Introduction

The Baltic Sea is the world's second largest brackish basin. The Baltic proper comprises a number of deep areas, of which the Gotland Deep is the largest. These areas are characterized by periodically anoxic conditions in their bottom waters. A stable halocline below 50–60 m separates the water column into an upper oxygenated layer and an underlying oxygen-deficient and anoxic/sulfidic layer (Lepland and Stevens, 1998; Neretin et al., 2003), resulting in a vertically stratified redoxcline that is responsible for a distinct prokaryotic stratification (Rheinheimer et al., 1989; Labrenz et al., 2007). For example, high dark CO2 fixation is consistently found in the upper sulfidic layer (Jost et al., 2010) and is driven mainly by chemolithoautotrophic Epsilonproteobacteria (Grote et al., 2008; Glaubitz et al., 2009). At the interface between nitrate and reduced sulfur compounds, Epsilonproteobacteria are also responsible for chemoautotrophic denitrification (Labrenz et al., 2005; Brettar et al., 2006), which represents probably the most important N loss process in the central Baltic Sea (Hannig et al., 2007). If the accordant nitrate is not laterally introduced, then it must be produced by nitrification.

Studying the oxic–anoxic interface of the Gotland Deep from 1998 to 2000, Bauer (2003) determined mean nitrification rates of 202 nmol NH4+ L−1 d−1. In that study, based on 16S rRNA as well as bacterial ammonia monooxygenase (amoA, the key enzyme in aerobic ammonia-oxidizing bacteria) gene analyses, only Betaproteobacteria phylogenetically related to Nitrosomonas or Nitrosospira could be identified as putative nitrifying organisms. However, the highest cell numbers determined by fluorescence in situ hybridization (FISH) corresponded to only 0.6–3.9% of total prokaryotic cell counts. In fact, until recently nitrification was thought to be exclusively a bacterial feature, but an amoA-like gene has been detected on an insert in a fosmid clone library from soil also carrying a crenarchaeotal 16S rRNA gene, thus demonstrating the nitrifying potential of non-thermophilic Crenarchaeota for the first time (Schleper et al., 2005; Treusch et al., 2005). In parallel, amoA genes on a presumptive archaeal metagenome fragment from the Sargasso Sea study were detected (Venter et al., 2004), and later on also in the marine sponge symbiont Cenarchaeum symbiosum (Hallam et al., 2006). The only free-living cultivated representative of the marine pelagic Crenarchaeota is still C. Nitrosopumilus maritimus, which is known to be able to oxidize ammonium aerobically to nitrite (Könneke et al., 2005; Martens-Habbena et al., 2009). Phylogenetically, C. N. maritimus is a member of marine group I (MGI) (Schleper et al., 2005). MGI Crenarchaeota are widely distributed throughout the mesopelagic ocean, reaching cell abundances accounting for up to 30% of the oceanic picoplankton (Massana et al., 2000; Karner et al., 2001; Church et al., 2009). While a considerable fraction of the MGI was suggested to be autotrophic (Herndl et al., 2005; Ingalls et al., 2006), MICRO-CARD-FISH (Herndl et al., 2005; Teira et al., 2006) and archaeal amoA analyses (Agogué et al., 2008; De Corte et al., 2008) demonstrated that members of this group are not restricted to an autotrophic lifestyle. In any case, it is likely that non-thermophilic Crenarchaeota have a significant role in the global nitrogen cycle (Francis et al., 2005; Wuchter et al., 2006; Erguder et al., 2009).

Recently, the relevance of nitrifying Crenarchaeota and Bacteria was also shown for the suboxic zones of Black Sea redoxclines. Applying gene abundance and expression analyses, reaction-diffusion modeling, and 15N incubation experiments, Lam et al. (2007) identified ammonium-oxidizing Crenarchaeota as well as gammaproteobacterial ammonia-oxidizing bacteria as important nitrifiers and proposed that nitrification is coupled to anammox (anaerobic ammonia oxidation). 16S rRNA gene analyses by Coolen et al. (2007) identified different crenarchaeotal phylotypes throughout the Black Sea suboxic zone. The importance of nitrifying organisms, especially in the suboxic zones of Black Sea redoxclines, is also supported by biomarker analyses indicating the presence of Crenarchaeota or nitrifying bacteria (Wakeham et al., 2007; Schubotz et al., 2009).

In principle, pelagic redoxclines of the central Baltic Sea are physicochemically comparable to those of the Black Sea, as these specific habitats share several important microbial organisms involved in the C-, N- and S-cycle (Labrenz et al., 2007). However, in contrast to the Black Sea, central Baltic Sea redoxclines are irregularly oxygenated after inflow events of saline- and oxygen-rich North Sea water (Reissmann et al., 2009), albeit eventually becoming re-established. Additional disturbances include small-scale turbulence, which can produce local mixing events (Lass et al., 2003). Thus, the extended suboxic zones characteristic of the Black Sea (Yakushev et al., 2008) are absent in Baltic redoxclines. Nevertheless, 16S rRNA fingerprinting analyses has indicated that MGI Crenarchaeota are also prominent in suboxic zones of the central Baltic Sea (Labrenz et al., 2007).

The main objective of this study was to assess the diversity and abundance of nitrifying assemblages throughout a central Baltic Sea redoxcline. Although amoA gene and transcript analyses failed to provide evidence of bacterial ammonium-oxidizing capacities, one highly abundant nitrifying member of the MGI Crenarchaeota was instead detected.

Materials and methods

The complete Materials & Methods part is given as Supplementary information.

Sampling

Water samples were obtained from the central Baltic Sea onboard the RV Maria S. Merian in February 2006 (Baltic Sea monitoring station 271, Figure 1). Samples from different depths around the chemocline, defined as the uppermost presence of sulfide, were collected in free-flow bottles attached to a conductivity, temperature and depth-rosette.

Figure 1
figure 1

Map of the central Baltic Sea and position of the sampling station 271 (57°19.2′N; 20°03′E, Gotland Deep).

Full size image

Physicochemical structure of the water column

Chemical profiles of oxygen, hydrogen sulfide, ammonia, phosphate, nitrite and nitrate were determined as described elsewhere (Grasshoff et al., 1983).

Bacterioplankton sampling

For catalyzed reporter deposition fluorescence in situ hybridization (CARD-FISH), 100 ml of water samples were directly transferred to glass bottles and fixed with particle-free formaldehyde (2% final concentration) at 4 °C for 2–5 h. Portions of 40–45 ml were filtered onto white polycarbonate membrane filters (Millipore GTTP, 0.2-μm pore size, 47-mm diameter, Molsheim, France), which were then rinsed with sterile seawater, air dried and stored at −80 °C until further processing. For nucleic acid extraction, water samples of 1–2 l were filtered onto white Durapore filters (Millipore GVWP, 0.22-μm pore size, 47-mm diameter) and stored frozen at −80 °C.

Nucleic acid extraction and complementary DNA (cDNA) synthesis

RNA and DNA were extracted from filters using an acidic phenol extraction protocol (Weinbauer et al., 2002). Before RT-PCR, RNA-containing extracts were purified from DNA by incubation with DNAse I (DNA-free-Kit, Ambion, Austin, TX, USA or amplification-grade enzyme, Invitrogen Corporation, CA, USA) at 37 °C for 30 min and the RNA concentrations determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). To retrieve cDNA the iScript cDNA Synthesis-Kit (Bio-Rad Laboratories, Hercules, CA, USA), the Access One-Step-RT-PCR kit (Promega, Madison, WI, USA), or Superscript III First-Strand Synthesis supermix (Invitrogen) were used following the protocols of the respective manufacturer. In each reverse transcription reaction, RNA samples without or with denaturized reverse transcriptase were used as controls in the PCR.

16S rRNA/rRNA gene-based diversity analyses

Single-strand conformation polymorphism (SSCP) fingerprinting. Archaeal w036/039-primers (Table 1) were used for 16S rcDNA and 16S rRNA gene amplification in samples obtained from eight different depths throughout the redoxcline (Figure 2). PCR, consisting of a total of 25 (cDNA) or 35 (DNA) cycles, was done following the method of Leclerc et al. (2001). Single-stranded DNA was generated and purified and SSCP analysis was carried out according to Schwieger and Tebbe (1998). Individual bands excised from the SSCP gels were re-amplified as described by Pöhler et al. (2002). PCR products were purified using the MiniElute kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol and sequenced by Seqlab (Göttingen, Germany).

Table 1 Primers used in this study
Full size table
Figure 2
figure 2

Depth profile throughout the water column of the Gotland Deep (station 271) in February 2006. (a) H2S, O2, PO43−; (b) NH4+, NO2− and NO3−. The dotted line denotes the chemocline. Sampling depths for molecular studies: (a) clone libraries (119, 122, 152 m); (b) SSCP fingerprinting (102, 109, 117, 122, 123, 124, 132, 152 m); (c) DGGE fingerprinting (111, 113, 115, 119, 121, 124, 130, 134 m).

Full size image

16S rRNA/rRNA gene clone library construction and RFLP. Archaeal 16S rRNA clone libraries were prepared from genomic DNA or RNA extracts originating from the suboxic zone, suboxic-sulfidic interface, and sulfidic zone (Figure 2). The primer systems Arch21F/1492R, w036/1492R and Arch21F/w039 (Table 1) were used to amplify 16S rRNA for PCR, which was performed as described above. The PCR products were purified also as described above and cloned following the manufacturer's instructions using the pGEM-T-Easy Vector system (Promega) and competent E. coli JM109 cells. For restriction fragment length polymorphism (RFLP) and sequencing, the inserted fragment was PCR-amplified with the vector-specific primers T7 and SP6 (Table 1). The unpurified PCR products were digested with restriction enzymes Hin6I and MspI and documented as described previously (Bond et al., 2000). Identical RFLP patterns were grouped and representative cloned fragments sequenced by Seqlab using the primers Arch21F, w036, w069 or 1492R.

Phylogenetic analysis. 16S rRNA/rRNA gene sequences were examined for accuracy using the software SeqMan (DNAstar) and checked for chimeras using the Bellerophon program Huber et al. (2004). Phylogenetic affiliations of the partial sequences were initially estimated with the program BLAST (Altschul et al., 1990). Sequences were aligned using the ARB software package (Ludwig et al., 2004) and those with a similarity >99% were grouped together. Basic trees were reconstructed based on related public sequences and generated clones of approximately 1300 nucleotides. These were reduced to 1284 unambiguously alignable positions using group-specific filters. An evolutionary-distance dendrogram was constructed using neighbor-joining (NJ), maximum likelihood (ML) and maximum parsimony (MP) algorithms. Partial SSCP sequences were individually imported using ARB parsimony and group-specific filters.

AmoA mRNA/amoA gene-based diversity analyses

DGGE fingerprinting. Ten nanograms of DNA or 20 ng of RNA, both originating from eight different depths throughout the redoxcline, were used as template for the PCR (Figure 2). Seven of these depths were slightly different from those depths used for the 16S rRNA-based analysis, but still from the same biogeochemical zone. The method of Coolen et al. (2007) was followed, with 25 (cDNA) and 35 (DNA) cycles in the PCR together with the archaeal amoA primers ammonia-oxidizing Archaea (AOA)-amoA-f and AOA-amoA-r (Table 1). DGGE was performed according to Coolen et al. (2004), using 9% polyacrylamide gels (ratio of acrylamide to bisacrylamide, 37.5:1) submerged in 0.5 × TAE buffer (20 mmol l−1 Tris, 20 mmol l−1 acetic acid, 0.5 mol l−1 EDTA; pH 7.4) at 60 °C. Individual bands excised from the DGGE gels were re-amplified using the above described amoA PCR protocol, purified using the MiniElute kit (Qiagen) as described by the manufacturer, and sequenced by Seqlab.

AmoA clone libraries. Archaeal and bacterial amoA clone libraries were prepared from extracts containing 10 ng of genomic DNA or 20 ng of RNA originating from the suboxic zone, suboxic-sulfidic interface and sulfidic zone (Figure 2). For PCR, the primer systems Arch-amoAF and Arch-amoAR (archaeal (Table 1; Francis et al., 2005)), AmoA-1F and AmoA-2R (β-proteobacterial (Rotthauwe et al., 1997)), and AmoA-3F and AmoB-4R (γ-proteobacterial (Purkhold et al., 2000)) were used (Table 1), with a total of 25 (cDNA) or 30 (DNA) PCR cycles. Cloning and sequencing steps were done as described for the 16S rRNA procedure.

Phylogenetic analyses. The procedures were the same as those described for 16S rRNA. Phylogenetic trees were reconstructed based on sequence information from DGGE bands and clones, and related to public sequences. These were reduced to unambiguously alignable positions using group-specific filters, resulting in 166 valid characters. An evolutionary-distance dendrogram was constructed using NJ, ML and MP.

Gene probe design

Based on the above-described 16S rRNA and rRNA gene clone libraries, a probe directed against crenarchaeotal cells of the suboxic zone was designed with the PROBE_FUNCTION tool of the ARB package. Probe specificity was validated with the PROBE_MATCH tool of the ARB package, the SILVA database (Pruesse et al., 2007) and BLAST. Horseradish peroxidase (HRP)-labeled probes were synthesized by Biomers (Ulm, Germany). The specificity of the newly designed probe S-*-Cren-0679-a-A-18 (Table 2), which was named according to the Oligonucleotide Probe Database nomenclature (Alm et al., 1996), was experimentally tested by whole-cell hybridization with different archaeal strains. With only one mismatch at the fourth base position of the gene probe, C. ‘Nitrosopumilus maritimus’ SCM1 served as positive control. With 9 and 7 mismatches, respectively, Halorubrum lacusprofundi (DSM5036T) and Pyrococcus furiosus (DSM3638T) served as negative controls (Table 2). All strains were cultivated following the recommendations of the DSMZ (Braunschweig, Germany).

Table 2 Probe sequence, appropriate target sites and corresponding sequences in other Archaea, including reference organisms
Full size table

Cellular abundances

Total cell numbers were determined by flow cytometric analyses as described previously (Labrenz et al., 2007). All HRP-labeled probes, appropriate sequences, target sites and hybridization conditions used in this study are described in Table 1 of the Supplementary information. Non-specific binding was determined using the NonEUB probe. By varying the formamide concentration, the hybridization conditions for the newly designed probe Cren679 were optimized (formamide concentrations tested: 0, 15, 20, 25, 30, 35, 50 and 60%). A concentration of 35% formamide in the hybridization buffer was found to be optimal for probe specificity.

The euryarchaeotal CARD-FISH enumeration was carried out according to the proteinase K protocol of Teira et al. (2004), but this method was not successful for the Cren537 and Cren679 gene probes. Therefore, the crenarchaeotal cell wall was permeabilized and the subsequent steps carried out as described by Sekar et al. (2003). Specifically, filter sections were washed in pre-warmed washing buffer and the tyramide signal was amplified with 5-(and 6-) carboxyfluorescein-labeled tyramides. The preparations were counterstained with a previously described mixture of 4′, 6′-diamidino-2-phenylindol (DAPI), Citifluor and VectaShield (Pernthaler et al., 2002). Stained filter sections were examined with an epifluorescence microscope (Axioskop 2 mot plus, Zeiss) as described previously (Grote et al., 2007). Ratios of gene probe versus DAPI-stained cells were determined. Group-specific cell numbers were calculated by adjusting the ratios to flow cytometry-based total cell numbers.

Quantification of amoA gene transcripts by real-time PCR

Quantitative PCR (qPCR) standards were prepared as described by Agogué et al. (2008). All qPCR analyses were conducted on an iCycler iQ5 thermocycler (Bio-Rad). Archaeal or β-proteobacterial amoA transcripts were determined in triplicate on non-diluted or 1:5 and 1:25 diluted cDNA and RNA samples. The methods of Coolen et al. (2006) and Agogué et al. (2008) were used for the qPCR, together with the primers Arch-amoA-for and Arch-amoA-rev and the primers amoA-1F and amoAr NEW (Table 1). AmoA abundance of the samples were calculated by simultaneously quantifying the standard, prepared as 10-fold serial dilutions ranging from 107 to 101 copies.

Nucleotide sequence accession numbers

Sequences were deposited in the GenBank database under accession numbers GU386315 to GU386337, and GU455423.

Results

Physicochemical structure of the redoxcline

The pelagic redoxcline was established from approximately 110 m to 130 m. Sulfide was detectable from a depth of 122 m onwards and increased to 17 μmol l−1 at a depth of 150 m (Figures 2a and b). At the chemocline, oxygen was still detectable and was present at concentrations of 3 μmol l−1, with an overlap of O2 and H2S at lower concentrations measurable at a depth ranging from 122 m to 130 m (Figure 2a). Phosphate concentrations were generally between 2.2 and 3.3 μmol l−1; however, at 119 m, they dropped to 1.2 μmol l−1 (Figure 2a). Nitrate concentrations decreased sharply between 110 and 120 m but, together with nitrite, peaked at 122 m. Ammonium increased strongly below 120 m, reaching a maximum value of 4.2 μmol l−1 at 150 m (Figure 2b).

Gene-based diversity

Archaeal 16S rRNA/rRNA gene SSCP fingerprinting revealed 10 different and identifiable bands for the whole redoxcline. In general, neither distinct differences between 16S rRNA and 16S rRNA gene fingerprints nor pronounced differences in the archaeal composition throughout the redoxcline were detected for bands of highest relative abundance (Figure 3a; Supplementary information Figure 1A, B). Others, as SSCP bands 7 and 11 or SSCP band 5 were more pronounced in the sulfidic or in the suboxic zone, respectively. Phylogenetically, most of the SSCP bands belonged to the Euryarchaeota, with SSCP bands 1, 6, 7, 10, 11 and 12 representing a predominantly Baltic-Sea-specific cluster. SSCP bands 3 and 4 were members of the crenarchaeotal MGI and closely related to C. Nitrosopumulus maritimus, but constituting a specific phylogenetic cluster (Figure 4a).

Figure 3
figure 3

Genetic fingerprints of samples covering the pelagic redoxcline. Sequenced bands are labeled by numbers. The phylogenetic affiliation of these bands is presented in Figure 4a and b. (a) Archaeal 16S rRNA/rRNA gene single-strand conformation polymorphism (SSCP); (b) Crenarchaeotal amoA mRNA/amoA gene denaturing gradient gel electrophoresis (DGGE). S, standard; R, cDNA; D, DNA.

Full size image
Figure 4
figure 4

Phylogenetic relationships of representative sequences from the studied redoxcline samples. The evolutionary distance dendrogram was reconstructed using the Jukes-Cantor correction and NJ. Branching points supported by NJ, ML and MP algorithms are marked by a filled circle. Branching points supported by two algorithms are marked by an open circle. (a) Unrooted 16S rRNA sequence-based phylogenetic tree of members of the domain Archaea. (b) Unrooted archaeal amoA mRNA sequence-based phylogenetic tree of members of the Crenarchaeota.

Full size image

From 16S rRNA/rRNA gene clone libraries based on different archaeal primer combinations, 91 clones were evaluated by RFLP analyses, but only one operational taxonomic unit could be detected for each of the three primer pairs. Subsequent sequencing of eight representative clones revealed that these were nearly 100% identical; thus, clone 9 was chosen as a representative sequence. Phylogenetically, clone 9 belonged to crenarchaeotal MGI, clustering with SSCP bands 3 and 4 and named Gotland Deep cluster 2 (GD2 (Figure 4a)). GD2 was 2% distant to C. N. maritimus. Euryarchaeotal 16S rRNA clones were not detected.

Archaeal amoA transcript and gene DGGE fingerprinting revealed eight distinct bands for the whole redoxcline (Figure 3b). No distinct differences between amoA transcript and gene fingerprints were detected (Figure 3b). Phylogenetically, all amoA sequences were related to C. Nitrosopumilus, but they again formed a distinct cluster (Figure 4b). With the exception of DGGE_band_7, all generated DGGE sequences were identical on the amino acid level. Their distances to DGGE_band_7 and C. Nitrosopumilus amoA were 2 and 3% on the amino acid and 3 and 8% on the nucleic acid level, respectively.

Of the clones generated by amoA transcript and gene clone libraries, 123 were evaluated by RFLP analyses, 9 OTUs identified and 31 representative clones finally sequenced. Comparable to the DGGE results, most of the clone sequences were identical or at least highly similar to each other (>98.5%). On the amino acid level, all amoA sequences were identical, with clones 14 and 15 chosen as representative sequences. Phylogenetically, these were similarly related to C. Nitrosopumilus and grouped with the amoA sequences generated by DGGE fingerprinting (Figure 4b).

β- or γ-proteobacterial amoA cDNA or genes could not be amplified.

Gene probe design

The group-specific rRNA-targeting oligonucleotide probe S-*-Cren-0679-a-A-18 (abbreviated as Cren679) was designed based on the eight nearly identical full-length crenarchaeotal 16S rRNA clones. Probe details and the number of mismatches with cultured organisms or related clone sequences are presented in Table 2. Cren679 was specific for all generated clones, SSCP_band_4, and, according to the SILVA database of August 2009, 44 other members of the GD2 cluster of organisms (Figure 4a). Within the target region of SSCP_band_3, two mismatches were found; however, it is unclear whether cells corresponding to SSCP_band_3 are detectable with Cren679. Despite one mismatch with C. N. maritimus SCM1 (Table 2), binding with this strain was recorded at all formamide concentrations tested, enabling the use of SCM1 as a positive control for the optimization procedure. Taking the C. N. maritimus mismatch into account, gene probe Cren679 matched 838 out of 1712 MGI sequences included in the SILVA database of August 2009, but only one crenarchaeotal sequence (out of 4540) outside this group. Therefore, the specificity of gene probe Cren679 should be described as detecting members of the C. Nitrosopumilus and MGI cluster. In situ, cells detected by Cren679 showed a peanut-shaped cell morphology (Figure 5), similar to the morphology of C. N. maritimus cells (Könneke et al., 2005). The probe did not hybridize with the negative control strains Pyrococcus furiosus (DSM3638T) and Halorubrum lacusprofundi (DSM5036T).

Figure 5
figure 5

Photomicrographs of DAPI- or CARD-FISH-stained microorganisms from the suboxic zone of the Gotland Deep redoxcline. Bar denotes 20 μm. (a) DAPI-stained cells; (b) Cells from the same microscopic field stained with the gene probe Cren679 exhibit a peanut-shaped morphology.

Full size image

Abundance of prokaryotes and specific groups

Total prokaryotic cell numbers (TCN) varied between 0.5 and 1.3 × 106 cells ml−1, with the highest numbers reached below the chemocline (Figure 6a). Bacterial cell numbers followed the prokaryotic pattern, with 60–80% EUB338-positive cells below the chemocline. Total crenarchaeotal cell numbers varied between 0.8 and 2.3 × 105 cells ml−1, reaching highest numbers 1 m above the chemocline. In this zone, 25–26% of the TCN were Cren537-positive. Below the chemocline, crenarchaeotal cells remained relatively stable, with abundances of 1.5–1.9 × 105 cells ml−1 that decreased only slightly in the deeper sulfidic zone. Above the suboxic zone, Cren537 cell numbers declined to less than 1.0 × 105 cells ml−1. Cren679-positive cells closely followed the total crenarchaeotal distribution (Figure 6a). Euryarchaeota could not be detected by CARD-FISH analyses.

Figure 6
figure 6

Prokaryotic cell numbers and amoA mRNA abundances along the Gotland Deep redoxcline in February 2006. The dashed horizontal line indicates the chemocline. (a) Vertical distribution of total cell numbers (DAPI counts), bacterial abundance, crenarchaeotal, and Cren679 abundance. (b) Vertical distribution of amoA mRNA abundances and theoretical amoA mRNA abundances per Cren679 cell.

Full size image

In a second profile of the Gotland Deep obtained in February 2006, the highest Cren679 abundance of 2.1 × 105 cells ml−1, accounting for 26% of the TCN, was also determined in the suboxic zone, where concentrations of 4 μmol O2 l−1 and 0.3 μmol H2S l−1 were measured (data not shown). As in the first profile, Cren679-positive cells closely followed crenarchaeotal abundance (Cren537-positive cells) and vertical distribution, whereas Euryarchaeota were not detectable (data not shown).

Quantification of amoA transcripts throughout the redoxcline

Comparable to the amoA cloning results, no β-proteobacterial but only archaeal amoA transcripts were amplifiable using RT-qPCR primer combinations. The highest number of archaeal amoA transcripts (6.5 × 103 ml−1) coincided with the highest number of Cren679-positive cells in the suboxic zone (Figures 6a and b). Calculation of the ratios of amoA transcript numbers per Cren679 cell showed the highest ratios (0.04) in the suboxic zone and generally following the distribution of number of transcripts per ml−1 (Figure 6b).

Discussion

Taken together, our data set shows that not only the nitrifying community but also the crenarchaeotal assemblage itself was characterized by an astonishing low diversity (Figure 4), consisting of only one crenarchaeotal subcluster (or potentially even a single species), which we named subgroup GD2. Phylogenetically, GD2 is closely related to, but clearly separated from, C. Nitrosopumilus maritimus (Figure 4a). GD2 abundance was highest in the suboxic zone, where it accounted up to 26% of the total prokaryotic abundance, but relatively high cell numbers were also detected in the deeper sulfidic waters (Figure 6a). Based on maximal archaeal amoA transcript abundances, archaeal ammonia oxidation occurred in a narrow suboxic zone close to the chemocline, where ammonium, oxygen, nitrate, nitrite and phosphate concentrations were below 5 μmol l−1. The observation that the crenarchaeal assemblage was almost exclusively composed of the GD2 cluster leads to the conclusion that these cells were also responsible for amoA expression. This corroborates results obtained for the isolate C. N. maritimus, identified as an ammonia oxidizer (Könneke et al., 2005; Martens-Habbena et al., 2009).

Archaeal diversity

According to the results of both 16S rRNA and rRNA gene SSCP fingerprinting, members of the Euryarchaeota and Crenarchaeota were present throughout the pelagic redoxcline (Figures 3a and 4a). Euryarchaeotal diversity was higher than that of Crenarchaeota, but summing up relative cren- and euryarchaeotal SSCP band areas as an indicator for accordant abundances, these were principally comparable throughout the redoxcline (Supplementary information Figure 1c and d). Moreover, potentially most abundant members of both phyla (SSCP bands 4 and 9) seemed to be uniformly distributed from the suboxic to the sulfidic zone, without pronounced differences between 16S rRNA and 16S rRNA gene fingerprints (Figure 3a; Supplementary information Figure 1a and b). The observed differences between eury- and crenarchaeotal diversities in the Baltic Sea might be a common phenomenon and is consistent with the findings of other studies, for instance, the lower diversity of pelagic MGI Crenarchaeota compared with Euryarchaeota described for the southern North Sea (Herfort et al., 2007). Interestingly, most of the euryarchaeotal sequences formed a nearly Baltic Sea-specific cluster that was at least 9% distant from the most similar environmental 16S rRNA sequence clone iBSZ2p.47 (Figure 4a). On the cellular level, Euryarchaeota were not detectable using the Eury806 gene probe. Euryarchaeotal 16S rRNA SSCP sequences were divers and showed 0–4 mismatches to gene probe Eury806. The most abundant euryarchaeotal SSCP band 9 (Figure 3a) holds 1 mismatch at the edge of the probe and should be detected, but could be below the detection limit or the cell walls not properly permeabilized by the fixation procedure. Based on the data set of this study the exact cellular euryarchaeotal abundance remains unclear; however, most recent metagenomic analyses of a comparable Baltic Sea redoxcline showed that euryarchaeotal gene abundance was below 2% of the total genes (unpublished data), supporting low euryarchaeotal abundances. This is in contrast to the redoxclines from the Black Sea and the Cariaco Trench, where both Eury- and Crenarchaeota were present in considerable numbers (Lin et al., 2006).

Only crenarchaeotal sequences were generated in the 16S rRNA/rRNA gene clone libraries but they were also detected in the SSCP fingerprints. The lack of euryarchaeotal clones is probably due to the lower coverage of the primer combinations used. For instance, according to the SILVA database of August 2009 Arch21F matches only 1176 out of 10936 sequences, in comparison with 5443 matches of the w036 SSCP forward primer. All gained crenarchaeotal sequences clustered together and, with the exception of the partial sequence SSCP_band_3 (98% sequence similarity), were >99% similar to each other. Assuming that 16S rRNA sequence similarities above 98.7–99% indicate an affiliation with a single species (Stackebrandt and Ebers, 2006), then nearly all of the Baltic crenarchaeotal 16S rRNA sequences could have arisen from one species. Archaeal amoA gene and transcript diversities supported the 16S rRNA analyses: the amoA sequences were also related to C. Nitrosopumilus (Figure 4b) and, with the exception of DGGE_band_7, all of the generated DGGE sequences were identical on the amino acid level. However, with a 16S rRNA distance of 2% and a phylogenetic separation supported by NJ, MP and ML algorithms (Figure 4a), the relationship to C. N. maritimus remains unexplained. Interestingly, the Baltic 16S rRNA sequences cluster together with sequences reported by Vetriani et al. (2003) (Figure 4a). Those sequences were obtained from comparable physicochemical conditions in the Black Sea, supporting recent observations that marine pelagic redoxclines share similar microbial key organisms (Labrenz et al., 2007).

Subcluster GD2 was quantified by CARD-FISH analyses. The highest cell numbers (2.3 × 105 cells ml−1) detected by the newly developed probe Cren679 were in the suboxic zone, accounting up to 26% of the TCN and matching the Cren537-positive cell abundance covering most of the known Crenarchaeota (De Corte et al., 2008, Figure 6a). These results were reproduced in a second profile, in which Cren679 contributed up to 26% of the TCN. Considering the C. N. maritimus mismatch (Table 2), Cren679 matches approximately 45% of all MGI sequences included in the SILVA database of August 2009. Thus, it potentially covers different members of the MGI Crenarchaeota. Nonetheless, the results of our ribosomal, and functional gene and transcript-based diversity analyses were obtained using several primer, fingerprinting and cloning systems, all of which identified one 16S rRNA and one amoA cluster, both closely related to C. N. maritimus (Figures 4a and b). Also, on the morphological level, the specific detection of peanut-shaped cells (Figure 5) was an indication that C. Nitrosopumilus relatives are targeted by probe Cren679. Hence, there is strong evidence that only subcluster GD2 was detected by both gene probes, Cren679 and Cren537 (Figure 6a). Taken together, these data point to the high likelihood that: (1) the crenarchaeotal assemblages consist only of one phylotype, named GD2, which dominates the archaeal community throughout the entire redoxcline, (2) the detected amoA gene cluster belongs to this group and, consequently, (3) GD2 is a nitrifying Crenarchaeota.

Nitrification zone

Higher nitrite concentrations combined with lower oxygen or ammonium concentrations are indicative of a nitrification zone in pelagic redoxclines. In our study, these conditions were fulfilled for a narrow zone ranging from a depth of 120 m to 122 m just above the chemocline (Figures 2a and b). This area was designated as the nitrification zone because it contained the maximal number of archaeal amoA transcripts (Figure 6b). Analogously, archaeal amoA expression in Black Sea redoxclines is confined to a narrow nitrification zone with low oxygen and ammonium concentrations and high crenarchaeotal amoA expression (Lam et al., 2007). Interestingly, the highest ratio of amoA transcript numbers per cell was in the nitrification zone (Figure 6b), but the value of 0.04 is, overall, lower than what might be expected. The rates of transcriptional regulation or amoA mRNA synthesis with respect to degradation are unknown for Cren679 cells, and absolute transcript numbers do not necessarily reflect their physiological state (Lam et al., 2007). However, we developed a new sampling device for deeper redoxcline waters, which is based on in situ fixation of water samples, and most recent analyses showed that amoA transcript numbers are approximately 20 times higher when this system is used (unpublished data). Transferring these data to our study, absolute amoA transcript abundances per cell should by around 1 (assuming 100% active cells) to 4 (20% active cells) in the nitrification zone. This is in the range what could be expected for a population actively oxidizing ammonia.

It is not surprising that archaeal amoA abundances were substantially lower in the sulfidic zone, which is inadequate for nitrification, than in the chemocline. It remains unclear, however, why the abundance of the GD2 subcluster is still high below the putative nitrification zone. This may be due to distribution processes (for example, sinking, mixing events) or to other metabolic pathways that enable a broad vertical distribution of the GD2 subgroup.

Selection factors

Despite the fact that different bacterial amoA primer systems were used (Table 1), no β- or γ-proteobacterial amoA genes or transcripts could be detected in any of the samples investigated. This finding was also supported by bacterial 16S rRNA clone libraries generated from the suboxic zone, which were devoid of β- or γ-proteobacterial nitrifiers (data not shown), and is in contrast to earlier studies of the Baltic Sea (Bauer, 2003) or the Black Sea (Lam et al., 2007), where proteobacterial ammonia oxidizers were also found in the lower oxic and suboxic zones of the pelagic redoxcline.

In our investigated redoxcline, the environmental factors favoring the dominance of the GD2 cluster are as yet unknown. Martens-Habbena et al. (2009) demonstrated that the high affinity of C. N. maritimus for ammonium enables it to out-compete bacterial nitrifiers at low ammonium concentrations. Therefore, the authors proposed that ammonia oxidation kinetics can determine the niche separation of nitrifying Archaea and Bacteria. In our study, the highest Cren679 cell numbers (Figure 6a) and archaeal amoA abundances (Figure 6b) occurred at ammonium concentrations of less than 1 μmol l−1, which would support the assumption of Martens-Habbena et al. (2009). However, similar concentrations have been determined for the Black Sea where β- and γ-proteobacterial nitrifiers were detected (Lam et al., 2007).

Erguder et al. (2009) reviewed the environmental factors shaping the ecological niches of AOA. These authors proposed that sulfide-containing waters favor AOA, because the inhibitory effect of sulfide or other S-containing compounds is more severe for bacterial than for archaeal nitrification. In our study, the overlap of sulfide- and oxygen-containing waters (Figure 2a), indicative of local mixing events or lateral intrusions, also supports a presumed inhibition of bacterial nitrifiers. It is known that these events are seasonally occurring in the Gotland Basin of the central Baltic Sea, but are more pronounced in fall and winter than in spring and summer (Axell, 1998). Thus, at the end of February, nitrifying GD2 could represent a sulfide-tolerating crenarchaeotal subcluster that is favored because of a longer period of sulfide pulses into the suboxic nitrification zones.