Introduction

Microbial members of all three domains of life (Archaea, Bacteria, and Eucarya) are ecologically and biogeochemically important in the ocean. Microscopic primary producers such as Prochlorococcus carry out 50% of global primary production (Field et al., 1998), while approximately half of marine primary production is transferred to the ‘microbial loop’—within which energy may eventually be transferred to other trophic levels or organic material may be broken down and remineralized (Azam et al., 1983). Certain processes are known to be important in the microbial loop—particularly heterotrophic respiration and nutrient-cycling processes such as nitrification—and a number of microbial groups are known to carry out these processes. However, early molecular studies uncovered substantial diversity within marine microbial assemblages (Giovanonni et al., 1990; Fuhrman et al., 1992), and recent metagenomic studies have revealed even greater diversity in the sea (Venter et al., 2004; Sogin et al., 2006).

Many of the groups identified in these studies have yet to be assigned to a particular ecological niche, although some potential functional roles have been identified. For example, the discovery of proteorhodopsin (PR; a putative light-driven proton pump) among the ‘SAR86’ group of alphaproteobacteria (Beja et al., 2000, 2001) indicated that these organisms may use light as a source of energy. SAR86 constitute up to 10% of microbial assemblages in the ocean (Eilers et al., 2000), and PR is now recognized to be present in a wide range of microbial groups—including the ubiquitous and abundant SAR11 or Pelagibacter clade (Giovannoni et al., 2005a). Pelagibacter spp. respire organic carbon and seem to gain little benefit from light in laboratory experiments (Schwalbach et al., 2005; Giovannoni et al., 2005a), however, and a paradigm is emerging by which ‘resourceful heterotrophs’ such as the SAR11 group may use light to supplement conventional heterotrophy (Moran and Miller, 2007). Some copiotrophs may use PR to extend survival times during starvation in the light compared with dark (Gomez-Consarnau et al., 2007), or PR may have an array of physiological roles, including sensory functions (Fuhrman et al., 2008). PR was recently identified within marine group 2 Euryarchaeota (Frigaard et al., 2006), yet PR phylogeny was inconsistent with 16S rRNA phylogeny, and Euryarchaeota at deeper depths did not seem to contain PR (Frigaard et al., 2006).

In contrast to group 2 Euryarchaeota, marine group 1 (MG1) Crenarchaeota are distributed in large numbers below the euphotic zone, where they comprise ca. 30% of microbial communities (Karner et al., 2001). Summed over the volume of the deep sea, MG1 Crenarchaeota constitute one the most abundant microbial groups in the ocean, and many of these organisms likely oxidize ammonia (reviewed by Francis et al., 2007; Prosser and Nicol 2008; Erguder et al., 2009). Several ammonia-oxidizing archaea have been isolated (Könneke et al., 2005; de la Torre et al., 2008; Hatzenpichler et al., 2008), and with a few exceptions (for example, Agogue et al., 2008; but also see Beman et al. (2010), Church et al. (2010) and Santoro et al. (2010)), many studies have shown that most marine Crenarchaeota have the ability to oxidize ammonia (Francis et al., 2005; Wuchter et al., 2006; Mincer et al., 2007), and that the MG1 Crenarchaeota actively do so in the ocean (Lam et al., 2007, 2009; Beman et al., 2008; Frias-Lopez et al., 2008; Church et al., 2010; Santoro et al., 2010).

These and other studies have helped place abundant microbial groups such as SAR86, SAR11, Euryarchaeota and Crenarchaeota into prospective ecological niches, but how they may function as a community is poorly understood. Recent results from the San Pedro Ocean Time series (SPOT) site located off the coast of Southern California have combined automated ribosomal intergenic spacer analysis (ARISA) to ‘fingerprint’ bacterial communities, with terminal restriction fragment length polymorphism patterns for Eukaryotes, and quantitative PCR (QPCR) assays for archaeal groups to generate an ecological ‘network’ spanning all three domains of life (J Steele and JA Fuhrman, unpublished). In this study, we identify co-occurrence patterns for abundant archaeal and bacterial groups in the chlorophyll maximum at SPOT based on variation through time.

Materials and methods

Sample collection

Samples were collected at 33° 33′ N, 118° 24′ W at the San Pedro Ocean Time Series Microbial Observatory site located off the coast of Los Angeles, CA, USA. Oceanographic measurements have been regularly made since 1998, and a Microbial Observatory program was established for euphotic zone depths in 2000. Samples were collected at the deep chlorophyll maximum (DCM) on a nearly monthly basis from August 2000 to December 2004 using 20 L Niskin bottles deployed on a CTD rosette from the R/V Sea Watch. Water samples were sequentially filtered (Brown et al., 2005) and DNA was extracted using previously described protocols (Fuhrman et al., 1988). Further details of sampling are available in Brown et al., 2005; Fuhrman et al., 2006 and Beman et al., 2010.

QPCR, automated ribosomal intergenic spacer analysis (ARISA), and local similarity analysis (LSA)

QPCR for the Archaea was performed using primers, probes, and conditions as reported by Takai and Horikoshi (2000). MG1 and archaeal amoA QPCR assays followed established protocols (Beman et al., 2008, 2010), and all samples from 2000–2004 were analyzed in triplicate reactions. ARISA (Fisher and Triplett, 1999) was performed using primers and reaction conditions as previously described (Brown et al., 2005; Fuhrman et al., 2006, 2008); all samples from 2000 to March 2004 were analyzed. Identification of operational taxonomic units (OTUs) within ARISA profiles is based on extensive cloning and sequencing of the linked 16S-ITS-23S region detailed in Brown et al. (2005) and Fuhrman et al. (2006). Correlations among microbial groups and with environmental variables were identified over time using local similarity analysis (LSA) (Ruan et al., 2006).

Oceanographic and biogeochemical data

Salinity, temperature, dissolved oxygen, chlorophyll fluorescence and concentrations of different nutrients (nitrite, nitrate, phosphate and silicate) were measured using standard oceanographic techniques as discussed in Fuhrman et al. (2006) and Beman et al. (2010). 3H-Thymidine incorporation (Fuhrman and Azam, 1982) and 3H-Leucine incorporation (Simon and Azam, 1989) were measured using previously published methods (Fuhrman et al., 2006).

Results and discussion

Temporal dynamics of archaeal groups in the DCM

Chlorophyll fluorescence at SPOT varied significantly in space, time and magnitude (Figure 1a). Maximum fluorescence values ranged from 0–1.99 μg l−1, and were typically higher in March–May when upwelling occurs along the CA coast. From 2000–2004, the depth of the DCM varied from 14 to 70 m, with a mean value of 32 m; the DCM was typically found at greater depths in the water column during summer months reflecting enhanced stratification. Time series were generated for archaeal groups at the chlorophyll maximum using QPCR of 16S rRNA and ammonia monooxygenase subunit A (amoA) genes. Three data sets were independently generated: 16S rRNA genes of all Archaea using primers developed by Takai and Horikoshi (2000) (Figure 1b); 16S rRNA genes from MG1 Crenarchaeota (Mincer et al., 2007; Beman et al., 2008) (Figure 1c); and amoA genes from Crenarchaeota (Beman et al., 2008) (Figure 1d). These three data sets may represent subsets of one another, in that the MG1 Crenarchaeota and amoA-containing crenarchaea both constitute a subset of the total archaeal community, whereas amoA-containing crenarchaea could form a subset of the marine MG1 community, may be essentially synonymous with MG1, or may include archaeal groups other than the MG1 crenarchaea (for example, the pSL12 clade identified by Mincer et al., 2007).

Figure 1
figure 1

Times series from 2000–2004 of (a) chlorophyll from 0 to 100 m; (b) all archaeal 16S rRNA genes in the DCM; (c) MG1 16S rRNA genes in the DCM; and (d) archaeal amoA genes in the DCM. Shaded regions indicate the months from July to December.

Full size image

These possibilities were evaluated based on the three data sets, and there were three instances where MG1 16S rRNA genes exceeded all archaeal 16S rRNA genes: these three discrepancies occurred during peaks in 16S rRNA gene copies in February 2001, September 2001 and February 2002 (Figure 1b and c). Although a recent review found the Takai and Horikoshi (2000) primer and probe set to be generally effective for quantifying MG1 16S rRNA genes (Teske and Sorensen, 2007), our findings indicate that they may not detect all MG1 Crenarchaeota present in the ocean water column. This was only the case when MG1 16S rRNA genes reached their highest levels (>100 000 genes ng DNA−1), and these datapoints fell just above a 1:1 line (Supplementary Figure S1). Other than these three exceptions, comparison of MG1 16S rRNA genes and those from all Archaea showed two broad patterns: (1) instances where data points fell on a 1:1 line, and the archaeal community was predominantly MG1 Crenarchaeota; and (2) those instances where archaeal 16S rRNA genes exceeded those of crenarchaea (Supplementary Figure S1). This latter pattern represents 81% of samples, and is likely due to the presence of Euryarchaeota or crenarchaeal groups other than the MG1 Crenarchaeota.

MG1 16S rRNA genes were strongly correlated with archaeal amoA genes (r2=0.978) over time in the DCM at SPOT (Supplementary Figure S2); this is consistent with a growing number of studies that have reported clear correspondence between amoA genes and crenarchaeal cell counts (Wuchter et al., 2006), archaeal lipids (Leininger et al., 2006), or MG1 16S rRNA genes (Wuchter et al., 2006; Lam et al., 2007; Mincer et al., 2007; Beman et al., 2008, 2010; Church et al., 2010; Santoro et al., 2010). Using a different archaeal amoA primer set, Agogue et al. (2008) reported discrepancies between MG1 16S rRNA genes and archaeal amoA genes, and inferred from this lack of correlation that low-latitude deep ocean crenarchaea lack amoA. In contrast, correlation between crenarchaeal 16S rRNA and amoA genes in the SPOT chlorophyll maximum indicates that all MG1 Crenarchaeota in the DCM are amoA-containing, and that amoA-containing crenarchaeal groups other than the MG1 are not present. This in turn suggests that differences between archaeal 16S rRNA genes and MG1 16S rRNA genes reflect the presence of 16S rRNA genes from marine Euryarchaeota.

Covariation among archaeal and bacterial groups

Archaeal QPCR time series were compared with ARISA bacterial community ‘fingerprints’ using LSA (Ruan et al., 2006). Three archaeal time series were used for this comparison: archaeal amoA, MG1 16S rRNA genes, and the difference between archaeal 16S rRNA genes and MG1 16S rRNA genes (referred to as ‘probable Euryarchaeota’). Comparisons with temperature, salinity, water density, chlorophyll concentrations, oxygen concentrations, nitrite concentrations, nitrate concentrations, phosphate concentrations, silicate concentrations, microbial cell counts and bacterial production were also assessed using LSA. All significant interactions (LSA P<0.05) between archaeal time series and other parameter with a time delay of 1 month or less are reported in Table 1.

Table 1 Significant LSA relationships among archaeal data sets and bacterial OTUs identified in ARISA community fingerprints
Full size table

Both amoA and MG1 16S rRNA data were included and serve as an internal check given the strong correlation between these data sets. In fact, amoA and MG1 16S rRNA exhibited the highest LSA score (0.972), with a value similar to the Spearman's correlation reported in Supplementary Figure S2. In contrast, none of the archaeal time series exhibited significant relationships with measured physical or biogeochemical variables, including depth and density. LSA identified significant correspondence between amoA and MG1 16S rRNA genes and the same bacterial OTUs—for example, 525, 528, 633, 709 and 779 (Table 1). However, there were four instances when amoA was significantly related to a given bacterial OTU and MG1 16S rRNA was not. In these instances, LSA scores were slightly lower (0.373–0.410 compared with 0.407–0.438) and P-values were greater than 0.05 for the comparison between 16S rRNA and OTU 744 (P=0.056), OTU 684 (P=0.061), OTU 774 (P=0.098) and OTU 654 (P=0.116).

LSA also identified three inverse relationships: archaeal amoA was inversely related to unknown bacterial OTU 774, and probable Euryarchaeota 16S rRNA genes were inversely related with OTU 684 (a SAR11 group) and OTU 1051 (a Synechococcus group; Table 1). Both of these relationships involved a month time lag, with increases in Synechococcus OTU 1051 and SAR11 OTU 684 leading decreases in Euryarchaeota. Archaeal amoA genes were positively related to SAR11 OTU 684 with no delay, the only case in which MG1 or amoA were related to the same OTU as were the probable Euryarchaeota. In total, the data presented here offer limited insight into the marine Euryarchaeota: few specific interactions were identified, none of which were positive and without time delay. This may be a function of the data set used for this analysis, which was generated by difference and captures all marine euryarchaeal groups (for example, groups 2, 3 and 4; Teske and Sorensen, 2007). Little is known about Euryarchaeota in the ocean other than the fact that group 2 is often abundant in surface waters (Massana et al., 1997; Frigaard et al., 2006) and some posses PR (Frigaard et al., 2006); results from LSA suggest that the abundance of Euryarchaeota may decrease following increases in SAR11 and Synechoccocus OTUs.

Cooccurrence patterns for Crenarchaeota and abundant bacterial lineages

MG1 and amoA-containing Crenarchaeota co-occurred with multiple bacterial OTUs identified in ARISA bacterial community fingerprints, including strong correspondence between amoA genes and OTU 684 from the SAR11 clade, SAR86 OTUs 525 and 528, and Bacteroidetes OTUs 709 and 779 (Figure 2). Genome streamlining in SAR11 is indicative of adaptation to oligotrophic conditions in the ocean (Giovannoni et al., 2005b), and the dominance of this group of heterotrophs within microbial communities suggests that they are important in oceanic carbon cycling (Morris et al., 2002; Giovannoni et al., 2005a). The SAR86 clade remains uncultivated and very little is known about the physiology and biogeochemical activity of this group in the ocean—although they may constitute 10% of microbial assemblages (Eilers et al., 2000). Bacteroidetes commonly number 10–20% of bacteria and archaea in seawater (Glockner et al., 1999; Cottrell and Kirchman, 2000a), and can be particularly abundant following phytoplankton blooms (Riemann et al., 2000; Pinhassi et al., 2004) when they may specialize in the degradation of organic polymers (Cottrell and Kirchman, 2000b; Bauer et al., 2006; González et al., 2008; Kirchman, 2008). Some cultivated marine members of this group contain PR and are stimulated by light (Gomez-Consarnau et al., 2007; González et al., 2008), leading to the possibility of a ‘dual life strategy’ proposed by González et al. (2008) based on the genome of Polaribacter MED152—however, it is unclear whether Bacteroidetes OTUs 709 and 779 use a similar strategy, because they have not been cultivated and their genomes have not been sequenced.

Figure 2
figure 2

MG1 Crenarchaeota time series compared with (a) SAR11 OTU 684; (b) SAR86 OTUs 525 and 528; and (c) Bacteroidetes OTUs 709 and 779. MG1 Crenarchaeota data are expressed in 16 rRNA genes ng DNA−1 (left axis), and bacterial OTUs are expressed as percent of total ARISA peak area (right axis). Shaded regions correspond the months from September to March.

Full size image

The relative abundance of Bacteroidetes OTUs 709 and 779 closely tracked SAR11 OTU 684, and were also correlated with MG1 Crenarchaeota over time (Figure 2a and c). Although SAR11 and the marine Crenarchaeota are typically characterized by different depth distributions—with SAR11 dominant in surface waters (Morris et al., 2002) and Crenarchaeota at depth (Karner et al., 2001)—SAR11 is comprised of multiple subclades that partition with depth in the marine water column (Field et al., 1997; García-Martínez and Rodríguez-Valera, 2000; Morris et al., 2002, 2005; Brown and Fuhrman 2005; Stingl et al., 2007): SAR11 groups Ia and Ib are typically most abundant in surface waters and group II in deeper waters. Crenarchaeota also display depth-related partitioning based on 16S rRNA (Massana et al., 1997; García-Martínez and Rodríguez-Valera, 2000), ITS (García-Martínez and Rodríguez-Valera, 2000), amoB (Mincer et al., 2007) and amoA (Hallam et al., 2006; Mincer et al., 2007; Beman et al., 2008; Santoro et al., 2010) sequences. García-Martínez and Rodríguez-Valera (2000) in fact used 16S rRNA sequences and ITS sequences from SAR11 and MG1 Crenarchaeota to examine both groups in the same samples, and found that SAR11 and MG1 partitioned into different ‘ecotypes.’ Significantly, MG1 Crenarchaeota often exhibit peak absolute abundance (cells ml−1 or 16S rRNA genes ml−1) and high temporal variability at the base of the euphotic zone (Karner et al., 2001; Mincer et al., 2007; Beman et al., 2008, 2010), and SAR11 may share these properties (Eiler et al., 2009). In the SPOT chlorophyll maximum, SAR11 OTU 684 and Crenarchaeota were correlated with one another through time (r2=0.55 for amoA, P<0.01; r2=0.51 for 16S rRNA, P<0.05). Peaks in MG1 Crenarchaeota occurred in February and September 2001, February and September of 2002, and May 2003, whereas peaks in SAR11 OTU 684 occurred in February 2001 and February, August and September of 2002 (peaks defined as greater than one standard deviation above the mean; Figure 2).

OTU 684 represents SAR11 group 1b, and in the Sargasso Sea, group Ib was ‘transitional’: it was most abundant when the water column was not stratified, and hence was not divided into surface and deep niches occupied by different ecotypes (Ia and II; Morris et al., 2005; Carlson et al., 2009). Our results from SPOT are consistent with this, as we found group 1b to be periodically abundant, whereas other ‘surface’ SAR11 OTUs were dominant for most of the time series—SAR11 group 1a OTUs 666 and 669 together comprised 14–37% of ARISA peak area, averaged 21%, and were correlated with a Prochlorococcus OTU (J Steele and JA Fuhrman, unpublished data). Our data furthermore indicate that ammonia-oxidizing archaea are transiently abundant in the DCM and closely track SAR11 OTU 684 (Figure 2a). It is possible that similar covariation among these groups occurs in the Sargasso Sea, as nitrite is the product of ammonia oxidation and can accumulate in high concentrations during periods of winter mixing (Lomas and Lipschultz, 2006) when SAR11 group Ib is also abundant (Morris et al., 2005; Carlson et al., 2009). However, to date, ours is the only study to have examined temporal variability in both of these groups concurrently.

Strong seasonal variability is typical for the Southern California Bight, as productivity is greatest in the months of March, April and May due to shoaling of the nutricline (Hayward and Venrick, 1998). SPOT also experiences seasonal changes in the strength of the Southern California Eddy or Inshore Current; the Southern California Eddy is located inshore of the equatorward flow of the California Current (an eastern boundary current) and results in maximum poleward flow along the coast of southern California in summer through early fall and also in winter (and weakens considerably in spring) (Hickey, 1979; Lynn and Simpson, 1987). These seasonal changes are known to affect phytoplankton community composition: distinct phytoplankton assemblages vary in time and space, and include an inshore group of abundant species that ‘reaches maximum abundances…in the spring, almost certainly responding to the increased nutrient concentrations’ (Venrick, 2002). The fluctuations of this group are superimposed on a less abundant background assemblage that is more important in offshore regions, and a few shared species form a ‘transitional’ phytoplankton community (Venrick, 2002).

Co-occurrence of Crenarchaeota, SAR11 OTU 684, SAR86 OTUs 525 and 528, and Bacteroidetes OTUs 709 and 779 occurred at a transition between periods of lower microbial abundance, production and nutrient concentrations to more replete conditions (Figure 3). The depth of the chlorophyll maximum did not drive these dynamics, as it was only weakly correlated with amoA based on Spearman's ranked correlation (r2=0.18, P>0.05), and neither depth nor density was identified by LSA as a significant correlate for any of these microbial groups. Instead, this pattern of increased abundance, production and nutrient concentrations in spring and summer was consistent from year to year (Figure 3). Carlson et al. (2009) hypothesized that different forms of dissolved organic carbon affect the abundance of different SAR11 ecotyopes in the Sargasso Sea, whereas Eiler et al. (2009) modeled the abundance of all SAR11 clades using environmental data: in the North Pacific Subtropical Gyre, dissolved organic carbon, combined nitrate and nitrite concentrations, depth, silicate and phosphate were the best predictors of SAR11 abundance (Eiler et al., 2009). Previous laboratory studies have also shown that the growth and activity of nitrifying bacteria may be enhanced in mixed cultures with heterotrophic bacteria (Gundersen, 1955; Clark and Schmidt, 1966; Steinmüller and Bock, 1976; Jones and Hood, 1980), and Jones and Hood (1980) specifically demonstrated that the ability of Nitrosomonas spp. to oxidize ammonia increased by 50–150% in the presence of two heterotrophs (whose growth was in turn enhanced by as much as an order of magnitude).

Figure 3
figure 3

Time series for SAR11 OTU 684 (a) compared with (b) cell counts and (c) nitrate concentrations for each month of the year at SPOT. Smaller gray symbols show the mean values averaged for each month; values for individual months in each year are shown in different colors.

Full size image

In the DCM off the Southern California Bight, we propose that ammonia-oxidizing archaea, and specific SAR11, SAR86, and Bacteroidetes OTUs constitute a specialized microbial loop that is present during seasonal transitions. Together these groups may be involved in the breakdown of organic material and recycling of nutrients, and given the similarities in the behavior of the SAR11 group 1b off the coast of Southern California and in the Sargasso Sea, we hypothesize that this community may assemble (that is, become relatively more abundant and active) during similar transition periods in other areas of the ocean. Such co-occurrence patterns are observationally identifiable using the time series approaches that we have used here, and interactions among these groups could be examined experimentally using stable isotope probing or similar techniques. In addition to examining the physiological properties that have made these microorganisms successful ecological competitors in the sea (Giovannoni et al., 2005b; González et al., 2008; Martens-Habbena et al., 2009), future research should determine whether similar ‘community assembly rules’ (Fuhrman, 2009) apply to these groups elsewhere in the ocean, and what ecological and biogeochemical processes shape these patterns.