Statistical Approaches for Estimating Actinobacterial Diversity in Marine Sediments

Abstract

Bacterial diversity in a deep-sea sediment was investigated by constructing actinobacterium-specific 16S ribosomal DNA (rDNA) clone libraries from sediment sections taken 5 to 12, 15 to 18, and 43 to 46 cm below the sea floor at a depth of 3,814 m. Clones were placed into operational taxonomic unit (OTU) groups with ≥99% 16S rDNA sequence similarity; the cutoff value for an OTU was derived by comparing 16S rRNA homology with DNA-DNA reassociation values for members of the class Actinobacteria. Diversity statistics were used to determine how the level of dominance, species richness, and genetic diversity varied with sediment depth. The reciprocal of Simpson's index (1/D) indicated that the pattern of diversity shifted toward dominance from uniformity with increasing sediment depth. Nonparametric estimation of the species richness in the 5- to 12-, 15- to 18-, and 43- to 46-cm sediment sections revealed a trend of decreasing species number with depth, 1,406, 308, and 212 OTUs, respectively. Application of the LIBSHUFF program indicated that the 5- to 12-cm clone library was composed of OTUs significantly (P = 0.001) different from those of the 15- to 18- and 43- to 46-cm libraries. F_ST and phylogenetic grouping of taxa (P tests) were both significant (P < 0.00001 and P < 0.001, respectively), indicating that genetic diversity decreased with sediment depth and that each sediment community harbored unique phylogenetic lineages. It was also shown that even nonconservative OTU definitions result in severe underestimation of species richness; unique phylogenetic clades detected in one OTU group suggest that OTUs do not correspond to real ecological groups sensu Palys (T. Palys, L. K. Nakamura, and F. M. Cohan, Int. J. Syst. Bacteriol. 47:1145-1156, 1997). Mechanisms responsible for diversity and their implications are discussed.

Deep-sea sediments cover 63.5% of the Earth's surface (12) and represent the most undersampled marine habitat (5). Deep-sea sediments, once thought of as uniform featureless environments, are extremely heterogeneous; recent reports suggest that species diversity in deep-sea sediments is greater than in coastal sediments (16, 26) and may even rival that of tropical rainforests (38). To date, few studies have focused on microbial diversity in nonhydrothermal vent system deep seas (8, 18, 21, 23, 27, 57), and fewer have investigated diversity through depth profiles (21, 57). Furthermore, all microbial diversity investigations made using marine sediments represent surveys of microbial diversity; none have attempted to estimate and compare bacterial diversity at different sites.

The estimation of bacterial diversity is required for understanding bacterial biogeography, community assembly, and ecological processes (9). To date the inability to accurately measure bacterial diversity has been limited by the size of samples needed to give adequate sample coverage of bacterial communities (11) and by the lack of a clear concept of the bacterial species as an evolving entity (49). Currently, species definition is based on the degree of DNA-DNA relatedness between two bacterial isolates (47). Diversity predictions can be made using statistical approaches that estimate species number from relatively small sample sizes. Hughes et al. (19) recently reviewed both rarefaction and richness estimators that have been applied to microbial data sets and in particular highlighted the utility of nonparametric estimators in predicting and comparing bacterial species number. Rarefaction and richness estimators rely on a species or operational taxonomic unit (OTU) definition. OTU definitions are usually based on 16S ribosomal DNA (rDNA) gene similarity under the assumption that a high degree of similarity will be reflected in the degree of DNA-DNA relatedness, i.e., likely to be the same species (48). The limitations of OTUs have previously been addressed and include inconsistent arbitrary cutoff values (1 to 5% 16S rDNA gene dissimilarities have been used [19]) and the fact that OTUs are counted equivalently despite the fact that some may be highly divergent and phylogenetically unique, whereas others may be closely related and phylogenetically redundant (32). Recently statistical analyses borrowed from population genetics and systematics have been employed and reviewed for use with microbial data sets (32). The benefits of such methods include the fact that OTU definitions are not required and that they do not rely on estimation of the frequency of different sequences i.e., it is phylogenetic diversity and not species richness that is estimated. The estimation of species richness combined with knowledge of phylogenetic diversity will advance our ability to test inferences as to the processes that govern bacterial diversity.

In this study we restricted our investigation to the class Actinobacteria, in order to improve coverage of species in reasonably sized 16S rDNA clone libraries and since diversity estimates are likely to be more applicable when applied to specific groups (17). Members of this class include marine species of value for the production of novel bioactive compounds for biotechnology (3, 33, 45); hence, knowledge of their diversity, distribution, and ecology in marine systems will aid bioprospecting strategies. This is the first study designed to investigate bacterial diversity in a deep-sea sediment by comparing phylogenetic diversity and estimating species number using a number of statistical methods. The principal results show that species richness decreased with sediment depth with a concomitant decrease in genetic diversity and an increase in the level of dominance. Each of the sediment core sections contained unique phylogenetic lineages, and furthermore, correlation of 16S rDNA and DNA-DNA reassociation combined with phylogenetic diversity comparisons showed that clones grouped into operational taxonomic units are likely to be derived from distinct species corresponding to separate ecological units.

MATERIALS AND METHODS

Sediment samples.

Atlantic ocean deep-sea sediment was collected from the edge of the Saharan debris flow near the Canary Islands (27^o02.39′N 18^o29.02′W) at a depth of 3,814 m using a piston corer during a scientific cruise aboard the RRS Charles Darwin in April 2001. The upper 38 cm of the core was a single turbidite (identified as turbidite A [54]) grading from fine/medium sand at the base to very soft mud at the top; below the turbidite the core was comprised of pelagic clay. Sections were taken at 5 to 12 cm (fine-grained muddy turbidite), 15 to 18 cm (sandy basal layer), and 43 to 46 cm (pelagic clay) below the surface of the sediment. The deposition time of the turbidite was radiocarbon dated according to the method of Thomson and Weaver (54). Approximate dating of the pelagic clay was conducted by visual inspection. Prediction of the age range of the initial sediment from which the turbidite originated was calculated according to the method of Weaver and Thomson (58). Total organic carbon contents were measured using the Walkley-Black method (1).

DNA extraction, amplification, and cloning.

DNA was extracted and concentrated from 10 g of the sediment sections using an UltraClean Mega DNA soil kit (Mo Bio Laboratories, Solana Beach, Calif.) according to the manufacturer's instructions. DNA was subjected to a second round of purification using a Wizard DNA spin-purification column (Promega, Madison, Wis.) and quantified against known concentration standards with Quantity-One software (Bio-Rad, Richmond, Calif.).

PCR amplification of actinobacterial 16S rDNA.

PCR was made in a total volume of 50 μl containing 1× PCR buffer with 1.5 mM MgCl₂, deoxynucleoside triphosphates (200 μM [each] dATP, dCTP, dGTP, and dTTP), 0.5 μM [each] S-C-Act-0235-a-S-20 and S-C-Act-0878-a-A-19 primers (46), 2.0 U of DNA polymerase (Expand-Taq; Roche Diagnostics, Mannheim, Germany), and 100 ng of DNA extract. To increase amplification efficiencies from sediment DNA, T4 gene 32 protein (Roche Diagnostics) was added at a concentration of 5 μg per reaction (52). Amplification was made using a “Touchdown” protocol (39), which consisted of an initial denaturation at 95°C for 4 min, followed by denaturation at 95°C for 45 s, annealing at 72°C for 45 s, and extension at 72°C for 1 min; 10 cycles in which the annealing temperature was decreased by 0.5°C/cycle from the preceding cycle; and then 15 cycles of 95°C for 45 s, 68°C for 45 s, and 72°C for 1 min, with the last cycle followed by a 5-min extension at 72°C. PCR products were separated on 2% agarose gels stained with ethidium bromide (41).

Cloning and sequencing.

PCR products were purified using QIAquick gel extraction columns according to the manufacturer (Qiagen, Crawley, United Kingdom). Purified DNA was blunt-end ligated into the plasmid vector pST-Blue-1 and used to transform NovaBlue competent cells using a Perfectly Blunt cloning kit (Novagen, Madison, Wis.). Transformed cells were serially diluted and plated onto Luria-Bertani agar containing 50 μg of carbenicillin ml⁻¹, 15 μg of tetracycline ml⁻¹, 70 μg of 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside ml⁻¹, and 80 μg of isopropyl-β-d-thiogalactopyranoside ml⁻¹. Small actinobacterial 16S rDNA clone libraries were made for each sediment section by selecting the dilution (for each sediment core section) that yielded between 40 and 70 positive transformants. Transformants were screened by PCR using the primers S-C-Act-0235-a-S-20 and S-C-Act-0878-a-A-19 (46), and plasmids containing actinobacterial 16S rDNA inserts were sequenced commercially (Qiagen) using S-C-Act-0235-a-S-20. Chimera analysis was done as previously described (46).

OTU definition for the class Actinobacteria.

An OTU definition for the class Actinobacteria was made by comparing and plotting DNA-DNA reassociation and 16S rRNA homology from sample data published between 1997 and 2001 (369 separate samples; data available from the corresponding author). Actinobacterial 16S rRNA sequences recovered from each of the sediment sections were placed into OTU groups by comparison using the SIMILARITY_MATRIX function of the Ribosomal Database Project (31). The closest representative for each OTU group was identified using BLAST (2).

Phylogenetic tree construction.

Actinobacterial 16S rDNA sequences recovered from the sediment sections were uploaded to the CLUSTAL X interface, where they were aligned to generate a guide dendrogram from which a final alignment was made (53). All phylogenic calculations were made using programs available in the phylogeny inference package PHYLIP (version 3.5.1; J. Felsenstein, Department of Genetics, University of Washington [http://evolution.genetics.washington.edu/phylip.html]). A distance matrix was constructed from the alignment using the DNADIST program. The phylogenetic tree was generated using the neighbor-joining method from the NEIGHBOR program under the Jukes-Cantor model of nucleotide substitution. Bootstrap analysis was conducted using the SEQBOOT and CONSENSE programs using 100 resamplings of the data.

Sample coverage and species richness estimation.

All species richness and sample coverage calculations were performed with the programs EstimateS (version 5.0.1; R. K. Colwell, University of Connecticut [http://viceroy.eeb.unconn.edu/estimates]) and SPADE (species prediction and diversity estimation; A. Chao and T.-J. Shen [http://chao.stat.nthu.edu.tw]). Each cloned sequence was treated as a separate sample, and 100 randomizations were conducted for all tests. Accumulation curves were calculated by plotting the proportion of individuals sampled against the proportion of OTUs observed. The actinobacterial species richness was calculated for each of the sediment sections using the nonparametric estimators ACE (abundance-based coverage estimator) and Chao1 (6, 7, 19). Since the distribution of estimates is not normal (6), Burnham's log transformation (6) was employed to calculate 95% confidence intervals so that the lower bound of the resulting interval is at least the number of observed species: {S + [(Ñ − S)/C], S + [(Ñ − S)C], where C = exp(1.96{log[1 + σ²/(Ñ − S)²]}^0.5), where Ñ is the estimate, S is the number of species observed, and σ is the variance (calculated from the closed-form solutions for the estimators) (6, 7). Extrapolation using best-fit regression analysis was performed (where necessary) to calculate the point at which 95% confidence intervals (CIs) did not overlap (19). Regression analysis was performed using SigmaPlot (version 7.1) (Jandel Scientific, San Rafael, Calif.).

Statistical comparison of coverage.

Actinobacterial 16S rDNA libraries from the three sediment sections were compared using the LIBSHUFF computer program (44) (http://www.arches.uga.edu/∼whitman/libshuff.html), which used the coverage formula of Good (15) to generate homologous and heterologous coverage curves from the 16S rDNA clone libraries. Sequences were randomly shuffled 999 times between samples prior to the distance between the curves being calculated using the Cramér-von Mises test statistic (37). The DNADIST program of PHYLIP (see above URL) using the Jukes-Cantor model for nucleotide substitution was used to generate the distance matrix analyzed by LIBSHUFF.

Reciprocal of Simpson's index.

The reciprocal of Simpson's index (1/D) was used as a measure of diversity, since it has been widely used for ecological studies, has good discriminating ability, and has previously been applied to microbial communities (61).

Lineage-per-time plots.

Plots were made according to the method of Martin (32). Cladograms were made by clustering of genetic distances corrected by the F84 (maximum-likelihood) model of nucleotide substitution. Genetic distances were calculated using DNADIST (see above URL for PHYLIP), and optimization of branch lengths was done using KITSCH (see above URL), enforcing a molecular clock (i.e., under the constraint that all sequences were contemporary).

Comparison of phylogenetic diversity.

The phylogenetic diversity within each community was compared using F statistics (F_ST) and phylogenetic grouping of taxa (P tests) (32). The F_ST test was used to compare the genetic diversity within each sediment section community to the total genetic diversity of the communities combined, using the equation F_ST = (θ_T − θ_W)/θ_T, where θ_T is the genetic diversity for all samples and θ_W is the genetic diversity in each community (32). Population differentiation using F_ST was calculated using the ARLEQUIN program (42); statistical significance was evaluated by randomly assigning sequences to populations and calculating the F_ST for 1,000 permutations. The P test was used to investigate whether the communities exhibited covariation with phylogeny (32). Neighbor-joining trees constructed under the HKY model of nucleotide substitution in PAUP (51) were used to estimate the minimum number of changes (switch from one community to the other) required to explain observed distribution. The significance of the covariation was determined by establishing the expected number of changes under the null hypothesis (29) that the community did not covary with phylogeny. The P test was conducted using MacCLADE software (28). F_ST and P tests were also applied to OTU groups to calculate whether such groups are comprised of sequences drawn from the same phylogenetic lineage or if they comprise distinct phylogenetic lineages that are arbitrarily considered as identical.

Nucleotide sequence accession numbers.

The nucleotide sequences determined in this study have been deposited in the National Institute for Biotechnology Information database under the accession numbers AF544254 to AF544369.

RESULTS

Sediment analysis.

Radiocarbon dating of the sediment revealed that the turbidite fraction (0 to 38 cm) of the core was deposited ca. 1,000 years ago (54). However, it is important to note that the turbidite was derived from failure of older, previously deposited sediment from further upslope. Modeling of the coccolith mixture in the turbidite indicted that it consisted of material covering an age range of approximately 200,000 years, suggesting failure of a slab of sediment 15 m thick. Below the turbidite (38 to 60+ cm), the sediment core consisted of pelagic clay. Visual inspection revealed that it contained foraminifera and was comprised of a mixture of fine-grained material: mainly coccoliths, clays, and volcanic minerals. The brown (oxidized) nature of the clay indicated that it was from the Holocene period and was younger than 12,000 years. Total organic carbon contents were 0.92% ± 0.14%, 1.05% ± 0.05%, and 0.98% ± 0.33% percent for the 5- to 12-, 15- to 18-, and 43- to 46-cm sediment sections, respectively.

Construction and analysis of actinobacterial 16S rDNA gene libraries.

DNA extracted from the three sediment sections was amenable to direct PCR amplification. The highest yields of genomic DNA were recovered from the 5- to 12-cm section, and amounts of recovered DNA decreased in proportion to sediment depth (data not shown). The 5- to 12-cm library (61 positive transformants) yielded 57 clones that tested positive for 16S rDNA genes (93%); the 15- to 18-cm library yielded 48 clones (90%); and the 43- to 46-cm library yielded 62 clones (85%). None of the actinobacterial 16S rDNA genes were considered to be chimeras under the methods employed. Sequencing revealed that all 167 16S rDNA clones belonged to the class Actinobacteria, giving a specificity of 100% for the S-C-Act-0235-a-S-20 and S-C-Act-0878-a-A-19 primers.

In order to conduct richness estimation, clones were placed into OTUs. OTUs were defined by comparing 16S rRNA homology with DNA-DNA reassociation values for members of the class Actinobacteria, using the ∼70% DNA-DNA reassociation cutoff point as the phylogenetic definition of a species (48). A 16S rRNA homology value of >99% minimized DNA-DNA reassociation values of <70% and incorporated 70% of all values of >70% (Fig. 1). Thus, an actinobacterial OTU was defined as a 16S rDNA sequence group in which sequences differed by ≤1%. Previous work has shown that 16S rRNA homology values calculated using the region amplified by the present primers are conservative by approximately 0.7% (46). Seventy-two percent of species with >99% 16S rRNA homology had <70% DNA-DNA reassociation values. Of the actinobacterial 16S rDNA sequences recovered, 116 OTUs were identified (Table 1). The closest relatives of representative actinobacterial OTUs, identified by searching in the GenBank database, are given in Table 1. The prefix T indicates that the 16S rDNA gene sequences was recovered from the 5- to 12-cm sediment section, M indicates the 15- to 18-cm section, and D indicates the 43- to 46-cm section. Phylogenetic analysis indicated that actinobacterial OTUs were most closely related to members of the suborders Corynebacterineae, Frankineae, and Streptomycineae (Fig. 2). However, only 9% of the OTUs showed 100 to 99% homology with cultured actinobacteria (i.e., likely to be known species), 85% had 98 to 95% homology with known species, and 17% had 94 to 89% homology. Using the DNA-DNA reassociation comparison with 16S rDNA homology, it is probable that 91% of the OTUs recovered from the deep-sea sediment represent novel species or genera.

FIG. 1.

Comparison of 16S rRNA homology and DNA-DNA reassociation values in the class Actinobacteria. The bar indicates the threshold value for species delineation (48). The horizontal line indicates the 16S rDNA homology value for which an actinomycete OTU was defined.

TABLE 1.

Actinomycete 16S rDNA sequences identified in a deep-sea sediment core in three clone libraries

Representative clone (accession no.)	No. of clones in library			Closest relative
	5-12 cm	15-18 cm	43-46 cm	Actinomycete	Accession no.	% Identity
T01 (AF544299)	1			Verrucosispora sp. strain IM-6907	AF131631	97
T02 (AF544256)	1			Streptomyces thermocarboxydovorans	U94489	98
T03 (AF544259)	1			Streptomyces thermocarboxydovorans	U94489	95
T04 (AF544298)	1			Streptomyces sp. strain KACC 20159	AF345862	89
T05 (AF544297)	1			Uncultured actinobacterium clone ASb07	AY124393	96
T06 (AF544294)	1			Streptomyces sp. strain 19504	AJ315072	92
T07 (AF544296)	1			Streptomyces capensis	AF452714	97
T08 (AF544295)	1			Dietzia maris	Y18883	94
T09 (AF544279)	1			Uncultured actinobacterium clone SDe09	AY124457	92
T10 (AF544293)	1			Corynebacterium ulcerans	X84256	96
T11 (AF544254)	1			Streptomyces sp. strain IM-6899	AF131529	95
T12 (AF544255)	1			Streptomyces sp. strain IM-6960	AF131527	95
T13 (AF544292)	1			Kineococcus-like bacterium AS2978	AF060676	95
T14 (AF544291)	1			Rhodococcus sp. strain ARG-BN062	AF420423	95
T15 (AF544290)	1			Frankia sp. strain	AF063642	93
T16 (AF544257)	1			Dietzia maris	Y18883	95
T17 (AF544298)	1			Mycobacterium sp. strain	U46146	97
T19 (AF544258)	1			Streptomyces sp. strain IM-1436	AF131505	96
T21 (AF544288)	1			Geodermatophilus sp. strain BC509	AJ296063	96
T22 (AF544260)	1			Streptomyces sp. strain IM-1436	AF131505	97
T23 (AF544287)	1			Streptomyces thermocarboxydovorans	U94489	99
T24 (AF544286)	1			Corynebacterium tuberculostearicum	AF510731	98
T26 (AF544285)	1			Streptomyces sp. strain IM-1436	AF131505	96
T27 (AF544261)	1			Streptomyces sp. strain IM-7585	AF131604	96
T28 (AF544284)	1			Amycolatopsis sp. strain GY109	AF466094	96
T29 (AF544262)	1			Uncultured bacterium ARFS-3	AJ277836	98
T30 (AF544263)	1			Tsukamurella inchonensis	AF283281	93
T31 (AF544283)	1			Micromonospora sp. strain IM-7020	AF131416	95
T32 (AF544264)	1			Streptomyces sp. strain NK1057	AF492844	98
T33 (AF544282)	3	1		Uncultured actinobacterium clone ASb07	AY124393	99
T34 (AF544265)	1			Dietzia maris	Y18883	94
T35 (AF544281)	1			Blastococcus sp. strain BC448	AJ316571	98
T36 (AF544266)	1			Streptomyces sp. strain IM-8073	AF131609	97
T37 (AF544280)	1			Streptomyces sp. strain 11AG8	AF233375	95
T38 (AF544267)	1			Streptomyces capensis	AF452714	98
T40 (AF544268)	1			Blastococcus sp. strain BC448	AJ316571	97
T41 (AF544269)	1		2	Dietzia maris	Y18883	94
T42 (AF544300)	1			Streptomyces pallidus	AJ399492	94
T43 (AF544270)	1			Rhodococcus tukisamuensis	AB067734	95
T44 (AF544301)	1			Amycolatopsis sp. strain GY109	AF466094	96
T45 (AF544271)	1			Streptomyces sp. strain IM-8073	AF131609	95
T47 (AF544272)	2			Streptomyces thermocarboxydovorans	U94489	99
T48 (AF544302)	1			Mycobacterium sp. strain STR-12	AF236839	96
T49 (AF544273)	1			Streptomyces somaliensis	AJ007403	95
T50 (AF544274)	1			Uncultured actinobacterium clone SDb12	AY124440	91
T51 (AF544303)	1			Streptomyces sp. strain IM-7585	AF131604	95
T52 (AF544275)	1			Streptomyces somaliensis	AJ007401	90
T54 (AF544304)	1			Geodermatophilus sp. strain	X92358	96
T56 (AF544276)	1			Geodermatophilus sp. strain IM-1092	AF131367	93
T57 (AF544305)	1			Streptomyces thermocarboxydovorans	U94489	99
T58 (AF544277)	1			Mycobacterium sp. strain STR-18	AF236841	95
T59 (AF544306)	1			Rhodococcus opacus	Y11892	92
T60 (AF544278)	1			Streptomyces thermocarboxydovorans	U94489	98
M03 (AF544336)		1		Corynebacterium appendicis	AJ314919	99
M06 (AF544335)		2		Rhodococcus sp. strain SRB1948-Z40	AB010913	96
M07 (AF544334)		1		Corynebacterium striatum	X81910	94
M08 (AF544308)		1		Rhodococcus ruber	X80625	95
M09 (AF544333)		1		Rhodococcus fascians	X79186	95
M11 (AF544307)		1		Pseudonocardia aurantiaca	AF325727	96
M12 (AF544332)		1		Dietzia maris	Y18883	95
M14 (AF544322)		1		Rhodococcus fascians	X79186	98
M15 (AF544331)		1		Rhodococcus koreensis	AF124342	96
M16 (AF544321)		1		Streptomyces sp. strain CN732	AF101415	92
M17 (AF544330)		1		Pseudonocardia alaniniphila	AF325726	98
M20 (AF544320)		1		Nocardioides sp. strain V4.BO.15	AJ244655	93
M22 (AF544319)		1		Rhodococcus fascians	X79186	95/PICK>
M23 (AF544329)		1		Turicella otitidis	X73976	99
M24 (AF544318)		1		Rhodococcus fascians	X79186	98
M25 (AF544328)		1		Rhodococcus fascians	X79186	97
M26 (AF544317)		1		Mycobacterium manitobense	AY082001	94
M27 (AF544327)		1		Rhodococcus fascians	X79186	97
M28 (AF544316)		1		Turicella otitidis	X73976	92
M29 (AF544326)		1		Actinosynnema sp. strain IM-1402	AF131351	95
M30 (AF544315)		1		Rhodococcus fascians	X79186	97
M33 (AF544325)		1		Rhodococcus fascians	X79186	96
M35 (AF544324)		1		Rhodococcus zopfii	AF191343	93
M36 (AF544314)		1		Uncultured actinobacterium clone ASb09	AY124394	94
M37 (AF544323)		1		Rhodococcus fascians	X79186	97
M38 (AF544337)		1		Mycobacterium manitobense	AY082001	96
M39 (AF544313)		1		Frigoribacterium sp. strain 301	AF157479	96
M40 (AF544312)		1		Rhodococcus fascians	X79186	95
M41 (AF544338)		2		Rhodococcus sp. strain UFZ-B520	AF235011	98
M43 (AF544311)		1		Rhodococcus fascians	X79186	98
M44 (AF544339)		1		Rhodococcus fascians	X79186	98
M45 (AF544310)		1		Rhodococcus fascians	X79186	98
M46 (AF544340)		1		Mycobacterium manitobense	AY082001	96
M47 (AF544309)		1		Rhodococcus fascians	X79186	97
D02 (AF544341)			1	Rhodococcus fascians	X79186	97
D05 (AF544342)	1		1	Mycobacterium manitobense	AY082001	98
D06 (AF544343)			1	Kitasatospora sp. strain IM-6832	AF131378	96
D11 (AF544344)			1	Rhodococcus fascians	X79186	98
D13 (AF544345)			1	Rhodococcus fascians	X79186	99
D16 (AF544346)			2	Rhodococcus sp. strain S9	AF260713	98
D17 (AF544347)			1	Rhodococcus fascians	X79186	97
D22 (AF544348)			1	Rhodococcus fascians	X79186	96
D25 (AF544349)			1	Arthrobacter sp. strain “SMCC G960”	AF197025	96
D26 (AF544350)			1	Rhodococcus fascians	X79186	98
D27 (AF544351)			1	Mycobacterium sp. strain STR-21	AF236837	96
D32 (AF544352)			1	Rhodococcus fascians	X79186	99
D35 (AF344353)			1	Uncultured actinobacterium clone ASc02	AY124397	95
D36 (AF544354)			1	Rhodococcus sp. strain YK2	AB070458	96
D41 (AF544355)			1	Rhodococcus fascians	X79186	98
D42 (AF544356)			1	Mycobacterium sp. strain STR-11	AF236838	97
D43 (AF544357)			1	Dietzia maris	Y18883	95
D44 (AF544358)			2	Rhodococcus fascians	AF205371	98
D45 (AF544359)			1	Rhodococcus fascians	X79186	98
D47 (AF544360)			1	Dietzia maris	Y18883	94
D48 (AF544361)			1	Rhodococcus fascians	X79186	99
D50 (AF544362)			1	Rhodococcus fascians	X79186	97
D52 (AF544363)			1	Blastococcus sp. strain BC412	AJ316574	98
D54 (AF544364)			1	Rhodococcus fascians	X79186	98
D55 (AF544365)			1	Rhodococcus fascians	X79186	96
D57 (AF544366)		11	31	Rhodococcus fascians	X79186	100
D58 (AF544367)			1	Mycobacterium manitobense	AY082001	96
D61 (AF544368)			1	Dietzia maris	Y18883	95
D62 (AF544369)			1	Rhodococcus fascians	X79186	97

FIG.2.

Phylogenetic relationship of partial 16S rDNA sequences generated in this study. See Materials and Methods for description of tree construction. Scale bar represents the number of changes per base position. Numbers at tree nodes represent the number of times the topology to the right of the node was recovered in 100 bootstrap resamplings. Clone prefixes: T = 5- to 12-cm (bold) actinomycete community; M = 15 to 18 cm (bold italics), and D = 43 to 46 cm (bold underlined). The numbers in parentheses indicate the number of times the OTU was found in each of the libraries e.g., D57 (0, 11, 31) was not found in the 5- to 12-cm community and was present 11 and 31 times in the 15-to-18-cm and 43- to 46-cm communities, respectively.

The OTU representation was distinct for each of the sediment sections. The 5- to 12-cm sediment actinobacterium community was dominated by bacteria most closely related to Streptomyces species (45%), though OTUs from this sediment section were distributed throughout the phylogenetic tree (Fig. 2). OTUs from the 15- to 18-cm sediment section were dominated by bacteria most closely related to Rhodococcus species (56%), with only one OTU (M16) being most closely related to a Streptomyces species. The 43- to 46-cm community was also dominated by Rhodococcus species (62%), with no Streptomyces species present. No OTU was found in all sediment sections; however, two OTUs (represented by T41 and D05) were present in both the 5- to 12-cm and 43- to 46-cm sediment sections; one OTU (represented by T33) was present in both the 5- to 12-cm and 15- to 18-cm sections; and one OTU (represented by D57) was present in the 15- to 18-cm and 43- to 46-cm sections. The most dominant OTU in the 5- to 12-cm community was T33 (most closely related to the uncultured actinobacterium clone ASb07 [99%]), which comprised 5% of all actinobacterial 16S rDNA sequences recovered from this section. The 15- to 18-cm and 43- to 46-cm sediment communities were dominated by an OTU represented by D57 (most closely related to Rhodococcus fascians [100%]), comprising 23 and 50% of the communities, respectively, and 34% of all OTUs described.

Actinobacteria species richness.

Typical accumulation curves for microbial communities (19) were found by plotting the cumulative number of OTUs observed against the sampling effort. Curves indicated a slight increase in sample coverage, in proportion to total richness, with depth. However, the curves closely matched those expected when every sequence identified is an individual; hence, each sediment section was sampled with roughly equal intensity relative to its overall richness (19). Nonparametric estimation revealed a trend of decreasing species richness concomitant with increasing sediment depth (Table 2; Fig. 3A). The 95% CIs for the last estimate in each sediment section overlap and therefore it is not possible to reject the null hypothesis, at the significance level of 0.05, that there is no difference between species richness in the three sediment communities. The range of CIs in the 5- to 12-cm community continued increasing with sample size, and therefore it was not possible to determine how much more sampling was required to detect a significant difference in species number between this community and the two others. However, the range of CIs stabilized in the 15- to 18-cm and 43- to 46-cm communities allowing estimation of the effort required to detect a significant difference between these two communities by estimating the size of the CIs in larger samples (19). Thus, approximately 100 samples would be needed to detect a significant difference between these two communities (Fig. 3B).

TABLE 2.

Comparison of diversity indices in three deep-sea sediment core communities

Community	Diversity estimate
	No. of distinct sequences	No. of OTUsa		1/Db	Nucleotide diversity	θ (π)c
		ACE	Chao1
5-12 cm	53	616 (289, 3,379)	1,406 (307, 7,280)	51	0.11 ± 0.05	63.9 ± 31.0
15-18 cm	36	325 (114, 1,105)	308 (106, 1,103)	15	0.07 ± 0.04	42.1 ± 20.7
43-46 cm	30	226 (82, 765)	212 (75, 762)	4	0.05 ± 0.03	32.4 ± 15.9

^aThe values in parentheses are the 95% confidence intervals.

^bReciprocal of Simpson's index.

^cTotal genetic variation.

FIG. 3.

(A) Nonparametric (ACE) estimates of actinobacteria species richness as a function of sample size; 5 to 12 cm (○) (n = 56), 15 to 18 cm (n = 48), and 43 to 46 cm (▵) (n = 63). Error bars are standard deviations. Curves are averaged over 100 simulations using EstimateS. (B) Average size of the 95% CIs of the Chao1 estimate for the 15- to 18-cm and 43- to 46-cm communities. The curves are fitted negative exponential curves [15 to 18 cm, f(x) = 9,861 × 10^−0.0596x and r² = 0.97; 43 to 46 cm, f(x) = 7,433 × 10^−0.0491x and r² = 0.99].

Significance of difference between coverage curves.

LIBSHUFF (44) analysis of homologous and heterologous coverage curves indicated that the 5- to 12-cm community was significantly different from both the 15- to 18-cm and 43- to 46-cm communities (P = 0.001 [Fig. 4A and B ]). Comparison of the 15- to 18-cm and 43- to 36-cm communities suggested that they were not significantly different (P = 0.855 [Fig. 4C]). Further information on the differences between the 5- to 12-cm community and the two other communities was gained by examination of the distribution of (C_X − C_XY)² with D (44). At a D value of <0.10, the actual values of (C_X − C_XY)² are greater than the comparable values at a P value of 0.05 obtained during the calculation of ΔC (Fig. 4A and B). This result suggests that the 5- to 12-cm library differs greatly from the other two libraries at low levels of genetic distance but shares all deep taxa (D > 0.10). Comparison of the 15- to 18-cm and 43- to 46-cm libraries show that the actual values of (C_X − C_XY)² do not exceed the comparable values at a P value of 0.05 and differ only at D < 0.01, suggesting that nearly all taxa are present in both libraries (Fig. 4C).

FIG. 4.

LIBSHUFF comparisons of the three actinobacterial 16S rDNA libraries. Solid lines indicate the value of (C_X − C_XY)² for samples at each value of D. D is equal to the Jukes-Cantor evolutionary distance determined by the DNADIST program of PHYLIP. Broken lines indicate the P = 0.05 value of (C_X − C_XY)² for the randomized samples. (A) Comparison of 5- to 12-cm library (○ [homologous]) with 15- to 18-cm library (□ [heterologous]). (B) Comparison of 5- to 12-cm library (○ [heterologous]) with 43- to 46-cm library (▵ [homologous]). (C) Comparison of 15- to 18-cm library (□ [heterologous]) with 43- to 46-cm library (▵ [homologous]).

Reciprocal of Simpson's index.

The reciprocal of Simpson's index is sensitive to the level of dominance in a community and indicated that diversity decreased in proportion to depth. The 5- to 12-, 15- to 18-, and 43- to 46-cm communities had 1/D index values of 51, 15, and 4, respectively. Zhou et al. (61) suggest that values for 1/D below about 50 indicate typical dominance profiles; therefore, both the 15- to 18- and 43- to 46-cm communities show dominance, and the 5- to 12-cm community lies on the border between a dominant profile and a uniform profile.

Phylogenetic diversity.

Phylogenetic diversity between the three communities (i.e., methods not based on OTUs) was made initially using lineage-per-time plots (32). The plots were similar for the 5- to 12-cm and 15- to 18-cm communities (Fig. 5) and showed correspondence with trees resulting from a constant birth and death rate model (32); the only difference between the curves was a slight excess of divergent lineages in the 5- to 12-cm community. The 43- to 46-cm community plot displayed a convex shape indicating an excess of closely related lineages (Fig. 5). Phylogenetic diversity in the three samples was further compared using genetic differentiation tests. Table 2 shows the diversity estimates calculated using OTU definitions for each community compared with the diversity statistics generated from population genetics. In general, the statistics show the same pattern; distinct OTUs, estimated number of OTUs, nucleotide diversity, and average sequence divergence all decreased in proportion to depth. Genetic differentiation and covariation with phylogeny among the three sediment section communities were assessed by using F_ST and P tests (32). Use of the F_ST test showed differentiation between all sediment sections: 5 to 12 cm and 15 to 18 cm (F_ST = 0.14858; P < 0.00001), 5 to 12 cm and 43 to 46 cm (F_ST = 0.24842; P < 0.00001), and 15 to 18 cm and 43 to 46 cm (F_ST = 0.06293; P < 0.00001). In addition, the P test was highly significant (P < 0.001) for the three communities, indicating that the actinobacterial communities present at different depths covaried with phylogeny. Significance for both F_ST and P tests signified that there was less genetic diversity present in each individual community than there was for the three communities combined and that each community harbored distinct phylogenetic lineages (32). The OTU group represented by sequence D57 contained 11 and 31 sequences from the 15- to 18-cm and 43- to 46-cm communities, respectively. F_ST and P tests applied to the whole OTU group showed that the group was in fact comprised of clades of closely related sequences that were unique to either the 15-to-18- or 43- to 46-cm communities and that these clades were interspersed throughout the OTU group (F_ST, P < 0.00001; P test, P = 0.21). Separate analysis of the sequences in the OTU group from either the 15- to 18- or 43- to 46-cm communities indicated that they were drawn from the same pool of sequences (F_ST, P > 0.9; P test, P ≥ 0.5 for both).

FIG. 5.

Lineage-per-time plot for the 5- to 12-cm (○), 15- to 18-cm (□), and 43- to 46-cm (▵) actinobacteria 16S rDNA libraries. The ordinate is the number of lineages; the abscissa is time (arbitrary units) measured from the common ancestor. See Materials and Methods for a description of plot construction.

DISCUSSION

Diversity estimates and statistical comparisons.

The methods employed in this study enabled us to (i) compare quantitatively the coverage of actinobacterial 16S rDNA genes in the separate libraries; (ii) estimate species richness; and (iii) compare the phylogenetic diversity in each community (i.e., the genetic relatedness of the actinobacteria in the different sediment sections). The nonparametric estimators, the LIBSHUFF calculations, and the reciprocal of Simpson's index are all reliant on an OTU definition. In this study we used experimental data to define an actinobacterial OTU to provide the highest likelihood that sequences placed into OTU groups were generated from identical species. Some limitations in the use of OTUs for diversity statistics have been highlighted previously (19, 32), and our study prompts a further caveat: the use of 16S rRNA homology and DNA-DNA reassociation correlations for actinobacteria showed that even with a nonconservative OTU definition (i.e., ≤1% sequence divergence) the majority of species diversity will be overlooked. It is evident from the present study that when an OTU definition of ≤1% sequence divergence is applied, >70% of sequences placed into an OTU group originated from different species. Moreover, this observation is supported by empirical predictions of DNA-DNA relatedness when the similarity of 16S rDNA genes is known; for 99.8% similarity there is only a 50% probability that DNA-DNA relatedness is >70% (24). Palys et al. (36) state that species cutoff values for DNA-DNA reassociation (such as 70%) are arbitrary and are not guaranteed to yield groups of bacteria that correspond to real ecological units. The results of this study support this contention; F_ST and P tests applied to an OTU group revealed that it was comprised of clades unique to either sampling location. Thus, species with >99% 16S rRNA homology from two differing environments can clearly be distinguished. The application of phylogenetic diversity statistics (32) may enable researchers to identify species that form monophyletic clades that are unique to their environment, thereby enabling practical implementation of microbial species concepts where the environment is implicit in the definition, i.e., the phylospecies (49). Clearly, it is not possible to state definitively what level of DNA-DNA reassociation should be used for defining bacterial OTUs; hence, diversity estimates based on OTUs should be interpreted cautiously. However, the use of OTUs is appropriate for comparing relative richness when applied to data sets representing the same length and region of the 16S rDNA gene.

Phylogenetic diversity comparisons recently proposed for microbial studies require no definition of an OTU; the phylogenetic tree provides the information necessary for comparison (32). These methods are complementary to species richness estimates and provide a way of assessing whether the communities under investigation contain different phylogenetic groups regardless of the species number. However, phylogenetic comparisons are dependent on sequencing; diversity comparisons involving thousands of clones (61) are currently too costly to achieve using sequence analysis alone. Martin (32) compared phylogenetic diversity in previously published 16S rDNA clone libraries and showed that conclusions regarding diversity based solely on the frequency of different species (30) do not account for actual differences in the phylogenetic composition of the communities.

This study is the first to apply both microbial species richness and phylogenetic diversity estimates to a marine environment. Comparison of all diversity measures (Table 2) combined with the F_ST and P tests clearly shows that diversity in the sediment section decreases with depth and that each sediment section was comprised of distinct actinobacterial lineages. The only exception was the LIBSHUFF result for the 15- to 18-cm and 43- to 46-cm communities, predicting that both communities were composed of the same taxa. The LIBSHUFF program is a good test of overlap, since it considers the distribution of pairwise differences instead of the mean and variance (i.e., the F_ST test [A. P. Martin, personal communication]). However, it is not sensitive to the phylogenetic grouping of taxa (P test). Results of the LIBSHUFF program are dependent on sample size; the minimum number of sequences necessary to distinguish between two dissimilar libraries increases with library complexity and decreases with the magnitude of dissimilarity (44). Therefore, libraries composed of closely related taxa will be treated as equal within small sample sizes; this was tested by constructing two small artificial actinobacterial communities in which equal numbers of identical genera were present but in which the species represented in each library differed (data not shown). Results of the LIBSHUFF comparison for these communities indicated that the communities were identical (P = 1.00). Therefore, the results of LIBSHUFF for the 15- to 18-cm and 43- to 46-cm communities are probably due to the fact that the two libraries are composed largely of members of the same genera. A greater number of sequences are needed for LIBSHUFF to distinguish between these communities.

Mechanisms responsible for diversity.

Four mechanisms were recently proposed for the production of noncompetitive diversity profiles (61): (i) superabundant resources, (ii) resource heterogeneity, (iii) spatial isolation, and (iv) nonequilibrium conditions. An absence of these mechanisms will result in competitive diversity profiles. It is reasonable to assume that in marine sediments spatial isolation is not a factor, due to constant aqueous contact. Rank abundance profiles represented by uniform low population densities (e.g., 5- to 12-cm section) are consistent with resource limitation (61); hence, resource superabundance is unlikely to be maintaining species diversity. Furthermore, in order for resource superabundance to maintain diversity it must be such that species saturation never occurs, i.e., continuously abundant, or species spatially isolated (both unlikely in deep-sea sediments). However, it is possible that all species are maintained, but at suboptimal growth rates, thereby maintaining a species-unsaturated environment, eliminating competition and thus supporting diversity.

The 5- to 12-cm community displayed a borderline noncompetitive diversity profile (1/D = 51) with high species diversity estimates, indicating that the diversity of the 5- to 12-cm community is maintained by nonequilibrium and heterogeneous resource conditions. Thus, the two deeper communities, with proportionally lower estimates, are exposed to a decreasing gradient of the two above mechanisms. These factors seem tenable, since the majority of the carbon in the sediment is deposited by “marine snow” (settling of organisms living in the photic zone), which varies in its composition and deposition frequency (55). The surface sediment will also be subjected to hydrodynamics (deep-sea storms) and patch dynamics (carcass falls). These factors contribute to the nonequilibrium and resource heterogeneity present in the surface layers that are ameliorated in the deeper sections. Carbon inputs that do not originate from marine snow (i.e., upwellings) are likely to be homogeneous (e.g., methane) and serve to increase competition in the deeper sediments. The fact that the total organic carbon concentrations in the sediment sections were equal indicates that if the carbon resource directly influences diversity, it is due to heterogeneity rather than abundance. Other resources, such as the type and abundance of electron acceptors, may also influence diversity. Oxygen penetration into marine sediments is dependent on a number of factors including water flow rate and topography at the sediment-water boundary layer (14, 62), with values of <3 mm to >200 mm being reported (4, 59). Therefore, decreasing oxygen abundance in proportion to depth may be responsible for the concomitant decrease in diversity, with the fluxes at the boundary contributing to nonequilibrium conditions and hence increased diversity.

Lack of competition is not the only proposed model for diversity maintenance; other models for supersaturated coexistence invoke competition as the promoting force of diversity. Huisman et al. (20) analyzed multispecies models incorporating different physiological scenarios and plausible tradeoffs and showed that >100 species could be maintained on just three resources. Similarly, models computed using nonlinear formulae showed that competition could either promote or reduce diversity depending on resource utilization rates (56). Furthermore, and of particular interest to bioprospectors, the production of antimicrobial compounds may promote stable coexistence of huge numbers of species (10).

Selection of different populations.

The highest similarity for two clones present in both the 5- to 12- and 15- to 18-cm sediment sections was 99.8% (OTU group represented by T33). Therefore, the two species from which the clones originate diverged approximately 5 million to 11 million years ago (depending on the model for 16S rDNA nucleotide base substitution employed) (34, 35). Thus, it is likely that the differing populations observed in these two sediment sections did not diverge after their secondary sedimentation. Some of the phylogenetically distinct populations present in these two turbidite sections would have diverged during the deposition of the initial sediment from which the turbidite originated (15 m thick covering 250 million years), and others (with >4.5% 16S rDNA difference) would have diverged prior to the deposition of that sediment. Distinct populations would have been mixed during the formation and deposition of the turbidite, which occurs in a matter of days. Therefore, distinct ecological groups that evolved in a specific niche would survive and proliferate upon reformation of that niche (or an analogous one) after sedimentation. As mentioned above, the number of exploitable niches will decrease with depth, and therefore, burial of populations upon sedimentation will cause a selective sweep of the total population being deposited. The lineage-per-time plots give preliminary evidence for this selective sweep, showing a trend toward an excess of closely related species. The fact that clones D57 and M04 were identical suggests possible movement of species between the turbidite and pelagic clay, although it is not possible to state in which direction. It is important to note that divergence of species post-turbidite formation (i.e., in the last 1,000 years) has probably occurred through periodic selection and that it is the use of the 16S rDNA gene that prevents us from observing divergence on this time scale. Palys et al. (36) used protein-coding genes as a means to delineate ecologically distinct bacteria. These authors concluded that the inability of 16S rRNA sequence data to distinguish between some taxa was a consequence of the low evolutionary rate of the 16S rRNA genes. Thus, the use of markers with rapid evolutionary rates would be required if we wished to examine divergence after turbidite formation, such as insertion sequence fingerprinting on bacteria cultured from different sediment sections that have identical 16S rRNA genes (60).

Implications of diversity estimates.

The combined use of species richness and diversity estimates provides information that enables a deeper understanding of microbial biodiversity in areas such as diversity management, biogeography, and bioprospecting. To date it is still not clear how microbial diversity influences environmental functions, such as nutrient cycling, degradation of xenobiotics, and ecosystem stability. Therefore, it is difficult to predict how to successfully manage microbial diversity or even if management is necessary (i.e., species redundancy). Statistical diversity estimates that account both for phylogeny and richness of species applied in concert with hypotheses that test ecological theory may provide new insights into how diversity affects ecosystem function.

Staley and Gosink (50) give three reasons for the importance of bacterial biogeography: determination of how many bacterial species exist, species preservation, and identification of ecological roles through knowledge of bacterial distribution. Phylogenetic diversity estimates aid in the identification of bacterial species that are cosmopolitan or endemic. Endemism is suggested when clades of bacteria unique to a particular environment are identified, i.e., the P test. Furthermore, the F_ST test enables species to be identified that contribute to differences in the bacterial genetic diversity between two environments. Once cosmopolitan or endemic species are identified, their distribution and number can be estimated using richness estimates.

The implementation of techniques like those described in this study will enhance bioprospecting strategies in several respects. Thus, the molecular census of species gives information on species presence or absence in the environment being sampled; this information can be employed in the design of cultivation strategies. This methodology is not novel, but the combination of such information with the statistical diversity estimation methods provides the bioprospector with additional resources. First, coverage curves and richness estimators provide a means to assess how much further sampling is required. Second, comparison of the cultured subset to the molecular data provides a means to assess how many of the target species are present in the screening stage, and correlation of these data to the success of the screening (i.e., how many novel compounds have been identified) will determine whether or not continued cultivation and screening efforts should be undertaken and allow analogous environments to be identified for future screens.

Niche adaptation in microbes through periodic selection implies that the use of composite samples (one sample containing distinct niches) for the cultivation of microbes will bias recovery in favor of those best adapted to the culture medium; those with large cell volumes, high ribosome concentrations, and multiple rRNA operons (13, 25). Application of diversity estimators coupled with diversity indices sensitive to the level of dominance in a community enables the discrimination of environments, either geographically isolated or as specific niches within one sample, and therefore, appropriate methods can be employed to reduce culture bias, such as physical separation of the delineated environments, use of low-nutrient or polymeric growth substrates (22, 40), or dilution to extinction culture (43).