Artificial neural network analysis of microbial diversity in the central and southern Adriatic Sea

Physico-chemical conditions

Sampling was performed at 6 stations representing the physical and chemical characteristics of the investigated area (Supplementary Table S1). Thermohaline properties were the result of horizontal advection of above-average salinities driven by a North Ionian cyclonic gyre controlled by the Adriatic Ionian Bimodal Oscillating System⁴⁶. September and the whole summer of 2016 was characterized by extremely high temperatures, and precipitation in the climatologic expected range. A cyclone with a cold front followed by a strong Bora wind passed over the Adriatic a week before the cruise, in the period between the 16th and 20th of September 2016. Heat and mass exchange in the air-sea boundary layer were responsible for the characteristic vertical thermohaline profiles measured in late summer. Over the investigated area, the mixed layer depth located between 20 and 25 m was horizontally homogenous. The coldest water mass (temperature 12.94 °C, salinity 38.68) was located at the bottom of Jabuka Pit.

Abundance of bacteria, autotrophic picoplankton and AAP

Bacterial abundances ranged between 0.05 and 0.46 × 10⁶ cell mL⁻¹ in all three areas, with a slightly higher average value in Jabuka Pit (0.31 × 10⁶ cell mL⁻¹). The bacterial abundances were the highest in the upper layers down to the 50 m deep layer and showed a decreasing trend towards the bottom (Supplementary Table S2). The portion of HNA bacteria ranged from 37.8 to 73.12% (on average 51.27%), with the prevalence of HNA over the LNA group below the epipelagic layer.

Marine Synechococcus dominated the autotrophic picoplankton community with abundances ranging from 0.08 to 23.86 × 10³ cell mL⁻¹. The presence of Prochlorococcus cells was also detected in all samples in a range from a few cells to 1.33 × 10³ cell mL⁻¹. Picoeukaryotes also showed a similar range from a few cells to 0.83 × 10³ cell mL⁻¹. The highest abundances of picophytoplankton were measured in the upper 50 m, with the exception of the Palagruža Sill (PS) area, where an increase in abundance was observed at 100 m depth. Bacterial production ranged from 0.2 × 10⁴ to 0.36 × 10⁴ cell mL⁻¹ h⁻¹, with increased values in the shallow layers and a mostly uniform vertical distribution in the water column (Supplementary Table S2).

AAP bacteria abundance ranged from 0.9 × 10³ to 22.3 × 10³ cell mL⁻¹, thus constituting 0.42% to 6.83% of the bacteria. Their highest average contribution was observed in the South Adriatic Pit (4.11%), while on the vertical scale, their highest contribution was observed in the upper 20 m of the seawater (see Supplementary Table S2).

Relationship between the picoplankton community and environmental parameters

Based on biological characteristics (total prokaryotes, Synechococcus, Prochlorococcus, picoeukaryotes, heterotrophic nanoflagellates, aerobic anoxygenic phototrophs, high and low nucleic acid bacteria, bacterial production), we distinguished five picoplanktonic clusters (PIC-BMUs) and then searched for explanations of the observed patterns (Fig. 2A,B). The mean values of biological and physico-chemical parameters for each cluster are shown in Table 1.

PIC-BMU1 described a very rare pattern, found in only two samples from 10 m depth in Palaguža Sill and Jabuka Pit. They were characterised by the highest abundances of total prokaryotes with a dominance of HNA and elevated AAP abundance. These samples were unique in terms of hydrological parameters, as they represented an N-limited environment (TIN < 1 µmol L⁻¹, TIN/SRP < 10, Si/TIN > 1), where the water temperature was high (22.02 °C).

PIC-BMU2 described 28% of the samples. This cluster included the picoplankton community from the water column at and below 100 m and is characterised by the dominance of HNA in total prokaryote abundance and a decrease of all other picoplankton parameters. This layer describes the lowest seawater temperature (14.42 °C) and the highest concentrations of nitrates, nitrites, ammonium ions, silicates and SRP.

PIC-BMU3 included only one surface sample from Palagruža Sill with the highest values of bacterial production and AAP portion. The measured temperature was 22.62 °C and the highest values of N-organic compounds were detected in this sample. This sample was P-limited.

PIC-BMU4 described the pattern mostly from 50 m, with the highest contribution of LNA to total prokaryote abundance, the highest abundances of Synechococcus and picoeukaryotes, and the lowest AAP abundance. These samples were collected below the well-developed thermocline in the area of deep chlorophyll maximum (DCM) and were characterised by a P-limited environment with elevated salinity values compared to the other BMU clusters.

PIC-BMU5 described the most frequent pattern in our samples, the frequency can be attributed to the sampling effort rather than to certain general features of the sampling area. It grouped the samples from the first 50 m with high values for all measured biological parameters, except for HNA contribution to total prokaryote abundance. This layer is characterised by a temperature of approximately 21.70 °C and P limitation.

The results of the above analyses (Fig. 2A,B) document that all observed parameters were differentiated by depth rather than location. For at least 81% of the data, the negative impact of salinity is visible through the opposite direction of its anomaly values compared to most other picoplankton variables (heterotrophic bacteria, HNA%, LNA%, Synechococcus, Prochlorococcus, picoeukaryotes, AAP, bacterial production). Finally, our results suggest that an increase in temperature had a positive impact primarily in terms of high picoplankton abundances and bacterial productivity, given that the anomaly values of total prokaryotes, bacterial production and temperature display the same direction.

Bacterial community composition

Proteobacteria (mainly Alpha- and Gammaproteobacteria) and Cyanobacteria were the most abundant phyla in all samples, followed by Bacteroidota and Actinobacteriota. The changes between sites and depths at phylum level were minimal, with higher relative abundances of Planctomycota and Myxococcota observed at Jabuka Pit at 75 m, and of Nitrospinota and Myxococcota in the South Adriatic Pit at 100 m (Fig. 3A). The changes in bacterial communities were more conspicuous with depth rather than between the different basins, as will be shown below at the genus level for the most abundant phyla.

‘Ca.’ Actinomarina and uncultured Microbacteriaceae (Actinobacteriota) co-dominated in the upper layers down to about 20 m (Fig. 3B). The relative abundance of Microbacteriaceae substantially decreased in deeper layers, where it was replaced either by ‘Ca.’ Actinomarina or by uncultured Microtrichaceae from the Sva0996 lineage. The latter dominated in the deepest layers at the South Adriatic Pit (Fig. 3B).

Uncultured NS4 and NS5 lineages of Bacteroidetes co-dominated up to approximately 50 m depth, with minor contributions of genus Balneola and uncultured Balneolaceae, Cryomorphaceae and, especially in the Jabuka Pit, Flavobacteriaceae (Fig. 3C). Uncultured lineages NS2b and NS9 became relatively more abundant in deeper layers, together with genera Marinoscillum and Muricauda.

Proteobacteria was the most diverse phylum in terms of the number of detected genera. Thus, Alphaproteobacteria and Gammproteobacteria are shown separately (Fig. 3D,E, respectively). The members of Alphaproteobacteria with the highest relative abundance in the euphotic zone (~ 20 m) in all basins were affiliated with uncultured AEGEAN-169 and SAR116 lineages, with minor contributions of Ascidiaceihabitans, SAR11 clade IV, uncultured Rhodobacteraceae and S25–593 lineage. Uncultured Magnetospiraceae, Methylobacterium and OCS116 clade were more abundant in deeper layers (Fig. 3D). Moreover, 29 of genera with a relative abundance < 4% contributed 10–25% of all reads in all the samples.

The most abundant Gammaproteobacterial lineages in the upper layers down to about 50 m were OM60 (NOR5) and SAR86, with a minor contribution of KI89A clade and genus Litoricola (Fig. 3E). The exception was 5 m depth at station PS2, Palagruža Sill, where a higher relative abundance of Alteromonas was observed. Higher relative abundances of uncultured HOC36 and HgCo23lineages, and Ectothiorhodospiraceae and Thiomicrospiraceae were observed below 75 m. Moreover, a substantial proportion of Gammaproteobacterial reads consisted of numerous other genera with individual contributions < 5% (Fig. 3E). Cyanobacteria were represented by only three genera, with distinct distribution at different depths. Marine Synechococcus and Cyanobium-like sequences dominated in the euphotic zone, while Prochlorococcus in deeper waters (Fig. 3F).

The neural gas analysis grouped the samples into four best matching units (BCC-BMU1, BCC-BMU2, BCC-BMU3, BCC-BMU4) representing heterogenic patterns of bacterial community composition that differed already at the phylum level (Fig. 4). Moreover, genera that showed higher importance in the samples were not always important for delineation of the BCC-BMU. The samples are grouped by depth rather than by basin.

BCC-BMU1 included a single sample from station SAP3 from 100 m depth. It was unique because of the significant contribution of Nitrospinota (LS-NOB), the high contribution of Actinobacteriota, and the lower contribution of Cyanobacteria and Bacteroidota (Fig. 4). The differences were even more pronounced at genus level (Figs. 5, 6). Actinobacteriota were dominated by Curtobacterium and the uncultured Sva0996 lineage. An interesting pattern was observed for Bacteroidota that, despite the lower relative abundance, show higher diversity at the genus level, with a relatively equal contribution of all genera, e.g. Mauricauda, Marinoscillum or Euzebyella. Similarly, Alphaproteobacteria were very diverse in this BMU, with equal contributions of Erythrobacter, Maricaulis, Martella, uncultured OM75 clade, Tistilia and Tistrella. In contrast, Gammaproteobacteria were less diverse in this deep-water BMU than in the other units, with a higher contribution of uncultured UP05 lineage and genus Woeseia (Fig. 5). Planctomycetota were dominated by the uncultured JL-ENTP-F27 lineage, with a minor contribution of Rubripirellula, while Verrucomicrobiota was dominated by Roseibacillus with a minor contribution of the uncultured SCGC_AAA164-Eo4 lineage. Prochlorococcus was the dominant genus of Cyanobacteria.

BCC-BMU2 represents 23% of samples, mostly from the DCM layers (JP-66, JP-75, SAP3-78). This BMU appears transitional between the BCC-BMU1 and the other BCC-BMUs. This layer had the highest proportion of Cyanobacteria, represented in relatively equal proportions by Synechococcus and Prochlorococcus (Fig. 5). The contribution of Actinobacteriota was similar to that of the BCC-BMU1 (Fig. 5), but the relative abundance of Curtobacterium and the uncultured Sva0996 lineage decreased, and Ca. Actinomarina contributed almost 50% to this phylum. Bacteroidota were dominated by uncultured NS lineages and genus Balneola (Fig. 5). Within Alphaproteobacteria, there was a higher contribution of Ascidiaceihabitans and Methylobacterium, but no genus was dominant. A similar pattern was noticed within Gammaproteobacteria, which showed a relatively equal contribution of all the genera, e.g. Ca. Tenderia, Dyella, Litoricola, SUP05 lineage or Woeseia, and Verrucomicrobiota, where Coraliomargarita and Lentimonas co-dominated with Roseibacillus and the SCGC_AAA164-E04 lineage. In contrast, the composition of Planctomycetota was very similar to that of BCC-BMU1 (Fig. 6).

BCC-BMU3 describes the most frequent pattern, grouping the samples from the surface layer down to 65 m depth, regardless of the area. It is characterised by the dominance of Proteobacteria and Cyanobacteria followed by higher contributions of Bacteroidota and Actinobacteriota (Fig. 4). Ca. Actinomarina and Curtobacterium dominated Actinobacteriota, with a minor contribution of the uncultured Sva0996 lineage (Fig. 5). The proportion of genus Balneola and uncultured NS4 and NS5 lineages of Bacteroidota further increased in this BCC-BMU. Ascidiaceihabitans contributed most to Alphaproteobacteria, with higher proportions of Ca. Puniceispirillum, uncultured HIMB11 and OM75 clades, and Paracoccus. Within Gammaproteobacteria no genus dominated and the contribution of e.g. Ca. Tenderia, Litoricola, Pseudoalteromonas, and OM60/NOR5 lineage were similar (Fig. 5). The composition of Planctomycetota was very different from that of the BCC-BMUs described in the deeper samples, with a clear dominance of Urania-1B-19 and CL005 lineages. Marine Synechococcus was the dominant Cyanobacterium, followed by Cyanobium and Prochlorococcus (Fig. 5).

Finally, BCC-BMU4 describes 32% of the observed samples from surface layers down to 40 m. This pattern is similar to that of the BCC-BMU3, but with a lower contribution of Cyanobacteria and a higher contribution of Proteobacteria and Bacteroidota (Fig. 4). The composition of most phyla at a finer taxonomic level was very similar to that of BCC-BMU3, but with lower proportions of the genera that dominated in BCC-BMU1 and BCC-BMU2 (Figs. 5, 6). Actinobacteriota were dominated by Ca. Actinomarina and Curtobacterium, but the uncultured Sva0996 lineage was absent, while Nocardioides and Rhodococcus showed increased proportions. The composition of Bacteroidota was almost identical to that of BCC-BMU3, with a visibly lower proportion of NS2b lineage. Alphaproteobacteria showed higher proportions of Aurantimonas, Croceicoccus, Erythrobacter and Marivivens, with Gammaproteobacteria of Alteromonas and Litoricola displayed the same pattern (Figs. 5, 6). A further increase of Urania-1B-19 lineage (Planctomycetota), Lentimonas, Coraliomargarita (Verrucomicrobia) and Synechococcus (Cyanobacteria) was observed compared to BCC-BMU3 (Figs. 5, 6).

Bacterial community diversity patterns

The rarefaction analysis indicated that this sequencing depth was sufficient to describe the diversity of bacteria in the investigated areas of the Adriatic Sea (Fig. S2). The average values of observed ASV numbers, Shannon diversity index (H′) and Pielou’s evenness (J′) showed the highest values at the DCM depth in the Jabuka Pit and Palagruža Sill, while at SAP the highest diversity and evenness was determined at 100 m depth and the highest number of ASVs was observed in the surface layer (0–50 m) (Fig. 7). All diversity indexes correlated with Chl a concentrations (approximated with fluorescence measurements): H′ (Spearman correlation: r = 0.45 n = 22, P < 0.05), J′ (r = 0.21, n = 22, p < 0.05) and number of ASVs (r = 0.56, n = 22, p < 0.05), indicating the relationship between bacterial diversity and phytoplankton biomass.

Source: Ecology - nature.com