Seasonal variability and regulation of bacterial production in a shallow eutrophic lake

The seasonal variation of bacterial production (BP) in a shallow, eutrophic Lake Kasumigaura was clarified from 2012 to 2016. During the studied period, BP fluctuated from 1.9 to 138 μg C L−1 d−1. There were no significant correlations between BP and bacterial abundance in any season, suggesting a strong top‐down regulation on BP throughout the year. On the other hand, BP was also related to bottom‐up regulation factors such as water temperature, phosphorus, and primary production (PP) annually. During winter, BP was positively correlated with chlorophyll a concentration, suggesting that autochthonous substrates were relatively important for BP. Moreover, BP was positively correlated with heterotrophic nanoflagellates, ciliates, and copepods, suggesting higher availability of substrates for BP. In summer, although there was no significant correlation between BP and PP, rainfall amount showed significant negative correlations with both BP and PP, suggesting depressed PP from relatively lower solar irradiance coupled with unfavorable weather conditions that decreased the substrate supply for bacteria. These results suggest that temporal variation of BP was regulated not by allochthonous, but by autochthonous substrates during both the highest (summer) and lowest (winter) productive seasons, even in a shallow, eutrophic lake. PP in autumn was approximately half that of spring due to lower solar irradiance, although water temperatures during both seasons were similar and nutrient concentrations during autumn were higher. On the other hand, BP in autumn was comparable with that in spring, and the bacterial carbon demand (= BP + bacterial respiration; 1.12 ± 0.79 g C m−2 d−1) was comparable to PP (1.16 ± 0.53 g C m−2 d−1), suggesting the relative importance of higher allochthonous substrates relative to other seasons.

Heterotrophic bacteria are important in the cycling of material and energy in pelagic environments (Azam 1998). Bacteria produce their biomass (= bacterial production [BP]) using dissolved organic matter (DOM), which most other organisms cannot utilize (recalcitrant DOM), and are linked to microbial and grazing food webs (Wylie and Currie 1991). Moreover, bacteria play a key role in the mineralization of organic matter through respiration (= bacterial respiration; BR), and it accounts for more than half of the ecological respiration (Williams 2000). BR and bacterial carbon demand (BCD = BP + BR) often show higher values than phytoplankton primary production (PP) in unproductive or humic lakes with high dissolved organic carbon (DOC) concentration (e.g., del Giorgio et al. 1997;Jansson et al. 2000;Fouilland and Mostajir 2010). These lakes are defined as heterotrophic lakes and are supported by inputs of allochthonous organic carbon. On the other hand, PP exceeds BR, or BCD, in relatively high productive, or clear, lakes Fouilland and Mostajir 2010). As the balance between BR, or BCD, and PP is important to assess whether a lake is a CO 2 source or sink relative to the atmosphere, understanding bacterial dynamics is critical to quantifying biogeochemical cycles in lacustrine environments.
Heterotrophic BP is regulated by many environmental factors, such as water temperature (e.g., Shiah and Ducklow 1994;*Correspondence: ktsuchiya@soka.gr.jp This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Additional Supporting Information may be found in the online version of this article. Coveney and Wetzel 1995), resources (organic and inorganic nutrients; Cole et al. 1988;Toolan et al. 1991;Carlsson and Caron 2001), predation (Bettarel et al. 2003), and viral infection (Weinbauer and Höfle 1998). Of the latter two bacterial loss processes, grazing mortality is dominant (Pedrós-Alió et al. 2000). Relative importance of these factors can vary depending on season, depth, and trophic state of aquatic environments (Ducklow 1992). In the case of shallow, eutrophic lakes, two regulating factors are commonly debated: top-down and bottom-up controls (McQueen Donald et al. 1989;Muylaert et al. 2002). Pace and Cole (1994) argued that BP appears to be limited primarily by resources in most situations (Pace and Cole 1994). On the other hand, Sanders et al. (1992) reviewed relationships between bacteria and heterotrophic nanoflagellates (HNF) and concluded that top-down control (predation) is more important in eutrophic than in oligotrophic environments (Sanders et al. 1992).
When sufficient organic matter and/or inorganic nutrients are supplied through frequent sediment resuspension, especially in shallow, eutrophic lakes, BP and PP are often relieved from resource limitations (e.g., Johengen et al. 2008). Under these conditions, top-down controls such as predation and viral infection are considered the dominant regulating factors of BP (Fischer and Velimirov 2002;Zingel et al. 2007). For example, the addition of inorganic nutrient did not stimulate BP in the eutrophic, shallow Lake Suwa in Japan; thus, predation and viral infection (top-down control) were likely more important in regulating BP than availability of resources in the lake (Watanabe 1996). Further, seasonal variability should play an important role in biological processes, especially in temperate lakes, where water temperature varies from~4 C to > 30 C (Nanazato and Yasuno 1985).
The primary factor for bottom-up control of BP is temperature because biochemical reaction rates, metabolic rates, and nearly all other rates of biological activity increase exponentially with temperature according to the kinetics described by the Boltzmann factor or the van 't Hoff-Arrhenius relation (Brown et al. 2004). However, high variability of BP in relatively high water temperatures has often been observed (Coveney and Wetzel 1995). Previous studies have explained the variances along temperature gradients with phosphorus concentration (Gurung and Urabe 1999), algal biomass (White et al. 1991), and bacterial communities (Hall and Cotner 2007).
Macrozooplankton, including copepods and cladocerans, also play a role in bottom-up regulation of BP. Their activities such as sloppy feeding, excretion, and feces can enhance BP (Lampert 1978;Güde 1988;Peduzzi and Herndl 1992;Hygum et al. 1997;Strom et al. 1997;Møller and Nielsen 2001). Sloppy feeding increases bacterial accessibility to DOM through the physical breakage of food source (Strom et al. 1997). In a short incubation experiment, excretion was the dominant mode of release of DOC, where up to 80% of total DOC release was attributed to the copepod Acartia tonsa (Saba et al. 2011). In other studies, fecal pellets of copepods A. tonsa and Calanus spp. were suggested to leach up to 50% of the total carbon content of the fecal pellets in dissolved form within the first 12-48 h of egestion (Urban-Rich 1999; Thor et al. 2003), and the leakage from fecal pellets of copepod Eudiaptomus graciloides and cladoceran Daphnia cucullate enhanced BP (Hygum et al. 1997). Although the relationships between BP and macrozooplankton activities have been examined by experimental approaches, there are few studies that examine the relationship through field investigations .
The objective of the present study was to clarify seasonal variability and regulation of bacterial carbon production in a shallow, eutrophic lake. Knowledge of what factors control bacterial carbon production furthers our understanding of how bacterial-derived recalcitrant DOM is supplied in aquatic ecosystems. Because BP is high in carbon production in Lake Kasumigaura (Kawasaki et al. 2013), the lake is appropriate for testing whether top-down or bottom-up effects control bacterial carbon production. Data were obtained through consecutive monthly investigation for 4 yr, and relationships between abiotic and biotic variables seasonally and annually were analyzed to clarify temporal variability of regulating factors of bacterial growth. Furthermore, carbon budgets in BCD and PP were estimated both seasonally and annually. In the present study, top-down and bottom-up regulations were examined from the viewpoint of regressions of bacterial abundance (BA) as a function of BP (Billen et al. 1990); there is a steep regression slope if bacteria are strongly controlled by resources (bottom-up regulation), and there is no relationship, or a lower regression slope, if other factors such as predation are more important in regulating bacteria (Pace and Cole 1994).

Materials and methods
Lake Kasumigaura and monitoring data Lake Kasumigaura, the second largest lake in Japan, is located in the eastern part of the Kanto Plain, 50 km northeast of Tokyo. About 1 million people live in the lake's watershed (1577 km 2 ). Land use in the watershed is 30% forest, 25% paddy field, 25% plowed field, 10% residential, and 10% others. The lake basin is smooth and shallow, with a surface area of 171 km 2 , a mean depth of 4.0 m, and a maximum depth of 7.3 m. Due to the extremely high loads of organic matter and nutrients, this lake is well known for eutrophication, with mean concentrations of chlorophyll a (Chl a), phosphorus, and nitrogen of 65 μg L −1 , 95 μg L −1 , and 1.15 mg L −1 , respectively, measured at the center of the lake from August 1992 to March 1993 (Imai et al. 2001;Imai et al. 2003). In addition, although Lake Kasumigaura is a shallow lake, diurnal stratification is formed in summer, and the bottom waters of this lake periodically become anoxic (Ishikawa and Tanaka 1993). Under such condition, phosphate is released at high rates from the bed sediments, leading to high phosphate concentration in the water column. Lake Kasumigaura has been registered as a core site of the Japan Long-Term Ecological Research Network and as a trend monitoring station of the United Nations Global Environmental Monitoring System Water Trend Monitoring Project (GEMS/Water).

Tsuchiya et al. Seasonal variability and regulation of BP
Five data sources were utilized in the present study: three publications from Ecological Research Data Paper Archives (Takamura and Nakagawa 2012a;Takamura and Nakagawa 2016;Takamura et al. 2017), Lake Kasumigaura Database (National Institute for Environmental Studies 2016, accessed via http://db. cger.nies.go.jp/gem/moni-e/inter/GEMS/database/kasumi/index on 30 July 2017), and the present study. All data used in the present study were obtained from the standardized monthly sampling protocol as part of Lake Kasumigaura Long-term Environmental Monitoring Program that has been conducted by National Institute for Environmental Studies since 1976. More specifically, data on (1) BA, HNF, and ciliates were obtained from Takamura and Nakagawa (2012a), (2) PP, Chl a, dissolved inorganic carbon (DIC), and photosynthetically active radiation (PAR) were obtained from Takamura and Nakagawa (2016), (3) cladocerans and copepods were obtained from Takamura et al. (2017), (4) water temperature, nutrients (N and P), and particulate organic carbon/nitrogen (POC/N) were obtained from Lake Kasumigaura Database, and (5) BP and DOC were obtained from this study.
Monthly sampling was conducted at the center of Lake Kasumigaura (Sta. 9; 36 02.142 0 N, 140 24.222 0 E) from March 2012 to March 2016 aboard the R/V NIES'94 of National Institute for Environmental Studies (Fig. 1). Water temperature was measured using a Hydrolab DS5 (OTT Messtechnik GmbH) at 0.5 m intervals, and the average from 0 to 2 m was defined as the surface-water temperature. PAR in the water column was measured every 0.5 m depth with an LI-192SA/B quantum photometer (LI-COR), and the vertical light extinction coefficient (K d ) was calculated from the slope of a linear regression of the natural logarithm of PAR vs. depth. A total of 40 liters of surface-water samples were collected using a 2-m vertical column sampler for nutrients, DIC, DOC, POC/N, Chl a, PP, BA, BP, HNF, ciliates, and macrozooplankton. Water samples for nutrients, DOC, POC/N, and Chl a were immediately cooled in an ice cooler and brought back to the laboratory, and samples were filtered through precombusted (450 C for 4 h) GF/F glassfiber filters (Whatman). The filtrate was used for nutrients and DOC measurements, and samples collected on the filters were analyzed for POC/N and Chl a measurement. Samples for DIC, PP, and BP were brought back to the laboratory in a dark box without cooling. Lake water samples for macrozooplankton (cladocerans and copepods [adults + copepodites + nauplius]) were filtered through a 40-μm-mesh plankton net and immediately preserved with sugar-containing formalin at a final concentration of 4% on the survey vessel. Samples for the enumeration of BA and HNF were immediately fixed on the survey vessel with glutaraldehyde (final concentration of 1%) and were fixed with Lugol's iodine solution for counting ciliates.
Briefly, the collected standard monthly samples were analyzed as follows: nutrients (PO 4 -P, dissolved total phosphorus, NH 4 -N, NO 3 -N, and NO 2 -N) were analyzed using a continuous flow analyzer (AACS-II and BRAN + LUEBBE) (Yukihiro 1987;Otsuki et al. 1993). POC/N was measured with a CHN analyzer (MT-5, Yanaco; Knap et al. 1996). DIC was measured with a total organic carbon (TOC) analyzer (TOC-V CSN , Shimadzu; Takamura and Nakagawa 2016). PP was measured via shortterm incubation using NaH 13 CO 3 in the laboratory, and the concentration of organic carbon and isotope ratios of 13 C and 12 C were determined with a mass spectrometer facilitated with a combustion furnace (Delta V Advantage, Thermo Fisher Scientific; Takamura and Nakagawa 2016). Chl a was extracted in 100% methanol, and the concentration was spectrophotometrically measured (Marker 1980). The sample for BA was filtered onto a 0.2-μm-pore-size nuclepore filter (stained with Sudan Black B), stained with 4 0 6-diamidino-2-phenylindole solution, and counted using an epifluorescence microscope (BX51, Olympus) equipped with a U-excitation system (BX-RFA, Olympus; Takamura et al. 1996). The sample for HNF was filtered onto a 1.0-μm-pore-size nuclepore filter (stained with Sudan Black B), stained with fluorescein isothiocyanate solution, and counted using an epifluorescence microscope (BX51, Olympus) equipped with a BV-excitation system (BX-RFA, Olympus; Takamura et al. 1996). The ciliates were counted using an inverted microscope (TSM, Nikon; TS10, Nikon; CKX42, Olympus) after the plankton in 3-10 mL of sample was settled for 24 h in a Utermöhl chamber (Takamura et al. 1996). Macrozooplankton were counted in the laboratory using an inverted microscope (TSM, Nikon; TS10, Nikon; CKX42, Olympus) and were identified to species or genus level (Takamura et al. 2017).
DOC measurements were conducted as nonpurgeable DOC with a TOC analyzer (TOC-V, Shimadzu) equipped with a Pt catalyst on quartz wool. At least three measurements were made for each sample, and analytical precision was typically less than AE 2%. Potassium hydrogen phthalate (Kanto Chemical) was used as a standard.

Bacterial production
Due to the regulation in Japan on outdoor use of radio isotopes such as 3 H-thymidine (Fuhrman and Azam 1982) and 3 H-leucine (Kirchman et al. 1985;Smith and Azam 1992), which are common BP measurement methods, we applied a stable isotope method ( 15 N-deoxyadenosine method; Tsuchiya et al. 2015b) to measure BP. The collected lake water was poured through acid-cleaned 60 mL polyethylene bottles, and incubated with 50 nmol L −1 final concentration of [ 15 N 5 ]-2 0 -deoxyadenosine ( 15 N-dA, NLM-3895-PK, Cambridge Isotope Laboratories) under dark condition at in situ temperature for a few hours to 24 h depending on the in situ temperature. After the incubations, 5 to 20 mL of water samples were filtrated onto a 0.2-μm polytetrafluoroethylene (PTFE) membrane filter (Omnipore, Millipore), and the filter was stored at −80 C until further analysis. Bacterial DNA was extracted from the filter using Extrap Soil DNA Kit Plus ver.2 (Nippon Steel & Sumikin Eco-Tech Corporation) according to the manufacturer's protocols. Briefly, each filter was cut into small pieces and placed in a bead-beating tube, to which 950 μL of extraction buffer and 50 μL of lysis solution were added. Bacterial cells were disrupted using a homogenizer (Fast Prep FP120, MP Biomedical) at a speed of 6.0 m s −1 for 40 s, after which DNA was purified by using magnetic beads to produce 100 μL of DNA extract.
After the DNA extraction, the DNA sample was hydrolyzed by three enzymes to nucleosides to quantify the 15 N-dA according to Nohara et al. (2011) with minor modification. The extracted DNA sample was denatured by heating at 98 C for 10 min, and the sample was then quenched rapidly on ice. The sample was incubated with 2 units of nuclease P1 (Wako) at 60 C for 2 h in 10 mmol L −1 ammonium acetate to hydrolyze 3 0 -5 0 phosphodiester bonds in heat-denatured DNA and 3 0 -phosphomonoester bonds in mono-and oligonucleotides terminated with a 3 0 -phosphate group. Next, the sample was incubated with 0.002 units of phosphodiesterase I (Worthington) at 37 C for 2 h in 100 mmol L −1 ammonium bicarbonate to hydrolyze 5 0 -mononucleotides from 3 0 -hydroxy-terminated deoxyribooligonucleotides. Thereafter, 0.5 units of alkaline phosphatase (Promega) was added to the sample and then incubated at 37 C for 1 h to catalyze the hydrolysis of 5 0 -phosphate groups from deoxyribonucleoside triphosphates.

DNA extraction efficiency
For examining DNA extraction efficiency of Extrap Soil DNA Kit Plus ver.2 (Nippon Steel & Sumikin Eco-Tech Corporation), lake water samples were collected at Sta. 1, Sta. 7, and Sta. 9 in Lake Kasumigaura in August 2015 (Fig. 1), and 1 and 5 mL of the lake water samples (3 stations × 2 filtration volumes = total 6 samples) were filtered using a 0.2-μm PTFE membrane filter (Omnipore, Millipore). When DNA of the filter samples were extracted, nec1 gene (polymerase chain reaction (PCR) amplified product) was added to bead tubes (final concentration 1.0 × 10 6 copies tube −1 ), and then the amount of nec1 gene in purified DNA samples were quantified by quantitative polymerase chain reaction (qPCR; Light Cycler480, Roche) after DNA extraction. Control samples, including no lake water filter samples, were also tested. In the quantification of nec1 gene, NECS-F and NECS-R were used as forward and reverse primers, respectively. Amounts of nec1 gene of each sample were measured in triplicate, and the standard deviations ranged from 2% to 12% (average 6%). The recovery of nec1 gene showed 87% to 110% (average 102%, n = 7), suggesting that added nec1 gene is recovered almost 100% through the DNA extraction. Therefore, DNA extraction efficiency was considered as 100% in the present study.

Comparison with tritiated leucine ( 3 H-Leu) method and calculation of BP
The 15 N-dA method was compared with the 3 H-Leu method (Kirchman et al. 1985) using lake water samples. Lake water samples were collected at Sta. 1, Sta. 7, Sta. 9, and Sta. 12 in Lake Kasumigaura in August 2013 (Fig. 1). That is, 1.5-5 mL of triplicate lake water subsamples from each station were incubated with 50 nmol L −1 final concentration of 15 N-dA or 3 H-Leu (L-[4,5-3 H(N)]-Leu, NET1166, PerkinElmer), respectively. Each sample was incubated at 10 C, 15 C, 20 C, 25 C, and 30 C for 5 h. Blanks of 3 H-Leu were prepared by adding formalin (final concentration 1%) to the samples collected from each station at 20 C at the start of incubation. At the end of incubation, the samples with 15 N-dA were filtrated onto a 0.2-μm PTFE membrane filter (Omnipore, Millipore), and the filters were stored at −80 C until further analysis. The 15 N-dA incorporation rate was measured as previously described. Incorporation assay of 3 H-Leu was performed according to the procedure described by Kirchman et al. (1985). Radioactivity was determined using a Perkin Elmer Wallac 1414 Scintillation Counter.
The incorporation rate of 15 N-dA was significantly correlated with that of 3 H-Leu (Fig. 2). The slope of a standard major axis (SMA) Model II linear regression (Legendre 2001) was 32.6 with 97.5% confidence intervals of 25.9 and 41.0, and we used the slope value (32.6) as a conversion factor from 15 N-dA incorporation rate (pmol-dA L −1 h −1 ) to 3 H-Leu incorporation rate (pmol-Leu L −1 h −1 ). The converted leucine uptake was converted into rates of carbon production, assuming a conversion factor of 3.1 kg C mol −1 leucine (Kirchman 1993). To examine the relative importance of autochthonous substrate (PP) for BP in the water column, BR (μg C L −1 h −1 ) was estimated using an empirical equation of temperate lakes (Log[BR] = 0.66 × Log[BP] + 0.68; Amado et al. 2013), and then BCD was calculated (BCD = BP + BR). The BCD was depth integrated (5.5-6.9 m depths in Lake Kasumigaura), and then depth-integrated BCD and PP were compared.

Definition of season and statistical analysis
Seasons at Lake Kasumigaura were defined based on water temperature: spring included April, May, and June (11.8-23.4 C); summer included July, August, and September (23.8-30.2 C); autumn included October and November (14.0-22.4 C); and winter included December, January, February, and March (4.0-10.4 C).

Statistical analysis
To examine statistical significance in correlations among variables in each season and throughout the year, the data were analyzed using the Excel 2016 software (Microsoft Corporation) with the add-in software Statcel2 (OMS Publishing) and the qvalue R package (Storey et al. 2004). The correlations were investigated by Pearson's correlation coefficients. Multiple testing corrections were conducted using the Storey's false discovery rates (FDR) to remove false positives and determine the appropriate α for interpreting the results as significant, given its higher power than other correction methods such as Benjamini-Hochberg FDR (Storey and Tibshirani 2003;Storey et al. 2004). We considered α = 0.021 to be an acceptable significance level.

Seasonal variations
Water temperature showed clear seasonal variations, and ranged from 4.0 C in February 2014 to 30.2 C in August 2013 (Table 1; Fig. 3a). DIC and DOC concentration did not show clear seasonal variations and displayed relatively stable values throughout the year at 13.6 AE 1.0 mg C L −1 and 3.19 AE 0.28 mg C L −1 , respectively (  Fig. 3c). In some cases, NH 4 concentrations were depleted in spring and summer. NO 2 + NO 3 concentrations increased from autumn to winter and showed a maximum of 708 μg L −1 in February 2014 (Table 1; Fig. 3d).
BP ranged from 1.88 μg C L −1 d −1 in December 2013 to 138 μg C L −1 d −1 in August 2015 (Fig. 4a) and showed relatively higher values in summer (Table 1). BA fluctuated from 4.5 × 10 9 cells L −1 in December 2012 to 30.3 × 10 9 cells L −1 in September 2014, and relatively higher values were observed in summer and autumn (Table 1; Fig. 4b). PP ranged from 0.3 g C m −2 d −1 in March 2013 to 6.9 g C m −2 d −1 in July 2013 (Fig. 4c). PP increased in summer, and showed similar and lower values in autumn and winter (Table 1). Chl a concentration did not show a clear seasonal trend (Table 1; Fig. 4d).
Ratio of BCD to PP (BCD/PP ratio) ranged from 0.148 (June 2013) to 2.50 (July 2015; see Supporting Information). The ratio occasionally showed more than 1, indicating that PP in the lake water column could not satisfy BCD, and relative importance of allochthonous substrate was significant. In spring, summer, and winter, PP (2.05, 3.19, and 1.18 g C m −2 d −1 ) was 1.8, 2.1, and 1.7 times higher than BCD (1.11, 1.53, and 0.69 g C m −2 d −1 ) compared to each seasonal average, respectively. In autumn, PP (1.16 g C m −2 d −1 ) was almost the same as BCD (1.12 g C m −2 d −1 ) (BCD/PP ratio = 0.97). Throughout the year, PP (1.90 g C m −2 d −1 ) was 1.8 times higher than BCD (1.08 g C m −2 d −1 ).

Relationships between variables
When all seasons are considered, BP was significantly positively correlated with PP, water temperature, phosphorus, cladocerans, and copepods (see Table 2, statistical analysis). Conversely, there were no significant correlations between BP and carbon species (DIC, POC, and DOC). Significantly positive correlations between PP and water temperature and DOC and POC and negative correlations between PP and NO 2 + NO 3 were observed. Chl a concentration was significantly positively correlated with POC and ciliates and negatively correlated with NO 2 + NO 3 . BP/PP ratio was negatively correlated with POC and positively correlated with phosphorus and copepods. The BP/BA ratio was significantly positively correlated with PP, NH 4 , and copepods.
Correlations between variables within each season exhibited different trends compared to correlations when analyzed throughout the year. In spring, BP showed a significant positive correlation to NH 4 concentration ( Table 2). No significant correlations between environmental variables and PP were observed in spring. The BP/BA ratio was positively correlated with NH 4 concentration. In summer, no environmental variables explained the variation of BP. PP was significantly positively correlated with POC and water temperature, and Chl a concentration was significantly negatively correlated with NO 2 + NO 3 . In autumn, BP and the BP/PP ratio was significantly positively correlated with NH 4 concentration. Interestingly, there was a significant positive relationship between BP/PP ratio and cladoceran abundance in autumn. In winter, BP was positively correlated with Chl a, HNF, ciliates, and copepods. During this season, the BP/BA ratio showed significantly positive correlations with Chl a, DIC, and POC and negative correlation with NO 2 + NO 3 . In addition, Chl a was negatively correlated with NO 2 + NO 3 concentration. In all seasons, BA did not show any significant correlations with HNF, ciliates, and cladocerans. There were no significant correlations between BA and BP both seasonally and annually (Table 2), suggesting that top-down regulations on bacteria could not be negligible throughout the year, although BP was significantly positively correlated with resources such as DTP, NH 4 , and Chl a concentration seasonally and annually.

Discussion
The remarkable finding in the current study is that, in the case of Lake Kasumigaura, the top-down regulation on BP is confirmed throughout the year based on the regression of BA and BP (Billen et al. 1990). This agrees with previous studies that concluded top-down control is more important in eutrophic environments (Sanders et al. 1992). However, the uncoupling between BA and BP might have been driven by resource regeneration by consumers, subsequently implying that substrate supply processes were also      Cili, ciliates; clad, cladocerans; cope, copepods (adults + copepodites + nauplius); NH 4 , ammonium; NO 2+3 , nitrite and nitrate; PO 4 , phosphate; WT, water temperature.
Corrections for multiple testing was conducted by Storey's false discovery rates (Storey et al., 2004), and correlation coefficients in bold were statistically significant (α < 0.021).

Tsuchiya et al. Seasonal variability and regulation of BP
BP was positively correlated with water temperature throughout the year ( Table 2). The finding agrees with previous studies (Goosen et al. 1997;Pomeroy and Wiebe 2001). However, water temperature explained the variation of BP just 24% of the time (Fig. 5). Increasing water temperature and residual errors of BP from the regression line tended to increase. The relationship suggests there could be regulation factors of BP other than water temperature. To minimize the effect of water temperature on the analysis, correlations between BP and other environmental variables were examined seasonally. As a result, it was found that the correlations within seasons were different from those analyzed annually.
In winter, BP was significantly positively correlated with Chl a concentration ( Table 2), suggesting that autochthonous substrates derived from PP mainly supported BP. During this season, rainfall was low compared to other seasons in the studied period (winter 62 mm month −1 ; other seasons 126-141 mm month −1 in average; AMeDAS 2017; http://www.data.jma.go.jp/obd/stats/ etrn/index.php), suggesting that the amount of inflow to Lake Kasumigaura was relatively limited. Furthermore, loading of nutrients and organic matter from the surrounding cultivations, such as lotus fields, was relatively low compared to other seasons (Kitamura et al. 2013). Therefore, relative importance of autochthonous substrates increased compared to allochthonous substrates for BP. During this season, there were significant negative correlations between Chl a concentration and NO 2 + NO 3 , suggesting that phytoplankton abundance increase due to utilization of NO 2 + NO 3 . Decrease in loading of such nutrients due to low rainfall will influence PP, which could result in the suppression of BP in winter.
Besides autochthonous substrates as bottom-up factors, we found a positive correlation between BP and copepods abundance in winter (Table 2). In this season, diatoms were dominant (Thalassiosiraceae and Ulnaria japonica), accounting for an average of 59% of the total phytoplankton biovolume, during the studied period (Takamura and Nakagawa 2012b). Dominant taxon of the copepod community was Cyclopoida in winter, and the dominant species was Cyclops vicinus especially in January, February, and March 2016 (Takamura et al. 2017), when BP showed the highest values in winter. This species utilizes a wide size range of algae and is considered as an explicit diatom feeder (Tóth and Zánkai 1985;László et al. 1987). The results suggest that Cyclopoida, especially C. vicinus, ingested relatively large-sized diatoms and contributed to DOC release through sloppy feeding, excretion, and/or feces, leading to enhancement of BP in Lake Kasumigaura, although the relative importance of them was unclear. In a laboratory experiment, bacterial growth rate was significantly stimulated in an incubation with copepod Calanus pacificus and diatom Thalassiosira weissflogii (Strom et al. 1997) and with copepod A. tonsa and diatom Ditylum brightwelli (Møller and Nielsen 2001). Moreover, in a mesocosm experiment, BP in the vat with additions of diatom T. weissflogii and copepod A. tonsa was enhanced compared to the vat without addition of A. tonsa (Roman et al. 1988). Yoshida et al. (2001) conducted incubation experiments with bacteria and natural zooplankton communities dominated by copepods and found that BA increased in proportion to zooplankton biomass . The study concluded that the increase in BP was not only due to decrease in HNF and ciliates, major consumers of bacteria, but also due to the increase in the availability of substrates (e.g., nutrient remineralization) through copepod's activities. Thus, the results from the present study suggest that in winter, microphytoplankton are major substrate sources for BP and bacterial accessibility to the substrates was enhanced by copepods.

Tsuchiya et al. Seasonal variability and regulation of BP
In both spring and autumn, BP was significantly positively correlated with NH 4 concentration ( Table 2). The result indicates that BP was limited by NH 4 and/or that NH 4 accumulated through bacterial degradation of organic matter. Usually, regenerated NH 4 is an important nitrogen source for PP and assimilated rapidly. However, the BP/PP ratio was significantly positively correlated with NH 4 concentration in both spring and autumn ( Table 2). The results suggest that relatively high BP leads to high NH 4 regeneration and that relatively low PP cannot utilize regenerated NH 4 rapidly, resulting in NH 4 accumulation during these seasons. On the other hand, nitrogen competition between phytoplankton and bacteria cannot be dismissed. Lake Kasumigaura is known as a nitrogen-limited lake because of high amounts of phosphorus supplied from sediments. Moreover, Matsuzaki et al. (2018) showed PP was limited by NO 3 , using a convergent cross-mapping statistical approach (Matsuzaki et al. 2018), although NH 4 was not included in the analysis. The present study could not determine the relative importance of the two processes of nitrogen accumulation and/or competition. However, nitrogen dynamics associated by the bacterial activity should be investigated to examine the productive structure of Lake Kasumigaura.
In summer, although BP did not show any significant correlations with environmental variables in the water column (Table 2), it was found that BP was significantly negatively correlated with integrated precipitation 3 d prior to the sampling day (Fig. 6a). Generally, BP can be enhanced after heavy rain due to allochthonous substrate supply from terrestrial runoff and/or sediment resuspension in coastal oceans (Tsuchiya et al. 2015a) and lake ecosystems . However, the results in the present study were different from the expected results (Fig. 6a). There can be two plausible explanations for the negative correlation between BP and rainfall: (1) the dilution of BA from increased rainfall and (2) the decrease of autochthonous substrate from PP due to increased cloud cover (decrease of surface PAR) and/or reduced PAR transmittance in the water column by introduced allochthonous matter after heavy rain. There was no significant correlation between the integrated precipitation and BA (R 2 = 0.0002-0.1113, data not shown), suggesting that there was no influence of dilution on BA and BP. However, PP showed a significantly negative correlation to the integrated precipitation (Fig. 6b). Surface PAR integrated from 1, 2, or 3 d prior to the sampling day was significantly negatively correlated with the integrated precipitation during the same period (p < 0.05, data not shown). There were no correlations between K d of PAR and integrated precipitation (R 2 = 0.010-0.067, data not shown). Therefore, lower surface PAR from increased cloud cover condition depressed PP leading to a lower substrate supply from phytoplankton to bacteria in the water column and might have caused negative correlations between BP and integrated precipitation. On the other hand, BP did not show significant correlations to PP (Table 2). Low irradiance conditions can lead to phytoplankton carbon-limited conditions, causing a lower excretion of DOC from phytoplankton through PP (Zlotnik and Dubinsky 1989). As bacteria usually produce their biomass using excreted DOC, the varying proportion of exudate from PP could mask the correlations between BP and PP.
In many cases, BR tends to exceed phytoplankton net production in unproductive systems (< 70-120 μg C L −1 d −1 ; del Giorgio et al. 1997). In Lake Kasumigaura, volumetric PP in the euphotic zone (= areal PP divided by PAR 1% depth) ranged from 163 to 2697 μg C L −1 d −1 , and these values were much higher than PP of "unproductive systems." Moreover, throughout the year, PP was sufficient for BCD in the water column (annual BCD/PP ratio = 0.568). The present results agree with previous studies that PP generally exceeds BR in relatively high productive systems (del Giorgio et al. 1997).
When our results were assessed from the seasonal fluctuation of the carbon budget, PP was about twofold higher than BCD in winter, spring, and summer. Due to the fact that Lake Kasumigaura is surrounded by farmland such as paddies and plowed fields, a large amount of anthropogenic nutrients should be loaded into this lake (Kitamura et al. 2013). Furthermore, nutrient regeneration from the lake sediment should also significantly affect the nutrient dynamics in the water column, especially when dissolved oxygen concentration of the bottom layer becomes low due to the formation of diurnal stratification (Ishikawa and Tanaka 1993). The fluxes of PO 4 and NH 4 from sediment at the center of Lake Kasumigaura were up to 0.8 μg P cm −2 d −1 and 21 μg N cm −2 d −1 (Imai et al. 2007). These nutrient-loading processes can relieve phytoplankton from nutrient limitation in spring and summer, when there is enough solar irradiance for PP. In autumn, PP was almost the same as BCD. Lower solar irradiance in this season (12.0 AE 4.6 MJ m −2 d −1 ) could have caused relatively lower PP relative to spring solar irradiance (17.6 AE 9.0 MJ m −2 d −1 ), although NO 2 + NO 3 and PO 4 concentrations in autumn were higher than those in spring (Table 1). Bacterial uptake of phytoplankton exudates averaged 39% (reviewed in Fouilland and Mostajir [2010]), suggesting that allochthonous substrates should be necessary in autumn. The DOC dissolution rate from sediments was approximately 10 mg C m −2 d −1 in autumn (Imai et al. 2007), suggesting that DOC loading accounted for just 1% of the autumn BCD (= 1.12 g C m −2 d −1 ). Moreover, the 100-d biodegradation rate of sedimentary pore-water DOC was found to be from 2% (winter) to 11% (summer). Thus, the contribution of DOC originating from sediments was lower than allochthonous substrates for bacterial growth. As to inputs from outside of the lake, we examined the correlation between integrated rainfall 1 d prior to the sampling date and the BP/BA ratio and found a significant positive correlation between the two (n = 8, R 2 = 0.74, p < 0.01). The result suggests that bacterial productivity was enhanced by, and quickly responded to allochthonous substrate inputs from sources outside of the lake such as from runoff, rainfall itself, and river in autumn when PP could not satisfy the BCD.

Tsuchiya et al. Seasonal variability and regulation of BP
Interannual variabilities in BP and PP are shown in Fig. 7. BP showed relatively high values in 2012 and 2015, and there was a statistically significant difference between those in 2012 and 2013 (Steel-Dwass multiple comparison test, p < 0.05). PP in 2012 was relatively high and statistically higher than those in 2014 and 2015 (Steel-Dwass multiple comparison test, p < 0.05). In the Delaware Estuary (U.S.A.), year-to-year variation (5 yr) in BP was apparently controlled by PP (using mean values; r = 0.94, n = 5, p < 0.01), and the ratio of the two productions ranged from 0.3 to 0.4, suggesting the fraction of PP processed by bacteria was relatively constant when averaged over a year (Hoch and Kirchman 1993). Contrary to the previous study, there was no significant relationship between BP and PP (r = 0.29, n = 4, p > 0.05), and the ratio of BP to PP ranged from 0.059 in 2013 to 0.24 in 2015 and varied > fourfold, suggesting the fraction of PP processed by bacteria was not constant in the present study. The DOC from phytoplankton to bacteria could vary depending on the presence of consumers (Jumars et al. 1989); their grazing activity would increase bacterial accessibility to organic substrates produced by PP and enhance BP (Güde 1988). In fact, NH 4 concentrations in 2012 and 2015 were relatively high, and BP was significantly positively correlated with NH 4 concentration (logarithmic regression; r = 0.958, n = 4, p < 0.05) despite limited sample number, suggesting that the higher concentration of NH 4 could be regarded as one of the proxies for higher activity of consumers as well as bacteria through their nutrient regeneration. Moreover, there was a significantly positive relationship between BP and cladoceran abundance over the year during the study period (r = 0.998, n = 4, p < 0.001), even though they did not covary tightly in each season. These results partly support that consumers play not only a topdown regulation on bacteria but also enhance bacterial activity through their grazing activities in Lake Kasumigaura.
In conclusion, empirical correlation analysis revealed seasonal change of abiotic/biotic factors involved in BP dynamics. The top-down regulation on bacteria was emphasized throughout the year (and in each season), but bottom-up regulation also controlled and explained bacterial dynamics through resource supply processes, including sloppy feeding, excretion, and feces by consumers in Lake Kasumigaura. The positive correlation between BP and Chl a concentration indicated higher relative importance of autochthonous substrate for bacterial activity in winter, and the negative correlation with precipitation implied that variations of BP were regulated by PP in the most productive season, summer, even in such a eutrophic lake. PP in autumn was approximately half of spring due to lower solar irradiance, although water temperature during both seasons was similar and nutrient concentrations during autumn were higher. BP in autumn was comparable with spring and BCD was comparable to PP, suggesting the relative importance of higher allochthonous substrates relative to other seasons. Our findings of the seasonal variability of BP regulation factors, the meteorological relation to BP, and the zooplankton-mediated substrate supply enhance our understanding of microbiological dynamics in eutrophic aquatic ecosystems and provide insight toward future consideration of biogeochemical processes even where water temperature is the primary regulation factor.