SEARCH

SEARCH BY CITATION

Keywords:

  • 16S rRNA gene;
  • biodiversity;
  • biogeography;
  • deep-sea bacteria;
  • Mediterranean microbiota

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The bacterial and archaeal assemblages at two offshore sites located in polar (Greenland Sea; depth: 50 and 2000 m) and Mediterranean (Ionian Sea; depth 50 and 3000 m) waters were studied by PCR amplification and sequencing of the last 450–500 bp of the 16S rRNA gene. A total of 1621 sequences, together with alignable 16S rRNA gene fragments from the Sargasso Sea metagenome database, were analysed to ascertain variations associated with geographical location and depth. The Ionian 50 m sample appeared to be the most diverse and also had remarkable differences in terms of the prokaryotic groups retrieved; surprisingly, however, many similarities were found at the level of large-scale diversity between the Sargasso database fragments and the Greenland 50 m sample. Most sequences with more than 97% sequence similarity, a value often taken as indicative of species delimitation, were only found at a single location/depth; nevertheless, a few examples of cosmopolitan sequences were found in all samples. Depth was also an important factor and, although both deep-water samples had overall similarities, there were important differences that could be due to the warmer waters at depth of the Mediterranean Sea.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The intrinsically low cultivability of marine prokaryotes, coupled with the relative ease of obtaining DNA from seawater samples, has made this environment one of the best studied by the use of molecular markers such as the 16S rRNA gene (Giovannoni et al., 1990; Britschgi & Giovannoni, 1991; Weisburg et al., 1991; Reysenbach et al., 1992; DeLong et al., 1994). More than a decade after the first publications on this subject, our view of the diversity of marine prokaryotes has expanded enormously (Furhman & Davis, 1997; Massana et al., 1997; Marchesi et al., 1998; Acinas et al., 1999; Rappe et al., 2000; Karner et al., 2001; Bano & Hollibaugh, 2002; Bano et al., 2004). The recent shotgun sequencing of 1.045 Gbp of DNA extracted from a superficial sample of the Sargasso Sea (Venter, 2004) has also contributed significantly to the development of this field. However, some important, basic questions remain unanswered. For example, the geographical distribution of marine prokaryotes remains largely unexplored (Fenchel, 2003; Whitaker et al., 2003) as most studies have concentrated on a few locations in coastal regions and the results obtained in different studies have often been difficult to compare. In addition, the majority of these studies have concentrated on superficial waters of the photic zone. The deep, aphotic (depth>200 m) zone of the open ocean is the largest aquatic habitat on Earth and has a critical role in nutrient cycling, which is essential for the ecology of the photic zone. However, the microbiota of this vast area has not been widely explored, although with the data available some trends have been detected. As was expected, the prokaryotic population density decreases significantly with depth (Fuhrman et al., 1992); however combining figures for the whole water column does show that deep ocean contains more prokaryotic cells than any other aquatic habitat (Whitman et al., 1998) and is, per unit of biomass, more productive than terrestrial ecosystems (Schut et al., 1997). Furthermore, the role of this deep microbiota is critical to the functioning of the ecosystem of the planet. Sinking of particulate material acts as a nutrient drain that needs to be replenished by upwelling from the deep ocean, and this essential nutrient turnover is dependent on the deep-sea microbiota (Schut et al., 1997).

Here we present the results of a large-scale comparison of the microbiota present at the most active depth (about 50 m) in the photic zone and at a deep level in the water column, but still far from the seafloor, in two offshore sites with very different oceanographic and climatic characteristics. One site is located in the Greenland Sea (North Atlantic Ocean), with permanently cold waters. The other is located in the Eastern Mediterranean (Ionian Sea), characterized as having an oligotrophic environment but with a slightly higher salinity and higher temperatures (c. 13°C) throughout the water column. In the Ionian Sea the samples were taken at 50 m (I50) and 3000 m (I3K) whereas in the Greenland Sea the samples were from 50 m (G50) and 2000 m (G2K). The comparison involves the study of more than 1500 sequences derived from PCR amplification of the 16S rRNA gene. The results have been also compared with those from other studies carried out around the world's oceans and particularly with the Sargasso Sea metagenome database, ‘Sso’ (Venter, 2004). As expected from previous more restricted studies (Acinas et al., 1999; Lopez-Lopez et al., 2005) the Mediterranean results were remarkably different and underline the uniqueness of this landlocked sea.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Sample collection

Seawater samples were collected in different cruises in the two sampling sites, located at co-ordinates 36°30′N, 15°50′E (Ionian Sea, Eastern Mediterranean) and 72°54′N, 4°45′E and 72°54′N, 4°42′E (Greenland Sea, Arctic Ocean). At both sites the maximum depth is around 3400 m. At each sampling site different volumes of seawater were collected (depending on the depth) by using Niskin bottles (General Oceanics Inc.). One hundred litres was collected from 50 m depth at both locations [in the Ionian coinciding with the deep chlorophyll maximum (DCM)]. From the deep waters of the Mediterranean Sea (3000 m) we collected 240 L of seawater whereas from Greenland Sea (2000 m) 100 L was retrieved. CTD data (conductivity, temperature, depth) from the throughout the water column were registered during collection in both cruises. For the Ionian Sea the data were as follows: 50 m depth: temperature, 15.2201°C; salinity, 38.6525 ppt; 3000 m depth: temperature 13.8519°C; salinity 38.7234 ppt. The corresponding data for the Greenland sample were: 50 m depth: temperature, 3.9007°C; salinity, 35.0204 ppt; 2000 m depth: temperature, −0.8329°C; salinity, 34.9154 ppt.

Nucleic acid extraction

Seawater samples were processed immediately after sampling. The samples were filtered by pressure filtration sequentially through a nylon mesh (retaining c.>200 μm particles) and 5 μm pore size polycarbonate filters. The remaining picoplankton was collected in 0.2 μm pore size Sterivex filters (Durapore, Millipore). Sterivex units were filled with 1.8 mL of lysis buffer (40 mM EDTA, 50 mM Tris/HCl, 0.75 M sucrose) and stored in liquid nitrogen or at −20°C, respectively, for the Greenland and Ionian samples, until DNA extraction. Following a lysozyme and proteinase K–sodium dodecyl sulphate lysis step, nucleic acids were extracted as previously described (Massana et al., 1997). For the Ionian samples, lysates were extracted twice with phenol–chloroform–isoamylalcohol, while in the case of Greenland samples, 5 M NaCl and CTAB was used. In all cases, DNA was washed once with chloroform–isoamylalcohol. For Ionian samples nucleic acids were concentrated with sterile water using a microconcentrator (Centricon 100, Amicon) whereas for Greenland samples the DNA was precipitated using isopropanol and the pellet washed using cold ethanol. DNA integrity was checked using agarose gel electrophoresis and samples were stored at −80°C until use.

Amplification, cloning and sequencing of the 16S rRNA gene and the ITS

Different lengths of environmental 16S rRNA molecules plus the ribosomal 16S–23S internal transcribed spacer (ITS) were amplified with 10 different combinations of primers for Bacteria and four for Archaea (Table 1). Combinations of primers were named as follows: combinations A, B, C, D and E used the forward primers ANT1, B1055, 16S5F, 785F and 338F, respectively, and the primer 23SR1 as reverse. Combinations A′, B′, C′, D′ and E′ used the same forward primers and 23S0R as reverse. In a first step we amplified the bacterial DNA extracted from Ionian samples (I50 and I3K) by using combinations A′ B′ C′, D′ and E′, in which primer 23S0R acted as reverse. The sequences obtained revealed that the use of this reverse primer introduced an important bias given that no sequences from the Alphaproteobacteria, a very widespread group in seawater samples, were retrieved. However we obtained a large amount of sequences related to the Gammaproteobacteria (567) and to Nitrospira (186), whereas other groups were retrieved in very low numbers (Actinobacteria, Deltaproteobacteria, Firmicutes, Bacteroidetes and Fibrobacteria). Therefore, we decided to change the reverse primer for all subsequent work and to use combinations A, B, C, D and E (Table 1), in which primer 23S1R acted as reverse. With this reverse primer we obtained 1440 sequences (361 from I50, 368 from I3K, 402 from G50 and 309 from G2K). The bacterial sequences obtained in the first survey (23S0R as reverse primer) were not included in the comparative analyses of the prokaryotic diversity presented here, although they have been used for some analyses of ITS variability (data not shown) and have been deposited in GenBank under accession numbers AY932826AY934425. For Archaea we amplified different lengths of the 16S rRNA gene plus the ribosomal 16S–23S ITS using three different combinations of primers (Table 1). We obtained 181 archaeal sequences with the oligonucleotides 21F, 344F and 915F as forward primers and an equimolar mixture of Euriclus and Crenarclus as reverse. Of the 181 clones, 108 were retrieved from I3K, 32 from G2K and 41 from G50. Preliminary results obtained with the Ionian samples showed that archaeal diversity was very low and therefore we decided to reduce the number of clones to be analysed in the Greenland samples.

Table 1.   rRNA gene primers used in this study
NameSpecificitySense* and positionSequence (5′–3′)Reference
  • *

    F, forward; R, reverse.

  • Referred to Escherichiacoli (for bacterial primers) or Halobacteriumhalobium (for archaeal primers) 16S or 23S rRNA gene nucleotide position.

  • Combinations of primers were named as follows: combinations A, B, C, D and E used the forward primers ANT1, B1055, 16S5F, 785F and 338F, respectively, and the primer 23SR1 as reverse. Combinations A′, B′, C′, D′ and E′ used the same forward primers and 23S0R as reverse.

ANT1BacteriaF: 8–27 (16S)AGAGTTTGATCATGGCTCAGMartínez-Murcia et al. (1995)
B1055BacteriaF: 1055–1074 (16S)AATGGCTGTCGTCAGCTCGTAmman et al. (1995)
16S5FBacteriaF: 514–533 (16S)CGTGCCAGCAGCCCGCGGTAAGurtler and Stanisich, (1996)
785FBacteriaF: 785–803 (16S)GGATTAGATACCCTGGTAGAmman et al. (1995)
338FBacteriaF: 338–355 (16S)ACTCCTACGGGAGGCAGCAmman et al. (1990)
23S1RBacteriaR: 227–247 (23S)GGGTTTCCCCATTCGGAAATCGarcía-Martínez et al. (1999)
23S0RBacteriaR: 21–38 (23S)TGCCAAGGCATCCACCGTGurtler and Stanisich, (1996)
21FArchaeaF: 7–26 (16S)TTCCGGTTGATCCTGCCGGAGarcía-Martínez et al. (2000)
344FArchaeaF: 344–363 (16S)ACGGGGYGCAGCAGGCGCGARaskin et al. (1994)
915FArchaeaF: 915-934 (16S)GTGCTCCCCCGCCAATTCCStalh and Amann, (1991)
EuryclusArchaeaR: 24–46 (23S)TCGCAGCTTRSCACGYCCTTCBenlloch et al. (2001)
CrenarclusArchaeaR: 21–38 (23S)TCGGYGCCCGAGCCGAGCCATCCMankin et al. (1985)

Five independent PCR reactions for each primer combination were carried out under the following standard conditions: 35 cycles (denaturation at 94°C for 15 s, annealing at 55°C for 30 s, extension at 72°C for 3 min) preceded by 2 min of denaturation at 94°C and followed by 5 min of extension at 72°C. In the case of the combinations A, C, E, A′ C′ E′, F and H (Table 1) the extension time was increased to 4 min. PCR products obtained with the five independent reactions of each primer combination were mixed and purified using the Qiaquick PCR purification kit (Qiagen). Clone libraries were constructed using the Topo TA cloning system (Invitrogen) according to the manufacturer's recommendations. The plasmids of positive clones were extracted and measured with Hoescht fluorescent dye. The concentration of the plasmid/PCR product was adjusted to 0.075 pM in 6 μL by dilution with sterile water. The sequencing reaction was performed by PCR amplification in a final volume of 10 μL using 0.05 pM of plasmid, 4 pmol of primer and 1–3 μL of Big Dye Terminators premix, according to the manufacturer's protocol. After heating to 98°C for 5 min, the reaction was cycled as follows: 30–40 cycles of 30 s at 96°C, 30 s at 55°C and 4 min at 60°C (9700 thermal cycler AB). Removal of excess of Big Dye Terminators was performed by ethanol precipitation. The samples were dried in a vacuum centrifuge, dissolved in 10 μL of formamide, loaded onto an Applied Biosystems 3730 sequencer and run for 2 h.

For sequencing, primers B1055 (for bacterial clones) and Arc915 (for archaeal clones) (Table 1) were used. Study of our environmental libraries was limited to the last portion of the 16S rRNA gene so as to obtain the ITS fragment and to compare the clones retrieved with each of the primer combinations used. All 16S rRNA gene sequences presented a minimum length of 450 nucleotides.

Sequence analysis

Sequences of the last portion of the 16S rRNA gene (450–500 nucleotides) were revised and corrected with Sequencher 4.1.4 software (Gene Codes Corp.). The corrected sets of sequences were compared with 16S rRNA gene sequences available in the GenBank using NCBI (National Centre for Biotechnology Information, http://www.ncbi.nlm.nih.gov) blastall (version 2.2.9) (Altschul & Koonin, 1998). The set from the Sargasso Sea environmental sequences (Venter et al., 2004) that presented a length longer than 200 bp and a similarity of more than 90% with any sequence from the Ribosomal Database Project (http://rdp.cme.msu.edu/html/) were selected for comparison. For construction of dendrograms, sequences coming from the same sample that were 100% identical were grouped and represented as individual sequences and only those sequences from the Sso with a length of more than 430 nucleotides were used. Sequences were aligned with ClustalW and manually corrected using the BioEdit Sequence Alignment Editor program version 5.09 [http://www.mbio.ncsu.edu/BioEdit/bioedit.html. The respective neighbour-joining dendrograms were generated by using the Jukes–Cantor correction of the Phylip package (http://evolution.genetics.washington.edu/phylip.html)]. Bootstrap analysis was performed with 100 replicates and the consensus values were added to the branches of all the trees. To facilitate viewing of the dendrograms, the sequences were grouped using Sequencher 4.1.4 software (Gene Codes Corp.) at 97% similarity, and are represented as triangles. All sequences included in the triangles are detailed in supplementary Table S1.

Diversity indexes

A range of diversity indexes (Dominance-D, Shannon, Margalef) were used to compare the prokaryotic populations from the two locations and different depths, and we also included the sequences from the Sso that were retrieved according to our search criteria (see above). PAST software (Paleontological Statistics, version 1.15, http://folk.uio.no/ohammer/past) was used to compute the statistical indices in the sequences analysed (1621 clones from our study and 479 sequences from the Sso).

Sequence accession numbers

All the sequences have been submitted to the GenBank database, under accession numbers AY867880AY869668 for Bacteria and AY929619AY929799 for Archaea.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Diversity of 16S rRNA gene sequences recovered at the different sampling sites

Table 2 shows the number of sequences obtained for the main taxonomic groups with each of the primer combinations used in this study. With a few exceptions, the numbers of sequences belonging to the different groups were quite similar for all primer combinations tested, so we may consider that they are reasonably reliable for indicating marine prokaryotic diversity.

Table 2.   Number and taxonomic affiliation of the sequences amplified with the different primer combinations
GroupCombination of primers
ABCDE
Alphaproteobacteria
 Rhodospirillaceae8893
 Rhizobiaceae43268
 Sphingomonadaceae1
 SAR11676274777
 Other Alphaproteobacteria1914121626
Gammaproteobacteria
 Alteromonadales3526375542
 Thiotrichales3982421
 Chromatiales62219
 Vibrionales12
 Oceanospirillales11
 SAR8616128317
Other Gammaproteobacteria3835295412
Deltaproteobacteria3927303121
Betaproteobacteria3431
Epsilonproteobacteria1
Cyanobacteria21111211
CFB1210124315
Verrucomicrobiales376
Plantomycetales2
Firmicutes2
Nitrospira22121
Green nonsulphur Bacteria2
Fibrobacter/Acidobacteria116737
Actinobacteria474122
GroupCombination of primers
FHJ
Crenarchaeota753391

Table 3 shows the number of identical sequences retrieved from each location (considering only those clusters that have three or more members), their taxonomic affiliation and the primer combination with which they were amplified. Most sequences that appeared redundantly were obtained with different combinations of primers, indicating that they do not represent PCR artefacts. One remarkable finding regarding this redundancy is the number of identical sequences retrieved from G50 that are highly related to the isolate Pelagibacter ubique (99–100% sequence similarity). In the Ionian samples the sequences that have been amplified in high numbers were related to the groups Alteromonadales, Thiotrichales and Chromatiales. By contrast, it is notable that among the 1621 analysed, nearly all identical sequences were retrieved from the same sample. Even sequences that clustered at over 97% similarity, which could be taken as indicative of belonging to a single species (Stackebrandt et al., 1997), nearly always came from the same sample. Only three pairs of bacterial sequences from different samples were 100% identical. One of these was formed by sequences that belonged to the Bacteroidetes division and were retrieved from G50 and G2K. Another pair, related to Alteromonas marina, was formed by sequences obtained in I50 and I3K (in both cases identical sequences came from the same sites but at different depths). Nevertheless, they showed less than 87% similarity in the ITS section determined (data not shown). Finally, one pair of 16S rRNA genes showing 100% sequence similarity belonging to the group SAR86 had 97% ITS similarity and were retrieved from G50 and I50 (same depth but different sites). This was the only case of a nearly identical sequence over the whole stretch determined here and could reflect a cosmopolitan representation of this particular sequence type. In the Archaea, redundancy was much higher (four groups comprising 24 sequences) and all the identical sequences came from I3K. Three of these (containing 20 sequences) matched with different crenarchaeotal sequences retrieved in Mediterranean deep waters (Garcia-Martinez & Rodriguez-Valera, 2000) and the other, including four sequences, matched with an uncultured archaeon retrieved from a deep-sea hydrothermal vent in the Japan Sea (AB193977).

Table 3.   Number of identical sequences found in each sampling site
Identical sequencesBest match and % similaritySample*Primer combination
  • *

    I50=Ionian 50 m, G50=Greenland 50 m, I3K=Ionian 3000 m and G2K=Greenland 2000 m.

  • Origin of the sequences is shown as well as the combination of primers that have amplified each group of identical sequences.

21Pelagibacter ubique (100%)G50(5A, 1B, 5C and 10E)
12Pelagibacter ubique (>99%)G50(5A, 1B, 5C and 1D)
8Pelagibacter ubique (>99%)G50(1A, 3B, 2C, 1D and 1E)
4Other SAR11 (>94%)I3K(2B, 1C and 1E)
3Other SAR11 (>91%)I3K(2B and 1C)
3Other SAR11 (>90%)I3K(2A and 1C)
37Thiotrichales (>93%)I3K(1A, 2B, 3C, 19D and 11E)
4Thiotrichales (>91%)I50(1A, 2B and 1E)
17Thiotrichales (>85%)I50(6B, 4C and 7E)
14Chromatiales (>93%)I3K(3A, 2B, 1C and8D)
5Chromatiales (>84%)I50(3C and 2E)
7Firmicutes (>90%)G2K(2B, 3C and 2D)
20Pseudomarynaantarctica (100%)I50(4A, 1B, 5C, 7D and 3E)
11Pseudomaryna antarctica (>99%)I50(3C, 4D and 4E)
3Alteromonas marina (100%)I50(1A, 1C and 1E)
5Shewanella frigidimarina (>99%)I50(5D)
6Other Alteromonadales (>90%)I3K(1C and 5D)
4Other Alteromonadales (>88%)G2K(2A, 1C and 1D)
9Nitromincula (>89%)I3K(5A, 1C and 3D)
10SAR406 (>93%)I3K(1A, 2B, 1C, 1D and 5E)
11SAR406 (>90%)I3K(7A, 3C and 1D)
5SAR86 (>99%)G50(1C and 4D)
4Desulfuromonadales (>86%)I3K(1A, 2C and 1D)
4Burkholderiales (>98%)I3K(2C, 1D and 1E)
6Rhizobiales (>90%)I3K(4A and 2E)
3Flavobacteria (>95%)I50(1A, 1D and 1E)
7Unc.Cren. M3-450-2(>97%)I3K(1F and 6J)
3Unc.Cren.M3-150-1(>97%)I3K(1F, 1H and 1J)
10Unc.Cren.M3-150-3(>97%)I3K(7F and 3J)
4Unc.Cren.SSM264-NA06(>95%)I3K(4J)

The combined library of the 1621 clones from this study plus the 478 sequences retrieved from the Sso (see Materials and methods) contained 437 sequence types or operational taxonomic units (OTUs), assuming that one OTU is formed by the sequences that present a similarity equal to or greater than 97% (Stackebrandt et al., 1997). Table 4 illustrates the number of OTUs and the number of sequences in each sample together with the values of the diversity indexes obtained. The number of OTUs was consistent with the highest diversity being: I50>Sso>G2K>I3K>G50. Consistent with this, the Dominance-D index in sample G50 was much higher than in the remaining samples, owing to the dominance of the SAR11 relatives in this sample. The rarefaction curves (Fig. 1) show even more clearly that G50 has the lowest diversity, I50 being the most diverse sample, and still far from reaching saturation.

Table 4.   Statistical indices calculated for all the sequences from this study and those retrieved from the Sargasso Sea metagenome database
 Sample*
I50G50I3KG2KSso
  • *

    I50=Ionian 50 m, G50=Greenland 50 m, I3K=Ionian 3000 m, G2K=Greenland 2000 m and Sso=Sargasso Sea metagenome.

  • 97% similarity clusters have been considered as operational taxonomic units (OTUs).

Number of OTUs1268495106113
Number of sequences361458474343479
Dominance-D0.040.160.050.030.06
Shannon index4.022.933.613.893.69
Margalef index21.2313.5415.2617.8218.15
image

Figure 1.  Rarefaction analysis for the five different samples compared in the study. Curves were calculated using the algorithm described by Hurlbert (1971) and plotted as the number of operational taxonomic units (OTUs) vs. the number of clones, assuming that one OTU is formed by the sequences that present a similarity equal to or greater than 97%.

Download figure to PowerPoint

Prokaryotic diversity at different depths and locations

The four libraries constructed with the different samples were clearly dominated by proteobacterial sequences, mostly from the Alphaproteobacteria (in Greenland samples) and Gammaproteobacteria (in Ionian samples). Other important groups retrieved were Deltaproteobacteria, Bacteroidetes, Actinobacteria and Fibrobacter, whereas members of Nitrospira, Cyanobacteria, Verrucomicrobia, Planctomyces and Betaproteobacteria were retrieved in very low numbers. Although all the sequences have been analysed and deposited in GenBank, only the main phylogenetic groups are discussed here in more detail.

Alphaproteobacteria

We found 344 sequences affiliated to this subdivision (Fig. 2) where the largest group was formed by sequences similar enough to be considered members of the SAR11 cluster (Morris et al., 2002), probably the most abundant prokaryotic group found in open-ocean superficial waters (Field et al., 1997; Lopez-Garcia et al., 2001b; Bano & Hollibaugh, 2002; Morris et al., 2002). As expected, SAR11-related sequences were found in very high numbers in Greenland samples, both from surface water and at depth, but surprisingly only three of 40 clones of alphaproteobacterial sequences were retrieved from the Mediterranean surface seawater, marking a very important difference in the bacterial composition of this sample. Overall, the sequences from the G50 sample and those retrieved from the Sso database (Groups A31–A33 in Fig. 2) were highly related to the only cultivated species of this group, Pelagibacter ubique. The majority of the sequences found in G2K and I3K, and some retrieved from the Sso database (Groups A19–A30) were close to different clones retrieved in a previous work from deep Mediterranean waters (Garcia-Martinez & Rodriguez-Valera, 2000). Heterogeneity of these groups appeared to be very high, and they probably contain many diverse species.

image

Figure 2.  (a) Rooted dendrogram depicting relationship among Alphaproteobacteria 16S rRNA gene fragments obtained from the different samples: Ionian 50 m (green), Ionian 3000 m (orange), Greenland 50 m (pink), Greenland 2000 m (blue) and Sargasso Sea metagenome (grey). In the case of identical sequences only one is represented and numbers indicate how many clones are identical. The archaeon Haloarcula hispanica has been used as outgroup. The scale bar indicates the number of substitutions by sequence position. (b) Clone distribution over the samples.

Download figure to PowerPoint

The sequences affiliated to the Alphaproteobacteria not related to the SAR11 cluster were retrieved from all the samples, including the Sso database, and were related to the uncultured cluster SAR116, and the genera Pseudovibrio, Erythrobacter and Phaeospirilllum (Groups A1–A11). Some other sequences also appeared in these clusters, showing that they represent consistent groups with a significant level of microdiversity (Groups A12–A18); however, levels of similarity to sequences deposited in GenBank were so low that it was not possible to affiliate them to any known microorganism, the most closely related genera being Azospirillum, Blastochloris and Devosia (associated with the nitrogen cycle), with sequence similarities below 91%.

Gammaproteobacteria

The majority of the bacterial sequences analysed were affiliated to the Gammaproteobacteria (30% of the total), in which the largest groups were formed by the sequences related to the uncultured cluster SAR86 and to the well-known genera Alteromonas and Pseudoalteromonas (Fig. 3). An interesting trend was observed contrasting with what we found with the SAR11 cluster. All the Pseudoalteromonas and almost all the Alteromonas sequences were retrieved from superficial waters and most of them were found in sample I50. Other sequences belonging to Alteromonadales were found in G50 and the Sso database but those were related to the genera Colwellia, Shewanella, Photobacterium or Alcanivorax. The second important group (60 sequences) comprised sequences affiliated to the SAR86 group and falls in three different clades. Two of these were mainly found in the surface waters of the Greenland and Sargasso seas and were related to SAR86 (L35461) or to clone EBAC20E09 (AY552545), a bacterial artificial chromosome (BAC) clone from Monterrey Bay. The other clade (29 sequences) is represented principally in the Greenland deep-water sample (and very few in the Ionian) and clustered together with the uncultured bacterium 311 EBAC750-02H09 (Y458632).

imageimage

Figure 3.  (a) Rooted dendrogram depicting relationship among Gammaproteobacteria 16S rRNA gene fragments obtained from the different samples. Code colour as in Fig. 2. In the case of identical sequences only one is represented and numbers indicate how many clones are identical. The archaeon Haloarcula hispanica has been used as outgroup. The scale bar indicates the number of substitutions by sequence position. (b) Clone distribution over the samples.

The Thiotrichales and Chromatiales were clearly associated with the deep Ionian sample; few sequences were also found at the surface and even fewer at the deep Greenland site. The presence of these groups in deep-water samples was unexpected, and although the possibility that these represent artefacts cannot be totally disregarded, it seems unlikely as they were retrieved with different combinations of primers. The Thiotrichales-related sequences were similar (up to 94%) to the sulphur bacterium Achromatium minus (AJ010596) and probably represent organisms involved in sulphur-oxidation processes. Even more surprising is the cluster containing sequences related (93% similarity) to Chromatium sp. NZ (AF384209), all but one (coming from I50) from the deep Ionian sample. Sequence similarity was high enough to infer a significant relationship but not to make any predictions about the physiology of this group. Nonetheless, it strengthens the impression that some form of sulphur-based metabolism might be important in the deep Mediterranean waters (see below).

Deltaproteobacteria

A relatively large number (142 sequences) of different clones affiliated to this subdivision was found in our libraries. This group appeared clearly associated with depth as 81% of the clones were retrieved from I3K and G2K. As shown in the dendrogram (Fig. 4) the sequences were related to the clone 159 EBAC750-03B02 (AY458631), to some uncultured bacteria retrieved from the redox interface of the Cariaco Basin (AF224847 and AF285613) (Madrid et al., 2001), to the uncultured SAR324 cluster (Wright et al., 1997), obtained in previous studies of deep-sea waters, or to clone MERTZ_2CM_117 (AF424200), found in sediments of the Antarctic continental shelf (Bowman et al., 2003). In the superficial samples, sequences related to this group appeared in very low numbers, I50 providing the most representatives. In this sample were found the few sequences distantly related to the cultured bacterium Pelobacter sp. (AJ296616).

image

Figure 4.  (a) Rooted dendrogram depicting relationship among Deltaproteobacteria 16S rRNA gene fragments obtained from the different samples. Code colour as in Fig. 2. In the case of identical sequences only one is represented and numbers indicate how many clones are identical. The archaeon Haloarcula hispanica has been used as outgroup. The scale bar indicates the number of substitutions by sequence position. (b) Clone distribution over the samples.

Download figure to PowerPoint

Fibrobacter/Acidobacteria

Another cluster found chiefly in the two deep samples was formed by sequences related to this group, found in previous studies in marine environments, although it has been regarded as characteristic of soils. In fact, sequences belonging to this cluster were rarely found in the Sso database and in other similar surveys (Bano & Hollibaugh, 2002; Venter, 2004). Among this group, three subclusters could be discerned (Fig. 5). Group CAF2 (Fig. 5) comprised sequences from I3K and G2K that were related to an uncultured bacterium retrieved from a deep-sea hydrothermal vent (AB193932). The second subcluster, comprising Groups CAF3 and CAF4, was formed by a few sequences from G2K and I50 with high sequence similarity to clone ZA3312c (AF382116), retrieved from Atlantic waters. The last subcluster comprised Groups CAF5, CAF6 and CAF7, formed by sequences retrieved from G2K and Sso that were related to the well-known uncultured cluster SAR406. It is remarkable that clones belonging to this cosmopolitan cluster were not found in the Ionian samples either at the surface or depth.

imageimage

Figure 5.  (a) Rooted dendrogram depicting relationship among Actinobacteria, Fibrobacter/Acidobacteria and Bacteroidetes 16S rRNA gene fragments obtained from the different samples. Code colour as in Fig. 2. In the case of identical sequences only one is represented and numbers indicate how many clones are identical. The archaeon Haloarcula hispanica has been used as outgroup. The scale bar indicates the number of substitutions by sequence position. (b) Clone distribution over the samples.

Actinobacteria

Although the Actinobacteria is also a typical soil group, they represent one of the most abundant group of freshwater bacterioplankton (Glockner et al., 2000; Sekar et al., 2003) and their presence in these environments seems to be due to their resistance to protistan predation (Pernthaler, 2001; Hahn et al., 2003). However, in marine environments they are found in smaller proportions (Bruns et al., 2003; Crump et al., 2004) and almost nothing is known about their role in this habitat. In our libraries only 22 sequences related to this group were found (Fig. 5). The 10 sequences coming from the three superficial samples analysed here (I50, G50 and Sso) clustered together and in a different clade from those retrieved from G2K. All were related (up to 98% sequence similarity) to the uncultured actinobacterium MB11C06 (AY033302) or to SAR432 (AF110142), both obtained in previous studies of marine coastal samples (Rappe et al., 1999; Suzuki et al., 2001). The similarity of these clones with cultured microorganisms was very low (93%), the closest match being with the filamentous organism Microthrix parvicella, often isolated from wastewater treatment plants (Blackall et al., 1996). It is notable that no clones from sample I3K were affiliated to this microbial group.

Bacteroidetes

We found 94 sequences affiliated to this phylum, which constitutes one of the most heterogeneous groups retrieved in our libraries (Fig. 5). The sequences were retrieved from all the sampling sites including the Sso database, but only one sequence belonging to this group was retrieved from I3K. The Bacteroidetes is the only phylogroup where nearly all sequences studied here, including the Sso database, were different (no repetitions), which indicates the high diversity contained within it. Overall, our clones were related to the genera Pedobacter, Polaribacter, Tenacibacilum, Gelidibacter and Owenweeksia, and to uncultured microorganisms found in previous studies of marine environments (AY354775, AF354621) (O'Sullivan et al., 2004).

Unaffiliated Mediterranean Group

Notably, we found a group of 34 bacterial clones in the I50 sample with a very low similarity to any previously described sequence. The best hit was with environmental clone CL500-48 (AF316757) coming from an ultraoligotrophic environmment (Crater Lake), with a similarity below 90%. This environmental clone belongs to a candidate division (OP10) that has been detected in only one previous study (Urbach et al., 2001). We consider that these sequences could belong to a new bacterial group and because they were amplified with different sets of primers and checked for possible chimera formation, they are unlikely to represent an artefact.

Archaea

One hundred and eighty-one clones of marine Archaea from the different sampling sites and depths were analysed and compared with the NCBI nucleotide database, mainly those coming from different seas around the world reported in a previous work (Garcia-Martinez & Rodriguez-Valera, 2000), to be able also to compare the ITS fragments.

Amplification products were obtained from superficial waters of the Greenland Sea but no amplicons were obtained from the superficial waters of the Ionian with any of the three primer combinations tested. All the clones retrieved were related to marine Archaea Group I (Crenarchaeota). The 16S rRNA gene had a mean sequence divergence of 6.4±0.6%, consistent with closely related organisms, and we therefore analysed the ITS fragments (data not shown). We found no variability in the size of this region, with a length of c. 147 bp. As expected, no tRNAs were detected, a common feature in cultured and uncultured Crenarchaea (Achenbach-Richer & Woese, 1998; Garcia-Martinez & Rodriguez-Valera, 2000), and the main differences among the ITS of the clones were due to small insertions and deletions (data not shown). In the majority of the marine prokaryotes the length and sequence of the ITS can vary among strains of the same species (Boyer et al., 2001; Rocap et al., 2002), but in marine crenarchaeota both parameters seem to be very homogeneous, suggesting that phylogenies obtained based on the 16S rRNA gene and ITS sequences were congruent, without significant differences among them (data not shown).

The dendrogram obtained by comparing the 16S rRNA gene is shown in Fig. 6. It was not possible to include the nine sequences of Archaea retrieved from the Sso database because the sequence stretches obtained did not correspond to those used here. In the dendrogram there are five defined clusters containing sequences that grouped depending on the depth and geographical origin of the clones. All the sequences retrieved from the superficial waters of the Greenland Sea formed a consistent cluster together with some reference sequences from Mediterranean, Atlantic and Antarctic superficial waters (5, 10, 15 or 25 m deep). This cluster, named as DSCA1 in Fig. 6, comprises also eight sequences obtained in G2K and the marine archaeon Cenarchaeum symbiosum (Preston et al., 1996). Clusters DCA2 and DCA3 are formed by sequences coming from the deep Ionian sample and some reference strains retrieved from depth in the Mediterranean Sea. Cluster DCA4 contains sequences from Greenland and Ionian deep-sea waters together with some reference sequences retrieved from the Mediterranean at depths down to 400 m. Finally, it is of note that we found a group of 29 sequences obtained exclusively in the deep Ionian sample (Cluster DCA1) that showed far more sequence variability than others. Comparison of 560 bp of their 16S rRNA gene showed that they were closely related (99% nucleotide identity) to crenarchaeotal clones retrieved from a deep-sea hypersaline anoxic basin in the Mediterranean (AY226371 and AY64299).

image

Figure 6.  (a) Rooted dendrogram depicting relationship among Archaea 16S rRNA gene fragments obtained from the different samples. Code colour as in Fig. 2. In the case of identical sequences only one is represented and numbers indicate how many clones are identical. The archaeon Haloarcula hispanica has been used as outgroup. The scale bar indicates the number of substitutions by sequence position. (b) Clone distribution over the samples.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We are aware of the multiple biases to which PCR amplification can be prone and the problem of chimeric PCR products. To decrease both, several combinations of primers matching with the 16S rRNA gene and two primers at the beginning of the 23S rRNA gene were used (both for Archaea and Bacteria). In this way, the 3′ end of the amplicon contained the 16S–23S spacer, normally highly variable in length and sequence even within close taxonomic units, and this sequence diversity decreases the risk of chimera formation. One of the 23S rRNA gene primers (23S0R) may have biased the results, and therefore we did not use the sequences obtained with it for most comparisons. The database produced by Sso shotgun sequencing was also introduced in the comparison. It must be noted, however, in the case of Sso, that the sample was collected directly from the surface (5 m), which could lead to a significant difference with our surface samples, collected from 50 m depth and corresponding to the DCM (maximum of photosynthetic primary productivity) in the Mediterranean. With these caveats in mind the three samples can be considered as representative of the photic zone microbiota at the three locations. The Mediterranean and the Sso are more similar regarding latitude (36 and 32°N, respectively) and they are also similar in their average temperature and oligotrophy, whereas the Greenland sample is more different, belonging to permanently cold polar waters (72°N). In spite of the different methodologies used, many similarities were found between our results, obtained by PCR, and those of the Sargasso database, which were produced by shotgun cloning, supporting the reliability of both approaches. But any methodology will have limitations. For example, we failed to obtain representatives of some bacterial and archaeal groups that could be present in our samples. The low-G+C Gram-positive bacteria, the Cyanobacteria and the Euryarchaeota are the most prominent near absences. Lack of the first can probably be attributed to the lysis protocol, which might be insufficient to break the robust cell walls of this group and their spores. The Cyanobacteria might be partially excluded by the known difficulty associated with PCR amplifying this group with the primers used (Nubel et al., 1997; Acinas et al., 1999). The absence of representatives of Euryarchaeota seems to indicate that they are less common in these samples, or restricted to particular niches, as pointed out by other authors (DeLong et al., 1994; Bintrim et al., 1997; Furhman & Davis, 1997; Maarel et al., 1998; Massana et al., 1998). However, the Sso database also has a poor representation of such groups (Venter, 2004). We consider that the numbers of clones should not be interpreted as a reliable quantification of the groups that they represent whereas the diversity of the sequences provides an absolute indication. The presence of many sequences with significant variation belonging to a given group can be interpreted as a reflection of its rich representation in the habitat.

Geography

The sequences found in this study reflect the presence of different communities with differences at the level of large taxonomic groups and, even more so, at the level of diversity within each of the groups. OTUs defined by >97% sequence similarity nearly always came from unique locations and depths, which could be taken as an indication of endemism at the species level. A few exceptional groups could be cosmopolitan, such as P. ubique relatives for which nearly identical sequences have been found worldwide (Garcia-Martinez & Rodriguez-Valera, 2000; Morris et al., 2002; Venter, 2004). Among the three superficial samples, the most different is the Mediterranean I50 sample. Here the relatively small input from the SAR11 cluster representatives is compensated for by an increase in the Alteromonadales. Previous studies have established that one of the most conspicuous members of the group Alteromonadales, Alteromonas macleodii, is far more common in warmer waters (Garcia-Martinez et al., 2002), but the very low representation of SAR11 in this sample was not expected. In previous studies SAR11 sequences had been amplified from Western Mediterranean offshore waters both using universal and using specific primers (Acinas et al., 1999; Garcia-Martinez & Rodriguez-Valera, 2000) and it was found that whereas SAR11 cluster representatives were relatively common in the surface waters (5 m) their numbers decreased markedly even at the DCM (c. 50 m). However, in this study the number of sequences was too small to extrapolate to any quantification of abundance. In fact, data from the literature indicate that SAR11 decreases in its relative abundance with depth also in other locations (Field et al., 1997; Morris et al., 2002), but this decrease is not significant enough to explain the small representation found here. This is not the case for the Greenland sample, where abundant SAR11 sequences both at 50 and at 2000 m were found. One possible explanation for this Mediterranean anomaly might reside in its warmer water at depth. The SAR11 cluster organisms have been described as typical k strategists with very slow growth rates (Malmstrom et al., 2004; Simu & Hagstrom, 2004), whereas the Alteromonadaceae tend to be more r strategists, with larger cell volume and higher growth rates. The warm Mediterranean deep water mass might give them a competition edge, allowing faster growth rates by the better uptake of available nutrients. A similar replacement in part of SAR11 by Roseobacter clade representatives has been found in coastal waters (Malmstrom et al., 2004). Even within the Alteromonadales there was a significant variation in the Mediterranean samples, with the absence of Colwellia relatives that were found in the cold Greenland waters. Among the Gammaproteobacteria it was also notable that SAR86 relatives in the Mediterranean samples were scarce, a group that is well represented in Greenland and Sargasso waters. Along the same lines, the finding of a new group (Unaffiliated Mediterranean Group) distantly related to all other sequences deposited in GenBank and only present in the Mediterranean surface sample strengthens the distinctiveness of this sampling site. The analyses of the archaeal sequences also reflected this uniqueness. The marine crenarchaeota are a highly homogeneous group, and it was therefore surprising that we detected 29 clones coming from depth in the Ionian Sea with 16S rRNA gene and ITS sequences different enough form the rest of the clones (37% differences in the ITS nucleotide sequence, data not shown) to consider them as a new group. Comparison of the last third portion of their 16S rRNA gene showed that they were closely related (99% nucleotide identity) to crenarchaeotal clones retrieved from a deep-sea hypersaline anoxic basin in the Mediterranean (AY226371 and AY64299) so they may represent a particular species adapted to the deep Mediterranean waters.

The strong continental influence of the Mediterranean, being a landlocked sea, might be a factor in differences of its microbial composition but it is important to underline that the sampling site was far from any major source of pollution such as large cities or industrial complexes. Distance to the Sicilian coast (40 nautical miles) and the 3400-m depth of the sampling site decreases the potential impact of continental influence.

By contrast, the two Atlantic samples were more similar than expected, as the composition of the large groups was remarkably similar. For example, both sites had large numbers of SAR11-related sequences and although sequences clustered separately, some groups that may represent the same species (>97% similarity) were retrieved from both locations. The Alteromonadales marked a certain difference, as Pseudoalteromonas and Alteromonas were found in the Sargasso but were totally absent in Greenland, and the opposite was apparent for Colwellia, a well-known psycrophilic bacterium (Huston et al., 2004). Contrastingly, the Shewanella clade was present in Sso and I50 and absent in G50. In the case of Fibrobacter, Sso sequences seem to be more related to G50 and G2K than to Ionian samples. It is interesting to note that the only case of identical sequences (including the stretch of ITS) from different sites corresponded to SAR86 clones from I50 and G50.

Depth

A number of studies have shown that the bacterial community structure in the open ocean varies significantly with depth (Lopez-Garcia et al., 2001a; Bano & Hollibaugh, 2002; Morris et al., 2002), and this is even more so for their archaeal counterparts (DeLong et al., 1994; Massana et al., 1997; Bano et al., 2004). The general trend of an increase of Gammaproteobacteria at the expense of Alphaproteobacteria (Furhman & Davis, 1997; Lopez-Garcia et al., 2001a) is confirmed in this work and might be partially explained by the decrease of the relative contribution of the SAR11 cluster. Within the SAR11 cluster the group of sequences that are more different from the cultured P. ubique (Groups A19–A30 in Fig. 2) contains sequences retrieved mostly from deep-water samples.

The presence of some groups in our deep-water samples was also remarkable and in some ways unexpected, for example the Thiotrichales- and Chromatiales-related sequences that were abundantly retrieved from I3K. Even the few sequences that were not retrieved from this deep sample corresponded to I50 (except for three) keeping a close association with the Mediterranean habitat. Although a single clone similar to Thiotrichales was retrieved at 400 m depth in the Western Mediterranean (Acinas et al., 1999) it is not usual to find this group in the deep ocean. In addition, the finding of a group of crenarchaeotal sequences coming from the deep Ionian that were only distantly related to the rest of the sequences supports the unique prokaryotic composition of the deep-sea Mediterranean.

Another group of sequences clearly associated with depth belonged to the Deltaproteobacteria. The majority of the sequences of this group were retrieved in samples G2K and I3K with a few clusters in I50 that are not present in G50 and Sso. The groups, Fibrobacter/Acidobacteria and Actinobacteria, which were also associated with deep waters, came largely from the G2K sample.

Overall, comparing the two deep samples from the Mediterranean and Greenland seas some important differences were found. Although the deep ocean is considered as relatively homogeneous around the world with regarding to physical conditions, the two samples studied here represent very different deep ocean environments. The Mediterranean deep sea is unique among this kind of environment, being uniformly warm (>13°C) compared with the 4–10°C of the temperate Atlantic areas (CIESM 2003). Although the depth of the Greenland sample was less than in the Ionian (2000 vs. 3000 m), the deep communities are probably more similar than their superficial counterparts. Unfortunately, most of the sequences from the deep samples correspond to organisms of unknown physiology. Members of the Deltaproteobacteria have been found in a number of molecular surveys of marine environments, particularly from deep samples (Lopez-Garcia et al., 2001a; Madrid et al., 2001; Bano & Hollibaugh, 2002), and although this group has been demonstrated to be extremely diverse physiologically, a certain tendency to anaerobic niches and chemolithotrophy is often present (Bowman et al., 2003; Dhillon et al., 2003; Abildgaard et al., 2004). In fact, the environmental clones Car168 and Car118rc were retrieved from the redox interface of the Cariaco Basin, an environment that is permanently anoxic and sulphidic. Unfortunately, the similarities detected with cultivated organisms are very low, hampering any assessment of their physiology.

In the case of Thiothricales-related sequences, similarity to cultivated relatives was also very low (94% with Achromatium minus) and again no inference about their physiology can be made. The same is true for the Chromatiales-related sequences that were 93% similar to Chromatium sp. NZ. These two groups have a tendency towards inorganic biochemistry often involving sulphur compounds. From this point of view, the Eastern Mediterranean basin possesses unique geological features. Compressive tectonic activity produces active fluid and reduced gas emissions, and several deep, hypersaline, anoxic basins rich in sulphide have developed (Bregant et al., 1990), all of which might lead to a particular microbiota. In the last few years comprehensive sequencing efforts aimed at retrieving as much genetic diversity as possible from the ocean have been initiated. This work underlines that the large variations in the microbiota found at different geographical locations and (even more so) at different depths should be considered as an important factor in the study of oceanic prokaryotic diversity.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

This work was funded by the MIRACLE (EVK3-2002-00087) project of the European Commission. Samples from the Ionian Sea were collected during oceanographic cruise Talastro of the R/V Urania (Italian CNR). Support from Laura Giuliano and Marcella Leonardi during the cruise is gratefully acknowledged. The Greenland Sea sample was collected during a cruise related to the SUBMAR project using the research vessel Håkon Mosby from the University of Bergen. We are very grateful to Jørn Einen and Bjørn Olav Steinsbu for performing the water collection and the filtration. We also acknowledge Alex Mira for his helpful comments.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References