Comparison of Bacterioneuston and Bacterioplankton Dynamics during a Phytoplankton Bloom in a Fjord Mesocosm

Abstract

The bacterioneuston is the community of Bacteria present in surface microlayers, the thin surface film that forms the interface between aquatic environments and the atmosphere. In this study we compared bacterial cell abundances and bacterial community structures of the bacterioneuston and the bacterioplankton (from the subsurface water column) during a phytoplankton bloom mesocosm experiment. Bacterial cell abundance, determined by flow cytometry, followed a typical bacterioplankton response to a phytoplankton bloom, with Synechococcus and high-nucleic acid content (HNA) bacterial cell numbers initially falling, probably due to selective protist grazing. Subsequently HNA and low-nucleic acid content bacterial cells increased in abundance, but Synechococcus cells did not. There was no significant difference between bacterioneuston and bacterioplankton cell abundances during the experiment. Conversely, distinct and consistent differences between the bacterioneuston and the bacterioplankton community structures were observed. This was monitored simultaneously by Bacteria 16S rRNA gene terminal restriction fragment length polymorphism and denaturing gradient gel electrophoresis. The conserved patterns of community structure observed in all of the mesocosms indicate that the bacterioneuston is distinctive and nonrandom.

Determining and understanding both spatial and temporal patterns in bacterioplankton community structure are a core aim of marine microbial ecology (15). Distributions of bacterioplankton over space and time can be correlated to environmental parameters, and subsequent links can therefore be made to ecosystem function. A broad range of spatial studies made on macro- (34), meso- (20), and microscales (27) have shown clear patterns in distribution of the bacterioplankton.

The sea surface microlayer is part of the air-sea interface and is generally considered to be the top 1 mm or less of the ocean (26). Surface microlayers have a fundamental role in regulating transport processes between the ocean and the atmosphere (26) and are often referred to as the neuston (28, 31). For more than 25 years it has been hypothesized that the sea surface microlayer is a hydrated gelatinous layer (40) that contains surface-active organic compounds such as carbohydrates, proteins, lipids, and humic substances in relatively high concentrations (17, 45, 48). Recently, gel-like transparent expolymer particles (TEP) have been shown to be enriched in the surface microlayer, supporting the concept of a gelatinous interfacial layer (46).

Bacteria present in surface microlayers or the neuston are regarded as the bacterioneuston. There are relatively few studies which have directly compared the community structure of the bacterioneuston with that of the cognate subsurface (bacterioplankton) in the marine environment. Analysis of Bacteria 16S rRNA gene clone libraries constructed using DNA isolated from surface microlayer and subsurface water (<1 m) samples from the North Sea revealed that the bacterioneuston was dominated by two operational taxonomic units which accounted for 81% of clones analyzed (13). Community structure profiling using denaturing gradient gel electrophoresis (DGGE) of the bacterioneuston at three sites around Oahu Island in the Pacific Ocean showed that the bacterioneuston forms consistent and distinct community structures. Conversely, the archaeal community structure of the same samples using Archaea 16S rRNA gene DGGE analysis did not show the same surface microlayer-specific response, indicating that bacteria and archaea respond to their environment in fundamentally different ways in the neuston (7).

Other studies have, however, reported no consistent differences between the bacterioneuston and the bacterioplankton. Samples collected from two separate sites in the Mediterranean Sea were analyzed using single-strand conformation polymorphism of Bacteria 16S rRNA genes (1). The authors did not report any significant differences between the surface microlayer and subsurface samples using this community profiling method.

Nonmarine studies of the bacterioneuston and Archaea communities in estuarine (10) and freshwater (5, 19) environments have also shown distinct microbial community structures present in the surface microlayer compared to those in subsurface water ≤1 m below.

Recurring phytoplankton blooms are a key feature of coastal waters and strongly influence bacterioplankton community structure and succession (4, 14, 38). Phytoplankton blooms stimulate the bacterioplankton by the release of dissolved organic matter (22) or affect bacterioplankton negatively by direct competition for resources (6). Bacterioplankton community structure may also be influenced by grazing flagellates or viral lysis (47).

Mesocosm experiments have been used to study plankton ecology for many decades (33). Mesocosms facilitate study of the effects of key environmental parameters, such as temperature, on plankton communities and allow the succession of natural plankton communities that resemble those found in the marine environment (11). The enclosed water mass means that experiments can be designed which manipulate physicochemical parameters to observe biological effects. Furthermore, with replicated mesocosms, the data collected can be analyzed with statistics rigorously. In this study we monitored the dynamics of the bacterioneuston and the bacterioplankton in mesocosms of fjord surface water during an artificially induced phytoplankton bloom and compared bacterial abundances and bacterial community structures in the surface microlayer and subsurface water.

MATERIALS AND METHODS

Mesocosm setup and sampling.

The experiment was carried out at the Marine Biological Field Station, Espeland, Norway (20 km south of Bergen, Norway) from 21 May 2008 to 1 June 2008. Twelve land-based mesocosms (1.5-m diameter and 1.5-m deep) were each filled (2,474 liters) with prefiltered (∼300 μm) water from the Raunefjorden. The water in the mesocosms was constantly mixed by means of submerged aquarium pumps. The mesocosms were contained in three larger open containers (Fig. 1A) that were filled and circulated constantly with pumped fjord water to maintain the mesocosms at ambient fjord temperature. The 12 mesocosms were divided into two treatment groups, control and nutrient amended, allowing six replicate mesocosms for each treatment. Each of the larger containers held two control mesocosms and two nutrient-amended mesocosms (Fig. 1B). Addition of nitrate and phosphate according to the Redfield stoichiometry (N/P = 16:1) (35), as 16 μM NaNO₃ and 1 μM KH₂PO₄, was used to induce the phytoplankton bloom at 2100 h on day zero.

FIG. 1.

(A) Photograph showing the mesocosms used in this study. Twelve mesocosms were divided into three larger containers. (B) Each mesocosm was filled sequentially, from A to L. Control mesocosms were A, C, E, G, I, and K. The phytoplankton bloom was induced in nutrient-amended mesocosms B, D, F, H, J, and L.

Sampling took place every day for 11 days at 0900 h. Subsurface waters were sampled from a depth of 0.75 m in the center of the mesocosms using a siphon. The surface microlayer was sampled using two different methods, a mesh screen (Garrett screen) and polycarbonate membranes taken from the center of the mesocosms. The methods sample two different depths: the mesh screen removes the top ∼400 μm, and the polycarbonate membrane removes the top ∼40 μm of the surface microlayer (7). The mesh screen (16-mesh stainless steel screen; size, 275 by 275 mm) was placed below the surface water and lifted horizontally through the surface microlayer, and the water was collected into a sterile bottle. A total of 250 ml was then filtered using a peristaltic pump through a Sterivex-GS filter unit (pore size, 0.2 μm; Millipore). After all the water had been evacuated from the filter unit, 1.6 ml of RNAlater (Ambion) was added, and the filter unit was stored at 4°C. Polycarbonate membranes (47-mm diameter; pore size, 0.2 μm; Isopore; Millipore) were placed onto the water surface using forceps and left for 10 s before being removed and stored in 2-ml screw-cap tubes at −20°C.

Dissolved inorganic nutrients.

Subsurface water samples were filtered (Sterivex-GS; pore size, 0.2 μm; Millipore) before being stored in polyethylene vials at −20°C until nitrate, nitrite, phosphate, and silicate quantities were determined using standard segmented flow analysis with photometric procedures (18).

Phytoplankton and bacterial cell counts.

Phytoplankton and bacterial cells in the mesocosms were enumerated with a Becton Dickinson FACSCalibur benchtop flow cytometer (BD Bioscience) equipped with a 488-nm laser line. Cells were enumerated in samples collected from the subsurface and mesh screens only since membrane-collected samples do not remove enough water for flow cytometry analysis. Two analyses were performed per sample to determine both phytoplankton and bacterial cell counts. Briefly, phytoplankton (picoeukaryotes, coccolithophorids, and small and large nanoplankton) and Synechococcus cell counts were enumerated on fresh unstained samples using modified flow rates (ca. 100 μl min⁻¹) and pre- and postaspiration sample weighing together with timed acquisition (5 min) (42). Bacterial cell counts (total count and subsets for high-nucleic acid content [HNA] and low-nucleic acid content [LNA] bacterial cells) were determined on paraformaldehyde-fixed, citrate-treated samples stained with SYBR Green I (Invitrogen) using timed acquisition (2 min) in concert with pre- and postaspiration weighing (50). For pre- and postaspiration weighing, all samples were weighed before and after analysis to determine sample volumes aspirated during the sample analysis, and internal 0.49-μm reference beads were used to account for flow and machine drift. All analyzed samples were exported as list mode files and analyzed using Cyflogic to gate major populations and calculate absolute cell concentrations from aspirated volumes.

Extraction of DNA for bacterial community structure analysis.

DNA was extracted from subsurface, mesh screen, and membrane samples collected on day 2, day 5, and day 10. DNA was extracted from three control mesocosms (replicates A, E, and K) and three nutrient-amended mesocosms (replicates B, F, and L) (Fig. 1B). DNA was extracted in a sucrose buffer using lysozyme, proteinase K, sodium dodecyl sulfate, and phenol-chloroform as described by Cunliffe et al. (10). The resuspended DNA was quantified using a spectrophotometer (ND-1000; NanoDrop) before all DNA samples were diluted to a concentration of 30 ng·μl⁻¹ and stored at −20°C.

Bacterial community structure analysis.

PCR amplification of Bacteria 16S rRNA genes for terminal restriction fragment length polymorphism (T-RFLP) analysis was performed using the fluorescently labeled primer, (6FAM)27F (5′-AGA GTT TGA TCM TGG CTC AG-3′; 6FAM is 6-carboxyfluorescein), and primer 536R (5′-GWA TTA CCG CGG CKG CTG-3′) (41). For PCR, a total volume of 50 μl contained 0.5 mM each of the deoxynucleoside triphosphates, 0.5 μM of each primer, 2 units of Taq DNA polymerase (Sigma), and 30 ng of template DNA. The PCR program consisted of initial denaturation at 94°C for 2 min, followed by 30 cycles of 94°C for 1 min, annealing at 52°C for 1 min, and elongation at 72°C for 3 min, with a final elongation step at 72°C for 10 min. PCR products were verified by agarose gel electrophoresis and stored at −20°C.

PCR products were purified using a QIAquick PCR Purification Kit (Qiagen) according to the manufacturer's instructions. A total of 20 μl of purified PCR product was digested for 4 h at 37°C using the restriction enzyme MspI (Promega). The digestion product (0.5 μl) was combined with 0.5 μl of denatured LIZ600 size standard (Applied Biosystems) and formamide before being run on a 3730 DNA sequencer (Applied Biosystems). The sizes of the terminal restriction fragments (T-RFs) were calculated and binned using Genemarker (Softgenetics). Bin widths were checked and manually adjusted to encompass all concordant peaks. To differentiate signal from background, a fluorescence unit threshold of 40 units was used to determine which T-RFs to include. Relative abundance was calculated for each T-RF by dividing individual T-RF fluorescence by total sample fluorescence.

PCR amplification of 16S rRNA genes from Bacteria for DGGE analysis was performed using primers 341F (5′-CCT ACG GGA GGC AGC AG-3′) and primer 518R (5′-ATT ACC GCG GCT GCT GG-3′) (30). The same PCR was set up as before for T-RFLP but using the different primers. The PCR program for DGGE consisted of an initial denaturation at 94°C for 5 min, followed by 35 cycles of 95°C for 1 min, annealing at 65 to 55°C for 20 cycles (reduction of −0.5°C per cycle) and at 55°C for 15 cycles, elongation at 72°C for 1 min, and then a final elongation step at 72°C for 10 min.

DGGE was performed with a DCode system (Bio-Rad). Gels were prepared with 10% (vol/vol) acrylamide-bisacrylamide with a 30 to 70% linear denaturant gradient (100% denaturant solution contains 6.9 M urea and 11.5 M formamide). The gel was run in 1× Tris-acetate-EDTA buffer at 60°C for a total of 1,008 volt-hours (constant voltage of 63 V for 16 h). Gels were stained with SYBR Gold nucleic acid stain (Invitrogen) before the image was captured on a UV transilluminator (Syngene).

DGGE bands that were relatively more abundant in the surface microlayer samples were selected and excised. The excised bands were washed in sterile molecular-grade water before being crushed in 20 μl of molecular-grade water and incubated at 4°C for 2 h. The eluted DNA was used to reamplify the DGGE band using the same PCR primers and conditions as before. DGGE band DNA sequences were obtained using the University of Warwick Molecular Biology Services Laboratory.

Statistical and ordination analysis.

Analysis of variance was used to identify statistical significance in the phytoplankton and bacterial cell count data (n = 6; P < 0.05). Where significant differences were seen, a Tukey test was used to compare data within a defined set. Both analysis of variance and Tukey's test were performed using SPSS statistical software (SPSS). Principal component analysis (PCA) was used to visualize the relationships between bacterial community structures from the T-RFLP data and was carried out using MINITAB statistical software (Minitab). PCA is used to reduce the complexity of multivariant data (T-RF relative abundance) by producing new variables that account for most of the variation in the original data (39). DGGE profiles of 16S rRNA genes from Bacteria were compared using GelCompare II (Applied Maths) by calculating similarity coefficients using a curve-based Pearson correlation, followed by the construction of unweighted pair group method with arithmetic mean dendrograms from the calculated similarity coefficients.

Nucleotide sequence accession numbers.

DGGE band DNA sequences determined in this study were deposited in the GenBank under accession numbers GQ902042 to GQ902046.

RESULTS

Phytoplankton abundance.

The phytoplankton bloom succession in the mesocosms progressed generally as expected based on previous experience from earlier experiments with water collected from Raunefjorden (6, 29). The nitrate and phosphate added to the nutrient-amended mesocosms were steadily depleted, and levels returned to background concentrations by day 9 (Fig. 2). The concentration of silicate remained constant throughout the experiment. Nitrite increased in the nutrient-amended mesocosms to 0.19 ± 0.01 μM at day 5 before returning to background levels by day 10 (Fig. 2).

FIG. 2.

Dissolved inorganic nutrient concentration changes in control (□) and nutrient-amended mesocosms (▪). Mean values were plotted (n = 6), with the error bars representing the standard errors.

Phytoplankton cells were divided into four groups by flow cytometry analysis: picoeukaryotes, large nanoplankton, small nanoplankton, and coccolithophorids (see Materials and Methods). Picoeukaryote numbers increased in both control and nutrient-amended mesocosms at the start of the experiment (Fig. 3). By day 5 a significant increase in picoeukaryote numbers was detected in the nutrient-amended mesocosms compared to control mesocosms. The artificially induced picoeukaryote bloom peaked on day 7 with a median cell density of ∼2 × 10⁵ cells·ml⁻¹. There was no detectable significant difference between picoeukaryote cell counts in the surface microlayer and their cognate subsurface water samples.

FIG. 3.

Changes in abundances of phytoplankton and bacterial cells in the surface microlayer (▴) and subsurface water (▪). The surface microlayer was sampled using a mesh screen. Open symbols, control mesocosm samples; filled symbols, nutrient-amended mesocosm samples. Mean values are plotted (n = 6), with the error bars representing the standard errors.

Phytoplankton cells designated as large nanoplankton showed a significant increase in numbers in the nutrient-amended mesocosms from day 5 onwards (Fig. 3). As with picoeukaryotes, there was no significant difference between numbers in the surface microlayer and subsurface water.

Small nanoplankton showed more variable cell counts during the time of the experiment than picoeukaryotes and large nanoplankton (Fig. 3). After day 6, a significant difference was detected between the counts in the nutrient-amended mesocosms and cell counts in control mesocosms. The bloom of small nanoplankton peaked on day 7 before returning to cell numbers similar to those in the control mesocosms by day 9.

As with the small nanoplankton, coccolithophorid abundance appeared stochastic in contrast to the picoeukaryotes and large nanoplankton cell counts and had no distinct trend. The intravariation between mesocosms was high for coccolithophorid counts, and this subsequently affected statistical analysis. At day 7 there was a significant difference between cell counts in the subsurface samples from the control and nutrient-amended mesocosms. For the remainder of the experiment the coccolithophorid counts were significantly higher in the nutrient-amended mesocosms. There was also some indication of weak enrichment of coccolithophorids in the surface microlayer (Fig. 3).

Bacterial abundance.

Flow cytometry was used to separate three bacterial cell groups: HNA bacterial cells, LNA bacterial cells, and Synechococcus cells. The dynamics of the three groups was different during the experiment (Fig. 3).

HNA bacterial cells showed a marked decrease in abundance at the start of the experiment, with the rate of decrease accelerating rapidly on day 3. On day 5 the HNA bacterial cell numbers had dropped from an initial ∼6 × 10⁵ cells·ml⁻¹ to ∼1 × 10⁵ cells·ml⁻¹. After day 5 the abundance of HNA bacterial cells began to increase in all mesocosms, and a significant difference between HNA bacterial cell counts in the nutrient-amended mesocosms and counts in the control mesocosms for the remainder of the experiment was detected (Fig. 3). At the end of the experiment HNA bacterial cell numbers reached similar levels to those at the start of the experiment. There was no significant difference in HNA bacterial cell abundances between surface microlayer and subsurface water samples.

Unlike the HNA bacterial cells, LNA bacterial cells did not show a drastic drop in abundance (Fig. 3). LNA bacterial cell abundance fluctuated from day 0 to day 8 with no overall pattern. At day 2 and day 3 there was a significant difference between subsurface and surface microlayer LNA bacterial cell abundances, with fewer cells in the surface microlayer sample. LNA bacterial cell abundance fluctuated until day 9, when there was a significant increase in the nutrient-amended mesocosms, peaking at ∼7 × 10⁵ cells·ml⁻¹.

As with the HNA bacterial cell abundance, Synechococcus cell abundance declined at two rates at the start of the experiment. Initially, cell abundance dropped slowly up to day 3 and then rapidly down to ∼4 × 10³ cells·ml⁻¹ on day 6 (Fig. 3). Unlike HNA bacterial cell numbers, Synechococcus cell abundance did not recover and remained low for the remainder of the experiment. There were no significant differences in abundances of Synechococcus cells between treatments or between surface microlayer and subsurface water.

Bacterial community structure.

We used two Bacteria 16S rRNA gene profiling methods (T-RFLP and DGGE) to monitor changes in the bacterial community structures in surface microlayer and subsurface water samples collected on day 2, day 5, and day 10.

PCA ordination of the structures of the bacterial communities from T-RFLP analysis of subsurface and surface microlayer DNA samples is shown in Fig. 4. On day 2, the samples collected from the subsurface and from the surface microlayer using the mesh screen clustered closely together relative to the surface microlayer samples collected using polycarbonate membranes. As the mesocosm blooms progressed, this pattern changed drastically. At day 5, samples from the subsurface showed a distinct cluster that was separate from the mesh screen samples. As with day 2, the membrane-collected surface microlayer samples remained distinct from the subsurface samples. Near the end of the experiment on day 10, bacterial community structure in the samples collected with the mesh screen clustered with the samples collected with membranes and not subsurface water samples. Ordinance analysis of the T-RFLP data in this experiment showed no evidence of bacterial community structural differences as a result of the induced phytoplankton bloom (Fig. 4).

FIG. 4.

Ordination diagram from PCA of bacterial T-RFLP profiles. Samples were collected on day 2 (black), day 5 (dark grey), and day 10 (light grey). Subsurface water (▪) was collected using a siphon, and the surface microlayer was sampled using two methods: a mesh screen (▴) and polycarbonate membranes (•). Open symbols, control mesocosm samples; filled symbols, nutrient-amended mesocosm samples.

DGGE analysis of the bacterial community structures showed similar results to those of the T-RFLP analysis. At day 2, subsurface and mesh screen-collected samples were similar, and membrane-collected samples showed some differences (Fig. 5). This was less pronounced with DGGE than with T-RFLP at day 2. By day 5, the membrane-collected samples were distinctly different from mesh screen and subsurface samples, forming a separate clade in the dendrogram. Also at day 5, some mesh screen collected-samples were different from their associated subsurface samples. By day 10, both the membrane- and mesh screen-collected samples were distinctly different from the subsurface samples, corroborating the results from the T-RFLP analysis. As with the T-RFLP analysis, DGGE analysis confirmed that the bacterial community structures were not affected by the phytoplankton bloom.

FIG. 5.

Bacterial 16S rRNA gene DGGE profiles from day 2, day 5, and day 10. DGGE profiles show each replicate from the subsurface water (SS) and from the surface microlayer sampled using a mesh screen (MS) and polycarbonate membranes (PC). Beside each DGGE profile is the associated unweighted pair group method with arithmetic mean dendrogram showing the similarity of the lanes in the DGGE profiles. The arrows show which DGGE bands were excised and sequenced (Table 1).

Five relatively dominant DGGE bands from the surface microlayer samples were excised and sequenced (Fig. 5). All five DGGE band DNA sequences were very similar (≥98%) to 16S rRNA gene sequences from isolated bacterial strains (Table 1). DGGE bands 1 and 2 were identical to the 16S rRNA gene sequences of Dokdonia donghaensis PRO95 (FJ627052) and Krokinobacter genikus Cos-13 (AB198086), respectively, from the Flavobacteria family Flavobacteriaceae. DGGE DNA sequences 3, 4, and 5 were almost identical to two genera, Alteromonas and Glaciecola, of the Alteromonadaceae (Table 1).

TABLE 1.

Sequence similarities of excised 16S rRNA gene DGGE bands shown in Fig. 5

Band	BLAST match (accession no.)	% Similarity (no. of bases)	Taxonomic grouping (class, order, family)
1	Dokdonia donghaensis PRO95 (FJ627052)	100 (158)	Flavobacteria, Flavobacteriales, Flavobacteriaceae
2	Krokinobacter genikus Cos-13 (AB198086)	100 (158)	Flavobacteria, Flavobacteriales, Flavobacteriaceae
3	Alteromonas sp. strain BCw006 (FJ889589)	100 (163)	Gammaproteobacteria, Alteromonadales, Alteromonadaceae
4	Alteromonas sp. strain Oct07-MA-2BB-3 (GQ215064)	100 (163)	Gammaproteobacteria, Alteromonadales, Alteromonadaceae
5	Glaciecola nitratireducens FR1064 (AY787042)	98 (161)	Gammaproteobacteria, Alteromonadales, Alteromonadaceae

DISCUSSION

Bacterial abundance.

Results show that the three bacterial cell types quantified in the mesocosms responded in three different ways (Fig. 3). Both HNA bacterial cells and Synechococcus cells decreased in numbers drastically at the start of the experiment. HNA bacterial cells and LNA bacterial cells then increased in numbers in the phytoplankton bloom.

An abrupt decrease, followed by an increase in bacterioplankton cell abundance, is a characteristic response frequently observed during phytoplankton blooms (4, 6, 29, 36). A previous Emiliania huxleyi-dominated mesocosm experiment using Raunefjorden fjord water showed a very similar bacterial cell response (6). Other mesososm experiments at Raunefjorden also reported the same reduction in Synechococcus cell abundance during an induced bloom (29), thus indicating that Synechococcus cells are not successful under these conditions and/or are out-competed.

One of the principal sources of bacterial mortality in the water column is protist predation, with many protists grazing selectively (32). Significantly, some protists target rapidly growing and dividing bacterial cells, such as HNA cells (16, 44). Furthermore, recent evidence suggests that the concomitant drop in bacterial numbers and bloom of small phytoplankton may be due to mixotrophic growth of phytoplankton (49). This may therefore account for the mortality of HNA bacterial cells and Synechococcus cells, whereas the LNA bacterial cells did not appear to be affected (Fig. 3).

In this study, cell numbers in the bacterioneuston and the bacterioplankton were not significantly different, indicating that there was no enrichment of cells in the surface microlayer. Surface microlayer and subsurface water samples collected from two sites in the Mediterranean Sea also showed that the numbers of Synechococcus cells in the surface microlayer were the same as those in subsurface (0.5 m) samples (23). Bacterial cell counts by flow cytometry analysis from the same samples did have low levels of enrichment in the surface microlayer, yet the enrichment of cultivable bacterial cells was much more variable, with enrichment factors ranging from 0.5 to 191 (23). High numbers of cultivable bacterial cells in the surface microlayer compared to subsurface waters are often reported (1, 2, 43).

Bacterial community structure.

Unlike bacterial cell abundance, bacterial community structure in the surface microlayer was consistently different from that of the subsurface water. Surface microlayer samples collected using both membranes and a mesh screen showed a reproducibly distinct bacterioneuston in the mesocosms. Previous studies have characterized the marine bacterioneuston and cognate subsurface bacterioplankton in the North Sea (13), the Mediterranean Sea (1), and Pacific Ocean (7). In the North Sea and Pacific Ocean studies, the bacterioneuston community structure was distinct compared to that of the bacterioplankton 1 m below the surface (7, 13). Conversely, the Mediterranean Sea study reported no consistent differences between communities (1).

The method of surface microlayer sampling is important in the study of the bacterioneuston (7). Even though the sea surface microlayer is considered the top 1 mm of the ocean, it is operationally defined by sampling depth (26). We used a mesh screen (sampling depth of ∼400 μm) and membranes (sampling depth of ∼40 μm) to determine bacterial community structure. Previous comparison of membrane-collected and mesh screen-collected samples from an estuarine surface microlayer showed that samples collected using a mesh screen underrepresent the bacterioneuston because samples also contain subsurface water, which “dilutes” the bacterioneuston sample (7). In this study, at the start of the experiment, the mesh screen-collected bacterial community structures were more similar to the subsurface (bacterioplankton) than to the membrane-collected samples (bacterioneuston). This, however, changed during the experiment, with mesh screen-collected samples becoming more similar to the membrane-collected samples (Fig. 4 and 5). This indicated an enrichment effect in the surface microlayer, causing the bacterial communities sampled using the mesh screen to change from bacterioplankton-like to bacterioneuston-like during the experiment.

The proposed enrichment of the surface microlayer and bacterioneuston may be due to the physical nature of the mesososms used in this experiment. Even though the mesocosms were mixed continuously, they were calmer than the open fjord. Examination of surface microlayer samples offshore of Barcelona showed that under calm conditions (low wind speed and cloudless skies), the enrichment of several parameters in the surface microlayer, including heterotrophic Bacteria counts, chlorophyll-a, and suspended particle matter, increases substantially (23), supporting our observations in the mesocosms.

The methodological approaches used to compare the community structures of the bacterioneuston and the bacterioplankton can also influence data interpretation. Agogue et al. (1) used similarity values based upon Jaccard coefficients of single-strand conformation polymorphism profiles from surface microlayer and subsurface water samples collected in the Mediterranean Sea. Jaccard coefficients are absence/presence based and do not consider relative abundances (21). Franklin et al. (13) and Cunliffe et al. (7) used 16S rRNA gene clone libraries and DGGE profiles assessed using Pearson correlations, both of which take into account the relative abundances between samples. In this study we also included changes in relative abundances (T-RFs and DGGE bands). The increased resolution of community structure comparisons made using relative abundances versus comparisons made using absence/presence data may, in part, account for the conclusions of Agogue et al. (1).

In this study, bacterial community structure dynamics in each mesocosm were synchronous, showing consistent patterns between replicates (Fig. 4 and 5). The bacterioneuston communities at two sites on either side of Oahu Island were more similar to each other than to their cognate subsurface water bacterioplankton communities just 0.4 m below, also indicating nonrandom assembly of the surface microlayer community (7). Synchronicity of discrete bacterial communities, although poorly understood, is very important, as concordant community dynamics suggest that the community structure patterns that emerge are controlled and are not random (24). Therefore, if the bacterioneuston community structure is controlled by the environment and is not random, as our data suggest, then the sea surface microlayer is, indeed, an important ecological zone of the water column.

Five dominant DGGE bands in the surface microlayer were sequenced and identified (Fig. 5 and Table 1). The bands were very similar to just two families, Flavobacteriaceae (bands 1 and 2) and Alteromonadaceae (bands 3, 4 and 5). The genera Alteromonas and Glaciecola (order Alteromonadales, family Alteromonadaceae) were also prevalent in surface microlayer samples collected from the marine end of Blyth Estuary on the Northeast Coast of the United Kingdom (10). A previous study has also showed that the closely related genus Pseudoalteromonas (order Alteromonadales, family Pseudoalteromonadaceae) dominated surface microlayer samples collected from the North Sea, close to the coast of the United Kingdom (13).

Bacterial cell abundance compared to community structure.

Bacterioplankton in the water column include both free-living cells and cells attached to several possible surfaces, including phytoplankton (25) and marine gels (3). Marine gels are a significant component of the sea surface microlayer, giving it a gelatinous structure (8, 40, 46). Surface microlayer samples collected from the same mesocosms in this study were enriched with TEP (9). Therefore, in the sea surface microlayer more microorganisms may be attached than are free living (8). Analysis of free-living and attached bacterioplankton communities cooccurring in the water column show that both temporal variability and diversity in the attached community are higher than in the free-living bacterial community (37), and specific attached bacterial communities can develop (12).

The two standard marine microbial ecology approaches used in this study, flow cytometry and community profiling (T-RFLP and DGGE), inherently analyze different components of the free-living and attached bacterial cell pools. We filtered the water samples for DNA extraction and subsequent community profiling; therefore, all particles in the water sample of >0.2 μm were analyzed by T-RFLP and DGGE in both free-living and attached bacterial cell pools. However, flow cytometry counts only the free-living bacterial cells. This may contribute to the observations that there are no differences in bacterial cell abundances between the surface microlayer and subsurface water (free-living cells only), yet there are distinct and consistent differences in the bacterial community structures (free-living and attached cells). This may also be responsible for the differences reported between flow cytometry bacterial cell counts and bacterial CFU counts in samples collected in the Mediterranean Sea by Joux et al. (23).

Conclusions.

The similar dynamics of bacterial cell numbers and community structure between replicate mesocosms described in this study show how conserved patterns can emerge in bacterial systems such as the sea surface microlayer. These data indicate that the bacterial community structure patterns witnessed in the sea surface microlayer are determined by environmental forces and are not idiosyncratic. This has important implications for marine microbiological research as it is empirical evidence that supports the hypothesis that the surface ocean, particularly the sea surface microlayer, is much more structured than previously thought.