Microbial Ecology of Expanding Oxygen Minimum Zones

Home , Actinobacteridae, Lentisphaerae, Oxygen minimum zone, Polaribacter

REVIEWS

Microbial ecology of expanding oxygen minimum zones

Jody J. Wright1, Kishori M. Konwar1 and Steven J. Hallam1,2 Abstract | Dissolved oxygen concentration is a crucial organizing principle in marine ecosystems. As oxygen levels decline, energy is increasingly diverted away from higher trophic levels into microbial metabolism, leading to loss of fixed nitrogen and to production of greenhouse gases, including nitrous oxide and methane. In this Review, we describe current efforts to explore the fundamental factors that control the ecological and microbial biodiversity in oxygen-starved regions of the ocean, termed oxygen minimum zones. We also discuss how recent advances in microbial ecology have provided information about the potential interactions in distributed co‑occurrence and metabolic networks in oxygen minimum zones, and we provide new insights into coupled biogeochemical processes in the ocean.

Ventilated Over geological time the ocean has evolved from being ecosystems and pelagic ecosystems reduces, changing Pertaining to the ocean: an anaerobic incubator of early cellular existence into the species composition and food web structure in supplied with atmospheric 15 a solar-powered emitter of molecular oxygen (O2), a these regions . Organisms that are unable to escape gases through processes transformation that has been punctuated by catastrophic O -deficient conditions may experience direct mortal- including exchange between 2 the air and sea, exchange extinctions followed by the iterative re-emergence of bio- ity (that is, the fish in these regions die) or decreased fit- 1,2 16,17 between the surface mixed logical diversity . Today, the ocean is being transformed ness . Even organisms that can escape to more highly layer and immediate in response to human activities. Indeed, the fourth assess- oxygenated refuges are susceptible to increased preda- subsurface layer, and ment report of the Intergovernmental Panel on Climate tion and density-dependent reductions in population circulation in the interior Change observed that the ocean is becoming substan- size18. OMZ expansion also causes changes in the cycling of the ocean. 3 tially warmer and more acidic . As these changes inten- of trace gases such as methane (CH4), nitrous oxide

sify, marine ecosystems will experience disturbances in (N2O) and carbon dioxide (CO2), which are important

the structure and dynamics of food webs, with resulting for metabolism and can have an effect on climate. CH4 4 radiative feedback on the climate system . Oxygen-starved regions and N2O are powerful greenhouse gases with of the ocean, known as oxygen minimum zones (OMZs), forcing effects that are approximately 25 and 300 times 5 are important bellwethers for these changes . the effect of CO2, respectively. Although oceanic CH4

OMZs are an intrinsic feature of water columns emissions are minor (<2% of natural CH4 emissions), the 1 Department of Microbiology that arise when the respiratory O2 demand during the ocean accounts for at least one-third of all natural N2O and Immunology, University degradation of organic matter exceeds O2 availability emissions, a large fraction of which are derived from of British Columbia, ventilated 6–9 − in poorly regions of the ocean . Increases OMZs via microbial respiration of nitrate (NO3 ) and Life Sciences Institute, − 19 2552–2350 Health Sciences in ocean temperature drive decreases in O2 solubility nitrite (NO2 ) . Moreover, OMZs account for up to 50% Mall, Vancouver, British and reduced ventilation owing to thermal stratification of oceanic fixed-nitrogen loss, and their expansion has Columbia V6T 1Z3, Canada. of the water column8,10, resulting in OMZ expansion. the potential to affect primary production, with resulting 2Graduate Program in Consistent with this, between 1956 and 2006 the O con- feedback on carbon transport processes20–23. Bioinformatics, University of 2 British Columbia, Life Sciences centrations in the OMZ of the northeast subarctic Pacific Although OMZs are inhospitable to aerobically Institute, 2552–2350 Health (NESAP) declined by 22%, and the hypoxic boundary respiring organisms, these zones support thriving micro-

Sciences Mall, Vancouver, layer (defined as ~60 μmol O2 per kg water) expanded bial communities that mediate cycling of nutrients and British Columbia V6T 1Z3, upwards from a depth of 400 m to 300 m (REF. 11). Similar radiatively active trace gases (which affect the climate). Canada. declines have been observed in the eastern tropical Therefore, systems-level investigations of microbial Correspondence to S.J.H. 12 12 13,14 e‑mail: Atlantic , the equatorial and northeast Pacific, and communities in the OMZ-containing water column 8 [email protected] in the Southern Ocean during the past 50 years. have great potential to enhance our mechanistic under- doi:10.1038/nrmicro2778 As O2 concentrations decline, the amount of habitat standing of a pervasive ecological phenomenon that Published online 14 May 2012 available to aerobically respiring organisms in benthic is integral to ocean productivity and climate balance.

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 381

Here, we review recent observations that have emerged western Pacific (where Pacific deep waters are formed) from the intersection of taxonomic and functional gene as well as to the length of time that these waters have surveys, gene expression studies and measurements of been isolated from the atmosphere (a result of global

process rates to better formulate hypotheses regarding circulation patterns). However, O2 deficiency is often the metabolic interactions that drive OMZ ecology and more intense in the Arabian Sea and in coastal Atlantic biogeochemistry on a global scale. We focus on bacte- waters on the African shelf than in the Pacific Ocean rial and archaeal contributions to these networks, with owing to unusually high levels of carbon export and sub- the understanding that microbial eukaryotes and viruses surface respiration in these naturally eutrophic waters of have biologically essential but as-yet physiologically the Arabian Sea and Atlantic Ocean29. Compounding the

uncharacterized roles in modulating matter and energy effects of respiratory demands for O2, many upwelling transformations in OMZs. systems experience episodic plumes of hydrogen sul-

phide (H2S) that can be attributed to diffusive flux from OMZ formation and expansion underlying sediments30. Such sulphidic events are toxic

OMZs are typically found on the western boundaries to most O2-respiring organisms. In addition to coastal of continental margins, where wind-driven circulation and open-ocean OMZs, enclosed or semi-enclosed patterns push nutrient-rich waters upwards to the sur- basins — including the Baltic Sea31, Black Sea32, Cariaco face in a process known as coastal upwelling. This process Basin33 and Saanich Inlet34 — experience varying degrees

Thermal stratification effectively fertilizes surface waters and results in high of O2 deficiency and sulphide accumulation, making A temperature-layering effect levels of photosynthetic primary production. During them useful model ecosystems for exploring microbial that occurs in water owing to photosynthesis, phytoplankton fix CO2. Much of the community responses to OMZ expansion. differences in water density: inorganic carbon that is fixed through photosynthesis is Human activities exacerbate the natural O2 defi- warm water is less dense than respired in surface and intermediate layers of the water ciency in shallow coastal and estuarine environments, cool water and therefore tends to float on top of the cooler, column through microbial degradation processes. A where nutrient run-off from agricultural and wastewater 35 heavier water. fraction of the product of primary production sinks as sources results in eutrophication . Moreover, changes in dead organisms and particles that are exported to depth. wind-driven circulation patterns can induce upwelling Benthic ecosystems Throughout the ocean this process, called the biological of O -deficient waters from coastal OMZs onto conti- Ecosystems residing at the 2 lowest level of a body of water carbon pump, has a large influence on the biogeochemi- nental shelves, increasing mortality of shelf-dwelling 36 such as an ocean or a lake, cal carbon cycle because carbon is sequestered in the organisms . Over the past two decades, shelf intrusions including the sediment surface interior of the ocean for long periods of time, during have produced ‘dead-zones’ off coastal Oregon37 and in and subsurface layers. which it cannot influence the climate24. Estimates of the Gulf of Mexico38 (USA), and off the coast of Chile28, carbon rain rates to carbon sediments in the northeast Africa39 and India40, contributing to a drop in produc- Pelagic ecosystems 35 Ecosystems residing in the Pacific suggest that the presence of an OMZ greatly tion from commercial fisheries . Regardless of the water region of a body of water that increases the amount of carbon exported to the deep body (estuary, basin, coastal waters or open ocean), O2 is neither close to the bottom ocean25. deficiency shifts energy away from a pelagic macrofauna nor near the shore. Persistent O2 deficiency occurs when the amount of towards microorganisms, decoupling predator–prey Radiative forcing effects dissolved O2 in the water column is consumed faster than interactions and changing the trophic exchanges that The change in net irradiance it is resupplied through air–sea exchange, photosyn- occur through existing food webs (FIG. 2). 15 between different layers of the thetic O2 production and ventilation . Global circula- atmosphere. Microbial energetics in OMZs tion patterns transport younger, more oxygenated waters throughout the deep ocean, resulting in deep oxycline Under oxic conditions (>90 μmol O per kg water), Coastal upwelling 2 The upwards movement of formation. Thus, in profile, OMZs resemble a band of 25–75% of the energy generated via oceanic primary 35 deep, nutrient-rich water along O2-deficient water inserted between two O2-containing production is transferred to mobile predators . As O2 a coast, caused by wind-driven (FIG. 1) water masses . The upper O2 thresholds chosen to levels decline, aerobic organisms escape to more oxy- currents. define OMZs have been manifold, ranging from <2 μmol genated refuges, resulting in habitat compression and a 6 Oxycline O2 per kg water to 90 μmol O2 per kg water . For the pur- concomitant diversion of energy into microbial metab- 35 (FIG. 2) A sharp gradient in oxygen poses of this Review, we adopt the criterion of <20 μmol olism in O2-deficient waters . Typically, energy concentration that is O2 per kg water, to include the maximum O2 level at flows according to a well-defined sequence of reduc- associated with a redoxcline (a which the use of alternative electron acceptors (in this tion–oxidation (redox) reactions, the order of which shift in electron donor and case, NO −) have been reported26. Using this definition, is determined by the amount of free energy available acceptor usage). 3 OMZs currently constitute 1–7% of the volume of the through each reaction. O2 is the most favourable electron 3 Eutrophic global ocean, occupying approximately 102 million km acceptor because it provides more energy (via its reduc- Pertaining to a body of water: (REFS. 6,7,27,28) (FIG. 1) . tion) than any other electron acceptor. CO2, the electron rich in mineral and organic Geographically, OMZs occur in the Pacific Ocean (in acceptor used by methanogenic archaea, yields the least nutrients. the NESAP, off western North America; the eastern trop- energy. Thus, the electron acceptors that are available in Eutrophication ical North Pacific (ETNP), off Mexico; and the eastern a given environment are reduced in a sequential order − Excessive nutrient input to a tropical South Pacific (ETSP), off Peru and Chile), the according to the free energy yield: O2, then NO3 and lake or other body of water − Atlantic Ocean (in the Northwest-African upwelling and NO2 , followed by manganese and iron, then sulphate (frequently owing to run-off the Namibian or Benguela upwelling) and the Arabian (SO 2−) and, finally, CO (REF. 41) (FIG. 2). This sequence from the land), resulting in 4 2 (FIG. 1) explosive plant growth and Sea . The Pacific OMZs are more voluminous than helps define specific metabolic niches and biogeochem- animal mortality owing to those in the Atlantic Ocean and the Arabian Sea. This ical potentials spanning oxic, dysoxic (20–90 μmol O2 oxygen starvation. is due to decreased ventilation at high latitudes in the per kg water), suboxic (1–20 μmol O2 per kg water) and

382 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/micro

a NESAP SI CB Baltic Black

0 Latitude (°N) –20 ≥200

–40 180

–80 160 140 –150 –100 –50 0 50 100 150 120 HOT ETSP NAM Arabian Longitude (°E) 100 mol per kg water)

b 0 μ

80 ( 2

Upper oxycline O 500 60

1,000 OMZ core 40

1,500 20 Deep oxycline 0 2,000 Depth (m) 2,500 Sea ﬂoor 3,000

3,500

4,000

Figure 1 | O2 concentrations in the ocean. a | Minimum molecular oxygen (O2) concentrations for different regions of the ocean. Locations highlighted in this Review are indicated and comprise the Hawaii OceanNature Time-series Reviews (HOT), | Microbiology the northeast subarctic Pacific (NESAP), Saanich Inlet (SI), the eastern tropical South Pacific (ETSP), the Cariaco Basin (CB), the Namibian upwelling (NAM; also known as the Benguela upwelling), and the Baltic, Black and Arabian seas. Oxygen data

were derived from REF. 115. b | Cross-section of the NESAP oxygen minimum zone (OMZ) , showing the O2 concentration from surface waters to the sea floor. Upper oxycline: transition from surface waters to the OMZ core. OMZ core: defined

by O2 concentrations <20 μmol per kg water. Deep oxycline: transition from the bottom of the OMZ core to abyssal waters.

− − anoxic (<1 μmol O2 per kg water) water column condi- bacteria, which convert the NO2 intermediate to NO3 . tions, under which multiple electron acceptors can be Nitrification typically takes place under dysoxic or sub- 42,43 used simultaneously to maximize the free energy yield oxic conditions, resulting in N2O production . The at different ecological scales (BOX 1). oxidizing nature of the modern ocean has resulted in − Examples of biogeochemical processes in which NO3 being the most abundant form of nitrogen in the − Chemoautotrophic sequential reactions are carried out by different organ- ocean. However, biological fixation of N2 to NO3 must Capable of using chemical isms can be found in the microbial pathways that drive be balanced by N2 production in order to maintain energy to synthesize organic 7,21 the nitrogen cycle . Nitrogen gas (N2) is the most abun- atmospheric N2 at a constant level over geological time- molecules from inorganic 44 − substances. dant form of nitrogen on earth, but few microorganisms scales . Under suboxic or anoxic conditions, NO3 and − are able to use (fix) N2, converting it to the more ame- NO2 are used as terminal electron acceptors by denitri- dissimilatory − Dissimilatory nable form of ammonia (NH3) or its protonated species, fying bacteria in NO3 reduction (denitrifi Metabolic processes through + 45 + ammonium (NH4 ), both of which can be terminally cation) and by anammox bacteria in anaerobic NH4 which elements are oxidized or oxidized to NO − by nitrifying bacteria. Nitrification is a oxidation (anammox)46. Both of these processes regener- reduced and for which the 3 chemoautotrophic organism uses the energy process that is carried out in two steps, ate N2, but denitrification also produces N2O, thus con- released in the process the first by NH3-oxidizing bacteria or archaea, which tributing to a loss in fixed nitrogen and to the production (catabolism). − − − convert NH3 to NO2 , and the second by NO2 -oxidizing of greenhouse gases. Dissimilatory NO3 reduction to

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 383

100% 0% well as the candidate divisions TM6, WS3, ZB2, ZB3, TEA GN0, OP11 and OD1, are also present in OMZs (FIG. 3) Energy to microorganisms O H O (note that the taxonomy used in this Review is taken 2 2 52 – – from the Greengenes database ). The distribution of NO3 NO2 Mn IV Mn III these taxa varies throughout the water column, with

Fe III Fe II different subdivisions partitioning along the oxycline. 2– H S SO4 2 These patterns reflect unique and overlapping interac- CO CH 2 4 tions among individual microorganisms, and among populations and communities of microorganisms. In the oxic surface waters overlying OMZs, sequences Energy to mobile predators Energy affiliated with the SAR11 cluster and the order Rhodobacterales (class Alphaproteobacteria), with the 0% 100% order Methylophilales (class Betaproteobacteria), and 200 150 100 50 0 with the SAR86 cluster and the clone Arctic96B‑1 (class

O2 (μmol per kg water) Gammaproteobacteria) are prevalent, as are sequences * affiliated with the phylum Cyanobacteria, with the Oxic Dysoxic Suboxic *Anoxic marine OM1 clade (phylum Actinobacteria), and with the clone Arctic97A‑17 and the genus Polaribacter Figure 2 | O2 concentration affects ecosystem energy flow. Alternative states of the sea water (oxic, dysoxic, suboxic and anoxic) and correspondingNature Reviews molecular | Microbiology oxygen (phylum Bacteroidetes). In dysoxic and suboxic waters, prevalent sequences include those that are affiliated (O2) concentrations are defined. The red–orange area indicates the range of energy transferred from pelagic nutrients to higher-level predators under oxic conditions. with the SAR11 cluster (class Alphaproteobacteria),

With declining O2, higher-level predation is suspended and the proportion of energy with the agg47 cluster (also known as ESP OMZ transferred to microorganisms rapidly increases (the yellow–green–blue area). This sequence accumulation cluster II (EOSA-II)53) and energy is generated via microbial respiration using a defined order of terminal the clones Arctic96B‑1, ZD0417 and ZA3412c (class − electron acceptors (TEAs), with O2 as the preferred TEA, followed by nitrate (NO3 ), Gammaproteobacteria), and with the SAR324 cluster manganese iv (Mn iv), iron iii (Fe iii), sulphate (SO 2−) and, finally, carbon dioxide (CO ). 4 2 and the genus Nitrospina (class Deltaproteobacteria). Figure is modified,with permission, from REF. 35 © (2008) American Association for Sequences affiliated with Microthrixineae (class the Advancement of Science. Actinobacteria), anammox bacteria (genus ‘Candidatus Scalindua’, phylum Planctomycetes), the phylum Chloroflexi and various Verrucomicrobia are also pre- + NH4 (DNRA), a process that takes place under suboxic sent in dysoxic and suboxic regions. In suboxic and or anoxic conditions, has the potential to moderate the anoxic waters (that is, OMZs), the dominant sequences loss in fixed nitrogen and to regenerate redox couples are affiliated with the SUP05 cluster (Suiyo Seamount − + 47–49 (REF. 54) (in the form of NO2 and NH4 ) for anammox . Taken hydrothermal plume group 5 ; also known as together, these microbial nitrogen transformations EOSA-I53; class Gammaproteobacteria). In addition constitute a distributed metabolic network linking the to these SUP05 sequences, prevalent sequences in metabolic potentials of different taxonomic groups to anoxic or sulphidic waters include those affiliated with higher-order biogeochemical cycling of nitrogen in the the sulphate-reducing family Desulphobacteraceae environment. (class Deltaproteobacteria), the sulphur-oxidizing Recent studies also posit an essential role for sul- Arcobacteraceae (class Epsilonproteobacteria), the Endemism phur cycling in OMZs, coupling the production and clone VC21_Bac22 (phylum Bacteroidetes), the phylum The ecological state of being consumption of reduced sulphur compounds to dis- Gemmatimonadetes and the phylum Lentisphaerae. The unique (endemic) to a defined − similatory NO3 reduction and the fixation of inor- presence of candidate divisions increases with decreas- geographical location, which ganic carbon50,51. The integration of carbon, nitrogen ing O concentrations, with most sequences affiliated can be a particular habitat, 2 zone or environment. and sulphur cycles represents a recurring theme in the with OP11 and OD1 identified in anoxic waters. In O2-deficient water column, where electron donors and addition to bacterial SSU rRNA genes, sequences affili-

Operational taxonomic acceptors are actively recycled between lower and higher ated with NH3-oxidizing marine group I (MGI) archaea units oxidation states (BOX 1). in the phylum Thaumarchaeota have been identified in (OTUs). Groups of organisms several locations, where they are most prevalent in the that are used in phylogenetic 55–59 studies. An OTU is tentatively OMZ microbiota oxycline . assumed to be a valid taxon for Taxonomic survey data based on small-subunit ribo- purposes of phylogenetic somal RNA (SSU rRNA) gene sequences indicates Taxa of emerging interest in OMZs. Although many of analysis. that there are conserved patterns in microbial com- the taxa identified in O2-deficient waters are ubiquitous endemism Ecotype munity composition between open-ocean and coastal throughout the ocean, patterns of emerge at A group of organisms within a OMZs and enclosed or semi-enclosed basins experi- the level of operational taxonomic units (OTUs) obtained species that are selectively encing water column O2 deficiency. The most abun- by clustering SSU rRNA gene sequences together at adapted to a particular set of dant phyla in the OMZs are (in order of abundance) specific identity thresholds. These OTU distribution environmental conditions and Proteobacteria, Bacteroidetes, marine group A (a patterns reinforce a model of ecological type (ecotype) therefore exhibit behavioural, structural or physiological candidate phylum), Actinobacteria and Planctomycetes selection in which genetically cohesive populations 60 differences from other (FIG. 3). The phyla Firmicutes, Verrucomicrobia, manifest distinct ecological or biogeochemical roles . members of the species. Gemmatimonadetes, Lentisphaerae and Chloroflexi, as These distinct roles in turn form the basis of distributed

384 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/micro

Box 1 | Redox-driven niche partitioning

Reduction–oxidation (redox)-driven niche partitioning in the molecular oxygen (O2)-deficient water column selects for shared metabolic capabilities across different ecological scales. Consistent with this observation, the chemical gradients found in marine oxygen minimum zones (see the figure, label 1) also exist in interior oceanic waters in the form of sinking organic particles or ‘marine snow’ (REF. 105) (see the figure, label 2). Particle association provides a nucleation point for otherwise suboxic or anoxic processes in oxygenated waters owing to the formation of microscale oxyclines83,106,107 (see the figure, label 3). The sulphate-reducing potential of such particles has been demonstrated82, and a similar relationship has been identified for methane production and transport in the North Pacific Ocean108. The interplay between particle- associated and free-living bacteria creates distributed networks of metabolite exchange between community members with alternative or competing nutritional or energetic needs109 (see the figure, label 4). The recent identification of the SAR324 cluster as particle-associated bacteria with genomic potential for inorganic carbon assimilation, sulphur oxidation and methane oxidation reinforces the ecological and biogeochemcial importance of microzone formation 65 − throughout the water column . CH4, methane; CO2, carbon dioxide; H2S, hydrogen sulphide; NH3, ammonia; NO2 , nitrite; − 2− NO3 , nitrate; TEA, terminal electron acceptor; SO4 , sulphate.

0 ≥200 TEA O H O 180 2 2 NO – NO – 500 3 2 160 Mn IV Mn III 1 Fe III Fe II 2– 1,000 140 SO4 H2S

CO2 CH4 120 1,500 100 3 mol per kg water) μ

80 (

2,000 2 O Depth (m) 60 2,500 40

3,000 20 1 mm 0 3,500 2 4

– 2– NO2 oxidizer Sulphur oxidizer SO4 reducer

NH3 oxidizer Heterotrophs Methanogen

Dimethylsulphonio- Anammox CH4 oxidizer propionate degrader Metabolite Substrate exchange competition

Nature Reviews | Microbiology

interaction networks that integrate the phenotypes of SUP05 and Arctic96BD‑19 exhibit overlapping but not different taxonomic groups. Here, we focus on the four identical distribution patterns, consistent with redox- most abundant bacterial taxonomic groups identified in driven niche partitioning. SUP05 is most abundant in surveys of OMZs (the SUP05–Arctic96BD‑19 group, the the slightly to moderately sulphidic waters at the base of − SAR11 and SAR324 clusters, and the candidate phylum the sulphide–NO3 transition zone, and the organisms in marine group A) for OTU distribution analysis (FIG. 4). this clade derive energy from the oxidation of reduced − SSU rRNA gene sequences affiliated with chemoauto sulphur compounds using NO3 as a terminal electron trophic, sulphur-oxidizing gill symbionts of deep-sea acceptor50,62. Arctic96BD‑19 is most abundant in dys- clams and mussels were first identified in open-ocean oxic and suboxic waters, as its members derive energy

OMZs in the Arabian Sea, the ETSP and the Namibian from reduced sulphur compounds using O2 as a terminal upwelling53,61,62. Phylogenetic analysis indicates that electron acceptor64,65. Both SUP05 and Arctic96BD‑19 these symbionts are part of a larger group of free-living members have the potential to use the energy gained symbionts (also referred to as the gammaproteobacterial from the oxidation of reduced sulphur compounds to fix sulphur-oxidizing cluster (or GSO)) consisting of two inorganic carbon via Rubisco50,65. Under more sulphidic closely related, co‑occurring and currently uncultivated water column conditions, SUP05 members are replaced lineages, SUP05 (REF. 54) (encompassing the clam and by Epsilonproteobacteria that exhibit similar meta- mussel symbionts) and Arctic96BD‑19 (REF. 63) (FIG. 4). bolic capabilities66–68. The presence of SUP05 species in

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 385

Oxygen status Oxic Dysoxic Suboxic O (μmol per kg water) 2 Anoxic

10 50 100 200 400 0 m 0 m 0 m 0 m 0 m 0 m 0 m 0 m 0 m 0 m 0 m 5 m 5 m 0 m 0 m 0 m 0 m 0 0 2 2 0 0 0 m 0 m 0 m 0 m 0 m 0 0 2 0 0 2 1 0 m 0 m 0 m 0 m 0 m 0 m 0 m 0 2 0 0 m* 0 m 0 0 0 0 0 m 1 1 1 1 1 2 2 0 m 0 m 1 1 2 1 0 m 0 m 7 0 0 ,0 3 0 m 5 0 m 0 10 m 1 1 2 0 0 0 1 1 1 2 7 8 7 7 8 7 8 0 m 0 m 0 m ,0 ,0 ,0 ,0 0 0 6 6 6 6 1 7 7 5 2 4 1 6 6 6 0 m 1 4 6 2 6 6 6 6 ,0 0 0 0 0 0 0 0 0 ,0 ,3 0 0 0 0 06 0 0 0 1 1 2 1 2 1 5 5 P P P P 0 0 0 0 1 1 2 6 5 1 6 2 2 6 2 1 6 OT OT OT OT OT OT OT AM 130 m 4 4 1 2 4 4 2 1 1 2 1 4 2 TS TS TS TS SI F P SI J P P H H SI A E E NAM 119 m H H H H H SI N SI A SI A N SI F SI A SI A SI J E E P P P P P P P P P SI N SI N SI F SI J SI N SI A SI A SI F SI J NAM 90 m P SAR11 OM38 NAC1-6

Alpha- Rhodobacterales Other Alpha Methylophilales

Beta- Other Beta Alteromonas Other Alteromonadales SUP05 Arctic96BD-19 SAR86 ZA2333c agg47 ZD0417 HTCC2207 Proteobacteria Xanthomonadales Arctic96B-1 OM60 Gamma- AEGEAN_245 Legionellales Marinobacter Pseudomonadaceae ZA3412c Methylophaga HTCC2089 Chromatiales Other Gamma Nitrospina SAR324 Delta- Other Delta Arcobacteraceae Other Epsilon VC21 Bac22 Epsilon- Other Bacteroidales Arctic97A-17 Chl112 Bacteroidetes F4C20 Other Cytophaga Polaribacter Other Flavobacteriales Other Bacteroidetes Microthrixineae OM1 Actinobacteria Other Actinobacteridae Other Acidimicrobidae Cyanobacteria Prochlorales Other Cyanobacteria Firmicutes Mollicutes Other Firmicutes Planctomycetes Anammoxales Other Planctomycetes Verruco- Verruco-3 microbia Opitutae Other Verrucomicrobia Spirochaetes Gemmatimonadetes Lentisphaerae Acidobacteria Chloroﬂexi Marine group A VHS-B5-50 TM6 Candidate WS3 divisions ZB2 ZB3 % SSU rRNA GN0 gene clone OP11 library OD1 Other bacteria 1 5 10 25 50 Nature Reviews | Microbiology

386 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/micro

◀ Figure 3 | Bacterial diversity in the ocean. Dot plot of the diversity of bacterial taxa at representation for subgroup ZA3312c. Six additional sub- various sample points and depths in Saanich Inlet (SI), the northeastern subarctic Pacific groups seem to be endemic to the NESAP, Saanich Inlet (NESAP; labelled P4, P12 and P26), the Hawaii Ocean Time-series (HOT), the eastern and the ETSP (information not shown). These results tropical South Pacific (ETSP) and the Namibian upwelling (NAM; also known as the are consistent with previous observations indicating Benguela upwelling), based on small-subunit ribosomal RNA (SSU rRNA) gene sequence that there is strong habitat selection for different marine profiles. ‘*’ indicates a sample taken from P4 1,000 m in June 2008; all other NESAP 78 samples were taken in 2009. Samples are organized according to the similarity of their group A subgroups . Despite these organisms being community composition, as revealed by hierarchical clustering of the distribution of widespread in the ocean, their metabolic capabilities remain unknown. taxonomic groups across environmental samples. The molecular oxygen (O2) concentration is shown for each oceanic sample, and the classification of the Similarly to marine group A, the SAR324 clade (also environment as oxic, dysoxic, suboxic or anoxic is also indicated in the colour bar. known as marine group B) of the class Deltaproteobacteria Names for identifying bacterial groups were selected according to the taxonomic level is also prevalent in the dark ocean79–81. The most com- at which the most relevant information was available. Data used to generate the dot mon SAR324 OTUs observed in OMZs (FIG. 4) are closely plot were derived from sequences deposited in Genbank. related to marine group B–SAR324 clade II, and there is also a minority representation for marine group B–SAR324 clade I; two additional clades are endemic to non-sulphidic OMZs serves as a biomarker for chang- Saanich Inlet and the ETSP (information not shown). ing ecosystem dynamics, indicating an increased poten- SAR324 members have the potential to oxidize one-carbon tial for toxic sulphur blooms and periodic or persistent (C1) compounds and reduced sulphur compounds, anoxia57,62,64. using the resulting energy to fix inorganic carbon via SAR11 is the most abundant and ubiquitous clade Rubisco65. Consistent with a functional role for Rubisco, of Alphaproteobacteria in the ocean, often constituting microautoradiography linked with catalysed reporter 30% of surface bacterioplankton communities69. The deposition fluorescence in situ hybridization (MAR– dominant SAR11 OTUs observed in OMZs (FIG. 4) are CARD–FISH) demonstrated that SAR324 members fix closely related to Pelagibacter ubique, a cultivated mem- inorganic carbon and undergo particle association in ber of SAR11 in subgroup Ia, and there is also a minority oxygenated waters of the North Atlantic65. representation for subgroups Ib and II. SSU rRNA gene surveys support the existence of several SAR11 ecotypes The symbiotic ocean that exhibit geographical and depth-specific water col- Recent advances in microbial ecology that combine cul- umn distributions70. Ecotype selection is particularly tivation-independent molecular methods with process apparent in the NESAP OMZ, where more than 30 dif- rate measurements are beginning to reveal previously (FIG. 4) ferent SAR11 OTUs have been identified . Current unknown metabolic interactions in the O2-deficient studies of cultivated SAR11 strains and free-living popu- water column. Parallel advances in exploring the dark lations indicate a genomic repertoire that is streamlined ocean have identified similar metabolic interactions for rapid heterotrophic growth71. Comparative genomics at different ecological scales. In the following case analyses suggest that there are differences in the glyco- study we touch on these observations, with particular lytic potential of coastal and open-ocean SAR11 popu- emphasis on the integration of carbon, nitrogen and lations. Consistent with this observation, carbon use sulphur cycles. and gene expression assays have measured preferential glucose use by coastal isolates that are associated with Unravelling a cryptic sulphur cycle. The identification a gene cluster encoding a variant form of the Entner– of the SUP05–Arctic96BD‑19 clade of Gammaproteo Doudoroff pathway72. However, the specific metabolic bacteria suggested an important role for sulphur cycling capabilities that enable SAR11 members to thrive in in the ecology and biogeochemistry of OMZs53,57,62.

O2-deficient waters remain unknown. The role of surface Metabolic reconstruction of the SUP05 metagenome water SAR11 populations in mediating demethylation identified numerous genes encoding components of

Heterotrophic of dimethysulphoniopropionate (DMSP) to methylmer- the sulphide oxidation and nitrate reduction pathways. Dependent on obtaining captopropionate (MMPA) may indicate a role for this Principal components of both pathways were found carbon for growth and energy group in OMZ sulphur cycling. Although this phenotype to be clustered in a 52 kb ‘metabolic island’ that also from complex organic is shared between a number of different pelagic bacte- contains a gene encoding the large subunit of form II compounds. ria73,74, SAR11 members are unable to perform dissimi- Rubisco, consistent with coordinated regulation of car- 50 Entner–Doudoroff pathway latory sulphate reduction, making them dependent on bon and energy metabolism . More recently, single-cell An alternative series of exogenous sources of reduced sulphur for growth and techniques were used to assemble genomic scaffolds for reactions for the catabolism of further reinforcing a model of distributed metabolite Arctic96BD‑19 from North Atlantic and Pacific waters, glucose to pyruvate, using a exchange75. uncovering sulphur oxidation and CO fixation genes65. different set of enzymes from 2 those used in either glycolysis The candidate phylum marine group A was first iden- The discovery of potential sulphur oxidizers in non- or the pentose phosphate tified almost 20 years ago in northeast Pacific Ocean sulphidic waters is enigmatic, bringing into question the pathway. waters from 100 m and 500 m depth intervals76,77. Since source of the reducing equivalents that are needed to fix that time, SSU rRNA gene surveys have identified marine inorganic carbon. Dark ocean group A as being ubiquitous in the dark ocean. The domi- Sinking particles have been proposed to be sources The depths of the ocean (FIG. 4) beyond which less than 1% of nant marine group A OTUs observed in OMZs of reduced compounds such as sulphide in suboxic or 82,83 2− sunlight penetrates; also are closely related to subgroups SAR406, Arctic95A‑2, anoxic waters . However, SO4 reduction in the water known as the aphotic zone. Arctic96B‑7 and ZA3648c, with an additional minority column is difficult to measure because sulphide rapidly

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 387

1,000 SUP05–Arctic96BD-19 SAR11 Colour key for heat map 400 Colour key for heat map 800 300 600 0 5 100 0 1 10 200 400 SSU rRNA gene sequences SSU rRNA gene sequences

200 100 Number of SSU rRNA sequences Number of SSU rRNA sequences 0 0 SUP05_01 SUP05_02 SUP05_03 SUP05_04 SUP05_05 SUP05_06 SUP05_07 SUP05_08 SUP05_09 SUP05_10 SUP05_11 SUP05_12 SUP05_13 SUP05_14 SUP05_15 SUP05_16 SUP05_17 SAR11_01 SAR11_02 SAR11_03 SAR11_04 SAR11_05 SAR11_06 SAR11_07 SAR11_08 SAR11_09 SAR11_10 SAR11_11 SAR11_12 SAR11_13 SAR11_14 SAR11_15 SAR11_16 SAR11_17 SAR11_18 SAR11_19 SAR11_20 SAR11_21 SAR11_22 SAR11_23 SAR11_24 SAR11_25 SAR11_26 SAR11_27 SAR11_28 SAR11_29 SAR11_30 SAR11_31 SAR11_32 SAR11_33 SAR11_34 NAM 90 m SI A07 100 m P26 500 m SI A07 120 m P12 500 m SI A08 120 m P4 1,000 m* SI N06 10 m P4 500 m SI A08 100 m P12 2,000 m SI F06 100 m P26 2,000 m SI J06 100 m P4 1,000 m SI J06 120 m P4 1,300 m SI N06 200 m P26 1,000 m SI F06 125 m SI N06 120 m P12 1,000 m SI F06 215 m P26 10 m SI N06 100 m P12 10 m SI A07 200 m SI N06 10 m SI A08 200 m SI A08 100 m SI J06 200 m SI A07 120 m ETSP 450 m SI A07 100 m ETSP 60 m SI N06 100 m ETSP 200 m P4 10 m SI F06 10 m SI N06 200 m SI J06 10 m SI J06 100 m SI A07 10 m SI F06 100 m NAM 130 m SI A08 120 m P4 10 m SI N06 120 m NAM 119 m ETSP 200 m P4 1,000 m* ETSP 450 m P4 500 m ETSP 60 m P4 1,000 m NAM 130 m P12 500 m NAM 119 m HOT 500 m HOT 4,000 m HOT 4,000 m SI J06 10 m P12 10 m SI A07 10 m P26 10 m SI F06 10 m P12 1,000 m SI F06 215 m P4 1,300 m SI J06 120 m P26 500 m SI A08 200 m P12 2,000 m SI F06 125 m P26 1,000 m SI A07 200 m P26 2,000 m SAR324 100 Marine group A 250 Colour key for heat map Colour key for heat map 80 200

60 150 0 1 10 0 1 10 SSU rRNA gene sequences SSU rRNA gene sequences 40 100

50 20 Number of SSU rRNA sequences Number of SSU rRNA sequences

0 0 MGA_01 MGA_02 MGA_03 MGA_04 MGA_05 MGA_06 MGA_07 MGA_08 MGA_09 MGA_10 MGA_11 MGA_12 MGA_13 MGA_14 MGA_15 MGA_16 MGA_17 MGA_18 MGA_19 MGA_20 MGA_21 MGA_22 MGA_23 MGA_24 MGA_25 MGA_26 MGA_27 MGA_28 MGA_29 MGA_30 MGA_31 SAR324_01 SAR324_02 SAR324_03 SAR324_04 SAR324_05 SAR324_06 SAR324_07 SAR324_08 SAR324_09 SAR324_10 SAR324_11 SAR324_12 SAR324_13 SAR324_14 SAR324_15 SAR324_16 SAR324_17 P4 1,000 m SI N06 200 m P26 1,000 m SI N06 100 m P4 500 m SI N06 120 m P12 500 m SI A07 200 m P26 500 m SI A08 200 m P4 1,300 m SI F06 215 m SI J06 200 m P12 1,000 m P4 1,300 m P12 2,000 m P26 1,000 m P26 2,000 m P26 2,000 m P4 1,000 m* P4 500 m SI F06 125 m P26 500 m SI A08 200 m P12 1,000 m SI F06 215 m P12 2,000 m SI J06 100 m P4 1,000 m* SI A08 100 m SI A07 120 m SI F06 100 m SI A08 100 m SI A07 120 m SI J06 100 m SI N06 200 m SI F06 100 m SI A07 100 m SI A08 120 m SI A08 120 m SI F06 125 m SI J06 120 m SI A07 100 m SI A07 200 m SI N06 10 m SI J06 200 m SI J06 120 m SI N06 120 m P12 500 m SI N06 100 m ETSP 200 m HOT 130 m ETSP 60 m P4 1,000 m ETSP 60 m HOT 770 m ETSP 200 m HOT 500 m NAM 130 m HOT 200 m HOT 500 m HOT 130 m P26 10 m SI A07 10 m SI N06 10 m SI J06 10 m SI A07 10 m ETSP 450 m NAM 119 m P12 10 m HOT 200 m NAM 90 m HOT 770 m NAM 119 m HOT 4,000 m P26 10 m HOT 4,000 m O2 (μmol per kg water) 1 10 25 50 75 >100

388 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/microNature Reviews | Microbiology

◀ Figure 4 | Diversity in the four most abundant bacterial groups identified in OMZs. In addition to providing reducing equivalents for The four most abundant groups that were identified in small-subunit ribosomal RNA − NO3 reduction, sulphur metabolism may contribute (SSU rRNA) gene surveys of oxygen minimum zones (OMZs) are the SUP05– to NH + production through the process of DNRA Arctic96BD‑19 group, the SAR11 and SAR324 clusters, and marine group A (MGA). Each 4 (also termed NO − or NO − ammonification). Both histogram bar represents a cluster of SSU rRNA gene sequences, or an operational 3 2 SO 2−-reducing and sulphur-oxidizing bacteria have taxonomic unit (OTU), generated at a 97% identity cutoff (clustered using the 4 furthest-neighbour algorithm). The height of the bar is equivalent to the sum of all been shown to carry out DNRA, in some cases using − − + sequences belonging to a specific OTU across all environments surveyed: the eastern the conversion of NO3 or NO2 to NH4 as an electron tropical South Pacific (ETSP), the Hawaii Ocean Time-series (HOT), the northeastern sink for substrate-level phosphorylation48, and in oth- subarctic Pacific (NESAP; labelled as P4, P12 and P26), Saanich Inlet (SI) and the Namibian ers coupling the conversion process to generate a proton upwelling (NAM; also known as the Benguela upwelling). ‘*’ indicates a sample taken motive force49. Although most of what we know about from P4 1,000 m in June 2008; all other NESAP samples were taken in 2009. Heat maps DNRA comes from sediment incubation or laboratory below the histograms represent the distribution of sequences in each OTU across all experiments with pure cultures49,87, several studies using environments surveyed. Heat maps were clustered by row using Euclidean distance and nitrogen‑15 incubations have measured potential DNRA the furthest-neighbour algorithm to highlight patterns of diversity among samples. Inset rates in the water columns of the Baltic Sea88, the ETSP22 colour scales depict the colour code for the number of SSU rRNA gene sequences in 89 heat maps. Data were derived from sequences deposited in Genbank. and the Namibian upwelling . In the ETSP OMZ, DNRA is estimated to provide a substantial portion of + 22 the NH4 that is required for anammox , with up to 22% 2− 51 derived from SO4 reduction . The potential contribu- auto-oxidizes in the presence of even trace amounts of tions of sulphur-oxidizing bacteria, including members

O2. Typically, when sulphide is detected in OMZs, it of the SUP05–Arctic96BD‑19 and SAR324 clades, to − − originates in rare pockets of NO3 - and NO2 -depleted DNRA in the O2-deficient water column remain to be water84 or is released by diffusive flux from sediments30,85. determined. Thus, the in situ component of the sulphur cycle in non- The existence of a cryptic sulphur cycle coupling the sulphidic waters has been described as cryptic because it metabolic activities of SUP05–Arctic96BD‑19 bacte-

lacks obvious chemical expression in the water column. ria and sulphate-reducing bacteria in the O2-deficient To resolve this enigma, researchers conducted process water column or in association with sinking particles is 2− rate measurements of SO4 reduction in the ETSP OMZ reminiscent of the symbiotic associations found at oxic– 2− 35 2− 90 using radiolabelled SO4 ( SO4 ) after a pulse of unla- anoxic interfaces . Indeed, chemoautotrophic symbioses belled sulphide to capture the formation of radiolabelled are a common innovation at hydrothermal vent and cold 51 2− sulphide . High rates of SO4 reduction were detected seep habitats, where eukaryotic hosts provide optimal (between 0.28 and 1.0 nmol per m2 per day) coupled to access to the redox couples that are needed to fix inor- − 91 the production of NO2 , N2O and N2. Consistent with ganic carbon on or near the sea floor . Similarly, symbi- 2− process rate measurements, metagenome sequenc- otic sulphur-oxidizing and SO4 -reducing bacteria have ing recovered a limited number of genes originating been described in association with the shallow-water from canonical sulphate-reducing bacteria, including sand-dwelling mouthless worm Olavius algarvensis; for Desulphatibacillum, Desulphobacterium, Desulphococcus, these bacteria, metabolite exchange between different Syntrophobacter, and Desulphovibrio spp. taxonomic groups balances out the fitness costs asso- − 92 Sulphur oxidation coupled to NO3 reduction in the ciated with resource competition in the host milieu . ETSP OMZ was supported by metatranscriptomic anal- Other forms of syntrophy, including direct electron trans- yses, which revealed that transcripts for dissimilatory fer, have been described between bacterial and archaeal 2− sulphite reductase (dsr) genes, sulphur oxidation (sox) cells such as acetogens and methanogens or SO4 reduc- genes (encoding proteins that mediate thiosulphate oxi- ers and methane oxidizers, resulting in the production dation), adenosine-5ʹ‑phosphosulphate (APS) reductase of reduced compounds that fuel subsurface metabolism (apr) genes (encoding proteins that mediate the con- and deep-sea chemolithoautotrophic communities93,94. 2− version of sulphur to SO4 ) and the gene encoding the Thus, the ecology and biogeochemistry of OMZs repre- − catalytic subunit of respiratory NO3 reductase (narG) sents one manifestation of the greater symbiotic ocean were highly expressed86. Although most of these tran- and its impact on the world around us. This impact is scripts originated from sulphur-oxidizing organisms, rooted in the collective metabolic capabilities of micro- including members of the SUP05 clade and close rela- bial cells that drive matter and energy transformations tives, a minority of aprA and dsrB transcripts affiliated throughout the depth continuum. 2− with canonical SO4 -reducing bacteria were detected, Syntrophy consistent with there being an active sulphur cycle in Co‑occurrence networks Metabolite exchange that this environment. Interestingly, in ETSP metagenomes, Just as cellular complexity arises through networks of occurs between two or more groups of organisms and is 32% of the top hits to aprA were affiliated with SAR11, genes, proteins and metabolites interacting across multi- 95–97 necessary for cell growth or and cognate aprA transcripts were highly expressed ple hierarchical levels , so ecological and biogeochemi- energy production. throughout the oxycline. Although the precise role of cal phenotypes arise from complex interactions between SAR11 in sulphur cycling in OMZs remains unknown, microbial community members. As Chisholm and Cary Chemolithoautotrophic the expression of Apr could indicate a link between state, “No single organism contains all the genes neces- Capable of obtaining energy from the oxidation of inorganic DMSP demethylation and the production of reducing sary to perform the diverse biogeochemical reactions that compounds and carbon from equivalents for sulphur oxidation in the surrounding make up ecological community function. Yet, distributed the fixation of carbon dioxide. water column. among the community, are all the functions necessary to

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 389

Box 2 | Network analysis The properties of many complex systems, including the cell, the brain and the internet, a Degree are the result of numerous pairwise interactions of individual components. Such a system can be represented by a set of nodes (components or subunits) connected by links (interactions between nodes) to form a network (see the figure). In the co‑occurrence network shown in FIG. 5 of this Review, nodes represent operational taxonomic units (OTUs) and links represent Pearson correlation coefficients greater than 0.4. In order for a network model to bolster understanding of a complex system such as a microbial A community, it is necessary to quantify the topological features or properties of the network. Important properties include: • Degree: the number of neighbours of a node, also known as connectivity. Nodes with high B degrees (many links to other nodes) are referred to as hubs. In the example network (see the figure, part a), node A has a degree of 12 and node B has a degree of 8. The degree distribution, P(k), of a network is the probability that a given node has exactly k links110. kA = 12 • Betweenness: the frequency at which a node is present on the shortest path between k = 8 all other nodes; in other words, a measure of how central a node is within a network. B Nodes with high betweenness have been shown to control the flow of information across a network111. In the example network (see the figure, part b), the dark-blue nodes have high betweenness relative to the light-blue nodes. b Betweenness Quantification of the properties of a network is the basis for distinguishing the network type, from which we can infer certain biological properties97. Many biological networks reported in the literature (including metabolic and protein networks) are scale-free networks95,112, meaning that they exhibit power-law degree distributions (that is, P(k) ~ k−γ, in which γ is the degree exponent, an experimentally observed quantity that typically ranges between 2 and 3)97. Scale-free distribution implies that a network consists of a small number of hubs in addition to numerous nodes with fewer links113. In scale-free networks (including the co‑occurrence network shown in FIG. 5), the hubs (or keystone nodes) display a high betweenness, suggesting that these nodes have important roles in regulating network interactions. Although co‑occurrence networks do not directly implicate specific modes of ≥0.05 metabolite exchange, they provide an excellent framework for generating hypotheses <0.05 regarding potential metabolic interactions that can be further tested using environmental parameter, in situ process rate and functional gene data. The box is adapted from REF. 114 © (2009) Macmillan Publishers Ltd. All rights reserved. Nature Reviews | Microbiology

define that community’s interaction with its environment” correlating with other bacterial OTUs and others with (REF. 98). These interactions form the basis of distributed environmental parameter data), consistent with ecotype networks in which nodes are taxa and links are the cor- selection and succession. The LSA approach has been relations between taxa. Microbial networks continuously extended to include three-domain interactions (between evolve by the arrival and departure of new nodes and links Archaea, Bacteria and protists (from Eukarya)) occur- through mutation, gene transfer or habitat selection, cre- ring at the SPOTS and in the English Channel, revealing ating functionally redundant modules that are separated in space and time. The application of network theory to discover and define co‑occurrence patterns among Figure 5 | Co‑occurrence networks: correlations ▶ so‑called ‘free-living’ microorganisms represents a new among bacterial OTUs in different OMZs. a | The network of interactions between the operational taxonomic frontier in microbial ecology99 (BOX 2). units (OTUs) identified in FIG. 4 and found in various oceanic oxygen minimum zones (OMZs) described in FIGS 3,4. Patterns of co‑occurrence. Recently, local similarity anal- Dominant bacterial OTUs are shown as per the key. Nodes ysis (LSA)100 was used to calculate co‑occurrence patterns are sized according to the weighted average O2 between bacterial OTUs and environmental parameter concentration across all samples where that OTU is found. data recovered from the chlorophyll maximum at the Each node represents a different OTU, although multiple San Pedro Ocean Time Series (SPOTS)101. The resulting OTUs can belong to the same taxa. The left side of the networks revealed both positive and negative correlations network consists of oxic subnetworks; dysoxic and between specific OTUs and water column conditions, suboxic subnetworks are present in the centre; and and these correlations were either direct or time-lagged in anoxic subnetworks are in the upper right corner. b | The betweenness data for this network. Nodes exhibiting a nature101. From an interpretive perspective, positive cor- betweenness centrality of ≥0.05 (that is, those that are relations could represent cooperative activities, includ- statistically likely to be central to the network) are ing distributed metabolism, cross-feeding or overlapping highlighted. Node sizes are based on the total number of habitat preference, whereas negative correlations could small-subunit ribosomal RNA (SSU rRNA) gene sequences represent resource competition, predation or alterna- belonging to that OTU summed across all OMZ samples. tive habitat preference101. For example, ten SAR11 OTUs Data for this figure were derived from sequences participated in different subnetworks over time (some deposited in Genbank.

390 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/micro

O2 (μmol per kg water) Major taxa SUP05 Nitrospina SAR11 Planctomycetes 5 25 50 100 250 SAR324 Microthrixineae Marine group A Other bacteria b

S23_91 SAR11

SAR11 Microthrixineae

Rhodobacter Arctic96B-7

Cytophaga

SAR324

Nitrospina SAR11

ZA3420c

SUP05 Nitrospina SAR324

Microthrixineae Cytophaga SSU rRNA gene sequences ZA3420c Betweenness centrality 10 100 500 1,200 2,400 ≥0.05 <0.05

Nature Reviews | Microbiology

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 391

progressive changes in microbial co-occurrence patterns which are both resilient and responsive to environmental over time and uncovering potential symbiotic associa- perturbation. Over geological timescales, recurring tions102,103. On a global scale, co‑occurrence analysis of changes in the oxygenation status of the ocean have 298,591 publicly available SSU rRNA gene sequences was resulted in multiple biotic crises with concomitant used to define nonrandom and recurring co-occurrence changes in marine ecosystems and climate balance. networks, consistent with habitat preference between Available monitoring data suggest that OMZ expansion specific taxonomic groups104. in the modern ocean is consistent with a renewed period Although networks based on phylogenetic infor- of change. When viewed from an Earth systems perspec- mation alone cannot explain underlying mechanisms tive, these observations take on immediate significance of metabolite exchange, they can help define putative as we consider the potential impacts of OMZ expansion metabolic interactions and enable more direct hypoth- on marine resources and global warming trends. These esis testing when combined with data about envi- impacts include reduced biodiversity and food secur ronmental parameters, process rates and functional ity and increased production and transport of radia- genes. We applied co‑occurrence network analysis to tively active trace gases owing to changes in microbial publicly available SSU rRNA gene sequences from the interaction networks. Hawaii Ocean Time-series, the NESAP, Saanich Inlet Determining how these interaction networks form, and the ETSP, and identified nonrandom patterns of function and change over time reveals otherwise hid- co‑occurrence between microbial taxa associated with den links between microbial community structure and oxic, dysoxic, suboxic or anoxic water column condi- higher-order ecological and biogeochemical processes. tions (FIG. 5). In this analysis, dysoxic, suboxic and anoxic Indeed, over the past few years, plurality sequencing subnetworks are dominated by OTUs representing the combined with process rate analyses and targeted gene SUP05–Arctic96BD‑19, SAR11, marine group A and surveys in coastal and open-ocean OMZs has identi- SAR324 clades, consistent with members of these clades fied conserved patterns of microbial community struc- having overlapping habitat preferences and the potential ture and function, and has uncovered novel modes of for metabolite exchange. Moreover, OTUs representing metabolic integration that couple carbon, nitrogen and these taxa were identified as hubs in the larger net- sulphur cycles. These findings have important implica- work on the basis of the number of connections run- tions for our understanding of the nutrient and energy ning through them, consistent with previous reports flow patterns in expanding marine OMZs. Looking of ‘keystone’ connectivity in marine ecosystems102,103. forwards, comparative studies are needed to define the Interestingly, the most abundant bacterial OTUs in the shared or specialized metabolic subsystems that mediate

network are not typically the most connected, and dif- microbial community responses to changing levels of O2 ferent OTUs for several taxonomic groups participated deficiency in the water column in different oceanic prov- in multiple subnetworks, suggesting that the overall net- inces. Additional time series monitoring studies combin- work consists of many functionally redundant modules ing gene expression and process rate measurements are with the potential to change over time. also needed to validate pathway predictions and provide parameters for regulatory and network dynamics Where do we go from here? for more effective ecosystem modelling. An effective We live on an ocean-dominated planet, and the collec- human adaptation and response to OMZ expansion, tive metabolic expression of cellular life in the ocean ranging from our environmental management to our has a profound influence on the evolution of the bio- policy towards Earth systems engineering, may depend sphere. Cellular life in the ocean is in turn dominated by on our collective capacity to understand and mimic the microbial communities that form interaction networks problem-solving power of the symbiotic ocean.

1. Kasting, J. F. & Siefert, J. L. Life and the evolution of 8. Helm, K. P., Bindoff, N. L. & Church, J. A. Observed 13. Bograd, S. J. et al. Oxygen declines and the shoaling Earth’s atmosphere. Science 296, 1066–1068 (2002). decreases in oxygen content of the global ocean. of the hypoxic boundary in the California Current. This concept paper outlines the interconnectedness Geophys. Res. Lett. 38, L23602 (2011). Geophys. Res. Lett. 35, L12607 (2008). between atmospheric transformation and 9. Karstensen, J., Stramma, L. & Visbeck, M. Oxygen 14. Emerson, S., Watanabe, Y. W., Ono, T. & Mecking, S. microbial metabolism now and throughout minimum zones in the eastern tropical Atlantic and Temporal trends in apparent oxygen utilization in the evolutionary time. Pacific oceans. Prog. Oceanogr. 77, 331–350 (2008). upper pycnocline of the North Pacific: 1980–2000. 2. Falkowski, P. G., Fenchel, T. & Delong, E. F. The 10. Keeling, R. F., Kortzinger, A. & Gruber, N. Ocean J. Oceanogr. 60, 139–147 (2004). microbial engines that drive Earth’s biogeochemical deoxygenation in a warming world. Annu. Rev. Marine 15. Rabalais, N. et al. Dynamics and distribution of cycles. Science 320, 1034–1039 (2008). Sci. 2, 199–229 (2010). natural and human-caused hypoxia. Biogeosciences 7, 3. Kumar, S. Fourth assessment report of the This paper reviews the evidence for global oxygen 585–619 (2010). Intergovernmental Panel on Climate Change: loss in the oceans over the past 50 years and 16. Breitburg, D. L., Hondorp, D. W., Davias, L.A. & Diaz, important observations and conclusions. Curr. Sci. 92, discusses potential consequences of continued R. J. Hypoxia, nitrogen, and fisheries: integrating 1034–1034 (2007). deoxygenation on the ecology and biogeochemistry effects across local and global landscapes. Annu. Rev. 4. Doney, S. C. The growing human footprint on coastal of marine ecosystems. Mar. Sci. 1, 329–349 (2009). and open-ocean biogeochemistry. Science 328, 11. Whitney, F. A., Freeland, H. J. & Robert, M. 17. Vaquer-Sunyer, R. & Duarte, C. M. Thresholds of 1512–1516 (2010). Persistently declining oxygen levels in the interior hypoxia for marine biodiversity. Proc. Natl Acad. Sci. 5. Falkowski, P. G. et al. Ocean deoxygenation: past, waters of the eastern subarctic Pacific. Prog. USA 105, 15452–15457 (2008). present and future. Eos Trans. AGU 92, 409–410 Oceanogr. 75, 179–199 (2007). 18. Ekau, W., Auel, H., Portner, H. O. & Gilbert, D. (2011). This is the first report to describe a direct Impacts of hypoxia on the structure and processes in 6. Paulmier, A. & Ruiz-Pino, D. Oxygen minimum zones relationship between thermal stratification and pelagic communities (zooplankton, macro- (OMZs) in the modern ocean. Prog. Oceanogr. 80, oxygen loss in the interior waters of the NESAP invertebrates and fish). Biogeosciences 7, 113–128 (2008). over a 50 year time interval. 1669–1699 (2010). 7. Lam, P. & Kuypers, M. M. Microbial nitrogen cycling 12. Stramma, L., Johnson, G. C., Sprintall, J. & Mohrholz, V. 19. Naqvi, S. W. A. et al. Marine hypoxia/anoxia as a

processes in oxygen minimum zones. Ann. Rev. Mar. Expanding oxygen-minimum zones in the tropical source of CH4 and N2O. Biogeosciences 7, 2159–2190 Sci. 3, 317–345 (2010). oceans. Science 320, 655–658 (2008). (2010).

392 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/micro

20. Codispoti, L. A. et al. The oceanic fixed nitrogen and 44. Deutsch, C., Sarmiento, J. L., Sigman, D. M., Gruber, 67. Grote, J., Jost, G., Labrenz, M. Herndl, G. J. & nitrous oxide budgets: moving targets as we enter N. & Dunne, J. P. Spatial coupling of nitrogen inputs Jürgens, K. Epsilonproteobacteria Represent the the anthropocene? Scientia Marina 65, 85–105 and losses in the ocean. Nature 445, 163–167 (2007). Major Portion of Chemoautotrophic Bacteria in Sulfidic (2001). 45. Zumft, W. G. Cell biology and molecular basis of Waters of Pelagic Redoxclines of the Baltic and Black 21. Gruber, N. in Nitrogen in the Marine Environment denitrification. Microbiol. Mol. Biol. Rev. 61, 533–616 Seas. Appl. Environ. Microbiol. 77, 7456–7551 (2008). (eds Capone, D. G., Bronk, D. A., Mulholland, M. R. & (1997). 68. Grote, J. et al. Genome and physiology of a model for Carpenter, E. J.) 1–50 (Elsevier, 2008). 46. Mulder, A., Vandegraaf, A. A., Robertson, L. A. & responsible Epsilonproteobacterium sulfide 22. Lam, P. et al. Revising the nitrogen cycle in the Kuenen, J. G. Anaerobic ammonium oxidation detoxification in marine oxygen depletion zones. Proc. Peruvian oxygen minimum zone. Proc. Natl Acad. Sci. discovered in a denitrifying fluidized-bed reactor. Natl Acad. Sci. USA 109, 506–510 (2012). USA 106, 4752–4757 (2009). FEMS Microbiol. Ecol. 16, 177–183 (1995). 69. Morris, R. M. et al. SAR11 clade dominates ocean 23. Ward, B. B. et al. Denitrification as the dominant 47. Lees, H. & Simpson, J. R. The biochemistry of surface bacterioplankton communities. Nature 420, nitrogen loss process in the Arabian Sea. Nature 461, nitrifying organisms. V. Nitrite oxidation by 806–810 (2002). 78–81 (2009). Nitrobacter. Biochem. J. 65, 297–305 (1957). 70. Field, K. G. et al. Diversity and depth-specific References 22 and 23 highlight the major 48. Cole, J. A. & Brown, C. M. Nitrate reduction to distribution of SAR11 cluster rRNA genes from marine metabolic processes and microbial players that ammonia by fermentative bacteria: a short circuit of planktonic bacteria. Appl. Environ. Microbiol. 63, mediate loss of fixed nitogen in disparate OMZs. the bacterial nitrogen cycle. FEMS Microbiol. Lett. 7, 63–70 (1997). 24. Siegenthaler, U. & Sarmiento, J. L. Atmospheric 65–72 (1980). 71. Giovannoni, S. J. et al. Genome streamlining in a carbon dioxide and the ocean. Nature 365, 119–125 49. Simon, J. Enzymology and bioenergetics of respiratory cosmopolitan oceanic bacterium. Science 309, (1993). nitrite ammonification. FEMS Microbiol. Rev. 26, 1242–1245 (2005). 25. Devol, A. H. & Hartnett, H. E. Role of the oxygen- 285–309 (2002). 72. Schwalbach, M. S., Tripp, H. J., Steindler, L., Smith, deficient zone in transfer of organic carbon to the deep 50. Walsh, D. A. et al. Metagenome of a versatile D. P. & Giovannoni, S. J. The presence of the glycolysis ocean. Limnol. Oceanogr. 46, 1684–1690 (2001). chemolithoautotroph from expanding oceanic dead operon in SAR11 genomes is positively correlated with 26. Smethie, W. M. Nutrient regeneration and zones. Science 326, 578–582 (2009). ocean productivity. Environ. Microbiol. 12, 490–500 denitrification in low oxygen fjords. Deep Sea Res. 51. Canfield, D. E. et al. A cryptic sulfur cycle in oxygen- (2010). Part 1 Oceanogr. Res. Pap. 34, 983–1006 (1987). minimum-zone waters off the Chilean coast. Science 73. Howard, E. C. et al. Bacterial taxa that limit sulfur flux 27. Ulloa, O. & Pantoja, S. The oxygen minimum zone of 330, 1375–1378 (2010). from the ocean. Science 314, 649–652 (2006). 2− the eastern South Pacific. Deep Sea Res. Part 2 Top. This paper provides the first report of SO4 74. González, J. M. et al. Silicibacter pomeroyi sp. nov. & Stud. Oceanogr. 56, 987–991 (2009). reduction and concomitant sulphide oxidation in Roseovarius nubinhibens sp. nov., 28. Fuenzalida, R., Schneider, W., Garces-Vargas, J., the ETSP OMZ. dimethylsulfoniopropionate-demethylating bacteria Bravo, L. & Lange, C. Vertical and horizontal extension 52. DeSantis, T. Z. et al. Greengenes, a chimera-checked from marine environments. Int. J. Syst. Evol. of the oxygen minimum zone in the eastern South 16S rRNA gene database and workbench compatible Microbiol. 53, 1261–1269 (2003). Pacific Ocean. Deep Sea Res. Part 2 Top. Stud. with ARB. Appl. Environ. Microbiol. 72, 5069–5072 75. Tripp, H. J. et al. SAR11 marine bacteria require Oceanogr. 56, 1027–1038 (2009). (2006). exogenous reduced sulphur for growth. Nature 452, 29. Estrada, M. & M, C. Phytoplankton biomass and 53. Stevens, H. & Ulloa, O. Bacterial diversity in the 741–744 (2008). productivity off the Namibian coast. S. Afr. J. Mar. Sci. oxygen minimum zone of the eastern tropical South 76. Fuhrman, J. A., McCallum, K. & Davis, A. A. 5, 347–356 (1987). Pacific. Environ. Microbiol. 10, 1244–1259 (2008). Phylogenetic diversity of subsurface marine microbial 30. Canfield, D. E. Models of oxic respiration, denitrification 54. Sunamura, M., Higashi, Y., Miyako, C., Ishibashi, J. & communities from the Atlantic and Pacific Oceans. and sulfate reduction in zones of coastal upwelling. Maruyama, A. Two bacteria phylotypes are Appl. Environ. Microbiol. 59, 1294–1302 (1993). Geochim. Cosmochim. Acta 70, 5753–5765 (2006). predominant in the Suiyo seamount hydrothermal 77. Gordon, D. & Giovannoni, S. Detection of stratified 31. Conley, D. J., Humborg, C., Rahm, L., Savchuk, O. P. & plume. Appl. Environ. Microbiol. 70, 1190–1198 microbial populations related to Chlorobium and Wulff, F. Hypoxia in the Baltic Sea and basin-scale (2004). Fibrobacter species in the Atlantic and Pacific oceans. changes in phosphorus biogeochemistry. Environ. Sci. 55. Sinninghe Damste, J. S. et al. Distribution of membrane Appl. Environ. Microbiol. 62, 1171–1177 (1996). Technol. 36, 5315–5320 (2002). lipids of planktonic Crenarchaeota in the Arabian Sea. 78. Rappé, M. S. & Giovannoni, S. J. The uncultured 32. Jorgensen, B. B. Ecology of the bacteria of the sulphur Appl. Environ. Microbiol. 68, 2997–3002 (2002). microbial majority. Annu. Rev. Microbiol. 57, cycle with special reference to anoxic-oxic interface 56. Lin, X. et al. Comparison of vertical distributions of 369–394 (2003). environments. Phil. Trans. R. Soc. Lond. B 298, prokaryotic assemblages in the anoxic Cariaco Basin 79. Fuhrman, J. A. & Davis, A. A. Widespread Archaea 543–561 (1982). and Black Sea by use of fluorescence in situ and novel Bacteria from the deep sea as shown by 33. Scranton, M. I., Astor, Y., Bohrer, R., Ho, T. Y. & hybridization. Appl. Environ. Microbiol. 72, 16S rRNA gene sequences. Mar. Ecol. Prog. Ser. 150, Muller-Karger, F. Controls on temporal variability of 2679–2690 (2006). 275–285 (1997). the geochemistry of the deep Cariaco Basin. Deep Sea 57. Zaikova, E. et al. Microbial community dynamics in a 80. Wright, T. D., Vergin, K. L., Boyd, P. W. & Giovannoni, Res. Part 1 Oceanogr. Res. Pap. 48, 1605–1625 seasonally anoxic fjord: Saanich Inlet, British S. J. A novel δ-subdivision proteobacterial lineage (2001). Columbia. Environ. Microbiol. 12, 172–191 (2010). from the lower ocean surface layer. Appl. Environ. 34. Anderson, J. J. & Devol, A. H. Deep water renewal in 58. Labrenz, M., Jost, G. & Jurgens, K. Distribution of Microbiol. 63, 1441–1448 (1997). Saanich Inlet, an intermittently anoxic basin. Estuarine abundant prokaryotic organisms in the water column 81. Brown, M. V. & Donachie, S. P. Evidence for tropical Coastal Marine Sci. 1, 1–10 (1973). of the central Baltic Sea with an oxic–anoxic interface. endemicity in the Deltaproteobacteria Marine Group 35. Diaz, R. J. & Rosenberg, R. Spreading dead zones and Aquat. Microb. Ecol. 46,177–190 (2007). B/SAR324 bacterioplankton clade. Aquat. Microb. consequences for marine ecosystems. Science 321, 59. Belmar, L. Molina, V. & Ulloa, O. Abundance and Ecol. 46, 107–115 (2007). 926–929 (2008). phylogenetic identity of archaeoplankton in the 82. Shanks, A. L. & Reeder, M. L. Reducing microzones This influential concept article describes the permanent oxygen minimum zone of the eastern and sulfide production in marine snow. Marine Ecol. ecological consequences of episodic or prolonged tropical South Pacific. FEMS Microbiol. Ecol. 78, Progr. Ser. 96, 43–47 (1993). oxygen starvation on coastal marine ecosystems. 314–326 (2011). 83. Alldredge, A. L. & Cohen, Y. Can microscale chemical 36. Helly, J. J. & Levin, L. A. Global distribution of 60. Koeppel, A. et al. Identifying the fundamental units of patches persist in the sea? Microelectrode study of naturally occurring marine hypoxia on continental bacterial diversity: a paradigm shift to incorporate marine snow, fecal pellets. Science 235, 689–691 margins. Deep Sea Res. Part 1 Oceanogr. Res. Pap. ecology into bacterial systematics. Proc. Natl Acad. (1987). 51, 1159–1168 (2004). Sci. USA 105, 2504–2509 (2008). 84. Dugdale, R. C. Goering, J. J., Barber, R. T., Smith, R. L. 37. Grantham, B. A. Upwelling-driven nearshore hypoxia 61. Fuchs, B. M., Woebken, D., Zubkov, M. V., Burkill, P. & & Packard, T. T. Denitrification and hydrogen sulfide in signals ecosystem and oceanographic changes in the Amann, R. Molecular identification of picoplankton the Peru upwelling region during 1976. Deep Sea Res. northeast Pacific. Nature 429, 749–754 (2004). populations in contrasting waters of the Arabian Sea. 24, 601–608 (1977). 38. Rabalais, N. N., Turner, R. E. & Wiseman, W. J. Aquat. Microb. Ecol. 39, 145–157 (2005). 85. Bruchert, V. et al. Regulation of bacterial sulfate Hypoxia in the Gulf of Mexico. J. Environ. Qual. 30, 62. Lavik, G. et al. Detoxification of sulphidic African shelf reduction and hydrogen sulfide fluxes in the central 320–329 (2001). waters by blooming chemolithotrophs. Nature 457, Namibian coastal upwelling zone. Geochim. 39. Monteiro, P. M. S., van der Plas, A. K., Melice, J. L. & 581–584 (2009). Cosmochim. Acta 67, 4505–4518 (2003). Florenchie, P. Interannual hypoxia variability in a This report is the first to detail abundant 86. Stewart, F. J., Ulloa, O. & DeLong, E. F. Microbial coastal upwelling system: ocean-shelf exchange, populations of SUP05 bacteria in association with metatranscriptomics in a permanent marine oxygen climate and ecosystem-state implications. Deep Sea process rate measurments for denitrification. minimum zone. Environ. Microbiol. 14, 23–40 (2012). Res. Part 1 Oceanogr. Res. Pap. 55, 435–450 63. Bano, N. & Hollibaugh, J. T. Phylogenetic composition This article provides the first gene expression (2008). of bacterioplankton assemblages from the Arctic profiles for sulphide oxidation, denitrification, 40. Naqvi, S. W. A. et al. Severe fish mortality associated Ocean. Appl. Environ. Microbiol. 68, 505–518 (2002). anammox and nitrification in the ETSP OMZ. with ‘red tide’ observed in the sea off Cochin. Curr. Sci. 64. Walsh, D. A. & Hallam, S. J. in Handbook of Molecular 87. Sorensen, J. B. Nitrate reduction in marine sediments: 75, 543–544 (1998). Microbial Ecology II: Metagenomics in Different Habitats pathways and interactions with iron and sulfur-cycling. 41. Zehnder, A. J. & Stumm, W. in Biology of Anaerobic (eds de Bruijn, F. J.) 253–267 (Wiley & Sons, 2011). Geomicrobiol. J. 5, 401–421 (1987). Microorganisms (ed. Zehnder, A. J.) 1–38 (Wiley & 65. Swan, B. K. et al. Potential for chemolithoautotrophy 88. Samuelsson, M. & Rönner, U. Ammonium production Sons, 1988). among ubiquitous bacteria lineages in the Dark by dissimilatory nitrate reducers isolated from Baltic 42. Codispoti, L. A. & Christensen, J. P. Nitrification, Ocean. Science 333, 1296–1300 (2011). Sea water, as indicated by 15N study. Appl. Environ. denitrification and nitrous oxide cycling in the easter This paper reveals the hidden metabolic powers of Microbiol. 44, 1241–1243 (1982). tropical South Pacific Ocean. Marine Chem. 16, Arctic96BD‑19 and SAR324 bacteria using 89. Kartal, B. et al. Anammox bacteria disguised as 277–300 (1985). single-cell genomics techniques. denitrifiers: nitrate reduction to dinitrogen gas via nitrite 43. Santoro, A. E., Buchwald, C., McIlvin, M. R., & 66. Lin, X. J., Scranton, M. I., Chistoserdov, A. Y., Varela, R. and ammonium. Environ. Microbiol. 9, 635–642 (2007).

Casciotti, K. L. Isotopic signature of N2O produced by & Taylor, G. T. Spatiotemporal dynamics of bacterial 90. Brune, A., Frenzel, P. & Cypionka, H. Life at the oxic– marine ammonia-oxidizing archaea. Science 333, populations in the anoxic Cariaco Basin. Limnol. anoxic interface: microbial activities and adaptations. 1282–1285 (2011). Oceanogr. 53, 37–51 (2008). FEMS Microbiol. Rev. 5, 691–710 (2000).

NATURE REVIEWS | MICROBIOLOGY VOLUME 10 | JUNE 2012 | 393

91. Stewart, F. J., Newton, I. L. G. & Cavanaugh, C. M. 104. Chaffron, S., Rehrauer, H., Pernthaler, J. & von Acknowledgements Chemosynthetic endosymbioses: adaptations to oxic- Mering, C. A global network of coexisting microbes This work was carried out under the auspices of the Natural anoxic interfaces. Trends Microbiol. 13, 439–448 from environmental and whole-genome sequence Sciences and Engineering Research Council (NSERC) of Canada, (2005). data. Genome Res. 20, 947–959 (2010). the Canada Foundation for Innovation (CFI) and the Canadian 92. Dubilier, N. et al. Endosymbiotic sulphate-reducing This paper describes a useful method for detecting Institute for Advanced Research (CIFAR). J.J.W. was supported and sulphide-oxidizing bacteria in an oligochaete co‑occurrence patterns within and between by the NSERC, and K.M.K. was supported by the Tula worm. Nature 411, 298–302 (2001). microbial communities using geo-referenced Foundation-funded Centre for Microbial Diversity and Evolution 93. Schink, B. Synergistic interactions in the microbial taxonomic and functional-gene information. (CMDE) at the University of British Columbia, Vancouver, world. Antonie Van Leeuwenhoek 81, 257–261 (2002). 105. Alldredge, A. L. & Silver, M. Characteristics, dynamics Canada. We thank N. Korniyuk, P. Macoun, V. Tunnicliffe, 94. Stams, A. J. & Plugge, C. M. Electron transfer in and significance of marine snow. Prog. Oceanogr. 20, F. Whitney, M. Robert, A. Hawley and N. Hanson for insightful syntrophic communities of anaerobic bacteria and 41–82 (1988). commentary; D. Walsh for assistance with data visualization; archaea. Nature Rev. Microbiol. 7, 568–577 (2009). 106. Karl, D. M., Knauer, G. A., Martin, J. H. & Ward, B. B. P. Tortell for help with figure 1; and the anonymous reviewers 95. Jeong, H., Tombor, B., Albert, R., Oltvai, Z. N. & Bacterial chemolithotrophy in the ocean is associated for insightful commentary and guidance. Barabasi, A. L. The large-scale organization of with sinking particles. Nature 309, 54–56 (1984). metabolic networks. Nature 407, 651–654 (2000). 107. Woebken, D., Fuchs, B. M., Kuypers, M. M. & Amann, R. Competing interests statement 96. Ravasz, E., Somera, A. L., Mongru, D. A., Oltvai, Z. N. Potential interactions of particle-associated anammox The authors declare no competing financial interests. & Barabasi, A. L. Hierarchical organization of modularity bacteria with bacterial and archaeal partners in the in metabolic networks. Science 297, 1551–1555 (2002). Namibian upwelling system. Appl. Environ. Microbiol. 97. Barabasi, A. L. & Oltvai, Z. N. Network biology: 73, 4648–4657 (2007). understanding the cell’s functional organization. 108. Karl, D. M. & Tilbrook, B. D. Production and transport FURTHER INFORMATION Nature Rev. Genet. 5, 101–113 (2004). of methane in oceanic particulate organic-matter. Steven J. Hallam’s homepage: 98. Chisholm, S. W. & Cary, C. Ecological genomics: the Nature 368, 732–734 (1994). http://www.cmde.science.ubc.ca/hallam/ application of genomic sciences to understanding the 109. Paerl, H. W. & Pinckney, J. L. A mini-review of microbial Center for Coastal Margin Observation and Prediction: structure and function of marine ecosystems. In consortia: their roles in aquatic production and http://www.stccmop.org/ Report, NSF workshop on Marine Microbial Genomics biogeochemical cycling. Microb. Ecol. 31, 225–247 Genbank: http://www.ncbi.nlm.nih.gov/genbank/ 1–20 (Chisholm Univ. Press, 2001). (1996). Greengenes taxonomy database: http://greengenes.lbl.gov 99. Raes, J. & Bork, P. Molecular eco-systems biology: 110. Barabasi, A. L. & Albert, R. Emergence of scaling in Institute of Ocean Sciences: towards an understanding of community function. random networks. Science 286, 509–512 (1999). http://www.pac.dfo-mpo.gc.ca/sci/sci/facilities/ios_e.htm Nature Rev. Microbiol. 6, 693–699 (2008). 111. Newman, M. E. J. Networks: An Introduction (Oxford Leibniz Institute for Baltic Sea Research: 100. Ruan, Q. et al. Local similarity analysis reveals unique Univ. Press, 2010). http://www.io-warnemuende.de/en_index.html associations among marine bacterioplankton species 112. Jeong, H., Mason, S. P., Barabasi, A. L. & Oltvai, Z. N. Monterey Bay Aquarium Research Institute: and environmental factors. Bioinformatics 22, Lethality and centrality in protein networks. Nature http://www.mbari.org/default.htm 2532–2538 (2006). 411, 41–42 (2001). The Eastern South Pacific Oxygen Minimum Zone: 101. Fuhrman, J. A. & Steele, J. A. Community structure of 113. Barabasi, A. L. & Bonabeau, E. Scale-free networks. http://omz.udec.cl/ marine bacterioplankton: patterns, networks, and Sci. Am. 288, 60–69 (2003). The Fourth Paradigm: Data-Intensive Scientific Discovery: relationships to function. Microb. Ecol. 53, 69–81 114. Yamada, T. & Bork, P. Evolution of biomolecular http://research.microsoft.com/en-us/collaboration/ (2008). networks — lessons from metabolic and protein fourthparadigm/ 102. Steele, J. A. et al. Marine bacterial, archaeal and interactions. Nature Rev. Mol. Cell Biol. 10, 791–803 Victoria Experimental Network Under the Sea (VENUS): protistan association networks reveal ecological (2009). http://www.venus.uvic.ca/ linkages. ISME J. 5, 1414–1425 (2011). 115. Garcia, H. E. et al. World Ocean Atlas 2009 Vol. 3 103. Gilbert, J. A. et al. Defining seasonal marine microbial (ed. Levitus, S.) (US Government Printing Office, ALL LINKS ARE ACTIVE IN THE ONLINE PDF community dynamics. ISME J. 6, 298–308 (2012). 2010).

394 | JUNE 2012 | VOLUME 10 www.nature.com/reviews/micro