Heterotrophic Archaea Contribute to Carbon Cycling in Low-pH, Suboxic Biofilm Communities

Biofilm samples were collected underground within the Richmond Mine at Iron Mountain, CA. Temperature was measured in situ, and the pH was determined shortly after sample collection. Samples for metagenomic analysis were frozen on site on dry ice, whereas those for cultivation experiments were maintained on ice water during transport to the laboratory, where they were used immediately for inoculation experiments. At the time of sampling, biofilm thickness was estimated visually to indicate growth (developmental) stage. Based on prior direct measurements, early growth stage biofilms are typically <30 μm thick, whereas late growth stage biofilms can be up to 200 μm thick (61).

Environmental samples were fixed and stained with lineage-specific fluorescent in situ hybridization (FISH) probes to determine archaeal and bacterial relative abundances, as described previously (9). Fixed cells were counted manually on a Leica DMRX epifluorescence microscope at ×630 magnification. At least five fields of view and >1,000 cells were counted for each sample. FISH cell counts were used to calculate species percentages from total cell counts obtained using DAPI (4′,6′-diamidino-2-phenylindole). Probes used in the present study are listed in Table 1.

For all DNA extractions, ∼1 g of frozen biofilm was washed twice with 0.5 ml of cold phosphate-buffered saline (pH 1.2) before being resuspended in 500 μl of TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]) and 500 μl of STEP buffer (0.5% sodium dodecyl sulfate, 50 mM Tris-HCl [pH 7.5] 400 mM EDTA, 1 mg of proteinase K/ml, 0.5% Sarkosyl). The samples were cycled three times between liquid nitrogen and a 50°C water bath to facilitate lysis. After the last cycle, an additional 500 μl of STEP buffer was added, and the sample was left at 50°C for 20 min. The samples were extracted three times with 1 volume each of a 25:24:1 phenol-chloroform-isoamyl alcohol mix (Fisher BioReagents, Pittsburgh, PA) using a 15-ml Phase Lock Gel tube (5 Prime, Gaithersburg, MD). The aqueous layer was precipitated with 1 volume of cold isopropanol and 0.1 volume of cold sodium acetate (3 M; pH 5.2). Pellets were washed with 1 ml of cold ethanol and resuspended in TE buffer.

For clone library construction, 16S rRNA gene sequences were amplified by PCR in mixtures containing 1× PCR Master Mix (Promega, Madison, WI), ∼40 ng of template DNA, 300 mM concentrations (each) of forward and reverse primers, and 0.5 mg of bovine serum albumin (New England BioLabs, Ipswich, MA)/ml. The primers for the Bacteria-specific libraries were 27F (5′-AGAGTTTGATCMTGGCTCAG) and 1492R (5′-TACGGYTACCTTGTTACGACTT) (15) and for the Archaea-specific libraries were 522F (5′-GGYAAGACSGGTGSCAGC) and 1354R (5′-GCGRTTACTASGGAWTCC), designed to bind to both archaea of the Thermoplasmatales and the ARMAN. Reactions were incubated in a DNA Engine Dyad thermocycler (MJ Research, Quebec, Canada) for 94°C (2 min), followed by 30 cycles of 94°C for 45 s, 53°C for 45 s, and 72°C for 90 s, with a final extension of 72°C for 10 min. PCR products were purified and cloned, as described previously (10). Sequencing was carried out using the same amplification primers as used in PCRs on a 3730XL DNA Analyzer (Applied Biosystems, Carlsbad, CA). Sequences were assembled with Phred/Phrap software, and chimeric sequences were screened using Mallard (3). Sequences were clustered using UClust (23) at 97% percent identity. Sequences were aligned with MAFFT (37) and curated with MEGA (57). Trees were constructed using the maximum-likelihood algorithm in MEGA using the general time reversible model. Bootstrap values were calculated from 1,000 iterations.

For metagenomic sequencing, DNA was extracted from samples, as described above. Samples were treated with 10 U of RNase (New England BioLabs)/ml for 15 min at room temperature to improve sequencing quality. DNA from both samples was sequenced with 454 FLX Titanium (454 Life Sciences, Branford, CT), according to the manufacturer's protocols, with the exception that fragment libraries were quantified with digital PCR rather than with titrations (60). The data were quality controlled by removing reads with at least one ambiguous base. Artificial duplicate reads formed as a result of the 454 FLX Titanium sequencing method were removed, as described previously (16, 30). For analysis of the community composition, reads over 100 nucleotides in length were compared by BLAST to genomes of organisms from the Richmond Mine recovered by prior metagenomic studies (6, 59). Only matches with >98% identity were reported, and the number of hits for each species was normalized by genome size.

Whole-cell fractions of proteins were extracted from frozen biofilms (∼500 mg) via sonication and trichloroacetic acid precipitation, as previously described (18). Briefly, the cells were washed with H₂SO₄ (pH 1.1) to remove residual iron, and then the pellets were resuspended in 6 ml of 20 mM Tris-SO₄ (pH 8.0) and sonicated on ice using a microtip sonicator. Next, 5 ml of 0.4 M Na₂CO₃ (pH 11) was added, and nonlysed cells and polymers were removed by centrifugation at 6,000 × g for 20 min. The protein was precipitated using trichloroacetic acid (1:10) for at least 3 h, centrifuged (14,000 × g for 10 min), and washed with cold methanol. The protein was resuspended in 6 M guanidine–10 mM dithiothreitol, and enzymatically digested with trypsin (Promega, Madison, WI). The peptides were separated using Nano-2D-LC for 24 h (strong cation exchange/reversed phase) and analyzed by tandem mass spectrometry on a hybrid LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA), as described previously (18, 42). At least two technical replicates were run for each sample. Unmodified peptides were identified by using the SEQUEST algorithm (26) to search tandem mass spectra against the predicted protein database amdv1_reads_11212008_arman_Biofilm_AMD_CoreDB_04232008_B constructed from previously published community genomic data sets (6, 59). The SEQUEST search results were filtered and sorted with the DTASelect algorithm (56). False-positive rates for this method of analysis have been determined to be between 1 and 5% (18, 45). The database is available online (http://compbio.ornl.gov/amd_gtl_ms_results/databases/).

Normalized spectral abundance factors (NSAF) were calculated as described previously (27, 42). For clustering analysis of Gplasma proteins, organism-specific NSAF values (orgNSAF) were calculated by normalizing NSAF values to the total detected spectral counts from that organism, after which they were arcsin transformed, mean centered, and scaled as described previously (42). Only samples with >150 Gplasma proteins were used in order to prevent bias from zeros during normalization. Hierarchical clustering was carried out in a Multiple Experiment Viewer (49, 50) using a Pearson correlation coefficient and average linkage. A Fisher exact test was performed in R (47) to compare enrichment of functional categories in the group of proteins enriched in sunken samples versus enrichment among other proteins in the data set. Functional categories were defined using the Clusters of Orthologous Groups (COG) database.

Tandem mass spectra were searched again for deamidated peptides with the SEQUEST algorithm by considering the dynamic modification of deamidation for all glutamine (Q) and asparagine (N) residues. DTASelect output files were analyzed with custom Ruby scripts to calculate the spectral counts of deamidated Q and deamidated N residues, as well as the total spectral counts of Q and N residues in identified peptides. These spectral counts were summed for each organism or domain. The deamidation frequencies of Archaea and Bacteria were calculated as the percentages of the run-averaged spectral counts of deamidated Q or N over the total spectral counts of Q or N, respectively, in each domain. In order to ensure that deamidation percentages were calculated from a representative number of peptides, runs in which a single peptide contributed more than 50% of the total deamidation percentage for either archaeal or bacterial domains were not included in the analysis. By searching the reversed database (i.e., all peptide sequences reversed), false-discovery rates of deamidated peptides (∼2%) and deamidated spectra (∼3%) were identified. For analysis of differences of deamidation between growth stages and domains, data were analyzed with a two-way analysis of variance (ANOVA) in R (47) after verification that the assumptions of normality and homoscedasticity were met.

The floating biofilm was cultivated in laboratory bioreactors as previously described (7, 8). After growing to a thickness consistent with late-growth-stage biofilms, the biofilm was dislodged from the reactor walls and submerged at the bottom of the reactor chamber under ∼2 cm of slowly flowing AMD solution. After 7 days, sunken biofilm samples were taken and either immediately fixed for FISH or frozen for later proteomic analyses.

To maximize diversity for anaerobic community culturing, inocula were derived from a variety of floating and sunken biofilm samples collected from several locations within the Richmond Mine (Redding, CA). Approximately 5-ml aliquots of the mixed inocula were added to replicate 120-ml serum vials filled with 50 ml of nitrogen-degassed ferric-9kBR medium containing 100 mM Fe₂(SO₄)₃, 1 mM (NH₄)₂SO₄, 0.5 mM KCl, 0.050 mM K₂HPO₄, 5 mM MgSO₄·7H₂O, 0.2 mM CaSO₄, and trace minerals (8), all adjusted to pH 1.1 with H₂SO₄.

Triplicate cultures were supplemented with one of each of the following organic substrates: glycolate, formate, lactate, acetate, and betaine (all at 10 mM; Sigma-Aldrich), Casamino Acids (0.02%, acid-hydrolyzed casein; Becton Dickinson, East Rutherford, NJ), or peptone (0.02%, enzymatic digest of protein; Becton Dickinson). In addition, some cultures were supplemented with naturally derived biofilm substrate obtained by lyophilization of a mature floating biofilm from the Richmond Mine, followed by redissolution into sterile deionized H₂O and sonication for 5 min with a microtip sonicator. This natural substrate was added to cultures at a concentration of 0.05% (wt/vol). The sterility of this substrate was monitored via DNA extractions and microscopy in uninoculated controls. Replicates of all cultures were cultivated with or without 0.01% yeast extract. A 10% volume of each culture was transferred three times into 5-ml volumes of fresh medium every 3 to 4 weeks and transferred into 40 ml in the fourth and final transfer. Ferrozine assays of the dissolved iron concentration were performed, and DNA was extracted from cultures on the fourth transfer 28 days after inoculation. The ferrozine assays were done as previously described (55), and DNA extractions were carried out as described above.

Three sunken biofilms were collected from the 5-way site (February 2008, pH 0.98, 38°C), and three were collected from the UBA site (June 2009, pH 1.1, 38°C) (see the location map in Fig. S1 in the supplemental material). At the 5-way site, we sampled a mature, thick (∼0.1-mm) flexible biofilm floating on a ∼5-cm-deep flowing AMD solution (floating growth stage 2 [GS2], Fig. 1A). In addition, we collected three fragile sunken biofilms of similar thickness that were stratified underneath the floating biofilm and suspended close to the stream base. We labeled these biofilms “5-way–Sunken1,” reflecting its location immediately below the floating GS2 biofilm; “5-way–Sunken2,” the biofilm a few millimeters below 5-way–Sunken1; and “5-way–Sunken3,” the biofilm resting on the stream channel (Fig. 1A). Fragile (easily disintegrated) sunken biofilm samples collected from the UBA site (UBA-Sunken1, UBA-Sunken2, and UBA-Sunken3) were recovered from three different regions of a slowly draining AMD pool about 25 cm deep. All of the samples collected for the present study, and those obtained in previous studies, are listed in Table S1 in the supplemental material.

FISH was used to determine the relative proportion of Bacteria to Archaea in sunken biofilms (Fig. 2; see Table S2 in the supplemental material). In contrast to previously characterized early-growth-stage (GS1) and late-growth-stage (GS2) floating biofilms, sunken biofilms were dominated by Archaea (Fig. 3). The relative community composition of the 5-way–Sunken3 biofilm was qualitatively similar to the other 5-way samples but with clearly reduced cell density. High autofluorescence in non-DAPI filter channels, presumably caused by degraded biofilm and minerals, prohibited quantitative analysis of this sample. In the other five sunken samples analyzed quantitatively, non-ARMAN archaeal species dominated the community. ARMAN and Sulfobacillus species were more abundant in the 5-way–Sunken samples than UBA Sunken samples (see Table S2 in the supplemental material).

Clone libraries were constructed for 5-way–Sunken1 to better determine the phylogenetic diversity of the Archaea and Bacteria present. Forty-five of the 89 sequences obtained from 5-way–Sunken1 using Archaea-specific 16S rRNA primers belonged to the deeply branching clade known as ARMAN, a euryarchaeal lineage (5, 13) (Fig. 4). Eleven clones were closely related (>99% identity) to Ferroplasma acidarmanus, a species previously isolated from the Richmond Mine (25). The remaining clones represented Aplasma, Bplasma, Gplasma, and Eplasma (Fig. 4), a radiation of closely related Thermoplasma archaea, referred to as the “alphabet plasmas” (4, 19).

The Bacteria-specific 16S rRNA library contained 89 clones: 36 belonged to Leptospirillum ferrodiazotrophum, a member of Leptospirillum group III; 8 clones were related to, but distinct from Leptospirillum group III (3.0% divergent); 15 belonged to Leptospirillum group II (Leptospirillum ferriphilum); and the remainder to species of the genus Sulfobacillus (Fig. 5). Within the Sulfobacillus lineage, 10 sequences were Sulfobacillus thermosulfidooxidans (100% identity), 14 were Sulfobacillus benefaciens (100% identity), and 4 derived from other Sulfobacillus species.

Proteomic analyses of the six sunken biofilms and the 5-way–Floating biofilm were used to analyze community structure in the context of previously sampled floating biofilms as well as to identify metabolic processes active in the sunken biofilms. The relative abundance of bacterial proteins was highest in early-growth-stage (93.0%) and late-growth-stage (84.9%) floating biofilms and was lowest in sunken biofilm samples (55.0%) (Fig. 6). Archaeal proteins were most abundant in the sunken biofilm samples (24.5%) and less abundant in the early-growth-stage (1.9%) and late-growth-stage (7.1%) floating biofilms.

The proteomes of the 5-way–Floating biofilm and the three stratified sunken biofilms showed a striking increase in the relative abundance of archaeal proteins with increasing depth, 5-way–Sunken1 contained relatively more archaeal protein (7.3%) than 5-way–Floating (1.9%). Below 5-way–Sunken1, a large increase in the relative abundance of the Archaea was found in 5-way–Sunken2 (35.8% Archaea), with a similar relative abundance in 5-way–Sunken3 (35.9% Archaea) (Fig. 7).

Overall, 2,334 distinct archaeal proteins were identified across all samples analyzed (see Table S3 in the supplemental material). Of those, 545 proteins were only identified in sunken samples, 761 in floating samples, and 1,128 proteins were identified in both. Proteins detected represented many major metabolic pathways, such as tricarboxylic acid (TCA) cycle, glycolysis, fatty acid oxidation, and electron transport. Furthermore, many proteins associated with the acquisition and breakdown of organic carbon were detected, including peptide transporters, sugar transporters, peptidases, and extracellular glucoamylases (see Table S4 in the supplemental material).

Gplasma was the only archaeal species with consistently high protein abundances across samples and growth stages, enabling proteome comparisons across growth stages by hierarchical clustering of orgNSAF values. The sunken samples formed a distinct cluster, as did a group of proteins showing relatively higher abundance in the sunken biofilms (Fig. 8A). This group of proteins had a distinct distribution of COG functional category counts (Fisher exact test, P = 0.0002), with a lower abundance of proteins involved in protein biogenesis (ribosomal) and higher abundances of proteins involved in amino acid metabolism (Fig. 8B). Elevated in abundance in the sunken biofilm protein cluster were two enzymes of the TCA cycle (malate dehydrogenase and aconitate hydratase), several enzymes associated with transformations of pyruvate (three subunits of the pyruvate dehydrogenase/2 oxoacid dehydrognease complex, a cytochrome-associate pyruvate dehydrogenase, pyruvate:ferredoxin oxidoreductase, pyruvate phosphate dikinase, and acetolactate synthase), and multiple proteins that likely participate in electron transfer reactions, including an oxidoreductase, and aldehyde dehydrogenase, and a flavoprotein (see Table S5 in the supplemental material).

The relative percentage of bacterial proteins in sunken samples was higher than what was expected based on FISH analyses, so we analyzed metagenomic sequence from a subset of samples (5-way–Sunken2 and UBA-Sunken2) as a third measure of community composition. These analyses used 50 Mb of the 454 FLX Titanium sequence data obtained for each of the UBA-Sunken2 and 5-way–Sunken2 biofilms (the read lengths averaged ∼290 bp for both samples).

BLAST comparisons of 454 sequencing reads to the genomic databases showed that the 5-way–Sunken2 and UBA-Sunken2 biofilms were dominated by archaeal species (Fig. 9). In the 5-way–Sunken2 biofilm, 98.3% of reads mapping to the database matched archaeal sequences, whereas bacterial sequences represented only 1.3% of reads. In the UBA-Sunken2 sample, archaeal species comprised 69.24% of genomic reads, whereas bacterial species made up 30.68%.

Compared to FISH and metagenomic analyses, proteomic measurements showed a striking overrepresentation of Bacteria in sunken communities. Although this discrepancy could be partially explained by differences in cell size (and thus protein content) between the bacterial and archaeal community members, we hypothesized that bacterial proteins may derive from cell lysates rather than living cells. It is known that glutaminyl and asparaginyl residues can undergo an acid-dependent deamidation (12, 36, 48). Thus, if our hypothesis is true, bacterial proteins in sunken communities should have more extensive deamidation than in floating communities. In contrast, archaeal proteins should have the same degree of deamidation in floating and sunken biofilms if the cells were equally viable in both communities. We measured the degree of deamidation in proteins extracted from both floating and sunken biofilms by proteomics. On average, 28.3% of the glutamines underwent deamidation in bacterial proteins from early-growth-stage biofilms, 52.5% in late-growth-stage biofilms, and 72.1% in sunken biofilms (Fig. 10A). On the other hand, glutamines in archaeal proteins were deamidated at frequencies of 17.0 and 19.0% in early- and late-growth-stage biofilms, respectively, and at 44.2% in sunken biofilms. Similarly, the frequencies of deamidation in bacterial asparagine residues were 22.2, 36.4, and 58.1% in early-growth-stage, late-growth-stage, and sunken biofilms, respectively, whereas archaeal asparagine deamidation remained relatively constant between 21.1 and 30.3% across all sample subsets (Fig. 10B). Deamidation was not limited to any particular subset of proteins and was evident for essential cytoplasmic proteins such as ribosomal proteins, TCA cycle enzymes, and glycolysis enzymes (data not shown). A two-way ANOVA was used to test for differences among deamidation frequencies between domains of life (Archaea and Bacteria) and growth stages (GS1, GS2, Sunken). Deamidation frequencies of asparagine residues differed significantly for domains [F(1,43) = 14.11, P < 0.001], as well as between growth stages [F(2,43) = 19.53, P < 0.001]. Analysis of deamidation frequencies of glutamine residues showed similar results, with significant difference among domains [F(1,44) = 52.88, P < 0.001] and between growth stages [F(2,44) = 36.97, P < 0.001]. Importantly, there was an interaction between growth stage and domain for both asparagine [F(2,43) = 6.74, P < 0.01] and glutamine [F(2,44) = 4.94, P < 0.05] residues, indicating that the two domains respond differently to increasing growth stage and submersion. A post hoc analysis using Tukey's honestly significant difference test was used to compare deamidation frequencies among growth stages and between domains for each residue (see Table S6 in the supplemental material).

We submerged a laboratory-cultured biofilm in our laboratory bioreactors to test the hypothesis that a bacterial dominated floating biofilm would shift toward archaeal dominance if submerged. FISH results showed an increase in the relative abundance of Archaea after 7 days of submersion (increasing from 23.5% of the community prior to sinking to 52.0% after submersion). Bacteria populations, on the other hand, showed a marked decrease from 76.5% of the community in the floating biofilm, to 47.9% of the sunken biofilm. Proteomic data also showed an increase in the relative abundance of archaea, with archaeal proteins compromising 11.3% of the floating biofilm and 17.2% of the sunken biofilm, while bacterial proteins compromised 77.2 and 67.0% of the floating and sunken biofilms, respectively. Measures of protein deamidation showed high frequencies of deamidation for bacterial asparagine (20.15% in floating and 34.8% in sunken) and glutamine (30.6% in floating and 52.6% in sunken) residues but not for archaeal residues (between 10.2 and 12.6% for glutamine and asparagine in either floating or sunken samples) (Fig. 11).

In order to determine whether archaea are capable of carrying out anaerobic carbon oxidation in sunken biofilms, we enriched for anaerobic iron reducers using ferric sulfate media and a variety of different carbon sources. Iron was reduced to similar degrees in yeast-extract supplemented cultures of peptone, betaine, Casamino Acids, and AMD biofilm extract (see Fig. S2 in the supplemental material). Cultures without yeast extract were not viable even after the first transfer, except for native-biofilm cultures, which did not grow after the third transfer. No growth was apparent in cultures containing glycolate, formate, acetate, or lactate with or without yeast extract. Analysis of 16S rRNA sequences from cultures cultivated on peptone, betaine, Casamino Acids, and AMD biofilm were dominated by Ferroplasma acidarmanus, with Aplasma comprising 0 to 47% of the sequenced clones (see Fig. S2 in the supplemental material). We also detected no hydrogen consumption or production and no methanogenesis (data not shown).