Content uploaded by Julia Mitzscherling
Author content
All content in this area was uploaded by Julia Mitzscherling on May 07, 2019
Content may be subject to copyright.
Available via license: CC BY 4.0
Content may be subject to copyright.
1
Microbial community composition and abundance after millennia of
submarine permafrost warming
Julia Mitzscherling1, Fabian Horn1, Maria Winterfeld2, Linda Mahler1, Jens Kallmeyer1, Pier P.
Overduin3, Lutz Schirrmeister3, Matthias Winkel4, Mikhail N. Grigoriev5, Dirk Wagner1,6 and Susanne
5
Liebner1,7
1GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam, Section 3.7 Geomicrobiology, 14473 Potsdam,
Germany
2Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Marine Geochemistry, 27570 Bremerhaven,
10
Germany
3Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Permafrost Research, 14473 Potsdam, Germany
4GFZ German Research Centre for Geosciences, Helmholtz Centre Potsdam, Section 3.5 Interface Geochemistry, 14473
Potsdam, Germany
5Siberian Branch, Russian Academy of Sciences, Mel’nikov Permafrost Institute, Yakutsk, Russia
15
6University of Potsdam, Institute of Geosciences, 14476 Potsdam, Germany
7University of Potsdam, Institute of Biochemistry and Biology, 14476 Potsdam, Germany
Correspondence to: Susanne Liebner (Susanne.Liebner@gfz-potsdam.de)
Abstract. Warming of the Arctic led to an increase of permafrost temperatures by about 0.3°C during the last decade.
20
Permafrost warming is associated with increasing sediment water content, permeability and diffusivity and could on the long-
term alter microbial community composition and abundance even before permafrost thaws. We studied the long-term effect
(up to 2500 years) of submarine permafrost warming on microbial communities along an onshore-offshore transect on the
Siberian Arctic Shelf displaying a natural temperature gradient of more than 10 °C. We analysed the in-situ development of
bacterial abundance and community composition through total cell counts (TCC), quantitative PCR of bacterial gene
25
abundance and amplicon sequencing, and correlated the microbial community data with temperature, pore water chemistry
and sediment physicochemical parameters. On time-scales of centuries, permafrost warming coincided with an overall
decreasing microbial abundance while millennia after warming microbial abundance was similar to cold onshore permafrost
and DOC content was least. Based on correlation analysis TCC unlike bacterial gene abundance showed a significant rank-
based negative correlation with increasing temperature while both TCC and bacterial gene copy numbers showed a negative
30
correlation with salinity. Bacterial community composition correlated only weakly with temperature but strongly with pore-
water stable isotope signatures and depth. Microbial community composition showed substantial spatial variation and an
overall dominance of Actinobacteria, Chloroflexi, Firmicutes, Gemmatimonadetes and Proteobacteria which are amongst the
microbial taxa that were also found to be active in other frozen permafrost environments. We suggest that, millennia after
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
2
permafrost warming by over 10°C, microbial community composition and abundance show some indications for proliferation
but mainly reflect the sedimentation history and paleo-environment and not a direct effect through warming.
1 Introduction
Temperatures in high-latitude regions have been rising twice as fast as the global average over the last 30 years (IPCC in
Climate Change 2013, 2013) and are predicted to experience the globally strongest increase in the future (IPCC in Climate
5
Change 2013, 2013; Kattsov et al., 2005). In the northern hemisphere, 24 % of the land surface (Zhang et al., 2003) and large
areas of the Arctic shelves are underlain by permafrost (Brown et al., 1997). With 1672 Pg carbon (Schuur et al., 2008), the
northern circumpolar permafrost zone stores about twice as much carbon as currently found in the atmosphere (Schuur et al.,
2009; Zimov et al., 2006). About 88% of this carbon occurs in permafrost soils and deposits (Tarnocai et al., 2009).
Permafrost harbours numerous ancient but viable cells (Bischoff et al., 2013; Gilichinsky et al., 2008; Graham et al., 2012;
10
Koch et al., 2009; Mackelprang et al., 2011; Wagner et al., 2007) that can remain active at extremely low temperatures
(Hultman et al., 2015; Rivkina et al., 2000). With increasing permafrost age, microbial communities show adaptations to the
permafrost biophysical environment and specialize towards long-term survival strategies such as increased dormancy, DNA
repair or stress response (Johnson et al., 2007; Mackelprang et al., 2017). Following the trend of air temperature increase in
the northern hemisphere, continuous permafrost warmed by about 0.3°C over the last decade at a global scale (Biskaborn et
15
al., 2019). Warming of permafrost can substantially increase liquid water content, sediment diffusivity and permeability
(Overduin et al., 2008; Rivkina et al., 2000; Watanabe and Mizoguchi, 2002) potentially mobilizing carbon in the form of
trapped methane (Portnov et al., 2013; Shakhova et al., 2010, 2014; Thornton et al., 2016). Microbial community
composition was reported to be responsive to temperature changes (Luo et al., 2014; Rui et al., 2015; Weedon et al., 2012;
Xu et al., 2015; Zhang et al., 2005; Zogg et al., 1997). However, results on the extent of these community changes and their
20
dependence on exposure time are contradictory (Allison et al., 2010; Schindlbacher et al., 2011; Walker et al., 2018; Weedon
et al., 2017; Xiong et al., 2014; Zhang et al., 2016). In general, the microbial community response to warming appears to be
delayed (DeAngelis et al., 2015) and the effect of warming might take decades to affect the microbial community
composition (Radujković et al., 2018; Rinnan et al., 2007). Not only microbial community composition can be responsive to
temperature but also microbial abundance especially in systems with weak energy constraints. Microbial abundance
25
correlates with enzymatic activities and methane production (Taylor et al., 2002; Waldrop et al., 2010), which are sensitive to
temperature. Microbial growth, respiration and carbon uptake can correlate with microbial biomass (Walker et al., 2018).
Thus, substantial permafrost warming on long time-scales could affect microbial community composition and abundance
before permafrost thaws.
Submarine permafrost provides an analogue for rising permafrost temperatures over time-scales of centuries and millennia.
30
Submarine permafrost of the Arctic Sea shelves originally formed under terrestrial (subaerial) conditions and was inundated
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
3
by post-glacial sea level rise during the Holocene (Romanovskii and Hubberten, 2001). Upon sea transgression, permafrost
degraded over thousands of years as the relatively warm ocean water warmed the submerged sea floor. Mean annual bottom
water temperatures in the Laptev Sea (East Siberian Arctic shelf) are 12 to 17 °C warmer than the annual average surface
temperature of terrestrial permafrost (Romanovskii et al., 2005). Even today, new submarine permafrost is created by erosion
of Arctic permafrost coasts (Fritz et al., 2017), which account for 34% of the coasts worldwide (Lantuit et al., 2012). In a
5
recent study, we compared submarine sediment cores from two locations on the Siberian Arctic Shelf and looked at the
combined effect of permafrost inundation time and seawater intrusion on microbial communities. We showed that flooding
by sea water reduced permafrost bacterial abundance and changed bacterial community composition due to the penetration of
seawater into a former freshwater habitat (Mitzscherling et al., 2017). It was suggested that in addition to the effect of
seawater infiltration, the sediment warming taking place over millennia could lead to proliferation. However, the specific
10
effect of long-term permafrost warming independent of thawing has not been assessed so far. Here we hypothesize that
millennial-scale permafrost warming directly increases microbial abundance and alters microbial community composition.
We used submarine permafrost sediments of comparable age and physicochemical properties that differed in temperature by
more than 10 °C due to different periods of inundation and sediment warming and assessed total microbial and bacterial
abundances and community composition relative to temperature, pore-water chemistry and sedimentation history.
15
2 Materials and Methods
2.1 Study site and drilling
The study area (~73°60’N, 117°18’E) is situated in the western part of the Laptev Sea, on the East Siberian Arctic Shelf (Fig.
1). Mean annual bottom water temperatures in the Laptev Sea range between -1.8 °C to -1 °C (Wegner et al., 2005) leading
to sediment temperatures of -1.0 °C and -2.0 °C within the largest part of the shelf (Romanovskii et al., 2004). We
20
investigated four cores (C1-C4, Fig. 2a) that were retrieved along an onshore-offshore transect in the coastal region of Cape
Mamontov Klyk in 2005 (Overduin, 2007; Rachold et al., 2007). Drilling was performed with a hydraulic rotary-pressure
mechanism (URB-2A-2) and without utilizing of drill fluid. All samples were frozen immediately after recovery and were
kept at -22 °C until further processing. Cores were named after the order of drilling and we kept this order (C1, C4, C3, C2)
for better comparability with previous studies (Koch et al., 2009; Mitzscherling et al., 2017; Overduin et al., 2008; Winkel et
25
al., 2018).
Assuming a constant mean annual coastal erosion rate of 4.5 m yr-1 (Grigoriev, 2008) the drill site located furthest offshore
(C2, 11.5 km off the coast) was inundated approximately 2500 years ago (Rachold et al., 2007). Accordingly, the drill sites
C3 and C4, located 3 km and 1 km off the coast, were inundated around 660 and 220 years ago, respectively. More recent
analysis based on remote sensing shows that 40-year coastal erosion rates for the same stretch of coastline between 1965 and
30
2007 were slower (about 2.9 m yr-1) (Günther et al., 2013), which would translate into even longer inundation periods.
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
4
However, in the present study we refer to Grigoriev (2008), which are based on direct observations of coastal erosion at the
C1 coring site. From onshore to offshore all cores were characterized by an increase in water depth, in depth to the ice-
bonded permafrost table (Fig. 2a, Table S1) and in ground temperature (Table S2) (Overduin, 2007; Rachold et al., 2007).
Temperature measurements at all sites were done using thermistors and infra-red sensors (Junker et al., 2008).
2.2 Sample selection
5
Each of the four drill cores exhibited different sedimentological units. Lithostratigraphic Unit II was identified in all cores
(Fig. 2a) and was entirely located within the ice-bonded permafrost. Depth location of Unit II within each core can be found
in Table S1. This unit was deposited during the late Pleistocene, was warmed without thawing, and had so far remained
unaffected of seawater infiltration. On the basis of a PCA analysis (see next chapter and Fig. 3) and previous
lithostratigraphic descriptions (Winterfeld et al. 2011) all further analysis was conducted on samples from Unit II. The ages
10
of the sediment are published in Winterfeld et al. (2011). The present study refers to sediment ages determined by optically
stimulated luminescence (OSL) on quartz and infrared optically stimulated luminescence (IR-OSL) on feldspars. OSL ages
of Unit II sediments from core C1 range from 30.5 ± 2.0 ka at 22 m below surface (m bs) to 114 ± 6 ka at 50 m bs. OSL ages
range from 97 ± 6 to 112 ± 8 ka between 23 and 30 m below sea floor (m bsf) in core C3 and from 133 ± 8 to 148 ± 14 ka
between 37 and 53 m bsf, and increase with depth. IR-OSL ages date back to 59 ± 5.8 ka at around 15 m bsf in C4 and 86 ±
15
5.9 ka at 44 m bsf and 111 ± 7.5 ka at 77 m bsf in C2. Consequently, sediments of Unit II were deposited during the early to
middle Weichselian (Winterfeld et al., 2011).
For molecular analyses we took 6 replicate samples from each of the cores C1 (C1-1 – C1-6), C4 (C4-1 – C4-6) and C3 (C3-
1 – C3-6) and 8 replicates from core C2 (C2-1, C2-2, C2-4, C2-5, C2-7, C2-8, C2-9, C2-10) (Fig. 2a). Those replicates were
located at different depths within Unit II (Table S4). The unit was mainly composed of sands with varying proportions of silt
20
and to a minor extent of clay, and a frequent occurrence of wood fragments, plant detritus interlayers and small peat
inclusions (Winterfeld et al., 2011). Both, sandy as well as organic-rich deposits were represented by three replicates in C1,
C4 and C3 and four replicates in C2 (Table S4). Furthermore, to check for reproducibility we included samples from C2
retrieved in a previous study (Mitzscherling et al., 2017) (sample names CK12xx). To minimize contamination we took the
subsamples from the centre of the core.
25
2.3 Pore water and sediment analyses
Pore water of segregated ground ice was extracted from thawed subsamples of the sediment cores using rinsed Rhizons™
(0.15 µm pore diameter). Electrical conductivity, salinity, cation and anion concentrations, stable isotope concentrations
(δ18O, δD), and pH were measured for 183 samples of C1, 67 samples of C2, 38 samples of C3 and 10 samples of C4 in Unit
II (Table S3). Electrical conductivity, salinity and pH were measured with a WTW MultiLab 540 using a TetraConTM 325
30
cell referenced to 20°C. Total dissolved element concentrations (Ba2+, Ca2+, K+, Mg2+, Na+, Siaq) were determined by
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
5
inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 3000XL, Perkin-Elmer, Waltham) (Boss and
Frieden, 1989). Dissolved anion concentrations (Cl-, SO42-, Br-, NO3-) were measured using a KOH eluent and a latex particle
separation column on a Dionex DX-320 ion chromatographer (Weiss, 2001). The stable water isotopes (δD andδ18O) of
segregated ground ice were determined following (Meyer et al., 2000) using a Finnigan MAT Delta-S mass spectrometer in
combination with two equilibration units (MS Analysetechnik, Berlin).
5
Dissolved organic carbon (DOC) was measured as non-purgeable organic carbon via catalytic combustion at 680 °C using a
Shimadzu TOC-VCPH instrument on samples treated with 20 µl of 30% supra-pure hydrochloric acid. The ice content was
determined gravimetrically. Grain sizes were measured with a Coulter LS 200 laser particle size analyzer. The total organic
carbon (TOC) was measured with the element analyser VARIO MAX C, while total carbon (TC), total nitrogen (TN) and
total sulfur (TS) contents were determined with a CNS analyzer (Elementar Vario EL III).
10
2.4 DNA extraction
Core subsamples were homogenized in liquid nitrogen and DNA was extracted from ~5 g of sediment using a modified
protocol of Zhou et al. (1996). The method was described before (Mitzscherling et al., 2017) and in the following we refer to
these samples as molecular samples. Quality of the extracted genomic DNA was assessed via gel electrophoresis. DNA
concentration was quantified with the Qubit2 system (Invitrogen, HS-quant DNA) and the crude DNA was purified using the
15
HiYield PCR Clean-Up & Gel-Extraction Kit (SLG) to reduce PCR inhibitors prior to PCR applications.
2.5 Quantification of the bacterial 16S rRNA gene
Quantitative PCR was performed using the CFX Connect™ Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.)
and the primers S-D-Bact-0341-b-S-17 and S-D-Bact-0517-a-A-18 for the bacterial 16S rRNA gene (Table S5). Each
reaction (20 µl) contained 2x concentrate of iTaq™ Universal SYBR® Green Supermix (Bio-Rad Laboratories), 0.5 µM of
20
each the forward and reverse primer, sterile water and 2 µl of template DNA. The qPCR assays comprised the following
steps: initial denaturation for 3 min at 95 °C, followed by 40 cycles of denaturation for 3 sec at 95 °C, annealing for 20 sec at
58.5 °C, elongation for 30 sec at 72 °C and a plate read step at 80 °C for 0.3 sec. Melt curve analysis from 65-95 °C with
0.5°C temperature increment per 0.5 sec cycle was conducted at the end of each run. The qPCR assay was calibrated using
known amounts of PCR amplified gene fragments from a pure Escherichia coli culture. For each sample three technical
25
replicates were analysed and DNA templates were diluted 5- to 100-fold prior to qPCR analysis. The PCR efficiencies based
on standard curves were calculated using the BioRad CFX Manager software. They varied between 93 and 99%. All cycle
data were collected using the single threshold Cq determination mode.
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
6
2.6 High throughput Illumina16S rRNA gene sequencing and analysis
Sequencing of each sample was performed in two technical replicates. Primers comprised different combinations of barcodes
(Table S6). PCR amplification was carried out with a T100™ Thermal Cycler (Bio-Rad Laboratories, CA, USA). The PCR
mixtures (25 µl) contained 1.25 U of OptiTaq DNA Polymerase (Roboklon), 10x concentrate buffer C (Roboklon), 0.5 µM
of the sequencing primers S-D-Bact-0341-b-S-17 and S-D-Bact-0785-a-A-21 (Table S5), dNTP mix (0.2 mM each),
5
additional 0.5 mM of MgCl2 (Roboklon), PCR-grade water, and 2.5 µl of template DNA. Thermocycler conditions as well as
clean up and quantification of PCR products, library preparation, Illumina MiSeq sequencing (GATC, Germany) and raw
sequence data analysis were performed as described before (Mitzscherling et al., 2017). The number of PCR cycles was
chosen to be 35. OTUs were taxonomically assigned employing the SILVA database (release 123) with a cutoff value of
97%.
10
2.7 Total cell counts
Preparation and quantification of the total cell abundance per g sediment were performed after (Llobet-Brossa et al., 1998).
The modified protocol was described before by Mitzscherling et al. (2017). Briefly, cells were fixed with 4%
paraformaldehyde in phosphate-buffered saline (PBS). After incubation, the sediment was pelleted by centrifugation for 5
min at 9600 g and washed in sterile filtered PBS. Two subsamples of each sample were diluted in PBS and filtered onto a
15
polycarbonate membrane filter (0.2 µm) by applying a vacuum. Total cell counts were determined by SYBR Green I.
Fluorescence microscopy was performed with a Leica DM2000 fluorescence microscope using the FI/RH filter cube. A
magnification of 100x was used to count cells of either 200 fields of view or until 1000 cells were counted. We counted two
filters per sample.
2.8 Statistics
20
Prior to statistical analysis, absolute singletons and OTU0.03 (operational taxonomic units of clustered sequences with 97%
similarity level) not classified as bacteria or classified as chloroplasts or mitochondria were removed. In addition, OTU0.03
with reads <0.5% of total read counts in each sample were removed to reduce background noise. The background noise was
estimated with the help of a positive control (E. coli), where the number of OTUs is known prior to sequencing. Absolute
read counts were transformed into relative abundances in order to standardize the data and to make technical replicates
25
comparable. Relative abundances of technical replicates were merged to mean relative abundances for bacterial community
analysis i.e. the bubble plot and CCA. Samples having < 15.000 raw reads were checked for divergent relative abundances
within duplicates (Table S7) and excluded from the calculation of mean relative abundances when the discrepancy was too
big. Variation in OTU0.03 composition, 16S rRNA gene and total cell abundance between samples and among drill sites, as
well as correlations of the abundance and OTU0.03 composition with environmental parameters were assessed using the Past
30
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
7
3.14 software (Hammer et al., 2001). Principal component analyses (PCA) based on Euclidean distance were used to assess
variation in environmental variables across the different sediment units and within Unit II. Prior to analysis, all
environmental data were standardized by subtracting the mean and dividing by standard deviation. To assess the correlations
of bacterial and microbial abundance with environmental parameters the rank-based Spearman correlation was calculated.
Mantel tests were used to study the relationship between environmental parameters and bacterial community composition
5
(Mantel, 1967). A canonical correspondence analysis (CCA) was conducted to visualize the dependence of the bacterial
communities on environmental parameters. PerMANOVAs were conducted (Anderson, 2001) to test whether communities or
abundances were significantly different between drill sites. Analysis of variance (ANOVA) and the Dunn’s post-hoc test
were conducted to test whether DOC concentrations of the cores differed.
3 Results
10
3.1 Physicochemical pore water and sediment properties
Temperature (Fig. 2b) of Unit II was lowest in the terrestrial borehole (C1, constantly at around -12.4 °C at the time of
drilling (Junker et al., 2008) and between -12.0 and -12.5 °C recently measured over a 2 year period (Kneier et al., 2018))
and increased with distance to the shore. According to (Junker et al., 2008) C4 exhibited a temperature range from -7.1 to -
5.8 °C. Ground temperatures of C3 and C2 were similar with mean values of -1.4 and -1.5 °C, respectively, and showed
15
marginal variation. C3 exhibited a slightly higher mean temperature than the longest inundated core C2.
Overall, the salinity of Unit II was low (Fig. 2b, (Winterfeld et al., 2011)). In C4, the drill site located closest to the coast,
Unit II had the highest pore water salinity (mean = 5.6 PSU) ranging from 0.9 to 17.6 PSU (Table S2), which spans
freshwater to mesohaline water but is much below seawater salinities. Salinity in C3 reached a mean value of 1.1 PSU. The
submarine core furthest offshore (C2) and the terrestrial core (C1) had a mean pore water salinity of around 0.8 and 0.5 PSU,
20
respectively. The stable water isotopes δD and δ18O of the sediment cores C1 and C4 exhibited similar mean values of -22 ‰
for δ18O and around -178 ‰ for δD, albeit a greater variance in C1 (Fig. 2c, Table S2). Sediments of C3 were characterized
by higher and constant isotope values of around -20 ‰ for δ18O and -158 ‰ for δD. In core C2, the isotope values were
smaller with mean values of -28 ‰ for δ18O and -213 ‰ for δD (Table S2).
DOC concentrations were lowest in Unit II of core C2, the core furthest offshore, and ranged from 4 to 41 mg C L-1, with a
25
mean value of 17 mg C L-1 (Fig. S1). Towards the coast the DOC content increased to mean values of 43 mg C L-1 in C3 and
96 mg C L-1 in C4. The terrestrial core C1 had a mean DOC concentration of around 48 mg C L-1 with values ranging from 4
to 305 mg C L-1, thereby having by far the highest measured DOC concentration of all cores. The TOC content in this Unit II
was generally very low with mostly < 0.5 wt%. While C1 and C4 had lowest mean values of 0.17 wt%, the TOC content
increased with distance to the coast to 0.22 wt% in C3 and 0.33 wt% in C2 (Table S4). The pH of Unit II sediments ranged
30
from slightly acidic to slightly alkaline values. In cores C1 and C4 the pH ranged from 5 to 7.9, whereas values of C2 and C3
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
8
were higher ranging from pH 6.5 to 8.0. Mean pH values of all cores were around pH 7 to 7.5. Other pore water data like
anion and cation concentrations, conductivity, CNS, grain sizes and the gravimetrically determined water content can be
found in Table S3.
All environmental, sedimentological and pore water data (Table S3) were used to conduct principal component analyses
(PCA) to check for the level of similarity within Unit II. Unit II formed a dense cluster relative to the other sediment units
5
(Fig. 3 Insert). Focusing on samples from Unit II only (Fig. 3) confirmed highly similar physicochemical characteristics of
this unit in all cores even though C2 and C3 clustered along the axis PC2, while C1 and C4 were more randomly scattered.
Variance between samples was mainly explained by grain sizes, stable water isotope concentrations and to a lesser extent by
pH.
3.2 Microbial abundance
10
Overall microbial abundance decreased from onshore to offshore (C1, C4, C3) and had increased again in the drill site
located furthest from the coast (C2). The terrestrial permafrost core C1 and the submarine core C2 had highest DNA
concentrations (Fig. S2), total cell counts (TCC) (Fig. 4a) and bacterial 16S rRNA gene copy numbers of all cores (Fig. 4b).
Lowest DNA concentrations and TCC were observed in core C3, whereas lowest numbers of bacterial 16S rRNA gene
copies were found in core C4. All three abundance measures (DNA concentrations, TCC, and bacterial 16S rRNA gene copy
15
numbers) significantly correlated with each other (Table S8). DNA concentrations reached mean values of 141.6 ng g-1 and
106.9 ng g-1 in C1 and C2, respectively, whereas the mean DNA concentration in C4 and C3 were 88.5 and 19.8 ng g-1 (Table
S9). Mean TCC reached a value of 5 x 107 g-1 in C1. C4 and C2 had similar values of 1.3 x 107 g-1 and 1.5 x 107 g-1, while
cell numbers of C3 were one order of magnitude lower (1.5 x 106 g-1). Bacterial 16S rRNA gene copy numbers usually
exceeded TCC by an order of magnitude, with mean values of 1.6 x 108 g-1 and 2.9 x 108 g-1 in C1 and C2, but lower mean
20
values of 3.6 x 107 g-1 and 1.7 x 107 g-1 in C4 and C3, respectively.
A correlation analysis (Table 1) revealed that microbial and bacterial abundance including DNA concentrations, 16S rRNA
bacterial gene copies and TCC showed a significant rank-based negative correlation with salinity (p < 0.05, Spearman -0.63
≤ rs ≤ -0.35), cations (K+, Mg2+, Na+) and anions (Cl-, Br-) (p < 0.05, -0.71 ≤ rs ≤ -0.39), and δ18O (p<0.05, -0.38 ≤ rs ≤ -0.37).
Furthermore, DNA concentrations negatively correlated with temperature (p < 0.05, rs = -0.37) and pH (p < 0.05, rs = -0.44),
25
while TCC negatively correlated with temperature (p < 0.01, rs = -0.64) and 16S rRNA gene copies with pH (p < 0.01, rs = -
0.24). Positive correlations were found for DNA and 16S rRNA gene copies with total organic carbon (TOC, p < 0.05, rs >
0.34) and the water content (p < 0.01, rs = 0.47).
3.3 Bacterial community composition
The most abundant bacterial taxa were Actinobacteria (class), Chloroflexi (Gitt-GS-136, KD4-96), Clostridia (class),
30
Gemmatimonadetes, and Proteobacteria (primarily Alpha- and Betaproteobacteria) (Fig. 5). Candidatus Aminicenantes
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
9
(candidate phylum OP8) and Candidatus Atribacteria (candidate phylum OP9) were highly abundant in core C3, where
Actinobacteria, Chloroflexi, and Gemmatimonadetes were almost absent.
In order to test for correlation between the bacterial community composition at each drill site with environmental parameters
like salinity and temperature, we performed Mantel tests (Table S11). We found no correlation with salinity, but a correlation
to temperature (p = 0.0001, correlation R = 0.25). However, the community formation was stronger influenced by stable
5
water isotopes (p = 0.0001, R = 0.40) and the sample depth (p = 0.0001, R = 0.36) in meters below sea floor (m bsf) and
below surface (m bs), respectively, than by temperature.
We included those environmental parameters that showed a significant correlation with microbial community composition in
a canonical correspondence analysis (CCA, Fig. 6). Accordingly, grouping patterns of the bacterial community based on the
OTU0.03 composition of the samples and the Bray-Curtis dissimilarity were visualized. The CCA showed a clustering of
10
samples according to their borehole location for C2 and C3, while communities of C1 and C4 were more scattered. Samples
located to the left side of the plot originated from a greater depth (C1 and C2) than samples to the right side (C3 and C4).
Variance of samples from the bottom left to the top right was explained by rising temperature, while variance of samples
from the top left to the bottom right are likely explained by decreasing values of the stable water isotopes δ18O and δD. The
bacterial community of C3 was most distinct and clustered furthest from communities of all other sites, and was linked with
15
stable water isotopes and sample depth. The variance between C1, C4 and C2 samples are explained by temperature
differences.
Despite the overlaps within the CCA ordination, a one-way PerMANOVA revealed that the variance between each of the
clusters was significantly higher than within single clusters (Table S12), i.e., the bacterial subpopulations of each drill site
were significantly different from each other.
20
4 Discussion
The present study aimed at understanding the effect of long-term permafrost warming independent of thaw on microbial
community composition and abundance. The observed significant negative rank-based correlation between increasing
temperature and total cell counts (TCC) contradicts our hypothesis that millennial-scale permafrost warming directly
increases microbial abundance. It is, however, in line with related studies on arctic and subarctic soil microbial communities
25
where a negative effect of increasing temperature on microbial abundance was assigned to freeze-thaw cycles (Schimel et al.,
2007; Skogland et al., 1988) and substrate depletion (Walker et al., 2018). Both effects are, however, unlikely here. Firstly,
sample depths were always more than 10 m below surface and sea floor, respectively, and freeze-thaw cycles within the
investigated Unit II can be excluded. Secondly, preservation, rather than depletion, of substrates was more likely in the two
submarine cores C3 and C4, where DOC contents were comparable to that of the cold terrestrial permafrost of C1 (Fig. S1).
30
The degradation of DOC can be used as measure for microbial carbon turnover (Seto and Yanagiya, 1983) and the DOC
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-144
Manuscript under review for journal Biogeosciences
Discussion started: 6 May 2019
c
Author(s) 2019. CC BY 4.0 License.
10
concentration usually correlates with microbial abundance (Junge et al., 2004; Smolander and Kitunen, 2002; Vetter et al.,
2010). The cores C3 and C4 had significantly lower TCC and bacterial gene copy numbers than the onshore core C1 and the
C2 core furthest offshore. Thus, microbial activity and substrate utilization were likely low in C3 and C4. A negative
influence of permafrost warming on microbial abundance is further challenged through some indication for microbial
proliferation in core C2, which had experienced longest warming of all cores. In detail, TCC in C2 were higher than in the
5
other submarine cores while DOC