Genetic Diversity of the Biofilm Covering Montacuta ferruginosa (Mollusca, Bivalvia) as Evaluated by Denaturing Gradient Gel Electrophoresis Analysis and Cloning of PCR-Amplified Gene Fragments Coding for 16S rRNA

Abstract

The shell of the bivalve Montacuta ferruginosa, a symbiont living in the burrow of an echinoid, is covered with a rust-colored biofilm. This biofilm includes different morphotypes of bacteria that are encrusted with a mineral rich in ferric ion and phosphate. The aim of this research was to determine the genetic diversity and phylogenetic affiliation of the biofilm bacteria. Also, the possible roles of the microorganisms in the processes of mineral deposition within the biofilm, as well as their impact on the biology of the bivalve, were assessed by phenotypic inference. The genetic diversity was determined by denaturing gradient gel electrophoresis (DGGE) analysis of short (193-bp) 16S ribosomal DNA PCR products obtained with primers specific for the domain Bacteria. This analysis revealed a diverse consortium; 11 to 25 sequence types were detected depending on the method of DNA extraction used. Individual biofilms analyzed by using the same DNA extraction protocol did not produce identical DGGE profiles. However, different biofilms shared common bands, suggesting that similar bacteria can be found in different biofilms. The phylogenetic affiliations of the sequence types were determined by cloning and sequencing the 16S rRNA genes. Close relatives of the genera Pseudoalteromonas, Colwellia, and Oceanospirillum (members of the γ-Proteobacteria lineage), as well as Flexibacter maritimus (a member of the Cytophaga-Flavobacter-Bacteroides lineage), were found in the biofilms. We inferred from the results that some of the biofilm bacteria could play a role in the mineral formation processes.

Bacteria are widespread in the sea and can colonize virtually any submerged substrate, whatever its nature (inert or living) (3, 5). In this context, a biofilm may develop and form a microbial community in which several types of microorganisms coexist and interact (5). Body surfaces of marine organisms are often colonized by bacteria which are termed epibionts and are considered symbionts sensu Kinne (this general term includes parasitism, commensalism, mutualism, and phoresis) (38). The epibiotic coating can be complex when several types of microorganisms share the same surface (33); moreover, the composition of the community may change with time. Observations of epibiotic microbial communities have raised questions about biotic interactions between the microorganisms and the living substrate (i.e., the host organism) and between the microorganisms themselves. Understanding these interactions is an important step in explaining the functioning of a microbial community on a particular substrate. Basically, to do this, the coexisting microorganisms must be identified and their metabolic abilities must be determined.

In studies of bacterial symbiosis and of microbial communities in general, microbiologists have long been constrained by the use of traditional methods (cultivation and/or microscopy). These traditional methods are essential for understanding the biology of microbial communities, but it is known that they have limited usefulness for revealing in situ microbial diversity (47). Consequently, the occurrence of epibiotic bacteria on marine organisms and the composition, structure, specificity, and stability of the epibiotic bacterial communities known today remain for the most part unexplored. Genetic characterization of microbial diversity through amplification and sequence analysis of the 16S rRNA genes (rDNAs) has been successful in a wide range of environments (4, 8, 9, 21, 29, 40, 44, 71). To date, only a few molecular studies of marine epibiotic microbial communities have been performed; some of the hosts that have been studied are the nematode Laxus sp. (55), the hydrothermal vent polychaete Alvinella pompejana (33), the shrimp Rimicaris exoculata (54), the gutless oligochaete Inanidrilus leukodermatus (13), and the seagrass Halophila stipulacea (73).

Montacuta ferruginosa is a small marine bivalve (length, ca. 7 mm) that lives in the burrow of the echinoid Echinocardium cordatum (22). Each burrow generally contains one to seven bivalves, but up to 18 bivalves may share the same burrow (42). A typical rust-colored coating covers the shell of M. ferruginosa and the specific epithet refers to this coating. The coating is a partially mineralized biofilm (ferric minerals rich in phosphate) made up of three superimposed layers: (i) a superficial layer that is essentially microbial and partially iron encrusted, (ii) a middle layer with microorganisms that are deeply iron encrusted, and (iii) a deep layer that is essentially mineral and apparently lacks microorganisms (27, 28). Although the predominant morphotypes of the biofilm are filamentous bacteria that resemble members of the Beggiatoales (28), none of the bacteria has been identified or cultured.

The aim of the present work was to study the phylogenetic diversity of the M. ferruginosa biofilm bacteria. To obtain a global view of the bacterial diversity of the biofilm, we used the denaturing gradient gel electrophoresis (DGGE) approach (48), and to obtain phylogenetic information, we used the classic cloning-sequencing approach. We restricted our analyses to the domain Bacteria (51) by using specific primers. On the basis of the phylogenetic information obtained, the major metabolic abilities of the epibiotic bacteria were tentatively inferred in order to understand the potential role of these organisms in iron precipitation and their impact on the biology of the bivalve.

MATERIALS AND METHODS

Samples.

Specimens of M. ferruginosa (Montagu, 1803) were collected intertidally from the burrows of E. cordatum (Pennant, 1777) (Echinoidea, Spatangoida) at Wimereux (Pas-de-Calais, France) during September and October 1996. The bivalves that shared the same burrow were noted. In the laboratory, the bivalves were briefly rinsed in sterile seawater (filter sterilized with 0.22-μm-pore-size filters) and measured (anteroposterior axis). The presence of filamentous bacteria (Beggiatoales-like bacteria) was determined for each bivalve with a binocular microscope, and the levels of these bacteria were estimated as follows: 0, no filaments; −, few filaments; and +, abundant filaments. The biofilms were then scraped off each shell with a sterile blade. Then, two sets of samples were prepared; in one set the biofilms were pooled (pooled biofilm samples), and in the other set they were kept individually (individual biofilm samples). Two samples of pooled biofilms (samples E1 and E2) were prepared from a mixture of 60 biofilms; the mixture was divided into two parts that were suspended in 500 μl of TE buffer (10 mM Tris, 1 mM EDTA; pH 8.0). The two samples of pooled biofilms (samples E1 and E2) were subjected to different genomic DNA extraction protocols. The individual biofilm samples were separately suspended in 15 μl of TE buffer. Twenty individual samples were prepared. All 22 samples were then stored at −80°C.

DNA extraction.

Different DNA extraction protocols, including both chemical disruption and physical disruption of bacterial cells, were tested previously (unpublished data). Only physical methods were used in this study because they successfully extracted DNA from very small quantities of material. The following two physical methods were used to extract DNA from the pooled biofilm samples: freeze-thawing for sample E1 and bead beating for sample E2. Genomic DNA of the major biofilm bacteria should have been obtained with these two methods (the bacteria were efficiently lysed by the treatments, particularly by the bead-beating treatment). As the freeze-thawing method was very simple and very efficient (it resulted in more bands during DGGE than the other method), it was used to extract DNA from the individual biofilm samples. The freeze-thawing method included four freeze-thaw cycles (−196 and 80°C). The bead-beating method included one cycle of bead beating with a Mini Beadbeater (Biospec Products). Bead beating was performed for 30 s at 5,000 rpm after 0.5 g of zirconium beads (diameter, 0.1 mm) was added. Lysis of cells was checked by microscopy. After the lysis treatments, 1- or 5-μl portions of the uncentrifuged mixtures were used directly as template DNA for the PCR.

PCR amplification.

The extracted DNA of the two pooled biofilm samples were used in two different PCR (62). For cloning, the nearly complete 16S rDNA was amplified with primers 27F and 1492R, which are specific for the domain Bacteria. For DGGE analysis, a 193-bp rDNA fragment was amplified with primers GM5F-GC-clamp and 518R. The 193-bp fragment spans the V3 region of the 16S rRNA (49). The GC clamp was a 40-nucleotide GC-rich sequence added so that the melting behavior of the DNA fragments during the DGGE analysis would be stable (48). The sequences of the primers are shown in Table 1.

TABLE 1.

Primers used in this study

Primer	Positiona	Sequence	Reference
27F	8-27	5′-AGAGTTTGATCATGGCTCAG-3′	39
1492R	1492-1509	5′-GGTACCTTGTTACGACTT-3′	39
GM5F	341-357	5′-CCTACGGGAGGCAGCAG-3′	48
518R	518-534	5′-ATTACCGCGGCTGCTGG-3′	48
GC-Clampb		5′-CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCACGGGGGG-3′	48
Sequencing primerc	1053-1070	5′-AGCTGACGACAGCCATGC-3′	19

^aThe numbering of the positions was the numbering used for the 16S rRNA of E. coli. 

^bThe GC-Clamp was attached to the 5′ end of the GM5F primer. 

^cTexas Red was attached to the 5′ end of the sequencing primer. 

The PCR amplification procedure used for cloning was performed with an OmniGene temperature cycler as follows. Each mixture containing 5 μl of template DNA, each primer at a concentration of 0.5 μM, each deoxynucleoside triphosphate at a concentration of 200 μM, 1.5 mM MgCl₂, 20 ng of bovine serum albumin, 5 μl of 10× PCR buffer (100 mM Tris-HCl [pH 9], 500 mM KCl), and 2.5 U of Taq DNA polymerase (Boehringer, Mannheim, Germany) was adjusted to a final volume of 50 μl with sterile water (Sigma) and overlaid with 3 drops of mineral oil (Sigma). The tubes were incubated for 3 min at 94°C and then subjected to 25 cycles consisting of denaturation at 94°C for 1 min, annealing at 55°C for 1 min, and primer extension at 72°C for 2 min. The tubes were then incubated for 10 min at 72°C.

The PCR amplification procedure used for the DGGE analysis was performed with a Perkin-Elmer model 480 thermal cycler. The concentrations of chemicals and the quantity of template DNA were the same as those described above. The tubes were first incubated for 5 min at 94°C. A touchdown PCR (12) was then performed by using 20 cycles consisting of denaturation at 94°C for 1 min, annealing at 65°C (the temperature was decreased by 1°C every second cycle until the touchdown temperature of 56°C was reached) for 1 min, and primer extension at 72°C for 1 min. Five additional cycles were carried out at an annealing temperature of 55°C. The tubes were then incubated for 10 min at 72°C.

Aliquots (4 μl) of the amplification products were analyzed first by electrophoresis in 1% (wt/vol) agarose gels and then by ethidium bromide staining. DNA isolation, preparation of PCR mixtures, and pipetting of template DNA into PCR tubes were performed in UV-treated hoods in separate rooms in order to avoid contamination of the PCR reagents. Different pipettes fitted with autoclaved filter tips were used in each hood.

DGGE analysis.

The PCR products obtained with primers GM5F-GC-clamp and 518R were analyzed by DGGE. DGGE was performed with a Bio-Rad Protean II system or a Ingeny system as described previously (47, 48). PCR samples were applied directly onto 8% (wt/vol) polyacrylamide gels in 0.5× TAE (20 mM Tris-acetate [pH 7.4], 10 mM acetate, 0.5 mM disodium EDTA). The denaturing gradients contained 30 to 70% denaturant (100% denaturant corresponded to 7 M urea and 40% [vol/vol] formamide). The gels were prepared with a Minipuls-2 pump (Gilson) and a gradient former. Electrophoresis was performed for 16 h at 75 V and 25 mA (Bio-Rad Protean II system) or at 100 V and 25 mA (Ingeny system). The temperature was set at 60°C. After electrophoresis, the gels were incubated for 30 min in water containing 0.5 mg of ethidium bromide per liter and photographed on a Mini-Transilluminator UV table equipped with a digital camera (Imager 2.03 system; Appligene Inc.).

After the DGGE analysis of the individual biofilm samples, the individual DGGE profiles were compared to each other by using the pairwise similarity coefficient C_s, which was determined as follows: C_s = 2j/(a + b) × 100, where a was the number of DGGE bands in biofilm 1, b was the number of DGGE bands in biofilm 2, and j was the number of common DGGE bands (46, 50). Two identical DGGE profiles had a C_s value of 100%, and two completely different profiles had a C_s value of 0%.

Cloning.

PCR products obtained with primers 27F and 1492R were purified with a Wizard DNA Clean-Up system (Promega) and cloned by the TA cloning method in pCR2.1 vectors (DNA from freeze-thawed biofilms) and pGemT vectors (DNA from bead-beaten biofilms) with TA cloning kits (Invitrogen and Promega, respectively). Ligation with T4 DNA ligase and transformation with One Shot competent cells were performed by using the protocols recommended by the manufacturer. Recombinants which were white when they were plated onto X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside)-IPTG (isopropyl-β-d-thiogalactopyranoside) indicator plates (20 μg of X-Gal per liter, 5 μg of IPTG per liter) were picked and grown overnight in liquid medium (Luria-Bertani broth). Luria-Bertani broth and indicator plates also contained ampicillin (20 μg/liter) and methicillin (80 μg/liter). For the PCR, 1 μl of each clone culture was used directly as a DNA template with primers GM5F-GC-clamp and 518R. The PCR conditions were the same as those described above for the DGGE analysis. The PCR products were used in a DGGE as described above. Clones that produced a DGGE band at the same position as a band in the environmental profiles were partially sequenced (the environmental profiles were the DGGE profiles of the pooled biofilm samples). For sequencing, plasmids were isolated and purified from each clone culture with a High Pure plasmid isolation kit (Boehringer).

Sequencing methods.

Cycle sequence reactions were performed with the plasmid-cloned material by using Thermosequenase (Amersham, Little Chalfont, United Kingdom) according to the manufacturer’s instructions. Fragment separation, detection, and base calling were performed with a Vistra model 725 automatic sequencer (Amersham). The sequences were determined in one direction with primer 1053-Tex, which is specific for the domain Bacteria (19), labelled with Texas Red. The sequence of the sequencing primer is shown in Table 1.

Sequence analysis.

The sequences obtained were compared to known sequences by using Mail-FASTA program, version GCG73 (53), and the SIMILARITY_RANK tool of the Ribosomal Database Project (RDP) (41), last updated on 17 May 1995. The sequences were checked for chimeric molecules by using the CHECK_CHIMERA tool of the RDP. Nucleotide sequences of close evolutionary relatives of our sequences were retrieved from the National Center for Biotechnology Information World Wide Web ENTREZ browser that maintains and distributes the GenBank sequence database. Sequences were aligned manually with the sequence alignment editor SeqApp (Macintosh version 1.9a) (26). Phylogenetic trees were constructed with programs of the PHYLIP program package (Macintosh version 3.5c) (18). For the distance matrix method we used DNADIST with the Jukes-Cantor model; phylogenetic trees were then constructed from evolutionary distances by the neighbor-joining method implemented through the program NEIGHBOR. For the parsimony and maximum-likelihood methods we used the programs DNAPARS and DNAML (under the default settings) implemented in the PHYLIP software package. A total of 100 bootstrapped replicate resampling data sets were generated with SEQBOOT. The bootstrap values indicate the resampling percentages which supported a specific branching pattern. The consensus tree was determined with CONSENSE. The similarity values given in the test (S) are noncorrected distances that were calculated from corrected-distance DNADIST matrices by reversing the Jukes-Cantor corrected distance formulae (35).

The 16S rRNA sequences of the following organisms were used in this study (the numbers in parentheses are GenBank nucleotide sequence accession numbers): Pseudoalteromonas atlantica (X82134), Pseudoalteromonas aurantia (X82135), Pseudoalteromonas carrageenovora (X82136), Pseudoalteromonas citrea (X82137), Pseudoalteromonas denitrificans (X82138), Pseudoalteromonas espejiana (X82143), Pseudoalteromonas haloplanktis subsp. haloplanktis (X67024), P. haloplanktis subsp. tetraodonis (X82139), Pseudoalteromonas luteoviolacea (X82144), Pseudoalteromonas nigricifaciens (X82146), Pseudoalteromonas piscida (X82141), Pseudoalteromonas rubra (X82147), Pseudoalteromonas undina (X82140), Beggiatoa alba (L40994), clone PVB_OTU_4 (U15116), Colwellia psychroerythrus (L10939), isolate SCB11 (Z31658), marine clone agg53 (L10950), Oceanospirillum linum (M22365), Pseudomonas aeruginosa (M34133), Sphaerotilus natans (Z18534), strain S51-W(gv)1 (U14581), Teredinibacter turnerae (M64339), Thiothrix nivea (L40993), Thiothrix ramosa (U32940), “Antarcticum vesiculatum” (M61002), Cytophaga flevensis (M58767), Cytophaga lytica (M62796), Flavobacterium odoratum (M58777), Flectobacillus glomeratus (M58775), Flexibacter canadensis (M62793), Flexibacter maritimus N⁰449, NCIMB 2154 (D14023), F. maritimus ATCC 43398 (M64629), F. maritimus JCM 8137 (D12667), and isolate 301 (U14586).

Nucleotide sequence accession numbers.

The sequences obtained in this study have been assigned in the GenBank database under accession no. AF017792 to AF017805.

RESULTS

DGGE analysis of the pooled biofilm DNA samples.

The two sets of equal-size 16S rDNA PCR products, samples E1 and E2, produced two different DGGE profiles (Fig. 1). The total numbers of bands detected in the profiles were 25 for sample E1 and 11 for sample E2. Replicates produced the same numbers of bands (data not shown). The profiles contained intense DNA bands, as well as faint DNA bands. Ten bands, including three intense bands, were found in both profiles.

FIG. 1.

(a) Negative image of ethidium bromide-stained DGGE gels. Lanes E1 and E2, DGGE profiles of pooled biofilm samples E1 (freeze-thaw extraction) and E2 (bead-beating extraction), respectively; lanes 1 to 20, DGGE profiles of 20 individual biofilm samples. (b) Graphic representation of the DGGE profiles shown in panel a. Column C indicates the clones obtained from the three screenings. Columns E1 and E2 indicate the bands produced by pooled biofilm samples E1 and E2, respectively (black DGGE bands represent intense bands). The Filaments line indicates the presence of Beggiatoales-like filamentous bacteria (0, no filaments; −, few filaments; +, abundant filaments). Column H, indicates the number of individual biofilm samples which produced each band. Column N indicates DGGE position. The Size (mm) line indicates the lengths (anteroposterior axis) of the bivalves (in millimeters). Line V indicates the number of bands produced by each sample or (for column C) the number of clones. Columns 1 to 20 show the DGGE profiles of the 20 individual biofilm samples.

DGGE analysis of the individual biofilm DNA samples.

Twenty individual biofilms were analyzed by DGGE. Each biofilm produced a different DGGE pattern containing intense and faint bands (Fig. 1). Each pattern was reproducible (data not shown). The number of individual bands varied from 9 to 21. Figure 1b shows that the DGGE profiles were independent of the size of the bivalve and that no particular profile occurred when Beggiatoales-like bacteria were present. Two biofilm samples (biofilm samples 7 and 15) produced no DGGE profiles; this may have been due to the poor development of these biofilms. When the 18 other DGGE profiles were examined, a total of 40 different bands were found. One band (band N27) was present in all of the biofilm samples examined; seven bands (bands N12, N14, N19, N24, N27, N33, and N34) were present in at least 15 biofilms; two bands (bands N9 and N22) were present in more than 10 but less than 15 biofilms; and eight bands (bands N5, N6, N8, N10, N15, N18, N37, and N40) were present in more than 5 but less than 10 biofilms. The most frequent band in the individual biofilm DGGE patterns (band N27) was probably an artifact due to fortuitous contamination by Escherichia coli DNA (band N27 migrated to the same position on DGGE gels as the band originating from E. coli 16S rRNA, and a faint band was also found at this position in the negative controls). As PCR amplifications were performed with the greatest care (see above), this contamination may have originated from the E. coli used during industrial production of the Taq enzyme.

The 18 DGGE profiles were compared to one another by using the similarity coefficient C_s (Table 2) without taking band N27 into account. The C_s values ranged from 32 to 84%, and the mean C_s value was 54%. The highest C_s values (C_s values greater than 75%) were always obtained for biofilm profiles originating from bivalves living in the same burrow; for example, the C_s value for biofilms 1 and 2 was 78%, the C_s value for biofilms 10 and 11 was 80%, and the C_s value for biofilms 16 and 17 was 84%. The data in Table 2 show that the mean intraburrow C_s values (calculated with the bold-faced C_s values in Table 2) were not significantly higher than the mean interburrow C_s values (calculated without the bold-faced C_s values in Table 2) (except for burrows B1 and B6, which contained only two bivalves).

TABLE 2.

C_s matrix for individual biofilm samples^a

Biofilm	Cs (%)
	Burrow 1		Burrow 2				Burrow 3	Burrow 4	Burrow 5					Burrow 6		Burrow 7
	Biofilm 1	Biofilm 2	Biofilm 3	Biofilm 4	Biofilm 5	Biofilm 6	Biofilm 8	Biofilm 9	Biofilm 10	Biofilm 11	Biofilm 12	Biofilm 13	Biofilm 14	Biofilm 16	Biofilm 17	Biofilm 18	Biofilm 19
2	78
3	64	62
4	45	61	50
5	72	69	65	64
6	48	45	44	67	67
8	52	50	48	61	54	45
9	46	52	56	69	62	56	67
10	40	38	52	56	50	33	54	62
11	52	43	61	52	60	38	50	58	80
12	58	56	60	58	59	52	64	57	52	55
13	64	70	71	64	64	38	52	54	48	52	75
14	55	43	57	64	64	57	70	62	64	74	75	55
16	48	45	44	48	50	20	36	48	50	46	35	57	38
17	60	57	46	60	61	32	48	58	52	56	45	60	50	84
18	44	57	42	52	47	38	36	39	53	44	41	44	30	46	48
19	65	63	59	58	71	53	44	57	47	61	61	58	52	47	55	72
20	42	60	32	63	45	44	50	61	45	42	48	42	42	33	47	67	57

^aBiofilm samples 7 and 15 were omitted from the analysis because they produced no DGGE profiles. 

^bThe mean intraburrow C_s values (calculated with the bold-faced values) were as follows: burrow 1, 78% (n = 1); burrow 2, 60% ± 16% (n = 6); burrow 5, 63% ± 17% (n = 10); burrow 6, 84% (n = 1); and burrow 7, 65% ± 8% (n = 3). The mean interburrow C_s values were as follows: burrow 1, 54% ± 18% (n = 32); burrow 2, 58% ± 26% (n = 56); burrow 5, 53% ± 20% (n = 65); burrow 6, 48% ± 16% (n = 32); and burrow 7, 49% ± 22% (n = 45). 

Cloning and sequencing.

Three DNA libraries were obtained; these libraries were designated libraries I, II, and III (the freeze-thaw DNA extraction method was used for libraries I and II, and the bead-beating DNA extraction method was used for library III). Twenty clones were picked from library I (S1 clones), 25 clones were picked from library II (S2 clones), and 17 clones were picked from library III (S3 clones). The S1 clones produced 9 different bands on DGGE gels, the S2 clones produced 8 different bands, and the S3 clones produced 10 different bands. The number of clones obtained at each DGGE position and the clones that were sequenced are indicated in Fig. 2. No S2 clones were sequenced (the bands were represented in the S1 clones or were not abundant). Only seven of the eight S1 clones sequenced produced a readable sequence (in which base peaks were well resolved) containing ca. 600 nucleotides; the S1C36 sequence was not readable. The nine S3 clones that were sequenced produced nine readable sequences containing ca. 500 nucleotides.

FIG. 2.

Graphic representation of the DGGE profiles obtained from the three screenings. Column N indicates DGGE position. Columns S1, S2, and S3 indicate the bands obtained from the first (S1), second (S2), and third (S3) screenings. The numbers between the columns indicate the numbers of clones obtained for each DGGE position.

Sequence analysis.

The 16 readable sequences obtained were submitted to the CHECK_CHIMERA service of the RDP. Two sequences (sequences S1C9 and S1C29) exhibited classic chimeric behavior; the CHECK_CHIMERA histogram values consistently rose and fell (the histogram for S1C29 was somewhat irregular), and the maximum oligo-gain values were high (about 40). Moreover, the S_ab values for the full-length sequences were lower than the S_ab values for the fragments. In addition, S1C9 was not found in the individual DGGE patterns (Fig. 1b). Two other sequences (sequences S1C26 and S1C35) were more problematic because they did not exhibit typical chimeric behavior; the histograms rose slowly and irregularly toward moderate oligo-gain values (about 25 and 19, respectively), and only S1C26 had a full-length sequence S_ab value lower than the S_ab values of the fragments. Many biofilms produced bands corresponding to sequences S1C26 and S1C35 in their DGGE patterns (Fig. 1b). These sequences could represent a novel lineage of the domain Bacteria because they possess all of the signature nucleotides and do not group with any of the lineages of that domain (data not shown). Before definite conclusions are made, clones S1C26 and S1C35 should be sequenced completely.

The 12 sequences for which there was no indication of chimerism were aligned with the sequences of close relatives and used to construct phylogenetic trees. The 16S rRNA region at E. coli base positions 508 to 991 (length, 484 nucleotides) was utilized in the analysis. This region spans the V5 hypervariable region of the 16S rRNA molecule (49). None of the sequences was identical to any of the known sequences obtained from the databases examined. The sequences fell into two major lineages of the domain Bacteria (51): the γ-Proteobacteria lineage (nine sequences) and the Cytophaga-Flavobacter-Bacteroides lineage (three sequences). Within the γ-Proteobacteria lineage, none of our sequences grouped with the Beggiatoales. All groups were identified by the three methods of phylogenetic analysis used and were supported by high bootstrap replication values. The phylogenetic trees obtained from the distance matrix analysis are shown in Fig. 3 to 5.

FIG. 3.

Phylogenetic tree showing the relationship between the Pseudoalteromonas-related 16S rDNA clones and Pseudoalteromonas species. All of the bacteria are members of the γ Proteobacteria. P. denitrificans served as the outgroup. The tree was generated by the distance matrix method. The numbers are the bootstrap values for the nodes, based on a total of 100 replicate resamplings (values less than 50 are not shown). The horizontal dotted lines do not represent phylogenetic distances. The nonpigmented Pseudoalteromonas species are surrounded with a dashed square. Bar = 1% nucleotide change. P., Pseudoalteromonas; P. hal. haloplanktis, P. haloplanktis subsp. haloplanktis; P. hal. tetraodonis, P. haloplanktis subsp. tetraodonis.

FIG. 5.

Phylogenetic tree showing the relationship between the 16S rDNA clones related to the genus Flexibacter and related bacteria. All of the organisms are members of the Cytophaga-Flavobacter-Bacteroides phylum. F. canadensis served as the outgroup. The tree was generated by the distance matrix method. The numbers are bootstrap values for the nodes, based on a total of 100 replicate resamplings (values less than 50 are not shown). Bar = 5% nucleotide change. A., Antarcticum; Cy., Cytophaga; F., Flavobacterium; Flc., Flectobacillus; Flx., Flexibacter.

γ-Proteobacteria-related sequences.

Sequences S3C7, S3C8, S3C13, and S3C14 were members of the γ3 subgroup. These sequences grouped with the robust monophyletic taxon which includes the 12 Pseudoalteromonas (formerly Alteromonas) species (23) (Fig. 3). More precisely, these four sequences grouped with the nonpigmented Pseudoalteromonas species. The 16S rRNA region examined here was shorter; a total of 390 positions (corresponding to E. coli positions 608 to 991), including both conserved and hypervariable domains, were included in the analysis. The four Pseudoalteromonas-related sequences differed at six positions (mutations) and by 10 deletion-insertion events in the region examined, so the existence of this set of related sequences is not artifactual (the highest error rate reported for Taq was about one mistake per 364 nucleotides [14]). Moreover, the E. coli clones corresponding to sequences S3C7, S3C8, S3C13, and S3C14 produced different bands on DGGE gels; this indicates that the sequences did not result from a sequencing artifact. The four sequences were also represented by different bands in environmental DGGE profiles, so the sequence differences were probably not the result of mutations during amplification of the plasmids by cloning.

Sequences S1C12, S1C20, and S3C15 were related to the Colwellia cluster. This group includes the psychrophilic marine bacterium C. psychroerythrus (10), gas vacuolate Antarctic strain S51-W(gv)1 (31), and clone AGG 53 (9) (Fig. 4). Sequence S1C18 was related to the bacteria O. linum and T. turnerae (11) (Fig. 4). S3C16 was related to the hydrothermal vent bacterium clone PVB_OTU_4 (44) (Fig. 4).

FIG. 4.

Phylogenetic tree showing the relationship between the 16S rDNA clones related to C. psychroerythrus, O. linum, and PVB_OTU_4 and related bacteria. All of the organisms are members of the γ Proteobacteria. S. natans served as the outgroup. The tree was generated by the distance matrix method. The numbers are bootstrap values for the nodes, based on a total of 100 replicate resamplings (values less than 50 are not shown). Bar = 5% nucleotide change. B., Beggiatoa; C., Colwellia; O., Oceanospirillum; P., Pseudomonas; T., Teredinibacter; Thx., Thiothrix; S., Sphaerotilus.

Cytophaga-Flavobacter-Bacteroides-related sequences.

Se- quences S3C18, S3C19, and S3C20 grouped together in the F. maritimus cluster of the Cytophaga subgroup (24) and, more precisely, with three F. maritimus strains, NCIMB 2154, JCM 8137, and ATCC 43398 (Fig. 5) (in the region examined, ATCC 43398 and NCIMB 2154 differed by only one nucleotide, and ATCC 43398 and JCM 8137 differed by four nucleotides). F. maritimus is the same organism as Cytophaga marina, but the name F. maritimus has priority (34). Like the Pseudoalteromonas-related sequences, the three F. maritimus-related sequences also represented a set of phylogenetically related organisms.

Except for DGGE band N27, which is an E. coli band, it appears from Fig. 1b that the most abundant DGGE bands in the individual biofilm patterns corresponded to Pseudoalteromonas-related sequence S3C7 (found in 17 of 18 bivalves), Colwellia-related sequence S3C15 (15 bivalves), and F. maritimus-related sequence S3C20 (15 bivalves). The DGGE bands obtained with the Oceanospirillum strain and PVB_OTU_4 appeared less frequently (seven bivalves and one bivalve, respectively). Three other bands were observed frequently in the patterns; two were not cloned (DGGE bands N33 and N34) (15 bivalves), and the third was the S1C35 band (17 bivalves), a sequence that could represent a novel lineage of the domain Bacteria. There were no clear relationships between the presence of Beggiatoales-like bacteria in the biofilm and a particular clone.

Sequence signature analysis.

The 12 sequences were checked for the presence of signature nucleotides in the 16S rDNA region examined. A total of 23 positions were examined for the γ-Proteobacteria lineage (74), and 21 positions were examined for the Cytophaga-Flavobacter-Bacteroides lineage (24). The signatures of our sequences were consistent with the phylogenetic placement of the sequences. There was only one difference between the S1C20, S1C12, S3C15, and S1C18 sequences and the γ-Proteobacteria consensus sequence (G instead of A at position 640). In this case, a compensatory base change always occurred at position 598 (positions 640 and 598 formed a base pair in the 16S rRNA secondary structure). Two positions differed in all of the Pseudoalteromonas and Pseudoalteromonas-related sequences (G instead of A at position 640; T instead of G at position 760). The following three positions differed in S3C16: position 554 (T instead of A), position 690 (A instead of G), and position 812 (C instead of G) (the same differences occurred in the close evolutionary relative PVB_OTU_4). The three Cytophaga-Flavobacter-Bacteroides sequences contained all of the sequence signatures that characterized the F. maritimus cluster of the Cytophaga subgroup (24).

DISCUSSION

DGGE analysis.

We concluded from our DGGE analyses that the biofilms studied were not monospecific; the individual biofilms on average produced 13 DGGE bands. Muyzer and de Waal found that the number of DGGE bands in a pattern is related to the number of bacterial species in the community (47). Although heteroduplex formation (20) and possible divergence within multicopy rRNA gene families (67) might increase the number of DGGE bands, many other factors should reduce it, including sampling biases, differential DNA extraction (61), PCR biases (59), and comigration on DGGE gels (46). To summarize, we view the biofilms studied as complex microecosystems containing at least 13 microorganisms belonging to the domain Bacteria. The influence of the DNA extraction protocol on the number of DGGE bands was obvious in this study; the freeze-thaw extraction method (sample E1) was more efficient than the bead-beating extraction method (sample E2). The complexity of the M. ferruginosa biofilm is not surprising because it was suggested previously by microscopic observations (28). In addition, many other studies of microbial communities have revealed complex consortia (4, 21, 32, 60). The fact that there was no apparent relationship between the size (age) of a bivalve and the number of bands in the DGGE pattern of its biofilm indicates that the complexity of the biofilm cannot be easily related to time. This can be explained by the biology of the bivalve; the shell, whatever its size, acts as a colonizing area for the microorganisms present in the immediate environment (i.e., microorganisms from the sediments, from sea urchin feces, and from other bivalves sharing the same burrow). The immediate environments are not identical for all of the bivalves living in the same area. Moreover, as bivalves are active burrowers, their biofilms are frequently abraded; this, along with the heterogeneity of the medium, results in differences among biofilms. On the other hand, M. ferruginosa is gregarious, and juveniles often attach to adults. This behavior may result in progressive homogenization of the biofilms in a burrow (although the error of the analysis was too great to be conclusive, the highest C_s values were obtained for biofilms originating from the same burrow). These phenomena (abrasion and homogenization) should be independent of the age of the bivalves.

It appears from our DGGE analyses that each biofilm is unique, because all of the individual DGGE profiles were different. However, some DGGE bands (bands N9, N12, N14, N19, N24, N33, and N34) were found in most of the DGGE profiles. These bands could represent common biofilm bacteria. The other bands may represent atypical biofilm bacteria. The presence of atypical bacteria in a biofilm is not surprising if the heterogeneity and high bacterial diversity of coastal marine sediments are considered (32). Another possibility is that the atypical DGGE bands represent common but less abundant 16S rRNA sequences and that these sequences are more influenced by PCR biases than the common and abundant sequences (for example, positively charged iron colloids, which are probably present in DNA samples, could inhibit amplification of the less abundant sequences by adsorption).

It should be noted that an intense DGGE band does not necessarily mean that the corresponding bacteria are abundant in the biofilm because band intensities are also influenced by 16S rRNA gene copy numbers (17), by differential DNA extraction (61), by PCR biases (59, 68), by comigration of two or more sequence types (46), or by a combination of these events. Therefore, the in vivo abundance of a sequence type must be confirmed by other methods, such as in situ hybridization.

Cloning and sequence analysis.

The following two sets of related sequences were identified in this study: four Pseudoalteromonas-related sequences and three F. maritimus-related sequences. Sets of phylogenetically related bacterial populations that coexist in relatively well-circumscribed microecosystems have also been observed in other environments (2, 19, 21, 29, 40, 63). The significance of this apparently frequent phenomenon is uncertain; it could be the result of microevolutionary divergence within a multicopy rRNA gene family or the result of microgeographical adaptation as a result of selection (29, 43, 67).

Since it is thought that major metabolic traits are shared by (phylogenetically) closely related species (52, 69, 74), we can tentatively infer the major metabolic traits of our unknown organisms by looking at their close evolutionary relatives. Such a phenotypic inference method reveals the potential metabolic traits of unknown organisms without the necessity of cultivation. If confirmation is needed, suggested traits may direct the design of subsequent research.

Although there are no rules for relating taxonomic units to 16S rRNA similarities, the high levels of similarity (S = 98.2 to 99.8%) between four of our sequences and the sequences of the nonpigmented Pseudoalteromonas species suggest that our sequences represent new species or subspecies belonging to the genus Pseudoalteromonas. Pseudoalteromonas species are widely distributed and frequently isolated from marine environments (1); these bacteria are strictly aerobic chemoorganotrophs (23). Interestingly, Pseudoalteromonas strains are known to produce siderophores with an exceptionally high affinity constant for ferric ion (57, 58). As suggested by Dahanayake and Krumbein (6), siderophore-producing bacteria could produce various iron minerals, depending on the microenvironmental redox conditions. Also interesting are the high levels of similarity (S = 99.5 to 99.8%) between our sequences and the sequence of P. haloplanktis subsp. tetraodonis. This organism, like a number of other marine bacteria, produces tetrodotoxin, a strong neurotoxin that is the cause of pufferfish poisoning. Such bacteria, which are frequently encountered in seawater and are associated with marine animals (64, 65), could protect the microbial community from biofilm grazers.

The close relatives of sequences S1C12 and S1C20 (Colwellia group) and sequence S1C18 (Teredinibacter-Oceanospirillum group), as well as sequences S3C18, S3C19, and S3C20 (F. maritimus group), are aerobic chemoorganotrophs that were found to be symbiotic; C. psychroerythrus was isolated from the surface of flounder eggs (7), T. turnerae was isolated from the gland of Deshayes in shipworms (Bivalvia, Teredenidae) (11, 72), and F. maritimus is epibiotic on marine fishes (70). The similarity values ranged from 93.4 to 99.0%. O. linum is not known to be symbiotic; it was isolated from coastal seawater. A strain related to the genus Oceanospirillum has been reported to be able to oxidize manganese (15). If this metal-oxidizing ability is present, the organism represented by S1C18 could contribute to iron deposition within biofilms (25).

S3C16 grouped with the PVB_OTU_4 sequence (S = 93.2%). The latter sequence was retrieved from the bacterial mat community of Pele’s Vents, an active deep-sea hydrothermal vent system in Hawaii (44). The exact phylogenetic placement of PVB_OTU_4 is still uncertain; the corresponding bacterium is probably chemolithoautotrophic (44). It is worth noting that bacterial mats collected from this vent system are made up of long filaments coated with amorphous iron precipitates rich in Ca and P (36, 37), like the biofilm covering M. ferruginosa (28). The vent field bacterial mats, which are apparently dominated by Thiovulum-related species and Xanthomonas-related species, also contain bacteria related to the genera Colwellia (sequence PVB 54) and Alteromonas (Pseudoalteromonas) (sequence PVB 18), like the biofilm in the present study. It is not known if there is any relationship between the simultaneous presence of these three sequences (sequences related to PVB_OTU_4, Colwellia, and Pseudoalteromonas) and the presence of an amorphous iron mineral deposit rich in P and Ca. If S3C16 represents a chemolithoautotrophic bacterium, this bacterium could participate in the formation of the iron deposit (by contributing an iron-oxidizing activity). However, the organism represented by sequence S3C16 probably does not occur frequently in the M. ferruginosa biofilm (Fig. 1b).

The epibiotic bacteria may have several effects on the biology of M. ferruginosa. Most of our sequences were related to the sequences of epibiotic chemoorganotrophic marine bacteria. The presence of such bacteria on the bivalve (we suppose that we may infer that this type of metabolism occurs with some confidence), which produces various hydrolytic enzymes, can be problematic; these bacteria could degrade the periostracum of the mollusk (the outer layer of mollusk shells is composed of a scleroprotein called conchiolin), as well as the organic shell matrix (which is rich in glycoproteins, polypeptides, and polysaccharides) (66). Actually, proteolytic and chitinase-producing microorganisms, like C. psychroerythrus and probably the organisms represented by our related sequences, have been found in degrading mollusk shells (56). This could explain why many M. ferruginosa have highly degraded shells under their biofilms (unpublished data). The epibiotic bacteria could also provide benefits to the bivalve by immobilizing toxic iron ions and sulfide (28) and by producing toxic compounds, such as tetrodotoxin. The chemoorganotrophic metabolism could also be related to iron deposition because most of the iron in seawater is probably complexed by organic ligands (30), and microbial degradation of these ligands could lead to subsequent deposition of ferric minerals (16).

In future studies we will determine the in vivo abundance of our sequences with general and specific oligodeoxynucleotide probes. We also plan to study the M. ferruginosa biofilm with cultivation techniques and to test the probes on the cultures. We hope that this integrated approach will lead to a better understanding of the roles of microorganisms in the processes of mineral deposition within the M. ferruginosa biofilm.