DNA hybridization to compare species compositions of natural bacterioplankton assemblages

Vol. 56, No. 3 APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Mar. 1990, p. 739-746 0099-2240/90/030739-08$02.00/0 Copyright © 1990, American Society for Microbiology DNA Hybridization To Compare Species Compositions of Natural Bacterioplankton Assemblagest SANGHOON LEElt* AND JED A. FUHRMAN2 Marine Sciences Research Center, State University of New York, Stony Brook, New York 11794-5000,' and Marine Biology Research Section, Department of Biological Sciences, University of Southern California, Los Angeles, California 90089-03712 Received 13 July 1989/Accepted 12 December 1989 culturing. However, its application to natural bacterioplankton communities is still at the stage of testing. An extension of this approach is single-cell identification with 16S rRNAbased probes (9). For this technique, one must devise and produce unique probes to identify each type of cell. Both of these techniques require extensive time and labor to analyze a few samples for a comprehensive communitywide species Much research has been performed to measure and combacterial activities in various marine environments, with the goal of understanding the roles of marine microorganisms. However, bacterioplankton assemblages have usually been treated as a single group, and most of the studies were done at the assemblage level. Although assemblagelevel measurements have their own virtue in the study and comparison of bacterioplankton processes, bulk parameters alone, without proper attention given to the identities of the components, do not allow us to understand natural variabilpare comparison. Complete cataloging of indigenous bacterial species is an ultimate answer to the question of "how much of what kind of bacteria are in the water column?" However, there is a need for a rapid and simple technique that does not necessarily identify individual species but that can differentiate bacterial communities in terms of their constituents. This technique is particularly suitable for answering basic questions about the variability of species composition. It is unclear how much species compositions of bacterioplankton assemblages vary over time and space. The measured bulk parameters are of more use in understanding the various aspects of marine microbial ecology if they are combined with information regarding the variability of the species composition. The technique also allows us to choose typical samples for more detailed and time-consuming species composition analyses. DNA hybridization has been used widely to investigate the phylogenetic relatedness of bacterial strains (8, 22, 29, 30). Palleroni et al. (26) reported that DNA-DNA hybridization has a high resolution for closely related organisms whose differences could not be distinguished by DNA-rRNA hybridization. They often found 0% DNA-DNA homology between Pseudomonas species, although DNA-rRNA homologies ranged from 80 to 95%. Because of the wellconserved rRNA genome (rDNA) compared with the con- ity. One of the major reasons that little is known about species composition and its variation is the difficulty of conventional identification. Conventional identification of bacteria needs pure cultures. However, most marine planktonic bacteria are noncultivable (CFU, <1% of total counts [11]), despite the observation that many (40 to 90% of total counts) can be shown to be metabolically active by microautoradiography (13). Comparison of the base sequences of rRNA has become a powerful tool to infer the evolutionary pathways of procaryotes (reference 33 and references therein). Recently, a procedure that uses rRNA sequence homology was proposed for defining and enumerating the components of a mixed natural population of bacteria (25). It should allow the definition of a population, both quantitatively (species diversity) and qualitatively (phylogenetic relatedness), without * Corresponding author. t Contribution no. 709 from the Marine Sciences Research Center, State University of New York at Stony Brook. t Present address: Department of Biological Sciences, University of Southern California, Los Angeles, CA 90089-0371. 739 Downloaded from http://aem.asm.org/ on July 23, 2020 by guest Little is known about the species composition and variability of natural bacterial communities, mostly because conventional identification requires pure cultures, but <1% of active natural bacteria are cultivable. This problem was circumvented by comparing species compositions via hybridization of total DNA of natural bacterioplankton communities for the estimation of the fraction of DNA in common between two samples (similarity). DNA probes that were labeled with 35S by nick translation were hybridized to filter-bound DNA in a reciprocal fashion; similarities (in percent) were calculated by normalizing the values to self-hybridizations. In tests with DNA mixtures of pure cultures, the experimentally observed similarities agreed with expectations. However, reciprocal similarities (probe and target reversed) were often asymmetric, unlike those of DNA from single strains. This was due to the relative complexity and G+C content of DNA, which provided a means to interpret the asymmetry that was occasionally observed in natural samples. Natural bacteria were collected by filtration from Long Island Sound (LIS), N.Y., the Caribbean and Sargasso seas, and a coral reef lagoon near Bermuda. The samples showed similarities of <10 to 95%. The LIS and Sargasso and Caribbean sea samples were 20 to 50% similar to each other. The coral reef sample was <10% similar to the others, indicating its unique composition. Seasonality was also observed; an LIS sample obtained in the autumn was 40% similar to two LIS samples obtained in the summer; these latter two samples were 95% similar. We concluded that total DNA hybridization is a rapid, simple, and unbiased method for investigating the variation of bacterioplankton species composition over time and space, avoiding the need of culturing. 740 APPL. ENVIRON. MICROBIOL. LEE AND FUHRMAN 0.1 point mutation per sequence position Ref. % G+C Ref. [Pseudavoms testosteronA 34 NJ "Crane Neck Isolate Ea * .ILN 150" * * N ProteLs wl/garis 7 40 19 Escherichia coli 6 50-51 24 [AgAobacteriwn &refacir,s Caulobacter VOC1 Ceulobacter "VC5O Caulobacter halobacteroides [F/avobacteni.rn heparimru 34 NJ * 64-67 28 * 64-67 28 * 64-67 28 32 NJ * * 46 "Crane Neck Isolate C" FIG. 1. Phylogenetic tree based on 16S rRNA base sequence homology. The phylogenetic distance between two species is proportional to the sum of the lengths of the horizontal lines linking the two. Strains within brackets were not used (NU) in this study. ND, Not determined. This study is the source for the references with asterisks. servation of the total genome (10), rRNA (or rDNA) hybridization is generally used for distantly related organisms up to the kingdom level (17), whereas DNA hybridization is used for relatively closely related organisms, e.g., at the species level (16). We adapted the DNA hybridization technique for use with mixed populations, exploiting this specificity of DNA hybridization. The degree of commonness of identical or very closely related bacterial strains shared between natural bacterioplankton assemblages was determined by community DNA hybridization. The natural samples showed a full range (from <10% to almost 100%) of similarities, with the results being related to sampling locations and seasons. In tests of this method with pure cultures, we found that the results reflect not only the amount of shared DNA but also the relative complexity, the percent molar G+C content of DNAs of two communities that were compared, or both. This information was useful in interpreting the results from natural samples. MATERIALS AND METHODS Bacterial strains and their phylogenetic relations. Bacterial strains used in this study were Caulobacter sp. strains VC1 and VC5; Caulobacter halobacteroides ATCC 15269; Escherichia coli cell line K-12; and Proteus vulgaris ATCC 13315, "LN150," "Crane Neck isolate C" ("CN-C"), and "Crane Neck isolate E" ("CN-E"). The caulobacters were from east Pacific deep hydrothermal vents (strains VC1 and VC5) or filtered seawater (C. halobacteroides) and were provided by J. Poindexter. Strain LN150 was isolated from the Pacific Ocean near San Diego by A. Carlucci. Strains CN-C and CN-E were isolated from Long Island Sound (LIS) by members of our laboratory. Strains CN-C and CN-E are oligotrophs that are capable of growth in unsupplemented filtered seawater. The strains used in this study are listed in Fig. 1, which is a phylogenetic tree built with a computer program written by G. Olsen based on 16S rRNA base sequence homology (23). Strains were sequenced by members of our laboratory by the reverse transcription method (18). The G+C contents reported in this study were determined by high-pressure liquid chromatography (21). Other strains not used in this study are included in the phylogenetic tree in Fig. 1 to help understand the relative positions. Preparation of DNA from cultured cells. Cells were grown in CPM medium (seawater filtered through glass fiber filters; diluted to 80% with deionized water, 0.05% Bacto-Peptone [Difco Laboratories, Detroit, Mich.], and 0.05% Casamino Acids [Difco]; and autoclaved) for the marine isolates or in LB medium (20) for the other isolates. Cells were harvested by centrifugation (15 min, 10,000 x g), and cell pellets (volume, about 100 to 200 IL) were rinsed with 1 ml of ice-cold 10% sucrose-50 mM Tris (pH 8.0) to remove the salts and were then pelleted again. Cells were suspended in 1 ml of 10% sucrose-50 mM Tris (pH 8.0), 0.3 ml of 0.5 M EDTA (pH 8.0) was added and mixed thoroughly, and 0.6 ml of 10% sodium dodecyl sulfate (SDS) was added. The tube was inverted actively until the liquid became clear. The liquid was extracted once with phenol (pH 8.0), three times with phenol-chloroform-isoamyl alcohol (24:6:1, by volume; pH 8.0), and once with chloroform-isoamyl alcohol (24:1, by volume). Ammonium acetate (final concentration, 2.5 M) and ethanol (two times the volume of the total aqueous phase) were added to the final aqueous phase and precipitated overnight at -20°C. The DNA was pelleted by centrifugation (10 min, 12,000 x g, 40C), rinsed with 70% ethanol, briefly dried, and dissolved in TE (10 mM Tris, 1 mM EDTA [pH 8.0]). The DNA was quantified by bisbenzimide (Hoechst 33258 dye; Sigma Chemical Co., St. Louis, Mo.) fluorometry (27). Preparation of probe and target DNA. Probe DNA was Downloaded from http://aem.asm.org/ on July 23, 2020 by guest I 64 VOL. 56, 1990 BACTERIAL COMMUNITY COMPARISON BY DNA HYBRIDIZATION TABLE 1. Dates and locations of natural bacterioplankton samples Sample (moldaylyr) Location Autumn LIS 11/18/87 Summer LIS-1 6/16/88 Summer LIS-2 6/23/88 Caribbean Sea 3/2/88 Sargasso Sea Coral reef lagoon 8/27/87 8/26/87 40058' N, 73°09' W, Crane Neck, Long Island, N.Y. 40058' N, 73009' W, Crane Neck, Long Island, N.Y. 40058' N, 73009' W, Crane Neck, Long Island, N.Y. 12039' N, 61047' W, 50 km northwest of Grenada 10 km southeast of Bermuda Coral reef, North Shore, Bermuda served similarities (SO) were compared with theoretical ones. When two mixtures had a species in common, the species contributed x (in percent) to one mixture and y (in percent) to another, whereas x may not have been the same as y. The smaller of the two, x and y, was used for the calculation of the theoretical similarity (S,); i.e., S, is the least common fraction between two mixtures. When two mixtures had n species in common, the calculation of S, (in percent) was: n St = pi, where n is the number of common species, and pi (in percent) is the percentage contribution of the ith common species to either of the two mixtures, whichever one was smaller. Hybridization kinetics. The progress of hybridization with incubation time was examined to understand the nature of the hybridization under the conditions we set. Hybridization kinetics of three different DNA types were examined: simple DNA (single-strain DNA), complex DNA (mixture of equal amounts of DNA from six different strains), and DNA of a natural bacterioplankton community (from an oligotrophic Pacific Ocean sample; 34049' N, 124008' W). Probes were made of the three types of DNA and hybridized to the DNA from which the probe was prepared (control) as well as to the other DNA samples. Four replicate sets were incubated for each DNA type; and one set was removed and processed after 6, 12, 24, and 48 h of incubation. The similarities at each time were calculated. Experiments with natural bacterioplankton. Naturally occurring near-surface marine planktonic bacteria were collected with Niskin bottles (Caribbean and Sargasso sea samples) or plastic buckets and carboys from various places and times (Table 1). Filtration and DNA extraction followed the protocol of Fuhrman et al. (14), with slight modifications. Collected seawater was immediately pressure filtered from a stainless steel pressure vessel (20 liters; Gelman Sciences, Inc., Ann Arbor, Mich.) to which two stainless steel filter holders (142-mm diameter; Fisher Scientific Co., Pittsburgh, Pa.) were attached in-line in series. With gentle pressure (ca. 70 kPa), the water was first prefiltered through glass fiber filters (type AE; Gelman) to remove larger particles, and bacterial cells were collected on 0.22-,um-pore-size filters (Durapore; Millipore Corp., Bedford, Mass.). The filtrations were done in a relatively short period of time (a total of 1 to 4 h after sample collection, depending on the volume being filtered), to prevent a possible change of the species compo- Downloaded from http://aem.asm.org/ on July 23, 2020 by guest labeled with [a-35S]dATP or [a-35S]dCTP (Dupont, NEN Research Products, Boston, Mass.) with a nick-translation kit (Bethesda Research Laboratories, Inc., Gaithersburg, Md.). The probes were purified with a nucleic acid cleaning kit (NENSORB; Du Pont Co., Wilmington, Del.), to remove unincorporated nucleotides and enzymes, and were dried with a stream of N2. The dried probes were reconstituted in 50 ,l of deionized water and denatured for 10 min in a boiling water bath before they were added to the hybridization solution. Target DNA, which was unrestricted, unlabeled total DNA, was loaded onto a nylon membrane (0.45-,um pore size; Magnagraph; Micron Separations Inc., Westboro, Mass.) in a series of dots (the amount ranged from 100 to 500 ng, each within a 5- by 5-mm square). Hybridization. After target DNA loading, the membrane was dried at room temperature, denatured (3 min in 1.5 M NaCl-0.2 N NaOH; 2 min in 0.5 M Tris [pH 7.5]; 2 min in 5x SSC [20x SSC is 3 M NaCl plus 0.3 M trisodium citrate; pH 7.0]), and baked at 80°C for 1.5 h. Filters were prehybridized at 64 to 69°C (mostly at 68°C) for 6 to 8 h with denatured salmon sperm DNA (1 mg/ml) in hybridization solution (6x SSC, 1% SDS, Sx Denhardt solution [50x Denhardt solution is 1% Ficoll {Pharmacia Fine Chemicals, Piscataway, N.J.}, 1% polyvinylpyrrolidone, and 1% bovine serum albumin]), and hybridized at the same temperature with probe added (concentration, 0.5 to 2 ,ug/ml; 2 x 107 to 10 X 107 dpm/ml) for 24 to 28 h. The filters were washed twice in each of 2x SSC (5 min at room temperature), 0.5% SDS in 2x SSC (30 min at the hybridization temperature), and 0.1x SSC (5 min at room temperature). The filters were dried at room temperature, and the squares (5 by 5 mm) of the filter containing each DNA dot were cut. The radioactivities of the DNA dots were counted by liquid scintillation with 2 ml of scintillation fluor (Ecoscint; National Diagnostics, Manville, N.J.). The blank was a filter square without DNA. Eucaryotic (calf thymus) DNA was also tested as a blank. The target dots of DNA from which the probe was prepared served as a control (control target); i.e., the radioactivity on these dots was defined as 100% similarity. For the calculation of observed similarity (SO), hybridization signals (in disintegrations per minute) from unidentical DNA dots (sample target) were normalized to the control: S, (%) = 100 x [(R - Rb)l(RC - Rb)], where R is the radioactivity (in disintegrations per minute) from the target DNA dot and the subscripts s, b, and c denote, respectively, sample DNA, blank (no DNA), and control DNA. Degree of cross-hybridization among cultured strains. The degree of cross-hybridization under the hybridization stringency condition we set was examined. DNAs of C. halobacteroides LN150, CN-C, and CN-E were used for the estimation of the cross-hybridization between distantly related strains. The degree of cross-hybridization between closely related strains was examined with DNA from the three caulobacters, E. coli, and P. vulgaris. The degree of crosshybridization was calculated in the same way as the S,, calculation described above. Hybridization of mixed DNA of defined composition. The DNAs from single strains were mixed in various manners to make mixtures with diverse compositions and complexities. Probes were made from the mixtures and hybridized to the other mixtures by using the same methods and conditions described above. Single-strain DNA was also hybridized to mixture DNA. For tests of reciprocity (reciprocal hybridizations), the probe and target were switched (i.e., probe A to target B and probe B to target A), and the reverse similarity was compared with the forward one. Experimentally ob- 741 APPL. ENVIRON. MICROBIOL. LEE AND FUHRMAN 742 4- TABLE 3. Comparison of SO with S, from hybridizations of single-strain DNA probe to mixture DNA targeta (%/0) 0) 100 A Mean ± SEM S0 (%) of the following probes: S, (%) o 4-' 04, 80 10 17 25 33 50 70 .0 a to 0o -o- single-strain DNA ---e----. mixture DNA (6 sp) // iEN 40 O ° 4- co 60 Co v1)X 20 20 30 40 (15), to monitor the filtering efficiencies. The filters were frozen immediately and kept frozen until DNA extraction. DNA from the cells collected on the Durapore filters was extracted in hot 1% SDS, concentrated by ethanol precipitation, and purified by phenol extraction (14). After a final ethanol precipitation and redissolution in TE (pH 8.0), DNA was quantified by Hoechst 33258 dye fluorometry (27) that was modified to decrease the amount of the sample DNA spent for the quantitation (flow-injection fluorometry; injection volume, 6 to 12 [LI of 1 to 10 ng of DNA per ,ul; final dye concentration, 1.5 x 10-6 M). Probe preparation and hybridization were the same as described above. Hybridizations were carried out reciprocally except for the coral reef lagoon sample, due to a shortage of the sample. natural DNA mixture DNA---- a) .0 .60 0E BB 100- U) ol 50- _- < / ~~~--------.;----- I-=L 10 0 30 20 40 12 1 20±1 21 4 37 2 64 3 69 5 5C Incubation Time (h) 150 NDb 17±2 32 ± 2 26 ± 3 64 ± 4 ND 50 Incubation Time (h) FIG. 2. (A) Effect of complexity on hybridization kinetics. Radioactivities of control dots were normalized to the radioactivity at the end of the incubation (48 h). The curves, Y = k1 + k2 (natural log of X), where k is constant, were fit by a least-squares method. Error bars indicate 2 standard errors of the mean. (B) Similarities determined from sample DNA dots after incubations of 6, 12, 24, and 48 h. The six lines are from six different samples. Error bars indicate 2 standard errors of the mean of each independent determination of the similarity. sition under confinement (11). Subsamples were taken from unfiltered seawater, the intermediate filtrate (between AE filter and Durapore filter), and the final filtrate; and cells were counted by the acridine orange direct counting method RESULTS Cross-hybridization among the pure strains. Cross-hybridizations of total DNA between distantly related strains were <1% to 2%. Cross-hybridization of bacterial DNA to eucaryotic DNA (calf thymus) was <0.1%, and it was not significantly different (paired comparison t test, 0.1 > P > 0.05) from the blank (no DNA) level, which was <7% (mode, 1%) of the control signal level during the experiments. The range of 16S rRNA sequence homologies among the distantly related strains was 0.67 to 0.73 (given as fraction, instead of as a percentage, to avoid confusion with total DNA similarity). The DNA hybridization among the closely related strains (16S rRNA homologies, 0.86 to 0.94) revealed low similarities, from 3 to 11%, except for VC5 and C. halobacteroides (16S rRNA homology, 0.90), which crosshybridized 30%. Cross-hybridization between E. coli and P. TABLE 2. Comparison of S,, with S, from hybridizations of mixture DNA probe to single-strain DNA targeta Probe Target DNA strain Mean ±SEM SO (%( CN-C CN-E LN150 Caulobacter strain VC5 C. halobacteroides 7 <1 144 ± 4 39 3 NP 175 ± 15 SI %) (%) 10 30 30 30 Mean +SEM Mea + SE 100 5 257 ± 30 53 4 NP 101 ± 12 Mixture D Mixture C Mixture B Mixture A Mean ± SEM SIMean ±(%)SEM Ma+SE, (%) S, S, 50 17 17 17 1 33 51 7 261 ± 6 NP 33 33 19 + Npb S St (%) <1 NDc 35 ± 4 NP NP 40 + See text for definitions of parameters. Mixture compositions, as a percentage of the total, can be read down the S, columns. All data were from n experiments, with two replicates for n = 1. b NP, Nonparticipating strain. c ND, Not determined. a 33 33 33 = 4 Downloaded from http://aem.asm.org/ on July 23, 2020 by guest 10 D P. vulgaris " See text for definitions of parameters. All data were from n = 4 experiments with two replicates for n = 1. b ND, Not determined. -o-- natural DNA a:- CN-E VOL. 56, 1990 BACTERIAL COMMUNITY COMPARISON BY DNA HYBRIDIZATION TABLE 4. Comparison of SO with S, from hybridizations of mixture DNA probe to mixture DNA target' TABLE 6. Species composition of mixture DNA Mixture 1 Mixture 2 Mixture 3 Mixture 1 Mean ± SEM S0 (%) S, Mixture 2 Mean ± SEM SI (%) S" (%) b 108 ± 11 60 ± 7 94 16 % Constituent bacterial strainsa Mixture Probe Target 60 75 83 + 60 17 743 75 a See text for definitions of parameters. See Table 6 for the species compositions of the mixtures. All data were from n = 4 experiments, with two replicates for n = 1. b_, Self-hybridization, defined as 100% similarity. C. halobacteroides CN-E CN-C LN150 1 2 0 0 30 17 30 17 10 50 30 17 3 4 5 6 0 17 33 0 25 33 67 0 25 17 0 33 25 17 0 33 7 8 25 33 50 0 25 17 0 33 0 25 33 0 33 0 a See Fig. 1 for the G+C content of the bacterial strains. However, hybridization done in one direction sometimes yielded >100% S,, whereas the reverse hybridization gave a number close to or lower than the S, (e.g., mixture 5 in Table 5). Table 5 presents extreme cases of asymmetric as well as symmetric reciprocities from a set of experiments designed to see this effect. S, was almost always between the two reciprocal SO values (Table 5). Table 6 shows the compositions of the DNA mixtures. Natural bacterioplankton. Loss of bacterial cells from the prefiltration was 8 to 10% of the total cell count in unfiltered seawater samples. About 1% or less passed through the Durapore filters. The species compositions of the natural bacterioplankton assemblages from a variety of places and times showed a wide range (from <10 to >90%) of similarity (Table 7). The two LIS summer samples taken from the same site but 1 week apart were indistinguishable (>90% similarity). The LIS autumn sample taken from the same site showed ca. 40% similarity to the two LIS summer samples, which was higher than its similarity to samples from the open ocean localities. The coral reef bacterioplankton assemblage was consistently least similar (<10%) to the other assemblages, even to nearby pelagic waters from the Sargasso Sea. The Caribbean Sea sample was the most similar to the Sargasso Sea sample. The reciprocal hybridizations were symmetric, in general, except for the Sargasso and the Caribbean sea samples and the Sargasso Sea and the LIS autumn samples (Table 7). DISCUSSION General considerations. Community DNA of bacterioplankton from diverse environments is likely to have a variety of complexities. The DNA reassociation rate is a TABLE 5. Symmetric and asymmetric hybridization reciprocity and comparison of SO, with SIa Probe Mixture 4 Target Mean + SEM Mixture 4 Mixture 5 Mixture 6 Mixture 7 Mixture 8 .b 201 18 110 41 S, (%) 40 14 1 4 4 Mixture 6 Mixture 5 Mean ± SEM 50 50 67 50 + 1 < 1+ <1 56 ± 3 20 ± 1 Mean ± SEM S 50 0 75 33 60 4 6 ±2 12 <1 65 6 Mixture 7 Mean ± SEM SI 50 0 25 67 NDc 325 45 7 1 41 2 Mixture 8 Mean ± SEM S, 67 75 25 ND 125 ± 19 66 ± 5 70 ± 5 S 50 33 67 50 50 a Reciprocal hybridizations (probe and target exchanged) can be seen as mirror images across the diagonal. See text for definitions of parameters. See Table 6 for the species compositions of the mixtures. All data were from n = 4 experiments, with two replicates for n = 1. b -, Self-hybridization, defined as 100% similarity. c ND, Not determined. Downloaded from http://aem.asm.org/ on July 23, 2020 by guest vulgaris (16S rRNA sequence homology, 0.94) was 4%. The higher degree of cross-hybridization between VC5 and C. halobacteroides, despite the lower 16S rRNA sequence homology, as compared with those between E. coli and P. vulgaris, was probably due to the high G+C content of caulobacters (Fig. 1; discussed below). Hybridization kinetics. Figure 2A shows the kinetics of the three DNA types with different complexities. Regression curves of Y = k, + k2. (natural log of X), where k is constant, were fit by a least-squares method. The regression curves of the three DNA types showed little difference overall, except that the curve of the simple DNA was slightly steeper than those of the other two DNA types. The curves of the complex (mixture of six species) DNA and the natural DNA were similar to each other. The similarities estimated at each incubation time (i.e., SO) remained relatively constant at all apparent similarities over the whole incubation period (Fig. 2B), despite the change of the absolute signal and the different kinetics, if any, of the different types of DNA (Fig. 2A). Hybridization between single-strain DNA and mixture DNA from cultures. Hybridization of a mixture DNA probe to single-strain DNA targets usually gave high measurements of SO, often >100% (Table 2). Hybridization of a singlestrain DNA probe to mixture DNA targets yielded a better estimation, with S0 being close to theoretical similarity, S, (Table 3). Hybridization of mixture DNA. Observed and theoretical similarities (SO and S,) were usually close within ca. 15%, or ±+1 standard error of mean (Table 4 and mixtures 6 through 8 in Table 5). Reciprocal similarities were usually symmetric; i.e., the forward and the reverse similarities agreed well with each other (e.g., mixtures 6 through 8 in Table 5). VC5 744 LEE AND FUHRMAN APPL. ENVIRON. MICROBIOL. TABLE 7. Species composition similarities of natural bacterioplankton assemblages 10 4 SO>St I 10 U) % Similarity of the following probes: Autumn Tart Target LIS LISan)' (mean)a Autumn LIS Summer LIS-1 Summer LIS-2 Caribbean Sea Sargasso Sea Coral reef Summer LIS-1 (mean + SEM)bc 34 ±2 57 43 <1 43 5 94 ± 8 12 ± 3 19 ± 2 4 1 Summer LIS-2 (mean ± SEM)b,d 33 1 92 ± 9 30 + 9 30 8 7 1 Caribbean Sea (mean + SEM)e 8 3 25 ± 13 32 + 9 (ea (mean)a 5 13 15 119 46± 5 4 1 '-- i Sargasso Sa 13 function of the complexity of the DNA, i.e., sequence diversity and size (1, 4). Unlike eucaryotic genomes, bacterial genomes do not generally contain repetitive sequences (4), and the genome size varies relatively less from species to species (within an order of magnitude [5]). Therefore, the complexity in this study means DNA sequence diversity or species diversity. If the similarity estimation is affected by DNA complexity or incubation time, it may not be appropriate for the samples of different complexities to be hybridized in one fixed incubation time. In our experiments, the relative signals (i.e., SO) did not change over the entire incubation period, regardless of probe complexity (Fig. 2A and B). According to Beltz et al. (2), the ratio of crosshybridization to self-hybridization stays constant over the extent of the reassociation reaction when the filter-bound target is in excess. In our study, the total amount of probe was about one-quarter of the total amount of the control DNA loaded onto the filter membrane. When the target is in excess, the hybridization rate is limited by the diffusion of the probe rather than by the complexity (12). Since the diffusion of the probe is nonselective for the kind of DNA, the hybridization (and SO) simply reflect the relative area of probe-binding sites in the targets. Community DNA similarity is an estimate of the fraction of identical DNA shared in common between two communities. For that purpose, it is ideal that different strains do not cross-hybridize with each other. Our tests showed that total DNA rarely cross-hybridized (from <1 to 11%) across the tested species (rRNA homologies, 0.67 to 0.94), except for one case (30%). This level of cross-hybridization would probably not cause a serious problem for general application of the method, unless the two communities being compared contained very closely related bacterial strains and these strains constituted major components. For the definition of species in bacteriology, it has been suggested that DNADNA cross-hybridization of .70% constitutes a single species (31), because the DNA hybridization technique is very discriminating even at the species level (16). With this criterion, a community DNA similarity of 70% can be interpreted as indicating that the two communities are the same at the species level. At the other extreme, it is possible that the communities are completely different at the strain or subspecies level. The most probable and realistic interpretation is between the two extreme examples. Given the observation that even congeners usually cross-hybridize less 0n (I)a).(j) lL 1 0 00 9 2 08 4So<St 8 ° o 0.1 0.1 ~~--8...0 0.o_s____o 0 o < complex probe simple target 1 sim-le probe complex target 10 Relative Complexity of Target to Probe FIG. 3. Goodness of fit of observed similarity (So) to theoretical similarity (S,) (S,, divided by 5,) as a function of the relative complexity of target DNA to probe DNA. The relative complexity was calculated by dividing the number of species constituting the target DNA by the number of species constituting the probe DNA. The dotted line indicates the 95% confidence limit of the regression line (solid line). than a few percent, it is likely that the major part of the similarity is from the presence of identical strains. Calibration with cultures. Hybridization of a single-strain DNA probe to a mixture DNA target gave a good estimate of S, (Table 3), whereas hybridization of a mixture DNA probe to a single-strain DNA target often resulted in an overestimation, sometimes >100% (Table 2). This is due to the complexity of DNA; for example, if a mixture probe contains 20% of strain A, a sample target of 100% of strain A would have five times the binding sites for the strain A DNA compared with that of the control (the mixture) target, which would contain strain A at only 20%. A reverse hybridization of the example, a probe of 100% of strain A to a mixture target containing 20% of strain A, would estimate So at close to 20%, because the sample (the mixture) target would contain 20% of strain A, whereas the control target would be 100% of strain A. When one is a subset (inclusion) or a partial subset (partial inclusion) of the other, reciprocal hybridization would easily be asymmetric. This is a limitation of an attempt to describe the relationship of species composition with a single number representing the similarity. However, reciprocal hybridizations help to infer the relative complexity of species compositions (as shown in Tables 2 and 3). A high percent molar G+C content (mol% G+C) also affected the reciprocity by increasing the melting temperature of the DNA-DNA hybrid (reference 1 and references therein). Since SO is the relative signal of a sample to a control, a probe that has a high-G+-C-content DNA would decrease SO by enhancing the control signal, whereas a target with a high G+C content DNA would result in an overestimation (mixture 5 in Table 5). The effect of high G+C content was reflected in the asymmetry of reciprocal hybridizations (Table 5). A significant observation was that hybridization of a simple probe to a complex target yielded an So value that was closer to the 5, value than did the reverse (complex probe to simple target) hybridization (Fig. 3; see Tables 2 and 3). The reciprocal similarities almost always bracketed the 5,, and the S, was always much closer to the underestimated similarity (Table 5). We suspect that most variations Downloaded from http://aem.asm.org/ on July 23, 2020 by guest a One determination (two replicates for n = 1). b Samples LIS-1 and LIS-2 were collected 7 days apart. c n = 4 (two replicates for n = 1). d n = 6 (two replicates for n = 1). e n = 3 (two replicates for n = 1). f-, Self-hybridization, defined as 100% similarity. 4- ° VOL. 56, 1990 BACTERIAL COMMUNITY COMPARISON BY DNA HYBRIDIZATION filters) was observed. However, prefiltration selectively removed larger cells (typically, ca. >0.8 ,um in diameter; 10% or less of the total count) and, probably, cells that were attached to larger particles, so such bacteria are excluded from these reported results. Cells that passed through the Durapore filters (0.22-p.m pore size) made up a very small fraction (<1% of the total cell count). Possible improvements. Tetraethylammonium chloride was introduced by Britten et al. (3) to suppress the G+C content effect. Although not yet used for procaryotic DNA hybridization, tetraethylammonium chloride may help to separate the complexity effect from the G+C content effect. We are investigating this possibility. Also, preliminary experiments indicated that closer examination of hybridization kinetics, measurement of the melting curves of hybrid DNA, or use of culturable strains isolated from the communities provides more unambiguous results. Conclusions. The differentiation of bacterioplankton assemblages has long been an unresolved problem. We chose total community DNA extracted from mixed populations of natural bacterioplankton to measure how close the species compositions are between pairs of samples, circumventing biases associated with culturing. We found that largely noncultivable natural bacterioplankton communities have species compositions with widely different patterns that vary over time and space. Our methodology for community DNA hybridization should be generally applicable for investigating the distribution patterns of bacterial communities. ACKNOWLEDGMENTS We thank G. Olsen for kindly providing a computer program written by him for sequence analysis and phylogenetic tree structure, J. Poindexter for helpful discussions and providing the Caulobacter strains, D. Dykhuizen for helpful advice, and R. Goodman for offering the E. coli and P. vulgaris strains. We also thank E. Carpenter and D. Capone for providing an opportunity for a Caribbean cruise on RV Columbus Iselin. This work was supported by grants OCE-8716988 and OCE8996117 from the National Science Foundation. LITERATURE CITED 1. Anderson, M. L. M., and B. D. Young. 1985. Quantitative filter hybridisation, p. 73-112. In B. D. Hames and S. J. Higgins (ed.), Nucleic acid hybridisation: a practical approach. IRL Press, Oxford. 2. Beltz, G. A., K. A. Jacobs, T. H. Eickbush, P. T. Cherbas, and F. C. Kafatos. 1983. Isolation of multigene families and determination of homologies by filter hybridization methods. Methods Enzymol. 100:266-285. 3. Britten, R. J., A. Cetta, and E. H. Davidson. 1978. The singlecopy DNA sequence polymorphism of the sea urchin Strongylocentrotus purpuratus. Cell 15:1175-1186. 4. Britten, R. J., and D. E. Kohne. 1968. Repeated sequences in DNA. Science 161:529-540. 5. Brock, T. D., and M. T. Madigan. 1988. Biology of microorganisms, 5th ed. Prentice-Hall, Inc., Englewood Cliffs, N.J. 6. Brosius, J., M. L. Palmer, P. J. Kennedy, and H. F. Noller. 1978. Complete nucleotide sequence of a 16S ribosomal RNA gene from Escherichia coli. Proc. Natl. Acad. Sci. USA 75:48014805. 7. Carbon, P., J. P. Ebel, and C. Ehresmann. 1981. The sequence of the ribosomal 16S RNA from Proteus vulgaris. Sequence comparison with E. coli 16S RNA and its use in secondary structure model building. Nucleic Acids Res. 9:2325-2333. 8. Cox, R. L., C.-C. Kuo, J. T. Grayston, and L. A. Campbell. 1988. Deoxyribonucleic acid relatedness of Chlamydia sp. strain TWAR to Chlamydia trachomatis and Chlamydia psittaci. Int. J. Syst. Bacteriol. 38:265-268. 9. DeLong, E. F., G. S. Wickham, and N. R. Pace. 1989. Phyloge- Downloaded from http://aem.asm.org/ on July 23, 2020 by guest from a good fit when the probe had equal or less complexity than the target were due to the percent G+C effect. Calibration summary. From the pure culture experiments, we can conclude the following. When reciprocal similarities are symmetric, (i) a high similarity (>90%) indicates a nearly identical species composition; (ii) a low similarity (<10%) indicates that few or no identical species are present; and (iii) an intermediate similarity indicates that a part of each assemblage is in common or that a large fraction is phylogenetically close species, but it is most likely that the major part of the similarity is from the identical strains that are present. When reciprocal hybridizations are asymmetric, (iv) the true similarity is closer to the lower of the two measured values, and (v) the higher similarity indicates that (a) the probe is more complex (more species each contributing a small percent) than the target (subset or partial subset of the probe, with fewer species each contributing a high percent), (b) DNA that is in common between the samples has a high percent G+C content and makes up a larger fraction of the target than of the probe, or (c) both. With the current protocol at a fixed hybridization temperature, situations i, ii, and iv give unambiguous answers, but situations iii and v have more than one possible interpretation. However, a higher similarity means that there is a greater chance that identical species are present than a lower similarity does. Ways to potentially resolve the ambiguities are discussed below. Natural assemblages. The two LIS summer communities taken 1 week apart were indistinguishable (>90%), while the LIS autumn community taken from the same site was about 40% similar to the two LIS summer samples. Thus, the species composition of the LIS samples changed significantly over time, yet the patterns in comparison with those of samples from other locations suggest that certain species were present in LIS throughout the studied period. Not surprisingly, species composition also showed geographical variations. The coral reef bacterioplankton assemblage was found to be unique (<10% similarity to all others), even when it was compared with assemblages from the nearby Sargasso Sea, with which it must periodically mix during rough weather. The Caribbean Sea sample was found to be the most similar to the Sargasso Sea sample among the samples that were compared, again, as might be expected. A few natural communities showed a significant asymmetric reciprocity. This suggests that the natural communities may be composed of heterogeneous groups and, thus, may have a comparable complexity and a balanced distribution of G+C content. Our measurements of the G+C content of natural bacterial community DNA from LIS (summer and autumn) and Sargasso Sea (summer) by high-pressure liquid chromatography (21) ranged from 40 to 50%, despite the potentially wide range of G+C content of procaryotic organisms (20 to 78% [5]). Asymmetric reciprocity between the Sargasso and the Caribbean sea samples may indicate, according to the results from pure cultures, that (i) the Caribbean Sea sample has a major component of high G+C content of DNA in common with the Sargasso Sea sample, (ii) the Caribbean Sea community is a subset or a partial subset (less complex) of the Sargasso Sea sample (more complex), or (iii) both. The same reasoning may be proposed for the asymmetric reciprocity between the Sargasso Sea and the autumn LIS samples. Since for our method we used extracted DNA to infer species variations, it is important that the sample DNA properly represents the community. No significant variation in filtration efficiencies (88 to 91% collected on Durapore 745 746 10. 11. 12. 13. 15. 16. 17. 18. 19. 20. 21. netic stains: ribosomal RNA-based probes for the identification of single cells. Science 243:1360-1363. Dubnau, D., I. Smith, P. Morell, and J. Marmur. 1965. Gene conservation in Bacillus species. I. Conserved genetic and nucleic acid base sequence homologies. Proc. Natl. Acad. Sci. USA 54:491-498. Ferguson, R. L., E. N. Buckley, and A. V. Palumbo. 1984. Response of marine bacterioplankton to differential filtration and confinement. Appl. Environ. Microbiol. 47:49-55. Flavell, R. A., E. J. Birfelder, J. P. M. Sanders, and P. Borst. 1974. DNA-DNA hybridization on nitrocellulose filters. 1. General considerations and non-ideal kinetics. Eur. J. Biochem. 47:535-543. Fuhrman, J. A., and F. Azam 1982. Thymidine incorporation as a measure of heterotrophic bacterioplankton production in marine surface waters: evaluation and field results. Mar. Biol. 66:109-120. Fuhrman, J. A., D. E. Comeau, i. Hagstrom, and A. M. Chan. 1988. Extraction from natural planktonic microorganisms of DNA suitable for molecular biological studies. Appl. Environ. Microbiol. 54:1426-1429. Hobbie, J. E., R. J. Daley, and S. Jasper. 1977. Use of Nuclepore filters for counting bacteria by fluorescence microscopy. Appl. Environ. Microbiol. 33:1225-1228. Kaneuchi, C., M. Seki, and K. Komagata. 1988. Taxonomic study of Lactobacillus mali Carr and Davis 1970 and related strains: validation of Lactobacillus mali Carr and Davis 1970 over Lactobacillus yamanashiensis Nonomura 1983. Int. J. Syst. Bacteriol. 38:269-272. Kwok, S., T. J. White, and J. W. Taylor. 1986. Evolutionary relationships between fungi, red algae, and other simple eucaryotes inferred from total DNA hybridizations to a cloned basidiomycete ribosomal DNA. Exp. Mycol. 10:196-204. Lane, D. J., K. G. Field, G. J. Olsen, and N. R. Pace. 1988. Reverse transcriptase sequencing of ribosomal RNA for phylogenetic analysis. Methods Enzymol. 167:138-144. Lautrop, H. 1974. Genus X. Proteus, p. 327. In R. E. Buchanan and N. E. Gibbons (ed.), Bergey's manual of determinative bacteriology, 8th ed. The Williams & Wilkins Co., Baltimore. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecular cloning: a laboratory manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. Mischke, C. F., and E. Wickstrom. 1980. Deoxynucleoside composition of DNAs and modified nucleoside composition of tRNAs determined at nanomole sensitivity by reversed-phase liquid chromatography. Anal. Biochem. 105:181-187. 22. Moore, R. L., J. Schmidt, J. Poindexter, and J. T. Staley. 1978. Deoxyribonucleic acid homology among the caulobacters. Int. J. Syst. Bacteriol. 28:349-353. 23. Olsen, G. J. 1988. Phylogenetic analysis using ribosomal RNA. Methods Enzymol. 164:793-812. 24. 0rskov, F. 1974. Genus I. Escherichia, p. 293. In R. E. Buchanan and N. E. Gibbons (ed.) Bergey's manual of determinative bacteriology, 8th ed. The Williams & Wilkins Co., Baltimore. 25. Pace, N. R., D. A. Stahl, D. J. Lane, and G. J. Olsen. 1986. The analysis of natural microbial populations by ribosomal RNA sequences. Adv. Microb. Ecol. 9:1-55. 26. Palleroni, N. J., R. Kunisawa, R. Contopoulou, and M. Doudoroff. 1973. Nucleic acid homologies in the genus Pseudomonas. Int. J. Syst. Bacteriol. 23:333-339. 27. Paul, J. H., and B. Myers. 1982. Fluorometric determination of DNA in aquatic microorganisms by use of Hoechst 33258. Appl. Environ. Microbiol. 43:1393-1399. 28. Poindexter, J. S. 1974. Genus Caulobacter, p. 153. In R. E. Buchanan and N. E. Gibbons (ed.), Bergey's manual of determinative bacteriology, 8th ed. The Williams & Wilkins Co., Baltimore. 29. Salyers, A. A., S. P. Lynn, and J. F. Gardner. 1983. Use of randomly cloned DNA fragments for identification of Bacteroides thetaiotaomicron. J. Bacteriol. 154:287-293. 30. Stackebrandt, E., B. Wunner-Fussel, V. J. Fowler, and K.-H. Schleifer. 1981. Deoxyribonucleic acid homologies and ribosomal ribonucleic acid similarities among sporeforming members of the order Actinomycetales. Int. J. Syst. Bacteriol. 31:420-431. 31. Wayne, L. G., D. J. Brenner, R. R. Colwell, P. A. D. Grimont, 0. Kandler, M. I. Krichevsky, L. H. Moore, W. E. C. Moore, R. G. E. Murray, E. Stackebrandt, M. P. Starr, and H. G. Truper. 1987. Report of the ad hoc committee on reconciliation of approaches to bacterial systematics. Int. J. Syst. Bacteriol. 37:463-464. 32. Weisberg, W. G., Y. Oyaizu, H. Oyaizu, and C. R. Woese. 1985. Natural relationship between bacteroides and flavobacteria. J. Bacteriol. 164:230-236. 33. Woese, C. R. 1987. Bacterial evolution. Microbiol. Rev. 51: 221-271. 34. Yang, D., Y. Oyaizu, H. Oyaizu, G. J. Olsen, and C. R. Woese. 1985. Mitochondrial origins. Proc. Natl. Acad. Sci. USA 82: 4443-4447. Downloaded from http://aem.asm.org/ on July 23, 2020 by guest 14. APPL. ENVIRON. MICROBIOL. LEE AND FUHRMAN