Skip to main content

Some NLM-NCBI services and products are experiencing heavy traffic, which may affect performance and availability. We apologize for the inconvenience and appreciate your patience. For assistance, please contact our Help Desk at info@ncbi.nlm.nih.gov.

Applied and Environmental Microbiology logoLink to Applied and Environmental Microbiology
. 1998 Nov;64(11):4522–4529. doi: 10.1128/aem.64.11.4522-4529.1998

Kinetic Bias in Estimates of Coastal Picoplankton Community Structure Obtained by Measurements of Small-Subunit rRNA Gene PCR Amplicon Length Heterogeneity

Marcelino Suzuki 1,2,, Michael S Rappé 2,, Stephen J Giovannoni 2,*
PMCID: PMC106679  PMID: 9797317

Abstract

Marine bacterioplankton diversity was examined by quantifying natural length variation in the 5′ domain of small-subunit (SSU) rRNA genes (rDNA) amplified by PCR from a DNA sample from the Oregon coast. This new technique, length heterogeneity analysis by PCR (LH-PCR), determines the relative proportions of amplicons originating from different organisms by measuring the fluorescence emission of a labeled primer used in the amplification reaction. Relationships between the sizes of amplicons and gene phylogeny were predicted by an analysis of 366 SSU rDNA sequences from cultivated marine bacteria and from bacterial genes cloned directly from environmental samples. LH-PCR was used to compare the distribution of bacterioplankton SSU rDNAs from a coastal water sample with that of an SSU rDNA clone library prepared from the same sample and also to examine the distribution of genes in the PCR products from which the clone library was prepared. The analysis revealed that the relative frequencies of genes amplified from natural communities are highly reproducible for replicate sets of PCRs but that a bias possibly caused by the reannealing kinetics of product molecules can skew gene frequencies when PCR product concentrations exceed threshold values.


Libraries of small-subunit rRNA gene (SSU rDNA) clones prepared by PCR are widely applied to study the microbial diversity of natural ecosystems. These studies have provided dramatic evidence that the majority of microbial communities are dominated by previously unknown organisms (4, 8, 18). However, quantitative comparisons using clone libraries to assess microbial community structure have been limited by several factors, including (i) undersampling of diversity and (ii) uncertainty about sources of bias in the cloning process, in particular bias by the PCR. Undersampling, often estimated by coverage values or by rarefaction curves, results from the difficulty of processing a large number of clones (11). The lack of an alternative means to quantitatively assess the composition of complex mixtures of rDNAs from in situ communities has made it difficult to evaluate methodological sources of bias by the cloning process. The method we describe here, length heterogeneity analysis by PCR (LH-PCR), overcomes some of these problems by quickly providing a profile of amplicon diversity in complex mixtures of PCR products.

LH-PCR is similar to the approach used in our earlier study of bias in the PCR (16) and the recently published terminal restriction fragment length polymorphism (9) and fluorescent restriction fragment length polymorphism (2) techniques. In both LH-PCR and these methods, the proportions of PCR amplicons originating from different genes are estimated from the fluorescence emission of labeled PCR primers. However, instead of identifying PCR amplicons based on restriction endonuclease sites, in LH-PCR the discrimination of amplicons originating from different organisms is based on natural variation in the lengths of SSU rDNAs.

In a previous study, we investigated biases introduced during the amplification of rDNAs by PCR (16). In that study, the templates consisted of pairwise mixtures of SSU rDNAs from bacteria belonging to three different phylogenetic groups. To estimate bias, we compared the proportions of genes in the PCR products with their proportions in the starting template mixtures. We observed that, above threshold product concentrations, PCR dramatically biased the frequency distribution among gene homologs relative to the original mixture. A kinetic model based on competition between primers and products which successfully explained the experimental results was developed. These results indicated that this type of bias by PCR might lead to an increase in net diversity estimates among amplicons relative to the gene diversity of the native DNA mixture. Evidence also indicated that artifacts resulting from this phenomenon could be controlled by limiting the number of replication cycles to maintain product levels below threshold values. However, the effect of this type of bias on the composition of PCR products amplified from natural community DNA was uncertain since, in a complicated mixture of genes, a single gene might not reach threshold concentrations at which competition between product and primer reannealing would have a pronounced effect.

Here we present the results of a study in which LH-PCR was used to estimate the community composition of bacterioplankton from a water sample collected off the Oregon coast. In order to trace the phylogenetic origin of the domains amplified by LH-PCR, we performed an analysis of the length variability in SSU rDNAs of bacterial strains cultivated or directly cloned from the same seawater sample, as well as sequences retrieved from gene sequence databases. We found that the relative gene frequencies obtained from natural communities by LH-PCR were highly reproducible when PCR product concentrations were limited to relatively low values but that, at high concentrations, the kinetic bias caused by template reannealing significantly skewed gene frequencies. SSU rDNA amplicons with sizes corresponding to the alpha subdivision of the class Proteobacteria (alpha-Proteobacteria) represented the largest fraction of the bacterial rDNA amplicons (ca. 65%), while no other size class of SSU rDNA amplicons represented greater than 10% of the bacterial rDNA amplicons. Overall, the results suggest that LH-PCR is an effective tool for assessing microbial community structure and that clone libraries may often overrepresent bacterioplankton diversity because the relative frequencies of dominant species have been reduced by a systematic bias.

MATERIALS AND METHODS

Sample collection, nucleic acid isolation, and clone library construction.

On 28 April 1993, a subsurface (10-m) water sample was collected by Niskin bottles at a station located 8 km off the mouth of Yaquina Bay, Oreg. (44°39.1′N, 124°10.6′W). The water was prescreened through 10-μm-pore-size Nitex mesh and transported in autoclaved polyethylene carboys to the laboratory for the remaining analyses. Picoplankton from 4-liter (subsample 1) and 16-liter (subsample 2) subsamples were collected by filtration onto 0.2-μm-pore-size polysulfone filters (Supor-200; Gelman Sciences Inc., Ann Arbor, Mich.). Total cellular nucleic acids were isolated from the picoplankton samples by lysis with proteinase K and sodium dodecyl sulfate, followed by phenol-chloroform extraction as previously described (6). A portion of the DNA sample isolated from subsample 2 was used as template in the amplification of nearly full-length SSU rDNAs by PCR and subsequently cloned into a plasmid vector as described elsewhere (12, 17). SSU rDNA clones recovered in this library have been partially described elsewhere (12, 17).

LH-PCR.

Ten nanograms of purified genomic DNA from each subsample was used as template for LH-PCR. The forward primer, 27F (5′-AGA GTT TGA TCM TGG CTC AG-3′) (3), was 5′ end labeled with the phosphoramidite dye 6-FAM (graciously supplied by Applied Biosystems Inc., Foster City, Calif.) or purchased from Genset (San Diego, Calif.). The reverse primers used were 355R (5′-GCT GCC TCC CGT AGG AGT-3′) (1) for domain A and 536R (5′-GWA TTA CCG CGG CKG CTG-3′) (5) for domain B, synthesized at the Central Services Laboratory, Center for Gene Research and Biotechnology, Oregon State University. In a final volume of 100 μl, reaction mixtures contained 0.2 mM premixed deoxynucleoside triphosphates (Stratagene, La Jolla, Calif.), 1.5 mM MgCl2, 5% acetamide, 0.5 μM forward primer, 0.5 μM (one) reverse primer, and 2.5 U of Taq DNA polymerase (Promega, Madison, Wis.). All reactions used the Ampliwax hot-start protocol (Perkin-Elmer Cetus, Norwalk, Conn.) in a PTC100 thermal cycler (MJ Research Inc., Watertown, Mass.) programmed to 16 cycles for primer 355R (except for the reactions evaluating PCR bias) or 21 cycles for primer 536R, each consisting of 96°C denaturation for 1 min, 55°C annealing for 1 min, and 72°C extension for 3 min.

The concentration of labeled PCR products was measured in a Shimadzu UV160U spectrophotometer (Shimadzu Co., Kyoto, Japan) or estimated after electrophoresis in an agarose minigel stained with ethidium bromide (50 μg/ml) and compared with mass standards. The PCR products were purified with Qiaquick spin columns (Qiagen, Chatsworth, Calif.). Approximately 10 fmol of the LH-PCR products was discriminated by Long Ranger (FMC, Rockland, Maine) polyacrylamide gel electrophoresis in a model 377 automated DNA sequencer (Applied Biosystems Inc.) with Genescan (Applied Biosystems Inc.), a software package that estimates the sizes of bands in the gel and their integrated fluorescence emission. The output of the software is electropherograms in which the bands are represented by peaks and the integrated fluorescence of each band is the area under the peaks (see Fig. 1). The integrated fluorescence increased linearly with concentrations of up to 50 fmol of PCR products, indicating that the relative proportion of the integrated fluorescence of each peak corresponded to the proportion of each amplicon in the PCR products (data not shown). The relative abundance of amplicons was estimated as the ratio between the integrated fluorescence of each of the peaks and the total integrated fluorescence of all peaks.

FIG. 1.

FIG. 1

Electropherogram of DNA fragments amplified by PCR with primer set A from genomic DNA isolated from seawater subsample 2. The letters A to W correspond to the peaks detected by the Genescan 2.1.2 software in at least one of triplicate reactions. The x axis represents the size of domains in base pairs estimated by comparison to the size standard GS2500 (Applied Biosystems Inc.). The y axis represents relative fluorescence units.

Length heterogeneity analysis of published sequences.

In LH-PCR, amplicons originating from different templates are identified by length heterogeneity in hypervariable regions of the SSU rDNA. Three such regions occur in the 5′ end of the gene, around locations homologous to Escherichia coli positions 90, 190, and 450. In order to verify the phylogenetic coherence of length heterogeneity contained in these variable regions, we compared the length heterogeneity of domains homologous to the domain between E. coli positions 8 and 355 (domain A) and positions 8 and 536 (domain B). The analysis included previously published sequences of bacterial species isolated from the same water sample as that used for the LH-PCR analysis or directly cloned from DNA extracted from subsample 2 (17), as well as SSU rDNA sequences of bacterial species isolated from seawater or directly cloned from DNA extracted from seawater, retrieved from the GenBank, Ribosomal Database Project (10), and ARB (14) sequence databases.

Bias by PCR.

Two experiments were performed to evaluate the introduction of bias by PCR. In order to evaluate the bias described by Suzuki and Giovannoni (16) in the amplification of domain A (16) and to optimize the number of cycles for LH-PCR, we performed a time course experiment in which PCRs of domain A, with DNA purified from subsamples 1 and 2 as templates, were stopped by freezing at 10 (only subsample 1), 12, 14, 16, 18, 20, and 25 cycles. Concentrations of LH-PCR products from subsample 1 were measured spectrophotometrically as described above. Concentrations of LH-PCR products from subsample 2 were estimated from the agarose minigel as described above, except for the products of reactions with subsample 2, and stopped after 12 cycles, which were calculated assuming an amplification efficiency of 85% per cycle (13).

To evaluate the introduction of reannealing bias by PCR in the amplification of full-length rDNAs from mixed populations of bacteria, we used the optimized LH-PCR protocol for domain A as described above to compare the genotypic bacterioplankton community structure of (i) a genomic DNA sample from the Oregon coast (subsample 2) and (ii) the nearly full-length PCR products from subsample 2 after 35 cycles of amplification, used to prepare the SSU rDNA clone library, as described previously (12, 17). Triplicate LH-PCRs were performed as described above with 10 ng of genomic DNA or 60 pg of nearly full-length SSU rDNA PCR amplicons as templates, calculated so that the reactions using genomic DNA and full-length PCR amplicons contained approximately the same numbers of copies of SSU rDNAs. For this calculation, we assumed a bacterial origin for 50% of the DNA, an average chromosome size of 2 Mbp, and an average of two copies of the ribosomal operon per chromosome.

Finally, to estimate the introduction of bias by the cloning per se, we compared the community structure estimated by LH-PCR from full-length PCR products amplified from subsample 2 to that inferred from the relative proportion of SSU rDNA clones recovered in the clone library, grouped according to the sizes of domain A obtained directly from their SSU rDNA sequences.

RESULTS

Predicted length heterogeneity in the 5′ region of SSU rDNAs.

Three regions at the 5′ end of the SSU rDNA (V1, E. coli SSU rDNA positions 72 to 101; V2, E. coli SSU rDNA positions 176 to 221; and V3, E. coli SSU rDNA positions 451 to 481) are variable between different phylogenetic groups of bacteria. Insertions and deletions in these variable regions cause natural variability in the nucleotide lengths of molecules amplified with the 27F and 355R primer pair (domain A, ca. 312 to 363 bp) and the 27F and 536R primer pair (domain B, ca. 472 to 574 bp).

The lengths of domains A and B of bacteria isolated from seawater or SSU rDNAs directly cloned from seawater DNA are shown in Table 1 and are generally coherent with phylogenetic relationships. Many discrete fragment lengths are monophyletic but are shared by multiple species (e.g., 316 bp). Alpha-Proteobacteria and cyanobacteria have the shortest lengths for both domains. Beta-, gamma-, and delta-Proteobacteria and the Flexibacter-Bacteroides-Cytophaga group have intermediate lengths, and the longest domains are those from genes of low- and high-G+C gram-positive bacteria and members of the Vibrio fischeri subgroup of the gamma-Proteobacteria. Most phylogenetic groups have a unique combination of lengths for domains A and B (i.e., alpha-Proteobacteria have a domain A length of 315 bp and domain B lengths between 470 and 472 bp). The lengths of domains A and B of genes with plastid origins were not included in this study and are described elsewhere (12).

TABLE 1.

Length in nucleotides between positions homologous to E. coli SSU rDNA positions 8 through 355 (domain A) or 8 through 536 (domain B) for a variety of marine bacteriaa

Accession no. Cellular strain or environ-mental gene clone Taxonomic affiliation Size of domain (bp)
A B
env.OCS28 [A] α 312 468
U70681 env.OM55 α 312 468
U75264 env.SAR490 α 312 ND
X52170 env SAR12 c 313 ND
U75260 env.SAR418 α 313 ND
U75262 env.SAR440 α 313 ND
U75263 env.SAR466 α 313 ND
X52169 env.SAR6 c 313 470
Synechococcus sp. strain WH8101 c 313 471
Synechococcus sp. strain WH8103 c 313 472
U78945 env.OCS122 [B] α 314 470
U75259 env.SAR414 α 314 ND
env.SAR420 α 314 472
X52171 env.SAR7 c 314 471
U64002 Rhizobium sp. strain BAL25 α 314 470
U63957 Zoogloea sp. strain BAL43 α 314 470
L10934 env.FL1 α 315 470
L10935 env.FL11 α 315 471
env.OCS24 [C] α 315 471
U70678 env.OM25 α 315 471
U75258 env.SAR241 α 315 471
U75254 env.SAR464 α 315 471
X78315 Roseobacter algicola α 315 471
L15345 Strain LFR α 315 472
U63935 Caulobacter sp. strain BAL3 α 316 472
U75252 env.OCS12 [D] α 316 472
env.OCS126 [D] α 316 472
U78942 env.OCS19 [D] α 316 ND
U78943 env.OCS84 [D] α 316 ND
U70684 env.OM136 α 316 472
U70585 env.OM143 α 316 471
U70686 env.OM155 α 316 472
U70692 env.OM299 α 316 ND
U70679 env.OM38 α 316 472
U70680 env.OM42 α 316 472
U70682 env.OM65 α 316 473
X52280 env.SAR1 α 316 473
X52172 env.SAR11 α 316 473
U75255 env.SAR203 α 316 472
U75256 env.SAR211 α 316 472
U75257 env.SAR220 α 316 472
env.SAR402 α 316 472
U64003 Erythrobacter sp. strain BAL26 α 316 472
U64005 Erythrobacter sp. strain BAL28 α 316 472
U64011 Erythrobacter sp. strain SCB34 α 316 472
U64025 Erythrobacter sp. strain SCB48 α 316 472
U63952 Erythromicrobium sp. strain BAL34 α 316 472
U63958 Flavobacterium sp. strain BAL44 α 316 ND
U63939 Rhizomonas sp. strain BAL11 α 316 472
U63934 Rhodobacter sp. strain BAL2 α 316 472
U63949 Rhodobacter sp. strain BAL27 α 316 472
U63956 Sphingomonas sp. strain BAL40 α 316 472
U63959 Sphingomonas sp. strain BAL45 α 316 472
U63960 Sphingomonas sp. strain BAL46 α 316 472
U63962 Sphingomonas sp. strain BAL48 α 316 485
U63937 Sphingomonas sp. strain BAL5 α 316 470
U63998 Sphingomonas sp. strain SCB21 α 316 472
U78913 Strain R2A114 [D] α 316 ND
U78918 Strain R2A163 [D] α 316 ND
U78919 Strain R2A166 [D] α 316 ND
U78910 Strain R2A62 [D] α 316 ND
U78912 Strain R2A84 [D] α 316 ND
env.OCS138 [E] α 317 473
env.OCS154 [E] α 317 473
env.OCS180 [E] α 317 473
env.OCS53 [E] α 317 473
U70687 env.OM188 α 317 473
U70689 env.OM242 α 317 ND
U75649 env.SAR193 α 317 473
env.SAR222 α 317 ND
env.SAR239 α 317 ND
env.SAR258 α 317 ND
U75253 env.SAR407 α 317 473
U14583 Strain 307 α 318 474
U64009 Strain BAL32 α 318 474
U78909 Strain R2A57 [F] α 318 ND
U64019 Strain SCB42 α 318 472
env.OCS14 [J] α 324 480
env.OCS27 [J] α 324 483
U70683 env.OM75 [J] α 326 482
U78944 env.OCS116 [M] α 328 484
U70714 env.OM110 γ 328 ND
U78917 Strain R2A153 [M] α 328 ND
U65915 env.SAR276 δ 329 ND
env.OCS124 [N] α 330 486
M58793 Microscilla marina f 330 505
U20797 env.SAR202 x 331 488
U20798 env.SAR307 x 331 488
env.OCS2 γ 334 ND
U63945 Aeromonas sp. strain BAL19 γ 336 ND
U05570 Methylobacterium pelagicum γ 338 518
U14585 Strain 34-p α 338 520
U70702 env.OM241 γ 339 520
U70696 env.OM60 γ 339 520
L35470 env.SAR160 γ 339 520
U85887 Flavobacterium sp. strain A103 f 339 515
U85888 Flavobacterium sp. strain A265 f 339 515
D32219 Strain K189C γ 339 ND
U64010 Aeromonas sp. strain SCB33 γ 340 ND
X82144 Alteromonas luteoviolacea γ 340 522
X82147 Alteromonas rubra γ 340 522
M93352 Deleya aquamarina γ 340 521
env.OCS111 [Q] β 340 521
env.OCS7 [Q] β 340 521
L35469 env.SAR156 γ 340 521
U85873 Halomonas variabilis γ 340 521
U85872 Halomonas variabilis γ 340 521
U85871 Halomonas variabilis γ 340 521
L42618 Halomonas variabilis γ 340 521
L35540 Methylobacterium pelagicum γ 340 521
X72775 Methylomicrobium pelagicum γ 340 521
RDP Oceanospirillum kriegii γ 340 521
U63961 Rhodoferax sp. strain BAL47 β 340 461
U85854 Strain C079 γ 340 521
D32220 Strain K189B γ 340 ND
D32221 Strain unid gamma-proteobacterium γ 340 ND
U78920 Strain R2A9 [Q] γ 340 ND
X82143 Alteromonas espejiana γ 341 522
X67024 Alteromonas haloplanktis γ 341 522
L10938 Alteromonas macleodii γ 341 522
X82140 Alteromonas undina γ 341 523
env.OCS43 [R] β 341 ND
L35471 env.SAR166 γ 341 519
env.SAR470 γ 341 ND
env.SAR471 γ 341 523
U63946 Flavobacterium sp. strain BAL22 f 341 ND
U63938 Flavobacterium sp. strain BAL9 f 341 ND
X87339 Methylophaga thalassica γ 341 521
X98336 Pseudoalteromonas antarctica γ 341 522
U85857 Pseudoalteromonas sp. strain MB6-03 γ 341 522
U85858 Pseudoalteromonas sp. strain MB8-02 γ 341 522
U64012 Aeromonas sp. strain SCB35 γ 342 ND
U64020 Aeromonas sp. strain SCB43 γ 342 523
U63953 Alcaligenes sp. strain BAL37 β 342 ND
X82141 Alteromonas piscicida γ 342 524
U63943 Cytophaga sp. strain BAL17 f 342 ND
U78946 env.OCS181 γ 342 523
env.OCS44 γ 342 523
env.OCS5 γ 342 523
env.OCS66 β 342 523
env.OCS98 β 342 521
U70718 env.OM111 γ 342 ND
U70698 env.OM133 γ 342 ND
U70694 env.OM23 γ 342 523
U70697 env.OM93 γ 342 523
U63954 Flavobacterium sp. strain BAL38 f 342 ND
X82134 Pseudoalteromonas atlantica γ 342 523
X82136 Pseudoalteromonas carrageenovora γ 342 523
U85856 Pseudoalteromonas sp. strain IC006 γ 342 523
U85859 Pseudoalteromonas sp. strain IC013 γ 342 523
U85860 Pseudoalteromonas sp. strain MB6-05 γ 342 523
U85861 Pseudoalteromonas sp. strain SW08 γ 342 523
U85862 Pseudoalteromonas sp. strain SW29 γ 342 523
U85870 Pseudomonas sp. strain A177 γ 342 523
U85868 Pseudomonas sp. strain ACAM213 γ 342 523
U63942 Pseudomonas sp. strain BAL16 γ 342 ND
U63944 Pseudomonas sp. strain BAL18 γ 342 ND
U63947 Pseudomonas sp. strain BAL23 γ 342 ND
U64001 Pseudomonas sp. strain BAL24 γ 342 523
U85869 Pseudomonas sp. strain IC038 γ 342 523
U65012 Pseudomonas stutzeri γ 342 523
U26420 Pseudomonas stutzeri ZoBell γ 342 523
U78922 Strain R2A30 γ 342 ND
U63941 Zoogloea sp. strain BAL15 β 342 ND
U85897 Arthrobacter sp. strain MB6-07 h 343 504
U85896 Arthrobacter sp. strain MB8-13 h 343 516
U85893 Arthrobacter sp. strain MB90 h 343 503
U63940 Cytophaga sp. strain BAL13 f 343 ND
U70693 env.OM10 γ 343 523
U65912 env.SAR248 δ 343 499
U65908 env.SAR324 δ 343 499
U85890 Flavobacterium sp. strain ACAM123 f 343 519
U85889 Flavobacterium sp. strain IC001 f 343 519
U85891 Marine psychrophile IC025 f 343 519
X95640 Methylophaga thalassica γ 343 524
U78924 Strain R2A103 [S] f 343 ND
U64027 Alteromonas sp. strain SCB50 γ 344 525
X82061 Corynebacterium glutamicum h 344 507
U70706 env.OM156 β 344 525
U70707 env.OM180 β 344 525
env.SAR267 x 344 502
M63811 env.SAR92 γ 344 525
L35461 env.SAR86 γ 344 525
U85879 Psychrobacter glacincola γ 344 525
U85878 Psychrobacter glacincola γ 344 525
U85877 Psychrobacter glacincola γ 344 525
U85876 Psychrobacter glacincola γ 344 525
U85875 Psychrobacter sp. strain IC008 γ 344 525
U85874 Psychrobacter sp. strain MB6-21 γ 344 525
U78941 Strain R2A170 h 344 ND
U64026 Strain SCB49 f 344 ND
M58794 Microscilla sericea f 345 520
U64021 Antarcticum sp. strain SCB44 f 345 521
X80629 Corynebacterium glutamicum h 345 508
X84257 Corynebacterium glutamicum h 345 508
Z46753 Corynebacterium glutamicum h 345 508
M62796 Cytophaga lytica f 345 521
L10948 env.AGG13 f 345 521
L10945 env.AGG41 f 345 520
L10946 env.AGG58 f 345 519
env.SAR242 x 345 ND
M58775 Flectobacillus glomeratus f 345 521
X67022 Marinobacter hydrocarbonoclasticus γ 345 527
U63999 Marinobacter sp. strain SCB22 γ 345 526
X67025 Marinomonas vaga γ 345 526
U14586 Strain 301 f 345 521
U85883 Strain IC054 f 345 ND
U85885 Strain IC063 f 345 521
U85884 Strain IC066 f 345 ND
U78933 Strain R2A10 f 345 ND
U78935 Strain R2A132 f 345 ND
U78940 Strain R2A160 h 345 ND
U78939 Strain R2A54 f 345 ND
U64015 Strain SCB38 f 345 462
M61002 Vesiculatum antarcticum f 345 521
X82135 Alteromonas aurantia γ 346 527
X82137 Alteromonas citrea γ 346 527
X82137 Alteromonas denitrificans γ 346 528
U85895 Arthrobacter sp. strain IC044 h 346 507
U64000 Chromohalobacter sp. strain SCB23 γ 346 ND
U85844 Colwellia sp. strain IC068 γ 346 527
U85845 Colwellia sp. strain ICP11 γ 346 527
L42615 Deleya cupida γ 346 527
M93354 Deleya marina γ 346 527
L42616 Deleya pacifica γ 346 527
env.OCS178 β 346 527
U70699 env.OM182 γ 346 527
U70704 env.OM43 β 346 527
U70705 env.OM58 β 346 ND
U70695 env.OM59 γ 346 527
env.SAR226 x 346 504
env.SAR250 x 346 503
env.SAR259 x 346 501
env.SAR269 x 346 504
U85863 Marinobacter sp. strain IC022 γ 346 527
U85864 Marinobacter sp. strain IC032 γ 346 527
RDP Marinomonas communis γ 346 527
RDP Marinomonas vaga γ 346 527
RDP Oceanospirillum beijerincki γ 346 527
U85880 Psychrobacter immobilis γ 346 527
U78930 Strain R2A148 γ 346 ND
U85881 Strain IC051 f 346 522
U78931 Strain R2A173 γ 346 ND
U78924 Strain R2A44 γ 346 ND
U78927 Strain R2A86 γ 346 ND
U78928 Strain R2A88 γ 346 ND
U64017 Strain SCB40 f 346 489
U64022 Strain SCB45 γ 346 472
M62788 Cyclobacterium marinus f 347 522
L10944 env.AGG32 f 347 522
U70703 env.OM252 γ 347 528
env.SAR251 x 347 504
env.SAR432 h 347 503
U63955 Flavobacterium sp. strain BAL39 f 347 ND
U64023 Flexibacter sp. strain SCB46 f 347 523
X87755 Kytococcus sedentarius h 347 508
DEW Serratia rubidacea γ 347 527
U85906 Shewanella frigidimarina γ 347 528
U85882 Strain IC076 f 347 523
Z25522 Strain purple γ 347 524
U78929 Strain R2A113 γ 347 ND
U78932 Strain R2A5 f 347 ND
U85900 Arthrobacter sp. strain MB6-20 h 348 509
U63951 Azospirillum sp. strain BAL31 γ 348 ND
U85841 Colwellia sp. strain IC062 γ 348 529
U85842 Colwellia sp. strain IC064 γ 348 529
L10950 env.AGG53 γ 348 528
X54745 env.WHB461 γ 348 529
X54744 env.WHB462 γ 348 529
M58784 Flexibacter litoralis f 348 525
RDP Oceanospirillum jannaschii γ 348 529
M22365 Oceanospirillum linum γ 348 529
U85855 Pseudoalteromonas sp. strain MB8-11 γ 348 529
U85905 Shewanella frigidimarina γ 348 529
U85904 Shewanella frigidimarina γ 348 529
U85903 Shewanella frigidimarina γ 348 529
U85902 Shewanella frigidimarina γ 348 529
U85907 Shewanella gelidimarina γ 348 529
X81623 Shewanella putrefaciens γ 348 529
U85886 Strain ACAM210 f 348 524
X76334 Vibrio vulnificus γ 348 529
X74727 Vibrio vulnificus γ 348 529
X74726 Vibrio vulnificus γ 348 529
X76333 Vibrio vulnificus γ 348 529
Z22992 Vibrio vulnificus γ 348 529
X56582 Vibrio vulnificus γ 348 529
U64004 Xanthomonas sp. strain BAL27 γ 348 ND
U85846 Colwellia sp. strain ACAM179 γ 349 530
U85843 Colwellia sp. strain IC072 γ 349 530
M58770 Cytophaga marinoflava f 349 525
U70708 env.OM271 f 349 525
U70709 env.OM273 f 349 525
env.SAR196 n 349 520
L35504 Nitrospina gracilis δ 349 508
RDP Oceanospirillum maris γ 349 530
U64014 Strain SCB37 f 349 480
X74685 Photobacterium angustum γ 350 531
D25310 Photobacterium phosphoreum γ 350 532
X74687 Photobacterium phosphoreum γ 350 531
Z19107 Photobacterium phosphoreum γ 350 531
U85908 Shewanella hanedai γ 350 531
X82133 Shewanella putrefaciens γ 350 532
U63948 Shewanella sp. strain BAL25 γ 350 ND
U14582 Strain 90-P(gv)1 γ 350 531
U85849 Strain IC004 γ 350 531
U85852 Strain IC085 γ 350 531
U78923 Strain R2A37 [U] γ 350 ND
U14581 Strain S51-W(gv)1 γ 350 531
RDP Vibrio marinus γ 350 531
X82134 Vibrio marinus γ 350 533
U85847 Colwellia sp. strain IC035 γ 351 532
U85853 Strain IC067 γ 351 532
U64018 Strain SCB41 f 351 ND
L35468 env.SAR145 γ 352 532
M96493 Nitrococcus mobilis β 352 533
X82132 Shewanella hanedai γ 352 533
env.OCS155 h 353 509
U70710 env.OM1 h 353 509
U70711 env.OM231 h 353 ND
U41450 env.OCS307 fb 353 535
U85851 Strain IC059 γ 353 534
X56756 Vibrio alginolyticus γ 353 535
U34043 env.SAR406 fb 354 536
D55729 Planococcus okeanokoites 1 354 534
U85899 Planococcus sp. strain IC024 1 354 534
X56578 Vibrio harveyi γ 354 536
X56581 Vibrio natriegens γ 354 536
D25308 Photobacterium histaminum γ 355 537
D25309 Photobacterium leiognathi γ 355 537
X62172 Planococcus citreus 1 355 535
U85898 Planococcus sp. strain MB6-16 1 355 535
D83367 Staphylococcus halodurans 1 355 534
Z26896 Staphylococcus halodurans 1 355 536
X66100 Staphylococcus halodurans 1 355 536
L37600 Staphylococcus halodurans 1 355 538
U78937 Strain R2A180 [V] 1 355 ND
X56575 Vibrio campbellii γ 355 537
env.SAR272 x 356 514
U14584 Flectobacillus sp. strain S38-W(gv)1 f 356 537
U78938 Strain R2A161 1 356 ND
env.SAR256 x 357 514
X70642 Listonella pelagia γ 357 ND
U85867 Marinobacter sp. strain IC065 γ 357 ND
X74722 Listonella pelagia γ 358 539
X74686 Photobacterium leiognathi γ 358 539
X74691 Vibrio alginolyticus γ 358 539
X74690 Vibrio alginolyticus γ 358 539
X74692 Vibrio campbellii γ 358 539
X74702 Vibrio fischeri γ 358 539
X70640 Vibrio fischeri γ 358 538
X74706 Vibrio harveyi γ 358 539
X74710 Vibrio mediterranei γ 358 539
X74714 Vibrio natriegens γ 358 539
X74716 Vibrio nereis γ 358 539
X74717 Vibrio nigripulchritudo γ 358 539
X74719 Vibrio orientalis γ 358 539
U64016 Vibrio sp. strain SCB39 γ 358 524
U64024 Vibrio sp. strain SCB47 γ 358 523
Z31657 Vibrio splendidus γ 358 538
X74724 Vibrio splendidus γ 358 539
U64013 Flexibacter sp. strain SCB36 f 360 527
U64006 Vibrio sp. strain BAL29 γ 360 ND
U64007 Vibrio sp. strain BAL30 γ 360 526
X62171 Marinococcus halophilus 1 363 545
X90835 Marinococcus halophilus 1 374 556
U85901 Halobacillus sp. strain MB6-08 1 392 574

a Lengths were calculated for bacteria and plastids isolated from seawater, as well as for SSU rDNAs cloned from community DNA. Published sequences were obtained from the RDP, GenBank, and ARB databases. We also included unpublished sequences of genes cloned from environmental DNA from the Oregon coast (12, 15) (prefix env.OCS) and the Sargasso Sea (prefix env.SAR). Letters in brackets correspond to the peaks in Fig. 1. α, β, γ, and δ, alpha, beta, gamma, and delta subdivisions, respectively, of the Proteobacteria; f, Flexibacter, Bacteroides, and Cytophaga phylum; l, low-G+C gram-positive phylum; h, high-G+C gram-positive phylum; x, Chloroflexus and Herpetosiphon phylum; c, cyanobacteria; fb, Fibrobacter phylum; n, Nitrospira phylum; ND, not determined. 

Analyses of coastal bacterioplankton diversity.

An example of an LH-PCR electropherogram is shown in Fig. 1. It shows the length heterogeneity of domain A for PCR products obtained directly from DNA extracted from seawater subsample 2. The 23 peaks are labeled A through W and correspond to amplicons with varying lengths in domain A. Organisms that produce amplicons corresponding in size to these peaks were identified by reference to a clone library prepared from the same seawater sample. These clones are indicated in Table 1 by the prefix “env.OCS.”

We were able to assign an OCS gene clone or R2A cellular clone to peaks from natural community DNA (Table 1). The domain A peaks of sizes 317 bp (E), 318 bp (F), 319 bp (G), and 321 bp (H) correspond to the sizes of SSU rDNAs of plastid origin (12). The domain A peak of 317 bp corresponds to sizes of both alpha-Proteobacteria and plastids (12). There is strong evidence that the 317-bp domain A peak corresponds mainly to alpha-Proteobacteria, since the ratio between the integrated fluorescence of peak E and all bacterial peaks is approximately the same for samples whether they were filtered through 0.8-μm-pore-size polycarbonate membranes, which removed the other plastid peaks, or not (12). Figure 2 presents electropherogram data in a histogram format. The data from Fig. 1 are shown in panel A, with error bars shown to represent the standard deviations for triplicate PCRs from the same DNA sample. Here, as in other measurements, we found the method to be highly reproducible.

FIG. 2.

FIG. 2

Comparison between LH-PCR and SSU rDNA clone library. The figure shows the percentage of integrated fluorescence of each individual domain A produced by the optimized LH-PCR (final product concentration, <1.5 nM) from genomic DNA isolated from subsample 2 (A) or nearly full-length SSU rDNA PCR amplicons used to construct the clone library (B). The x axis represents the size of domains in base pairs, estimated by comparison to the size standard GS2500 (Applied Biosystems Inc.). The relative abundance of clones recovered in the OCS clone library classified by the length of domain A is shown in panel C. Error bars each represent one standard deviation from the mean of triplicate reactions.

Bias by PCR.

The possibility that a kinetic bias caused by template reannealing could occur during the amplification of domain A from bacterioplankton samples was investigated by examining the relationship between the final concentration of products obtained and the relative frequency of dominant genes in the population. A portion of this analysis is provided in Fig. 3, which shows the relative frequency of the 317-bp fragment (alpha-Proteobacteria and prymnesiophyte plastids) as a function of the total product molarity. The prediction for the kinetic bias effect is that the proportion of the dominant peak (the percentage of integrated fluorescence) should decrease with increasing product molarity, as observed in Fig. 3. This prediction assumes that the dominant peaks are composed primarily of genes of one or a few types. The final concentrations of product amplicons for the reactions used for Fig. 3 varied for the two samples of DNA isolated independently from the same water sample (subsamples 1 and 2) and also varied according to the number of cycles used for the amplification (12 to 25 cycles for subsample 1 and 12 to 18 cycles for subsample 2). We considered the possibility that biases caused by primer selection, which are dependent on the number of cycles, might have caused the bias. A plot of the same data shown in Fig. 3 with the number of amplification cycles replacing the final product concentrations on the abscissa revealed no relationship (data not shown). The observed bias is in accord with predictions for the kinetic bias effect.

FIG. 3.

FIG. 3

An example of PCR bias fitting the kinetic model, with PCR amplicons obtained from natural community DNA. The molar ratio of the dominant fragment (317 bp) to total products is plotted as a function of the final product concentration. Primer set A was used for the amplification from environmental DNA subsamples 1 (solid circles) and 2 (open circles).

Figure 1B shows the results of an optimized domain A LH-PCR analysis (i.e., the final product concentration was lower than 1.5 nM) of the full-length PCR products that were used to prepare the clone library. The template for this reaction was PCR amplicons obtained with the 27F-1542R primers from the same natural DNA sample that was used as template for the LH-PCR in Fig. 1A. There is a significant difference between the profiles in Fig. 1A and B. The comparison between panels A and B shows that the relative contributions of some peaks (A, F, G, K and L, N, P, and Q) increased when the full-length PCR amplicons were used as a template instead of genomic DNA, while other peaks (M, S, T, and U and V) decreased. We noted that the relative proportion of peak F (319 bp) is higher than that of peaks D and E (317 and 318 bp) in the PCR amplicons, although both are major peaks that contribute significantly to the total population of molecules.

It is evident that the reamplification reactions provided reproducible results and that therefore the difference between Fig. 1A and 1B could be attributed to the original PCR used to prepare amplicons for clone library construction. To explain these results, we reasoned that some of the differences between Fig. 1A and B (i.e., the decrease of the dominant peak at 317 bp) might be attributed to bias caused by the kinetic (template reannealing) effect, since 35 cycles of amplification were used for the PCR, and final product concentrations were greater than 10 nM.

Figure 1C shows the distribution of clones from the gene clone library, discriminated by the length of the domain A of their SSU rDNA, for comparisons to Fig. 1A and B. It is evident that the profiles in Fig. 1B and C are very different. The causes of these differences are uncertain but probably include a combination of three sources of error. The random error resulting from the sample size (number) of clones retrieved and identified in the library imposed limitations on the expected correspondence between LH-PCR results and clone library gene frequencies. Systematic biases by LH-PCR or by one of the steps of the cloning process are also plausible explanations for the observations.

Comparison of domain A with domain B.

In general, there was good agreement between the community structure estimated by LH-PCR for domains A and B (data not shown). The main difference between LH-PCR for domains A and B was the resolution of different peaks by the Genescan software. Genescan resolved more peaks in the analysis of domain A and tended to merge domain B peaks, especially peaks for larger fragments (peaks > 520 bp). Some adjacent peaks of domain A were also merged in some of the electropherograms (peaks C and D, E and F, K and L, and Q and R).

Phylogenetic composition of the community.

The relative proportions of LH-PCR peaks conform to previous observations that alpha-Proteobacteria dominate SSU rDNA clone libraries of surface samples. Peaks A to E and K to M, which correspond to alpha-Proteobacteria and plastids, respectively (Table 1), represent about 65% of the total fluorescence. Peaks F, G, and L correspond to plastid sequences (12). Most of the remaining peaks do not correspond to coherent phylogenetic groups when reference is made to all SSU rDNA sequences of bacteria isolated or environmental clones from seawater. However, most of the peaks had a corresponding isolate or environmental SSU rDNA clone from the same seawater sample. Peaks P and Q, which represented about 7% of the total fluorescence, corresponded to the sizes of beta-, gamma-, and delta-Proteobacteria, Flexibacter-Bacteroides-Cytophaga, and high-G+C gram-positive bacteria, many of which are cultivated strains. Peak S represented about 5% of the total fluorescence and corresponded to the sizes of previously cultivated members of the gamma-Proteobacteria. Finally, peaks T to V represented about 9% of the total fluorescence and corresponded to the sizes of several phylogenetic groups.

DISCUSSION

The adoption of molecular techniques for assessing microbial diversity has engendered a far-reaching appreciation of the importance of uncultured microbes but has also led to concerns about the limitations of the new methodologies. The approach we employed here, LH-PCR, was designed to address some of these concerns in the context of the study of complex natural communities.

We found that gene frequencies measured by PCR amplification can be highly reproducible. We also observed a bias that selectively reduced the relative frequency of a dominant size class of fragments as a function of increasing final product concentrations, which can be explained by the template reannealing bias described by Suzuki and Giovannoni (16). This bias is caused by the fact that, as the reaction progresses, amplicons increase in concentration and primers decrease in concentration. At a certain point, amplicons should reanneal, inhibiting the primers from annealing and stalling the reaction. Assuming that different SSU rDNAs do not cross-reanneal, genes with higher initial concentrations in the original sample should experience template reannealing at lower combined product concentrations—which for reactions run at the same initial conditions should be dependent on the number of replication cycles—than genes with lower initial concentrations. Therefore, in reactions experiencing this bias, dominant species should be underrepresented and rarer species should be overrepresented. The fact that the observed trend was related to combined product concentration but not to the number of amplification cycles supports the idea that this bias is due to reannealing kinetics and not to some other form of bias, such as primer selection. This reannealing bias became significant above product concentrations of about 2 nM, and therefore, we recommend that amplification reactions be stopped before reaching this value.

Differences between the LH-PCR electropherograms obtained from natural community DNA and those obtained from the reamplification of full-length SSU rDNA amplicons indicate that the PCR may significantly bias the composition of clone libraries. However, such biases do not appear to occur randomly but rather are systematic. The shift in the dominance from the peak of 317 bp to the peak of 319 bp contradicts our previous expectation (16) that reannealing bias would lead amplicons originating from different templates to reach similar concentrations. A possible explanation for this discrepancy is the fact that each of the LH-PCR peaks represents SSU rDNAs originating from several different organisms. Reannealing between domains originating from different organisms could explain the observed shift in dominance between the two peaks. If the degree of similarity between the sequences with a domain A length of 317 bp were high enough for PCR amplicon cross-hybridization and kinetic inhibition, while the degree of similarity between sequences with a domain A length of 319 bp were low enough to not lead to cross-hybridization, one could envision that each of the 319-bp amplicons would experience lower levels of kinetic inhibition than the 317-bp amplicons. The average similarity among four SSU rDNA clones with a 317-bp domain A was 0.93 (0.89 to 1.00), while the degree of similarity between two SSU rDNA clones with a 319-bp domain A was 0.85, supporting this hypothesis. Another hypothesis which might explain the observed peak shift would be a large difference in the degree of diversity among the organisms corresponding to each of the peaks. Template reannealing inhibition should theoretically be lower for peaks with more gene types or peaks lacking a dominant gene type. This hypothesis cannot be tested with the data included in the current analysis.

Uncertainties about the numbers of ribosomal operons in different bacterioplankton species and differences in the relative sizes of genomes preclude the extrapolation from gene frequencies to cell abundance. Nevertheless, relative gene frequencies offer some advantages as a measurement for assessing the composition of natural microbial communities. In particular, rDNA frequency histograms (electropherograms) should be relatively insensitive to short-term variation in growth rates, which may affect rRNA abundance significantly under some circumstances (7).

LH-PCR is a promising method for the analysis of natural microbial populations. The main advantages of LH-PCR are that it surveys relative gene frequencies within complex mixtures of DNA, is reproducible, requires small sample sizes, and can be performed with many samples simultaneously. Furthermore, some of the size classes emerging from LH-PCR analyses can be related at the group level to environmental rDNA sequences. However, overlapping size classes leave ambiguities that require further analyses to resolve. The relative proportions of the electropherogram peaks from seawater are in agreement with previous findings that alpha-Proteobacteria members are dominant components of clone libraries constructed from DNA extracted from surface seawater. The observation that most SSU rDNAs of organisms cloned or cultivated from the same water sample have sizes that correspond to the peaks in the LH-PCR electropherograms also supports the validity of the method.

The main technical problem associated with LH-PCR is the accuracy of peak detection, especially when longer domains are used. Improvements in automated DNA sequencers and in Genescan software may increase the accuracy of the method for longer domains. This problem notwithstanding, domain B was useful to confirm the results obtained with domain A and, in some cases, to differentiate between phylogenetic groups with identical sizes in domain A, like the alpha-Proteobacteria and prymnesiophyte plastids.

The attributes of LH-PCR make it useful for quick assessments of the diversity of natural microbial communities for comparative purposes, for experimental designs that involve the manipulation of natural microbial communities (15), and for experiments, such as those we describe here, aimed at investigating the properties of PCR in applications employing complex mixtures of gene homologs.

ACKNOWLEDGMENTS

We are grateful to the staff of the Oregon State University Center for Gene Research Central Services Laboratory, and particularly to Anne-Marie Girard, for technical assistance.

This research was supported by grant OCE-9618530 from the National Science Foundation.

REFERENCES

  • 1.Amann R I, Krumholz L, Stahl D A. Fluorescent-oligonucleotide probing of whole cells for determinative, phylogenetic, and environmental studies in microbiology. J Bacteriol. 1990;172:762–770. doi: 10.1128/jb.172.2.762-770.1990. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Bruce K. Analysis of mer gene subclasses within bacterial communities in soils and sediments resolved by fluorescent-PCR-restriction fragment length polymorphism profiling. Appl Environ Microbiol. 1997;63:4914–4919. doi: 10.1128/aem.63.12.4914-4919.1997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Giovannoni S J. The polymerase chain reaction. In: Stackebrandt E, Goodfellow M, editors. Nucleic acid techniques in bacterial systematics. New York, N.Y: John Wiley & Sons, Inc.; 1991. pp. 177–203. [Google Scholar]
  • 4.Giovannoni S J, Britschgi T B, Moyer C L, Field K G. Genetic diversity in Sargasso Sea bacterioplankton. Nature. 1990;345:60–63. doi: 10.1038/345060a0. [DOI] [PubMed] [Google Scholar]
  • 5.Giovannoni S J, DeLong E F, Olsen G J, Pace N R. Phylogenetic group-specific oligodeoxynucleotide probes for identification of single microbial cells. J Bacteriol. 1988;170:720–726. doi: 10.1128/jb.170.2.720-726.1988. . (Erratum, 170:2418.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Giovannoni S J, DeLong E F, Schmidt T M, Pace N R. Tangential flow filtration and preliminary phylogenetic analysis of marine picoplankton. Appl Environ Microbiol. 1990;56:2572–2575. doi: 10.1128/aem.56.8.2572-2575.1990. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Kramer J G, Singleton F L. Measurement of rRNA variations in natural communities of microorganisms on the southeastern U.S. continental shelf. Appl Environ Microbiol. 1993;59:2430–2436. doi: 10.1128/aem.59.8.2430-2436.1993. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Liesack W, Stackebrandt E. Occurrence of novel groups of the domain Bacteria as revealed by analysis of genetic material isolated from an Australian terrestrial environment. J Bacteriol. 1992;174:5072–5078. doi: 10.1128/jb.174.15.5072-5078.1992. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Liu W T, Marsh T L, Cheng H, Forney L J. Characterization of microbial diversity by determining terminal restriction fragment length polymorphisms of genes encoding 16S rRNA. Appl Environ Microbiol. 1997;63:4516–4522. doi: 10.1128/aem.63.11.4516-4522.1997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Maidak B L, Olsen G J, Larsen N, Overbeek R, McCaughey M J, Woese C R. The RDP (Ribosomal Database Project) Nucleic Acids Res. 1997;25:109–111. doi: 10.1093/nar/25.1.109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Mullins T D, Britschgi T B, Krest R L, Giovannoni S J. Genetic comparisons reveal the same unknown bacterial lineages in Atlantic and Pacific bacterioplankton communities. Limnol Oceanogr. 1995;40:148–158. [Google Scholar]
  • 12.Rappé M S, Suzuki M T, Vergin K L, Giovannoni S J. Phylogenetic diversity of ultraplankton plastid small-subunit rRNA genes recovered in environmental nucleic acid samples from the Pacific and Atlantic coasts of the United States. Appl Environ Microbiol. 1998;64:294–303. doi: 10.1128/aem.64.1.294-303.1998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Sardelli A. Plateau effect—understanding PCR limitations. Amplifications. 1993;9:1. ,3,5,6. [Google Scholar]
  • 14.Strunk, O., O. Gross, B. Reichel, M. May, S. Hermann, N. Stuckman, B. Nonhoff, M. Lenke, A. Ginhart, A. Vilbig, T. Ludwig, A. Bode, K.-H. Schleifer, and W. Ludwig. ARB: a software environment for sequence data. Submitted for publication.
  • 15.Suzuki, M. T. The effect of protistan bacterivory on bacterioplankton diversity. Submitted for publication.
  • 16.Suzuki M T, Giovannoni S J. Bias caused by template annealing in the amplification of mixtures of 16S rRNA genes by PCR. Appl Environ Microbiol. 1996;62:625–630. doi: 10.1128/aem.62.2.625-630.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Suzuki M T, Rappé M S, Haimberger Z W, Winfield H, Adair N, Ströbel J, Giovannoni S J. Bacterial diversity among small-subunit rRNA gene clones and cellular isolates from the same seawater sample. Appl Environ Microbiol. 1997;63:983–989. doi: 10.1128/aem.63.3.983-989.1997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Ward D M, Weller R, Bateson M M. 16S rRNA sequences reveal numerous uncultured microorganisms in a natural community. Nature. 1990;345:63–65. doi: 10.1038/345063a0. [DOI] [PubMed] [Google Scholar]

Articles from Applied and Environmental Microbiology are provided here courtesy of American Society for Microbiology (ASM)

RESOURCES