We have applied a simple methodology based on restriction endonuclease digestion of the total bacterial and archaeal 16S rDNA amplified from sea water samples to compare the prokaryotic diversity present along a transect from coastal to offshore waters in the Mediterranean basin (from Barcelona to the Balearic Islands). Samples from the surface and deep-chlorophyll maximum (DCM) were obtained. Temporal variation during a short span was also investigated by repeating the sampling after 48 h. The patterns of digestion bands from the samples were compared to detect major changes in the bacterial or archaeal groups present. For bacteria the main difference was found between the free-living community and that retained by the eukaryotic filter and assumed to be particle attached. By contrast Archaea collected by the same means did not appear as a separate group. The other main variation was depth-related, with remarkably different communities at the surface, at the DCM, and a single sample taken at 400 m depth. On the other hand, the variation along the transect from the continental platform to an offshore (2000 m maximum depth) station was relatively small. Temporal (48 h) and small scale spatial (a few km) variation was minimal.
Development in the late 1970s of reliable techniques to measure bacterial abundance and activity fostered a profound change in the understanding of planktonic food chains [1, 2]. Bacterioplankton is now known to account for a significant fraction of planktonic biomass  and to process around 50% of primary production  in many aquatic ecosystems. Phototropic picoplankton organisms are taken to account for a major portion of primary production in the ocean . Bacteria, however, have been studied as a ‘black box’, without any knowledge of the identities of the organisms involved. One of the main problems with bacterial identification is the difficulty in describing the bacterial species present in a single sample of sea water. Most of the cells are non-culturable by traditional means and their morphology differences are insufficient to discriminate among different prokaryotic species. In fact, it is not known how many species of bacteria are present in ecosystems.
The application of several molecular biological techniques to field samples is changing this situation [6–8]. These techniques avoid the need for isolation of pure cultures by focusing directly on the nucleic acids present in the environment. From the diversity of such nucleic acids, the biological (phylogenetic) diversity can be inferred. One set of techniques aims at amplifying a highly conserved (or sometimes group-specific) sequence by polymerase chain reaction (PCR), most frequently 16S rDNA, constructing a library of clones and sequencing such clones [9–17]. It has provided very interesting information about the presence of some clones in different systems, particularly when hybridization is used afterwards to check for the distribution of the organisms represented by such clones in natural populations [18, 19]. This approach, however, is very labour-intensive and time-consuming. Also, the number of samples/clones to be studied by sequencing is still normally very limited.
Simpler techniques which can be applied to a large number of samples in a study may be more effective to answer ecological questions about the diversity of picoplankton. Several molecular techniques exist to compare the diversity of different communities. For example, DNA can be studied directly without prior PCR amplification. DNA/DNA hybridization  and DNA reassociation  have been used to compare marine samples and to estimate the genetic diversity present in a soil sample, respectively. Low molecular mass RNA fingerprinting has also been used as a method to describe bacterial diversity in both freshwater [22, 23] and marine environments . The relevant 16S rDNA molecules may also be amplified by PCR prior to analysis. These molecules can then be separated by denaturing gradient gel electrophoresis [25, 26] or single strand conformation polymorphism  and sometimes the bands are retrieved from the gel, re-amplified and sequenced [28, 29]. In this work we have used PCR to amplify prokaryotic 16S rDNA, and the analysis of the amplified genes has been carried out by restriction enzyme digestion and electrophoresis of the fragments, a technique sometimes called amplified ribosomal DNA restriction analysis (ARDRA) [30–32]. The number and pattern of bands reflect the genetic diversity of the sample and patterns from different samples can be easily and quantitatively compared. We have already applied this technique successfully to study hypersaline waters. Its main advantage is that the community (targeted by the PCR primers) is studied as a whole and many samples can be compared.
Large areas of the world's oceans are characterized by the presence of a deep-chlorophyll maximum (DCM) [33–37]. In many cases the primary producers found at this DCM are different from those found at the surface [38–41]. The photosynthetic characteristics of these assemblages were different . The distribution of heterotrophic protists was also different . Carbon flux seems to be predominantly through the microbial food web at the DCM, while the classical path to crustacean zooplankton may be more important at the surface (unpublished results). Based on this evidence it seems reasonable to assume that two different communities exist, adapted to life in surface waters and at the DCM respectively. Our hypothesis, therefore, was that picoplankton assemblages inhabiting the two regions would also be different. We analyzed changes with depth, with distance from the shore, and during short time intervals (in the order of days) in the western Mediterranean Sea. In this area, from spring to autumn, a DCM is a permanent feature of the water column. From each sample we have separated two sub-assemblages of bacterioplankton, free-living and attached, and from every sample archaeal and bacterial 16S rDNAs were amplified independently.
2Materials and methods
2.1Study area and sampling
Three stations were sampled along a transect perpendicular to the coast off Barcelona (Spain) during 16–25 June 1995 (Fig. 1, Table 1). The work was done during the cruise FRONTS95 of the B/O Garcı́a del Cid. Temperature, salinity and fluorescence were determined with a CTD EG and G model MkIIIC. Water samples were collected with a 30 liter double Van Dorn bottle and dispensed into plastic carboys.
Table 1. Location of stations, depth, time and date of sampling, and abundance of bacteria in the different filtrates (see Fig. 1)
S1 (cells ml−1)
S2 (cells ml−1)
S3 (cells ml−1)
S4 (cells ml−1)
ND, not determined.
The filtration protocol is shown in Fig. 2. Sampled sea water was filtered through a 90 mm diameter Millipore AP20 glass fiber filter to remove larger particles and eukaryotes. Microorganisms were then collected by positive pressure filtration on 0.22 μm pore diameter filters (Durapore, Millipore, Bedford, MA, USA). Forty liters were filtered per sample. The filtrate from this second filtration was then used to rinse the AP20 filters. For this purpose, the filter was placed with the organisms-side down on top of a second AP20 filter and 10–20 l of the 0.22 μm filtered water passed through this ‘sandwich’. Finally, this second filtrate, presumably containing the bacteria attached to particles and cells and easily rinsed from them, was filtered on 0.22 μm membrane filters to collect them. Both 0.22 μm filters from each sample were placed in sterile Petri dishes and immediately frozen at −20°C during the cruise and at −80°C afterwards. Sub-samples were taken from the initial sample (S1 in Fig. 2 and Table 1), the ‘eukaryote-free’ sample (S2), the ‘bacteria-free’ sample (S3) and the ‘attached-bacteria’ sample (S4), to monitor the abundance of bacteria throughout the filtration process.
2.2Abundance of bacteria
Water was dispensed into 20 ml plastic scintillation vials and immediately fixed by addition of formalin to a final concentration of 2% formaldehyde and stored at 4°C in the dark. Aliquots of 10–20 ml were filtered through black stained Nuclepore filters (0.2 μm pore diameter). Samples were stained with DAPI (0.1 μg ml−1 final concentration) for 5 min before sucking the filters dry . Filters were then mounted on microscope slides with non-fluorescent oil (R.P. Cargille Lab.). Bacteria were counted by epifluorescence microscopy with a Nikon Diaphot microscope. Approximately 200–400 bacteria were counted per sample.
2.3DNA extraction and purification
DNA extraction and purification followed the protocol of Fuhrman et al. , with slight modifications. Thawed filters were cut into small pieces with a sterile razor blade and vortexed in 5 ml of TE buffer (50 mM Tris-HCl (pH 8); 20 mM EDTA), in 50 ml polyallomer centrifuge tubes. Acid-washed glass beads (≤106 μm) (Sigma) were added and vortex mixed until resuspension of the sample. The sample was incubated at 37°C for 1 h after addition of 50 μl of lysozyme (Boehringer) in TE buffer (1% w/v). A 0.1 volume of 10% sodium dodecyl sulfate (SDS) was added drop-wise while swirling. The tubes were placed into a boiling water bath for 2 min. Cellular debris suspended in the liquid was pelleted by centrifuging at 10 000 rpm, for 10 min at 12–15°C. The supernatant fluid was poured into a 50 ml centrifuge tube, and 3 volumes of absolute ethanol plus 1.5 ml of 10.5 M ammonium acetate were added to precipitate the DNA. Then the tubes were kept overnight at −20°C. After centrifugation (20 min, 14 000 rpm, 4°C), the supernatant was poured off and the pellet was dried and suspended in 0.5 ml of Milli-Q purified water. The liquid was transferred into a 1.5 ml centrifuge tube (Eppendorf) and the DNA was extracted once with a volume of Tris-saturated phenol (pH 8) and a volume of water-saturated chloroform. The DNA in the final aqueous extract was precipitated with 1 ml of ice-cold ethanol plus 150 μl of 10.5 M ammonium acetate for at least 1 h at −20°C. The DNA was pelleted (20 min, 14 000 rpm, 4°C), rinsed with 70% (v/v) ethanol and dried under vacuum (Speed-Vac) and then gently suspended in 0.1 ml of Milli-Q purified water. Quantification and purity (referred to protein content) of the DNA obtained were estimated (GeneQuant, Pharmacia). 5 μl of DNA at approximately 0.1 μg μl−1 was electrophoresed on a 1% (w/v) agarose gel electrophoresis to check for DNA integrity.
2.4PCR amplification of 16S rDNA
Reaction mixtures contained 50 mM KCl, 10 mM Tris-HCl (pH 9), 1.5 mM MgCl2, 0.1% Triton X-100, 200 mM of each deoxyribonucleotide triphosphate (dATP, dCTP, dGTP, dTTP; Pharmacia LKB Biotechnology), 2 U of TaqI DNA polymerase (Promega), 0.2 mM of each oligonucleotide primer and 100 ng of template DNA in a total volume of 50 μl. The sequences of the forward primers were 5′-TTCCGGTTGATCCTGCCGGA-3′, nucleotide positions 2–21 of the Halobacterium salinarium numbering system  for the Archaeal Domain, and 5′-AGAGTTTGATCATGGCTCAG-3′ (ANT-1), slightly modified (changing an M to an A at position 12) from Lane , at positions 8–27 of the E. coli numbering  for the Bacterial Domain. As reverse primers, for both Domains the oligonucleotide 5′-GGTTACCTTGTTACGACTT-3′ (S) was used (positions 1509–1491, E. coli numbering), obtained from Lane , changing a Y for a T at nucleotide 3. All primers were subjected to CHECK PROBE SSU Prok (RDP)  to confirm their adequacy. Reaction mixtures were overlaid with mineral oil (Light White Oil; Sigma) prior to a thermal cycling regime (35 cycles on a Perkin Elmer Cetus 480) of 94°C for 1 min, 55°C for 1 min, and 72°C for 2 min. After the final cycle was completed, the amplicons were allowed to extend at 72°C for 10 min. Negative controls were included with no addition of template DNA. 5 μl of the PCR products was analyzed on 1% (w/v) agarose gels to check for purity. PCR products were precipitated, dried and resuspended in 25 μl of Milli-Q purified water.
2.5Restriction endonuclease digestions
Enzymatic digestions were performed by incubating 15–20 μl of the amplified products with 5 U of each endonuclease and the corresponding enzyme buffer. Double digestions were performed for every sample with the combinations AluI+RsaI and MboI+HinfI (Boehringer). Digestions were carried out at 37°C overnight, to minimize partial digestion. Digested products were precipitated with ethanol, dried and resuspended in 3 μl of dye loading buffer.
The digested products were heated for 2 min at 96°C and chilled on ice before loading on 55 cm wedge-shaped (0.2–0.6 mm) 6% (w/v) polyacrylamide denaturing (7 M urea)-TBE gels. Electrophoresis was performed in TBE buffer at 55°C using a LKB Macrophor 2010 sequencing unit operated at 50 W per gel. The bromophenol blue was allowed to migrate to the bottom; the gel was washed in 10% glacial acetic acid for 20 min, and stained with the Silver Sequence™ Kit (Promega) following the manufacturer's protocol.
Pairwise comparisons of the band patterns obtained were manually performed, and a matrix (presence-absence for each digestion band) constructed. Some samples were amplified by PCR twice to check the consistency of the amplification step. At least two separate digestions (and the corresponding electrophoresis) were carried out for each combination of enzymes. Only the reproducible bands were considered in the numerical analysis. Comparisons were always made among samples run in the same gel (maximum 36 lanes). The data were computed using the NTSYS-pc program version 1.80 (Exeter Software) on an IBM PC. The procedure is to first run the SAHN program (UPGMA) to cluster the data and then the COPH program to compute a cophenetic value matrix using the tree matrix produced by SAHN. The MXCOMP program can then be used to compare the cophenetic value matrix with the original matrix that was clustered. The ‘cophenetic correlation’ (r) was used as a measure of goodness of fit for a cluster analysis.
Typical vertical profiles of temperature, salinity and fluorescence at three stations were examined in the present work (Fig. 3). Station C (coastal station) had a depth of 70 m and the water column was slightly stratified. There was a DCM at 40 m. A continental slope front separates coastal from oceanic waters. Station S (slope station) was placed above the continental slope, on 970 m of water. From the salinity and temperature data, this station appeared to be on the offshore side of the slope front. Finally, station D (deep station) was located in the middle of the western Mediterranean basin, above 2000 m of water. Both stations S and D had a sharp DCM at about 55 m. The profiles in Fig. 3 correspond to the first sampling at each station. Profiles for the second sampling were identical to those obtained at the first sampling.
Vertical distribution of bacteria in the three stations was determined (Fig. 4). The average values (and standard errors) of two diel cycles are presented for each station to indicate the range of variation that can be expected between samplings. Bacteria were distributed uniformly with depth at station C. At the other stations bacterial abundance was intermediate in the mixed layer (2–4×105 cells ml−1), slightly higher at the DCM and much lower in deeper waters (around 1×105 cells ml−1).
The fractionation procedure used was monitored by following the abundance of bacteria in the different filtrates (Table 1). The first filtration through AP20 filters did not cause a significant reduction in the total number of bacteria. This means that the ‘attached bacterial’ fraction (and thus, retained in the filter) was smaller than the error of the method. In fact, the number of ‘attached bacteria’ recovered after rinsing the filters was usually lower than 10% of the initial number. The sample CDCM2 gave anomalously high numbers in both the 0.2 μm filtrate and the rinsed AP20 sample.
The second filtration through Millipore 0.2 μm reduced the bacterial numbers in the filtrate by almost two orders of magnitude. Therefore, we recovered close to 99% of the prokaryotic cells present in the samples for the PCR analysis. Finally, the rinsing procedure recovered about 10% of the total cell count, which had been retained in the AP20 filters.
DNA was extracted from all the samples except for the two surface coastal samples CSUR1 and CSUR2 (both free-living and attached communities) that were accidentally defrosted. Bacterial 16S rDNA could be amplified from all the samples. With the primers for Archaea, only the DCM and deep samples gave amplification products. The amplicons were first run on a standard agarose gel always obtaining a single band of the expected size. The number of bands obtained from the combination of both double digestions varied widely with an average value of 124 bands for free-living bacteria and 57 for attached bacteria. The number of bands can be considered a gross estimation of total biodiversity of the group under consideration. The application of this assumption would give increasing diversity in the order attached bacteria<archaea<free-living bacteria.
As in a previous work, in which samples from hypersaline environments were analyzed with the same methodology , the stability of the clusters was remarkable regardless of the enzyme combination used. The trees generated had almost identical topology (data not shown), strengthening the view that they really reflect the similarity of the prokaryotic assemblages present. Therefore, the dendrograms shown in Fig. 6 represent the addition of the similarity matrix generated by both enzyme combinations (AluI+RsaI and HinfI+MboI). The cophenetic value was r=0.92539 for bacteria and r=0.95082 for archaea which is considered a very good fit. The first thing that the dendrogram of bacteria shows up clearly is the relative homogeneity in time. Repeated samples, although taken 48 h apart, grouped together very closely. All the samples with more than 80% coincidence in their band patterns were repetitions of the same samples after 48 h. This similarity is remarkable considering that although the station was accurately pinpointed by GPS, due to the changing sea conditions, drifting could cause a significant change in the actual water mass present at the sampling site. The similarity of the results indicates that the bacterial assemblage was stable in the time span considered and also in the scale of at least a few km.
The largest differences appeared between the free-living and the ‘attached’ assemblages. Again, this pattern was essentially the same regardless of the combination of restriction enzymes used (data not shown). The bacterial biomass obtained by washing the eukaryotic filter showed very little similarity to the cells directly collected in the 0.2 μm filter even from the same sample. In fact, most of the attached populations formed a cluster with low within-cluster similarity (Fig. 5A, Fig. 6A). Two of the attached populations, CDCM2(ATT) and DDCM2(ATT), clustered separately at a very low similarity with the free-living 400 m sample. The high number of bands obtained for those two samples could indicate an artifactual origin for this similarity, i.e., in these two samples part of the free-living biomass has been collected as attached. In addition to these two main groups, both sets of enzymes grouped the DCM samples together on the one hand and the surface (5 m deep) samples on the other hand. Among the samples of free-living bacteria the most distinct assemblage was that of the single 400 m deep sample analyzed.
Archaeal 16S rDNA could be amplified from the deep samples (DCM and 400 m) only. Again, the similarity of the two samples from the same station and depth was the highest (Fig. 5B, Fig. 6B). The C and S samples were somewhat more alike than the offshore D samples and the 400 m depth sample was again the most different. Only two samples from ‘attached cells’ yielded amplified DNA with the Archaeal primers, and they did not group together. The number of samples, however, is too small to consider the result significant. The number of bands per digestion, with an average value of 75, was significantly smaller than in the equivalent samples corresponding to bacteria revealing a lower diversity of Archaea.
The first point in the present work is the relative homogeneity of the picoplankton at the time and space scales analyzed. Apparently the picoplankton diversity present in the marine ecosystem studied was relatively constant at the surface and at the DCM, within the scale of a few km and days. Even at the large scale (offshore vs. coastal waters) the assemblages from the same depths showed high similarities. In a previous study comparing the populations from the different ponds of a solar saltern with different salinities, the assemblages were substantially more variable . For example, ponds with marine salinity showed only 30% similarity to the ponds with the highest salinities. And even contiguous ponds, with very similar salinities, showed only 50–70% similarity. Repeated DNA extraction, amplification and digestion from the sample always gave similarities over 95% but, more often than not, less than 100% (Acinas and Rodriguez-Valera, unpublished results). Thus, the apparent uniformity in the Mediterranean samples is not due to poor discrimination of the technique. We conclude that picoplankton assemblages were fairly similar horizontally at the scales from a few to hundred kilometers in space and a few days in time.
Naturally, this finding does not imply homogeneity at the microscale level. In fact, our own data indicate that loosely bound to the eukaryotic cells, or forming aggregates, there is a prokaryotic assemblage substantially different from the free-living assemblage. In fact, it is widely accepted that attached bacteria represent a special assemblage in marine bacterioplankton  and, although they are relatively few in the open ocean (compared to free-living cells), could have an important role in carbon cycling [52, 53]. The fact that these samples cluster together in the case of bacteria, but do not do so with archaea supports our interpretation that they actually represent a different population from that collected after the first filtering operation (free-living). At any rate, it appears that life attached to particles requires very specific adaptations  and, therefore, it was to be expected that different phylogenetic groups would be found. DeLong et al.  reached a similar conclusion studying clones amplified from marine snow aggregates and clones amplified from the water surrounding the aggregates.
The samples from the DCM formed a cluster with high within-cluster similarity. It is interesting to note that DCM assemblage from the slope (S) station resembled the offshore DCM assemblages (about 60 km apart) more than its own surface assemblage (only 40 m away). The single 400 m sample also had a markedly different diversity profile from those of the DCM and surface samples, indicating a strong discontinuity between the relatively shallower depths (over 100 m) and the deep water mass. This pattern was already observed by a different molecular technique . It appears clear that vertical stratification is a very important factor determining the prokaryotic species living in the ocean. This conclusion is further supported by the results obtained with the Archaeal primers. Representatives of the Archaeal domain could be amplified only from deep waters (DCM and 400 m), reflecting the very different environmental conditions found at the surface and at deep waters. The predominant presence of Archaea in deep waters has already been reported in the Pacific and Atlantic oceans [11, 19, 45, 55] and a shift from euryarchaeota-dominated archaea in the surface to a crearchaeota-dominated population at deeper waters has been described .
Similar results with respect to short time and space scales were obtained by Lee and Fuhrman  in an oligotrophic area of the Pacific Ocean, using DNA/DNA hybridization. This technique is very different from ours and thus results must be compared with caution. In the DNA hybridization technique the whole DNA of the organisms in the sample is compared. Therefore, the resultant similarities are probably a combination of the different species present plus the relative abundance of each one. With the technique used in the present work, only the 16S rDNA genes from each taxon are compared. Thus, our similarities are a consequence of the different species present, but only to a very limited extent of their relative abundance. However, both approaches aim at revealing the genetic similarities (or lack of them) among picoplankton assemblages in a relatively quick and straightforward manner.
Lee and Fuhrman  compared samples collected at 25, 100, 500 and 1000 m, from different stations and on different (but consecutive) dates. A DCM was also present at about 100 m. Photic zone samples were very different from deep samples. Similarity within depths was always higher than similarity between depths, and the 25 m samples were different from the DCM samples. The presence of the same pattern in such widely distant ocean basins suggests that it is a major characteristic of bacterial diversity in oligotrophic marine areas. Murray et al.  have found, by DGGE analysis of the PCR-amplified 16S rDNA, that the bacterial population in the surface water of two California bays (San Francisco and Tomales) was relatively stable with time and space within one bay. However, wide differences were found between the two bays that represent very different situations regarding the quality and quantity of the surrounding land input.
By studying the western Mediterranean we had the advantage of the existing data set on phytoplankton species composition [38, 56, 57]. Previously, no other study of bacterioplankton diversity has been able to compare the results to those of other components of the plankton. The similarities found among the bacterial assemblages studied here are consistent with what has been described about the phytoplankton of the area [38, 39, 56]. Estrada et al.  summarized the information on phytoplankton species composition in the following manner. Phytoplankton species could be grouped into three assemblages. Group A occupied the upper euphotic zone during stratification and consisted mostly of dinoflagellates and small flagellates. Typical species were Ceratium furca and Torodinium robustum. Group B was found in the DCM layer, more or less uniformly distributed throughout it. Some typical species were Coscinodiscus radiatus and Oxitosum margalefi. Finally, group C was composed of large diatoms and formed patches of high chlorophyll and cell abundance within the DCM. At the beginning of stratification, the phytoplankton assemblages are different on each side of the continental slope front. But, as stratification proceeds, a layer of warm water extends over the front and group A becomes established on both sides, from the surface to about the DCM. Group B forms a kind of permanent background at the DCM on the offshore side of the continental slope front, but it is absent from the inshore side of the front. Group C, finally, forms patches throughout the DCM. When these diatoms are present, they form most of the phytoplankton biomass.
We can interpret our results in the light of the phytoplankton studies. The similarities between all our surface samples would suggest a picoplankton assemblage specifically adapted to live with group A phytoplankton, and would appear to include mostly bacteria, not archaea. Clearly, it is very different from the assemblage found at the DCM. The fact that the DCM samples from the coastal station were not very different from those of the other two stations indicates that this assemblage is probably associated with phytoplankton group C, i.e., with the patches of large diatoms that can be found on both sides of the continental slope front, but not with the group B species which are only abundant offshore of the front. This picoplankton assemblage included archaea. Finally, the 400 m deep assemblage, including both bacteria and archaea, is obviously adapted to completely different conditions from those prevailing in the photic zone and is the most different one.
The differences in species composition of phytoplankton reflect the variations in environmental conditions. In turn, the picoplankton reflects such variations in its most likely food source: the phytoplankton.
This work was supported by CICYT Grants BIO 93/0750 and AMB94/0853, DGICYT Grants PB93/0930 and PB91/0153, and Grant EV5V-CT92-0080 from the European Commission. We are grateful to Jordi Salat and the captain and crew of the B/O Garcı́a del Cid for invaluable help during the cruise. We thank Magalı́ Brunel for doing the bacterial counts. Secretarial assistance by K. Hernández is gratefully acknowledged.