AQUATIC MICROBIAL ECOLOGYAquat Microb Ecol
Vol. 71: 141–153, 2013doi: 10.3354/ame01674
Published online December 16
INTRODUCTION
Microorganisms are crucial to the function of all eco -systems and biogeochemical cycles. In aquatic eco-systems, microbes typically make up the majority ofbiomass, carbon fixation, and nutrient cycling (Azamet al. 1983). Due to their extremely diverse metabolicpathways, prokaryotic microbes are major contribu-tors to many of these biogeochemical processes (New-man & Banfield 2002). Eukaryotic microbes (protists)also play major roles in decomposition, nutrient recy-
cling, and ecosystem energy flow, including carbonfixation (Paul 2007). Bacterivorous protists remineral-ize nutrients otherwise tied up in bacterial biomass,are the link between prokaryotic microbes and highertrophic levels, and are a major factor in regulatingbacterial population size and community structure inaquatic ecosystems (Sherr & Sherr 2002).
The use of molecular tools in microbial ecology hasbloomed over the last 25 yr, allowing the discovery ofdiversity, taxonomic affinity, and ecology of manyunculturable or morphologically indistinct prokary-
© Inter-Research 2013 · www.int-res.com*Corresponding author. Email: [email protected]
Linking bacterivory and phyletic diversityof protists with a marker gene survey and
experimental feeding with BrdU-labeled bacteria
Scott A. Fay1,3, Rebecca J. Gast2, Robert W. Sanders1,*
1Temple University, Biology Department, Philadelphia, Pennsylvania 19122, USA2Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543, USA
3Present address: Berkeley Initiative in Global Change Biology, University of California, Berkeley, California 94720, USA
ABSTRACT: Over the last few decades, molecular methods have vastly improved our ability tostudy the diversity of microbial communities. In molecular diversity surveys, the function of protistsis often inferred from phylogeny. Yet these surveys are unable to distinguish between differenttrophic modes among closely related taxa. Here we present results from a culture-independentstudy linking bacterivory to the diversity of pelagic protists from 3 depths of a stratified mesotrophiclake. Bacteria were labeled with bromodeoxyuridine (BrdU) and added to lakewater samples; afterincubation, total DNA was extracted from filtered samples. Part of the DNA extract was subjected toimmunoprecipitation with anti-BrdU antibodies, and then both whole DNA and BrdU-labeled sam-ples were analyzed using 454-pyrosequencing of the v9 region of 18S small subunit rRNA gene am-plicons. The results show that a different community of protists exists at each depth, with limitedoverlap of taxonomic composition between depths. The community of BrdU-labeled protists,deemed putative bacterivores, is largely a subset of the community found in the whole DNA sam-ples. Many of these BrdU-labeled taxa are poorly represented in GenBank and thus are probablyrarely isolated and/or uncultured species. Several of the taxa identified as bacterivores are alsophototrophs, highlighting the important role of mixotrophy among eukaryotic microbes. Definitiveidentity of functional traits among taxa requires careful experimentation, yet this method allows afirst-pass assay of the trophic role of microbial eukaryotes from environmental samples.
KEY WORDS: Molecular methods · Microbial community · Mixotrophy · Bromodeoxyuridine · Culture-independent · Eukaryotic microbes · Pyrosequencing · Lake microbes
Resale or republication not permitted without written consent of the publisher
FREEREE ACCESSCCESS
Aquat Microb Ecol 71: 141–153, 2013
otes and protists (Olsen et al. 1986). Pyrosequencinghas further improved our ability to rapidly character-ize whole microbial communities and identify rarelineages (Sogin et al. 2006). Increasingly thesemarker gene surveys, such as amplicon pyrose-quencing of a community’s rRNA genes, have beenapplied to examine eukaryotic microbial communi-ties (Stoeck et al. 2010, Bik et al. 2012). Althoughamplicon pyrosequencing has limited utility in deter-mining relative abundance between species becauseof gene copy number variation and non-linear PCRamplification (Medinger et al. 2010), next-generationamplicon sequencing is effective for rapidly analyz-ing microbial taxon richness and community struc-ture from environmental samples. However, the abil-ity of marker gene surveys to discern the function ofa particular microbial taxon typically depends onlyon the taxon’s phylogenetic affinity with well-studiedorganisms, i.e. those in culture.
Culture-independent methods such as stable iso-tope probing can be combined with marker gene sur-veys to more directly infer the function of recoveredtaxa (Gutierrez-Zamora & Manefield 2010). Anotherrelated method uses bromodeoxyuridine (BrdU), athymidine analog commonly applied in molecularand cell biology studies of cell proliferation, whichbecomes incorporated as a cell undergoes DNA syn-thesis. In microbial ecology, BrdU has been used toidentify metabolically active microbes and determinephylotype-specific growth rates from environmentalsamples (Urbach et al. 1999). In these methods, BrdUis added to environmental samples, where the com-pound is taken up by active microbes and incorpo-rated into their DNA. By immunoprecipitation (IP),BrdU-labeled DNA is separated from a subsample ofthe whole DNA extract, preparing for further analy-sis by standard molecular ecology techniques. Thismethod has also been used to study bacterivory:organisms that ingest bacteria labeled with BrdUthemselves become labeled (Randa 2007).
We coupled the ability to trace the transfer ofBrdU-labeled bacterial DNA into bacterivores (Randa2007) with environmental marker gene survey pyro -sequencing to investigate protistan bacterivorediversity at 3 distinct depth layers (epilimnion, met-alimnion, and hypolimnion) in a thermally stratifiedlake. We compared these data with overall microbialeukaryote diversity at the same depths. The com-munities of bacterivorous protists were less diverse,but largely overlapped with the whole communityof protists in the thermally stratified lake, with thegreatest diversity of both groups in the mid-depthmetalimnion.
MATERIALS AND METHODS
Preparation of BrdU-labeled bacteria
Cultures of Pasteurella sp. (isolated from Ice HousePond, Massachusetts, USA, size 0.6−0.7 × 1.2 µm)and Planococcus sp. (isolated from Barnegat Bay,New Jersey, USA, diameter ≤1 µm) were maintainedin 1% yeast extract (YE) solution and 0.1% YE solu-tion in 32 psu artificial sea water, respectively (Kempet al. 1993). For BrdU-labeling experiments, 5 ml ofYE supplemented with 20 µM 5-bromo-2’-deoxyuri-dine (BrdU, Sigma B5002) were inoculated with bac-teria and incubated for 36 h at room temperature (RT)in the dark. BrdU-labeled bacterial stocks were enu-merated using epifluorescence microscopy fromsamples collected onto black 0.2 µm polycarbonatefilters (Millipore GTBP02500) and stained with DAPI-Vectashield (Vector Labs H-1200). BrdU-labeledPlanococcus sp. bacteria were used for assessment ofIP efficiency as described below. The BrdU-labeledPasteurella sp. were prepared for environmentallakewater experiments and were harvested by cen-trifugation (3000 × g, 10 min), washed 3 times, anddispersed and resuspended by pipetting with coldphosphate-buffered saline.
DNA extraction
A hot detergent, bead homogenization protocoladapted by Gast et al. (2004) from Kuske et al. (1998)was used for all DNA extractions. Bacterial cultureswere harvested by centrifugation, and DNA wasdirectly extracted from cell pellets. Organisms inenvironmental samples and the field experimentwere collected by vacuum filtration on 47 mm poly-carbonate 0.8 µm (Millipore ATTP04700) filters.
Dot blots to determine positive labeling of bacteria
Before use, bacterial strains were tested to verifyincorporation of BrdU using dot blots in a protocoladapted from Ueda et al. (2005). For each sample,10 µl of 25 ng µl−1 genomic DNA was denatured byincubation with 40 µl of 0.4N NaOH for 30 min at RT.Samples were then placed on ice to prevent anneal-ing and neutralized with 50 µl cold 2M ammoniumacetate. Using a microfiltration device (Bio-Rad 170-6545), samples were dot-blotted onto nitrocellulosemembrane pre-wetted with 6× saline sodium citratebuffer (SSC, 0.9M NaCl, 90 mM sodium citrate, pH
142
Fay et al.: Culture-independent identification of bacterivores
7.0). Each well was then rinsed once with 2× SSC, themembrane was removed, rinsed again with 2× SSC,and baked at 80°C under vacuum for 2 h. After block-ing for 30 min at RT with 1% dry nonfat milk in TBS-T (50 mM Tris, 150 mM NaCl, 0.05% Tween 20, pH7.6), membranes were probed by adding monoclonalanti-BrdU mouse IgG (Sigma B-8434) at a 1:5000dilution, washed 3× 10 min with TBS-T, incubated1 h with anti-mouse HRP-conjugated IgG (Cell Sig-naling 7076), and then washed 3× 20 min with TBS-T. Blots were visualized with ECL Plus (Pierce 32132)and X-OMAT LS film (Kodak 868-9358).
Immunoprecipitation
IP of BrdU-labeled bacterial DNA was performedfollowing the protocol published by Urbach et al.(1999). One modification was the use of unlabeledbacterial genomic DNA from Pasteurella sp. (insteadof eukaryotic salmon sperm DNA, as used in theUrbach protocol, Urbach et al. 1999) in the blockingstep. This was to prevent contamination of the sam-ples with exogenous eukaryotic (salmon) DNA thatwould subsequently amplify with eukaryotic riboso-mal gene primers.
qPCR assessment of IP efficiency
SYBR-based quantitative PCR (qPCR) was used toassess the efficiency of the IP protocol. DNA extractsfrom Planococcus sp., labeled and unlabeled withBrdU, underwent IP as above. Primer3 (Rozen &Skaletsky 2000) was used to design quantitative PCRprimers for the 16S rRNA gene of Planococcus bacte-ria, Plano16SqPCRf, 5’-GTG TGT AGC CCA GGTCAT AAG G-3’ and Plano16SqPCRr, 5’-GAT CTTAGT TGC CAG CAT TCA GT-3’. Unknown sampleand standard curve (using purified genomic Pla no -coccus DNA) reactions were done in triplicate.
qPCR assessment of uptake of BrdU by protistsfrom bacteria
Cultures of the bacterivorous protist Paraphyso -mo nas sp. (Macaluso et al. 2009) were fed BrdU-labeled Pasteurella bacteria over a time course toexamine incorporation of BrdU into the protist DNA.Uptake was determined by qPCR of Paraphyso -monas DNA after immunoprecipitating whole DNAextracts from samples after 0, 6, 12, and 24 h. Para-
physomonas-specific primers were designed asabove (Paraphyso _18S_qF: 5’-GCC TGC GGC TTAATT TGA CT-3’, and Paraphyso_18S_qR: 5’-CAACTA AGA ACG GCC ATG CA-3’), and reactionswere done in triplicate.
Field site and bacterial feeding
Our field site was Lake Lacawac (41° 22.912’ N,75° 17.543’ W) in the Pocono Mountains, Pennsylva-nia, USA. Lacawac is a 13 000 yr old glacial lakeformed by ice scour with a maximum depth of 13 m,with surface chlorophyll ranging from 2 to 5 µg l−1.This 21 ha mesotrophic lake and its watershedremain undeveloped and are protected by theLacawac Sanctuary Foundation. The annual phyto-plankton community in Lake Lacawac is dominatedby chrysophycean algae (Siver & Chock 1986), attimes including known mixotrophs (bacterivorousalgae) like the colonial flagellates Dinobryon andUroglena. Small mixotrophic flagellated nanophyto-plankton are the dominant bacterivores in the lakeduring winter (Berninger et al. 1992), and occur inthe lake during other seasons (R.W. Sanders pers.obs.).
On 1 October 2010, the water column oxygen andtemperature profiles in the lake were characterizedusing a dissolved oxygen−temperature meter (YSIModel 85) to identify the depths of the mixed surfacelayer (epilimnion), the thermocline (metalimnion),and the deeper, cool hypolimnion (Fig. 1). Near dusk,
143
[O2] (mg–1) or temperature (Cº)
0
2
4
6
8
10
Meta
Epi
0 5 10 15 20
Hypo
Dep
th (m
)
Fig. 1. Depth profile of Lake Lacawac (Pennsylvania, USA)taken on 1 October 2010, indicating thermal and chemical(O2) stratification. Open squares represent dissolved oxygen(mg l−1), and circles represent water temperature (°C). Threesampling depths are indicated: Epi: epilimnion, Meta:
metalimnion, Hypo: hypolimnion
Aquat Microb Ecol 71: 141–153, 2013
samples were taken from these distinct layers of thethermally stratified water column using a Van Dornbottle, and were gently prefiltered through a 250 µmNitex mesh into incubation flasks to remove thelarger crustaceans that prey on the protist commu-nity. BrdU-labeled bacteria at a final concentration of1 × 106 cells ml−1 (~30% of the typical native bacterialabundance) were added to the 500 ml samples, andthe closed flasks were immediately returned to theirrespective sampled depths attached to a moored ropeand incubated overnight and into the morning for atotal of 16 h. Light levels were determined just priorto sample retrieval with a Li-Cor LI-250A light meterand LI-193 spherical quantum sensor; the lightextinction coefficient (η) was 0.73 m−1. Recoveredflasks were kept on ice in the dark until filtered,within 2 h of recovery.
From each incubation flask, 250 ml were filteredonto a 0.8 µm polycarbonate filter, DNA was isolated,and a 2 µg subsample of DNA from each whole DNAsample was used for BrdU-IP as described above.Thus 6 DNA samples were obtained for downstreamanalyses: ‘whole DNA’ (i.e. not treated by BrdU-IP)and ‘BrdU-IP’ from each of the 3 depths: epilimnion,metalimnion, and hypolimnion (Fig. 1).
Denaturing gel gradient electrophoresis (DGGE)
Before proceeding with the pyrosequencing run,DGGE was used to examine how the BrdU-IP com-munity profiles compared to those from the wholeDNA sample. Protocol was followed as published byGast et al. (2004).
Primers, PCR amplicon preparation, and pyrosequencing
We incorporated forward PCR primer 1391F (Stoecket al. 2010) and reverse primer 1510R (Amaral-Zettleret al. 2009) into fusion primers with Lib-L 1-way readsequencing adaptors and keys as described by 454Life Sciences Corp. (2010). Forward Primer A 1391Fincorporated 6 unique 10 bp Multiplex IDs (MIDs) todistinguish the 6 different amplicon libraries. Wechose these primers because, among all of the possi-ble pairs of small subunit (SSU) rRNA V9 regioneukaryotic universal primers reported by Amaral-Zettler et al. (2009) and Stoeck et al. (2010), this pair,1391F (5’-CCA TCT CAT CCC TGC GTG TCT CCGAC-TCA G-[MID]-GTA CAC ACC GCC CGT C-3’)and 1510R (5’-CCT ATC CCC TGT GTG CCT TGG
CAG TC-TCA G-CCT TCY GCA GGT TCA CCTAC-3’), has the greatest coverage against a UCLUST70% clustered subset (consisting of 3164 centroids)of the SILVA eukaryotic SSU rRNA database.
For each sample, we performed 3 PCR reactions,each at a volume of 50 µl that contained 0.2 µM ofeach primer, 1× reaction buffer, 200 µM each dNTP,and 0.5U Phusion DNA Polymerase (New EnglandBiolabs F-553). Cycling conditions were as follows:98°C for 120 s; 10 cycles of 98°C for 15 s, 67°C decre-mented by 1°C cycle−1 for 20 s, and 72°C for 15 s; 25cycles of 98°C for 15 s, 57°C for 20 s, and 72°C for15 s; and a final extension step of 72°C for 120 s. Eachsample was purified with AMPure XP beads (Beck-man Coulter A63880) following the manufacturer’sprotocol. The University of Pennsylvania DNASequencing Facility performed the sequencing reac-tion using a ‘1-way read’ approach to ampliconpyrosequencing with the GS Junior Titanium emPCRLib-L kit, as recommended by 454 Life Sciences(2010). The resulting read data and our associatedanalytical results were deposited in the NationalCenter for Biotechnology Information (NCBI)Sequence Read Archive (SRA) under accession num-ber SRA048528.
Pyrosequencing marker gene survey data analysis
We removed barcodes, primers, and low-qualitysequences and de-multiplexed the data using QIIME(Caporaso et al. 2010). Sequences were de-noisedusing the algorithm of Reeder & Knight (2010) imple-mented in QIIME. Chimeras were removed by denovo detection using UCHIME, and operational tax-onomic units (OTUs) were clustered at 95% identityusing UCLUST, both implemented in USEARCH(Edgar 2010). It is critical to implement an analyticalpipeline that filters out pseudodiversity, especiallyfrom PCR chimeras (Behnke et al. 2011). A second denovo check for chimeras was performed after thefinal list of OTUs was established. Many of thechimeras may have been removed earlier in the pro-cess of clustering the OTUs at 95% identity andremoving taxa that did not identify as ‘eukaryote’against the Ribosomal Database Project (RDP) data-base. Linux bash shell scripts were written whennecessary to prepare data for different applications.QIIME was used to generate rarefaction plots, andOTU tables were exported to PRIMER 6 (Clarke &Warwick 2001) for analysis of community similarityusing a variety of metrics. To incorporate phyloge-netic signal in our analyses, we used UniFrac
144
Fay et al.: Culture-independent identification of bacterivores
(Lozupone et al. 2007) implemented in QIIME. OTUswere assigned as putative bacterivores if >70% ofthe total sequence reads for a given OTU were foundin the BrdU-IP treatments; only those OTUs with>0.05% sequence read abundance, i.e. >38 readstotal, were considered.
RESULTS AND DISCUSSION
Effectiveness of labeling and precipitation
Not all strains of bacteria will take up and incorpo-rate thymidine or its analog BrdU into DNA(Hamasaki et al. 2007). Consequently, it is importantto test experimental bacteria to ensure that sufficientlabeling will occur. Anti-BrdU dot blots demon-strated that DNA of both Pasteurella and Planococ-cus was effectively labeled when grown in mediasupplemented with BrdU.
Quantitative PCR was used to verify that IP effec-tively isolated BrdU-labeled DNA. IP of BrdU-labeledDNA from Planococcus bacteria yielded 13.0 ± 0.7 ngµl−1 (95% CI) of Planococcus DNA, while IP of unla-beled control DNA from Planococcus bacteria yieldedonly 53 ± 6.6 pg µl−1 (95% CI) This represents a 245-fold enrichment of labeled over unlabeled DNA. Thetime course of a bacterivorous protist, Paraphyso -monas, feeding on BrdU-labeled bacteria demon-strated that these protists take up BrdU from theirfood (Fig. 2). In laboratory experiments, Randa (2007)also found that a marine strain of Paraphysomonasand a marine ciliate, Uronema sp., ingested BrdU- labeled bacteria and incorporated the label into theirgenomic DNA in laboratory experiments.
Analysis of the DGGE gels showed that the BrdU-IP samples in our field experiment had fewer totalbands than the whole DNA samples. This wasexpected since the putative bacterivores (BrdU-labeled taxa, BrdU-IP) should be a subset of thewhole community of microbial eukaryotes (wholeDNA untreated with IP). While some of the DGGEbands appearing in the BrdU-IP samples wereunique, most were shared with the whole DNA sam-ples. In these instances of shared band presence, theBrdU-IP bands were often brighter, suggestinggreater representation in that recovered DNA pool.
Diversity of protists
After initial quality filtering, which removed 12.6%of sequences, clustering OTUs at 95% identity
yielded 581 OTUs. Ten of these (1.7%) were identi-fied as chimeras and removed. After assigning taxon-omy using the RDP database, we identified a total of312 protist OTUs in the combined data set. Rarefac-tion curves from the pyrosequencing data showedthat taxonomic sampling was more complete for theBrdU-IP samples than for the whole DNA samples atall depths (Fig. 3). This confirmed the expectationthat the protists actively grazing bacteria repre-sented a subset of the total microbial community. Themajority of the OTUs recovered from each of the‘BrdU-IP’ samples were also found in their respective‘whole DNA’ samples (Fig. 4A), including all of theOTUs with high sequence abundance (>0.1% of thetotal) from the BrdU-IP samples (Table 1). The rar-efaction curves (Fig. 3) also predict that the rankingof diversity for both BrdU-IP and whole DNA sampleswill remain the same with deeper sequencing.
The greatest diversity was found in the metal-imnion and least diversity in the hypolimnion (Figs. 3& 4A). Studies of protistan communities in a stratifiedanoxic fjord also found increased diversity at anintermediate depth within the pycnocline wheredensity and chemistry was changing rapidly with
145
75
50
25
0
0 6 12 24
Time (h)
DN
A (n
g)
Fig. 2. Quantitative PCR results of bromodeoxyuridine(BrdU)-immunoprecipitated DNA from Paraphysomonascultures fed BrdU-labeled bacteria, with primers specific toParaphysomonas 18S rRNA sequence. Error bars represent
1 SE of the mean
Aquat Microb Ecol 71: 141–153, 2013
depth (Behnke et al. 2006). The metalimnion is alayer of high biomass and fine-scale habitat differen-tiation, and is therefore expected to have high biodi-versity (Finlay et al. 1997). Additionally, suspendedorganic macroaggregates that are numericallyenriched in both heterotrophic and phototrophic pro-tists are common in the metalimnion of this lake(Caron 1991) and offer a different habitat from thesurrounding water. The epilimnion is a mixed sur-face layer, without the gradient of nutrients, temper-ature, and oxygen typical of the metalimnion, andthus may have reduced range of habitat type. Find-ing the lowest diversity in the hypolimnion wasexpected due to its more ‘extreme’ environmentalconditions with low oxygen (Fig. 1) and higherhydrogen sulfide as evidenced by its distinct odor(S.A. Fay pers. obs.). Photosynthetic protists alsowould be more rare in the low light hypolimnion(1.0 µmol s−1 m−2). Despite the lower diversity in thehypolimnion, there was considerable overlap ofOTUs identified from the layers in both BrdU-IP andwhole DNA samples (Fig. 4B); <7% and <21% of theBrdU-IP and whole DNA OTUs, respectively, wereunique to the hypolimnion. The proportion of OTUsshared between layers was always greater for thewhole DNA than for BrdU-IP samples, regardless ofwhich layers were compared (Fig. 4B). This likelyreflects DNA from metabolically inactive cells orcysts that could sink through the water column thatwould have been identified in whole DNA samples,but not in samples labeled by feeding on BrdU-labeled bacteria.
Community composition
An analysis of community composition differentiat-ing between the 6 samples using non-metric multi- dimensional scaling (NMDS) shows that the commu-nities of eukaryotic microbes in the BrdU-IP samplesare more similar to the communities of their respectivewhole DNA samples than they are to each other(Fig. 5). This pattern holds true using a variety of beta-diversity distance metrics available in PRIMER 6, including ones that disregard relative abundances(Jaccard), incorporate joint absences (Euclidean), andincorporate phylogenetic signal (Unifrac, both weightedand unweighted for relative abundances). The relativesequence abundance of higher taxonomic categoriesamong the samples (Fig. 6) is consistent with theNMDS analyses. The finding of different protist com-munities at different depths is not unique to Lake Lacawac and in fact is likely a widespread phenome-
146
01000
20003000
40005000
60007000
80009000
100000
50
100
150
200
EpiBrdU
EpiWholeHypoBrduHypoWholeMetaBrdUMetaWhole
Sequences per sample
No.
of O
TUs
sam
ple
d
Fig. 3. Rarefaction curves of the 6 samples (mean of 20 valuesper step), showing the diversity of operational taxonomicunits (OTUs). Data were sampled at 200 sequence incrementsup to 104 sequences. EpiBrdU: epilimnion bromodeoxyuri-dine immunoprecipitated (BrdU-IP), EpiWhole: epilimnionwhole DNA, HypoBrdU: hypolimnion BrdU-IP, HypoWhole:hypolimnion whole DNA, MetaBrdU: metalimnion BrdU-IP,
MetaWhole: metalimnion whole DNA
Total
130
23
159
Whole DNA
BrdU-IP
Hypolimnion
106
9
35
Metalimnion
115
21
104
Epilimnion
72
14
72
Brdu-IP by depth
35
68
18
35 4
10
12
Whole DNA by depth
37
70
37
505
47
52
Epilimnion
Hypolimnion
Metalimnion
Epilimnion
HypolimnionMetalimnion
Whole DNA
BrdU-IP
Whole DNA
BrdU-IP
Whole DNA
BrdU-IP
B
A
Fig. 4. Overlap in recovered operational taxonomic units(OTUs) between samples. (A) Comparison of OTUs identi-fied from whole DNA samples and bromodeoxyuridine im-munoprecipitated (BrdU-IP) samples for each depth sam-pled and for the combined lake community (total). (B)Overlap OTUs identified from all depths for BrdU and whole
DNA samples
Fay et al.: Culture-independent identification of bacterivores 147
No.
of
seq
uen
ces
reco
vere
dC
lose
st G
enB
ank
BL
AS
TR
DP
con
sen
sus
lin
eag
eE
pi
Hyp
oM
eta
Su
mB
LA
ST
hit
wit
hid
enti
tyB
rdU
-IP
Wh
ole
Brd
U-I
PW
hol
eB
rdU
-IP
Wh
ole
Brd
U-I
PW
hol
e>
85%
id
enti
ty(%
)
365
1831
1011
014
612
8119
8011
456
Cry
pto
mon
as10
0E
uk
aryo
ta24
5249
3611
712
326
173
528
3057
94C
hry
sosp
hae
rell
a99
Eu
kar
yota
; str
amen
opil
es; S
ynu
rop
hyc
eae;
Syn
ura
les;
Syn
ura
ceae
58
961
2535
1052
976
2595
Pel
agot
hri
x10
0E
uk
aryo
ta; A
lveo
lata
; Cil
iop
hor
a; I
ntr
amac
ron
ucl
eata
; Pro
stom
atea
1196
212
110
791
1561
1988
1783
Syn
ura
93E
uk
aryo
ta; s
tram
enop
iles
02
166
1365
1832
184
1399
Cry
pto
mon
as98
Eu
kar
yota
65
066
2771
533
786
Un
cult
. alv
eola
te10
0E
uk
aryo
ta; A
lveo
lata
01
309
664
814
317
679
Mon
osig
a93
Eu
kar
yota
00
00
959
539
959
539
Uro
gle
na
93E
uk
aryo
ta; s
tram
enop
iles
532
04
291
454
296
490
Un
cult
. cil
iate
98E
uk
aryo
ta0
010
447
317
212
147
5H
isti
obal
anti
um
100
Eu
kar
yota
; Alv
eola
ta; C
ilio
ph
ora;
In
tram
acro
nu
clea
ta;
Oli
goh
ymen
oph
orea
112
205
9216
430
128
141
3U
ncu
lt. e
uk
aryo
te90
Eu
kar
yota
; Rh
izar
ia17
911
913
811
026
530
239
2O
ikom
onas
91E
uk
aryo
ta; s
tram
enop
iles
; Ch
ryso
ph
ycea
e3
825
430
672
6832
938
2C
ilia
te96
Eu
kar
yota
; Alv
eola
ta; C
ilio
ph
ora;
In
tram
acro
nu
clea
ta45
315
10
2189
920
913
5238
1G
onyo
stom
um
100
Eu
kar
yota
; str
amen
opil
es; R
aph
idop
hyc
eae;
Ch
atto
nel
lale
s;V
acu
olar
iace
ae83
5421
3966
260
170
353
Str
omb
idiu
m95
Eu
kar
yota
186
4854
287
08
240
343
Arc
uos
pat
hid
ium
99E
uk
aryo
ta; A
lveo
lata
; Cil
iop
hor
a; I
ntr
amac
ron
ucl
eata
; Lit
osto
mat
ea18
472
051
3517
621
929
9O
ikom
onas
89E
uk
aryo
ta; s
tram
enop
iles
00
174
252
232
197
254
Ath
alam
ea89
Eu
kar
yota
; Rh
izar
ia89
163
12
174
9123
9C
hro
mu
lin
a89
Eu
kar
yota
; str
amen
opil
es0
751
223
14
5223
4U
ncu
lt. e
uk
aryo
te10
0E
uk
aryo
ta; A
lveo
lata
; Cil
iop
hor
a; I
ntr
amac
ron
ucl
eata
86
03
101
218
109
227
Mon
osig
a87
Eu
kar
yota
00
147
206
175
164
211
Cry
pto
mon
as10
0E
uk
aryo
ta; C
ryp
top
hyt
a; C
ryp
tom
onad
ales
; Cry
pto
mon
adac
eae;
Cry
pto
mon
as17
918
40
11
1818
020
3U
role
ptu
s10
0E
uk
aryo
ta; A
lveo
lata
; Cil
iop
hor
a; I
ntr
amac
ron
ucl
eata
; Sp
irot
rich
ea14
217
02
322
175
464
194
Un
cult
. fre
shw
ater
100
Eu
kar
yota
euk
aryo
te0
427
8549
7976
168
Un
cult
. cry
pto
ph
yte
98E
uk
aryo
ta45
190
110
313
914
815
9U
ncu
lt. l
ake
100
Eu
kar
yota
pic
opla
nk
ton
9812
10
121
1411
913
6M
allo
mon
as10
0E
uk
aryo
ta; s
tram
enop
iles
; Syn
uro
ph
ycea
e; S
ynu
rale
s0
11
618
128
1913
5U
ncu
lt. c
ilia
te10
0E
uk
aryo
ta; A
lveo
lata
205
900
331
2423
611
7P
erid
iniu
m99
Eu
kar
yota
; Alv
eola
ta; D
inop
hyc
eae
02
00
8410
284
104
Par
aph
ysom
onas
99E
uk
aryo
ta; s
tram
enop
iles
06
10
195
210
1U
ncu
lt. K
atab
le-
100
Eu
kar
yota
ph
arid
acea
e0
042
970
142
98A
ntr
odia
81E
uk
aryo
ta0
00
026
596
265
96S
pon
gom
onas
88E
uk
aryo
ta0
037
630
2937
92S
tok
esia
92E
uk
aryo
ta; A
lveo
lata
143
270
760
528
484
Ch
lore
llac
eae
99E
uk
aryo
ta; V
irid
ipla
nta
e; C
hlo
rop
hyt
a; T
reb
ouxi
oph
ycea
e3
155
4672
2113
068
Nol
and
ia94
Eu
kar
yota
; Alv
eola
ta; C
ilio
ph
ora;
In
tram
acro
nu
clea
ta
Tab
le 1
. Op
erat
ion
al t
axon
omic
un
its
(OT
Us)
wit
h t
he
hig
hes
t to
tal n
um
ber
of
seq
uen
ces
reco
vere
d. P
roti
st t
axa
are
ord
ered
by
the
tota
l nu
mb
er o
f se
qu
ence
s re
cove
red
.In
clu
ded
are
tax
a w
ith
>0.
075
% o
f to
tal
seq
uen
ces
reco
vere
d.
Ep
i: e
pil
imn
ion
, M
eta:
met
alim
nio
n,
Hyp
o: h
ypol
imn
ion
, W
hol
e: w
hol
e D
NA
sam
ple
, B
rdU
-IP
: b
rom
odeo
xyu
rid
ine
imm
un
opre
cip
itat
ed s
amp
le, R
DP
: Rib
osom
al D
atab
ase
Pro
ject
Aquat Microb Ecol 71: 141–153, 2013
non, having been noted in both shallow ponds anddeep lakes (Finlay et al. 1988, Müller et al. 1991).
The most abundant OTUs from whole DNA(Table 1) and those with >70% enrichment in theBrdU-IP samples (Table 2) come from a broadphyletic diversity of eukaryotes, including stra-menopiles (particularly chrysophytes and synuro-phytes), alveolates (particularly ciliates and dinofla-gellates), rhizarians, green algae, cryptophytes, andhaptophytes. As indicated above, some groupsshowed especially large differences in their relativeabundance at different depths (Fig. 6). Synurophytes,all of which have chloroplasts, are most common inthe surface waters, while cryptomonads are morecommon in the hypolimnion (Fig. 6).
In Lake Lacawac, 9 of the 36 (25%) most abundantOTUs in whole DNA samples and 13 of the 32 (41%)
most abundant OTUs in the BrdU-IP sam-ples are most similar to uncultured organ-isms from environmental clone libraries ordo not have any close homolog (≥85% se-quence identity) in GenBank. This breadthof higher-level taxa from a single sampleis known for several other planktonic sys-tems, and a large proportion of the OTUsidentified with molecular methods are of-ten most similar to uncultured organisms(Luo et al. 2010). Note that close homologsto our sequences may have been recov-ered from other eukaryotic pyrosequenc-ing projects, but tag sequences <200 bpunfortunately cannot be deposited inGenBank as per NCBI rules; unlike Gen-Bank, the NCBI SRA can currently besearched only by project. Randa (2007)used a cloning procedure to identify bac-terivores from 5 eukaryotic lineages (cer-cozoans, alveolates, strameno philes, meta-zoans, and cryptophytes) that ingestedBrdU-labeled bacteria in a coastal marinesystem. The majority of phylotypes labeledwith BrdU in those experiments were alsonot closely related to previously identifiedprotist species (Randa 2007).
Recovery of poorly studied taxa
There are apparently many bacteri-vores whose SSU rRNA genes have stillnever been studied. Compared to thewhole DNA set of protistan taxa, agreater proportion of taxa labeled with
148
2D Stress: 0.01
BrdU-IPWhole DNA
Epilimnion
Hypolimnion
Metalimnion
NMDS plot: Bray-Curtis similarity
Fig. 5. Analysis of community similarity, showing non-metricmultidimensional scaling (NMDS) 2D ordination of a Bray-Curtis community dissimilarity matrix for the 6 samples.Close proximity of data points indicate high similarity
0 10 20 30 40 50 60 70 80 90 100
EpiBrdU
EpiWhole
MetaBrdU
MetaWhole
HypoBrdU
HypoWhole
Ciliates (Cil)
Dinoflagellates
Other Alveolates
Cryptomonas + other cryptophytes (Cryp) Euglenids
Prymnesiophytes
Rhizarians
Chlorophytes
Choanoflagellates
Synurophytes (Syn)
Chrysophytes
Diatoms
Other Stramenopiles (OS)
Raphidophytes
Other Eukaryotes (O)
Cil
Cil
Cil
Cil
Cil
Cil
Syn
Syn
Syn
Syn
OS
OS
OS
OS
Cryp
Cryp
Cryp
Cry
p
O
O
O
O
O
O
Fig. 6. Taxon plots, where bars show the relative abundance of sequencesrecovered from each higher-level taxon within each sample. EpiBrdU:epilimnion bromodeoxyuridine immunoprecipitated (brdU-IP), EpiWhole:epilimnion whole DNA, MetaBrdU: metalimnion BrdU-IP, MetaWhole:metalimnion whole DNA, HypoBrdU: hypolimnion BrdU-IP, HypoWhole:
hypolimnion whole DNA
Fay et al.: Culture-independent identification of bacterivores 149
No.
of
seq
uen
ces
reco
vere
d%
seq
s.C
lose
st G
enB
ank
BL
AS
TR
DP
con
sen
sus
lin
eag
eE
pi
Hyp
oM
eta
Su
mfr
omB
LA
ST
hit
wit
hid
enti
tyB
rdU
-IP
Wh
ole
Brd
U-I
PW
hol
eB
rdU
-IP
Wh
ole
Brd
U-I
P>
85%
id
enti
ty(%
)
151
453
210
209
899
1733
78G
onyo
stom
um
100
Eu
kar
yota
; str
amen
opil
es; R
aph
idop
hyc
eae;
Ch
atto
nel
lale
s; V
acu
olar
iace
ae17
142
20
175
322
658
71U
ncu
lt. f
resh
wat
er10
0E
uk
aryo
taeu
kar
yote
314
7627
05
036
877
Ch
lore
llac
eae
99E
uk
aryo
ta; V
irid
ipla
nta
e; C
hlo
rop
hyt
a; T
reb
ouxi
oph
ycea
e0
00
096
265
361
73S
pon
gom
onas
88E
uk
aryo
ta54
208
00
122
276
76M
allo
mon
as92
Eu
kar
yota
; str
amen
opil
es1
08
045
209
263
79N
one
–E
uk
aryo
ta6
844
012
7418
088
Non
e–
Eu
kar
yota
00
633
1182
132
87C
ryp
tom
onas
97E
uk
aryo
ta; C
ryp
top
hyt
a; C
ryp
tom
onad
ales
; Cry
pto
mon
adac
eae;
Cry
pto
mon
as27
720
05
1712
174
Tra
chel
ius
98E
uk
aryo
ta; A
lveo
lata
; Cil
iop
hor
a; I
ntr
amac
ron
ucl
eata
; Lit
osto
mat
ea0
012
100
10
113
88B
ryop
hyr
a91
Eu
kar
yota
; Alv
eola
ta16
430
013
4111
374
Syn
ura
100
Eu
kar
yota
; str
amen
opil
es; S
ynu
rop
hyc
eae;
Syn
ura
les;
Mal
lom
onad
acea
e3
791
02
2010
594
Gym
nod
iniu
m10
0E
uk
aryo
ta; A
lveo
lata
; Din
oph
ycea
e1
980
00
099
99C
hry
sosp
hae
rell
a92
Eu
kar
yota
; str
amen
opil
es3
940
02
099
95U
ncu
lt. a
lveo
late
95E
uk
aryo
ta; A
lveo
lata
; Din
oph
ycea
e2
940
00
096
98U
ncu
lt. G
ymn
odin
ium
95E
uk
aryo
ta; A
lveo
lata
; Din
oph
ycea
e1
860
00
087
99P
roro
cen
tru
m94
Eu
kar
yota
; Alv
eola
ta; D
inop
hyc
eae
768
00
47
8687
Un
cult
. alv
eola
te94
Eu
kar
yota
; Alv
eola
ta0
160
017
4477
78N
one
–E
uk
aryo
ta0
00
013
5770
81S
pon
gom
onas
86E
uk
aryo
ta; R
hiz
aria
; Cer
cozo
a0
00
03
6568
96U
roce
ntr
um
100
Eu
kar
yota
; Alv
eola
ta; C
ilio
ph
ora;
In
tram
acro
nu
clea
ta8
00
553
066
83C
hry
soch
rom
uli
na
99E
uk
aryo
ta; H
apto
ph
ycea
e; P
rym
nes
iale
s; P
rym
nes
iace
ae; I
man
ton
ia0
00
016
4460
73N
one
–E
uk
aryo
ta0
00
04
5256
93S
pon
gom
onas
89E
uk
aryo
ta; R
hiz
aria
; Cer
cozo
a; C
erco
mon
adid
a; S
pon
gom
onas
00
054
00
5410
0N
one
-E
uk
aryo
ta1
410
03
954
93C
hlo
rell
idiu
m89
Eu
kar
yota
; str
amen
opil
es1
211
04
2451
88N
one
–E
uk
aryo
ta0
490
00
049
100
Man
ton
iell
a97
Eu
kar
yota
; Vir
idip
lan
tae;
Ch
loro
ph
yta;
Pra
sin
oph
ycea
e0
00
03
4548
94D
inob
ryon
div
erg
ens
91E
uk
aryo
ta; s
tram
enop
iles
; Ch
ryso
ph
ycea
e0
00
430
043
100
Non
e–
Eu
kar
yota
142
00
00
4398
Un
clas
sifi
ed c
ilia
te93
Eu
kar
yota
; Alv
eola
ta; C
ilio
ph
ora;
In
tram
acro
nu
clea
ta0
00
09
3443
79V
orti
cell
a92
Eu
kar
yota
00
734
00
4183
Non
e-
Eu
kar
yota
Tab
le 2
. Su
bse
t of
tax
a m
ost
hig
hly
lab
eled
wit
h b
rom
odeo
xyu
rid
ine
(Brd
U).
Pro
tist
tax
a la
bel
ed w
ith
Brd
U, o
rder
ed b
y to
tal n
um
ber
of
seq
uen
ces
reco
vere
d. I
ncl
ud
ed a
reta
xa w
ith
>0.
05%
seq
uen
ce a
bu
nd
ance
(>
38 s
equ
ence
s) w
ith
>70
% o
f th
e to
tal
nu
mb
er o
f se
qu
ence
s co
min
g f
rom
Brd
U i
mm
un
opre
cip
itat
ed (
Brd
U-I
P)
sam
ple
s. S
um
: to
tal n
um
ber
of
seq
uen
ces
reco
vere
d. E
pi:
ep
ilim
nio
n, M
eta:
met
alim
nio
n, H
ypo:
hyp
olim
nio
n, W
hol
e: w
hol
e D
NA
sam
ple
, RD
P: R
ibos
omal
Dat
abas
e P
roje
ct
Aquat Microb Ecol 71: 141–153, 2013
BrdU were poorly described or rare species. For 8 ofthe taxa in the list of BrdU-labeled taxa combined forall depths, we found no match to a GenBank se -quence with ≥85% identity, whereas the 31 mostcommonly recovered taxa from the whole DNA sam-ples all had some match within at least 85%. TheseOTUs with no close homolog in GenBank are likelyto be uncultured taxa. They may be more difficult toculture because they are fragile, small, difficult todistinguish as distinct taxa, or have specialized meta-bolic requirements. The relative proportion of un -identified OTUs found in the BrdU-labeled se quencessuggests that BrdU is also useful for finding addi-tional rare protist diversity.
Identification of bacterivores
We have taken a conservative approach to estimat-ing those taxa that are truly BrdU-labeled and thuspotentially bacterivorous. Factors such as non-linearPCR amplification and contamination from unlabeledtaxa are possible, so only taxa with the highest pro-portion of BrdU-labeled versus whole DNA sequencesrecovered are listed in Table 1, i.e. taxa with >0.05%sequence abundance and >70% of the total numberof sequences coming from BrdU-IP samples. We iden-tify these taxa as putative bacterivores, either pureheterotrophs or mixotrophs. Confirmation of bacteri-vory as a trait may require continued experimentalwork in culture or field experiments. For example, theprasinophyte Mantoniella has not previously been re-ported as bacterivorous, but our results (Table 1), andthe fact that our Antarctic Mantoniella isolate hasbeen observed to ingest particles (McKie-Krisberg etal. 2011), suggest that more work should be done toexamine bacterivory in this genus. Conversely, only afew diatom OTUs were recovered. A small proportionof the diatom sequences were in the BrdU-IP samples,and it is unlikely that these were bacterivorous. Wesee the approach taken in this paper as a hypothesis-generating mode of research, analogous to transcrip-tomic assays (such as microarrays), which identifypromising genes (or in this case particular taxa) forfurther study of function.
Among the most frequently identified BrdU-labeledprotists, 3 of the 5 top hits were phytoplankton; an un-cultured eukaryote and a heterotrophic flagellaterounded out the top 5 bacterivores (Table 1). The mostsequences recovered in the BrdU-IP samples were forthe raphidophyte Gonyostomum. Although mixotro-phy, i.e. the combination of photosynthesis and feed-ing in an individual, has not been demonstrated for
this genus, several species of marine raphidophyteswere recently identified as mixotrophic (Jeong 2011).The 2 next most recovered algae were not previouslyidentified as bacterivorous either, but known mixotro-phic genera that were BrdU positive in these experi-ments included Cryptomonas, Chrysosphaerella,Chrysochromulina, Uro glena, and Dinobryon. Thesedata highlight the potential importance of bacterivoryas a mode of nutrition for many microalgae (Sanders1991, Jones 2000, Zubkov & Tarran 2008). It ispossible that incubating the samples overnight en-hanced phagotrophy in the mixotrophs, althoughdarkness has no effect on some common mixotrophspecies (Bird & Kalff 1987, Sanders et al. 2001).
Some of the BrdU-positive dinoflagellate generaidentified as bacterivores contain both heterotrophicand mixotrophic species (Sanders & Porter 1988). Wedo not identify these specifically as mixotrophs,although mixotrophy is well known for the group(Sanders 1991). Prasinophytes, previously identifiedas potentially bacterivorous by González et al. (1993),Bell & Laybourn-Parry (2003), and Sanders & Gast(2012), made up from ~1 to 5% and ciliates made up~8 to 25% of the bacterivores from all 3 depths. Het-erotrophic flagellates, including choano flagellates,also were identified with BrdU at all depths (Fig. 6).A closer look at stramenopiles shows that, within thisgroup, the 4 OTUs with the highest sequence abun-dance were Chryso sphaerella, Synura, Gonyosto-mum, and Uroglena. Over 60% of the sequencesrecovered for Uroglena were from BrdU-IP treat-ments, strongly supporting its ecological role as abacterivore. Most stramenopile taxa found in thisstudy fall within the chrysophyte/synurophyte group(their systematics remains in flux; Andersen 2007),and Fig. 6 emphasizes the fact that this taxonomicallyrich group of algae holds great opportunity for thecomparative study of modes of nutrition.
Certain factors may confound the use of the BrdUmethod by identifying ‘false positive’ bacterivoroustaxa. False positives may occur if osmotrophs take upfree BrdU, and free BrdU may result from cycling ofbiomolecules. In axenic culture with 1 µM BrdU, thephototrophs Nannochloris atomus, Gymnodidiniumsanguineum, and Phaeodactylum tricornutum didnot incorporate BrdU directly from the medium, butthe osmotrophic labyrinthulomycete Schizochytriumaggregatum did become labeled (Randa 2007). Evengreen plants can take up free DNA (Paungfoo-Lonhi-enne et al. 2010). Although bacteria are better com-petitors for the dissolved DNA than protists (Løvdalet al. 2007) and likely for BrdU as well, we recom-mend low centrifugation speeds, gentle methods for
150
Fay et al.: Culture-independent identification of bacterivores
dispersion of cells, and minimization of the timebetween washing/centrifugation of BrdU-labeledbacteria and their addition to environmental samplesto minimize potential leakage of free BrdU or free-BrdU-labeled-DNA.
Non-protistan bacterivores can also be identifiedby this method. Though not recorded in the tables ofmost commonly recovered sequences, rotifers andcopepods were appropriately identified as bacterivo-rous in the analysis. Predators of bacterivorous pro-tists (including other protists) at higher trophic levelsmight also become labeled, again generating falsepositives. Since this requires time for ingestion of thelabeled bacterivore followed by growth and incorpo-ration into the secondary predator’s DNA, minimiz-ing incubation times is one way to ameliorate thisproblem.
An important assumption in this work is that bac-terivores would not be identified if they did not incor-porate BrdU into their DNA after ingestion. While notall bacteria take up thymidine or BrdU, little is knownabout incorporation of complete nucleotides by pro-tists. Caron et al. (1993) found that 3H-thymidine fromlabeled bacteria did accumulate in protists, but notquantitatively over periods longer than several hours;this implies that some of the nucleotide is brokendown. However, Taylor & Sullivan (1984) found thatwhen bacteria were labeled with 14C-thymidine andfed to the ciliate Euplotes, respiration and exudationwere less than in experiments using 14C-glucose, im-plying that a consumer is more likely to use nucleo-tides in anabolic processes. Protists that failed toingest the labeled bacteria for any reason, and thosewith very low feeding rates, could also lead to falsenegatives. Size-selective feeding by protists is wellknown (González et al. 1990), but this is most likely tobe a problem if a strain of large bacteria is used andthe grazers are very small (Sanders & Gast 2012). Fu-ture work with BrdU could be directed towards label-ing the natural bacterial community to reduce the ef-fect of prey size or type, although as noted previously,not all bacteria take up BrdU or thymidine. The datapresented here and by Randa (2007) show that manyprotists incorporate ingested nucleotides directly intotheir DNA, and that this method can be utilized togain an important perspective into the link betweentaxonomic diversity and function.
CONCLUSION
The large phylogenetic diversity of eukaryoticmicro organisms is well recognized and ascribed to a
large extent to uncultured species (Medinger et al.2010). Consequently, the ability to assign functionalimportance to various groups in an aquatic food webis still limited. Culture-independent molecular meth-ods, including the BrdU-IP method used here, canpartition microbial taxa by function. Most other cul-ture-independent studies involve stable isotopeprobing (SIP, reviewed by Gutierrez-Zamora &Manefield 2010), whereby microbes are fed withheavy stable isotope labeled substrates. Like anytechnique, SIP has its drawbacks. It can be difficult toachieve sufficient incorporation of the heavy isotope,especially with in situ levels of the substrate (Chen &Murrell 2010). Additionally, centrifugal separation oflabeled nucleic acids is not complete, and back-ground levels of template are present in all fractions(Lueders et al. 2004).
One advantage of using halogenated nucleotideslike BrdU for labeling and separation by IP is thatmonoclonal antibodies have very high binding affin-ity and will bind regardless of the molecular weightof the antigen. Our study demonstrates the utility ofcombining BrdU-labeling and next-generation am -plicon sequencing to study microbial trophic ecology.These data show that bacterivores from a small set oflakewater samples are distributed widely across theeukaryotic tree of life. Furthermore, many of the pro-tists identified as bacterivores are algae. The occur-rence of mixotrophic algae in pelagic ecosystemnutrient flux is now well established (Sanders &Porter 1988, Jeong et al. 2010). Culture-independentmethods such as BrdU-labeling add to the toolboxavailable for studying natural protist communities,and models of aquatic microbial food webs must con-tinue to incorporate and expand on knowledge of thefunctional diversity of the many unculturable pro-tists, including mixotrophs.
Acknowledgements. We thank the Lake Lacawac Founda-tion for maintaining access to the lake for research, M.Randa and E. Lim for their advice on and discussion of theBrdU method, which they first developed, I. Akan for adviceon immunochemistry, and S. DeVaul for assistance in thelaboratory. This work was funded in part by NSF grantsOPP-0838847 and OPP-0838955. The funding agency hadno role in study design, data collection and analysis, deci-sion to publish, or preparation of the manuscript.
LITERATURE CITED
Amaral-Zettler LA, McCliment EA, Ducklow HW, Huse SM(2009) A method for studying protistan diversity usingmassively parallel sequencing of V9 hypervariableregions of small-subunit ribosomal RNA genes. PLoSONE 4: e6372
151
Aquat Microb Ecol 71: 141–153, 2013
Andersen RA (2007) Molecular systematics of the Chryso-phyceae and Synurophyceae. In: Brodie J, Lewis J (eds)Unravelling the algae: the past, present, and future ofalgal systematics. CRC Press, Boca Raton, FL, p 285−315
Azam F, Fenchel T, Field JG, Gray JS, Meyer-Reil LA,Thingstad F (1983) The ecological role of water-columnmicrobes in the sea. Mar Ecol Prog Ser 10: 257−263
Behnke A, Bunge J, Barger K, Breiner HW, Alla V, Stoeck T(2006) Microeukaryote community patterns along anO2/H2S gradient in a supersulfidic anoxic fjord (Fram-varen, Norway). Appl Environ Microbiol 72: 3626−3636
Behnke A, Engel M, Christen R, Nebel M, Klein RR, StoeckT (2011) Depicting more accurate pictures of protistancommunity complexity using pyrosequencing of hyper-variable SSU rRNA gene regions. Environ Microbiol 13: 340−349
Bell EM, Laybourn-Parry J (2003) Mixotrophy in the Antarc-tic phytoflagellate, Pyramimonas gelidicola (Chloro-phyta: Prasinophyceae). J Phycol 39: 644−649
Berninger U, Caron DA, Sanders RW (1992) Mixotrophicalgae in three ice covered lakes of the Pocono Moun-tains, U.S.A. Freshw Biol 28: 263−272
Bik HM, Porasinska DL, Creer S, Caporaso JG, Knight R,Thomas WK (2012) Sequencing our way towards under-standing global eukaryotic biodiversity. Trends Ecol Evol27: 233−243
Bird DF, Kalff J (1987) Algal phagotrophy: regulating factorsand importance relative to photosynthesis in Dinobryon(Chrysophyceae). Limnol Oceanogr 32: 277−284
Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data. Nat Methods 7: 335−336
Caron DA (1991) Heterotrophic flagellates associated withsedimenting detritus. In: Patterson DJ, Larsen J (eds) Thebiology of free-living heterotrophic flagellates. Claren-don Press, Oxford, p 77−92
Caron DA, Lessard EJ, Voytek M, Dennett MR (1993) Use oftritiated thymidine (TdR) to estimate rates of bacterivory: implications of label retention and release by bacteri-vores. Mar Microb Food Webs 7: 177−196
Chen Y, Murrell JC (2010) When metagenomics meets sta-ble-isotope probing: progress and perspectives. TrendsMicrobiol 18: 157−163
Clarke K, Warwick R (2001) Change in marine communities: an approach to statistical analysis and interpretation, 2ndedn. PRIMER-E, Plymouth
Edgar RC (2010) Search and clustering orders of magnitudefaster than BLAST. Bioinformatics 26: 2460−2461
Finlay BJ, Clarke KJ, Cowling AJ, Hindle RM, Rogerson A,Berninger UG (1988) On the abundance and distributionof protozoa and their food in a productive freshwaterpond. Eur J Protistol 23: 205−217
Finlay BJ, Maberly SC, Cooper JI (1997) Microbial diversityand ecosystem function. Oikos 80: 209−213
Gast RJ, Dennett MR, Caron DA (2004) Characterization ofprotistan assemblages in the Ross Sea, Antarctica, bydenaturing gradient gel electrophoresis. Appl EnvironMicrobiol 70: 2028−2037
González JM, Sherr EB, Sherr BF (1990) Size-selective graz-ing on bacteria by natural assemblages of estuarine flag-ellates and ciliates. Appl Environ Microbiol 56: 583−589
González JM, Sherr EB, Sherr BF (1993) Digestive enzymeactivity as a quantitative measure of protistan grazing: the acid lysozyme assay for bacterivory. Mar Ecol ProgSer 100: 197−206
Gutierrez-Zamora ML, Manefield M (2010) An appraisal ofmethods for linking environmental processes to specificmicrobial taxa. Crit Rev Environ Sci Technol 9: 153−185
Hamasaki K, Taniguchi A, Tada Y, Long RA, Azam F (2007)Actively growing bacteria in the inland Sea of Japan,identified by combined bromodeoxyuridine immunocap-ture and denaturing gradient gel electrophoresis. ApplEnviron Microbiol 73: 2787−2798
Jeong HJ (2011) Mixotrophy in red tide algae raphido-phytes. J Eukaryot Microbiol 58: 215−222
Jeong HJ, Yoo YD, Kim JS, Seong KA, Kang NS, Kim TH(2010) Growth, feeding and ecological roles of the mixo-trophic and heterotrophic dinoflagellates in marineplanktonic food webs. Ocean Sci J 45: 65−91
Jones RI (2000) Mixotrophy in planktonic protists: anoverview. Freshw Biol 45: 219−226
Kemp PF, Sherr BF, Sherr EB, Cole JJ (eds) (1993) Handbookof methods in aquatic microbial ecology. CRC Press,Boca Raton, FL
Kuske CR, Banton KL, Adorada DL, Stark PC, Hill KK, Jack-son PJ (1998) Small-scale DNA sample preparationmethod for field PCR detection of microbial cells andspores in soil. Appl Environ Microbiol 64: 2463−2472
454 Life Sciences Corp. (2010) GS Junior System guidelinesfor amplicon experimental design. 454 Life Sciences,Branford, CT
Løvdal T, Tanaka T, Thingstad TF (2007) Algal-bacterialcompetition for phosphorus from dissolved DNA, ATP,and orthophosphate in a mesocosm experiment. LimnolOceanogr 52: 1407−1419
Lozupone CA, Hamady M, Kelley ST, Knight R (2007) Quan-titative and qualitative beta diversity measures lead todifferent insights into factors that structure microbialcommunities. Appl Environ Microbiol 73: 1576−1585
Lueders T, Manefield M, Friedrich MW (2004) Enhancedsensitivity of DNA and rRNA based stable isotope prob-ing by fractionation and quantitative analysis of isopyc-nic centrifugation gradients. Environ Microbiol 6: 73−78
Luo W, Bock C, Li HR, Padisák J, Krienitz L (2010) Molecularand microscopic diversity of planktonic eukaryotes in theoligotrophic Lake Stechlin (Germany). Hydrobiologia661: 133−143
Macaluso AL, Mitchell DL, Sanders RW (2009) Direct effectsof UV-B radiation on the freshwater heterotrophic nano-flagellate Paraphysomonas sp. Appl Environ Microbiol75: 4525−4530
McKie-Krisberg Z, Fay SA, Sanders RW (2011) Competitiveassays of two mixotrophs and two diatoms from the RossSea, Antarctica. J Phycol 47: S67
Medinger R, Nolte V, Pandey RV, Jost S, Ottenwälder B,Schlötterer C, Boenigk J (2010) Diversity in a hiddenworld: potential and limitation of next generationsequencing for surveys of molecular diversity of eukary-otic microorganisms. Mol Ecol 19: 32−40
Müller H, Schöne A, Pinto-Coelho RM, Schweizer A, WeisseT (1991) Seasonal succession of ciliates in Lake Con-stance. Microb Ecol 21: 119−138
Newman DK, Banfield JF (2002) Geomicrobiology: howmolecular-scale interactions underpin biogeochemicalsystems. Science 296: 1071−1077
Olsen GJ, Lane DJ, Giovannoni SJ, Pace NR, Stahl DA(1986) Microbial ecology and evolution: a ribosomal RNAapproach. Annu Rev Microbiol 40: 337−365
Paul EA (2007) Soil microbiology, ecology, and biochemistry,3rd edn. Academic Press, Boston, MA
152
Fay et al.: Culture-independent identification of bacterivores
Paungfoo-Lonhienne C, Lonhienne TGA, Mudge SR,Schenk PM, Christie M, Carroll BJ, Schmidt S (2010)DNA is taken up by root hairs and pollen, and stimu-lates root and pollen tube growth. Plant Physiol 153: 799−805
Randa MA (2007) The role of temperature, salinity, and pro-tozoan predation on the population dynamics of Vibriovulnificus in Barnegat Bay, New Jersey. PhD disserta-tion, Temple University, Philadelphia, PA
Reeder J, Knight R (2010) Rapidly denoising pyrosequenc-ing amplicon reads by exploiting rank-abundance distri-butions. Nat Methods 7: 668−669
Rozen S, Skaletsky H (2000) Primer3 on the WWW for gen-eral users and for biologist programmers. Methods MolBiol 132: 365−386
Sanders RW (1991) Mixotrophic protists in marine and fresh-water ecosystems. J Eukaryot Microbiol 38: 76−81
Sanders RW, Gast RJ (2012) Bacterivory by phototrophicpicoplankton and nanoplankton in Arctic waters. FEMSMicrobiol Ecol 82: 242−253
Sanders RW, Porter KG (1988) Phagotrophic phytoflagel-lates. In: Marshall KC (ed) Advances in microbial ecol-ogy. Plenum, New York, NY, p 167−192
Sanders RW, Caron DA, Davidson JM, Dennett MR, MoranDM (2001) Nutrient acquisition and population growth ofa mixotrophic alga in axenic and bacterized cultures.Microb Ecol 42: 513−523
Sherr EB, Sherr BF (2002) Significance of predation by pro-
tists in aquatic microbial food webs. Antonie Leeuwen-hoek 81: 293−308
Siver PA, Chock JS (1986) Phytoplankton dynamics in achrysophycean lake. In: Kristiansen J, Andersen RA (eds)Chrysophytes: aspects and problems. Cambridge Uni-versity Press, Cambridge, p 165−183
Sogin ML, Morrison HG, Huber JA, Welch DM and others(2006) Microbial diversity in the deep sea and the under-explored ‘rare biosphere’. Proc Natl Acad Sci USA 103: 12115−12120
Stoeck T, Bass D, Nebel M, Christen R, Jones MDM, BreinerH, Richards TA (2010) Multiple marker parallel tag envi-ronmental DNA sequencing reveals a highly complexeukaryotic community in marine anoxic water. Mol Ecol19: 21−31
Taylor GT, Sullivan CW (1984) The use of 14C-labeled bacte-ria as a tracer of ingestion and metabolism of bacterialbiomass by microbial grazers. J Microbiol Methods 3: 101−124
Ueda J, Saito H, Watanabe H, Evers BM (2005) Novel andquantitative DNA dot-blotting method for assessment ofin vivo proliferation. Am J Physiol 288: G842−G847
Urbach E, Vergin KL, Giovannoni SJ (1999) Immunochemi-cal detection and isolation of DNA from metabolicallyactive bacteria. Appl Environ Microbiol 65: 1207−1213
Zubkov MV, Tarran GA (2008) High bacterivory by thesmallest phytoplankton in the North Atlantic Ocean.Nature 455: 224−226
153
Editorial responsibility: Klaus Jürgens,Rostock, Germany
Submitted: March 26, 2013; Accepted: October 14, 2013Proofs received from author(s): November 25, 2013