Seasonal Changes of Freshwater Ammonia-Oxidizing Archaeal Assemblages and Nitrogen Species in Oligotrophic Alpine Lakes

Abstract

The annual changes in the composition and abundance of ammonia-oxidizing archaea (AOA) were analyzed monthly in surface waters of three high mountain lakes within the Limnological Observatory of the Pyrenees (LOOP; northeast Spain) using both 16S rRNA and functional (ammonia monooxygenase gene, amoA) gene sequencing as well as quantitative PCR amplification. The set of biological data was related to changes in nitrogen species and to other relevant environmental variables. The whole archaeal assemblage was dominated by phylotypes closely related to the crenarchaeal 1.1a group (58% ± 18% of total 16S rRNA gene sequences), and consistent structural changes were detected during the study. Water temperature was the environmental variable that better explained spring, summer, and winter (ice-covered lakes) archaeal assemblage structure. The amoA gene was detected year round, and seasonal changes in amoA gene composition were well correlated with changes in the archaeal 16S rRNA gene pool. In addition, copy numbers of both the specific 1.1a group 16 rRNA and archaeal amoA genes were well correlated, suggesting that most freshwater 1.1a Crenarchaeota had the potential to carry out ammonia oxidation. Seasonal changes in the diversity and abundance of AOA (i.e., amoA) were better explained by temporal changes in ammonium, the substrate for nitrification, and mostly nitrite, the product of ammonia oxidation. Lacustrine amoA gene sequences grouped in coherent freshwater phylogenetic clusters, suggesting that freshwater habitats harbor typical amoA-containing ecotypes, which is different from soils and seas. We observed within the freshwater amoA gene sequence pool a high genetic divergence (translating to up to 32% amino acid divergence) between the spring and the remaining AOA assemblages. This suggests that different AOA ecotypes are adapted to different temporal ecological niches in these lakes.

For almost 30 years, the ubiquity of nitrification, particularly in the most oligotrophic environments, had constituted an enigma for ecologists (28, 52). First, because of very low natural abundances of ammonia-oxidizing bacteria (AOB), the a priori unique microorganisms with the potential to catalyze the first step of nitrification had been reported (e.g., 0.1% of total bacterial assemblage in oceanic habitats) (9, 38), and second, because field measurements indicated that the oxidation of ammonia (NH₄⁺) to nitrate was feasible in situ even in the most oligotrophic habitats, where NH₄⁺ concentrations were below the affinity threshold for AOB (8). Five years ago, metagenomics provided an answer to the enigma and radically changed the general perception of the nitrification process (65, 67). The presence of putative ammonia monooxygenase subunits (amoA, amoB, and amoC) within the genomes of widespread and abundant microorganisms of the domain Archaea (65, 67), as well as the chemolithoautotrophic growth using ammonia as an energy source observed in the recently isolated Crenarchaeota Nitrosopumilus maritimus (37), has fuelled the research on the role of archaea in the global nitrogen cycle. Recent studies have shown the dominance of ammonia-oxidizing archaea (AOA) over AOB in most natural habitats (29, 39, 40, 51, 68). This suggests a pivotal role of archaea in nitrification, particularly in oligotrophic environments (47).

PCR-based amoA gene surveys have revealed the ubiquity of AOA in natural habitats (see recent reviews in references 21 and 56) and have quickly expanded the shape of the archaeal amoA phylogenetic tree (56). However, most studies focused on marine samples, sediments, or soils, whereas freshwater habitats have remained poorly explored. Interestingly, freshwater habitats recently have been identified as one of the largest reservoirs of archaeal genetic diversity (3, 43). So far, the few studies that have investigated the environmental factors shaping the diversity of amoA genes in oceans, sediments, and soils have indentified pH (51), salinity (49), fertilization regimen (29, 61), and nitrite levels (58) as key environmental parameters for AOA.

In the present work, we have studied seasonal changes in environmental parameters and changes in the genetic composition and abundance of freshwater AOA in a few selected high-altitude lakes. These remote lakes are very poor in nutrients and are characterized by a large variability in both the concentration and the distribution of the different N oxidation states that are not directly related to general limnological parameters (physical and chemical). In these lakes, atmospheric deposition represents the main source of nitrogen (15). These features and previous observations of the changes in dissolved organic nitrogen associated with high microbial activities (15) suggest strong interactions between microbial communities and the nitrogen cycle. Recently, Crenarchaeota closely related to the ammonia-oxidizing 1.1a group have been detected in the surface of these lakes (4). Thus, these lakes are very convenient model systems to explore the dynamics of archaeal diversity and abundance related to N biogeochemistry.

MATERIALS AND METHODS

Study site and sampling.

For this study, we selected three connected oligotrophic high mountain lakes (HML), i.e., Lake Llebreta, Lake Llong, and Lake Redó AT, located in the same catchment area between 1,620 m (Lake Llebreta) and 2,117 m (Lake Redó AT) above sea level. The lakes belong to the Limnological Observatory of the Pyrenees (LOOP; Spanish Pyrenees; 42°33′N, 00°53′E) within the protected area of the Aigüestortes National Park. These lakes are small (<10 ha), are shallow (<13 m), are mixed by wind, and are ice covered for 4 to 7 months every year. More details can be found in Auguet and Casamayor (4) and Hervàs et al. (34). Samples were collected monthly from the air-water surface microlayer (SML; during the ice-free period) and the underlying water (UW; first 1 m integrated) at the deepest point of the lakes between May 2007 and April 2008. Collected samples were representative of surface waters according to a previous work showing no significant horizontal variability in the archaeal community composition and abundance for the same set of lakes (4). Lakes were frozen from November to April and were inaccessible during January and February (data for these 2 months are missing in the graphs). SML samples (first 400 μm) were collected with a nylon screen sampler (1, 4). As atmospheric nitrogen deposition (i.e., rain or snow) is the main source of nitrogen in remote mountain catchments (15), a bulk precipitation collector placed in the vicinity of Lake Llebreta was used to fortnightly collect atmospheric precipitation samples and determine the volume of rain and atmospheric ammonium deposition (see Camarero and Catalan [10] for details on the sampling devices).

Chemical analysis.

Samples for dissolved organic carbon (DOC) were filtered right at the sampling site through an ignited Whatman GF/F filter and stored in the dark at 4°C until analyzed (between 3 and 5 h after sampling) by high-temperature combustion/direct injection into a gas analyzer (model TOC-5000; Shimadzu). Sample stability and blank tests were run to check that the storage did not affect DOC measurements. NH₄⁺ content was colorimetrically determined according to Solórzano (62). NO₃⁻ content was measured by capillary electrophoresis (model Quanta-4000; Waters). NO₂⁻ content was determined spectrophotometrically at 560 nm after reaction in an acid medium of H₃PO₄ with a solution of sulfanilamide and N-(1-naphthyl) ethylene diamine dihydrochloride (NEDA). Previous tests indicated no significant differences in nitrogen species and DOC concentrations between the SML and UW. The analysis of atmospheric ammonium deposited followed the method of Camarero and Catalan (10).

Genetic analyses.

For DNA analyses, water samples were prefiltered in situ through a 40-μm-pore-size net to avoid large zooplankton and algae, and 300 to 500 ml subsequently was filtered on 0.2-μm-pore polycarbonate membranes (47-mm diameter; Nuclepore), and the filter was stored at −20°C until further processing. Filters were incubated with lysozyme, proteinase K, and sodium dodecyl sulfate (SDS) in lysis buffer (40 mM EDTA, 50 mM Tris, pH 8.3, 0.75 M sucrose), and DNA was extracted with phenol-chloroform-isoamyl alcohol at 25:24:1 (vol/vol/vol) and with chloroform-isoamyl alcohol at 24:1 (vol/vol) (20).

We initially used denaturing gradient gel fingerprinting (DGGE) for a rapid screening of the temporal and spatial shifts in both archaeal community composition (i.e., 16S rRNA gene) and amoA gene diversity. We then applied PCR cloning in six selected samples from Lake Llebreta for analyzing changes in detail (Table 1 lists PCR conditions). For 16S rRNA gene DGGE, a nested PCR was run with the archaeal primers set 21f-958r (18), followed by ARC344f-ARC915r (13). For the amoA gene (635-bp fragment), we used the primer set Arch-amoAF-Arch-amoAR (25). At the 5′ end of each forward primer, an additional 40-bp GC-rich nucleotide sequence (GC clamp) was added to stabilize the migration of the DNA fragments in the DGGE gel (50). DGGE was carried out in the D-Code equipment (Bio-Rad) as previously described (13). PCR products were run for 4 h at a constant voltage of 200 V in a 6% polyacrylamide gel (with a denaturing gradient of 30 to 65% [16S rRNA gene] or 35 to 70% [amoA gene] being 100% denaturing conditions, 7 M urea, and 40% deionized formamide). Prior to loading, DNA was quantified on agarose gels using a quantitative DNA ladder (low-DNA-mass ladder; Eurogentec). PCR-amplified products (500 and 1,000 ng for 16S rRNA and amoA genes, respectively) were loaded and run in TAE buffer (40 mM Tris, pH 8, 20 mM sodium acetate, and 1 mM EDTA) at 60°C. A DGGE ladder composed of a mixture of known SSU rRNA gene fragments was loaded as an internal standard for gel-to-gel comparison. Gels were stained with SYBR gold (2,000× final concentration; Molecular Probes) for 45 min in the dark. Digitized images were obtained with the UVIDoc system (UVItec, Cambridge, United Kingdom) and saved as computer files. Prominent bands were aseptically excised and reamplified (13) and sent for sequencing at external facilities (Macrogen).

TABLE 1.

PCR and qPCR conditions used in this work

Assay	Target	Primer pairb	No. of cycles	PCR conditionsa						Reference
				Denaturation		Annealing		Elongation
				Temp (°C)	Time (s)	Temp (°C)	Time (s)	Temp (°C)	Time (s)
PCR	Archaeal 16S rRNA gene	21f/958r	30	94	60	56	60	72	120	18
	Archaeal 16S rRNA gene (nested)	ARC344f/ARC915r	20/15	94	60	71c/61	60	72	90	12
	Archaeal amoA gene	Arch amoAf/amoAr	35	94	45	55	60	72	60	25
qPCR	Group 1.1a group	MCGI-391f/MCGI-554r	60	94	5	61	20	72	15	64
	Archaeal amoA gene	AOA-amoA-f/AOA-amoA-r	60	94	5	59	20	72	15	68

^aA GC-rich clamp was attached to the 5′ end of each forward primer used in all amplification reactions to generate amplicons for DGGE analysis.

^bFor conventional PCR, the temperature was held at 95°C for 4 min before each run of cycles and kept at 72°C for 15 min after all cycles were completed to allow final template elongation.

^cThe program consisted of a touchdown protocol where the initial annealing temperature decreased by 0.5°C each cycle during the first 20 cycles.

DGGE banding patterns were compared using a similarity matrix based on the presence/absence of each band within each sample for the whole data set. The similarity matrix (Dice coefficient) was used for a dendrogram analysis following the unweighted-pair group method with arithmetic means (UPGMA).

Because Archaea previously were reported to be abundant and genetically diverse in Lake Llebreta (4), these samples were characterized in detail by PCR cloning with the primer set 21f-958r (937-bp fragment) according to the shifts in archaeal populations previously detected in the fingerprints. PCR products were purified with the QIAquick PCR purification kit (Qiagen) and cloned with the TOPO TA cloning kit (Invitrogen) by following the manufacturer's instructions. The presence of inserts was checked by ampicillin resistance and blue/white selection in LB plates supplemented with 100 μg/ml ampicillin and 40 μg/ml 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) as previously reported (23). Sequencing reactions were carried out using external facilities (Macrogen).

Phylogenetic analysis.

The 16S rRNA gene sequences were automatically aligned with the NAST aligner (19). Ribosomal sequences were clustered at an identity threshold of 97% (60) and imported into the Greengenes database (http://greengenes.lbl.gov/) based on the ARB package (46) (http://www.arb-home.de). The Archaea base frequency filter in ARB was applied to exclude highly variable positions before sequences were added, using the ARB parsimony insertion tool, to the original Greengenes tree (7,682 positions) provided by default.

amoA gene sequences were checked manually with BioEdit (27) and submitted for searching in the protein database using translated nucleotide sequences (BLASTX; www.ncbi.nlm.nih.gov [2]) to get a first indication of the proteins retrieved, as well as checking for protein identity. Sequences were clustered at 99% identity in nucleotides with CD-hit (42). Multiple-sequence alignment was performed using the software MAFFT (35). Poorly aligned positions and divergent regions of the DNA alignment were removed using the Gbloks software (14), ending with 579-bp-length fragments for the final analysis. Phylogenetic inference was carried out with RAXML (63), which estimates large phylogenies by maximum likelihood. The best phylogenetic tree estimated by PHYML was drawn with iTOL (41).

amoA identity matrices were calculated with BioEdit (27) and MatGAT (11) for both nucleotide and protein datasets.

qPCR.

Gene copy numbers for the archaeal group 1.1a 16S rRNA gene and for the archaeal amoA gene were determined by quantitative real-time PCR (qPCR) amplification on water samples collected in Lake Llebreta, Lake Llong, and Lake Redo AT. The qPCR assays were run on 96-well white qPCR plates with adhesive seals (Bio-Rad) in a DNA engine thermal cycler (Bio-Rad, Hercules, CA) equipped with a Chromo 4 real-time detector (Bio-Rad) using the primers sets and conditions listed in Table 1. The reaction mixture (20 μl) contained 10 μl of SsoFast EvaGreen supermix (Bio-Rad), 5 μl of template DNA (2 ng), 10 μM primers, and molecular biology-grade water (Sigma) (Table 1). The qPCRs were run for 2 min at 98°C, followed by 60 cycles as detailed in Table 1. The fluorescence signal was read in each cycle after the elongation step at 78°C for 25 s for accurate stringent product quantification. All reactions were run in triplicate with standard curves spanning from 10¹ to 10⁷ and from 10² to 10⁸ copies of DNA for 16S rRNA and amoA genes, respectively. Standard curves were obtained after serial dilutions of previously titrated suspensions of each gene amplified by conventional PCR from environmental clones (FN691587 for 16S rRNA and FN773417 for crenarchaeal amoA) and further purification (QIAquick; Qiagen) and quantification. Overall, average efficiencies for all quantification reactions ranged from 76% for archaeal amoA to 92% for group 1.1a archaea, with R values of >0.99. The specificity of reactions was confirmed by melting-curve analyses and by agarose gel electrophoresis to identify unspecific PCR products, such as primer dimers or gene fragments of unexpected length (data not shown).

Diversity indices and statistical analyses.

Phylogenetic diversity (PD) was calculated as the sum of the branch length associated with the 16S rRNA gene sequences for each habitat (22). To correct for unequal numbers of sequences, we calculated the mean PD of 1,000 randomized subsamples of each habitat (5). Phylogenetic structure was estimated with the phylogenetic species variability (PSV) index (30). PSV quantifies how phylogenetic relatedness decreases the variance of a hypothetical neutral trait. The value is 1 when all species are phylogenetically unrelated (i.e., a star phylogeny) and approaches 0 as species become more related.

Shannon and Chao diversity indices were calculated using the FastGroupII server (http://biome.sdsu.edu/fastgroup/). The coverage of libraries and the ACE diversity index was calculated using the diversity tool (http://www.aslo.org/lomethods/free/2004/0114a.html) developed by Kemp and Aller (36).

Distance matrices for 16S rRNA and amoA genes were constructed using the UniFrac metric (http://bmf.colorado.edu/unifrac). UniFrac is a betadiversity metric that quantifies community similarity based on the phylogenetic relatedness (45). To assess the sources of variation in the UniFrac matrices (16S rRNA and amoA), we used permutational multivariate analysis of variance (PERMANOVA) based on 1,000 permutations (48) with function adonis in the vegan package (http://vegan.r-forge.r-project.org).

Spearman rank (r_s) correlations were run to investigate the relationship between environmental parameters and gene copy numbers. All statistical analyses were run in the R environment (http://www.r-project.org/) with the pvclust package for cluster analysis and the ade4 package for Mantel tests.

Nucleotide sequence accession numbers.

Sequences were deposited in GenBank under the accession numbers FN691483 to FN691759 for the archaeal 16S rRNA gene and FN773346 to FN773486 for the archaeal amoA gene. Archaeal 16S rRNA gene sequences were deposited following the MIENS (minimal information about an environmental study standard) procedure, allowing the association of metadata with each sequence. Metadata can be found under the accession numbers ERS008538 to ERS008543.

RESULTS

Seasonal environmental changes.

Water temperatures followed a typical seasonal pattern for high mountain lakes (HML), ranging from 0.8°C (December) to 12.5°C (July), and all three lakes were within the same temperature range with a ±2°C difference (Fig. 1A). Annual atmospheric precipitations (rain or snow) ranged between 16 and 111 mm, with higher values during the ice-covered period (i.e., November to April) and lower in summer (Fig. 1A). NH₄⁺ concentrations in atmospheric precipitation waters were, on average, of 18 ± 13 μmol liter⁻¹ (data not shown). During the sampling period, NH₄⁺, NO₂⁻, and NO₃⁻ concentrations ranged in the three lakes from 0.02 to 1.68 μmol liter⁻¹, 0.03 to 0.13 μmol liter⁻¹, and 4.2 to 16.9 μmol liter⁻¹, respectively (Fig. 1B, C, and D). DOC concentrations were between 0.5 and 2.4 mg liter⁻¹ (Fig. 1E). The dynamics of NO₂⁻, NH₄⁺, and DOC were quite similar and opposite that of NO₃⁻ in the three lakes, as shown in Fig. 1 (see also the Spearman correlation analysis in Fig. S1 in the supplemental material). For instance, Lake Llong showed a consistent gradual decrease of NO₂⁻ and NH₄⁺ concentrations from August to March, while the NO₃²⁻ concentration increased accordingly (Fig. 1B, C, and D). In spring, large concentrations of inorganic nitrogen and DOC, which previously had accumulated in the snow pack, enter the water column along the melting period. This input of nutrients is processed during summer, resulting in a decrease of inorganic nitrogen species and DOC concentrations to the lowest values, generally at the end of winter.

FIG. 1.

(A) Changes in temperature (lines) and precipitation (shadow bars) throughout the year in the sampled lakes. (B to D) Evolution of NH₄⁺, NO₂⁻, and NO₃²⁻ concentrations during the annual period sampled. (E) Annual variations of DOC concentrations. Standard deviations were <1% for nitrite and DOC and <10% for ammonium and nitrate.

Seasonal changes in archaeal 16S rRNA gene diversity.

The DGGE analysis of PCR-amplified archaeal 16S rRNA fragments (see an example in Fig. S2 in the supplemental material) produced a reproducible genetic fingerprint for each sample in different runs. Overall, the number of DGGE bands per sample ranged from 6 to 14. The similarity analysis of the banding pattern clustered samples mostly according to temporal (month and sampling season) rather than spatial components (Fig. 2). A PERMANOVA analysis confirmed this result (R² = 0.57, P < 0.001). The lake of origin was also a significant source of variation for archaeal diversity (R² = 0.05, P < 0.01). From the DGGE banding pattern alone, no significant differences were observed between surface microlayer (SML) and underlying water (UW) archaeal communities.

FIG. 2.

Dendrogram based on the similarity matrix generated from archaeal 16S rRNA gene DGGE fingerprints and UPGMA. Archaeal 16S rRNA gene diversity was studied further by cloning using the samples in boldface. The sample code corresponds to the format lake.month.layer. LL, Llebreta; Llo, Llong; RAT, Redó Aigüestortes.

Seasonal shifts in the archaeal assemblage composition of Lake Llebreta were analyzed further by cloning 16S rRNA genes from three selected samples with different fingerprints (June, September, and November, respectively). Sequences were grouped at 97% identity, and overall we obtained a good coverage (>71%) of the archaeal richness present in the samples. Interestingly, we observed consistent differences between SML and UW. The UW showed very similar values of diversity and estimated richness indexes (H′ and Chao, respectively; Table 2) for all three samples. SML, in turn, showed a higher variability, with the lowest values of PD (phylogenetic diversity) and PSV (phylogenetic species variability) in September, suggesting more-dynamic archaeal populations being present in the SML than in the UW.

TABLE 2.

Diversity indicators of archaeal assemblages in Lake Llebreta (based on the 16S rRNA gene)^a

Clone library	Layer	Clone nos.	OTU (97%)	Coverage (%)	Shannon index	Chao1 index	ACE	PD	PSV
June	SML	48	22	71	2.61	37	62	2.00 ± 0.15	0.77
	UW	47	19	77	2.33	30	31	1.96 ± 0.11	0.83
								1.15 ± 0.07	0.68
September	SML	48	14	77	1.9	20	29	1.47 ± 0.09	0.6
	UW	44	18	75	2.3	27	46	2.12 ± 0.12	0.75
								0.86 ± 0.16	0.36
November	SML	42	12	83	1.5	17	22	2.13 ± 0.00	0.74
	UW	47	17	74	2.33	31	/	1.98 ± 0.10	0.88
								1.13 ± 0.00	0.70

^aThe calculation of OTU number and diversity indices are at a 3% distance level. Values of phylogenetic diversity (PD) and phylogenetic species variability (PSV) in boldface were calculated only for the Crenarchaeota phylum combining SML and UW libraries each month.

Phylogenetic 16S rRNA gene-based analysis showed that archaeal lineages known to contain AOA members (i.e., 1.1a and 1.1b) (Fig. 3) dominated most of the samples, accounting for 60% ± 13% of the total archaeal 16S rRNA gene sequences. When the data subset containing the known AOA lineages was used for the community phylogenetic analyses (Table 2), the samples from June showed the highest PD value while November had the highest PSV value. In September, the crenarchaeal 1.1a group reached up to 92% of the SML sequences and 55% of the UW sequences. All of these sequences grouped in monophyletic clusters within the 1.1a group (see Fig. S3 in the supplemental material). Members of other crenarchaeal groups were present sporadically in the six libraries but never accounted for a high contribution. In particular, sequences from group 1.1b were detected only in the UW sample in June, where they represented up to 12.5% of the sequences. For Euryarchaeota, most of the clones belonged to Methanomicrobiales, HV-Fresh (also named LDS, for Lake Dagow sediment), and PlSA1 (also named RCV, for rice cluster V) (see reference 3 for proposed cluster renaming).

FIG. 3.

(A) Archaeal community structure in Lake Llebreta, obtained from the 16S rRNA gene libraries of the SML and UW in June, September, and November. Communities are represented by relative abundances of clones in different phylogenetic groups. Crenarchaeal groups where the presence of the amoA gene has been documented (i.e., 1.1a and 1.1b) are represented in white, other crenarchaeal groups are in black, and euryarchaeal groups are in grey. (B) Collapsed tree representing the phylogenetic position of the archaeal groups detected each month. The tree is based on maximum-parsimony analysis of the data set with the ARB program package (http://www.arb-home.de). The nomenclature follows the labeled clusters or divisions provided by default in the Greengenes general tree and immediately subordinate to the Crenarchaeota or Euryarchaeota phylum (also see reference 3).

Seasonal changes in archaeal amoA gene diversity.

Archaeal amoA gene fragments were detected in all of the samples. The fingerprinting richness (i.e., number of DGGE bands) ranged from 2 to 14. One hundred forty bands were excised from the DGGE gels and sequenced. Based on a 1% cutoff, we recovered a total of 103 unique operational taxonomic units (OTUs), with 65 to 89% nucleotide identity to the amoA sequence of Nitrosopumilus maritimus. We often observed that single lake samples contained more than one amoA OTU type. The whole set of sequences fell into three distinct but environmentally coherent freshwater clusters (A, B, and C) (Fig. 4). Indeed, the closest relatives in GenBank for most of the recovered amoA sequences were obtained from rhizosphere, groundwater, and drinking waters (33, 57, 66) (Table 3).

FIG. 4.

Phylogenetic relationships between archaeal amoA gene sequences from Pyrenean lakes and additional environments. The nucleotide maximum-likelihood tree was inferred by RaxML and based on 579 nucleotide positions. The scale bar represents 0.1 change per site. Bootstrap values (>50%; 1,000 replicates) are indicated at the root nodes. Habitats are color coded in the collapsed tree. More-detailed phylogenetic trees for the archaeal amoA gene clusters found in this study are shown. The nomenclature of the sequences retrieved in this study was clone-lake-layer-month. Seasons are color coded as follows: blue, winter; green, spring; and red, summer. The number of sequences within each cluster is provided. Accession numbers for all sequences are given in Table S1 in the supplemental material.

TABLE 3.

Description of clusters formed by the amoA gene sequences and values of PD and PSV

Season	Cluster	% Identity with N. maritimus	Closest BLAST	% Identity to closest BLAST	PD	PSV
Winter	A	75-79	EU667840 (rhizosphere)a	92-94	0.93 ± 0.13	0.74
	B	74-80	FJ543350 (groundwater)b	90-97
	C	87-88	EU667924 (rhizosphere)a	96-99
Summer	A	69-73	EU667867 (rhizosphere)a	84-95	1.89 ± 0.20	0.52
	B	77-79	EU852682 (drinking water)c	83-95
	C	69-87	EU852672 (drinking water)c	73-95
Spring	A	65-71	EF530112 (forest soil)d	84-91	2.10 ± 0.00	0.67
	C	74-77	HM191510 (wastewater)e	82-83

^aHerrmann et al. (33).

^bReed et al. (57).

^cvan der Wielen et al. (66).

^dS. A. Boyle, P. J. Bottomley, and D. D. Myrold (2007), unpublished.

^eT. Limpiyakorn, P. Sonthiphand, and C. Polprasert (2010), unpublished.

Consistently, we observed that the main source of variation in amoA gene diversity was the sampling period (R² = 0.41, P < 0.001; PERMANOVA test). Indeed, independently of the lake or the water layer, sequences grouped into phylogenetically distinct and often well-supported (bootstrap, >70%) subclusters characteristic of each season (Fig. 4). PD for amoA was higher in spring and summer (Table 3), as illustrated by the length of the branches in the phylogenetic tree (Fig. 4). In spring we also found the most divergent amoA sequences, with only 65 to 71% identities with the amoA sequence of N. maritimus at the nucleotide level (Table 3). The translation of amoA nucleotide sequences to amino acid sequences also revealed that the transition from the ice-covered period to spring and from spring to summer resulted in a change in the translated proteins of up to 32% sequence divergence. In contrast, sequences from the summer and ice-covered period shared 80% ± 9% identity at the amino acid level.

Overall, when we compared the whole data set of the annual survey, changes in seasonal composition for both 16S rRNA and amoA genes were significantly correlated (r = 0.45, P < 0.001; Mantel test). PERMANOVA analyses showed that changes in water temperature generally explained most changes in the diversity of both genes (P < 0.01). For the amoA gene, however, ammonia and nitrite explained better than temperature (P < 0.01) the changes in AOA composition. DOC concentrations and the volume of atmospheric precipitation (P < 0.05) were minor components of the observed AOA community changes.

Seasonal changes in archaeal group 1.1a 16S rRNA and amoA gene copy abundances.

Abundances of group 1.1a 16S rRNA and amoA genes are depicted in Fig. 5 as determined by qPCR in UW samples from the three lakes. The abundances of both genes were significantly correlated (r_s = 0.89, P < 0.01, n = 28; all data are from the three lakes) with a slope of 2.3. Thus, each archaeal 1.1a cell may contain 2.3 copies of the amoA gene. Copy numbers ranged from 0.25 × 10³ to 12.68 × 10³ copies ml⁻¹ for the group 1.1a 16S rRNA gene and from 0.40 × 10³ to 32.15 × 10³ copies ml⁻¹ for amoA genes (Fig. 5). This range of variability may be more related to temporal changes than spatial differences within the water column. Indeed, gene abundances presented a marked seasonal increase from spring to summer, with 2.1-fold for group 1.1a 16S rRNA and 1.7-fold for amoA genes, respectively (Fig. 5). Conversely, from summer to winter, copy numbers decreased back to spring abundances. We observed that gene abundances of group 1.1a and archaeal amoA were significantly correlated with water temperature (r_s = 0.61 and 0.68, P < 0.01, n = 28), nitrites (r_s = 0.60 and 0.64, P < 0.01, n = 28), and ammonium (r_s = 0.62 and 0.54, P < 0.05, n = 28). However, only nitrites significantly explained the variability in the archaeal amoA gene copy numbers (P < 0.05) after multiple linear regression analysis.

FIG. 5.

Abundance of group 1.1a 16S rRNA and amoA copies throughout the year in the three study lakes. Data are missing when not enough DNA was collected to run qPCR analysis.

DISCUSSION

Seasonal changes in ribosomal (16S rRNA) and functional (ammonia monooxygenase, amoA) genes reflected dynamic changes in the diversity and abundance of AOA assemblages inhabiting oligotrophic high mountain lakes. These changes were significantly related to the dynamics of ammonium and nitrite concentrations and included a succession of AOA freshwater ecotypes containing marked sequence divergences in the amoA gene.

The consistent lines of evidence that closely link ammonia-oxidizing archaea (AOA) with the nitrogen cycle recently have encouraged a large number of detailed studies that have significantly expanded the current knowledge on the diversity and worldwide distribution of AOA (see reviews in 21 and 56). In contrast, in situ studies exploring seasonal changes in archaeal amoA gene composition and the environmental factors controlling them are scarce, mainly because most of the available studies are a molecular snapshot in a specific habitat. Remarkably, we observed seasonal changes in the amoA gene composition for a single ecosystem that were as large as the differences existing among habitats in previous works. The transition between archaeal communities from spring, summer, and ice-covered lakes agrees with previous studies showing seasonal changes in the composition of planktonic archaeal assemblages, such as in karstic lakes (12, 43) and North Sea waters (32, 53, 68). We observed that seasonal changes in the concentrations of NH₄⁺, the substrate for ammonia oxidation, and mostly in NO₂⁻, the product of ammonia oxidation, explained a significant part of the observed changes in both AOA diversity and abundance (amoA genes). A similar relationship between nitrite and AOA diversity was found in the Westerschelde Estuary (58). In addition, the correlations found in our study between group 1.1a 16S rRNA or archaeal amoA gene copy numbers and nitrite or ammonium also had been previously reported in other ecosystems (7, 17, 32, 49, 59, 68). Taken together, these results suggest a close link between ammonia oxidation and both the distribution and composition of crenarchaeota in oligotrophic high mountain lakes. Nonetheless, caution must be taken, because our study only shows the potential for ammonia oxidation. A detailed survey of archaeal amoA transcript abundance and simultaneous bulk nitrification measurements would provide more conclusive evidence that ammonia oxidation is carried out by crenarchaeota in the study lakes.

Pyrenean lakes and, by extension, high mountain lakes show remarkable variability in both nitrogen content and composition (15). Such variability is not directly linked to a single geomorphological, chemical, or trophic characteristic of the lakes (15) but rather results from a combination of processes. Among them, bacterial/archaeal activities may have an important role in modulating the nitrogen cycle of these ecosystems. Interestingly, changes in the amount of atmospheric precipitation also significantly explained part of the AOA community shifts. This is closely related to the fact that atmospheric nitrogen deposition (i.e., rain or snow) is the main source of nitrogen in remote mountain catchments (15). During the survey presented here, NH₄⁺ concentrations in the rain were ca. 30-fold higher than mean values measured in the water of the lakes. High mountain lakes also show a high sensitivity to atmospheric dust depositions (34). Regular atmospheric deposition episodes probably directly affect the freshwater archaeal populations, especially those inhabiting the very top surface layers (SML), and may explain the vertical community shifts (i.e., differences between layers) observed in the clone libraries and the larger variation of the diversity indices found in the SML compared to those of UW samples. Thus, because ammonia inputs are mostly atmospheric in these remote environments, the local archaeal dynamics may be regulated by regional and global processes.

Overall, ultraoligotrophic waters from mountain areas may constitute an ideal habitat for AOA, since in situ ammonium concentrations still are four times higher than the recently reported AOA K_m uptake for NH₄⁺ (130 nM [47]). These concentrations are at the lower end of AOB affinity (>1 μM NH₄⁺ [8]), and AOA may easily outcompete AOB and heterotrophs under NH₄⁺-limiting conditions (47). This is in agreement with the fact that AOA were a stable (recurrent detection of archaeal amoA genes in all the samples throughout the whole year) and abundant (8.7 × 10³ copies ml⁻¹; average for the three lakes throughout the year) component of archaeal assemblages in high mountain lakes. Again, and as illustrated in a recent study (17), caution is needed, because amoA gene copy numbers cannot be considered a proxy of either amoA gene expression or ammonia oxidation rates.

In agreement with the findings obtained from other ecosystems (6, 7, 17, 32, 59), most of the freshwater AOA in alpine lakes probably belongs to the 1.1a group, although we can only indirectly argue for that. The good correlation found between group 1.1a 16S rRNA and archaeal amoA gene copy numbers and the slope value of 2.3 are strong evidence that most freshwater 1.1a Crenarchaeota had the amoA gene. Other lines of evidence add further support to this assumption. First, the 1.1a lineage was by far the dominant group present in our 16S rRNA gene libraries. Second, seasonal changes in the diversity indices PD and PSV for both Crenarchaeota 16S rRNA and amoA genes behave similarly, with the highest PD value in spring and the highest PSV value in the ice-covered period. Finally, the limited archaeal diversity restricted to almost only members of group 1.1a and the parallel reduction of amoA diversity to one specific cluster in the sample September SML strongly support that the amoA gene detected in this sample is linked to group 1.1a.

Only a few freshwater planktonic environments have been prospected so far, but AOA have been detected in rivers (26, 31), in the oligotrophic African Lake Kivu (44), in groundwater (57), and in drinking water (66). Here, we have extended the survey to ultraoligotrophic mountain lakes, showing that freshwater amoA sequences mainly formed coherent freshwater clusters (Fig. 4), supporting the idea of sequence clustering according to habitat segregation as previously suggested (6, 24, 25). Although freshwater amoA gene sequences currently are underrepresented in the global amoA phylogenetic tree, this result also indicates that planktonic freshwater have a typical amoA signature, as in other habitats (e.g., soil, sediments, and marine water column). In fact, we have shown that several freshwater amoA ecotypes with up to 32% divergence in amino acid sequence can coexist in the same lake. Analyses of the freshwater amoA genes showed consistent seasonal rather than spatial differences in the diversity and abundance of AOA. This pattern disagrees with the variability observed between two meromictic lakes in the Artic (55). The permanent mixed nature of the shallow lakes studied here and the permanent water connection existing among them probably have blurred differences in altitude or catchment area.

The seasonal switch in amoA genetic composition also deserves further attention. Spring-to-summer changes in AOA communities had been observed previously in the Westerschelde Estuary (58), but the magnitude of the genetic divergence between the amoA sequences was much lower than the values we observed here (32% amino acid divergence). In addition, we observed a strong increase in amoA gene copy numbers correlated with an increase in ammonium and nitrite concentrations. Conversely, the winter-to-spring transition in amoA was associated with a change in genetic composition but not a significant change in amoA gene copy numbers. This is in agreement with the fact that the ice-break period is a major promoter of community transition (15, 54), and in shallow lakes the planktonic community can be entirely washed out (16). In a previous work, Francis and coworkers (25) found an unexpected compositional overlap between amoA sequences from very different environments characterized by a large variability in oxygen content. Curiously, we have found the opposite in alpine lakes, in the sense that we observed a lack of compositional overlap between amoA gene sequences within the same habitat. The translation of gene sequences to amino acid sequences unveiled nonsynonymous changes that may influence the ammonia oxygenase performance. Whether or not these nonsynonymous changes occurred on active sites of the enzyme and improved the adaptation to specific environmental conditions remain to be elucidated.