Planktonic Euryarchaeota are a significant source of archaeal tetraether lipids in the ocean

Significance

All three domains of life—Eukarya, Bacteria, and Archaea—are known to inhabit the marine water column from surface waters to great depth. Planktonic marine Archaea are comprised of two dominant groups—the Thaumarchaeaota and Euryarchaeota. The Thaumarchaeota contain characteristic biomarker lipids known as tetraethers that are thought to be diagnostic for this group, and are used as paleotemperature proxies, since these lipids are well preserved in marine sediments. In this study, we show that planktonic Euryarchaeota produce the same types of archaeal tetraether lipids as do Thaumarchaeaota. Our results have important implications for environmental surveys of marine Archaea, and the use of their lipids for interpretation of the sedimentary record.

Abstract

Archaea are ubiquitous in marine plankton, and fossil forms of archaeal tetraether membrane lipids in sedimentary rocks document their participation in marine biogeochemical cycles for >100 million years. Ribosomal RNA surveys have identified four major clades of planktonic archaea but, to date, tetraether lipids have been characterized in only one, the Marine Group I Thaumarchaeota. The membrane lipid composition of the other planktonic archaeal groups—all uncultured Euryarchaeota—is currently unknown. Using integrated nucleic acid and lipid analyses, we found that Marine Group II Euryarchaeota (MG-II) contributed significantly to the tetraether lipid pool in the North Pacific Subtropical Gyre at shallow to intermediate depths. Our data strongly suggested that MG-II also synthesize crenarchaeol, a tetraether lipid previously considered to be a unique biomarker for Thaumarchaeota. Metagenomic datasets spanning 5 y indicated that depth stratification of planktonic archaeal groups was a stable feature in the North Pacific Subtropical Gyre. The consistent prevalence of MG-II at depths where the bulk of exported organic matter originates, together with their ubiquitous distribution over diverse oceanic provinces, suggests that this clade is a significant source of tetraether lipids to marine sediments. Our results are relevant to archaeal lipid biomarker applications in the modern oceans and the interpretation of these compounds in the geologic record.

Early cultivation-independent molecular surveys led to the discovery of planktonic marine Euryarchaeota (1) and Thaumarchaeota (formerly called Crenarchaeota) (1, 2), and have since been used to describe the abundance and ecological distributions of archaeal groups in diverse ocean biomes (3). Metagenomic analyses and the isolation of several marine Thaumarchaeota have provided further insight into their physiology and biogeochemistry. Distinctive archaeal tetraether membrane lipids (SI Appendix, Fig. S1) have also been reported throughout the oceans (4). These compounds, collectively referred to as glycerol dialkyl glycerol tetraethers (GDGTs), have been useful tracers of archaeal biomass (5) and, via their isotopic composition, have provided new information about archaeal community carbon metabolism (6–8). GDGTs are relatively recalcitrant; they can be exported with little alteration to marine sediments, where their distributions have been exploited to develop proxies for reconstructing sea surface temperature (9) and terrigenous organic matter input (10). On the basis of GDGT abundances in a black shale dated at 112 million years (11), it was suggested that archaea have been significant members of marine ecosystems since at least the Mesozoic Era.

Although tetraether lipids are used with increasing frequency in paleoceanography and microbial ecology, their specific taxonomic sources in the water column are not well constrained. Of the four groups of planktonic archaea identified in the oceans, representatives of only one—the Marine Group I (MG-I) Thaumarchaeota (1, 2, 12)—have been isolated in pure culture. All MG-I strains isolated to date are chemolithoautotrophic, fixing inorganic carbon via energy obtained from the oxidation of ammonia to nitrite (13). Recent evidence suggests that MG-I also contribute to the flux of potent greenhouse gases nitrous oxide (14) and methane (15) from the water column to the atmosphere. The membrane lipid assemblage of MG-I includes GDGTs with zero through four cyclopentyl moieties and crenarchaeol, a GDGT containing one cyclohexyl and four cyclopentyl moieties (16). Crenarchaeol has been considered uniquely diagnostic for Thaumarchaeota (17) and, by extension, has been postulated as a biomarker for archaeal nitrification (18).

In addition to MG-I Thaumarchaeota, three other groups of archaea—all Euryarchaeota—inhabit the marine water column: Groups II (1), III (19), and IV (20). Of these, Marine Group II (MG-II) are the most abundant and frequently detected, often but not exclusively in near-surface waters. MG-II inhabit diverse oceanic provinces including the oligotrophic North Pacific Subtropical Gyre (21, 22), coastal California (23–25), the North Sea (26), Arctic (27, 28) and Antarctic (29–31) waters, the coastal Mediterranean Sea (32), the eastern tropical South Pacific oxygen minimum zone (33), waters surrounding a tropical atoll (34), the deep North Atlantic (35), and the East China Sea (36). The potential contribution of this cosmopolitan group to the marine tetraether lipid pool has been debated (37–40), but the lack of cultivated representatives of MG-II has precluded direct analysis of their membrane lipids, and incomplete knowledge of the genetic basis of archaeal tetraether lipid biosynthesis limits the ability of metagenomic studies to address this question.

We conducted parallel characterization of planktonic archaeal lipids and archaeal community DNA composition to constrain the taxonomic origins of GDGTs in marine plankton and their distribution in the marine water column. Metagenomic analyses, quantitative polymerase chain reaction (qPCR) of Thaumarchaeota marker genes, and small subunit ribosomal RNA (SSU rRNA) gene amplicon pyrosequencing were used to characterize the overall archaeal community composition in size-fractionated suspended particulate matter (SPM). Parallel measurements of archaeal tetraether lipids from the same filter samples enabled us to associate GDGTs with particular members of the planktonic archaeal community.

Results and Discussion

SPM for this study was collected in the North Pacific Subtropical Gyre (NPSG) in September 2011 using in situ pumps deployed at depths ranging from 11 to 559 m (SI Appendix, Figs. S2 and S3). SPM was fractionated into two size classes (0.3–3 µm and 3–57 µm) in an effort to separate the archaeal community into free-living and particle-associated/detrital components. Lipids and DNA (SI Appendix, Fig. S4) were extracted from the same filters.

Depth Distributions of Archaeal Phyla.

We characterized archaeal communities in our sample set using two independent techniques: metagenomics and pyrosequencing of amplicons of the V1−V3 region of the SSU rRNA gene in DNA. Counts of euryarchaeotal and thaumarchaeotal 16S rRNAs and functional gene sequences in metagenomic datasets (Table 1) indicated that Euryarchaeota outnumbered Thaumarchaeota at depths above 157 m in the 0.3- to 3-µm size fraction and in the 3- to 57-µm size fraction throughout the entire profile. Notably, at 83 m depth, the archaeal population was comprised nearly exclusively of MG-II.

Table 1.

Archaeal community composition and relative abundances of crenarchaeol and GDGTs with zero to three cyclopentyl moieties in suspended particulate matter samples

Size, µm	z, m	% Eury		Relative abundance of GDGT cores
		rDNA	MGA	0	1	2	3	Cren
0.3–3	83	97.2	94.4	22.4	8.2	8.1	0.1	61.1
	131	74.4	66.7	21.8	8.3	12.1	6.0	51.7
	131*	74.4	66.7	19.6	7.5	10.6	4.3	58.1
	157	35.3	26.4	19.1	6.0	10.0	3.9	61.0
	231	47.0	44.7	25.9	7.0	13.2	0.9	53.0
	331	36.2	28.9	40.3	5.6	8.6	0.0	45.5
	450	20.1	12.8	9.8	7.4	68.8	0.0	14.0
	559*	ND	ND	25.5	10.3	13.7	0.4	50.1
3–57	83	100.0	96.1	38.6	12.8	1.2	0.0	47.4
	131	85.3	87.9	17.5	6.7	8.6	3.3	63.9
	157	75.0	74.3	18.0	6.7	11.4	3.6	60.3
	231	86.3	78.5	19.7	5.8	14.3	1.9	58.3
	331	71.0	66.1	21.2	5.0	16.8	0.7	56.3
	450	85.2	60.9	26.3	7.3	11.2	1.5	53.7

Archaeal community composition was determined from MGA and 16S rDNA gene representation in metagenomic datasets. % Eury, % Euryachaeotal; Cren, crenarchaeol; MGA, metagenomic read annotations; z, depth.

^*Hydrolysates of total lipid extracts.

The 16S rRNA amplicon data provided a more detailed taxonomic breakdown of MG-I and MG-II phylotypes in the different size fractions and depths. We focused on samples from eight depths that yielded >1,000 amplicons in each size fraction (SI Appendix, Figs. S5 and S6 and Table S2), and the 0.22- to 1.6-µm size fraction from 800 m and 1000 m. In total, this sample set contained 323,924 archaeal amplicon sequences after curation and quality control. Archaeal rRNA sequences were clustered into operational taxonomic units (OTUs) using the software program QIIME (41) and taxonomy assigned through comparison with the ARB-SILVA SSU reference database using BLASTn.

The 10 most prevalent archaeal OTUs across the entire dataset (each representing >2,000 amplicon sequences and, in total, 82% of archaeal amplicons) were assigned to either the Thaumarchaeota (primarily MG-I) or MG-II Euryarchaeota (Fig. 1). Among the Thaumarchaeota, OTU D was predominant in both size fractions of SPM and represented a higher percent of total archaeal OTUs with increasing depth. At 1,000 m, OTU A had a higher relative abundance than OTU D, and OTU C rose in abundance, consistent with previous reports of genetic diversity of Thaumarchaeota in depth profiles (22).

Fig. 1.

Archaeal representation in different sample depths and size fractions. (A) Phylogenetic analysis (maximum likelihood; see SI Appendix) of the top 20 OTUs recovered by rRNA PCR (labeled in orange; 97% similarity level) that represent the majority of sequences across the dataset. (B and C) The depth and size class distribution of the 10 most prevalent archaeal OTUs. Darkest gray represents the most abundant OTU in a sample.

In sharp contrast to Thaumarchaeota, MG-II Euryarchaeota represented a much higher proportion of total archaeal OTUs at shallower depths, and the MG-II population was not dominated by a single phylotype (Fig. 1). Four OTUs—E, F, H, and I—were abundant in the 131- and 157-m samples, and OTU I became more prominent with depth. Such intragroup stratification may reflect niche partitioning, since some MG-II photic zone ecotypes encode the light-driven proton pump proteorhodopsin (42, 43), whereas mesopelagic MG-II ecotypes tend to lack the proteorhodopsin gene (42). Patterns of MG-II OTU abundance also varied between size classes; at 131 m, for instance, OTU F was more highly represented in the 3- to 57-µm fraction than in the 0.3- to 3-µm fraction.

The near-binary nature of the archaeal community in the sample set enabled us to estimate MG-II cell densities using MG-I rRNA qPCR data and the relative abundances of MG-I and MG-II determined from metagenomes (Eq. 1 and Fig. 2 A and B). Overall, MG-I and MG-II distributions were similar to those reported in a previous NPSG study (22). Two independent metrics we report here also support the dominance of MG-II in the upper ocean: At 83 m, MG-I were not detectable by qPCR (Fig. 2 A and B), and Euryarchaeota comprised >94% of the 16S rRNA and archaeal metagenomic read annotations (MGAs) (Table 1).

Fig. 2.

Inferred cell densities of Thaumarchaeota and Euryarchaeota and core GDGT lipid concentrations in small (A and C) and large (B and D) size fractions of suspended particulate matter in the NPSG. MG-I cell densities (solid blue line) were determined by qPCR, and were below limits of detection at depths <100 m. MG-II cell densities (dashed red line) were inferred from archaeal community composition as determined from MG-I cell densities using SSU rRNA gene representation in metagenomic datasets (Eq. 1). Mean values of replicate GDGT analyses are shown; error bars indicating the range of measurements are smaller than the data points. Note variable scales on x axes.

Both MG-I and MG-II DNAs were more abundant in the 0.3- to 3-µm than in the 3- to 57-µm size fraction, consistent with a predominantly free-living existence (FISH) (21, 24). Both size class MG-I profiles (Fig. 2 A and B) showed a gradual increase in cell density with depth, and suggest a small percentage (10–15%) of a predominantly free-living thaumarchaeotal population was captured on the large filters.

MG-II depth profiles showed more pronounced differences between size fractions, with MG-II cells captured on the large filter ranging from 2% to 90% of the total MG-II detected at a given depth. This variability may be related to differences in cell size and the extent of particle association between ecotypes, as MG-II have been reported to be pleomorphic and larger than MG-I cells (44).

Tetraether Lipid Distributions and Partitioning.

Core GDGT molecules (SI Appendix, Fig. S1) are composed of two isoprenoidal C₄₀ hydrocarbon cores linked to glycerol by ether bonds at each end. Many GDGTs contain internal cyclopentyl moieties, and crenarchaeol has an additional six-membered ring (SI Appendix, Fig. S1). In intact polar lipid (IPL) GDGTs, glycerol is linked to polar head groups, often hexose moieties. Because polar head groups have been found to degrade rapidly after cell death (45), IPLs in sediments have been proposed as indicators of live biomass. It has been presumed that only a small proportion of GDGTs in live cells exist as free core GDGTs, but how the relative proportion of these lipids varies with environmental conditions, taxa, or growth states, is not well constrained.

We measured core GDGTs in lipid extracts using high-performance liquid chromatography−atmospheric pressure chemical ionization (APCI) mass spectrometry (46, 47). In the small size fraction samples, free core GDGT concentrations were 1–20 pg/L and peaked at 157 m. Free core GDGTs (Fig. 2 C and D) were most abundant in the large size fraction, where their concentrations ranged from 16 to 490 pg/L. They reached an upper water column maximum at 331 m, but the highest concentration was measured at 559 m. Similar partitioning, with a higher proportion of core GDGTs detected in the larger size fraction, was reported in a study of size-fractionated SPM collected in Puget Sound (48). Notably, concentrations of core GDGTs (including crenarchaeol) in the 83-m samples dominated by MG-II (94–100% Euryarchaeota) were similar to those in deeper (>300 m) samples that had high MG-I representation (Fig. 2 C and D).

Intact Polar GDGTs.

Several series of very late eluting peaks were prominent in extracted ion chromatograms of masses corresponding to GDGTs 0–3 and crenarchaeol in the small size fraction samples from 131 and 559 m (Fig. 3 and SI Appendix, Fig. S7). Within each series, the order of retention times was identical to that of commonly detected free GDGTs. Because they eluted at the stage of analysis at which HPLC eluents are most polar, we postulated that these peaks represented IPL GDGTs that lost their head groups during in-source fragmentation in the mass spectrometer. A hydrolysis experiment supported the IPL GDGT hypothesis: After the extracts were treated with hydrochloric acid (49), the late-eluting peaks disappeared and the peak areas of core GDGTs increased (Fig. 3 and SI Appendix, Fig. S7). This difference was most pronounced in the 131-m sample, in which concentrations of IPL-bound GDGTs were 123 pg/L, exceeding those of free core GDGTs (5 pg/L) by more than an order of magnitude. Euryarchaeotal DNAs were predominant in this sample (74% euryarchaeotal DNAs, 26% thaumarchaotal). At 559 m, the depth at which Thaumarchaeota reached a maximum, IPL-bound GDGT concentrations were 21 pg/L. Comparisons with samples in which IPL GDGTs had previously been reported using established techniques (50) suggested a large contribution from monoglycosyl GDGTs (SI Appendix, Fig. S1) at both 131 and 559 m. Identification of monoglycosyl crenarchaeol was confirmed on an accurate-mass quadrupole time-of-flight mass spectrometer (51) (SI Appendix, Fig. S8).

Fig. 3.

HPLC-APCI-MS composite extracted ion chromatograms (EIC) of total lipid extract (TLE) from 131 m, 0.3–3 µm size fraction sample before (A) and after (B) acid hydrolysis. Colored traces are EICs of m/z values of individual core GDGTs and the hydrolysis-resistant internal standard. Late eluting peaks in A represent three putative series of core GDGTs released from IPL GDGTs by fragmentation in the APCI source. After acid hydrolysis, late-eluting peaks disappeared and the peak area of core GDGTs increased, supporting this interpretation.

Marine Euryarchaeota Contain Crenarchaeol and Other Tetraether Lipids.

Two lines of convergent evidence indicate that MG-II Euryarchaeota are a major source of tetraether lipids in the NPSG and, by extrapolation, throughout the open ocean and in coastal waters where Euryarchaeota are found (3, 23–25). First, core GDGTs—including crenarchaeol—were detected in both SPM size fractions throughout the profile, most notably at depths at which MG-II dominated the archaeal population (Fig. 2). The 83-m samples composed nearly exclusively of Euryarchaeota (95–100%) contained as much core GDGTs as deeper samples with higher MG-I representation (Fig. 2 C and D). Because the turnover time of cells in the upper water column is on the order of a few days (52), the free core GDGTs we measured must have been derived from living or recently living archaea found at these depths. IPL-bound GDGTs (53–55) provide a second line of evidence for a MG-II contribution to the tetraether lipid pool. Concentrations of these lipids at 559 m (dominated by MG-I) were less than one-fifth those measured at 131 m (dominated by MG-II). It therefore appears unlikely that MG-I, less abundant at 131 m than at 559 m, could be the sole source of the much higher IPL GDGT concentrations detected in the shallower sample.

Similar distributions of MG-I and MG-II were reported in a previous metagenomic study using samples collected in 2002–2004 (22); this continuity suggests that the pattern of MG-II dominance in the upper open ocean is not ephemeral. To improve the temporal resolution of the historical record, we queried datasets from four metagenomic profiles and one metatranscriptomic profile generated from the NPSG against a comprehensive database that included a draft MG-II metagenome (43). Subsequent analysis of the relative contributions of Euryarchaeota and Thaumarchaeota to total protein-coding reads (SI Appendix, Fig. S9) revealed a pattern similar to that seen in the amplicon dataset. Thus, fosmid, amplicon, and pyrosequenced metagenomic datasets from samples spanning nearly a decade all indicated that euryarchaeotal dominance in the upper ocean is a persistent feature of the NPSG.

Our finding that Euryarchaeota are a major source of GDGTs in the water column is consistent with the observation that tetraether lipid biosynthesis—a broadly but not universally distributed trait among Euryarchaeota—is common among members of the Thermoplasmatales (56), the most closely related cultivars to MG-II. Moreover, biosynthesis of crenarchaeol by MG-II was previously proposed on the basis of water column GDGT and DNA distributions (37–40). A role for the cyclohexyl moiety in maintaining tetraether membrane fluidity at low temperatures, possibly enabling the expansion of thermophilic MG-I from hydrothermal environments to the open ocean, was postulated (17) but later contradicted by reports of crenarchaeol in hot springs (57, 58) and thermophilic MG-I isolates (18, 59).

Although crenarchaeol biosynthesis has been confirmed in MG-I isolates (e.g., ref. 16), determining its sources—and sources of other tetraether lipids—in environments in which multiple archaeal groups coexist has been difficult. Studies examining correlations between crenarchaeol and MG-I SSU rRNA or amoA gene transcript copy numbers in the water column may have been complicated by an early focus on core rather than IPL GDGTs, and the use of filters with different pore sizes for collection of lipids and DNA, and for FISH counts. The absence of an apparent correlation between community composition and the relative proportions of individual GDGTs detected in this study (Table 1) suggests that it may not be possible to structurally differentiate marine planktonic GDGT pools in the case of mixed MG-I and MG-II archaeal populations.

Geological and Ecological Implications.

Together with the widespread predominance of MG-II in the epipelagic zone, where exported GDGT signals have been reported to originate (60, 61), our evidence that MG-II are a source of tetraether lipids suggests that a significant proportion of GDGTs delivered to sediments are likely synthesized by Euryarchaeota. This finding has important implications for past and future oceanographic and organic geochemical studies, and may necessitate a reinterpretation of archaeal tetraether lipid distributions in the geologic record.

The paleotemperature proxy TEX₈₆ (TetraEther indeX of tetraethers containing 86 carbon atoms) relies on an empirical relationship between sea surface temperature (SST) and relative abundances of GDGTs in sediments that have been presumed to be derived from planktonic Thaumarchaeota (9, 62). Although TEX₈₆ is now widely used in paleoceanography and core top calibration studies have found correlations between temperatures inferred using it and SSTs to be strong (62), mechanisms underlying the proxy are not well understood (63). One critical but as yet unresolved issue is how SST signals can be closely reflected in the membrane lipid composition of planktonic Thaumarchaeota, which are typically less abundant in epipelagic waters, where much exported organic matter originates. Tetraether lipid biosynthesis by MG-II provides a possible solution: If temperature affects the membrane lipid composition of MG-II Euryarchaeota, MG-II GDGTs are more likely than those of deeper-dwelling MG-I Thaumarchaeota to reflect sea surface temperatures. Significant physiological differences between the groups, however, may complicate this scenario. Diverse heterotrophic MG-II populations may be influenced by very different ecological factors than those that impact chemolithoautotrophic MG-I populations. Heterotrophic MG-II populations likely respond to changes in organic matter availability and primary productivity, and their abundances have been observed to vary seasonally in a variety of habitats (24, 64).

Because TEX₈₆ relies on relative rather than absolute abundances of GDGTs to estimate SST, it may in theory be immune to changes in GDGT flux or changes in community composition, provided that temperature is the only factor influencing archaeal membrane lipid composition in the ocean. This assumption is probably too simplistic, because the tetraether lipid composition of cultured archaea is known to vary with factors such as pH (65).

Analyses of one sample in this study (440 m, small size fraction) underscored the fact that our understanding of factors influencing relative proportions of GDGTs in the water column remains incomplete. This deep sample showed a GDGT pattern atypical of marine waters (SI Appendix, Fig. S10), with GDGT-2 more abundant than GDGT-0 or crenarchaeol—yet the archaeal community was dominated by the same MG-I OTU as at other depths with similar in situ temperatures. Neither temperature nor community composition seem likely explanations for this unusual GDGT signal, although the full genetic diversity of MG-I ecotypes is not entirely reflected in rRNA data.

A mixed biological origin for GDGTs detected in marine sediments may confound efforts to use universally applicable paleoceanographic tools. A recent attempt to detect marine carbon isotope excursions by measuring the carbon isotopic composition (δ¹³C) of GDGTs (66) made the assumption that these compounds were synthesized by autotrophic Thaumarchaeota and argued that the δ¹³C of GDGTs must covary with that of the dissolved inorganic carbon (DIC) substrate. If, however, heterotrophic Euryarchaeota also contributed to the GDGT pool (as now appears likely), the isotopic signal would be mixed, reflecting the δ¹³C of their organic carbon substrates as well as that of DIC.

Conclusions

The results presented should help guide future studies of planktonic archaea and their lipids in the marine water column and sediments. The enrichment and isolation of representative strains of MG-II would greatly aid attempts to understand its ecology, physiology, and membrane lipid composition, but further cultivation-independent studies are also likely to yield new information about this important group of marine plankton.

Isotopically labeled tracer experiments have high potential for probing MG-II populations. Early microcosm studies using ¹³C labeled bicarbonate amendments (67) were valuable in predicting the physiology of MG-I communities before any representatives were isolated. More recently, ¹³C tracer experiments have been successful in attributing the production of specific IPL classes to their autotrophic/heterotrophic bacterial sources (68). Similar experiments using labeled organic substrates may prove equally informative regarding the physiology and membrane lipid composition of MG-II Euryarchaeota.

In addition to better-constrained GDGT sources, a more complete knowledge of export processes and residence times will be necessary to interrelate the distributions of GDGT-containing archaea in the water column with their lipid profiles in the sedimentary record. Sediment trap studies, coupled with lipid and nucleic acid analyses of SPM and sediments, may reveal the depths from which exported GDGTs originate. Such investigations, conducted in both open ocean and coastal environments, in transects, profiles, and time series, should also help constrain the balance of euryarchaeotal/thaumarchaeotal lipids exported to the seafloor in different oceanic provinces and seasons. Further analyses of size-fractionated SPM, combining archaeal community and lipid composition measurements with compound-specific radiocarbon analyses and thorium-based measures of particle flux, may also yield more insight into the mechanisms and timescales of GDGT export.

Materials and Methods

Sample Collection.

Size-fractionated SPM samples (0.3–3 µm and 3–57 µm) were collected in September 2011 (SI Appendix, Figs. S2 and S3; details in SI Appendix) near Hawaii Ocean Time-series Station ALOHA using in situ pumps fitted with glass fiber filters.

DNA Methods.

DNA was extracted from filter sections corresponding to 20–70 L of seawater (SI Appendix, Fig. S4). The qPCR assays used primers MGI_751F; 5′-GTCTACCAGAACAYGTTC and MGI_956R; 5′-HGGCGTTGACTCCAATTG and generally followed the program of Mincer et al. (24) (details in SI Appendix). We amplified the V1−V3 variable region of the archaeal SSU rRNA gene using the primers 20F 5′-TCCGGTTGATCCYGCCRG-3′ (23) and a barcoded reverse primer 519R, 5′-GGTDTTACCGCGGCKGCT-3′ (69). Pooled amplicons were sequenced using the 454 Genome Sequencer (Roche; details in SI Appendix). We used AmpliconNoise (70) to remove 454 sequencing errors, QIIME (41) to demultiplex the dataset and cluster sequences into OTUs at the 97% identity level using OTU-picking scripts based on uclust (71), and the Ribosomal Database Project classifier to assign taxonomy for 4,900 OTUs. A matrix of predominant OTUs was generated; darkest gray indicates the most abundant OTU in a single sample (Fig. 1). A maximum likelihood tree of the 20 most abundant OTUs was constructed in ARB (72) and visualized in ITOL (73) (see SI Appendix).

Metagenomic data and analyses.

A subset of DNA samples were sequenced on an Illumina MiSeq Sequencer. Historical contributions of MG-I and MG-II to total protein-coding reads were determined from metagenomic datasets generated from SPM profiles collected on Hawaii Ocean Time-series (HOT) research cruises 179 [March 2006 (74)], 186 [October 2006 (75)], 194 (August 2007), and 215 (September 2009) and a metatranscriptomic dataset from HOT 186.

Illumina datasets.

Reads were curated with Trimmomatic (76) and joined using PandaSeq (77); unjoined read pairs were retained and tracked as single records. Trimmed reads were filtered using SortMeRNA (78) to identify those containing rRNA sequences (SI Appendix); taxonomy was determined by searching against SILVA release 115 using lastal (79).

454 datasets.

Duplicate reads (100% identity) were removed and FASTA files divided into rRNA and non-rRNA reads by using BLASTn (top hit, e-values <0.0001, bit scores >50) to compare with a reference database comprised of combined SILVA release 106 SSU and LSU databases and a 5S database (80). Using BLASTx or LAST (79), non-rRNA reads were compared with a reference database (SI Appendix) and their taxonomic affiliation determined in MEGAN (81). The proportion of thaumarchaeal and euryarchaeal reads was calculated as a percent of total reads assigned at or below phylum level.

Lipid extraction, hydrolysis, and analysis.

Lipids were extracted from glass fiber filters containing SPM following a modified Bligh−Dyer protocol (50) (details in SI Appendix). A C₄₆ tetraether (82), which lacks a polar head group and is unmodified by acid hydrolysis, was used as an internal standard. Aliquots of TLE were subjected to acid hydrolysis to cleave head groups of intact polar GDGTs, converting them to core GDGTs (50) (SI Appendix). Core GDGTs were analyzed by high-performance liquid chromatography−APCI mass spectrometry (47, 48) (details in SI Appendix). IPLs in TLEs of the 0.3- to 3-µm size fraction of SPM from 131 and 559 m were analyzed by ultra-high-performance liquid chromatography−electrospray ionization mass spectrometry(51) (SI Appendix).

Cell density inference.

We inferred MG-II cell densities based on qPCR and community composition (SI Appendix) using Eq. 1, where n = number and f = fraction:

$$n_{\mathit{MG}\text{-}\mathit{II}}=f_{\mathit{MG}\text{-}\mathit{II}}\frac{n_{\mathit{MG}\text{-}\mathit{I}}}{f_{\mathit{MG}\text{-}\mathit{I}}}.$$ [1]