Introduction

The study of biodiversity of marine microbial communities has become commonplace in microbial ecology, with the 16S ribosomal RNA (rRNA) gene being the most frequently used phylogenetic marker (Amann et al., 1995; Pernthaler and Amann, 2005). This marker gene has revealed the great genetic diversity of bacteria that is now assembled in the Ribosomal Database Project-II (RDP-II; http://rdp.cme.msu.edu/). This currently stores approximately 418 000 partial 16S rRNA gene sequences, though it is likely that there is some degree of redundancy and sequence anomalies within the database (Ashelford et al., 2005). This genetic diversity also means that it is very unlikely that the whole range of diversity can be detected in a single sample, especially where communities are complex. Attempts to estimate the total bacterial community diversity have used partial analysis of the total community (that is clone library screening) combined with theoretical models. This approach, however, does not reveal the identity of the less abundant components of the assemblage. Increasing the number of clones per clone library has been successful in detecting novel bacterial clades (for example, Chouari et al., 2005) or comparing different marine environments in terms of bacterial community composition (for example, Rappé et al., 2000; Zaballos et al., 2006). However, despite the decreasing costs for nucleotide sequencing, the success of this approach is still limited because of the huge scale of bacterial diversity — perhaps as many as 2 × 106 different species in the oceans (Curtis et al., 2002). The recent arrival of pyrosequencing of 16S rRNA tags (Sogin et al., 2006) may represent an alternative because of the much lower cost per sequence; but pyrosequencing still does not allow analyses and comparison of the bacterial diversity in different environments on a routine basis.

Alternative approaches, such as denaturing gradient gel electrophoresis (DGGE), are routinely used to determine diversity because they avoid large-scale sequencing efforts. However, these are also likely to detect only a small subfraction of the total diversity. The use of bacterial PCR primers is likely to miss minor fractions of the microbial community because most of the PCR product will be composed of the more abundant species. Faint DNA bands on DGGE gels are unlikely to be detected or their identity determined. To overcome this limitation and to detect less abundant sequence clones, Holben et al. (2004) fractionated 16S rRNA gene sequences from a microbial community according to their G+C content before DGGE analysis. However, with the possible exception of the high G+C-containing Actinobacteria, this method has limited application and does not separate bacteria by phylogeny. Combining bromodeoxyuridine immunocapture and DGGE has been proposed to separate the DNA of the actively growing bacteria from the rest of the environmental DNA (Hamasaki et al., 2007). However, this results in the analysis of subgroups of bacteria that are not defined on phylogenetic criteria, and does not allow screening of the whole range of genetic bacterial diversity.

An alternative approach is the application of group-specific oligonucleotide probes. This has mainly been used in combination with fluorescence in situ hybridization (FISH). FISH has been a very effective technique but, in contrast to the analysis of clone libraries produced from PCR fragments, FISH cannot reveal unknown bacterial groups. It only detects and quantifies those bacteria that the probes were designed to detect. Group-specific PCR primers have already been used for the analysis of complex microbial communities. For example, Mehling et al. (1995) designed and used PCR primers for the 16S rRNA gene sequences of Streptomycetes. Driven by the biotechnological potential of compounds and activities found in Actinobacteria in general, Stach et al. (2003) developed new primers to detect a wide range of Actinobacteria in terrestrial and marine environments. Cyanobacterial and chloroplast 16S rRNA gene fragments can also be amplified specifically (Nübel et al., 1997). PCR primers have also been designed to amplify more defined groups of bacteria, such as those of the genus Pseudomonas (for example, Widmer et al., 1998). Many studies on pathogens have focused on well-defined species or strains of pathogens in the environment (for example, Miyamoto et al., 1997). Other microbial ecology studies have involved PCR primers specific for functional groups, such as the ammonia-oxidizing bacteria (McCaig et al., 1994; Freitag and Prosser, 2003; Calvo et al., 2004) and sulphate-reducing bacteria (Stubner, 2004; Dar et al., 2005). The study by Dar et al. (2005) provides also a convincing case that group-specific primers can be used in combination with DGGE to detect less abundant bacterial groups in a complex community. The novelty of that approach was to use group-specific primers in a PCR followed by reamplification of the PCR product with nested, universal bacterial DGGE primer. This overcomes the difficulty in designing novel DGGE primers and establishing the optimal conditions for those primers (Dar et al., 2005).

As 16S rRNA gene sequence databases continue to grow, group-specific PCR primers must be continually re-evaluated for their specificity and range of sequence matches (Baker et al., 2003). Furthermore, new and more sophisticated software packages, such as PRIMROSE (Ashelford et al., 2002) and ARB (Ludwig et al., 2004), can help researchers to design 16S rRNA gene probes. These software packages use different sequence databases and each has its specific strengths. For example, as well as differences in the algorithm used for searching for priming sites, it is possible to use PRIMROSE to design degenerate probes, a function that is not available in ARB. ARB, however, allows the search for probes to be based on a greater number of specific parameters.

To analyse the total bacterial community during changing environmental conditions and seasons, we have developed group-specific primers for seven different taxonomic groups. Apart from expensive large-scale sequencing of clone libraries, DGGE remains one of the few fingerprinting techniques currently available that allows analysis of the whole microbial community and identification, by sequence analysis of DGGE bands, of specific members of that community. Therefore, we used a similar DGGE approach to that of Dar et al. (2005), that is reamplification of the bacterial group-specific PCR products with a nested or semi-nested universal DGGE-PCR primer set. We tested these primers for their specificity and applicability by sequence analysis of clone libraries and comparative DGGE analysis of total environmental DNA from a coastal nutrient-rich and an oligotrophic open ocean environment.

Materials and methods

PCR primer design

PCR primers were designed using the ARB (Ludwig et al., 2004) and PRIMROSE (Ashelford et al., 2002) programs; the latter is part of the TOOLKIT software package (http://www.cf.ac.uk/biosi/research/biosoft). The name given to each primer consists of a short name indicating the target group, followed by a number, representing the position of the first base of the primer within the Escherichia coli 16S rRNA gene sequence and ‘f’ or ‘r’ indicating whether the primer is the forward or reverse primer, respectively. The sequences of the primers and their specificity are summarized in Table 1. The number of target and non-target matches of the primers was tested in silico using the PROBE MATCH function within the RDP-II database (http://rdp.cme.msu.edu) and the ROSE function of TOOLKIT which identifies the number of sequences, among the 15 104 Bacterial sequences in its database, to which the primer sequence is homologous.

Table 1 Summary of group-specific 16S rRNA gene PCR primers, their specificity towards taxonomic groups as revealed by in silico analysis and the annealing temperatures used in the PCR reactions
Full size table

A nested Bacteria primer pair was selected for each group-specific primer pair to be used in a nested re-PCR (Table 2). Only one nested Bacteria primer was available in the cases of the Betaproteobacteria-, the Bacteroidetes- and the Cyanobacteria/chloroplast-specific primers. Therefore, one of the group-specific primers was paired with the Bacteria primer for a semi-nested re-PCR (Table 2), and a GC clamp was attached to the 5′-position of one of the two primers (indicated in the primer name with ‘-GC’; Table 2).

Table 2 Nested PCR approach with DGGE primers
Full size table

Sampling

Samples from two contrasting environments were used in this study — a mesotrophic coastal and an oligotrophic open ocean environment. The coastal sample was collected in May 2003 from a 20-m3 mesocosm after 12 days of incubation during the ‘Pelagic Ecosystem CO2 Enrichment Study’ (PeECE; http://peece.ifm-geomar.de/) at the EU Large-Scale-Facility, University of Bergen, Norway, located at Espeland in the Raunefjord (60.27°N 5.22°E), 20 km south of Bergen. During the experiment, nitrate, phosphate and silicate were added at typical winter concentrations to stimulate a phytoplankton bloom. The open ocean, oligotrophic sample was obtained from 15 m depth in the Northern Atlantic Gyre (35°N 20°W) collected during the Atlantic Meridional Transect (AMT-15) cruise in September 2004 (Robinson et al., 2006).

Samples (4 l) from the mesocosm and samples (7 l) from the Northern Atlantic Gyre, respectively, were filtered through Sterivex cartridges with 0.2 μm pore size filter membrane (Millipore, Watford, UK). The cartridges were stored at −70 °C until analysis. Total environmental DNA was isolated from the cartridges using the method described by Jameson et al. (2007). This involved lysozyme treatment with 1.5 ml of lysis buffer (200 mM Tris-HCl, pH 8.0; 1.5 mg lysozyme; 125 mM EDTA), incubated at 37 °C on a rotary shaker (10 r.p.m.) for 30 min, with a 90° turn of the cartridge every 6 min to ensure coverage of the whole inner surface area of the cartridge. After addition of 0.25 ml of proteinase K lysis buffer (to a final concentration of 1.25% w/v SDS, 300 μg proteinase K, 70 mM NaCl), the cartridge was incubated at 37 °C for 15 min followed by another 60 min at 55 °C, again with regular rotation of the cartridge. Subsequently, the lysis mix was transferred to microcentrifuge tubes and extracted twice with one volume of chloroform/isopropanol (24:1). The DNA was precipitated (1 vol isopropanol; 0.4 vol 7.5 M ammonium acetate), and the pellet was washed with 70% (v/v) ethanol. The DNA pellet was resuspended in sterile, double-distilled water.

PCR and preparation of 16S rRNA gene clone libraries

Clone libraries were prepared from PCR-amplified 16S rRNA gene fragments for each of the group-specific primer pairs. A nested PCR approach was required for three of the primer pairs to obtain sufficient PCR product for subsequent cloning (see Table 1). In these cases, aliquots of the PCR products obtained with the Bacteria primer pair 9bfm/1512uR were used as templates for a reamplification with a nested group-specific primer set. A typical PCR reaction was carried out in 25-μl volumes and contained 2 mmol l−1 MgCl2, 200 μmol l−1 dNTPs, 2.5 U of Taq DNA polymerase (Promega, Southampton, UK) and 50 nmol l−1 of each of the primers. The temperatures for primer annealing were dependent on the primer pair used (summarized in Table 1). They were determined using the equation Tm=(4 × GC)+(2 × AT). A range of annealing temperatures (ATs) similar to the calculated Tm were then tested in a gradient PCR to determine empirically the temperature that resulted in the most specific PCR product. All PCRs used the same basic cycle protocol except for the AT: following an initial denaturation step of 4 min at 96 °C, 30 PCR cycles were performed (96 °C for 60 s, AT for 60 s, 74 °C for 60 s) followed by a final extension step at 74 °C for 10 min. The extension time was increased to 90 s for primer pair 9bfm/1512uR. In all cases, the PCR yielded only specific products, that is single bands as judged by electrophoresis of the PCR products on agarose gels.

Aliquots of the products from PCRs or re-PCRs were cloned into a TA vector using the pGEM-T Easy Vector System I cloning kit (Promega). Twenty or 50 clones were picked from each of the group-specific clone libraries prepared from the mesocosm and the Northern Atlantic Gyre samples, respectively, and the 16S rRNA gene fragments were reamplified using vector primers (M13). The PCR products were used for the sequencing reactions.

Sequence analysis

PCR products were treated with ExoSapIT (Amersham Biosciences, Little Chalfont, UK) according to the manufacturer's instructions and used directly for sequence analysis. Nucleotide sequencing was performed using the BigDye Terminator v3.1 cycle sequencing kit (ABI). An M13 primer was used in the cycle sequencing reaction. Sequences were analysed on an ABI 3100 automatic sequencer. Generally, only one strand of the DNA fragments was sequenced. This proved to be sufficient for the taxonomic identification of the cloned 16S rRNA gene fragments using the BLAST search function within the NCBI database. Both DNA strands were sequenced in those cases where sequences could not be assigned easily to taxonomic groups or where the quality of the nucleotide sequence could not be determined unambiguously.

Group-specific PCR-DGGE analysis

Denaturing gradient gel electrophoresis (DGGE) was based on Muyzer et al. (1998) and Dar et al. (2005). In essence, group-specific PCR products were used as template in a re-PCR with a nested Bacteria DGGE-PCR primer pair (Table 2). Semi-nested re-PCRs were used in the case of the Cyanobacteria/chloroplasts, the Bacteroidetes and the Betaproteobacteria (Table 2). Only one bacterial primer was available within the 16S rRNA gene fragment that was amplified by the primers specific for these phylogenetic groups. PCR conditions and cycle protocol for the nested PCRs with the DGGE primer pairs were the same as those used for the PCRs with the group-specific primer pairs, except for the ATs which are shown in Table 2.

DGGE of the PCR products was performed on a 8% (w/v) polyacrylamide gel with urea and formamide as denaturants. The denaturing gradients varied with the DGGE primer pair used for the PCR but were generally between 40% and 60% (Table 2). Electrophoresis was performed in 1 × Tris-acetate EDTA buffer at 60 °C at constant voltage of 60 V for 18 h. Subsequently, gels were stained in 1 × SYBR Gold nucleic acid gel stain (Molecular Probes) for 45 min and rinsed in distilled water prior to image analysis on a Syngene GelDoc station. Individual DGGE bands were cut out from the gel and incubated overnight at 4 °C in 30 μl H2O. An aliquot (ca 5 μl) was used in a PCR with the same primer set used for DGGE (Table 2; but without the GC clamp attached to one primer) to reamplify the insert. The nucleotide sequences of these 16S rRNA gene fragments were determined as described above.

Phylogenetic analysis

Sequences from the clone library prepared with the Gammaproteobacteria-specific primers were first compared to sequences stored in GenBank using the BLAST algorithm. Subsequently, the sequences from the clone libraries, and sequences with high similarity as identified by the BLAST searches, were imported into the ARB software program (http://www.mikro.biologie.tu-muenchen.de/pub/ARB) and aligned to other Gammaproteobacteria 16S rRNA gene sequences using the automated alignment tool within ARB (Ludwig et al., 2004). The alignment was further corrected ‘by eye’, taking the secondary structure prediction of the ARB program into account. Phylogenetic analyses were based on the alignment of the 16S rRNA gene sequences from the mesocosm and Northern Atlantic Gyre clone libraries with representative sequences from the sequence databases. Calculation of the phylogenetic trees was based on these sequence alignments using the neighbour-joining method with Jukes–Cantor corrections, as well as the maximum likelihood and parsimony algorithms. Analyses were carried out using a maximum frequency filter. For each of the phylogenetic analyses in this study, the grouping of strains and environmental clones within the different clusters of the tree was identical for all of the above three phylogenetic methods for calculating trees (maximum likelihood, maximum parsimony, neighbour-joining). However, those branching points within a tree that were not supported by each of the three algorithms were collapsed within the neighbour-joining tree using a strict consensus rule until the branching was supported in all three analyses. The neighbour-joining tree was chosen for depicting the phylogenic relationship of the 16S rRNA gene clones and strains, and bootstrap values were calculated from 100 trees using the neighbour-joining method.

Nucleotide sequence accession numbers

The sequence data of 16S rRNA gene fragments have been submitted to the EMBL database with accession numbers AM706671-707020 (Northern Atlantic Gyre clone library), AM706537-AM706670 (mesocosm clone library), AM747394-AM747468 (sequences of DGGE bands).

Results and discussion

In silico analysis of the group-specific PCR primers

Blackwood et al. (2005) used ARB to develop five group-specific primers including primers for four of the groups of bacteria investigated in this study. Ashelford et al. (2002) compared in detail the PRIMROSE and ARB programs and concluded that in many cases it was possible to identify better oligonucleotide probes (judged by in silico analysis) using PRIMROSE rather than ARB. We based the development of our primers on the independent use of both software packages to ensure the best design. The theoretical specificities of all primers were tested with the ROSE program of the TOOLKIT software package, as well as the PROBE MATCH function within the RDP database (http://rdp.cme.msu.edu/). The sequences of the primers and their specificity are summarized in Table 1. For the purpose of comparison (see below under ‘Group-specific clone libraries’), we preferred the use of the ROSE program of the TOOLKIT software package, since this uses exactly the same sequences as a basis for the comparison. In contrast, results of sequence searches in the current version of the PROBE MATCH function within the RDP database depend on the search window set for the priming site, which is different for different primer locations along the 16S rRNA gene.

PCR primers were developed for the amplification of 16S rRNA gene fragments that provide valuable additions to existing primers. In silico analyses indicate that they have generally a higher number of exact matches to the 16S rRNA gene sequences from members of the target group of bacteria for which they were designed, while their specificity is generally similar to that of published primers (Table 3).

Table 3 Comparison of 16S rRNA gene group-specific primers developed in this study with those used in previous studies by Blackwood et al. (2005) and Nübel et al. (1997)
Full size table

However, some of the most suitable primers were identical to the FISH probes suggested by Ashelford et al. (2002)—Alf28f, Beta359f and Beta682r. Other probes suggested by Ashelford et al. (2002) or previously published primers (for example, Nübel et al., 1997; Blackwood et al., 2005) exploit similar priming sites to those used in our study but were, for example, of different length (for example, Alf684r, CFB555 and CYA785r) and degeneracy (for example, CYA361f). Blackwood et al. (2005) and Blümel et al. (2007) paired their group-specific primers with either the universal primer 1392r or the Bacteria primers Eub338 or 27f. In contrast, we designed two group-specific primers (that is a primer pair) for each taxonomic group to increase specificity of the PCR, or to ensure that a more group-specific primer of the primer pair would compensate for lower specificity of the other primer (for example, CFB555f/CFP968r compared to 27F/Cyt1020R of Blümel et al., 2007).

A new bacterial forward primer used for nested reamplification of PCR products

The group-specific primer pairs were generally used to amplify 16S rRNA gene fragments directly from environmental DNA. However, primer pairs for the Alphaproteobacteria, the Planctomycetes and the Firmicutes resulted in only low yield when environmental DNA was used as template. Therefore, a nested PCR approach was designed to overcome this problem. Obviously, the forward primer used in the first PCR had to be upstream of the Alphaproteobacteria-specific forward primers (Alf28f), which were located close to the 5′-end of the 16S rRNA gene. Concerns have been raised for some time when the bacterial forward primers 8f and 27F are used because of limited amplification efficiency and potential mismatches with newly discovered strains or environmental 16S rRNA gene sequences (for example, Marchesi et al., 1998). These primers were designed when the sequence databases consisted only of a few thousand clones (for example, see under http://rdp.cme.msu.edu/misc/history.jsp). Testing Bacteria primers for specificity revealed that the sequences were homologous to only 56.5% (8f; Hicks et al., 1992) or 72.9% (27F; Giovannoni et al., 1996) of the over 15 000 sequences within the PRIMROSE database. The sequence of the universal (bacteria and archaea) reverse primer 1512uR (Weisburg et al., 1991) was homologous to 78.5% of the sequences tested (Table 1).

We decided to use reverse primer 1512uR but modify the forward primer to account for sequence differences at this priming site. In silico analysis showed that the sequence of the new Bacteria primer 9bfm (Table 1) is homologous to 77.7% of the bacterial sequences in the PRIMROSE database but also to one Archaeon (Methanobrevibacter sp strain MB-9; accession AB017514). However, an archaeal 16S rRNA gene sequence was never detected, either in any of the libraries screened in this study (see below) or in any other analyses in our laboratory. Although there is only a small increase (5%) in the number of sequences to which the sequence of primer 9bfm is homologous, this is likely to lead to a far higher amplification rate. The greater degeneracy of the primer is likely to result in the successful amplification of more novel and yet unknown genotypes since many of the sequences that are not homologous to primer 9bfm will differ only in one base (as tested using the OLIGOCHECK function within the TOOLKIT software package). Furthermore, this one-base difference is not within the last three or four 3′-end bases, which are important for the polymerase to start extending the DNA strand during PCR.

However, despite the high percentage of exact matches to bacterial 16S rRNA gene sequences, primers may still be biased against certain groups of bacteria while matching the 16S rRNA gene sequence of most strains of other phylogenetic groups. Therefore, primers should always be tested with respect to the specific bacterial group of interest prior to use.

In contrast to group-specific primers (see below), the application of these general Bacteria primers aims to amplify the whole genetic diversity of the domain. Therefore, use of less stringent conditions in the PCR reactions, for example a lower AT than the melting temperature of the primer, will allow for mismatches during the annealing, thus overcoming some of the above-mentioned biases. This further increases the genetic range of bacterial 16S rRNA gene fragments that are amplified in a PCR. In practice, it was found that, despite the increased degeneracy of primer 9bfm and the lower than optimal AT used in our PCRs (52 °C instead of ⩾54 °C Tm), all PCRs resulted in specific amplification of 16S rRNA gene fragments as judged by sequence analyses of clone libraries prepared with these primers (unpublished data). It should be noted that none of the PCRs with this primer pair in our laboratory has yet led to the amplification of nonspecific products.

Group-specific 16S rRNA gene clone libraries

The newly developed primers were tested with environmental DNA samples from contrasting environments—a mesocosm experiment in a cold, coastal surface environment and a subtropical oligotrophic environment of the Northern Atlantic Gyre. As mentioned above, the following comparisons concerning specificity of the primers are based on the program ROSE of the TOOLKIT software package, though the specificity of the primers when compared using the PROBE MATCH function within the RDP database are also provided (Table 1).

Alphaproteobacteria

The oligonucleotides designed to specifically amplify the Alphaproteobacteria were identical or similar (Alf28f and Alf684r, respectively) to those suggested for use as molecular probes by Ashelford et al. (2002), and in one case (Alf684r; Table 3) similar to a primer proposed by Blackwood et al. (2005). The primer sequence of reverse primer Alf684r was homologous to that of a relatively large number (242) of bacteria outside the Alphaproteobacteria, mainly Fusobacteria and bacteria belonging to the orders Desulfovibrionales, Desulfobacterales and Desulfuromonadales of the Deltaproteobacteria. In contrast, the sequence of forward primer Alf28f was homologous only to 13 16S rRNA gene sequences outside the target group, equally distributed among the Gammaproteobacteria, Deltaproteobacteria and the Verrucomicrobia. Due to the high specificity of the forward primer and the high AT, the application of both primers as a primer pair in a PCR resulted in the amplification of 16S rRNA gene fragments from organisms that all belonged to the target group (Table 4). This primer pair therefore demonstrated the advantage of both primers of a primer pair being biased towards a particular bacterial group of interest, rather than one group-specific primer combined with a second, Bacteria or universal 16S rRNA gene primer.

Table 4 Results from the BLAST searches with the sequences of the 16S rRNA gene fragments obtained from the screening of the two clone libraries by sequence analysis
Full size table

Most of the sequences that were detected in either of the two clone libraries belong to members of the Roseobacter group and the genus Sphingomonas, but several were also members of the genera Rhodobium and Brucella. However, the rest of the sequences belonged to a wide range of different groups of the Alphaproteobacteria (Supplementary Table S1). This demonstrates the value of the group-specific primer approach in detecting specifically a wide genetic diversity within each taxonomic group.

Betaproteobacteria

Ashelford et al. (2002) also suggested a number of oligonucleotides specific for the Betaproteobacteria. Our independent search led us to two primers (Beta359f, Beta682r; Table 1), which were identical to two of those suggested by Ashelford et al. (2002). Reverse primer Beta682r exploits the same discriminatory sequence stretch as primer Beta680F used by Blackwood et al. (2005) as a forward primer, although it is different in length and precise sequence position (Table 3). All sequences analysed from the mesocosm sample were from Betaproteobacteria (Table 4). In contrast, despite the high specificity of the primers, only one-third of the sequences from the Northern Atlantic Gyre clone library were from Betaproteobacteria (Table 4). Interestingly, three of the clones detected in the clone libraries from the Northern Atlantic Gyre show sequence similarities (at 92–95% sequence similarity) to 16S rRNA gene sequences of Burkholderia (Supplementary Table S1)—also prevalent in the Sargasso Sea data set (Venter et al., 2004).

To test whether this low yield of positive hits could be improved using a nested reamplification-PCR approach, we also prepared and screened a second clone library prepared using a two-step nested PCR approach. This used an aliquot of the PCR product obtained with primers 9bfm/1512uR as template for nested reamplification with primers Beta359f/682r in a second PCR. Again, only about one in four of the 25 sequences screened from this clone library were from Betaproteobacteria. The reason for this low yield of positive hits from this oligotrophic environment is not known. We believe that the use of primer Beta682r generally results in the group-specific amplification of 16S rRNA gene fragments as the last 3′-end (G) base of the primer is highly specific for the vast majority of Betaproteobacteria sequences in the ROSE database. This is supported by the fact that the clone libraries prepared from the coastal sample proved to be composed entirely of target sequences (Table 4). However, the low specificity at the 3′-end of primers Beta359f (most 16S rRNA gene sequences have three guanosines at E. coli positions 376–378), combined with the potentially low abundance of Betaproteobacteria in the Northern Atlantic Gyre sample may have led to the low percentage of target hits in this clone library (Table 4).

Gammaproteobacteria

As in the case of the Alphaproteobacteria, one of the primers (Gamma359f) appears to have a high number (203) of matches outside the target group (Table 1). However, more than 160 of these sequence hits are due to a large number of sequences for a small number (five) of particular bacterial species. For example, 45 non-target hits are due to homologous sequences within the 16S rRNA gene of Acetobacter, Gluconobacter and Gluconacetobacter. Furthermore, 118 hits are due to the Ralstonia (54) and Burkholderia (64) sequence hits. The reason for the high number of sequences obtained from these bacteria in the databases is because they are important pathogens. Given this, and the higher specificity of the reverse primer Gamma871r (Table 1), it is not surprising that the specificity found in both clone libraries is 100% (Table 4). This is further confirmed by the fact that the application of the Betaproteobacteria-specific primers demonstrated the presence of sequences with 92–95% sequence similarity to 16S rRNA gene sequences of Burkholderia as the next nearest isolated species (Supplementary Table S1), but no such sequences were detected within the clone libraries produced with the Gammaproteobacteria-specific primers (Supplementary Table S1).

Many of the sequences obtained from the mesocosm sample have a high sequence similarity to 16S rRNA gene sequences that fall within phylogenetic clades that belong to the oligotrophic marine Gammaproteobacteria (OMG) group (Figure 1; Supplementary Table S1). This group of Gammaproteobacteria was introduced by Cho and Giovannoni (2004) to indicate that all isolates were able to grow only in low-nutrient (oligotrophic) media. Phylogenetic analysis of their 16S rRNA gene sequences further confirmed that they form independent phylogenetic clades, which together comprise the OMG group of Gammaproteobacteria (Cho and Giovannoni, 2004). The fact that Gammaproteobacteria 16S rRNA gene sequences from the mesocosm (Figure 1) cluster within two (SAR92 and OM182) of the five OMG clades identified by Cho and Giovannoni (2004) may indicate that this phylogenetic clade represents a group within the Bacteria that is genetically diverse and occurs also in coastal nutrient-rich environments. In fact, Stingl et al. (2007) state for the SAR92 clade, into which several of the 16S rRNA gene sequences from the mesocosm sample cluster, that ‘the peak of abundance correlates with the relatively high nutrient concentrations found in an upwelling region off the Oregon coast. In the lower nutrient regions farther off the coast, the abundance of the SAR92 was low, close to the limit of detection’. A possible explanation for this difference may be that under laboratory conditions, SAR92 isolates grow only under low-nutrient conditions while in their natural environment they thrive at higher nutrient concentrations such as provided by coastal environments. Alternatively, the phylotypes that we detected in the mesocosm samples may be physiologically different from those isolated by Cho and Giovannoni (2004).

Figure 1
figure 1

Phylogenetic analysis of representative 16S rRNA gene nucleotide sequences from the Gammaproteobacteria clone libraries prepared from samples from mesocosm (coastal) and Northern Atlantic Gyre (open ocean) samples and representative sequences from the NCBI and ARB sequence database. The tree was calculated from a nucleotide alignment of 16S rRNA gene fragments (356 bases) using the neighbour-joining method within ARB, with Jukes–Cantor corrections and a maximum frequency filter (Ludwig et al., 2004). Escherichia coli (accession J01859) was used as outgroup. The confidence of branch points was determined by three separate analyses (maximum likelihood, neighbour-joining, maximum parsimony), with multifurcations indicating branch points that were collapsed using a strict consensus rule until supported in all three analyses. Values of 100 bootstrap replicates (calculated using the neighbour-joining method) are given as numbers at branching points, but those <70 are omitted. Clade OM182 is split into two on this tree (a and b) as the two subclades grouped differently in one of the three methods used, and therefore the branching was collapsed to represent the most stringent consensus branching. The strains that are grouped together in the two OM182 clades are also grouped together into subclades in the phylogenetic analysis of Cho and Giovannoni (2004). Filled triangles indicate clades containing formally described species, and filled circles indicate clades or subclades for which no formally described species is available.

Full size image

It should be noted in this context that the phylogenetic analyses carried out in this study were rigorous. These included independent analyses based on the three main algorithms (maximum likelihood, neighbour-joining, parsimony; individual trees not shown) followed by comparison of the resulting trees. The phylogenetic tree shown in Figure 1 is a strict consensus of the three trees that was achieved by collapsing the tree on those branching points where there were differences between the three different trees. Also, the sets of sequences that form the individual clades and the branching within these clades were identical in all of the three different trees (individual trees not shown), thus confirming the robustness of the consensus tree (Figure 1).

The success of this Gammaproteobacteria-specific primer pair is demonstrated by the high specificity achieved by primers Gamma395f/871r and the resulting discovery of 16S rRNA gene sequences in a nutrient-rich environment that appear to belong to the OMG group of Gammaproteobacteria. In addition, a relatively small sample (70 sequences) from the open ocean and the mesocosm clone libraries revealed sequences that cluster in a wide range of Gammaproteobacteria clades (Figure 1). In this context, it will be interesting to analyse a larger number of clones per clone library than has been possible in this study, and from a variety of different geographic origins and seasons; a more extensive screening programme has the potential to discover novel phylogenetic clades of Gammaproteobacteria.

Bacteroidetes

This phylum was previously referred to as ‘Cytophaga–Flavobacterium–Bacteroides’ or ‘CFB’, hence the primer names. Ashelford et al. (2002) suggested an oligonucleotide (CFB560) as a FISH probe that exploits a similar discriminatory position within the 16S rRNA gene sequence to that of our forward primer CFB555f, which is located five bases further downstream. Primer CFB555f is highly specific with only two non-target hits (Table 1), while its sequence is identical to 84% of the Bacteroidetes sequences in the PRIMROSE database. In contrast, the reverse primer CFB968r has a relatively high number (305) of non-target hits, but the sequence is identical to over 90% of those in the database. However, 298 of the 305 non-target hits are due to identical sequences within the 16S rRNA gene of three members (Borrelia, Spirochaeta and Treponema) of the phylum Spirochaeta, and these pathogens are unlikely to be sufficiently abundant in the marine environment to be amplified by PCR. This, and the high specificity of the forward primer, seems to be responsible for the fact that all of the sequences in the clone libraries produced using this primer pair derived from members of the Bacteroidetes.

Planctomycetes

Planctomycetes species in general, and marine members of this phylum in particular, are one of the least studied groups of bacteria, mainly due to the lack of successful laboratory cultivation. Only members of two (Planctomyces and Pirellula) of the four recognized genera have been grown in culture. The Sargasso Sea shotgun sequencing project (Venter et al., 2004) revealed only a small number of molecular marker genes identifying the genomic DNA clone as deriving from the Planctomycetes. FISH studies (Glöckner et al., 1999; Brinkmeyer et al., 2003) and 16S rRNA gene clone library-based approaches (Brinkmeyer et al., 2003) have indicated low abundance (up to 3%) of Planctomycetes in a variety of environments. However, these studies may have underestimated Planctomycetes genetic diversity and abundance since the 16S rRNA gene sequences of the Planctomycetes may contain mismatches to several commonly used bacteria primers (Vergin et al., 1998). We were interested to develop new PCR primers that specifically amplify the 16S rRNA gene fragments of the Planctomycetes with the potential to discover novel sequences and possibly phylogenetic lineages.

As with Alphaproteobacteria, Gammaproteobacteria and the Bacteroidetes, one primer (Plancto920r) was highly specific for the target group, while the sequence of the other primer (forward primer Plancto350f) was homologous to 140 non-target sequences (Table 1). However, 130 of these 140 non-target hits belonged to the genera Chlamydia and Chlamydophila, which are pathogens that have not yet been detected and are unlikely to occur in the marine environment. The application of these oligonucleotides as a primer pair therefore led to specific amplification of members of the Planctomycetes only (Table 4).

Cyanobacteria and chloroplasts of eukaryotic algae (oxygenic photoautotrophs)

The primers developed here exploited similar priming sites to those developed by Nübel et al. (1997) but differ in length and degeneracy (Tables 1 and 3); in particular, forward primer CYA361f matched a larger number of target sequences and had fewer non-target hits (Tables 1 and 3). The screening of the 16S rRNA gene clone libraries demonstrated that the primers were very selective and they specifically amplified only members of the target group. The two environment studies showed contrasting results. The sample from the mesocosm experiment was dominated by 16S rRNA gene fragments of chloroplasts (Supplementary Table S1) and had a high abundance of eukaryotic photoautotrophs (data not shown), but 96% of the sequences from the Northern Atlantic Gyre sample had high similarity with Prochlorococcus isolates or environmental sequence clones. Prochlorococcus is known to dominate the oligotrophic regions of the oceans due to its temperature range, high surface to volume ratio and an efficient nutrient uptake mechanism (Scanlan and West, 2002), and would not be expected to occur in Norwegian waters—as indicated by the results obtained here.

Firmicutes

We were intrigued by the fact that the vast majority of marine bacteria detected routinely in 16S rRNA gene clone libraries are derived from Gram-negative bacteria, although a considerable part (up to 0.05% by weight of clones) of the bacterial metagenome in the Sargasso Sea were derived from the Firmicutes (Venter et al., 2004). Results from the screening of PCR-amplified 16S rRNA gene fragment clone libraries generally contained only a small fraction of 16S rRNA gene sequences from Gram-positive bacteria (for example, Zaballos et al., 2006; Martín-Cuadrado et al., 2007), suggesting that Gram-negative bacteria far outnumber Gram-positive bacteria in the sea. However, this large difference in the abundance of these two groups may be exacerbated by using a molecular approach. For example, the considerably thicker peptidoglycan layer in the Firmicutes and the higher GC content of the Actinobacteria are likely to result in lower genomic DNA yield after DNA isolation from environmental samples and lower yields during PCR amplification, respectively.

To investigate specifically the diversity of Gram-positive bacteria in the oceans, we designed specific primers for this group. Actinobacteria-specific primers are already available and proved to be successful for marine samples (Stach et al., 2003). Interestingly, all our attempts to provide improved PCR primers for the Actinobacteria resulted in exactly the same primers as those suggested by Stach et al. (2003) or Blackwood et al. (2005). Due to the specificity and extensive testing by Stach et al. (2003) and Blackwood et al. (2005), we decided to concentrate further efforts on the Firmicutes group. This group consists of three classes with 33 families of bacteria (Garrity et al., 2003), so we attempted to design primers that amplify 16S rRNA gene sequences from the whole genetic range of this phylum, thus avoiding the need for multiple PCRs for the screening of the many different phylogenetic groups. Given the huge genetic range among the Firmicutes and the low number of matches of the primers to 16S rRNA gene sequences outside the target group (Table 1), it is not surprising that the sequences of the primers are identical only to about 25% of the 2267 Firmicutes sequences within the PRIMROSE database. However, despite the high specificity of the primers (Table 1), 6 of the 20 sequences of the clone library prepared from the mesocosm DNA sample using these primers were from Cyanobacteria (two) or Alphaproteobacteria (four). Four of the seven non-target hits of forward primer, Firm350f, were to 16S rRNA gene sequences from chloroplasts and two were to Alphaproteobacteria. The sequence of the reverse primer, Firm814r, is homologous to 17 non-target hits, 14 of which are members of the Geobacter group within the Deltaproteobacteria and only one from an Alphaproteobacterium (Ehrlichia), but none to chloroplasts or Cyanobacteria. The clone library from the oligotrophic Northern Atlantic Gyre sample consisted entirely of cyanobacterial (Prochlorococcus) 16S rRNA gene fragments. The precise reason for this is unknown. We believe that the bias of the last 3′-end (C) base of primer Firm350f for a wide range of the Firmicutes sequences in the ROSE database was sufficient to amplify 16S rRNA gene fragments mainly from the target group in the case of the mesocosm sample. However, in the North Atlantic Gyre sample, the low specificity at the 3′-end of primer Firm814r for their target groups (that is most 16S rRNA gene sequences have three adenosines at E. coli positions 814–816) combined with the potential low abundance of bacteria of the target group may have led to the absence of any target hits in the Firmicutes clone libraries (Table 4). This explanation is supported by the fact that the last three 3′-end bases of forward primer Firm350f are also present in all Cyanobacteria/chloroplast 16S rRNA gene sequences in the ROSE database, and all of the unspecific sequences derived from the cyanobacterium Prochlorococcus, which was the dominant autophototroph in this sample (6.4 × 104 cells ml−1; Jameson et al., 2007).

DGGE analysis to detect differences in assemblage diversity between a coastal and an open ocean environment

As outlined above, DGGE with group-specific PCR primers is a very powerful method for the detailed study of microbial assemblages. All of the group-specific primers described in this study were originally designed as DGGE primers, that is a GC clamp was attached to one of the primers of each pair. However, the primers did not provide reproducible DGGE patterns and we decided to adapt the PCR-DGGE approach of Dar et al. (2005). This requires a nested PCR with Bacteria primers (one primer with a GC clamp attached to it) subsequent to the PCR with the group-specific primers, thus resulting in a two-step nested or, in the cases of the Alphaproteobacteria, the Planctomycetes and the Firmicutes, a three-step nested PCR-DGGE approach. Tests of the two-step nested PCR approach showed that low yield of the PCRs that sometimes occurred with the specific primers was not a problem, since the nested reamplification with the bacterial DGGE primer sets resulted in an amount of PCR product sufficient for subsequent DGGE analysis.

This nested PCR approach was successful with all primer pairs when applied to environmental DNA samples from both environments tested. In the case of the Betaproteobacteria, Bacteroidetes and the Cyanobacteria/chloroplast, Bacteria primers are not available for nested PCRs. Therefore, a semi-nested approach was used in the second PCR, with one Bacteria primer and one that was specific for the group (Table 2). The results of the DGGE analysis are summarized in Figure 2. In all cases, there were detectable differences in the microbial assemblages between the Northern Atlantic Gyre and the mesocosm sample. However, the separation of the PCR fragments appeared to be better in those cases where a nested PCR was used as compared to a semi-nested PCR (that is for the Betaproteobacteria, the Bacteroidetes and the Cyanobacteria/chloroplast; see above). This was particularly the case with the Bacteroidetes and the Cyanobacteria/chloroplast-specific primer pair (Figure 2). In the latter case, it may therefore be advisable to use the primer pair developed by Nübel et al. (1997) for DGGE analysis rather than the one used here, but to consider primers CYA361f/CYA785r for the preparation of clone libraries.

Figure 2
figure 2

Denaturing gradient gel electrophoresis (DGGE) analyses of 16S rRNA gene fragments amplified from DNA samples from the Northern Atlantic Gyre and the mesocosm experiments. Details on the primers used for the two- or three-step nested PCR approach are outlined in the text. The sequence identity of the numbered bands is given in the Supplementary data. The abbreviations indicate the target group of the corresponding primers: alpha, Alphaproteobacteria; beta, Betaproteobacteria; gamma, Gammaproteobacteria; Firm, Firmicutes; Plancto, Planctomycetes; Bacter, Bacteroidetes; Cyano, Cyanobacteria; Bac, Bacteria.

Full size image

The identification of the DGGE bands by sequence analysis showed that the PCR fragments mostly belonged to the bacterial group that the primers specifically targeted. An exception was the DGGE analysis of Betaproteobacteria. As with the results from clone library screening, a large fraction (60%) of the bands on the DGGE gel from the Northern Atlantic Gyre DNA sample were from prokaryotes other than the Betaproteobacteria. In contrast to this, all of the bands on the Firmicutes DGGE gel that were identified by sequence analysis derived from strains belonging to the target group (Supplementary Table S2). However, this may only be an artefact due to the sequence analysis of only a relatively small number of DGGE ‘bands’. Increasing this number may again reveal non-target 16S rRNA gene fragments.

This research has shown the clear benefits of using group-specific PCR primers. Much more information is obtainable on bacterial diversity than with general bacteria primers as it is possible to target specific bacterial groups for more detailed investigation of diversity, either by screening clone libraries or by using these primers in association with the well-tried methodology of DGGE; there is greater discrimination on individual gels and more bands are visible and can be sampled for sequencing. Although these primers have been designed for use in studies of the bacterial diversity in marine aquatic environments, they are also likely to be useful for the analysis of the bacterial diversity in other environments.