Sampling sites were in the red seabream (Pagrus major) aquaculture area, which is located in the Uwa Sea, a coastal area of Shikoku Island in southwestern Japan (Figure 1). The water depth in the area is 30–50 m. Water sampling was conducted at three sites near the net pen (aquaculture site ID; EH-1, EH-2, and EH-3) on October 24 and December 26, 2012 and March 5, April 19, May 15, June 4, July 2, September 18, and October 16, 2013. All samplings were performed at morning (9:00–11:00 am). The net pen has dimensions of 11 m × 11 m with a depth of 7.5–10 m and is used for red seabream rearing. The monthly and cumulative feeding amount given to one net pen from April to October is summarized in Supplementary Figure S1, indicating cumulative feeding amount was over 30,000 kg in this duration. This shows a high concentration of organic matter was administered, thus we defined this area as an eutrophic one. Sampling was also conducted at three sites about 600 m away from the aquaculture area (far site ID; EH-4, EH-5, and EH-6). Water at these six sites was collected with a Van-Dorn or Niskin sampling bottle at depths of 1 and 20 m. Concentration of dissolved oxygen (DO) ranged from 87 to 100% in these depths. In addition, the bottom of sampling sites around pen net is sandy with shell exoskeleton. Even if deep waters bring up the bottom materials, Anoxic-PB could not affect on the AAnPB cell number measurement.

Environmental parameters (water temperature and salinity) were measured using field equipment (pH/COND METER D-54, HORIBA, Japan; Hand-Held Refractometer ATC-S/Mill-E, ATAGO, Japan). Sample water was dispensed into acid-washed 500 or 1000 ml polycarbonate bottles that had been rinsed three times with small amounts of collected seawater. Water temperature in aquaculture sites from October 2012 to October 2013 ranged from 15.6 to 27.3°C, and the salinity ranged from 31.5 to 37.0. Environmental characteristics of each sampling site are shown in Supplementary Table S1.

All samples (10 ml aliquots) were fixed with neutral formalin (1.0%, final concentration), incubated overnight at 4°C in the dark, and filtered through Nuclepore black polycarbonate membrane filters (0.2 μm in pore size) under gentle vacuum (≤20 cm Hg). The filters were dried and then stained with 4′,6-diamidino-2-phenylindole (DAPI) prepared at 1 μg ml^-1 in a 3:1 mixture of Citifluor AF1 (Citifluor Ltd, London, United Kingdom) to Vectashield (Vector Labs, Burlingame, ON, Canada). Bacterial cells were enumerated on images taken on a Zeiss Axioplan 2 epifluorescence microscope (Carl Zeiss, Germany) equipped with a mercury lamp (USH-102D, Ushio, Japan) and a Photometrics CH-250 cooled, slow scan, infrared (IR)-sensitive CCD camera (iKon-M, ANDOR, Belfast, Northern Ireland) and connected to a Windows PC. Images were taken with scanning at 1,024 pixels × 1,024 pixels with resolution of 0.1305 μm per pixel. The following three epifluorescence filter and mirror sets were used: (1) BChl a (excitation short pass filter, 400–530 nm; emission long pass filter, >850 nm, RG850, Edmund Optics, Barrington, NJ, USA; short pass dichroic mirror, >650 nm, XF-2072, Omega Optical, Brattleboro, VT, USA); (2) Chl a (excitation, 546 ± 12 nm, emission long pass filter, >590 nm; short pass dichroic mirror, >580 nm, ZEISS Filter set 15, 488015 – 0000, Carl Zeiss, Oberkochen, Germany); (3) DAPI (excitation, 365 ± 12 nm, emission long pass filter, >397 nm; short pass dichroic mirror, <395 nm, ZEISS Filter set 01, 488001 – 0000, Carl Zeiss). First, all DAPI-stained bacteria were recorded with the DAPI filter set (100 ms exposure). Then, Chl a autofluorescence was recorded with the Chl a filter set (100 ms exposure) to identify Chl a-containing organisms. Finally, IR emission (>850 nm) images were captured with the BChl a filter set, showing both AAnPB and phytoplankton (10 s exposure). Generally, between 10 and 15 sets of images were acquired from each DAPI-stained filter. The acquired images were saved and semi-manually analyzed with the aid of MetaMorph Software (MetaMorph, Molecular Device, Sunnyvale, CA, USA) to distinguish between heterotrophic bacteria, Synechococcus and AAnPB, for each sample. The contrast and brightness of images were manipulated using the imaging software (Meta Morph) with the ‘Top Hat’ process to differentiate the cell image from the background. AAnPB were identified as cells having DAPI and IR fluorescence but not Chl a fluorescence (Figure 2). The cell abundance and cell size for total bacteria (total count of 350∼1550 cells sample^-1) and AAnPB (total count of 20–169 cells sample^-1) were calculated from DAPI stain image, because both cells can be estimated with the same condition. The mean values of all of the measured cells were taken. Cell length was measured on the DAPI-stained image as the length of the long axis. Since the cell in the present data set were almost spherical or short rods, cell volume was calculated as a spherical shape using one-half of the cell length, r, and the following equation: V = 4/3 × π ×r³. For calculation of cell volume as a rod, the calculation value is estimated to be 3.3 times less than that calculated as a sphere (Zeder et al., 2011). Although these methods can detect Anoxic-PB also, as mentioned above, our sampling sites and depth do not contain bottom materials brought up from bottom, and microscopic observation did not show large particles where Anoxic-PB would present. Cell number and biomass should be results of AAnPB.

Abundances of total bacteria and AAnPB were calculated as the mean of the three aquaculture sites (EH-1, EH-2, and EH-3) and far sites (EH-4, EH-5, and EH-6). Change in bacterial abundance through the year is shown in Figure 3. Total bacterial number ranged from 9.4 × 10⁵ to 2.7 × 10⁶ cells ml^-1 (Figure 3A) and that of AAnPB ranged from 6.8 × 10⁴ to 5.9 × 10⁵ cells ml^-1 (Figure 3B). This indicates that AAnPB accounted for 4.7–24% of total bacteria (Figure 3C). Increasing AAnPB abundance from March occurred coincidently with raising water temperature (Figures 3B,D), whereas total bacteria abundance was not correlated with water temperature. AAnPB abundance is reported to respond to temperature, although the correlation is not strong (Mašín et al., 2006; Zhang and Jiao, 2007). The present result might be caused by difference of temperature response of total bacteria and AAnPB. Cell abundance in May–July (summer season) at the aquaculture sites was not statistically significantly different from the cell abundances of total bacteria (P = 0.537, two-tailed t-test) and AAnPB (P = 0.417, Mann–Whitney Rank Sum Test, U test). Differences with depth (1 and 20 m) between May and July were observed, suggesting an abundance of cells at the surface during the stratified period. Zhang and Jiao (2007) mentioned that high abundance patches of heterotrophic bacteria, including AAnPB, were supported by dissolved organic carbon (DOC) rather than temperature. In previous studies from various ocean environments, the percentage of AAnPB was 0.1–24% (Schwalbach and Furhman, 2005; Lami et al., 2007). The highest abundance of 24% of total bacteria was observed in an oligotrophic area of the South Pacific Ocean (Lami et al., 2007). AAnPB are suggested to have supplementary energy production utilizing light when low organic matter conditions are encountered. Although AAnPB were detected in epicontinental seas (Zhang and Jiao, 2007; Chen et al., 2011) and inland seas (Mašín et al., 2006), abundance under such high organic matter conditions was much lower than in oligotrophic seas (Mašín et al., 2006; Zhang and Jiao, 2007; Chen et al., 2011). Observations in our present study showed that the percentage of AAnPB was 4.7–24%, which was the highest among reports available to date (Mašín et al., 2006; Zhang and Jiao, 2007; Chen et al., 2011). Our previous report showed 3.5–7.9% in open areas of the Uwa Sea (Sato-Takabe et al., 2015). This indicates that AAnPB contribute greatly to the bacterial carbon stock in the high organic matter concentration area of the Uwa Sea. Abundance, growth, cell volume and species diversity are factors that determine the contribution of bacteria in the ecosystem.

Previous studies reported that AAnPB grow faster than heterotrophic bacteria in the euphotic zone across the Atlantic Ocean (Koblížek et al., 2007) and consumed leucine with a high rate in the bacterial community in the Delaware Estuary (Stegman et al., 2014). Based on the results of our study and previous studies taken together, it is suggested that AAnPB can utilize organic matter more efficiently than other bacteria, which leads to fast growth and results in high abundance.

To estimate cell volume, we measured cell length. The cell length for total bacteria was 1.00 to 1.58 μm (median value, 1.280 μm; 25th percentile value, 1.207 and 75th percentile value, 1.370; n = 75) and that for AAnPB bacteria was 1.26 to 2.37 μm (median value, 1.647; 25th percentile value, 1.518; and 75th percentile value, 1.749; n = 75), showing that AAnPB were significantly larger than other bacteria (P < 0.001, Mann–Whitney rank sum test). Average cell volume of total bacteria was 1.17 ± 0.356 μm³ (n = 75), whereas that of AAnPB was 2.58 ± 1.15 μm³ (n = 75). We previously demonstrated that AAnPB were larger among total bacteria from open areas of the Uwa Sea (Sato-Takabe et al., 2015). The results of the present study focused on an aquaculture site are consistent with findings from a non-aquaculture area of the Uwa Sea, where the biovolume of AAnPB was 9.7–22% (Sato-Takabe et al., 2015). The AAnPB biomass was estimated to make up 10–53% of the total bacteria, suggesting that AAnPB account for a large part of bacterial carbon stock in the aquaculture area. Organic matter input (Supplementary Figure S1) would make the cell size of AAnPB larger in the present study area. Former reports (Sieracki et al., 2006; Lami et al., 2007) obtained from oligotrophic open oceans showed one third to three order magnitude smaller biovolume than our result.

The temporal and spatial changes in bacterial community composition were monitored by DGGE targeting the 16S rRNA gene, whereas changes in AAnPB were monitored with the pufM gene. All DGGE gel profiles are shown in Figure 4 and the similarity is shown in Supplementary Figure S2. For the total bacterial community (16S rRNA gene), the profiles at depths of 1 and 20 m at all sampling sites (EH-1 ∼ 6) were similar (Figure 4A) and the similarity was >90% (Supplementary Figure S2A). This suggests that the bacterial community in this area is rather homogeneous in both depths of 1 and 20 m. Since high similarity was found among sampling sites throughout the year, representative data for year-long samples at a site (EH-1 at 1 m depth) were shown as Figure 4B. Among the bands in the DGGE gel, the season-common bands (C series bands) and season-specific bands (S series bands) were detected. Non-aquaculture sites are 600 m away from the aquaculture sites and no inlet of river and wastewater. However, residue of fish feed might draft to the non-aquaculture sites, which probably gave similar nutrient condition and bacterial flora.

In the case of AAnPB diversity, all sites showed quite similar profiles also (Figure 4C), suggesting that the same AAnPB are uniformly abundant in this area. Temporal changes in the pufM profile, however, were found as shown in Figure 4D (site EH-1, 1 m depth). Ferrera et al. (2014) reported seasonal succession in the Mediterranean Sea, suggesting the distribution of different subpopulations over time. Our results also indicate that AAnPB are temporally highly dynamic. The phylogenetic relationship of the season-common bands (C-pufM series) and season-specific bands (S-pufM series) are discussed below.

Denaturing gradient gel electrophoresis bands of 16S rRNA gene were excised from the gel in order to identify the major bacteria at the EH-1 site (1 m depth) (Figure 5A). Season-common bands included Flavobacteria (band# C1, 2, 3, and 6), Gammaproteobacteria (C4), Alphaproteobacteria (C7 and 8) and cyanobacteria (C5). Season-specific bands were identified as Flavobacteria (S4, S5, and S7), Gammaproteobacteria (S1, S2, S3, S6, and S9), and Alphaproteobacteria (S8, S10, and S11). These dominant phylogenetic groups are in accordance with other reports from coastal seawater areas without aquaculture site (Yokokawa and Nagata, 2010), and aquaculture facilities (Martins et al., 2013). The season-common bands (Flavobacteria and Gammaproteobacteria) are known to be active growers (Schäfer et al., 2001; Taniguchi and Hamasaki, 2008; Ferrera et al., 2011), suggesting their role in organic matter transport to the upper trophic level by grazing. DGGE sensitivity was reported by Muyzer et al. (1993) and Murray et al. (1996), who showed that 1–2% of bacterial populations in the mixed assemblage of selected bacterial strains appeared as bands. The bands showing high abundance in the present study are estimated to represent cell density of 10⁴∼10⁵ cells ml^-1. These bands suggest that this area has sufficient abundance of major heterotrophic bacteria, which represent key groups in organic matter cycling in the microbial loop in this ecosystem.

Partial pufM sequences were determined for four bands excised from the DGGE gel for the EH-1 site (1 m depth) sample. The season-common band (C1-pufM) was detected at all sites (Figure 4C), and the season-specific bands (S1 ∼ S3-pufM) were also detected. The sequences of these bands showed that all were closely related to uncultured IBEA_CTG clones (Figure 5B). The sequence identities to IBEA_CGT SKBBG42TR clone were as follows: C1-pufM 75%, S2-pufM 76%, S1-pufM 79%, and S3-pufM 72%. Similarity to other clones was lower than SKBBG42TR case (Yutin et al., 2007, Yutin et al., 2008). Yutin et al. (2007) reported that the phylogroups including IBEA_CTG clones were the main members of AAnPB found in oligotrophic regions. The present results suggest that the phylogroup closely related to IBEA_CTG clones would be abundant in not only the open ocean but also coastal eutrophic sea areas. These bacteria were the first described phylogroups based on analysis of the pufM sequence. The present study indicates that the new AAnPB phylogroups are abundant in the Uwa Sea, which should play a role to transport of organic matter to higher eutrophic levels.

In summary, the present study shows that AAnPB contribute 10–53% of the bacterial biomass and account for a large part of the bacterial carbon stock around eutrophic aquaculture areas. The abundant phylogroups might be ubiquitously abundant in the oligotrophic open ocean and eutrophic coastal sea areas. These AAnPB groups are possibly some of the key bacteria for organic matter transfer to protists as well as major proteobacteria.