Accepted Article
Received Date : 04-Dec-2013 Revised Date : 21-Mar-2014 Accepted Date : 28-Mar-2014 Article type
: Research Paper
Editor
: Tillmann Lueders
Bacterial community structure and dissolved organic matter in repeatedly flooded subsurface karst water pools
Tanja Shabarova1, Jörg Villiger1, Oleg Morenkov3, Jutta Niggemann2, Thorsten Dittmar2 & Jakob Pernthaler1*
1
Limnological Station, Institute of Plant Biology, University of Zurich, Seestrasse 187, CH-8802
Kilchberg, Switzerland 2
Max Planck Research Group for Marine Geochemistry, Institute for Chemistry and Biology of the Marine Environment (ICBM), University of Oldenburg, Carl-von-Ossietzky-Strasse 9-11, D-26129 Oldenburg, Germany 3
Hewlett-Packard Company, Schwarzenburgstrasse 160, CH-3097 Liebefeld, Switzerland
Running head: microbes and DOM in subsurface karst water pools
Key words: subsurface, microbial diversity, species sorting, dissolved organic matter, pyrosequencing, FT-ICR-MS
*corresponding author: Phone: 0041 634 92 10; Fax: 0041 634 92 25; Email: [email protected]
This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1111/1574-6941.12339 This article is protected by copyright. All rights reserved.
Accepted Article
Abstract Bacterial diversity, community assembly, and the composition of the dissolved organic matter (DOM) were studied in three temporary subsurface karst pools with different flooding regimes. We tested the hypothesis that microorganisms introduced to the pools during floods faced environmental filtering towards a ‘typical’ karst water community, and we investigated if DOM composition was related to floodings and the residence time of water in stagnant pools. As predicted, longer water residence consistently led to a decline of bacterial diversity. The microbial assemblages in the influx water harbored more ‘exotic’ lineages with large distances to known genotypes, yet these initial communities already appeared to be shaped by selective processes. Betaproteobacterial operational taxonomic units (OTUs) closely related to microbes from subsurface or surface aquatic environments were mainly responsible for the clustering of samples according to water residence time in the pools. By contrast, several Cytophagaceae and Flavobacteriaceae OTUs were related to different floodings, which were also the main determinants of DOM composition. A subset of compounds distinguishable by molecular mass and O/C content were characteristic for individual floods. Moreover, there was a transformation of DOM in stagnant pools towards smaller and more aromatic compounds, potentially also reflecting microbial utilization.
Introduction The subsurface aquatic karst environment is important for water resource protection and management (Ford & Williams, 2007). It represents a landscape of exclusive microbial habitats (Engel, 2010), some of which are readily accessible. While there is information about microbes in karst springs (Farnleitner et al., 2005, Pronk et al., 2009, Wilhartitz et al., 2009), in vadose (unsaturated) and phreatic (saturated) karst systems (Simon et al., 2001, Hunter et al., 2004, Gray & Engel, 2013), the epiphreatic (periodically flooded) realms are largely unexplored. Temporary epiphreatic pools harbor diverse microbial communities that are collected from karstic and karst-related ecosystems during floodings, and that transform during the interjacent periods, in parallel with decreasing concentrations of dissolved organic carbon (DOC) (Shabarova et al., 2013). Since the inocula to neighboring pools and the local physicochemical conditions are highly similar, these systems are models to assess the role of deterministic and stochastic processes for the development of planktonic bacterial assemblages (Stegen et al., 2012). Microbial communities in such habitats are shaped by different assembly processes related to temporal changes of local factors (often referred to as “species sorting”), or to microbes introduced by the influx, i.e. by “mass effects” (Langenheder et al., 2012, Shabarova et al., 2013). Depending on ecotone connectivity, inocula may originate from a single or various habitats (Fazi et al., 2008) and might, moreover, be transformed by the transport process itself. Selective local processes may also act upon microbial assemblages in temporary habitats, e.g., the quality and quantity of the dissolved organic matter (DOM) (Kirchman et al., 2004). Aquatic DOM is a complex mixture of compounds more or less accessible to microbial degradation (Amon & Benner, 1994), and that are transformed (or even de novo produced) by microbial metabolism (Stoderegger & Herndl, 1998, Berggren et al., 2009). Analytical techniques such as Fourier-Transform Ion Cyclotron Resonance Mass Spectrometry (FT-ICR-MS) provide information
This article is protected by copyright. All rights reserved.
Accepted Article
about the molecular composition of DOM. Due to the high mass accuracy molecular formulae can be assigned to individual detected masses within the complex DOM mixture (Kujawinski et al., 2002, Koch et al., 2005). Characteristic fingerprints, consisting of many thousand molecular formulae, allow the distinction of DOM of different origin (Koch et al., 2005), and to investigate its degradation (Gonsior et al., 2009). A multitude of parameters influence DOM composition in surface water habitats, e.g., allochthonous influx (Berggren et al., 2009) or light-dependent processes (Bertilsson & Tranvik, 2000, Hama & Yanagi, 2001). Thus, the role of microbes in DOM transformation might be most easily apprehended in natural aquatic systems that are temporarily stagnant and not exposed to light.
The 900 m deep Bärenschacht karst cave (Switzerland) provides perfect conditions for the study of subsurface pool systems (Shabarova & Pernthaler, 2010, Shabarova et al., 2013). It is a part of the well-known Siebenhengste cave system, comprising around 300 km of documented passages. The catchment area of this system is around 32 km2 and is characterized by high flow rates, reaching up to 540 m/h, resulting in 38 h flow time for 20 km distance (Häuselmann et al., 2003). The anual precipitation from snow and rain (nivo-pluvial regime) is between 1500 and 2000 mm per year. The geology of lower part of the area (up to 1700 m) consists of sandstone and flysch and of sandstone and limestone in the upper part; it allows for both, surface drainage with sinks and percolation and subsurface drainage. The most common landscape forms in the region are forests, bogs and karren fields. The lower part of Bärenschacht contains exits from the extended phreatic zone that induce flooding events in the cave shortly after rainfalls and snow melt. The rising of the phreatic water table up to 100 m leads to the progressive filling of normally dry realms through large karst conduits (Häuselmann et al., 2003). At the end of such events the water exits the epiphreatic zone through conduits aquifers of different size, leaving some pools in the unevenness of the bedrock relief (Fig.1). We recently presented information about the relationship between microbial community structure, DOC concentrations and flooding frequency in three such pools (Shabarova et al., 2013): Using classical sequencing of 16S rRNA genes we observed a concomitant decrease of DOC and of microbial diversity during stagnant conditions in one of the pools, which led us to hypothesize that these habitats might act as environmental filters on largely unstructured inflow assemblages. Unfortunately, a major phylogenetic lineage of karst water microbes, the Betaproteobacteria, had to be excluded from a community-wide molecular fingerprinting analyses in that study due to low resolution within this bacteria class. Thus it was not possible to address central questions related to community assembly, e.g., to distinguish between bacterial lineages that were favoured by residence in stagnant pools or by specific floodings. We, therefore performed additional molecular analyses on selected samples from these epiphreatic pools to (i) more comprehensively assess changes of microbial diversity in this endokarst system, to (ii) test the hypothesis of species sorting in several of the stagnant pools and using phylogenetically based indices, and (iii) to address the relationship between Betaproteobacteria and the residence time of water in the pools. In addition, the molecular composition of DOM was determined by FT-ICR-MS in order to distinguish between signals that were specifically related to floodings and changes due to the residence of water in the pools (potentially indicative of microbial DOM transformation).
This article is protected by copyright. All rights reserved.
Accepted Article
Materials and Methods Sampling procedure and hydrological monitoring Three shallow epiphreatic pools of the Bärenschacht cave systems that were hydrologically isolated during stagnant periods (‘Longs Couteaux’ [LC], ‘GaIery du Nord III’ [NGIII], and North Sump [NS]) were sampled during seven expeditions between Dec 16, 2009 and March 18, 2010. Material for the analysis of water chemistry, bacterial abundances and community structure were obtained at every sampling (see Shabarova et al 2013). Eleven additional water samples (1 l each) were obtained during samplings 1,2,3,5 and 7 in pool LC (subsequently referred to as LC 1, LC 2 etc)., during samplings 1,2,3,5 and 7 in pool NGIII, and during samplings 1 and 7 in pool NS. They were collected in Teflon bottles and frozen at -20°C to analyse the molecular DOM composition by FT-ICR-MS (Fig. 1). Eight duplicate bacterial biomass samples (LC 1, 7; NGIII 1, 3, 5, 7; NS 1, 7; Fig. 1) were filtered onto Sterivex-GP membrane filter units (Merck Millipore) in the cave for DNA extraction and subsequent Next Generation Sequencing (NGS). During the sampling period the water level of pools -indicating flooding events- was monitored by pressure loggers (Ingenieurbüro Ziegler GmbH) and an online reporting system (Cave-Link, Ingenieurbüro Ziegler GmbH; (Shabarova et al., 2013)). The residence time of water in the pools (time after the flood) was calculated as the difference between sampling time point and the time point when the last flood had subsided (i.e., the water table in the pool had reached pre-flooding levels). A complete exchange of water in the cave pools was assumed for all flooding events but one (3.04.2010), based on the observation that conduit-dominated systems such as Bärenschacht are best described by pipe flow models (Jeannin, 2001). This implies the complete renewal of water in the pools due to the high volumes passing through them during floods once flow is ensured (which was the case for all events but one [3.04.2010] according to a topographical analysis).
DNA extraction, pyrosequencing of 16S rRNA genes DNA extraction from the biomass samples prepared in the cave was performed within 2 days after sampling using the UltraClean® Water DNA isolation kit (MO BIO Laboratories, Inc.) (Shabarova et al., 2013). DNA extracts of 8 microbial biomass samples and of 4 replicates (LC 7, NGIII 7, and NS 1 and 7) were selected for 454 pyrosequencing (Roche FLX) of partial 16S rRNA genes using primers 341F (Muyzer et al., 1993) and 907R (Lane et al., 1985). Subsamples of 300 µl of DNA suspension (4-10 ng µl-1) were processed by Research and Testing Laboratory, Inc. (Lubbock, TX, USA).
Analysis of pyrosequencing data Raw data of the samples (95516 reads, mean per sample: 11940 [range 8891-19069], mean raw read length 571 base pairs, replicates not included) were processed by a custom-made pipeline realized in DELPHI on a 128 CPU core computing cluster. Flowgrams were denoised (Quince et al., 2011) by first truncating them at the second instance of a signal intensity between 0.5 and 0.7, indicative of a low quality position in a later segment of the flowgram, or if there were 4 consecutive flows with values 95% of total peak area. Masses detected in procedural blanks (212 of 5560 peaks), or only detected in single samples with peak areas 0.66), II - soil-derived polyphenols and PCAs with aliphatic chains (0.66 >= AImod > 0.50), III - soil-derived "humics", i.e. phenolic and highly unsaturated compounds (AImod H/C >= 1.5), V - saturated fatty and sulfonic acids, carbohydrates including amino sugars (H/C >= 2.0 or O/C >= 0.9), and VI - peptides (as previous group, but containing N).
Statistical analysis The number of shared OTUs between pairs of samples and θ similarity coefficients (Yue & Clayton, 2005) were calculated. Agglomerative hierarchical clustering (AHC) of relative abundances of OTUs and of DOM compounds was conducted by Pearson correlation (complete linkage algorithm) with the software XLSTAT 11 (Addinsoft). To assess the robustness of clustering results the matrices of θ similarity coefficients and of AHC were compared by Mantel tests. We identified the OTUs and molecular formulae responsible for the observed clustering patterns by a feature selection algorithm that implements a ‘random forest’ algorithm using the R package Boruta (Kursa & Rudnicki, 2010) (default confidence level, 0.999). To reduce noise, small OTUs and rare molecular formulae were excluded prior to this analysis, to the extent that the overall clustering patterns remained unaffected (Aagaard et al., 2012).
This article is protected by copyright. All rights reserved.
Accepted Article
Results Phylogenetic distance to known genotypes Almost half of all sequences from the pooled dataset (32027 of 68264) were found in
Received Date : 04-Dec-2013 Revised Date : 21-Mar-2014 Accepted Date : 28-Mar-2014 Article type
: Research Paper
Editor
: Tillmann Lueders
Bacterial community structure and dissolved organic matter in repeatedly flooded subsurface karst water pools
Tanja Shabarova1, Jörg Villiger1, Oleg Morenkov3, Jutta Niggemann2, Thorsten Dittmar2 & Jakob Pernthaler1*
1
Limnological Station, Institute of Plant Biology, University of Zurich, Seestrasse 187, CH-8802
Kilchberg, Switzerland 2
Max Planck Research Group for Marine Geochemistry, Institute for Chemistry and Biology of the Marine Environment (ICBM), University of Oldenburg, Carl-von-Ossietzky-Strasse 9-11, D-26129 Oldenburg, Germany 3
Hewlett-Packard Company, Schwarzenburgstrasse 160, CH-3097 Liebefeld, Switzerland
Running head: microbes and DOM in subsurface karst water pools
Key words: subsurface, microbial diversity, species sorting, dissolved organic matter, pyrosequencing, FT-ICR-MS
*corresponding author: Phone: 0041 634 92 10; Fax: 0041 634 92 25; Email: [email protected]
This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1111/1574-6941.12339 This article is protected by copyright. All rights reserved.
Accepted Article
Abstract Bacterial diversity, community assembly, and the composition of the dissolved organic matter (DOM) were studied in three temporary subsurface karst pools with different flooding regimes. We tested the hypothesis that microorganisms introduced to the pools during floods faced environmental filtering towards a ‘typical’ karst water community, and we investigated if DOM composition was related to floodings and the residence time of water in stagnant pools. As predicted, longer water residence consistently led to a decline of bacterial diversity. The microbial assemblages in the influx water harbored more ‘exotic’ lineages with large distances to known genotypes, yet these initial communities already appeared to be shaped by selective processes. Betaproteobacterial operational taxonomic units (OTUs) closely related to microbes from subsurface or surface aquatic environments were mainly responsible for the clustering of samples according to water residence time in the pools. By contrast, several Cytophagaceae and Flavobacteriaceae OTUs were related to different floodings, which were also the main determinants of DOM composition. A subset of compounds distinguishable by molecular mass and O/C content were characteristic for individual floods. Moreover, there was a transformation of DOM in stagnant pools towards smaller and more aromatic compounds, potentially also reflecting microbial utilization.
Introduction The subsurface aquatic karst environment is important for water resource protection and management (Ford & Williams, 2007). It represents a landscape of exclusive microbial habitats (Engel, 2010), some of which are readily accessible. While there is information about microbes in karst springs (Farnleitner et al., 2005, Pronk et al., 2009, Wilhartitz et al., 2009), in vadose (unsaturated) and phreatic (saturated) karst systems (Simon et al., 2001, Hunter et al., 2004, Gray & Engel, 2013), the epiphreatic (periodically flooded) realms are largely unexplored. Temporary epiphreatic pools harbor diverse microbial communities that are collected from karstic and karst-related ecosystems during floodings, and that transform during the interjacent periods, in parallel with decreasing concentrations of dissolved organic carbon (DOC) (Shabarova et al., 2013). Since the inocula to neighboring pools and the local physicochemical conditions are highly similar, these systems are models to assess the role of deterministic and stochastic processes for the development of planktonic bacterial assemblages (Stegen et al., 2012). Microbial communities in such habitats are shaped by different assembly processes related to temporal changes of local factors (often referred to as “species sorting”), or to microbes introduced by the influx, i.e. by “mass effects” (Langenheder et al., 2012, Shabarova et al., 2013). Depending on ecotone connectivity, inocula may originate from a single or various habitats (Fazi et al., 2008) and might, moreover, be transformed by the transport process itself. Selective local processes may also act upon microbial assemblages in temporary habitats, e.g., the quality and quantity of the dissolved organic matter (DOM) (Kirchman et al., 2004). Aquatic DOM is a complex mixture of compounds more or less accessible to microbial degradation (Amon & Benner, 1994), and that are transformed (or even de novo produced) by microbial metabolism (Stoderegger & Herndl, 1998, Berggren et al., 2009). Analytical techniques such as Fourier-Transform Ion Cyclotron Resonance Mass Spectrometry (FT-ICR-MS) provide information
This article is protected by copyright. All rights reserved.
Accepted Article
about the molecular composition of DOM. Due to the high mass accuracy molecular formulae can be assigned to individual detected masses within the complex DOM mixture (Kujawinski et al., 2002, Koch et al., 2005). Characteristic fingerprints, consisting of many thousand molecular formulae, allow the distinction of DOM of different origin (Koch et al., 2005), and to investigate its degradation (Gonsior et al., 2009). A multitude of parameters influence DOM composition in surface water habitats, e.g., allochthonous influx (Berggren et al., 2009) or light-dependent processes (Bertilsson & Tranvik, 2000, Hama & Yanagi, 2001). Thus, the role of microbes in DOM transformation might be most easily apprehended in natural aquatic systems that are temporarily stagnant and not exposed to light.
The 900 m deep Bärenschacht karst cave (Switzerland) provides perfect conditions for the study of subsurface pool systems (Shabarova & Pernthaler, 2010, Shabarova et al., 2013). It is a part of the well-known Siebenhengste cave system, comprising around 300 km of documented passages. The catchment area of this system is around 32 km2 and is characterized by high flow rates, reaching up to 540 m/h, resulting in 38 h flow time for 20 km distance (Häuselmann et al., 2003). The anual precipitation from snow and rain (nivo-pluvial regime) is between 1500 and 2000 mm per year. The geology of lower part of the area (up to 1700 m) consists of sandstone and flysch and of sandstone and limestone in the upper part; it allows for both, surface drainage with sinks and percolation and subsurface drainage. The most common landscape forms in the region are forests, bogs and karren fields. The lower part of Bärenschacht contains exits from the extended phreatic zone that induce flooding events in the cave shortly after rainfalls and snow melt. The rising of the phreatic water table up to 100 m leads to the progressive filling of normally dry realms through large karst conduits (Häuselmann et al., 2003). At the end of such events the water exits the epiphreatic zone through conduits aquifers of different size, leaving some pools in the unevenness of the bedrock relief (Fig.1). We recently presented information about the relationship between microbial community structure, DOC concentrations and flooding frequency in three such pools (Shabarova et al., 2013): Using classical sequencing of 16S rRNA genes we observed a concomitant decrease of DOC and of microbial diversity during stagnant conditions in one of the pools, which led us to hypothesize that these habitats might act as environmental filters on largely unstructured inflow assemblages. Unfortunately, a major phylogenetic lineage of karst water microbes, the Betaproteobacteria, had to be excluded from a community-wide molecular fingerprinting analyses in that study due to low resolution within this bacteria class. Thus it was not possible to address central questions related to community assembly, e.g., to distinguish between bacterial lineages that were favoured by residence in stagnant pools or by specific floodings. We, therefore performed additional molecular analyses on selected samples from these epiphreatic pools to (i) more comprehensively assess changes of microbial diversity in this endokarst system, to (ii) test the hypothesis of species sorting in several of the stagnant pools and using phylogenetically based indices, and (iii) to address the relationship between Betaproteobacteria and the residence time of water in the pools. In addition, the molecular composition of DOM was determined by FT-ICR-MS in order to distinguish between signals that were specifically related to floodings and changes due to the residence of water in the pools (potentially indicative of microbial DOM transformation).
This article is protected by copyright. All rights reserved.
Accepted Article
Materials and Methods Sampling procedure and hydrological monitoring Three shallow epiphreatic pools of the Bärenschacht cave systems that were hydrologically isolated during stagnant periods (‘Longs Couteaux’ [LC], ‘GaIery du Nord III’ [NGIII], and North Sump [NS]) were sampled during seven expeditions between Dec 16, 2009 and March 18, 2010. Material for the analysis of water chemistry, bacterial abundances and community structure were obtained at every sampling (see Shabarova et al 2013). Eleven additional water samples (1 l each) were obtained during samplings 1,2,3,5 and 7 in pool LC (subsequently referred to as LC 1, LC 2 etc)., during samplings 1,2,3,5 and 7 in pool NGIII, and during samplings 1 and 7 in pool NS. They were collected in Teflon bottles and frozen at -20°C to analyse the molecular DOM composition by FT-ICR-MS (Fig. 1). Eight duplicate bacterial biomass samples (LC 1, 7; NGIII 1, 3, 5, 7; NS 1, 7; Fig. 1) were filtered onto Sterivex-GP membrane filter units (Merck Millipore) in the cave for DNA extraction and subsequent Next Generation Sequencing (NGS). During the sampling period the water level of pools -indicating flooding events- was monitored by pressure loggers (Ingenieurbüro Ziegler GmbH) and an online reporting system (Cave-Link, Ingenieurbüro Ziegler GmbH; (Shabarova et al., 2013)). The residence time of water in the pools (time after the flood) was calculated as the difference between sampling time point and the time point when the last flood had subsided (i.e., the water table in the pool had reached pre-flooding levels). A complete exchange of water in the cave pools was assumed for all flooding events but one (3.04.2010), based on the observation that conduit-dominated systems such as Bärenschacht are best described by pipe flow models (Jeannin, 2001). This implies the complete renewal of water in the pools due to the high volumes passing through them during floods once flow is ensured (which was the case for all events but one [3.04.2010] according to a topographical analysis).
DNA extraction, pyrosequencing of 16S rRNA genes DNA extraction from the biomass samples prepared in the cave was performed within 2 days after sampling using the UltraClean® Water DNA isolation kit (MO BIO Laboratories, Inc.) (Shabarova et al., 2013). DNA extracts of 8 microbial biomass samples and of 4 replicates (LC 7, NGIII 7, and NS 1 and 7) were selected for 454 pyrosequencing (Roche FLX) of partial 16S rRNA genes using primers 341F (Muyzer et al., 1993) and 907R (Lane et al., 1985). Subsamples of 300 µl of DNA suspension (4-10 ng µl-1) were processed by Research and Testing Laboratory, Inc. (Lubbock, TX, USA).
Analysis of pyrosequencing data Raw data of the samples (95516 reads, mean per sample: 11940 [range 8891-19069], mean raw read length 571 base pairs, replicates not included) were processed by a custom-made pipeline realized in DELPHI on a 128 CPU core computing cluster. Flowgrams were denoised (Quince et al., 2011) by first truncating them at the second instance of a signal intensity between 0.5 and 0.7, indicative of a low quality position in a later segment of the flowgram, or if there were 4 consecutive flows with values 95% of total peak area. Masses detected in procedural blanks (212 of 5560 peaks), or only detected in single samples with peak areas 0.66), II - soil-derived polyphenols and PCAs with aliphatic chains (0.66 >= AImod > 0.50), III - soil-derived "humics", i.e. phenolic and highly unsaturated compounds (AImod H/C >= 1.5), V - saturated fatty and sulfonic acids, carbohydrates including amino sugars (H/C >= 2.0 or O/C >= 0.9), and VI - peptides (as previous group, but containing N).
Statistical analysis The number of shared OTUs between pairs of samples and θ similarity coefficients (Yue & Clayton, 2005) were calculated. Agglomerative hierarchical clustering (AHC) of relative abundances of OTUs and of DOM compounds was conducted by Pearson correlation (complete linkage algorithm) with the software XLSTAT 11 (Addinsoft). To assess the robustness of clustering results the matrices of θ similarity coefficients and of AHC were compared by Mantel tests. We identified the OTUs and molecular formulae responsible for the observed clustering patterns by a feature selection algorithm that implements a ‘random forest’ algorithm using the R package Boruta (Kursa & Rudnicki, 2010) (default confidence level, 0.999). To reduce noise, small OTUs and rare molecular formulae were excluded prior to this analysis, to the extent that the overall clustering patterns remained unaffected (Aagaard et al., 2012).
This article is protected by copyright. All rights reserved.
Accepted Article
Results Phylogenetic distance to known genotypes Almost half of all sequences from the pooled dataset (32027 of 68264) were found in
