ORIGINAL ARTICLE

Microbial community successional patterns in beachsands impacted by the Deepwater Horizon oil spill

Luis M Rodriguez-R1, Will A Overholt1, Christopher Hagan2, Markus Huettel2,Joel E Kostka1,3 and Konstantinos T Konstantinidis1,4

1School of Biology, Georgia Institute of Technology, Atlanta, GA, USA; 2Department of Earth, Ocean andAtmospheric Sciences, Florida State University, Tallahassee, FA, USA; 3School of Earth and AtmosphericSciences, Georgia Institute of Technology, Atlanta, GA, USA and 4School of Civil and EnvironmentalEngineering, Georgia Institute of Technology, Atlanta, GA, USA

Although petroleum hydrocarbons discharged from the Deepwater Horizon (DWH) blowout wereshown to have a pronounced impact on indigenous microbial communities in the Gulf of Mexico,effects on nearshore or coastal ecosystems remain understudied. This study investigated thesuccessional patterns of functional and taxonomic diversity for over 1 year after the DWH oil wasdeposited on Pensacola Beach sands (FL, USA), using metagenomic and 16S rRNA gene amplicontechniques. Gamma- and Alphaproteobacteria were enriched in oiled sediments, in corroboration ofprevious studies. In contrast to previous studies, we observed an increase in the functional diversityof the community in response to oil contamination and a functional transition from generalistpopulations within 4 months after oil came ashore to specialists a year later, when oil wasundetectable. At the latter time point, a typical beach community had reestablished that showed littleto no evidence of oil hydrocarbon degradation potential, was enriched in archaeal taxa known to besensitive to xenobiotics, but differed significantly from the community before the oil spill. Further,a clear succession pattern was observed, where early responders to oil contamination, likelydegrading aliphatic hydrocarbons, were replaced after 3 months by populations capable of aromatichydrocarbon decomposition. Collectively, our results advance the understanding of how naturalbenthic microbial communities respond to crude oil perturbation, supporting the specialization-disturbance hypothesis; that is, the expectation that disturbance favors generalists, while providing(microbial) indicator species and genes for the chemical evolution of oil hydrocarbons duringdegradation and weathering.The ISME Journal advance online publication, 17 February 2015; doi:10.1038/ismej.2015.5

Introduction

The oil spill caused by the blowout of the DeepwaterHorizon (DWH). Drilling rig in April 2010 constitu-tes the largest accidental release of oil into themarine environment in recorded history. Oil con-tamination from the DWH spill had a profoundimpact on indigenous microbial communities, andall available studies recognize shifts in the composi-tion of microbial communities in direct contact withoiled seawater and sediments in comparison withpristine environments (Atlas and Hazen, 2011; Joyeet al., 2014; Kostka et al., 2014; King et al., 2015).Moreover, consistent patterns were observed inmicrobial communities exposed to DWH oil in the

Gulf of Mexico including an increase in the relativeabundance of members of the Gammaproteobacteria, aprevalence of known hydrocarbon-degrading popula-tions, and the enriched abundance and expression ofgenes related to hydrocarbon degradation (Joye et al.,2014; Kostka et al., 2014; King et al., 2015). Thesepatterns and microbial responses are also in accor-dance with observations from laboratory studies andprevious accidental releases of oil in marine environ-ments (Roling et al., 2002; Head et al., 2006; Yakimovet al., 2007; Berthe-Corti and Nachtkamp, 2010; Greer,2010; McGenity et al., 2012).

The Unified Area Command estimated thatapproximately one-half of the B4.9 million barrelsof oil released from the DWH blowout reached theocean surface (Lubchenco et al., 2010), and a portionof this surfaced oil transported to nearshore andcoastal ecosystems was buried in the sediments(Hayworth et al., 2011; Wang and Roberts, 2013),impacting approximately 850 km of beaches fromeast Texas to west Florida (Michel et al., 2013). Oilstarted depositing on the Pensacola Beach sands

Correspondence: KT Konstantinidis, Civil and EnvironmentalEngineering, Georgia Institute of Technology, 311 Ferst Dr, FordES&T Building, Suite 3224, Atlanta, GA 30332, USA.E-mail: kostas@ce.gatech.eduReceived 12 September 2014; revised 16 December 2014; accepted23 December 2014

The ISME Journal (2015), 1–13& 2015 International Society for Microbial Ecology All rights reserved 1751-7362/15

www.nature.com/ismej

studied here on 22 June 2010. The input of largeamounts of crude oil, including an array of poten-tially toxic compounds, posed a potential distur-bance for benthic microbial communities (Valentineet al., 2012). Available studies to date were primarilyfocused on the water column and/or deep seaecosystems, and less is known about the responseor adaptation of sedimentary communities to oiling(Huettel et al., 2014). Studies characterizing thetaxonomic shifts between contaminated and non-contaminated beach sediments recognized that theoil input strongly affected the beach sand microbialcommunities, which responded with increasedbacterial cell densities (Kostka et al., 2011), reducedtaxonomic diversity, and a succession of microbialpopulations that paralleled the changes in abun-dance and composition of deposited hydrocarbons(Kostka et al., 2011; Bik et al., 2012; Lamendellaet al., 2014). Consistent responses have beenobserved across study sites, although other factorssuch as site heterogeneity and seasonal fluctuationsin environmental parameters have been shown tosomewhat confound assessments of the oil impact incertain beaches (Newton et al., 2013), sometimesmaking them undetectable (Roling et al., 2004). Ingeneral, an initial increase in the relative represen-tation of known oil degraders, mostly of theGammaproteobacteria class (most notably Alcani-vorax), was observed together with a temporalsuccession characterized by an increase in relativeabundance of Bacillus, Microbacterium and mem-bers of the Alphaproteobacteria class at later stages,when recalcitrant oil hydrocarbons predominate(Kostka et al., 2011). Moreover, the increase in oildegraders was concomitant with an increasedexpression of polycyclic aromatic hydrocarbons,n-alkane and toluene degradation genes as assessedby metatranscriptomics (Lamendella et al., 2014).Although these findings provided importantinsights into the effects of oil on benthic microbialcommunity composition, the gene functionsselected for and the genomic adaptations inresponse to the presence of oil remained mostlyuncharacterized in the Gulf coast.

Previously identified shifts in microbial commu-nities in response to DWH oil, both in the watercolumn and sediments, indicated significant sus-ceptibility of these communities; susceptibilitydefined as the degree to which community composi-tion changes in response to disturbance (Shadeet al., 2012). These observations are in accordancewith the majority of ecological studies addressingthe effect of disturbances such as carbon inputs onmicrobial communities, which have found evidenceof susceptibility (reviewed by Allison and Martiny,2008). However, the magnitude, stability and sto-chasticity of functional responses, as well as themechanisms driving the taxonomic and functionalcomposition of the microbial community afterdisturbance are not well understood (Reed andMartiny, 2007). For example, it has been recognized

in plant and animal communities that generalistpopulations better withstand disturbances, whereasspecialist populations tend to be favored in stableenvironments (specialization-disturbance hypo-thesis; Vazquez and Simberloff, 2002). Accordingto the disturbance-specialization hypothesis, mostspecialist taxa are selected against when commu-nities experience a severe disturbance, as they areadapted to relatively narrow niches in their naturalecosystem. In contrast, generalists are more resilientto disturbances altering the niches. In turn, thetaxonomic diversity of the community is negativelyimpacted by a disturbance, but the functionaldiversity can increase as an effect of the disturbance.Although some previous studies applied ecologicaltheory to describe the response and recovery ofcommunity dynamics to disturbance (cf. Prosseret al., 2007; Shade et al., 2012), the relationship ofdisturbance and specialization remains largelyunexplored in microbial communities. Disturbedcommunities are typically observed to encompassreduced taxonomic and/or phylogenetic diversitycompared with undisturbed controls, but whetherthis pattern translates to reduced functional diver-sity or increased specialization remains largelyunknown. In this study, we aimed to characterizethe response of sedimentary microbial communitiesfrom Pensacola Beach to the DWH oil spill, as anin-situ experiment of the effects of disturbance onfunctional and taxonomic diversity.

Materials and methods

Beach sands were collected at Pensacola MunicipalBeach, FL, USA (30119.57 N, 087110.47 W) on 6, 10,20 and 24 May 2010 (before arrival of the oil plumeto the shoreline; hereafter, termed pre-oil commu-nities/samples), 30 July 2010 (one month after theoil reached the beach; oiled), 20 October 2010 (whenoil constituents were still present in the sand;weathered oiled), and 14 June 2011 (when oil wasnot visually detectable; recovered; Table 1 andSupporting methods). Samples were collected fromaerobic beach sediments (oxygen concentrationsbetween 210 and 230 mmol l!1 down to 55 cm depth,which represents 450% of air saturation level)above groundwater level.

16S rRNA gene amplicons were sequenced, andthe resulting sequences were analyzed as describedrecently (Poretsky et al., 2014). Trimmed sequenceswere clustered into operational taxonomic units(OTUs) at 97% similarity using UCLUST (Edgar,2010), OTUs that represented o0.005% of the totalsequences were discarded (Bokulich et al., 2013)and representative sequences of each OTU wereclassified using the RDP Classifier at 50% confi-dence (Wang et al., 2007). Shotgun community DNAwas sequenced, and the resulting metagenomicreads were quality checked, assembled and anno-tated as described in the supplementary online

Response of benthic microbes to oilingLM Rodriguez-R et al

2

The ISME Journal

material. The level of coverage of the communityachieved by each metagenomic dataset was esti-mated and projected using Nonpareil with defaultparameters (Rodriguez-R and Konstantinidis,2014b). Assembled contigs were taxonomicallyannotated using MyTaxa (Luo et al., 2014). 18SrRNA gene-encoding reads were identified byMetaxa (Bengtsson et al., 2011) with e-value o0.1and taxonomically annotated using pplacer andtaxtastic (Matsen et al., 2010). Read mapping toestimate the relative abundance of genes and taxawas performed using BLAT with default parameters(Kent, 2002), considering only the best match withalignment length X80 bp and identity X97%.Annotation terms and taxa with significantly differ-ent abundance between groups of samples wereidentified using the negative binomial test asimplemented in DESeq2 (Anders and Huber, 2010).

To measure the average number of genes per cellwith a given functional annotation (genome equiva-lents), a set of universally conserved single-copygenes were identified among the assembled genesequences from the metagenomes. All genes werecompared against a collection of 101 HMMs (Dupontet al., 2012), using HMMER3 (http://hmmer.jane-lia.org/) with default settings and trusted cutoff,excluding genes for which more than one modelrepresented the same gene family. The mediansequencing depth (in reads/bp) of the remaining 91models was used as the normalizing factor for eachdataset. The sequencing depth of genes with a givenannotation (see below) was estimated for each dataset(in reads/bp), added up and divided by the normal-izing factor of the corresponding dataset.

To identify genes related to oil degradation,gene-specific databases were compiled and manuallycurated. Sequences for AlkB (alkane hydroxylase)

and CYP153 (cytochrome P450 family) were derivedfrom the annotated datasets by Wang et al. (2010);sequences for NahA (naphthalene 1,2-dioxygenase)were derived from the set compiled by Lu et al., 2012;and sequences for ArhA (polycyclic aromatic hydro-carbon dioxygenase) and BBS (benzylsuccinyl-CoAdehydrogenase) were derived from UniRef50 clus-ters (Suzek et al., 2007). Putative proteins of theassembled metagenomes were functionally identi-fied using blastp (Altschul et al., 1990) against eachreference dataset, with a 250 bit-score threshold.The resulting dataset for AlkB was aligned usingMuscle v3.8.31 with default parameters (Edgar,2004), and the gene phylogeny was reconstructedusing RAxML v7.7.2 with GTR model (proteins),gamma parameter optimization, and ‘-f a’ algorithm(Stamatakis, 2006). Putative coding fragmentspredicted with FragGeneScan (Rho et al., 2010) onsequence reads were subsequently placed onto thereconstructed tree based on a sequence-to-profilealignment built with Clustal Omega v1.1.0 (Sieverset al., 2011), using the evolutionary placementalgorithm (Berger et al., 2011). The same placementstrategy was independently applied to the partialsequences of AlkB reported in the study by Smithet al. (2013) (GenBank entries KF613175-KF613575).

Diversity was calculated as the true diversity oforder one (1D; equivalent to the exponential ofShannon index). The a and g components wereestimated from the abundance of categories in asample and in all samples, respectively, andadjusted for unobserved fractions using the Chao-Shen correction (Chao and Shen, 2003) as imple-mented in the R package entropy (Hausser andStrimer, 2013). Richness was estimated using theChao1 index (Chao, 1984), and evenness wascalculated as the corrected true diversity of order

Table 1 Samples used in this study

Designation Reads after trimminga Statusb Depth (cm) Sampling date Sediment temp. (1C)c

S1 2 937 972 Pre-oil 0 6 May 2010 29.96±2.66d

S2 7 951 456 Pre-oil 0 10 May 2010S3 7 837 390 Pre-oil 0 20 May 2010S4 6 710 972 Pre-oil 0 24 May 2010A 32 840 836 Oiled 30–40 30 Jul 2010 30.49±2.72B 32 392 430 Oiled 35C 25 024 134 Oiled 30–40D 21 469 632 Weathered oil 48–62E 26 279 070 Oiled 40–45 20 Oct 2010 23.73±2.95F 34 830 190 Oiled 25–47G 39 208 672 Oiled 24–36H 25 224 316 Weathered oil 50–55I600 33 188 686 Recovered 30–40 14 Jun 2011 31.02±2.79I606 31 477 910 Recovered 30–40J598 31 724 116 Recovered 50–65J604 28 119 496 Recovered 50–65

aReads after quality trimming with maximum probability of error of 1% and minimum length of 50 bp, and removal of contamination with adaptorsequences.bSamples with oiled and weathered oil status were distinguished based on visual assessment of oil presence. Recovered status was defined basedon undetectable levels of hydrocarbons at depths similar to (previously) oiled samples.cSediment temperature between 0 and 50 cm depth presented as mean±one standard deviation.dData for May 2010 not available, presented values were measured in June 2010. Cf. temperatures in May 2011: 25.21±2.07.

Response of benthic microbes to oilingLM Rodriguez-R et al

3

The ISME Journal

one (number of equivalent groups) divided by theestimated richness (number of groups).

Crude oil hydrocarbons in the sediment sampleswere identified by gas chromatography-mass spec-trometry using an Agilent 7890A Series GC (SantaClara, CA, USA), coupled to an Agilent 7000 triplequadrupole MS system, as described previously(Zuijdgeest and Huettel, 2012). The supplementaryonline material provides further information aboutprocedures and analytical techniques.

All sequencing datasets were deposited in theNCBI Sequence Read Archive under projectPRJNA260285 and additional material is availableat http://enve-omics.ce.gatech.edu/data/oilspill.

Results

Description of samples and their metagenomesConcentrations of total petroleum hydrocarbonsquantified by gas chromatography-mass spectro-metry and visible oil stains monotonically decreasedbetween sampling dates (P-values p0.05, one-sidedt-test; Figure 1a). Specifically, the depth-integratedsedimentary inventories of small molecular weightaliphatic and aromatic compounds decreasedrapidly from 6 and 1 mg kg! 1, respectively, in Julyto less than 0.5 mg kg!1 in October. In contrast,

gas chromatography-mass spectrometry profilesrevealed that sedimentary inventories of aromaticcompounds greater than C8 remained unchangedduring this same time frame, whereas aliphaticcompounds greater than C6 displayed only amarginal reduction.

A total of 16 metagenomic samples, ranging insize from 3 to 78 million reads after trimming(paired-end reads with average length of 90–190 bpper dataset), were recovered from each of the foursampling time points, with at least three replicatesper time point (Table 1). The metagenomes from pre-oil samples had an estimated abundance-weightedaverage coverage (Rodriguez-R and Konstantinidis,2014b) of 18–39%, the oiled samples a coverage of35–60% and the samples from recovered commu-nities an average coverage of 20–25%. Nonpareilcurves indicated that the communities in therecovered samples had a higher sequence complex-ity than both pre-oil and oiled communities, withpre-oiled communities displaying a slightly lowersequence complexity (Supplementary Figure S1A).The described trend in sequence complexity corre-sponded to the estimated richness of these commu-nities based on OTUs from 16S rRNA gene amplicondata (Supplementary Figure S1B). In general, allmetagenomes showed lower sequence complexitythan previously determined metagenomes from

S1

S2S3S4

AB

C

D EFG

H

I600I606J598 J604

S1

S2

S3S4

AB

C

D

EFG

H

I600 I606J598J604

Genera Subsystems

A B C D E F G H J5980

50

J604I606I600

TPH(mg/Kg)

p-value ≈ 0.05

Sample ID

p-value ≈ 0.01

Figure 1 Shifts in taxonomic and functional profiles in relation to oil concentration. (a) The concentration of total petroleumhydrocarbons was significantly higher in July samples (A, B, C) relative to October 2010 (E, F, G), and June 2011 samples (I600, I606,J598, J604). The comparisons between groups (July to October 2010, and October 2010 to June 2011) were performed using one-sidedt-tests (P-values in grey boxes), and the average per group is indicated as horizontal lines. The non-metric multidimensional scaling of(b) genera and (c) subsystems reveals non-overlapping regions between pre-oiled (green), oiled (shadowed grey) and recovered (olive)samples. The two-dimensional stresses for genera and subsystems are 3.361% and 3.358%, respectively, and the origins are indicatedwith grey lines. Distance matrices were generated using Bray-Curtis dissimilarities of normalized read counts and ordination wasselected by minimizing stress on two dimensions. Note that the heavily oiled samples also form non-overlapping areas by sampling date(dark green and brown), and are distinguishable from weathered oil samples (pink and dark red).

Response of benthic microbes to oilingLM Rodriguez-R et al

4

The ISME Journal

clayey or silty soils such as rain forest andpermafrost but were more complex than freshwateror ocean planktonic metagenomes (SupplementaryFigure S1A; cf. Rodriguez-R and Konstantinidis,2014a). The July and October 2010 samples (oiledand weathered oil) were assembled into B56 000contigs per sample with N50 of B1400 bp; whilethose from recovered samples resulted in B12 000contigs per sample with N50 of B745 bp(Supplementary File S1). These results furthersupported the Nonpareil estimates of highersequence complexity in the latter samples. In total,B6 70 000 contigs were obtained with an overallN50 of 1101 bp (723 Mbp in total, from 37 Gbp ofsequencing reads), on which B1.2 million geneswere predicted, resulting in an average codingdensity of 87% (Supplementary File S1).

Microbial community specialization in response tooilingTo assess the temporal effects of the oil spill on themicrobial community composition and its recovery,the functional and taxonomic profiles at differenttime points were compared. Four main groups wereidentified which significantly differed in bothtaxonomic and functional distributions (P-valuesp0.003, ANOSIM based on Bray-Curtis dissimilar-ity; Figures 1b and c) and were consistent with theoil concentrations measured in-situ: Pre-oil (S1, S2,S3, and S4), Oiled July 2010 (A, B, C), Oiled October2010 (E, F, G) and Recovered (I600, I606, J598, J604).16S rRNA gene amplicon data also demonstratedthat sample depth played a limited role in structu-ring microbial communities (Supplementary FigureS2; ADONIS: 3% variance explained by depth vs75% explained by oiling status and collection date),which was consistent with the facts that the beachsands studied here are subjected to high levels oferosion, and high levels of oxygen (450% of airsaturation level) were detectable at all samplingdepths. Hence, our pre-oiled datasets, eventhough originated from different depths (surficial)compared to oiled datasets (30–65 cm), representedreliable controls for assessing the oiled and recoveredmicrobial communities.

Most notably, the communities exhibited anincrease in the functional diversity in oiled sampleswith respect to pre-oil samples, and a reductionin functional diversity in recovered sampleswith respect to oiled samples (SupplementaryFigure S3A), revealing a different state of lowerfunctional diversity in the recovered communities(Supplementary Figure S3B; DECORANA analysis).Interestingly, this pattern was not observed in thetaxonomic diversity, richness or evenness levels(Supplementary Figure S3C-E), indicating that itwas primarily due to a decrease in functionalspecialization of the communities in the oiledsamples. This interpretation is further supportedby a concomitant decrease in the estimated

minimum doubling time in the oiled communities(Supplementary Figure S4A), as expected forbacteria with more generalist strategies (Dethlefsenand Schmidt, 2007). More generalist prokaryotes tendto have larger genomes (Konstantinidis and Tiedje,2005), but no significant changes in the estimatedaverage genome size were detected (SupplementaryFigure S4B). Nevertheless, these results suggestedthat the oil disturbance caused community shiftscharacterized by a decrease in functional speciali-zation and a consequent increase in functionaldiversity, which were reversed in the post-disturbancerecovery process as the succession advanced.

Oil degradation and toxicity drives communityphylogenetic compositionDifferences in the composition of the communitiesfrom pre-oil, oiled and recovered sediments weredetected at various levels of taxonomic resolution(Figure 2; Supplementary File S3). At the mostgeneral level (domain), recovered communitiesexhibited higher fractions of eukaryotic and archaealmembers than oiled and pre-oiled communities(Figure 3a), although no differences in the taxo-nomic composition of the eukaryotic fraction wereobserved (Supplementary File S3). The higherfraction of eukaryotic sequences is also consistentwith the lower coding potential of May, Julyand October 2010 metagenomes (B89% oftotal sequence length was protein-coding) vs theRecovered (June 2011) metagenomes (70%;Supplementary File S1). The higher representationof dominant taxa and lower evenness in commu-nities from oiled samples was also evident at theclass level, where Gamma- and Alphaproteobacteriaincreased in abundance, with a concomitantdecrease of novel taxa (represented by the unclassi-fied fraction; Figure 2b). The genera significantlymore abundant in oiled than in pre-oiled and/orrecovered samples were primarily well-known andsuspected hydrocarbon degraders, including Alca-nivorax, Pseudomonas, Hyphomonas, Parvibacu-lum, Marinobacter and Micavibrio (Figure 2c). Incontrast, groups significantly enriched in recoveredsamples included taxa typically found in marineenvironments and known to be highly susceptible toxenobiotics such as the archaeal genera Nitrosopu-milus and Cenarchaeum.

Functional gene content shift in response to oilTo further investigate the specific functional traitsselected by oil presence and, presumably, accountedfor the community compositional shifts observed,the abundances of genes associated with alkaneand aromatic degradation pathways were comparedbetween pre-oil, oiled and recovered samples.In all evaluated cases, oiled communities displayeda larger prevalence of gene annotations associatedwith aromatic and alkane hydrocarbon degradation

Response of benthic microbes to oilingLM Rodriguez-R et al

5

The ISME Journal

as well as beta-oxidation than pre-oil and recoveredcommunities (Figure 3). Interestingly, the relativeabundance of most genes associated with aliphaticsdegradation dropped from July to October 2010, inparticular those associated with rubredoxin-NADþ /NADP reduction and aldehyde oxidation (top panel

in Figure 3). In contrast, the abundance of genesassociated with aromatics degradation was roughlymaintained or, in some cases, increased from July toOctober 2010 (second panel in Figure 3). In addition,functions related to nutrient scavenging such asallantoicase and nitrogenase (low nitrogen response),

S1 S2 S3 S4 A B C D E F G H I600 I606 J598 J604

20

40

10

30

0

50

Class:

Abun

danc

e(%

nor

mal

ized

cou

nts)

10-3

10-2

0.1

1

10

2.26 2.5 2.15 2.68 2.17

2.33

3.043.69

3.043.33

4.73.12

2.373.13 2.42

2.16

2.63 2.22

2.9

5.72

5.84

3.31

5.09

2.644.22 3.68

8.42

3.273.44

3.38

10.5

12.2

7.83

8.06

12.912.5

10.1

5.86

3.76 4.15

5.695.58

1312.9

13.3 13.4

23.8

19.4

12.7

16.5

13.1

14.8

15.8

7.48

7.34 6.94

8.118.02

GammaproteobacteriaAlphaproteobacteriaBacteroidiaThaumarchaeota (n.c.)Cyanobacteria (n.c.)BetaproteobacteriaFlavobacteriiaSpirochaetiaActinobacteria

DeltaproteobacteriaBacilliVerrucomicrobiaeAnaerolineaePlanctomycetiaCytophagiaClostridiaSphingobacteriiaChloroflexi

NitrosopumilusT

●●

●

Parvibaculumα

●

●

Pseudomonasγ●

●

Thioalkalivibrioγ●

●

Nitrosococcusγ●

●

Wolbachiaα

●

●

Glaciecolaγ ●

●

Alcanivoraxγ●

●Vibrioγ●

●

Marinobacterγ●

●

Ca. NitrosoarchaeumT●

●

Micavibrioα

●

●

Hirschiaα

●

●

Burkholderiaβ●●

Hyphomonasα

●

●

Azospirillumα

●

●

PediococcusB

●

●α: Alphaproteobacteriaβ: Betaproteobacteriaγ: GammaproteobacteriaT: Thaumarchaeota (n.c.)B: Bacilli

M A A S O N D J F M A M JM J JFJ2010 2011

Blowout Oil onshore Mechanic cleaning

Abun

danc

e(%

read

s m

appi

ng to

gen

es)

24.2 21.0 16.5 15.610.2 14.2

93.5 92.9 94.0 93.7 93.4 93.0 95.4 83.8 90.1 87.1 90.7 79.3 70.2 73.4 75.9 76.6

20

40

60

80

0

100Bacteria Archaea Eucarya ViraDomain:

Abun

danc

e(%

cla

ssifi

ed re

ads)

NitrospiraMollicutesChlamydiiaSpartobacteriaOpitutaeChlorobiaEpsilonproteobacteriaOther Bacteria (32)Other Archaea (11)

S1 S2 S3 S4 A B C D E F G H I600 I606 J598 J604

Significantly different generawith base abundance over 0.1%

Figure 2 Taxonomic shifts in the microbial community in response to oil. The distribution of metagenomic reads in (a) domains and(b) classes is displayed for taxa that recruited more than 10% and 2% of the total reads, respectively (white numbers). (c) Generawith abundance above 0.1% and significantly different between pre-spill and oiled or between oiled and recovered samples(P-value adjustedp0.01) are also displayed. The minimum and maximum abundance of each genus is indicated with open and filledcircles, respectively, and the class is indicated with superscripts.

Response of benthic microbes to oilingLM Rodriguez-R et al

6

The ISME Journal

and siderophores (iron chelation) were observed to beincreased in oiled samples, whereas functions relatedto primary production such as iron-responsive ele-ments (iron-responsive binding), as well as functionsrelated to photosynthesis (possibly transported fromneighboring marine communities), were enriched inthe recovered communities. Notably, most functionalcategories exhibiting statistically significant differ-ence in abundance in the oiled communities returnedto the pre-oil state in the recovered communities, insome cases exceeding their pre-oil levels (Figure 3and Supplementary File S4).

To explore the phylogenetic diversity of genesrelated to oil degradation, we selected AlkB (alkanehydrolase) as a marker for alkane degradation andreconstructed a high-quality gene phylogeny basedon 66 reference genes (mostly based on Wang et al.,2010) and 43 genes recovered from metagenomicassemblies. In addition, individual metagenomicreads from all datasets were assigned to the mostlikely node in the tree to provide a quantitativepicture of the shifts of AlkB variants over time(Figure 4). This dataset included sequences from14 different genera in five different classes, hence

004764600337210004029000402200087260008860001504600150440018685

00477990052881005026900185090055041001862500087560016682004756900185110018774004707000186280008412001861800186260018623001868700186240018619001866700337670033776001845600186950018577001858300186660018662

000398800038570080023001881200164010016508000430000084700017099000446600709910003995000399700319550004321

0008408000461800037350004826

00303500008198001534300151090004037001616300152040015112

0016168

0016984

0045156

Oiled / Jul 2010 (A, B, C)Oiled / Oct 2010 (E, F, G)Recovered / Jun 2011 (I600, I606, J598, J604)

●●●●

p-value adjusted ≤ 0.01p-value adjusted ≤ 0.05p-value adjusted > 0.05

preoiloil

More abundant inGO IDAbundance

across time points oilrecov.

Log2 fold-change0 5-5

Genome equivalents(mean copies per cell)

-1010110-110-2 010001 -5 5Preoiled / May 2010 (S1, S2, S3, S4)

Beta

-oxi

datio

nAl

ipha

tics

degr

adat

ion

Arom

atic

s de

grad

atio

n

ferrous iron bindingsiderophore transmembrane transporter

allantoicasenitrogenase

nitrate transmembrane transporterNutrientsscavenging& response

chlorophyll binding

ribulose-bisphosphate carboxylase

e- transporter, cyclic e- transport pathway

Photosynthesis

Gene Ontology term

3'-5' exonucleasephosphoglycerate kinase

structural constituent of ribosomephenylalanine-tRNA ligase

House-keepinggenes

N

Fe

urea transmembrane transporter

chromate transmembrane transporter

iron-responsive element binding

AlkaneInitial oxidationAlcohol oxidationAldehyde oxidation

alkanal monooxygenase (FMN-linked)

aldehyde dehydrogenase (NAD)alcohol dehydrogenase (NAD)

alkanesulfonate monooxygenaseferredoxin-NAD+ reductase

rubredoxin-NADP reductaserubredoxin-NAD+ reductase

alkane 1-monooxygenase

Fatty acid-CoA

Oxo-acyl-CoA

Fatty acyl-CoA

Enoyl-CoA

Hydroxy-acyl-CoA

acetyl-CoA C-acyltransferase3-hydroxyacyl-CoA dehydrogenase

3R-hydroxyacyl-CoA dehydratase3-hydroxyacyl-CoA dehydratase

palmitoyl-CoA oxidaselong-chain-enoyl-CoA hydratase

enoyl-CoA hydrataseisovaleryl-CoA dehydrogenase

very-long-chain-acyl-CoA dehydrogenaselong-chain-acyl-CoA dehydrogenase

medium-chain-acyl-CoA dehydrogenaseacyl-CoA dehydrogenase

acyl-CoA oxidasefatty-acyl-CoA synthase

Acetyl-CoA

Aromatics cyclopentanone monooxygenase4-hydroxyphenylacetate 3-monooxygenase

coniferyl-aldehyde dehydrogenasecis-2,3-dihydrobiphenyl-2,3-diol dehydrogenase

cyclopentanol dehydrogenasenaphthalene 1,2-dioxygenase

o-succinylbenzoate-CoA ligasediphenols oxidoreductase (acceptor: oxygen)

3-oxoadipate CoA-transferasedihydroxy-dihydro-p-cumate dehydrogenase2,6-dioxo-6-phenylhexa-3-enoate hydrolase

3-carboxyethylcatechol 2,3-dioxygenaseterephthalate 1,2-dioxygenase

4-hydroxybenzoate octaprenyltransferase

2-chlorobenzoate 1,2-dioxygenasebenzoate 1,2-dioxygenasebiphenyl 2,3-dioxygenase

toluene dioxygenasebenzene 1,2-dioxygenase

cyclohexanone monooxygenase4-hydroxyacetophenone monooxygenase

phenylacetone monooxygenase

catechol 2,3-dioxygenasebiphenyl-2,3-diol 1,2-dioxygenase

2,4-dichlorophenol 6-monooxygenasephenol 2-monooxygenase

Phenols

Ketones

Benzenes

Benzoates

Othercarboxylic

acids

aldehyde dehydrogenase (NADP+)

4-cresol dehydrogenaseAryl-alcohol dehydrogenase (NAD+)

Anthranilate1,2-dioxygenase

Figure 3 Microbial community functional shifts in response to oil. Selected molecular functions related to hydrocarbon degradation,nutrient scavenging and response, photosynthesis, and some house-keeping genes are listed (left) along with the mean genomeequivalents per group of samples (middle) and the log2 of Preoil/Oiled and Oiled/Recovered fold changes (right). The rightmost columnindicates the GO ID of the terms. The abundance was assessed as average genome equivalents (mean copies per bacterial/archaeal cell) oneach sampling time (downwards; see legend). The triangles indicate values below the plotted range. The log2-fold-change was estimatedas the log2 of the ratio of normalized counts between pre-oiled samples (S1, S2, S3, S4) and oiled samples (A, B, C, E, F, G); and betweenoiled samples and recovered samples (I600, I606, J598, J604). P-values were estimated using a negative binomial test.

Response of benthic microbes to oilingLM Rodriguez-R et al

7

The ISME Journal

spanning a large diversity of known alkane degra-ders. Additionally, this dataset covered the diversityof the partial AlkB sequences reported by Smithet al. (2013) for the northern Gulf of Mexico, most of

which were assigned to clusters IV (73.5%) andII (20.9%). As expected, very few reads fromrecovered samples were placed in the tree, and mostplaced reads were derived from oiled or weathered

D-3317

S6-7 (EU853333) [cVI]

Alcanivorax borkumensis AP1 AlkB2 (CAE17295)

S6-1 (EU853332) [cI]

A-13753

A-21483

S14-10 (GQ145212)

B-114387

D-46790A-22278

C-78859

B-59514

S10-17 AlkB1 (EU853351) [cI]

Alcalinovorax hongdengensis A-11-3T AlkB (EU438898) [cIII]

S6-12 (EU853335) [cV]

B-41998

B-35204

Pseudomonas fluorescens CHA0 AlkB (CAB51045)

S12-2 (EU853353) [cI]

S8-3 (EU853339) [cII]

S4-6072

Gordonia sp. TF6 AlkB (AB112870)

A-122070

B-22100

A-105931

G-162111

S15-9 AlkB2 (EU853364) [cVI]

S10-17 AlkB2 (EU853350) [cIV]

S6-13 (EU853336) [cVI]

A-29995

S9-14 (EU853354)

S10-8 AlkB2 (EU853349) [cV]

Marinobact VT8 AlkB2 (ABM19918)

Nocardia farcinica IFM 10152 AlkB1 (BAD58168)

S17-4 Al (GQ145213) [cIV]

A-132156

Pseudomonas aeruginosa PAO1 AlkB1 (AAG05962)

Alcanivorax borkumensis SK2 AlkB1 (BAC98365) [cIV]

S5-10 (EU853331) [cV]

B-25138

G-36873

S15-8 (EU853362

S9-18 (EU853346) [cIII

S16-12-2 (EU853367) [cVI]

Pseudo s aeruginosa PAO1 AlkB2 (AAG04914)

S9-11 (EU853344) [cIII]

S12-4 (EU853354) [cIV]

S7-5 (EU853338) [cV]

A-15789

Alcanivorax dieselolei B-5 AlkB1 (AAT91722) [cI]

D-22801

Acinetobacter )

S16-2 (EU853365) [cI]

G-159976

S20-13 AlkB1 (EU853378) [cIII]

A-22280

Pseudomonas putida mt-2 XylM (P21395)

S4-11063

S4-4 (EU853326) [cVI]

C-7041

S9-8 (EU853343) [cIV]

A-87946

S4-22309

S2-11245

S4-23470

C-45126

S3-5250

C-21170

S3-8321

B-104580

S17-16 (EU853371) [cIV]

Marinobacter aquaeolei VT8 AlkB1 (ABM17541) [cIV]

E-7971

S4-8 (EU853327) [cII]

G-61137

S8-11 (EU853341) [cV]

S3-10 (EU853324) [cVI]

S16-2 (EU853365) [cI

S4-9 (EU853328) [cI]

S8-5 (EU853340) [cIII]

S3-5 (EU853322) [cIII]

A-122322

Alcanivorax borkumensis AP1 AlkB1 (CAC38027) [cIV]

Pseudomonas putida P1 AlkB (CAB51047) [cIV]

B-27295

A-92019

S17-8 (EU853369) [cIII]

]

Acinetobacter sp. ADP1 AlkM (CAA05333)

S19-3-2 (EU853374) [cII]

C-33435

Nocardia farcinica IFM 10152 AlkB2 (BAD59469)

S10-8 AlkB1 (EU853348) [cI]

Acinetobacter calcoaceticus EB104 AlkM (CAB51020)

S20-13 AlkB2 (EU853377) [cIV

F-76495

Alcanivorax borkumensis SK2 AlkB2 (BAC98366)

S17-15 (EU853370) [cVI]

B-3884

S4-2 (EU853325) [cI]

S19-10 (EU853375) [cV]

Burkholderia cepacia RR10 AlkB (AJ293306)

]

S14-14 (EU853359) [cVI]

S20-3 (EU853376) [cVI]

Thalassolituus oleivorans AlkB (CAD24434) [cIV]

Alcalinovorax venustensis ISO4T AlkB (AY683535) [cI]

A-11148

S18-5 (EU853372) [cVI]

+7.0

Metagenomic datasetABCD

EFGH

I600I606J598J604

Reads mapping(reads per million reads)

Pie radius

500

0.1

1,000

400

200100

600700800900

300

Tree scale(subsitutions per site)

4.01.0

S1S2S3S4

2.0 3.0

Cluster II

Cluster III

Cluster I

Cluster V

IC

luster IVC

luster V“O

S-II”

“OS

-I”

0.1

10.0

1,000

Response of benthic microbes to oilingLM Rodriguez-R et al

8

The ISME Journal

oil samples. However, an intermediate abundancewas detected in pre-oil samples (Figure 3, first row:alkane 1-monooxygenase). In fact, only cluster IIIwas undetectable in pre-oil samples, whereas allother clusters followed the general trend observedfor the entire gene abundance (cf. right panel onFigure 4 and first row on Figure 3). More impor-tantly, the reads from different oiled and pre-oilsamples were distributed across different clades,with larger concentrations in a few clades spanningthe entire tree, that is, an uneven but phylogeneti-cally unconstrained distribution. Notably, we iden-tified a cluster formed exclusively by genes fromthis study (labeled ‘OS-I’ in Figure 4) most abundantin the pre-oil samples and negatively impacted bythe oil spill.

Population successional patterns and communityrecoveryIn addition to the large differences observed incommunity composition (both taxonomic and func-tional) between pre-oil, oiled and recovered sam-ples, the microbial communities characterized inJuly 2010 differed from those in October 2010(Figures 1b and c), concurrent with a significantreduction in total petroleum hydrocarbons(Figure 1a). Examination of the taxonomic distribu-tion revealed that some populations respondedrapidly, reaching high abundances in July 2010,with large reductions in abundance by October, andbeing barely detectable in the recovered samples ofJune 2011 (Figure 3c). These populations includedmembers of the Alcanivorax, Borrelia, Spirochaeta,Micavibrio and Bacteroides genera. However, somepopulations were observed to peak in abundance inthe October 2010 samples and significantly drop inthe recovered samples, such as Hyphomonas,Treponema, Sphingopyxis and Hirschia. Most oil-associated genera did not maintain their abundancein oiled samples of July and October 2010 with thenotable exceptions of Marinobacter and Parvibacu-lum. The abundance profiles probably reflectedorganisms with different metabolic properties withrespect to oil degradation such as fast responders toeasily degradable oil constituents, organisms spe-cialized in degradation of aromatic and morerecalcitrant oil fractions, and oil degradation gen-eralists. Finally, we identified a significant increasein minimum doubling time between oiled samplesof July and October 2010 based on codon usage bias

patterns (Vieira-Silva and Rocha, 2010) (differenceof means: 3 h 9 m, P-value o10!16, two-sided t-test;Supplementary Figure S4A). The increase indoubling time observed from October 2010 to therecovered samples (June 2011) was much smallerand not statistically significant (difference of means:27 m, P-value: 0.19, two-sided t-test). Altogether,these results indicate that the community responseto the oil spill involved well-defined successionaltrends: a rapid response (from May to July 2010),with a peak growth rate in July 2010, followed by acontinued decrease in taxonomic diversity (betweenMay and October 2010) and, finally, a reduction inabundance of several known and suspected oildegraders, concomitant with the increase in abun-dance of several typical marine groups undetectableor rare in oiled samples, a large increase intaxonomic diversity and a decrease in functionaldiversity.

Discussion

The sands of the Pensacola Municipal Beachreceived repeated pulses of oil deposition for overa month, and, after about a year, oil was stilldetected in the beach sands, although it hadconcentrations below 5 mg kg!1 (Figure 1a). Thisindicates that the microbial community facedlargely a press (long-term) disturbance given thetime scale of microbial generation cycles andmigration processes (Shade et al., 2012). Pressdisturbances often result in community shifts drivenby the response traits of individual populations tothe disturbance, presumably sensitivity to toxiccompounds and hydrocarbon degradation capabil-ities in the case of oil contamination. The diversityand abundance of indigenous alkane-degraderspreceding the oil spill in the beach ecosystem, aswell as the origin of the degraders observed after thespill, was not robustly assessed in previous studiesmostly owing to the incomplete diversity recoveredin cultures of alkane-degraders and lack of completeunderstanding of their ecophysiology. The observa-tion of a large and phylogenetically unconstraineddiversity of alkB genes in the oiled and pre-oilsamples supports the hypothesis that the responseto the oil spill was primarily caused by shifts inabundances of pre-existing populations, as pre-viously observed in the deep-sea oil plume (Hazenet al., 2010). In other words, the alkB genes present

Figure 4 Phylogenetic reconstruction of AlkB protein sequences and putative sequences recovered from the metagenomes. The tree displaysreference AlkB (alkane hydrolase) proteins (text colored by clusters, following the nomenclature of Wang et al., 2010) along with variantsassembled from the metagenomes (black text). Proteins with experimental evidence of activity (from heterologous expression or geneknockouts) are indicated by þ . Reads mapping to different nodes of the tree are displayed as pie charts. The radius of the pie charts indicatesthe fraction of the metagenomes mapping to the node (expressed as reads per million, Reads mapping legend), and the different colors of theslices indicate the dataset of origin (Dataset legend). The terminal branch of the sequence A-87946 (cluster V) was shortened by 7.0 units, asindicated by a discontinuity. The right panel indicates the total abundance of each cluster averaged per group of datasets (in reads per million):Pre-oil samples (green), Oiled samples from July 2010 (mauve), Oiled samples from October 2010 (sea green) and recovered samples (olive).Reference sequences (including out-group sequence XylM from Pseudomonas putida) and clusters nomenclature (in squared parenthesis) arebased on Wang et al. (2010), but the definition of the clusters (colored backgrounds) was broaden to include all sequences in the analysis, andtwo additional clusters were defined (‘OS-I’ and ‘OS-II’).

Response of benthic microbes to oilingLM Rodriguez-R et al

9

The ISME Journal

in the oiled communities were not derived from asingle or a few recent gene alleles but, instead, alarge diversity of degraders was latent in the sandand/or surface waters seeping into the sands beforethe oil spill.

Initial responders (July 2010) included membersof the genera Alcanivorax, Borrelia, Spirochaeta,Micavibrio and Bacteroides, all members of theabundant fraction (X1% of the total community)in the oiled samples. Alcanivorax is a genus knownfor its hydrocarbonoclastic capabilities that canutilize alkanes but not aromatic hydrocarbons(Schneiker et al., 2006); the metabolic capabilitiesof the other genera in oil hydrocarbon degradationremain speculative. Interestingly, we found putativealkB genes (alkane hydrolase) in contigs classifiedas Alcanivorax, Borrelia and Bacteroides, but noevidence of arhA (polycyclic aromatic hydrocarbondioxygenase) in any of these genera, and putativenahA genes (naphthalene 1,2-dioxygenase) only inAlcanivorax. The former populations were replacedin the abundant fraction in October 2010 bymembers of the genus Treponema and theclass Alphaproteobacteria (including Hyphomonas,Sphingopyxis and Hirschia), suggesting a succes-sional dynamic as previously observed based on16S rRNA gene amplicon data (Kostka et al., 2011;Lamendella et al., 2014). Members of the Hyphomonasgenus have been reported as abundant members inconsortia degrading aromatic compounds, which aretypically more recalcitrant components of the crudeoil than alkanes and hence, more prevalent inlater post-spill stages (Maeda et al., 2009, 2010).Similarly, Sphingopyxis is known to have aromatichydrocarbon degradation capabilities (Kertesz andKawasaki, 2010) and was previously detected as adominant group in soil-derived oil-degradingconsortia amended with natural organic matter(Hassan et al., 2011).

Very few microbial groups, including members ofthe genera Marinobacter and Parvibaculum, wereconsistently enriched in the oiled samples with nonoticeable change in abundance between July andOctober 2010. Putative alkB and cyp153 (cyto-chrome P450 family) genes, associated with alkanedegradation, were identified in assembled contigsassigned to both genera, and putative nahA genes,associated with aromatic hydrocarbon degradation,were identified in contigs classified as Marinobacter.Members of the Marinobacter genus are able todegrade a large variety of aliphatic and aromatichydrocarbons (Gauthier et al., 1992). Similarly,members of the Parvibaculum genus exhibit meta-bolic capabilities for both aliphatic and aromaticdegradation (Schneiker et al., 2006; Wang et al.,2010; Lai et al., 2011). In contrast to previousanalyses based on 18S rRNA gene amplicons (Biket al., 2012), no consistent, statistically significantshifts in the taxonomic composition of the eukar-yotic fraction were detected between sampling datesor degree of oiling (Supplementary File S3).

Finally, in June 2011, Synechococcus, Pediococcusand archaeal genera including Nitrosopumilus,Cenarchaeum and Nitrosoarchaeum dominated theabundant fraction (in contrast to oiled samples), andan overall increase in the eukaryotic fraction wasobserved. Many of the former microbial groups areabundant in oligotrophic or nutrient-poor marineecosystems, indicating that they represent thesensitive fraction of the community to the oil spill,but to a large extent the community was resilient, asgenerally observed in microbial communities(Allison and Martiny, 2008). Notably, the observedsuccession process exhibited signs of communityrecovery, but the community in June 2011, 1 yearafter the oil reached the shoreline, significantlydiffered from that in May 2010, before oiling, similarto the results of previous microcosm experiments onoil amendment of beach sediment inocula (Rolinget al., 2002). The differences between the recoveredcommunity and its counterpart before the oil spillmay be due to the long-term effects of the oildisturbance (for example, establishment of newtaxa), stochastic events or other environmentalfactors such as organic matter input, nutrient inputand salinity changes. Clearly, more samples andanalyses would be required to obtain furtherinsights into the latter issue. Nonetheless, ourresults also suggest that these sensitive marinegroups could serve as indicator species of oilpresence and toxicity in future oil spill studies,and thus, potentially provide useful information forguiding bioremediation efforts and decisions by sitemanagers.

In general, microbial communities changed bothtaxonomically and functionally after exposure to arange of petroleum hydrocarbon concentrations. Thecommunity shifts caused a decrease in taxonomicdiversity during May to October 2010, with asignificant recovery by June 2011 (SupplementaryFigure S3C). Interestingly, the functional diversitywas observed to follow a contrasting trend: itincreased between May and July 2010, was main-tained between July and October 2010, and signifi-cantly decreased in June 2011 (SupplementaryFigure S3A-B). We hypothesize that several oligo-trophic (specialized) taxa were strongly outcom-peted upon deposition of oil onshore. Growth arrestdue to limited hydrocarbon degradation capabilitiesand/or sensitivity to toxic compounds from thecontinued presence of oil onshore would impactmore severely oligotrophic and/or specialist thancopiotrophic and/or generalist populations. Hence,a significant reduction in taxonomic diversity butnot functional diversity was expected, as observedin these communities. Moreover, we providedevidence indicating that specific fast-growing organ-isms (typically assumed to be copiotrophic) thrivedin the presence of relatively high concentrations ofpetroleum hydrocarbons. In other words, the dis-turbance favored generalist organisms in the com-munities, and the post-disturbance communities

Response of benthic microbes to oilingLM Rodriguez-R et al

10

The ISME Journal

were characterized by a narrower set of morespecialized functions. This observation seems coun-terintuitive because a common expectation is that apress disturbance would exclusively select for fewhighly specialized organisms, in this case oil-degraders. Nevertheless, this trend is predicted bythe disturbance-specialization hypothesis (Vazquezand Simberloff, 2002), is consistent with ecologicalsuccession theory and was previously observed inplant communities (for example, Bazzaz and Pickett,1980). It should be mentioned, however, that thepatterns observed here might be specific to distur-bances from crude oil and sand beach ecosystemsand not immediately generalizable to other, moreselective disturbances (Roling and van Bodegom,2014). Crude oil is composed of tens of thousands ofdifferent carbon sources that would favor generalistsin early succession, as well as toxic compounds thatwould preferentially select against specialists. Sandbeaches in the Gulf of Mexico are characterized bylow carbon content and nutrient-poor conditionsrelative to marshes or other coastal sediments(Huettel et al., 2014). This could explain the relativeabundance of putative chemolithoautotrophicarchaea (Nitrosopumilus, Cenarchaeum) in therecovered communities and suggests a suppressionof a range of organisms that are adapted to carbon-limited conditions.

In summary, the community response was pri-marily characterized by two concomitant trends.First, most of the community is selected based onthe ability to survive under disturbed conditions,that is, the response to the disturbance correlatesnegatively with the level of specialization. Second,few organisms with traits selected by the distur-bance become highly abundant, as niche opportu-nity promotes invasion (Shea and Chesson, 2002;Pintor et al., 2011). Overall, our results provideevidence of complex successional patterns in thestudied communities, involving invasion promotedby capabilities for oil hydrocarbon degradation, aswell as population survival generally hindered byspecialization and susceptibility to oil toxicity, anda general recovery of diversity, specialization andsensitive marine groups a year after the disturbance.

Conflict of Interest

The authors declare no conflicts of interest.

Acknowledgements

This work was supported in part by the U.S. NationalScience Foundation (NSF) award no. 1241046 (to KTK),OCE-1057417 and OCE-1044939 (to MH, JEK), the NSFgraduate research fellowship no. 2013172310 (to WAO),and by a grant from BP/The Gulf of Mexico ResearchInitiative to the Deep-C Consortium (#SA 12-12,GoMRI-008). We thank Patrick Chain and the personnelof the Los Alamos National Laboratory for sequencing ofthe samples.

References

Allison SD, Martiny JBH. (2008). Resistance, resilience,and redundancy in microbial communities. Proc NatlAcad Sci USA 105: 11512–11519.

Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ.(1990). Basic local alignment search tool. J Mol Biol215: 403–410.

Anders S, Huber W. (2010). Differential expression analysisfor sequence count data. Genome Biol 11: R106.

Atlas RM, Hazen TC. (2011). Oil biodegradation andbioremediation: a tale of the two worst spills in U.S.history. Environ Sci Technol 45: 6709–6715.

Bazzaz FA, Pickett STA. (1980). Physiological ecology oftropical succession: a comparative review. Annu RevEcol Syst 11: 287–310.

Bengtsson J, Eriksson KM, Hartmann M, Wang Z, Shenoy BD,Grelet G-A et al. (2011). Metaxa: a software tool forautomated detection and discrimination amongribosomal small subunit (12 S/16 S/18 S) sequences ofarchaea, bacteria, eukaryotes, mitochondria, andchloroplasts in metagenomes and environmentalsequencing datasets. Antonie Van Leeuwenhoek 100:471–475.

Berger SA, Krompass D, Stamatakis A. (2011). Perfor-mance, accuracy, and web server for evolutionaryplacement of short sequence reads under maximumlikelihood. Syst Biol 60: 291–302.

Berthe-Corti L, Nachtkamp M. (2010). Bacterial commu-nities in hydrocarbon-contaminated marine coastalenvironments. In: Timmis KN (ed), Handbook ofHydrocarbon and Lipid Microbiology. Springer: BerlinHeidelberg, pp 2349–2359.

Bik HM, Halanych KM, Sharma J, Thomas WK. (2012).Dramatic shifts in benthic microbial eukaryotecommunities following the Deepwater Horizon oilspill. PLoS One 7: e38550.

Bokulich NA, Subramanian S, Faith JJ, Gevers D,Gordon JI, Knight R et al. (2013). Quality-filteringvastly improves diversity estimates from Illuminaamplicon sequencing. Nat Methods 10: 57–59.

Chao A. (1984). Nonparametric estimation of the numberof classes in a population. Scand J Stat 11: 265–270.

Chao A, Shen T-J. (2003). Nonparametric estimation ofShannon’s index of diversity when there are unseenspecies in sample. Environ Ecol Stat 10: 429–443.

Dethlefsen L, Schmidt TM. (2007). Performance of thetranslational apparatus varies with the ecologicalstrategies of bacteria. J Bacteriol 189: 3237–3245.

Dupont CL, Rusch DB, Yooseph S, Lombardo M-J, Richter RA,Valas R et al. (2012). Genomic insights to SAR86, anabundant and uncultivated marine bacterial lineage.ISME J 6: 1186–1199.

Edgar RC. (2004). MUSCLE: multiple sequence alignmentwith high accuracy and high throughput. NucleicAcids Res 32: 1792–1797.

Edgar RC. (2010). Search and clustering orders of magni-tude faster than BLAST. Bioinformatics 26: 2460–2461.

Gauthier MJ, Lafay B, Christen R, Fernandez L, Acquaviva M,Bonin P et al. (1992). Marinobacter hydrocarbonoclas-ticus gen. nov., sp. nov., a new, extremely halotolerant,hydrocarbon-degrading marine bacterium. Int J SystBacteriol 42: 568–576.

Greer CW. (2010). Bacterial diversity in hydrocarbon-polluted rivers, estuaries and sediments. In: Timmis KN(ed), Handbook of Hydrocarbon and Lipid Microbiology.Springer: Berlin Heidelberg, pp 2329–2338.

Response of benthic microbes to oilingLM Rodriguez-R et al

11

The ISME Journal

Hassan A, Taqi Z, Obuekwe C, Al-Saleh E. (2011).Characterization of crude oil-degrading bacteria in acrude oil-contaminated and uncontaminated site inKuwait. WIT Trans Ecol Envir 167: 177–186.

Hausser J, Strimer K. (2013). Entropy: estimation ofentropy, mutual information and related quantities.http://cran.r-project.org/web/packages/entropy/index.html. Accessed August 16, 2014.

Hayworth JS, Clement TP, Valentine JF. (2011). DeepwaterHorizon oil spill impacts on Alabama beaches. HydrolEarth Syst Sci 15: 3639–3649.

Hazen TC, Dubinsky EA, DeSantis TZ, Andersen GL,Piceno YM, Singh N et al. (2010). Deep-sea oil plumeenriches indigenous oil-degrading bacteria. Science330: 204–208.

Head IM, Jones DM, Roling WFM. (2006). Marine micro-organisms make a meal of oil. Nat Rev Microbiol 4:173–182.

Huettel M, Berg P, Kostka JE. (2014). Benthic exchange andbiogeochemical cycling in permeable sediments.Annu Rev Mar Sci 6: 23–51.

Joye SB, Teske AP, Kostka JE. (2014). Microbial dynamicsfollowing the macondo oil well blowout across Gulf ofMexico environments. BioScience 64: 766–777.

Kent WJ. (2002). BLAT–the BLAST-like alignment tool.Genome Res 12: 656–664.

Kertesz MA, Kawasaki A. (2010). Hydrocarbon-degradingSphingomonads: Sphingomonas, Sphingobium, Novo-sphingobium, and Sphingopyxis. In: Timmis KN (ed),Handbook of Hydrocarbon and Lipid Microbiology.Springer: Berlin Heidelberg, pp 1693–1705.

King GM, Kostka JE, Hazen TC, Sobecky PA. (2015).Microbial responses to the Deepwater Horizon oil spillin the northern Gulf of Mexico: from coastal wetlandsto the deep sea. Annu Rev Mar Sci 7: 377–401.

Konstantinidis KT, Tiedje JM. (2005). Genomic insightsthat advance the species definition for prokaryotes.Proc Natl Acad Sci USA 102: 2567–2572.

Kostka JE, Prakash O, Overholt WA, Green SJ,Freyer G, Canion A et al. (2011). Hydrocarbon-degrading bacteria and the bacterial communityresponse in Gulf of Mexico beach sands impacted bythe Deepwater Horizon oil spill. Appl Environ Microbiol77: 7962–7974.

Kostka JE, Teske A, Joye SB, Head IM. (2014). Themetabolic pathways and environmental controls ofhydrocarbon biodegradation in marine ecosystems.Front Microbiol 5: 471.

Lai Q, Wang L, Liu Y, Yuan J, Sun F, Shao Z. (2011).Parvibaculum indicum sp. nov., isolated fromdeep-sea water. Int J Syst Evol Microbiol 61: 271–274.

Lamendella R, Strutt S, Borglin SE, Chakraborty R, Tas N,Mason OU et al. (2014). Assessment of the DeepwaterHorizon oil spill impact on Gulf coast microbialcommunities. Aquat Microbiol 5: 130.

Lubchenco J, McNutt M, Lehr B, Sogge M, Miller M,Hammond S et al. (2010). Deepwater Horizon/BP OilBudget: What happened to the oil? http://www.noaanews.noaa.gov/stories2010/PDFs/OilBudget_description_%2083final.pdf.

Luo C, Rodriguez-R LM, Konstantinidis KT. (2014). MyTaxa:an advanced taxonomic classifier for genomic andmetagenomic sequences. Nucleic Acids Res 42: e73–e73.

Lu Z, Deng Y, Van Nostrand JD, He Z, Voordeckers J, Zhou Aet al. (2012). Microbial gene functions enriched inthe Deepwater Horizon deep-sea oil plume. ISME J 6:451–460.

Maeda R, Ito Y, Iwata K, Omori T. (2010). Comparison ofmarine and terrestrial carbazole-degrading bacteria.Curr Res Technol Educ Top Appl Microbiol MicrobBiotechnol 2: 1311–1321.

Maeda R, Nagashima H, Widada J, Iwata K, Omori T.(2009). Novel marine carbazole-degrading bacteria.FEMS Microbiol Lett 292: 203–209.

Matsen FA, Kodner RB, Armbrust EV. (2010). pplacer:linear time maximum-likelihood and Bayesianphylogenetic placement of sequences onto a fixedreference tree. BMC Bioinformatics 11: 538.

McGenity TJ, Folwell BD, McKew BA, Sanni GO. (2012).Marine crude-oil biodegradation: a central role forinterspecies interactions. Aquat Biosyst 8: 10.

Michel J, Owens EH, Zengel S, Graham A, Nixon Z, Allard Tet al. (2013). Extent and degree of shoreline oiling:Deepwater Horizon Oil Spill, Gulf of Mexico, USAChin, W-C (ed). PLoS One 8: e65087.

Newton RJ, Huse SM, Morrison HG, Peake CS, Sogin ML,McLellan SL. (2013). Shifts in the microbial commu-nity composition of Gulf Coast beaches followingbeach oiling. PLoS One 8: e74265.

Pintor LM, Brown JS, Vincent TL. (2011). Evolutionarygame theory as a framework for studying biologicalinvasions. Am Nat 177: 410–423.

Poretsky R, Rodriguez-R LM, Luo C, Tsementzi D,Konstantinidis KT. (2014). Strengths and limitationsof 16 S rRNA gene amplicon sequencing in revealingtemporal microbial community dynamics. PLoS One9: e93827.

Prosser JI, Bohannan BJM, Curtis TP, Ellis RJ, Firestone MK,Freckleton RP et al. (2007). The role of ecological theoryin microbial ecology. Nat Rev Microbiol 5: 384–392.

Reed HE, Martiny JBH. (2007). Testing the functionalsignificance of microbial composition in naturalcommunities. FEMS Microbiol Ecol 62: 161–170.

Rho M, Tang H, Ye Y. (2010). FragGeneScan: predictinggenes in short and error-prone reads. Nucleic AcidsRes 38: e191.

Rodriguez-R LM, Konstantinidis KT. (2014a). Estimatingcoverage in metagenomic data sets and why it matters.ISME J 8: 2349–2351.

Rodriguez-R LM, Konstantinidis KT. (2014b). Nonpareil:a redundancy-based approach to assess the level ofcoverage in metagenomic datasets. Bioinformatics 30:629–635.

Roling WFM, van Bodegom PM. (2014). Toward quantita-tive understanding on microbial community structureand functioning: a modeling-centered approach usingdegradation of marine oil spills as example. AquatMicrobiol 5: 125.

Roling WFM, Milner MG, Jones DM, Fratepietro F,Swannell RPJ, Daniel F et al. (2004). Bacterialcommunity dynamics and hydrocarbon degradationduring a field-scale evaluation of bioremediationon a mudflat beach contaminated with buried oil.Appl Environ Microbiol 70: 2603–2613.

Roling WFM, Milner MG, Jones DM, Lee K, Daniel F,Swannell RJP et al. (2002). Robust hydrocarbondegradation and dynamics of bacterial communitiesduring nutrient-enhanced oil spill bioremediation.Appl Environ Microbiol 68: 5537–5548.

Schneiker S, dos Santos VAM, Bartels D, Bekel T, Brecht M,Buhrmester J et al. (2006). Genome sequence of theubiquitous hydrocarbon-degrading marine bacteriumAlcanivorax borkumensis. Nat Biotechnol 24:997–1004.

Response of benthic microbes to oilingLM Rodriguez-R et al

12

The ISME Journal

Shade A, Peter H, Allison SD, Baho DL, Berga M,Burgmann H et al. (2012). Fundamentals of microbialcommunity resistance and resilience. Front Microbiol3: 417.

Shea K, Chesson P. (2002). Community ecology theory as aframework for biological invasions. Trends Ecol Evol17: 170–176.

Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W et al.(2011). Fast, scalable generation of high-qualityprotein multiple sequence alignments using ClustalOmega. Mol Syst Biol 7: 539.

Smith CB, Tolar BB, Hollibaugh JT, King GM. (2013).Alkane hydroxylase gene (alkB) phylotype composi-tion and diversity in northern Gulf of Mexicobacterioplankton. Aquat Microbiol 4: 370.

Stamatakis A. (2006). RAxML-VI-HPC: maximum like-lihood-based phylogenetic analyses with thousands oftaxa and mixed models. Bioinformatics 22: 2688–2690.

Suzek BE, Huang H, McGarvey P, Mazumder R, Wu CH.(2007). UniRef: comprehensive and non-redundantUniProt reference clusters. Bioinformatics 23:1282–1288.

Valentine DL, Mezic I, Macesic S, Crnjaric-Zic N, Ivic S,Hogan PJ et al. (2012). Dynamic autoinoculation andthe microbial ecology of a deep water hydrocarbonirruption. Proc Natl Acad Sci USA 109: 20286–20291.

Vazquez DP, Simberloff D. (2002). Ecological speciali-zation and susceptibility to disturbance: conjecturesand refutations. Am Nat 159: 606–623.

Vieira-Silva S, Rocha EPC. (2010). The systemic imprint ofgrowth and its uses in ecological (meta)genomics.PLoS Genet 6: e1000808.

Wang L, Wang W, Lai Q, Shao Z. (2010). Gene diversityof CYP153A and AlkB alkane hydroxylases inoil-degrading bacteria isolated from the AtlanticOcean. Environ Microbiol 12: 1230–1242.

Wang P, Roberts TM. (2013). Distribution of surficial andburied oil contaminants across sandy beaches alongNW Florida and Alabama coasts following theDeepwater Horizon oil spill in 2010. J Coast Res 291:144–155.

Wang Q, Garrity GM, Tiedje JM, Cole JR. (2007). Naıvebayesian classifier for rapid assignment of rRNAsequences into the new bacterial taxonomy. ApplEnviron Microbiol 73: 5261–5267.

Yakimov MM, Timmis KN, Golyshin PN. (2007). Obligateoil-degrading marine bacteria. Curr Opin Biotechnol18: 257–266.

Zuijdgeest A, Huettel M. (2012). Dispersants as used inresponse to the MC252-spill lead to higher mobility ofpolycyclic aromatic hydrocarbons in oil-contaminatedGulf of Mexico sand. PLoS One 7: e50549.

Supplementary Information accompanies this paper on The ISME Journal website (http://www.nature.com/ismej)

Response of benthic microbes to oilingLM Rodriguez-R et al

13

The ISME Journal