This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Zo, Y.-G.
Right arrow Articles by Colwell, R. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Zo, Y.-G.
Right arrow Articles by Colwell, R. R.
Agricola
Right arrow Articles by Zo, Y.-G.
Right arrow Articles by Colwell, R. R.

 Previous Article  |  Next Article 

Applied and Environmental Microbiology, January 2009, p. 135-146, Vol. 75, No. 1
0099-2240/09/$08.00+0     doi:10.1128/AEM.02894-07
Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Diversity and Seasonality of Bioluminescent Vibrio cholerae Populations in Chesapeake Bay{triangledown}

Young-Gun Zo,1,2,{dagger} Nipa Chokesajjawatee,1,3 Christopher Grim,1,2 Eiji Arakawa,4 Haruo Watanabe,4 and Rita R. Colwell1,2*

Center of Marine Biotechnology, University of Maryland Biotechnology Institute, 701 E. Pratt St., Baltimore, Maryland 21202,1 Center of Bioinformatics and Computational Biology, University of Maryland Institute of Advanced Computer Studies, University of Maryland—College Park, College Park, Maryland 20742,2 National Center for Genetic Engineering and Biotechnology, 113 Phahonyothin Rd., Klong 1, Klong Luang, Pathumthani 12120, Thailand,3 Department of Bacteriology, National Institute of Infectious Diseases, Toyama 1-23-1, Shinjuku-ku, Tokyo 162-8640, Japan4

Received 21 December 2007/ Accepted 6 November 2008


arrow
ABSTRACT
 
Association of luminescence with phenotypic and genotypic traits and with environmental parameters was determined for 278 strains of Vibrio cholerae isolated from the Chesapeake Bay during 1998 to 2000. Three clusters of luminescent strains (A, B, and C) and two nonluminescent clusters (X and Y) were identified among 180 clonal types. V. cholerae O1 strains isolated during pandemics and endemic cholera in the Ganges Delta were related to cluster Y. Heat-stable enterotoxin (encoded by stn) and the membrane protein associated with bile resistance (encoded by ompU) were found to be linked to luminescence in strains of cluster A. Succession from nonluminescent to luminescent populations of V. cholerae occurred during spring to midsummer. Occurrence of cluster A strains in water with neutral pH was contrasted with that of cluster Y strains in water with a pH of >8. Cluster A was found to be associated with a specific calanoid population cooccurring with cyclopoids. Cluster B was related to cluster Y, with its maximal prevalence at pH 8. Occurrence of cluster B strains was more frequent with warmer water temperatures and negatively correlated with maturity of the copepod community. It is concluded that each cluster of luminescent V. cholerae strains occupies a distinct ecological niche. Since the dynamics of these niche-specific subpopulations are associated with zooplankton community composition, the ecology of luminescent V. cholerae is concluded to be related to its interaction with copepods and related crustacean species.


arrow
INTRODUCTION
 
Historically, Vibrio cholerae has been recognized solely as a human pathogen, causing severe watery diarrhea in humans that leads to either serious morbidity or death. Regardless of the severity of the symptoms in individual cases, however, the most harmful characteristic of the bacterium is its capacity to cause pandemic cholera, so that the number of individuals within a human population affected by this bacterium can rise quickly, causing a severe impact on the daily lives of a population, unless proper preventive measures are swiftly enacted (39). Based on a growing understanding of the ecology of V. cholerae, the pandemic capacity of pathogenic strains can be understood to result from its characteristic life cycle within its natural habitat (39). V. cholerae occurs naturally in aquatic environments throughout all seasons, even under conditions unfavorable for its growth and multiplication. It proliferates when conditions become favorable (10). The ecology of this species derives from its being an indigenous, natural inhabitant of the aquatic environment (10, 54). Therefore, an improved understanding of the ecology of V. cholerae will give greater possibilities for modeling and accurate prediction of the dynamics of cholera epidemics.

V. cholerae has been suggested to be moderately clonal (5), comprising diverse and distinctive subpopulations. Each subpopulation is furnished with unique genetic resources so that its ecological niches are diverse (28). Indeed, only a subpopulation of V. cholerae, mainly those strains producing a heat-labile cholera toxin, causes pandemic cholera, whereas other subpopulations can cause illness by producing a heat-stable enterotoxin (3, 29). The ecological niches of V. cholerae, therefore, may well be related to many other traits that vary among these distinctive subpopulations.

Bioluminescence occurs in some strains of V. cholerae (32, 53), and this observation prompts various hypotheses related to ecological function and a potential role in pathogenesis or symbiosis (19, 30, 51). The enzyme responsible for luminescence, luciferase, can account for as much as 5% of cellular protein and 20% of total oxygen consumption. In some strains, the energy committed to luminescence is 10% of the total metabolism (30). Yet luminescence is widely distributed among marine and estuarine bacteria (51). Therefore, luminescence must be an important ecological characteristic for an aquatic bacterium, since it carries a high energy cost. It is a trait expressed via quorum sensing and occurs in symbiosis, both of which are specialized responses for alternative life cycles (27, 55). In the case of V. cholerae, luminescence has been reported to occur frequently in a variety of nontoxigenic environmental isolates. However, luminescence is less frequent among cholera-causing strains and their relatives. Numerical taxonomy studies have determined the frequency to be approximately 10% of strains tested (25, 53). Bioluminescence in V. cholerae, therefore, poses an ecological question with respect to cholera epidemics, that is, whether it is a key property distinguishing survival strategies of cholera epidemic versus nonepidemic strains. In any case, the ecological role of luminescence in V. cholerae is intriguing.

In this study, we investigated the ecological significance of luminescence in V. cholerae strains isolated from a brackish water ecosystem by analyzing the incidence of luminescence among diverse V. cholerae lineages. By examining the coupling of luminescence with other bacterial traits, we found that luminescence is distributed differentially by phylogenetic lineage, showing a significant association with heat-stable enterotoxin in luminescent strains. Furthermore, the occurrence of luminescent V. cholerae strains was found to be highly seasonal, and certain environmental variables were found to be associated with presence/absence for specific luminescence lineages of V. cholerae.


arrow
MATERIALS AND METHODS
 
Ecological survey.
Strains of V. cholerae were isolated during field surveys carried out in the Chesapeake Bay from 1998 to 2000. Details of sampling stations and methods used to isolate the strains were reported previously (24). In brief, V. cholerae was isolated at five shore stations (B [Baltimore Inner Harbor], F [Susquehanna river flat], S [subestuary of the Rhode River], K [Kent Island], and H [Horn Point]) in the upper Chesapeake Bay (39°17.00'N to 38°58.84'N, 76°02.20'W to 76°36.32'W) and nine mid-bay, north-south transect stations. Shore sampling was done biweekly during the summer and monthly during other seasons of the year. Samplings at the north-south transect stations were conducted aboard the research vessel Cape Henlopen on four occasions: once in each month of August and September 1999 and in June and August 2000. The transect stations were named as follows: 908 (39°/08.00'N, 76°/20.00'W); 858 (38°/58'N, 76°/23'W); 845 (38°/45.00'N, 76°/26.00'W); 834 (38°/34.00'N, 76°/26.00'W); 818 (38°/18.00'N, 76°/17.00'W); 804 (38°/04.00'N, 76°/13.00'W); 744 (37°/44.00'N, 76°/11.00'W); 724 (37°/24.00'N, 76°/05.00'W); 707 (37°/07.00'N, 76°/07.00'W).

Station F is a freshwater site, based on salinity measurements. At the 13 other stations, a salinity gradient was observed, increasing from north to south in the Chesapeake Bay. Mean salinity ranged from 7 to 20{per thousand}. For purposes of data analysis, each sample collected was labeled using a string of characters made up of a one-letter site name, followed by the last digit of the year, the last digit of the month, and two digits for the day of sampling.

Suspended particulates, including planktonic organisms, in the surface water (0- to 1-m depth) were collected by filtering 250-ml water samples through 0.2-µm-pore membrane filters or by sieving or towing ca. 150 liters of surface water sequentially through two nylon nets (64-µm and 20-µm mesh sizes, respectively), resulting in three size-fractionated samples (W = particles of ≥0.02 µm; P = 20- to 64-µm particles; Z = particles of ≥64 µm). Particulate matter retained on the filters was inoculated into alkaline peptone water enrichment flasks, from which presumptive V. cholerae strains were isolated by streaking on thiosulfate-citrate-bile salts-sucrose agar plates after incubation for 6 to 8 h. Presumptive colonies were identified as V. cholerae using a simple biochemical test method (8) and PCR to target a phylogenetic signature in the intergenic transcribed sequence (ITS) region between 16S rRNA and 23S rRNA (8, 9). In 1999 and 2000, samples of particulate matter were also used as inocula in direct plating on alkaline-peptone agar plates (1% peptone, 1% NaCl, pH 8.6, 1.5% agar). For the latter, a 32P-labeled oligonucleotide probe with the ITS sequence was hybridized to colony lift blots to identify presumptive V. cholerae (8). From this survey, a collection of 278 V. cholerae isolates was established.

Characterization of V. cholerae isolates.
The Sakazaki O serotype (40) for each isolate was determined using a pool of 210 standard polyclonal antibodies developed and maintained in the Department of Bacteriology, National Institute of Infectious Diseases, Tokyo, Japan. In addition to serotyping, 24 phenotypic and 8 genotypic characteristics were employed: growth in nutrient broth (Difco, Detroit, MI) containing 0%, 1%, 3%, 6%, or 8% of NaCl and at 42°C; acid production from sucrose, arabinose, mannose (Mns), and mannitol (Mnt); production of enzymes, i.e., ornithine decarboxylase, lysine decarboxylase, arginine dihydrolase, oxidase, gelatinase, amylase, lipase, and chitinase; response to the methyl red test (MR) and Voges-Proskauer test; hydrolysis of esculin; and sensitivity to 10 and 150 mg of vibriostatin agent O/129 (V150) and 50 U of polymyxin B.

Isolates were also tested for the presence/absence of six toxin-related genes and two outer membrane protein genes, using dot blot hybridization. The toxin genes included those encoding cholera toxin subunit A (ctxA), toxin coregulated pilus subunit A (tcpA), zonular occludens toxin (zot), heat-stable enterotoxin (stn), and hemolysin subunit A (hlyA). The genes for the inner membrane transmembrane protein, ToxR, an important regulatory gene for cholera toxin, and the genes encoding two outer membrane proteins (OmpU and OmpW) were targeted.

For dot blotting, genomic DNA was extracted using the DNeasy tissue kit (Qiagen Inc., Valencia, CA) and eluted in 200 µl elution buffer AE (10 mM Tris-HCl [pH 9.0], 0.5 mM EDTA). Genomic DNA (500 ng) was denatured in 0.4 M NaOH-10 mM EDTA and heated to 100°C for 10 min to ensure complete denaturation. DNA dots were blotted onto a MagnaCharge nylon membrane (MSI [Micron Separations Inc.], Massachusetts), using the Bio-Dot microfiltration apparatus (Bio-Rad Laboratories, Hercules, CA). After blotting, the DNA was immobilized on nylon membranes by UV cross-linking (UV cross-linker; Fisher Scientific, Pittsburgh, PA) at 120 mJ cm–2 for 30 s. Genomic DNA from V. cholerae strains (ATCC 14035T, O1 classical biotype; N16961, O1 El Tor biotype; RC4, O139 El Tor biotype; and RC66, O14 stn+), Vibrio mimicus ATCC 33653T, and Aeropyrum pernix K1, an archaebacterium, were included on the blots, serving as positive or negative controls.

Hybridization probes were generated using a PCR digoxigenin (DIG) probe synthesis kit (Roche Diagnostics, GmbH, Mannheim, Germany). Primer sequences for amplification of each gene probe are listed elsewhere (56). All probes were amplified using 40 ng of template genomic DNA in 50-µl reaction volumes. V. cholerae N16961 ctx+ tcpA+ zot+ hlyA+ ompU+ ompW+ {Delta}stn, where "{Delta}" denotes the absence of a trait, was used as the template to generate gene probes, except for the stn probe, in which case V. cholerae RC66 stn+ was used. Hybridization and detection were done using DIG hybridization and the CSPD protocol (Roche Diagnostics, GmbH, Germany), respectively. In brief, the membranes were prehybridized at 42°C in DIG-EasyHyb solution for 30 min and hybridized in the presence of each gene probe (1 µl ml–1) at 42°C overnight. Membranes were subjected to a low-stringency wash with 2x SSC (1x SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.1% sodium dodecyl sulfate for 5 min twice at room temperature and a high-stringency wash with 0.5x SSC-0.1% sodium dodecyl sulfate for 15 min twice at 65°C. Signals were detected using the CSPD chemiluminescent substrate, followed by autoradiography at room temperature for 5 to 30 min.

Determination of luminescent V. cholerae.
Luminescence was determined by two methods: detection of the luxA gene directly and detection of luminescence in broth cultures. The luxA+ strains were identified by dot blot hybridization. The 650-bp-long, DIG-labeled DNA fragment was generated via PCR on genomic DNA of V. cholerae ATCC 14547 luxA+, using the PCR primer pair VCluxA108F and 757R, the sequences of which are reported elsewhere (15). Dot blot hybridization of the luxA probe was performed with the same protocol used for the other genes listed above. Expression of luminescence (Lum) by each strain was determined using the luminescence assay method of Grim et al. (15). Test strains were subcultured on marine agar 2216e (Difco, Detroit, MI) plates and incubated overnight at 30°C. Single colonies were inoculated into marine broth 2216e (Difco) medium and cultured overnight at 30°C with shaking at 200 rpm, after which the broth cultures were diluted 1:500 or 1:1,000 using fresh broth. Luminescence was measured at incremental time points using an LB96P luminometer (EG&G Berthold, Germany).

Phylogenetic analyses.
Results of the phenotypic and genotypic tests were coded as binary or multistate characters for each given trait of interest. In the case of O serotyping, an isolate was assigned to one of 212 O serogroups, determined by adding the two untypeable cases, i.e., rough strain and new serogroup, not recognized in the existing 210 Sakazaki serogroups. For genotypes assayed by dot blot hybridization, five states were established, namely, absence (e.g., {Delta}hlyA) and four allele groups (hlyA1+, hlyA2+, hlyA3+, or hlyA4+), determined as four levels of signal intensity. We interpreted the absence of targeted genes from insignificant difference (one-way analysis of variance [ANOVA]; P ≥ 0.05) of hybridization signals from the negative control, i.e., A. pernix K1, an archaebacterium lacking homologues for any of the assayed genes. Four groups of alleles were determined for an isolate when hybridization signals were significantly higher than that for A. pernix K1. The groups were determined by recoding the signal intensity semiquantitatively on a scale of 1 to 4, in steps of 0 to 25%, 25 to 50%, 50 to 75%, and 75 to 100% of signal intensity of the positive control. Two alleles in a different allele groups were interpreted as displaying phenotypic differences. An exception to this five-state recoding of genotype data was luxA. Because luxA produced either 0% or >90% signal intensity, binary coding was used. For luminescence, we created a three-state variable (state 0 = {Delta}luxA; state 1 = luxA+ {Delta}Lum; state 2 = luxA+ Lum+) by merging Lum and luxA into one character variable. For all other phenotypic tests, binary coding was used for presence or absence of responses.

Using the entire set of traits for each isolate, an operational taxonomic unit (OTU) was established. Isolates sharing an identical set of characters were recorded as belonging to the same OTU and were treated as identical clones. Distances between OTU pairs were expressed as numbers of different phenotypic characteristics, or gene loci, and hierarchal cluster analysis was performed, using the complete linkage method. This method was determined to be appropriate for maximizing homogeneity within a cluster while minimizing the influence of inequality of phylogenetic signals in three kinds of traits, i.e., phenotypic, genotypic, and serotypic. To test a cluster for the presence of significant phylogenetic signals, the permutation tail probability (PTP) test was used (2, 14). To compare support for two different tree topologies, the topology-dependent permutation tail probability (T-PTP) test (13), implemented as the "compare-2" method in the PAUP* version 4.0b10 software program (44), was used. For both PTP tests, the minimum tree length of the maximum-parsimony (MP) trees under the Wagner model was used as the test statistic for probability calculation. To test the significance of a split in a phylogenetic tree, Sokal and Rohlf's randomization test for distinctness (12, 42) and T-PTP test for monophyly were performed for each bifurcating split. For the former, 20,000 permutations by the Monte Carlo simulation method were used, and significance was assessed by using a one-tailed test.

Genomic DNA-DNA hybridization.
As a measure of relatedness among isolates, genomic DNA of isolates randomly selected to represent phylogenetic clusters were dot blotted and hybridized with alkaline phosphatase-labeled genomic DNA of V. cholerae O1 El Tor N16961 and V. cholerae O43 RC395.

The genomic DNA (500 ng) was dot blotted onto nylon membranes, as described above. The DNA of the probe genome was sheared to an approximate size of 400 to 600 bp by sonication and labeled using thermostable alkaline phosphatase (GeneImages AlkPhos Direct labeling kit; Amersham Biosciences Ltd., United Kingdom). The hybridization buffer and washing solution were prepared following the manufacturer's protocol. The membrane was prehybridized at 60°C for 30 min and hybridized (10 ng ml–1 probe) at 60°C overnight in a rotary hybridization tube. Each membrane was subjected to a high-stringency wash twice with primary wash buffer for 10 min at 70°C, followed by a low-stringency wash twice with secondary wash buffer for 5 min at room temperature. Chemifluorescent signals were generated using the ECF substrate (Amersham Biosciences). The fluorescent signals were recorded using a Storm 840 instrument (Molecular Dynamics Inc., Sunnyvale, CA), and the signal intensity was quantified by ImageQuant software, version 5.1 (Molecular Dynamics, Inc.). The results are expressed in relative binding units (relative binding ratio [RBR]), giving the ratio of signal from the target DNA to that from the probe DNA itself (i.e., positive control) as the target DNA. Mean RBR from duplicated blotting and probing was used to infer similarity of genomes of isolates to the probe strains.

Environmental variables.
Water temperature, salinity, pH, chlorophyll a concentration, and total bacterial number (TBN) for the 31 water samples that yielded at least one V. cholerae isolate in the survey reported by Louis et al. (24) were included in the analysis. Normality in the distribution of each variable was tested by using the Shapiro-Wilk's W statistic after appropriate transformation. For TBN, logarithmic transformation was used, with pH values normalized by taking the negative power of 10, yielding values of acidity of water in H+ ion concentration (20).

The zooplankton community structure of each water sample (24) was considered to comprise a separate set of environmental variables explaining variation in the V. cholerae population structure. Zooplankton composition data were available in units of relative abundance for 15 taxa. Taxonomic levels of the 15 taxa ranged from phylum to suborder. While adult copepods were identified as calanoid, cyclopoid, or harpacticoid (order-level taxa), copepods in the premature instar stages were pooled as subclass Copepoda, without further identification. Instar stages were classified to two levels: copepod nauplii and copepodites. According to conventional taxonomic schemes (4, 7), other crustaceans enumerated at the level of order were amphipods and cumaceans, within the class Malacostraca. Crustaceans enumerated and identified to the class level were ostracods. Barnacles (cirripedes), the subclass Cirripedia, were also enumerated as their nauplii: therefore as barnacle larvae. Cladocerans, i.e., suborder Cladocera, were the only members identified and enumerated within the class Branchiopoda. Insect larvae comprised noncrustacean arthropods, while oligochaetes, polychaetes, nematodes, and rotifers were included in the nonarthropod taxa. Zooplankton taxonomic composition was converted to prevalence in a sample by dividing by the total number of zooplankton in a sample.

Statistical analyses.
In the statistical analyses and tests, significance was determined at the 5% type I error level. To explore relationships between variation of the V. cholerae OTUs and environmental variables, the ANOVA framework, implemented by redundancy analysis (RDA) (36) with the F-ratio-based permutation test (6), was used by employing the software program CANOCO for Windows, version 4.5 (48). This multivariate analysis is more sensitive than performing a set of bivariate analyses for each of pairs of a response variable and explanatory variable because it can anticipate interactions among response variables, which is typical in a community of mixed populations with ecological similarity.

RDA was performed on the prevalence of V. cholerae subpopulations as the multivariate response variable. The prevalence of each subpopulation was defined as the ratio of the number of OTUs to the number of all OTUs in the V. cholerae population isolated from a given enrichment flask or sample. Prevalence values were transformed by the square root to calculate the Hellinger distance during ordination analyses (23). Explanatory variables were either the five environmental variables or the relative prevalence of zooplankton taxa. The significance of the canonical environmental axes was tested using the permutation test on F-ratio values by generating 499 data sets permuted with toroidal shift, keeping the temporal order of sampling at a site.

When a significant association was identified by RDA, bivariate analyses for each pair of a response variable and significant explanatory variable were performed for the purposes of validating the RDA result and finding an explanation for the significance of RDA. Correlation analysis, linear regressions, and logistic regressions were performed on prevalence and presence/absence of OTUs or on RDA residuals by employing appropriate packages in the statistical computing software environment R version 2.4.1 (35).


arrow
RESULTS AND DISCUSSION
 
Bioluminescence.
Of 278 strains, 149 possessed the luxA gene, and the luminescence assay detected expression of the luminescence gene in 136 of these strains. Thus, 13 strains in which the luxA gene was present did not express luminescence in this assay. Of 32 bacterial traits, excluding serotyping, luminescence was the most variable. Approximately half of the strains (54% and 49%) were positive by the luxA probe and the luminescence assay, respectively. The next-most-variable traits were Mnt and Mns, for which 58% and 62% were positive, respectively. In the case of the serotyping results, 49 Sakazaki O-serogroup antigen types were found, together with two strains carrying an O-serogroup antigen that was not typeable by the existing pool of polyclonal antibodies. Ten rough strains were detected, and these were not typeable. Within the 49 serogroups, O135 was predominant, with 35 strains belonging to this serogroup. Because serotyping yielded the largest number of alleles, it was judged the most variable trait of those tested in this study.

The collection of strains included redundant clones, that is, more than one presumed identical strain was present, most likely a result of the enrichment culture method used to isolate V. cholerae. In subsequent analyses, clonal redundancy was removed by defining those clones as OTUs that had a unique combination of the traits tested in this study. Of 278 strains, 180 combinations of traits were found, indicating a clonal redundancy of 1.5. That is, there were two kinds of clones for every three strains. Of 180 types, 80 were luxA+. Therefore, luminescence was concluded to be a phylogenetically and ecologically important characteristic of environmental isolates of V. cholerae, based on its wide distribution among the different OTUs. This observation also implies the presence of a positive natural selection for luminescence trait for some V. cholerae populations, since this energetically demanding trait is widely distributed among environmental V. cholerae strains. Considering that expression of the bioluminescence trait is unlikely to be neutral due to the high energy demand, together with the fact that most of the other species in the family Vibrionaceae are luxA+ by vertical inheritance (50), the observation that distribution of the luminescence trait was limited to only some V. cholerae subpopulations may also imply the presence of a negative natural selection against some V. cholerae subpopulations being luxA+. Therefore, understanding the diversity of luminescent V. cholerae in the context of overall diversity of the species is a prerequisite for the study of the ecological role of the luminescence trait in the species.

Diversity of luminescent V. cholerae.
Diversity was determined by complete linkage cluster analysis, based on the number of different characteristics between a pair of OTUs. We used several distance-based methods and the MP method in cluster analysis and found the complete linkage method produced clusters in best agreement with the ABY|XC split, described in the legend to Fig. 1a and c. The cholera toxin-related genes (ctxA, tcpA, and zot) were absent in all isolates of V. cholerae. Phenotypic traits that were invariable among the isolates were growth in nutrient broth containing 0%, 1%, 3%, and 8% NaCl, acid production from arabinose, production of enzymes such as arginine dihydrolase, oxidase, and gelatinase, hydrolysis of esculin, and sensitivity to 10 mg of vibriostatic agent O/129. Based on variability in serogrouping (Fig. 1a) and the 21 variable traits (Fig. 1b), the diversity of the V. cholerae isolates was analyzed. The PTP test was used to determine the terminal cluster in which no significant phylogenetic structure could be inferred, and 17 such clusters (P ≥ 0.10) were found (Fig. 1a). Luminescent V. cholerae OTUs tended to form homogenous clusters, but these are better understood as three distinct clusters (cluster A, comprised of terminal clusters A1 to -6, cluster B, and cluster C) rather than a single cluster. The nonluminescent cluster Y, comprised of terminal clusters Y1 to Y4, was present as a sister cluster to cluster A or B. Cluster X (terminal clusters X1 to X5) was distinguished from the others mainly by low signal strength in DNA-probe hybridization with toxR and ompU (Fig. 1b).


Figure 1
View larger version (43K):
[in this window]
[in a new window]

  		FIG. 1. Phylogeny of Vibrio cholerae based on phenotypic and genotypic characteristics. (a) Hierarchical clustering by complete linkage analysis. Terminal clusters (A1 to A6, B, C, X1 to X5, and Y1 to Y4) were determined by the PTP test (P ≥ 0.05). The O serogroup for each cluster, including rough strain (R) and new serogroup (OUT) not recognized in the existing 210 Sakazaki serogroups, is listed, with the number of OTUs as a superscript. Open triangles, luxA+; closed triangles, {Delta}luxA; the scale bar shows the number of characters. All splits shown in the dendrogram were significant by both the distinctness and T-PTP tests (P < 0.05). The split between clusters ABY and XC, corresponding to the split in panel c (dashed line), is marked with a rectangle. MP analyses were performed with subclusters of clusters ABY and XC, and splits supported by MP trees are marked with filled circles. (b) Pie charts showing prevalences (%) of phenotypic and genotypic characteristics for OTUs of V. cholerae in the terminal clusters. Columns: leftmost column, cluster label, ordered as in the dendrogram; rightmost column (n), number of OTUs in the clusters; middle columns, phenotypic and genotypic traits, showing variability among clusters. Pie chart colors are as follows: black (Absent), prevalence of OTUs lacking the given trait; dark gray (Group 1), prevalence of OTUs showing 0 to 25% hybridization signal with a DIG-labeled DNA probe; light gray (Group 2), prevalence of OTUs showing 25 to 50% hybridization; very light gray (Group 3), prevalence of OTUs showing 50 to 75% hybridization; white (Group 4/Present), prevalence of OTUs showing 75 to 100% hybridization or positive response for binary tests (columns with nonitalic font abbreviations). Pie charts for binary phenotypic tests comprised only "Absent" or "Group 4/Present." The color key is located at the bottom of panel b. In cluster analysis and PTP tests, the Lum and luxA traits (boxed columns) were merged into a three-state trait (state 0 = {Delta}luxA; state 1 = luxA+ {Delta}Lum; state 2 = luxA+ Lum+). Traits are as follows: NaCl6, growth in nutrient broth containing 6% of NaCl; G42oC, growth at 42°C; Suc, Mns, and Mnt, acid production from sucrose, mannose, and mannitol; production of enzymes, i.e., ornithine decarboxylase (Orn), lysine decarboxylase (Lys), amylase (Amy), lipase (Lip), and chitinase (Chi); MR and VP, response to the methyl red test and Voges-Proskauer test; V150, sensitivity to 150 mg of vibriostatin agent O/129; PB, sensitivity to 50 U of polymyxin B. (c) Similarity of the V. cholerae genome to those of the probe strains N16961 and RC395, based on DNA-DNA hybridization. RBR, relative binding ratio. Letters used inside the figure show major clusters of the isolates (A, B, C, X, and Y); the dashed line indicates a slope of 1. The k-means clustering method produced two distinct clusters: one comprising all strains above the dashed line and the other all strains below the dashed line, with centers at (0.26,0.70) and (0.62,0.40), respectively, corroborating the significance of the split between clusters ABY and XC. (d) Cladograms for three possible splits of clusters A, B, and Y4. The arrow points to the tree with the split best supported by the compare-2 T-PTP test. P, tail probability; n, number of random permutations.

When these clusters were compared with clusters formed on the basis of genome similarity estimates by k-means clustering, support for the split between clusters ABY and XC was strongly reinforced (Fig. 1c). The V. cholerae O1 El Tor N16961 strain, used as a probe strain in the analysis, is a representative of the seventh cholera pandemic strains (18), and cluster Y1 is determined to be the closest relative of N16961 in cluster analysis. Cluster Y1 harbored two other OTUs with the phenotypic and genotypic profiles identical to those of V. cholerae N16961, except for serotype, namely, O28 and R, and pathogenesis traits encoded by the two mobile elements CTX{Phi} prophage and the toxin coregulated pilus island. The other probe strain, V. cholerae O43 RC395, was selected as a random pick from the most abundant serotype of cluster X, belonging to cluster X1. Thus, the RBR values shown in Fig. 1c provide the relative similarities of the genomes of the OTU to those of clusters X1 and Y1. Six strains of 22 cluster X isolates showed a higher RBR to N16961 than to the RC395 probe. The genomic similarity of the 85 other isolates indicated a split between clusters ABY and XC with respect to a higher similarity to probe strains representing clusters X and Y. However, the RBR data did not resolve the internal structure of cluster ABY.

Without the Lum and luxA traits, 140 out of 180 OTUs maintained the A, B, C, X, and Y cluster while 40 OTUs changed cluster membership. Among these were four OTUs of cluster X that moved to cluster A, violating the empirically determined split between clusters ABY and XC in Fig. 1c. As demonstrated by a recent lux operon phylogenetics study (50), lux genes are nearly entirely vertically inherited in Vibrionaceae and thus have strong phylogenetic signals. Therefore, the phylogenetic signals in the lux operon contributed to delineation of the 40 OTUs in Fig. 1a, yielding clusters congruent with the split shown in Fig. 1c.

Since the results of cluster analysis indicated that cluster Y comprised a nonluminescent counterpart to clusters A and B, the branching order of clusters A, B, and Y became of particular interest, notably to provide an understanding of whether luminescent V. cholerae strains comprise a monophyletic cluster or multiple phylogenetic groups. According to results of the cluster analyses, Y4 was the nonluminescent cluster most closely related to the luminescent clusters. Subtle differences in phenotypic traits (amylase and MR) and genotypic traits (hlyA and ompU) were noted between clusters B and Y4 (Fig. 1b), while the strongest contrast was observed in luminescence and serogroup (Fig. 1a). According to results of the PTP tests, these differences provided sufficient evidence of heterogeneity between the two clusters. The compare-2 T-PTP test was applied to compare the three possible branching orders involving cluster Y4 (Fig. 1d). The phylogenetic standing of cluster B is best supported as a sister cluster of Y4, and this result supported a significant phylogenetic split and gradient between clusters A and Y. When the same set of hypotheses was tested with each terminal cluster, A1 to A6, instead of the entire set of OTUs of cluster A, the hypothesis of cluster Y4 being closer to cluster B than cluster A1 to A6 was generally accepted (data not shown). Notable was the absence of any evidence for relatedness of terminal clusters of cluster A to cluster B, while significant evidence supporting relatedness between cluster B and cluster Y4 was found. Therefore, it is concluded that luminescence occurs within many phylogenetic subsets of V. cholerae. Furthermore, cluster B of this study represents an intermediate luminescent cluster between clusters A and Y, with a higher similarity to the nonluminescent cluster Y. This is further supported by the observation that 1 OTU in cluster B lacked the luxA gene and 8 of 18 luxA+ OTUs in cluster B did not express luminescence (Fig. 1b). Thus, it is suggested that regulation for lux gene expression in cluster B strains differs from that of the cluster A strains.

The presence of multiple luminescent clades also provides an interpretation of the evolution of luminescence in V. cholerae. With respect to distribution of luminescence in clusters A, B, and Y, two competing hypotheses can be proposed, that luminescent OTUs in cluster B are relatives of OTUs of cluster Y, having acquired luminescence by gene transfer, or that luminescence was lost in the OTUs of cluster Y via divergence from a common luminescent ancestor of clusters A, B, and Y, arising from counterselective pressure in the ecological niche of cluster Y. The relatively low rate of expression of luminescence by cluster B isolates (Fig. 1b) supports both hypotheses in different ways. For the hypothesis of selective acquisition, the observation implies a difference in the regulatory system for the luminescence gene expression. For the hypothesis of a counterselective loss, the observation can be interpreted as a relatively high prevalence of a nonfunctional lux operon consisting of one or more pseudogenes. Comparative studies of the structure and regulation of the lux operon of cluster A and B strains should be able to determine which of the two alternative evolutionary pathways occurs within V. cholerae populations in nature.

Another important question in the evolutionary population biology of V. cholerae is whether the cholera-pathogenic clade can carry the luminescence trait or whether the trait is not compatible with the ecological niche of this V. cholerae clade. Previous studies showed that V. cholerae forms clonal subpopulations and that cholera epidemic clades are highly clonal (5, 57). The V. cholerae O1 El Tor N16961 strain, a representative of the V. cholerae El Tor lineage responsible for the current, seventh pandemic of cholera, revealed 21 genotypic and phenotypic traits identical to those of two OTUs of cluster Y1. Differences of the cluster Y1 strains from the O1 El Tor reference strain were serogroup (O28 and rough type versus O1 and O139) and pathogenicity. When the presence/absence of the pathogenicity genes was included in the cluster analysis, V. cholerae N16961 formed a long branch within cluster Y1, suggesting phylogenetic separation between the closest relatives of epidemic lineages and luminescent clades (clusters A, B, and C) of V. cholerae.

Traits linked to bioluminescence.
After all trait data were coded in binary format, the association coefficient {phi} (42) was calculated and a chi-square test of independence was performed between luxA and other traits (Table 1). To distinguish a luxA-specific association of traits from a general phylogenetic cohesion of traits arising from fixation of traits in large phylogenetic clusters, we analyzed the association of luxA with other traits by comparing a pair of sister clusters: one cluster with luminescent strains and its sister cluster without luminescent strains. Since clusters A, B, and Y were monophyletic, comparisons of cluster A versus cluster Y and cluster B versus cluster Y met the criterion. In cluster AY, the luminescence trait was positively associated with Mns+, {Delta}Mnt, MR+, ompU4+, stn3+, and stn4+, where "{Delta}" denotes the absence of a trait in an OTU. Since ompU4+ genes of toxigenic strains, including V. cholerae N16961, are known to code for bile resistance (34), intestinal adherence (43), and acid tolerance (26) and stn4+ is the gene encoding enterotoxin (31), the pattern of cohesion of traits in cluster AY suggested an association of luminescence with enteropathogenic properties for environmental V. cholerae strains. In cluster BY, luxA+ was linked with {Delta}Mns and {Delta}Mnt, suggesting cohesion of luminescence with limited sugar utilization. Thus, the single trait in common with luxA in strains of both clusters A and B was a lack of the ability to utilize mannitol ({Delta}Mnt). The two OTU types in cluster C (Fig. 1b) were based on only three isolates from the ITS oligonucleotide hybridization on colonies from the S9830 water sample; therefore, association analysis was not statistically meaningful. However, since all three cluster C strains utilized mannitol, it is concluded that none of the traits analyzed in this study were linked to luxA across the three luminescent clusters.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Prevalence and association coefficients ({phi}) of luxA with other traitsa

In conclusion, several traits were found to be significant in distinguishing a luminescent cluster from its nonluminescent sister cluster. However, none of the traits were linked to luxA across all of the luminescent clusters, indicating lineage-specific cohesion of the lux operon with other traits. One mechanism for this lineage specificity is that bioluminescence in each luminescent cluster plays a different ecological role, i.e., a different ecological niche for each of the luminescent clusters. The microhabitat and genomic composition of the three clusters are different. For cluster A, with luxA cohesion to ompU4+ and stn4+, interaction with eukaryotic organisms may be an important characteristic, since OmpU in pathogenic V. cholerae strains encodes bile resistance, intestinal adherence, and acid tolerance and stn4+ encodes enterotoxin. Significant association with specific O serogroups (Table 1) suggests luminescent subpopulations of V. cholerae are determined by their interaction with eukaryotic hosts. Understanding the dynamics and ecological role of luminescent V. cholerae can be achieved by studying subpopulations of luminescent V. cholerae, not in the aggregated set of entire subpopulations.

Spatial and temporal variation.
Since it was recognized, from the results described above, that the population of luminescent V. cholerae strains comprises phenotypic, genotypic, and possibly ecological subpopulations, localization of luminescent subpopulations at specific sites or in a specific season was analyzed. To compare V. cholerae population composition across different sites and samples, the entire set of OTUs isolated from 64 enrichment flasks from 31 samples collected at three shore sites (K, H, and S) from April to September was included in the analysis. Isolates obtained by the colony-lift hybridization method or from enrichment flasks from the nine midbay, north-south transect stations were excluded from the analysis of spatiotemporal dynamics because only a few of the samples from each site yielded V. cholerae isolates. Cluster C strains were also excluded because all had been isolated from a single sample, S9830W, only by the ITS-oligonucleotide probe hybridization method.

The composition of V. cholerae populations in each enrichment flask was calculated as a prevalence of clusters A, B, X, and Y by converting presence/absence of OTUs in each flask into the relative proportion of OTUs each cluster comprised. Therefore, the conversion was done in two steps: tabulation of presence/absence of each OTU in each flask and enumeration of OTUs present in a flask by cluster.

At the three sites, K, H, and S, at least seven samples yielded isolates by enrichment. The composition of OTUs according to the four major clusters is presented in Fig. 2 by time. For sites H and S, a common pattern of seasonal succession was observed: that is, a succession of the predominant V. cholerae nonluminescent OTUs in April to June 1998 to luminescent OTUs in July and August, 1998. On 5 August 1998, isolates were entirely luminescent strains, whereas isolates were almost entirely nonluminescent in samples collected on 16 April and 28 May 1998. The seasonal succession observed at the sites H and S in 1998 was not detected in 1999, mainly because V. cholerae could be isolated from only a single sample at each site in 1999. Samples in 1999 contained microorganisms competing with V. cholerae in the isolation method used, resulting in a low success rate for V. cholerae isolation. Therefore, the interannual stability of the observed seasonality could not be assessed from the results of this study, but the spatial stability of bioluminescence seasonality was supported, since the same pattern of seasonality was observed at the two different sites located on opposite sides of the Chesapeake Bay. Notable was that sites H and S showed similar ranges of variation in environmental parameters (24) and are located in subestuaries of the Choptank River and Rhode River, respectively. At site K, seasonal succession was not observed, but the presence of a mixed population from June to August 1998 was noted.


Figure 2
View larger version (27K):
[in this window]
[in a new window]

  		FIG. 2. Relative composition of V. cholerae OTUs, by cluster, in each of the enrichment flasks receiving the indicated size-fraction inoculum. Labels for fractions: W, particles of ≥0.02 µm; P, 20- to 64-µm particles; Z, particles of ≥64 µm. Labels for samples: site is indicated by the first letter, the last digit of the year by the first number, the month by the second number, and the day by the last two numbers. Height of bar = 100% when isolates were obtained or 0% when no isolate was obtained from a flask. Numbers under fraction labels indicate total numbers of OTUs in each flask.

To quantify and test the significance of graphical seasonal succession at sites H and S in 1998 (Fig. 2), a logistic regression, employing seasons as an independent variable and the presence of luminescent OTUs as a response variable, was performed. Sampling dates were divided into two seasons: spring (April to June) and summer (July to August). The log-odd ratio of the summer population being luminescent was 2.8 ± 1.1 (standard error) and 3.5 ± 1.2 at sites H and S, respectively, implying that the odds for V. cholerae being luminescent during the summer was 17 and 34 times higher, on average, than that for its being nonluminescent (P < 0.01; likelihood ratio test). For site K, the regression coefficient of 0.16 ± 0.2 was not significant (P = 0.85), implying the absence of seasonal succession at this site.

The significance of a pattern of site-specific seasonal succession of the V. cholerae populations was also tested for the four groups (clusters A, B, X, and Y) as response variables, employing RDA with the permutation test. The prevalences for the four groups were square root transformed and used as response variables in the RDA, while site, date, and inoculum fraction were used as explanatory factors in a three-way factorial design. V. cholerae compositions of samples collected at site K were significantly different from those for the other sites (P = 0.04; permutation test on F ratio; n = 499). The effects of sampling date were significant for sites H and S (P < 0.03) but not for site K (P = 0.45). The effect of the three-kind inoculum size fractions (W, P, and Z) was not significant, either for all sites or for each site. The interaction term between the date and the fraction was not significant (P = 0.42). Since the inoculum fraction was not a significant factor, the three fractions of each sample were treated as replicates from the same sample, a logical interpretation for the P and Z fractions because they were from the same volume of water sample and were sequentially filtered. The reason that fraction W and the other fractions were comparable is that the effect by difference in volume and particle size was not large enough to cause a significant difference in the bacterial composition detected by enrichment, which is believed to have a relatively high sampling error. Another explanation is that the high success rate for the W fraction compared to the others overcame the volume difference between W and the other two inoculum preparations. Even though sample volumes for the W fraction were smaller, enrichment flasks receiving W fraction inocula yielded a larger number of isolates and more frequently than the other fractions, as shown by Q statistics in the work of Louis et al. (24), by having smaller numbers of competitors.

In conclusion, seasonal succession from nonluminescent to luminescent populations was observed for shore site samples collected at H and S, and luminescent V. cholerae were more prevalent during midsummer (June to August) than in the spring. However, this seasonal pattern differed according to site, exemplified by site K. Therefore, significant spatial and temporal variation was observed in the dynamics of luminescent V. cholerae populations, with site-specific seasonal succession. The possibility of an association of luminescence with selected parameters of the estuarine environment undergoing site-specific seasonal changes was therefore analyzed.

Association of environmental parameters with bioluminescence.
The relationship of environmental parameters to the dynamics of luminescent V. cholerae subpopulations was analyzed by RDA. Environmental parameters recorded for the shore survey sites were water temperature, salinity, pH, chlorophyll a concentration, and TBN. Results for the midbay transect sampling stations were not included in this analysis because only a few of the environmental parameters were recorded and only for a limited number of samples. The composition of V. cholerae populations in each shore sample was expressed as the prevalence of clusters A, B, and Y among the entire set of unique OTUs detected in the three enrichment flasks (W, P, and Z) for each shore sample. The prevalence of cluster X was excluded in this analysis to avoid artificial colinearity among the response variables and to take its substantial phylogenetic distinction into account.

When the data for all five shore sites were analyzed, the only significant factor accounting for the distribution of the three clusters was the interaction term for salinity and pH, which explained 25% of the total variance in V. cholerae composition (P < 0.01; permutation test on F ratio; n = 499). The significant interaction originated from the fact that site F was essentially freshwater throughout the year and the other environmental parameters of site F varied in concert with their variations at the other brackish water sites. In other words, pH and salinity were coupled at the brackish water sites but not at the freshwater site. To analyze relationships between the V. cholerae population structure and environmental parameters without the effect of pH and salinity, site F was excluded from subsequent analyses.

For sites B, K, H, and S, forward selection in multiple regressions in the RDA framework of CANOCO identified three significant explanatory factors: pH, TBN, and interaction between temperature and TBN. The pH explained the first 29% of the total variance in the prevalence of A, B, and Y clusters, with high significance (P < 0.01). The interaction of temperature and TBN explained the next 17% of the total variance (P = 0.03), whereas TBN alone explained 7% of the total variance (P = 0.03).

Based on the RDA results, the relationship of V. cholerae composition with pH was examined in detail by bivariate regression analyses. As shown in Fig. 3a, the prevalences of clusters A and Y showed a monotonic relationship with pH while being antagonistic to each other. In contrast, cluster B showed unimodal distribution with an optimum range intermediate to those of the other clusters. This relationship was also confirmed by results of logistic regression employing the presence/absence of each cluster in each sample as the response variables (Fig. 3a). The probability of occurrence of clusters A and Y was related to pH as half-sections of the Gaussian response curve (P = 0.31 and P = 0.72, respectively; goodness-of-fit test). The probability of occurrence of cluster B appeared as a typical Gaussian logit distribution, with the optimum pH 8.0 (P = 0.29; goodness-of-fit test). A Gaussian logit distribution of a species along an environmental gradient is the most general form of an environment-species relationship (47). Prevalence values of each cluster showed a good fit to the Gaussian logit distribution of probability along a pH gradient (Fig. 3a), validating the view of a unimodal distribution of bacterial subpopulations along an environmental gradient, e.g., the pH gradient.


Figure 3
View larger version (17K):
[in this window]
[in a new window]

  		FIG. 3. Distribution of prevalence of V. cholerae clusters along an environmental gradient. (a) Relationship of cluster prevalence with pH. Symbols: as shown in panel b. Lines: probability of presence (p) of each cluster, determined by logit regression analysis of presence/absence of a cluster, i.e., logit(p) = b0 + b1x + b1x2, where x = 10–pH. (b) Relationship between TBN and residual prevalence (y) of RDA to pH. Lines show y for each cluster, determined by the quadratic regression equation y = b0 + b1x + b1x2, where x is TBN. (c) Relationship between temperature and residual prevalence (y) of RDA to pH. Symbols are as shown in panel b. Line = y for B cluster, determined by linear regression (y = b0 + b1x, where x is temperature).

Based on phylogenetic analyses, cluster B is more closely related to cluster Y than to cluster A (Fig. 1a and d). However, cluster B was luminescent, a characteristic absent in cluster Y. Cluster B can be thought of as an OTU that was added, with luminescence, to the genomic foundation of cluster Y. The responses of clusters A, B, and Y to pH, demonstrated in Fig. 3a, indicate that the habitat of cluster B is also an intermediate between those of luminescent cluster A and nonluminescent cluster Y. Therefore, we could observe congruence between the phylogenetics and ecological characteristics of V. cholerae. This result indicates that evolution of the clades of V. cholerae is driven by ecological selection, and both phylogenetic analysis and the RDA approach employed in this study cross-validate the appropriateness of the two methods.

TBN can be interpreted as an indicator of the carrying capacity of a body of water for heterotrophic bacteria, and in this study, it significantly accounted for residual variances of V. cholerae composition, after pH, in two forms: interaction with temperature and TBN alone. When residuals from the effect of pH were examined, with regard to TBN or temperature variation (Fig. 3b and c, respectively), the source of interaction between temperature and TBN was determined to be the single outlying sample B9422. This sample was the only one that yielded V. cholerae isolates among samples collected during a period of low temperature, in spring, with a low TBN. The relationship between environmental conditions and the composition of V. cholerae in the early spring sample was different from those of other samples collected from late spring to early fall. When regression analysis with residual prevalence of the latter group of samples was done, clusters A and Y revealed a significant quadratic relationship with TBN (P < 0.04; likelihood ratio test) (Fig. 3b) while cluster B showed a linear proportional relationship with temperature (P = 0.01; likelihood ratio test) (Fig. 3c).

Based on these observations, the pattern of seasonal succession among V. cholerae subpopulations described in the previous section can be explained by the seasonal variation in environmental parameters promoting or restricting the presence of a given subpopulation. TBN, temperature, and pH were significant factors controlling diversity within the V. cholerae population. It is well known that clinical V. cholerae strains have an optimum pH in the alkaline range (49). Typical clinical and epidemic V. cholerae strains were more closely related to cluster Y than other clusters of strains isolated from the Chesapeake Bay, and this major nonluminescent cluster was also predominant when the pH of the water was high in the shore waters of the upper Chesapeake Bay. To the contrary, luminescent clusters A and B had an optimal pH close to neutral. Luminescence in V. cholerae is concluded to be associated with bodies of water with neutral or slightly alkaline pH, while cholera-pathogenic V. cholerae is likely to be found in microhabitats within bodies of water with a selectively high pH (>8.0). A high pH in land and estuarine water is typically linked to conditions of high water temperature, which reduces the solubility of atmospheric CO2 (52), high planktonic primary production, which converts dissolved CO2 into organic carbon, and a high input of NH4+ via decomposition of aminogenic organic matter carried out by bacteria and other consumers in the aquatic food web (41).

Analysis of the habitat characteristics of cluster B showed a significant linear proportional relationship between the residual prevalence of cluster B and the water temperature (Fig. 3c), occurring together with a slightly reverse proportional relationship between cluster A prevalence and temperature, implying that the relative abundance of cluster B, with respect to that of cluster A, varied according to seasonal variation of water temperature during late spring to early fall. The underlying mechanisms for the response of V. cholerae to the seasonal and spatial variation in pH and temperature can be either a direct physiological modulation of bacterial activity by the environmental parameters or indirect. In the case of TBN, a direct physiological modulation of community composition is less likely.

Association with zooplankton composition.
Many laboratory and field studies have shown that V. cholerae strains can be found associated with phytoplankton (21), zooplankton (37, 46), and insects (16). Based on these established facts, the association of the prevalences of luminescent V. cholerae subpopulations with pH, TBN, and temperature suggests the possibility that aquatic fauna and flora might provide or function as microhabitats for V. cholerae and variation in the abundance of the flora and fauna of the estuary is an indirect mechanism of this association. Since the aquatic fauna and flora also undergo seasonal succession, a specific association of a plankton taxon with a specific V. cholerae subpopulation can also explain the seasonal shifts in prevalence of luminescent V. cholerae. From the previous study of Louis et al. (24), the zooplankton community composition, in terms of relative abundance, was available. Thus, we analyzed the taxon-specific association of zooplankton taxa and luminescent V. cholerae subpopulations by RDA.

For samples from sites B, K, H, and S, both V. cholerae subpopulation prevalences and zooplankton by taxon prevalences were available for 22 samples. Based on the mean prevalence of each taxon, the two crustacean nauplii (copepod nauplii and cirripede nauplii) were judged predominant (62% ± 27% [standard deviation] and 11% ± 12%, respectively), followed by calanoids at a mean prevalence of 8% ± 20%. In general, adult copepods (calanoids, cyclopoids, and harpacticoids) ranked within the top nine predominant taxa with >1% mean prevalence. Rotifers ranked as the fourth predominant group and cladocerans as the eighth, with mean prevalences of 5% ± 10% and 2% ± 5%, respectively. Based on these statistics, the predominant taxa were determined as the taxonomic hierarchy of the crustacean class, the copepod subclass, and the calanoid order. Rotifers and cladocerans were also recognized as exceptions to this pattern since they were abundant groups depending on the season. From these data, the a priori hypothesis developed was that crustacean prevalence is a determinant of luminescent V. cholerae distribution, and this was tested.

Explanatory variables employed in the ANOVA model for RDA were juvenile copepods (nauplii and copepodites), the three adult copepod taxa (calanoids, cyclopoids, and harpacticoids), the other crustacean taxa (cladocerans, ostracods, cumaceans, amphipods, and cirripede nauplii), rotifers, and their first-order interactions. The model explained 73% of the total variance with a P value of 0.03 (permutation test on F-ratio; n = 499), which is close to the amount explained by all zooplankton taxa without interaction terms (79%). The significant factor providing the most variance in the model for prevalences of clusters A, B, and Y was interaction between calanoids and cyclopoids (hereinafter denoted "cal x cycl"). In an examination of the relationship between cyclopoids and calanoids, two kinds of trends were observed: a group of samples showed covariation of calanoids and cyclopoids, and the other samples harbored calanoids without any correlation with cyclopoids. The interaction term corresponded to the former group of samples, showing correlation with cyclopoid prevalence (Fig. 4a). The latter group of samples was represented in the RDA model as the effect of calanoids after accounting for the effect of "cal x cycl." Reports of the predominance of two species of calanoid copepods (Acartia tonsa and Eurytemora affinis) in the mesozooplankton communities of the Chesapeake Bay, with a strong seasonal and spatial variation in a mutually exclusive manner (22, 38), indicate the presence of two kinds of predominant calanoid distribution in the Chesapeake Bay. The first canonical axis of the RDA model was significant (P < 0.05; permutation test on F-ratio; n = 499) and reflected the heterogeneity in the relationship of calanoids to V. cholerae composition by showing significant correlation between sample scores of the axis and calanoid or cyclopoid prevalence (Pearson coefficient; r = 0.52, P < 0.05 or r = –0.55, P < 0.08 for calanoids or cyclopoids, respectively). The axis was positively correlated with prevalences of clusters B and Y, while a negative relationship was observed with cluster A prevalence. Considering that the two kinds of calanoids are predominant, this result indicates that the canonical axis represents the gradient for which each calanoid species predominates in the system. These results indicate that phylogenetic diversity among bioluminescent V. cholerae populations can be related to the species-level composition of the calanoids.


Figure 4
View larger version (18K):
[in this window]
[in a new window]

  		FIG. 4. Relationship of the prevalence of V. cholerae clusters with zooplankton composition. (a) Biplot for correlation of zooplankton taxa prevalence (thin arrows) and RDA scores for prevalence of V. cholerae cluster (thick arrows) relative to the first two canonical axes. Symbols: *, larvae or nauplii; "cal x cycl" (cal*cycl), interaction term between calanoids and cyclopoids. (b) Relationship of the prevalence of V. cholerae with the copepod maturity index for the samples (maturity 1 = nauplii; maturity 2 = copepodites; maturity 3 = adult copepods). The line shows the probability of presence (p) for each cluster, determined by logit regression of the presence/absence of cluster B, i.e., logit(p) = b0 + b1x, where x is the copepod maturity index.

For copepods, a gradient comprised of the prevalence of adult copepods versus that of copepod juveniles was determined and was correlated with the V. cholerae population structure. A copepod maturity index for each sample was calculated by taking the mean of the maturity value for each copepod in each sample (Fig. 4b) and computing the bivariate relationship as presence/absence or prevalence of each V. cholerae cluster. By using logistic regression, a significant negative trend for the cluster B copepod maturity index was determined (log-odd ratio = –4.9 ± 2.4 [standard error] [P = 0.04]; likelihood ratio test). Cluster B included strains isolated from a sample in which adult copepods were predominant. Thus, a distinctive ecological niche for V. cholerae subpopulations appears to be related to the age distribution of the copepod population with which it is associated.

Role of luminescence in V. cholerae.
In the case of transparent planktonic animals, the luminescence of bacterial prey can make them an easy target for carnivorous predators. Luminescent bacterial populations can be protected by having their predator numbers reduced by the next consumer in the food web or by their own pathogenicity. Therefore, luminescence in bacterial cells can function as an important modulator of ecological relationships among consumers in an aquatic food web.

In this study, a numerical correspondence of V. cholerae subpopulations and zooplankton community composition was demonstrated. The distinction of cluster A versus cluster Y in association with the dynamics of calanoid copepods is notable. A specificity of the cluster A strains toward a calanoid copepod that covaries with cyclopoids was observed, while cluster Y strains showed specificity toward other calanoid species. This taxon-to-taxon specific association between bacterial clones and aquatic plankton can moderate the food web structure of the ecosystem by affecting predation behaviors of plankton and animals. The best-documented cases of taxon-specific associations are symbioses with animals possessing light organs. In the case of copepod zooplankton, attachment of luminescent V. cholerae to zooplankton or their detritus is sufficient to cause a change in the food web structure by discouraging predation by translucent predators (e.g., planktonic crustaceans) or by promoting predation by nontranslucent predators (e.g., fish). The age-specific association of cluster B with the juvenile copepod community suggests such a taxon-specific association can be fine-tuned, depending on the chemical and nutritional composition of the organisms or the ecological standing of the host organisms in a specific season.

The association of luminescent V. cholerae clusters with the enterotoxin gene stn suggests another hypothesis for the ecological role of bioluminescence: a parasitic life cycle of luminescent V. cholerae. As in the case of a bioluminescence-induced increase in predatory mortality of copepods grazing on dinoflagellates (1), luminous plankton biomass or particles are more likely to be eaten by fish, namely trout, herring, mackerel, and others with visually selective predation, using vision tuned to 480-nm-wavelength light (1, 33). By making a substrate to which it is attached luminous, the bacterium can be more rapidly transported into a nutrient-rich environment of the animal gut and initiate a new cycle of parasitic growth. The quorum sensing of V. cholerae (17), in which high cell density, as in biofilms, permits expression of the lux operon, also supports this life cycle.

Considering results of this study together with the observation that bioluminescence in vibrios functions as an intracellular light source to stimulate light-induced DNA damage repair mechanisms (11) and to detoxify deleterious oxygen derivatives (45), it is concluded that luminescence in V. cholerae can have several effects, ranging from the physiology of individual cells to food-web-wide ecology. Taxon-specific and age-specific association of the life histories (i.e., seasonal dynamics) between a luminescent subpopulation of V. cholerae and zooplankton and linkage of luminescence with the enterotoxin (stn) and cell surface (O-serotype) genes suggest that an essential role of luminescence in V. cholerae is to moderate interactions of V. cholerae with aquatic animals, including both plankton and nekton.


arrow
ACKNOWLEDGMENTS
 
This research was supported by National Institutes of Health grant no. 1R01A139129-01 and National Science Foundation grant no. 01200677. Young-Gun Zo was supported by the second-phase Brain Korea 21 Project in 2008.

We gratefully acknowledge helpful comments provided by Petr Smilauer concerning the permutation methods employed in this research.


arrow
FOOTNOTES
 
* Corresponding author. Mailing address: Center for Bioinformatics and Computational Biology, University of Maryland Institute of Advanced Computer Studies, University of Maryland—College Park, College Park, MD 20742. Phone: (301) 405-9550. Fax: (301) 314-6654. E-mail: rcolwell{at}umiacs.umd.edu Back

{triangledown} Published ahead of print on 14 November 2008. Back

{dagger} Present address: Department of Environmental Science, Kangwon National University, Chuncheon 200-701, Republic of Korea. Back


arrow
REFERENCES
 
    1
  1. Abrahams, M. V., and L. D. Townsend. 1993. Bioluminescence in dinoflagellates: a test of the burglar alarm hypothesis. Ecology 74:258-260.[CrossRef]
    2. 2
  2. Archie, J. W. 1989. A randomization test for phylogenetic information in systematic data. Syst. Zool. 38:239-252.[Abstract/Free Full Text]
    3. 3
  3. Arita, M., T. Takeda, T. Honda, and T. Miwatani. 1986. Purification and characterization of Vibrio cholerae non-O1 heat-stable enterotoxin. Infect. Immun. 52:45-49.[Abstract/Free Full Text]
    4. 4
  4. Barnes, R. D. 1987. Invertebrate zoology, 5th ed. Saunders College Publishing, Philadelphia, PA.
    5. 5
  5. Beltran, P., G. Delgado, A. Navarro, F. Trujilto, R. K. Selander, and A. Cravioto. 1999. Genetic diversity and population structure of Vibrio cholerae. J. Clin. Microbiol. 37:581-590.[Abstract/Free Full Text]
    6. 6
  6. Besag, J., and P. Clifford. 1989. Generalized Monte Carlo significance tests. Biometrika 76:633-642.[Abstract/Free Full Text]
    7. 7
  7. Bisby, F., Y. Roskov, M. Ruggiero, T. Orrell, L. Paglinawan, P. Brewer, N. Bailly, and J. van Hertum. 2007. Species 2000 & ITIS catalogue of life: 2007 annual checklist taxonomic classification CD-ROM. Species 2000, Reading, United Kingdom.
    8. 8
  8. Choopun, N., V. Louis, A. Huq, and R. R. Colwell. 2002. Simple procedure for rapid identification of Vibrio cholerae from the aquatic environment. Appl. Environ. Microbiol. 68:995-998.[Abstract/Free Full Text]
    9. 9
  9. Chun, J., A. Huq, and R. R. Colwell. 1999. Analysis of 16S-23S rRNA intergenic spacer regions of Vibrio cholerae and Vibrio mimicus. Appl. Environ. Microbiol. 65:2202-2208.[Abstract/Free Full Text]
    10. 10
  10. Colwell, R. R., and W. M. Spira. 1992. The ecology of V. cholerae, p. 107-127. In D. Barua and W. B. Grennough III (ed.), Cholera. Plenum Medical Book Co., New York, NY.
    11. 11
  11. Czyz, A., K. Plata, and G. Wegrzyn. 2003. Stimulation of DNA repair as an evolutionary drive for bacterial luminescence. Luminescence 18:140-144.[CrossRef][Medline]
    12. 12
  12. Deutsch, R., M. Cherner, and I. Grant. 2006. Significance testing of a cluster of multivariate binary variables: comparison of the tripartite T index to three common similarity measures. Stat. Methods Med. Res. 15:285-299.[Abstract/Free Full Text]
    13. 13
  13. Faith, D. P. 1991. Cladistic permutation tests for monophyly and nonmonophyly. Syst. Zool. 40:366-375.
    14. 14
  14. Faith, D. P., and P. S. Cranston. 1991. Could a cladogram this short have arisen by chance alone? On permutation tests for cladistic structure. Cladistics 7:1-28.[Medline]
    15. 15
  15. Grim, C. J., E. Taviani, M. Alam, A. Huq, R. B. Sack, and R. R. Colwell. 2008. Occurrence and expression of luminescence in Vibrio cholerae. Appl. Environ. Microbiol. 74:708-715.[Abstract/Free Full Text]
    16. 16
  16. Halpern, M., Y. B. Broza, S. Mittler, E. Arakawa, and M. Broza. 2004. Chironomid egg masses as a natural reservoir of Vibrio cholerae non-O1 and non-O139 in freshwater habitats. Microb. Ecol. 47:341-349.[Medline]
    17. 17
  17. Hammer, B. K., and B. L. Bassler. 2007. Regulatory small RNAs circumvent the conventional quorum sensing pathway in pandemic Vibrio cholerae. Proc. Natl. Acad. Sci. USA 104:11145-11149.[Abstract/Free Full Text]
    18. 18
  18. Heidelberg, J. F., J. A. Eisen, W. C. Nelson, R. A. Clayton, M. L. Gwinn, R. J. Dodson, D. H. Haft, E. K. Hickey, J. D. Peterson, L. Umayam, S. R. Gill, K. E. Nelson, T. D. Read, H. Tettelin, D. Richardson, M. D. Ermolaeva, J. Vamathevan, S. Bass, H. Qin, I. Dragoi, P. Sellers, L. McDonald, T. Utterback, R. D. Fleishmann, W. C. Nierman, O. White, S. L. Salzberg, H. O. Smith, R. R. Colwell, J. J. Mekalanos, J. C. Venter, and C. M. Fraser. 2000. DNA sequence of both chromosomes of the cholera pathogen Vibrio cholerae. Nature 406:477-483.[CrossRef][Medline]
    19. 19
  19. Herring, P. 2002. Marine microlights: the luminous marine bacteria. Microbiol. Today 29:174-176.
    20. 20
  20. International Union of Pure and Applied Chemistry. 1985. Definitions of pH scales, standard reference values, measurement of pH, and related terminology. Pure Appl. Chem. 57:531-542.[CrossRef]
    21. 21
  21. Islam, M. S., Z. Rahim, M. J. Alam, S. Begum, S. M. Moniruzzaman, A. Umeda, K. Amako, M. J. Albert, R. B. Sack, A. Huq, and R. R. Colwell. 1999. Association of Vibrio cholerae O1 with the cyanobacterium, Anabaena sp., elucidated by polymerase chain reaction and transmission electron microscopy. Trans. R. Soc. Trop. Med. Hyg. 93:36-40.[CrossRef][Medline]
    22. 22
  22. Kimmel, D. G., and M. R. Roman. 2004. Long-term trends in mesozooplankton abundance in Chesapeake Bay, USA: influence of freshwater input. Mar. Ecol. Prog. Ser. 267:71-83.[CrossRef]
    23. 23
  23. Legendre, P., and E. D. Gallagher. 2001. Ecologically meaningful transformation for ordination of species data. Oecologia 129:271-280.[CrossRef]
    24. 24
  24. Louis, V. R., E. Russek-Cohen, N. Choopun, I. N. G. Rivera, B. Gangle, S. C. Jiang, A. Rubin, J. A. Patz, A. Huq, and R. R. Colwell. 2003. Predictability of Vibrio cholerae in Chesapeake Bay. Appl. Environ. Microbiol. 69:2773-2785.[Abstract/Free Full Text]
    25. 25
  25. McNicol, L. A., S. P. De, J. B. Kaper, P. A. West, and R. R. Colwell. 1983. Numerical taxonomy of Vibrio cholerae and related species isolated from areas that are endemic and nonendemic for cholera. J. Clin. Microbiol. 17:1102-1113.[Abstract/Free Full Text]
    26. 26
  26. Merrell, D. S., C. Bailey, J. B. Kaper, and A. Camilli. 2001. The ToxR-mediated organic acid tolerance response of Vibrio cholerae requires OmpU. J. Bacteriol. 183:2746-2754.[Abstract/Free Full Text]
    27. 27
  27. Miller, M. B., K. Skorupski, D. H. Lenz, R. K. Taylor, and B. L. Bassler. 2002. Parallel quorum sensing systems converge to regulate virulence in Vibrio cholerae. Cell 110:303-314.[CrossRef][Medline]
    28. 28
  28. Miller, M. C., D. P. Keymer, A. Avelar, A. B. Boehm, and G. K. Schoolnik. 2007. Detection and transformation of genome segments that differ within a coastal population of Vibrio cholerae strains. Appl. Environ. Microbiol. 73:3695-3704.[Abstract/Free Full Text]
    29. 29
  29. Morris, J. G., T. Takeda, B. D. Tall, G. A. Losonsky, S. K. Bhattacharya, B. D. Forrest, B. A. Kay, and M. Nishibuchi. 1990. Experimental non-O group 1 Vibrio cholerae gastroenteritis in humans. J. Clin. Investig. 85:697-705.[Medline]
    30. 30
  30. Nealson, K. H., and J. W. Hastings. 1979. Bacterial bioluminescence: its control and ecological significance. Microbiol. Rev. 43:496-518.[Free Full Text]
    31. 31
  31. Ogawa, A., J. Kato, H. Watanabe, B. G. Nair, and T. Takeda. 1990. Cloning and nucleotide sequence of a heat-stable enterotoxin gene from Vibrio cholerae non-O1 isolated from a patient with traveler's diarrhea. Infect. Immun. 58:3325-3329.[Abstract/Free Full Text]
    32. 32
  32. Palmer, L. M., and R. R. Colwell. 1991. Detection of luciferase gene sequence in nonluminescent Vibrio cholerae by colony hybridization and polymerase chain reaction. Appl. Environ. Microbiol. 57:1286-1293.[Abstract/Free Full Text]
    33. 33
  33. Porter, K. G., and J. W. Porter. 1979. Bioluminescence in marine plankton: a coevolved antipredation system. Am. Nat. 114:458.[CrossRef]
    34. 34
  34. Provenzano, D., and K. E. Klose. 2000. Altered expression of the ToxR-regulated porins OmpU and OmpT diminishes Vibrio cholerae bile resistance, virulence factor expression, and intestinal colonization. Proc. Natl. Acad. Sci. USA 97:10220-10224.[Abstract/Free Full Text]
    35. 35
  35. R Development Core Team. 2006. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
    36. 36
  36. Rao, C. 1964. The use and interpretation of principal component analysis in applied research. Sankhyaá Ser. A 26:329-358.
    37. 37
  37. Rawlings, T. K., G. M. Ruiz, and R. R. Colwell. 2007. Association of Vibrio cholerae O1 El Tor and O139 Bengal with the copepods Acartia tonsa and Eurytemora affinis. Appl. Environ. Microbiol. 73:7926-7933.[Abstract/Free Full Text]
    38. 38
  38. Roman, M., X. Zhang, C. McGilliard, and W. Boicourt. 2005. Seasonal and annual variability in the spatial patterns of plankton biomass in Chesapeake Bay. Limnol. Oceanogr. 50:480-492.
    39. 39
  39. Sack, D. A., R. B. Sack, G. B. Nair, and A. K. Siddique. 2004. Cholera. Lancet 363:223-233.[CrossRef][Medline]
    40. 40
  40. Shimada, T., E. Arakawa, K. Itoh, T. Okitsu, A. Matsushima, Y. Asai, S. Yamai, T. Nakazato, G. B. Nair, M. J. Albert, and Y. Takeda. 1994. Extended serotyping scheme for Vibrio cholerae. Curr. Microbiol. 28:175-178.[CrossRef]
    41. 41
  41. Soetaert, K., A. F. Hofmann, J. J. Middelburg, F. J. R. Meysman, and J. Greenwood. 2007. The effect of biogeochemical processes on pH. Mar. Chem. 105:30-51.[CrossRef]
    42. 42
  42. Sokal, R. R., and F. J. Rohlf. 1995. Biometry: the principles and practice of statistics in biological research, 3rd ed. W. H. Freeman & Co., New York, NY.
    43. 43
  43. Sperandio, V., J. A. Giron, W. D. Silveira, and J. B. Kaper. 1995. The OmpU outer membrane protein, a potential adherence factor of Vibrio cholerae. Infect. Immun. 63:4433-4438.[Abstract]
    44. 44
  44. Swofford, D. L. 1998. PAUP*. Phylogenetic analysis using parsimony (*and other methods). 4.0 ed. Sinauer Associates, Sunderland, MA.
    45. 45
  45. Szpilewska, H., A. Czyz, and G. Wegrzyn. 2003. Experimental evidence for the physiological role of bacterial luciferase in the protection of cells against oxidative stress. Curr. Microbiol. 47:379-382.[Medline]
    46. 46
  46. Tamplin, M. L., A. L. Gauzens, A. Huq, D. A. Sack, and R. R. Colwell. 1990. Attachment of Vibrio cholerae serogroup O1 to zooplankton and phytoplankton of Bangladesh waters. Appl. Environ. Microbiol. 56:1977-1980.[Abstract/Free Full Text]
    47. 47
  47. ter Braak, C. J. F., and C. W. N. Looman. 1986. Weighted averaging, logistic regression and the Gaussian response model. Vegetatio 65:3-11.[CrossRef]
    48. 48
  48. ter Braak, C. J. F., and P. Smilauer. 2002. CANOCO reference manual and CanoDraw for Windows user's guide: software for canonical community ordination (version 4.5). Microcomputer Power, Ithaca, NY.
    49. 49
  49. Tison, D. L. 1999. Vibrio, p. 497-504. In P. R. Murray, E. J. Baron, M. A. Pfaller, F. C. Tenover, and R. H. Yolken (ed.), Manual of clinical microbiology, 7th ed. American Society for Microbiology, Washington, DC.
    50. 50
  50. Urbanczyk, H., J. C. Ast, A. J. Kaeding, J. D. Oliver, and P. V. Dunlap. 2008. Phylogenetic analysis of the incidence of lux gene horizontal transfer in Vibrionaceae. J. Bacteriol. 190:3494-3504.[Abstract/Free Full Text]
    51. 51
  51. Wegrzyn, G., and A. Czyz. 2002. How do marine bacteria produce light, why are they luminescent, and can we employ bacterial bioluminescence in aquatic biotechnology? Oceanologia 44:291-305.
    52. 52
  52. Weiss, R. F. 1974. Carbon dioxide in water and seawater: the solubility of a non-ideal gas. Mar. Chem. 2:203-215.[CrossRef]
    53. 53
  53. West, P. A., P. R. Brayton, T. N. Bryant, and R. R. Colwell. 1986. Numerical taxonomy of Vibrios isolated from aquatic environment. Int. J. Syst. Bacteriol. 36:531-543.[Abstract/Free Full Text]
    54. 54
  54. West, P. A., and J. V. Lee. 1982. Ecology of Vibrio spp. including V. cholerae in natural waters of Kent, England. J. Appl. Bacteriol. 52:435-448.[Medline]
    55. 55
  55. Zhu, J., M. B. Miller, R. E. Vance, M. Dziejman, B. L. Bassler, and J. J. Mekalanos. 2002. Quorum-sensing regulators control virulence gene expression in Vibrio cholerae. Proc. Natl. Acad. Sci. USA 99:3129-3134.[Abstract/Free Full Text]
    56. 56
  56. Zo, Y.-G., N. Chokesajjawatee, E. Arakawa, H. Watanabe, A. Huq, and R. R. Colwell. 2008. Covariability of Vibrio cholerae microdiversity and environmental parameters. Appl. Environ. Microbiol. 74:2915-2920.[Abstract/Free Full Text]
    57. 57
  57. Zo, Y. G., I. N. G. Rivera, E. Russek-Cohen, M. S. Islam, A. K. Siddique, M. Yunus, R. B. Sack, A. Huq, and R. R. Colwell. 2002. Genomic profiles of clinical and environmental isolates of Vibrio cholerae O1 in cholera-endemic areas of Bangladesh. Proc. Natl. Acad. Sci. USA 99:12409-12414.[Abstract/Free Full Text]

Applied and Environmental Microbiology, January 2009, p. 135-146, Vol. 75, No. 1
0099-2240/09/$08.00+0     doi:10.1128/AEM.02894-07
Copyright © 2009, American Society for Microbiology. All Rights Reserved.




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Zo, Y.-G.
Right arrow Articles by Colwell, R. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Zo, Y.-G.
Right arrow Articles by Colwell, R. R.
Agricola
Right arrow Articles by Zo, Y.-G.
Right arrow Articles by Colwell, R. R.

Copyright © 2009 by the American Society for Microbiology.   For an alternate route to Journals.ASM.org, visit: http://intl-journals.asm.org | More Info»