma deposits are widely distributed
across northeastern Siberia and their distribution relates to a number of
geomorphological settings, for example (1) low-elevation coastal mountains as
the main sediment sources for Yedoma, (2) Yedoma areas comprising separate
bedrock elevations (100–400 m high), and (3) extended lowland areas
dominated by large and numerous thermokarst basins and lakes with Yedoma
hills present between them (Schirrmeister et al., 2013). The question of which
processes formed these deposits has been a matter of debate in the last several
decades. Several hypotheses have been proposed about the origin of the late
Pleistocene Ice Complex, including eolian (Tomirdiaro et al., 1984;
Tomirdiaro and Chernen'k'ii, 1987), alluvial (Rozenbaum, 1981), and
polygenetic (Konishchev and Kolesnikov, 1981; Sher et al., 1987) formation.
Different opinions on the origin of these deposits have been summarized in
the recent publications of Schirrmeister et al. (2011, 2013). It was suggested that the ice-rich syngenetic permafrost of
the late Pleistocene Ice Complex developed in a cold-arid climate under
less hydromorphic conditions than the lake and lake-alluvial sediments
(Schirrmeister et al., 2013; Rivkina et al., 2006). Cold-arid climate was
presumably an important factor for Yedoma formation and Bischoff et al. (2013) found that Yedoma formed on Kurungnakh Island during cold and dry
period from ∼ 32 000 to 16 000 years ago had a less abundant community of
methanogenic archaea and low amounts of methane trapped within the sediments. The Yedoma sediments located on the Yana–Indigirka,
and Kolyma lowlands have been characterized by the absence of methane
(Rivkina et al., 2007; Rivkina and Kraev 2008) and much lower redox potential
and iron (Fe2+) content in comparison to permafrost layers of lake and
lake-alluvial origin in the same area (Rivkina et al., 2006).
A comparison of the two late Pleistocene permafrost metagenomes from samples
of different genesis, i.e., of lake origin from the floodplain and from Yedoma, revealed
differences in the composition of the microbial community that reflect the
conditions under which these deposits were formed. These data uncovered
significant distinctions in microbial community compositions between Yedoma
and lake-alluvial sediments. The relatively low abundance of methanogenic
archaea, the limited presence of enzymes from the carbon and nitrogen cycles,
the higher abundance of enzymes from the sulfate reduction process, and the
presence of methanotrophic bacteria could explain the absence of methane in
Yedoma deposits and provide evidence that the formation of these sediments
took place under much more aerobic conditions. In other words, we may assume
that during the late Pleistocene period, nearly 30 000 years ago, different
environmental conditions predetermined biogeochemical regimes and composition
of microbial communities of the studied samples. Similar to our study, a
recent survey of soil taxonomic, phylogenetic, and functional diversity has
demonstrated that metagenomic approaches can be used to build a predictive
understanding of variations in microbial diversity and functions across
terrestrial biomes (Fierer et al., 2012).
Potential limitation of analysis
Shotgun metagenomic sequencing provides insight into the identity and
functional diversity of microbial communities. Although this approach is a
powerful tool for understanding microbial structure and function, there are a
number of potential limitations that may bias conclusions. The limitations
and biases, which may be introduced throughout sampling and processing of
environmental samples, have been reviewed recently (Hazen et al., 2013).
Keeping in mind that obtaining a representative sample is the most
fundamental step in monitoring microbial communities, we selected samples
based on their similar age, close total carbon content, and significant
differences in methane concentrations. We expected that minimal biases
were introduced during coring, transportation, and storage, because the
permafrost samples were collected following the same technique (Shi et al.,
1997) and were transported frozen and stored at -20 ∘C in the
Laboratory of Soil Cryology, Pushchino. To obtain average gDNA probes and
generate enough gDNA for downstream applications, we used eight replicates
from each permafrost sample and the
PowerSoil® DNA Extraction Kit (MO BIO
Laboratories Inc., USA), which showed high gDNA extraction efficiency
(Vishnivetskaya et al., 2014). To ensure gDNA integrity, the gDNA solutions
were properly stored at -20 ∘C in the laboratory and shipped on dry
ice to the CRG Genomics Core Facility (Centre for Genomic Regulation,
Barcelona, Spain) for sequencing. To decrease biases during metagenome
sequencing we did not utilize multiple displacement amplification (MDA) to
amplify the total community gDNA from low biomass permafrost samples; rather,
we used multiple replicates to obtain enough gDNA for downstream
applications. Keeping in mind that interpretation of metagenomic data is
highly dependent upon the depth of sequencing, accurate annotation, and
comprehension of database (Hazen et al., 2013), bioinformatic analyses were
performed in the same fashion for both samples that allowed accurate
comparison of these metagenomes.