Taihu Lake (from the Chinese for “Great Lake”) is China's third-largest freshwater lake. Due to industrial development and urbanization in the watershed, Taihu Lake has shifted from a diatom-dominated, mesotrophic lake to a hypereutrophic lake experiencing cyanoHABs (Paerl et al., 2014; Qin et al., 2007). Historically, these blooms have been associated with toxin-producing, non-N2-fixing Microcystis spp., which can form surface scums on the lake for up to 10 months per year (Chen et al., 2003; Duan et al., 2009; Ma et al., 2016; Otten and Paerl 2011). The surface blooms have a well-documented negative impact on fisheries, tourism, and local economies, including a drinking water shutdown in 2007 (Qin et al., 2007; Steffen et al., 2017; Xu et al., 2010).

Taihu Lake is a large (2338 km2), shallow (mean depth =1.9 m) lake in southeast China, situated in the Yangtze river delta about 150 km west of Shanghai. The lake is an important source of freshwater and resources for the ∼ 40 million people within the watershed. Taihu Lake has a complicated hydrology, with 172 rivers and channels connected to the lake (Qin et al., 2007). This network of rivers carries nutrient loads from agricultural runoff, factories, and household wastewater. Taihu Lake has a relatively long residence time of approximately 280–300 days (Paerl et al., 2014; Xu et al., 2010).

Water samples were collected from four locations: stations 1 and 3 in Meiliang Bay, station 7 in the north-central part of the lake, and station 10 on the western side of the lake basin (Fig. 1). In previous studies (e.g., McCarthy et al., 2007), sampling stations 1, 3, and 7 followed a discharge gradient from the Liangxihe River in the northeast part of Meiliang Bay to the central lake, and station 0 (“river”) was located at the Liangxihe River discharge. However, in 2007, the Yangtze River was diverted into Taihu Lake in an effort to decrease the lake residence time and flush Microcystis spp. and nutrients out of the lake (Qin et al., 2010). Diverted water from the Yangtze River now flows into Gonghu Bay, the easternmost of the three northern bays. This diversion resulted in intermittent flow reversals through Meiliang Bay, where the Liangxihe River now mainly serves as an outflow. Since the discharge gradient from station 1 to 7 was no longer consistent in Meiliang Bay, station 0 was replaced with a new river input (station 10) on the western side of the lake near the Dapugang River mouth. Environmental variables (temperature; dissolved oxygen – DO; pH; total dissolved solids – TDSs; and chlorophyll a) were measured in situ at each site using a YSI 6600 multi-sensor sonde.

Water samples were collected in August 2013, June 2014, March 2015, and July 2016. Each of these sampling events corresponded with a pronounced Microcystis bloom at all sites (Ma et al., 2016; Deng et al., 2014; Li et al., 2017; Su et al., 2017; Qian et al., 2017), except stations 7 and 10 in March 2015 (visual observation). Our sampling dates were representative of seasonal conditions in the region, specific to this subtropical climate zone, and did not correspond with any extreme weather patterns (e.g., typhoons, droughts). Temperature and precipitation patterns were average for this climate region. Water was collected into 4 L carboys at the surface (top 20 cm) and near the bottom (approximately 2 m depth) to investigate any changes in nutrient dynamics associated with depth. Samples for nutrient analyses (NO3-, NO2-, o-PO43-, and urea) were filtered immediately in the field using 0.2 µm nylon syringe filters (GE Millipore) into 15 mL snap-cap tubes (Falcon) and stored frozen at -20 ∘C. Nutrient samples were analyzed on a Lachat QuikChem 8000 nutrient analyzer at the University of Texas Marine Science Institute (UTMSI; August 2013, June 2014) or a Lachat 8500 nutrient analyzer at Wright State University (WSU; March 2015, July 2016) according to manufacturer directions. Ambient NH4+ concentrations were determined by ammonium isotope retention time shift (AIRTS) high-performance liquid chromatography (HPLC) at UTMSI (Gardner et al., 1995). Briefly, the atom % 15N–NH4+ and total NH4+ concentration are determined by comparing the retention time shift of the sample relative to the natural abundance NH4+ standard (Gardner et al., 1996)

NH4+ uptake and regeneration rates were determined following the protocol of McCarthy et al. (2013). Water collected in 4 L carboys was returned to the Taihu Laboratory for Lake Ecosystem Research (TLLER) for isotope amendments and incubations. Five hundred milliliters from each site/depth was amended with 98 % 15NH4Cl (Isotec; concentration added 8–96 µM) and distributed into six (triplicates for light and dark) 70 mL, clear tissue culture bottles (Corning; McCarthy et al., 2007). The goal of the substrate additions in these uptake/regeneration experiments was to add more-than-trace levels to ensure that all of the label was not taken up during the incubations; our goal was to add the label concentration at an equivalent value to the most recent monitoring data we could obtain for NH4+ concentrations, or at least 8 µM (even when concentrations are low, recycling rates can be quite high). Dark bottles were wrapped with thick aluminum foil. Initial samples (T0) were withdrawn from each bottle with a rinsed syringe, filtered (0.2 µm filters) immediately into 8 mL glass vials (Wheaton), and frozen until analysis at UTMSI. Light and dark bottles were then submerged (approximate depth: 0.2 m) in a mesh bag at in situ light and temperature levels in the lake. After ∼ 24 h, final samples (Tf) were filtered in the same manner as the T0 samples. Total NH4+ concentrations and atom % 15N for all samples were determined by AIRTS HPLC (Bruesewitz et al., 2015; Gardner et al., 1995). Potential uptake and actual regeneration rates were calculated using the Blackburn–Caperon isotope dilution model (Blackburn, 1979; Caperon et al., 1979; McCarthy et al., 2013). The uptake rate is considered a potential rate, which includes nitrification, assimilation, and other consumption processes; regeneration is an actual rate that encompasses remineralization, decomposition of dead organic matter, heterotrophic excretion, respiration, biodegradation, and sloppy feeding by zooplankton (Saba et al., 2011).

Nitrification rates were measured directly using the 15NH4+ tracer addition method. Five hundred milliliters of water from each station and depth was distributed into 750 mL polycarbonate bottles, enriched with a tracer amount (approximately 20 % of the total pool) of 98 % 15NH4Cl (Isotec), mixed thoroughly by inverting 10 times, and distributed into three 125 mL polycarbonate incubation bottles. Unenriched samples for each station and depth were distributed into 125 mL incubation bottles. Initial samples (T0) were filtered using 0.22 µm syringe filters into 30 mL polycarbonate bottles and frozen until analysis. Final samples were collected as described after incubating for 24 h at in situ light and temperature levels. Samples were returned frozen to WSU for analysis.

Accumulation of 15NO2- was measured using the sodium azide (NaN3) reduction method (Heiss and Fulweiler, 2016; McIlvin and Altabet, 2005; Newell et al., 2011). Briefly, 7.5 mL from each sample was distributed into a 12 mL Exetainer vial (Labco, UK) and capped tightly. Each sample was then injected (with gastight syringe) with 0.25 mL of 1:1 (v:v) 2 M NaN3 : 20 % CH3COOH solution (previously purged with Ar for 30 min), followed by incubation for 1 h at 30 ∘C (McIlvin and Altabet, 2005). All NO2- accumulated in the sample from NH3 oxidation was transformed chemically to N2O. After 1 h, the reaction was stopped by injection of 0.15 mL of 10 M NaOH.

Accumulation of 15NO3- was measured using the Cd reduction/NaN3 reduction method (Heiss and Fulweiler, 2016). Approximately 25 mL from each sample was transferred into 50 mL centrifuge tubes. First, in situ NO2- was removed with 0.25 mL of 0.4 M sulfamic acid (H3NSO3). After 10 min, the reaction was neutralized with 0.125 mL of 2 M NaOH (Granger and Sigman, 2009). NO3- was reduced to NO2- by addition of 100 mg of MgO, 6.6 g of NaCl, and 0.75–1 g of acidified Cd powder to each sample, followed by 17 h incubation on a shaker table (McIlvin and Altabet, 2005). Samples were centrifuged at 1000× g for 15 min, and 7.5 mL of supernatant was carefully transferred into 12 mL Exetainers. Cadmium-reduced NO2- was further reduced to N2O with the previously described NaN3 method.

Samples were sent inverted to the University of California Davis Stable Isotope Facility for isotopic analysis of 45/44N2O using a Thermo Finnigan GasBench + PreCon trace gas concentration system interfaced to a Thermo Scientific Delta V Plus isotope-ratio mass spectrometer (Bremen, Germany). Nitrification rates were corrected for NaN3 reduction efficiency, and 15NO2- production was calculated as NH3Oxin nM day-1=15N/14N⋅NO2-24h-15N/14N⋅NO2-0h/α⋅t, where α=[15NH4+]/([15NH4+]+[14NH4+]).

15NO3- production was calculated as NO2-Oxin nM day-1=15N/14N⋅NO3-24h-15N/14N⋅NO3-0h/α⋅t, where α=[15NO2-]/([15NO2-]+[14NO2-]).

Total nitrification rates were calculated from the sum of 15NO2- and 15NO3- accumulation.

During the 2014–2016 sampling events, environmental DNA for AOO abundance was collected using 0.2 µm Sterivex filters (EMD Millipore, MA, USA) and preserved with Ambion RNAlater (Invitrogen, Carlsbad, CA, USA). Approximately 60–120 mL of site water was pushed through the filter for each station and depth and then stored filled with 5 mL RNAlater. Preserved filters were frozen at -80 ∘C and transported to WSU. DNA was extracted using the Gentra Puregene Kit (Qiagen Inc., USA) extraction protocol with slight modifications (Newell et al., 2011). Sterivex filters were first washed with phosphate-buffered saline 1X solution (Fisher BioReagents, USA) to remove any residual RNAlater. Lysis buffer (0.9 mL) and Proteinase K (4 µL) were added to the filters, followed by 1 h incubation at 55 ∘C and 1 h incubation at 65 ∘C. The solution was removed to a 1.5 mL tube, and the incubation was repeated with fresh lysis buffer and Proteinase K.

Concentration and purity of the DNA were measured spectrophotometrically (Nanodrop 2000, Thermo Scientific). AOA were targeted with Arch-amoAF and Arch-amoAR primers targeting the 635 base pair (bp) region of the amoA gene, subunit A of the ammonia monooxygenase enzyme (AMO; Francis et al., 2005). Bacterial amoA was quantified using amoAF and amoA2R primers (Rotthauwe et al., 1997) to target the 491 bp region of amoA. qPCR standards were prepared by cloning the fragment of interest for AOA and AOB with the TOPO TA Cloning Kit (Invitrogen, USA), inserting it into a competent cell plasmid (One Shot E. coli cells, Invitrogen, USA), and isolating the plasmid containing the amoA gene using the UltraClean Standard Mini Plasmid Prep Kit (Mo Bio Laboratories Inc., Carlsbad, CA, USA).

AOA and AOB qPCR assays were conducted within a single 96 well plate for each year (2014, 2015, and 2016). Each run included three negative controls (no template), five standards from serial dilution in triplicates, and the environmental DNA samples in triplicate. Each sample and standard received 12.5 µL of SYBR Green Fast Mastermix (Qiagen Inc., USA), 0.5 µL of each 10 µM primer, and 2–15 ng of template DNA.

All PCR work was performed in a PCR fume hood after cleaning the surface with DNA AWAY (Thermo Scientific, USA) and engaging the UV light (20 min) to prevent contamination. qPCR protocol followed the method of Bollmann et al. (2014) for AOA (95 °C initial denaturation for 5 min, 95 ∘C denaturation for 30 s, 53 ∘C annealing for 45 s, and 72 ∘C extension for 1 min; 45 cycles) and AOB (95 ∘C initial denaturation for 5 min, 95 ∘C denaturation for 30 s, 56 ∘C annealing for 45 s, 72 ∘C extension for 1 min; 45 cycles), followed by the melting curve. Automatic settings for the thermocycler (Realplex, Eppendorf) were used to determine threshold cycle (Ct values), efficiency (85–95%), and a standard curve with R2 values above 0.9. Gene copy number was calculated as (ng × number mol-1) / (bp × ng g-1 × g mol-1 of bp) and is reported in gene copies per milliliter of sample water. The detection limit was 980 copies ml-1 for AOB and 4807 copies mL-1 for AOA. These calculated detection limits do not represent the greatest sensitivity possible with our method, as the standard concentrations were selected to bracket the expected environmental concentrations. Indeed, our reported values are above the detection limit for both AOA (by 2 orders of magnitude) and AOB.

All statistical analyses were performed using RStudio software (R Version 3.3.1). Prior to statistical analysis, data were checked for normality using the Shapiro–Wilk normality test. The only variables that were normally distributed were DO, pH, and TDSs. To explore potential environmental drivers of the rates, a multivariate correlation analysis was performed using the Kendall correlation method for nonparametric data. A p value of < 0.05 was considered statistically significant. Additionally, stepwise multiple-regression models were run using the MASS package (R Version 7.3). The best-fitting model was selected based on the minimum Akaike's information criteria (AIC; Akaike, 1974). To normalize data for parametric analysis, all non-normally distributed variables were log⁡(x+1)-transformed prior to running the model.

Physicochemical parameters in Taihu Lake varied seasonally and spatially (Table 1). The most pronounced seasonal variations were observed in temperature and DO, with highest water temperature recorded in August. DO varied significantly, with highest values in March and lowest in August (p< 0.01). pH varied significantly with season, with lowest values in March and highest in August (p< 0.01). TDS values were highest in July 2016 and lowest in August 2013 (p< 0.001). Chlorophyll a concentrations were lowest in March 2015 (mean =11.1 µgL-1), but bloom conditions (> 20 µgL-1; Xu et al., 2015) were observed at some locations (e.g., 20.3 µgL-1 at station 3, and visual confirmation at stations 1 and 3 and several other areas of the lake). Bloom conditions were also present and observed at all sites in June 2014 (mean =36.6 µgL-1), July 2016 (mean =58.1 µgL-1), and August 2013 (43.7 µgL-1).

Ammonium concentrations remained high throughout all sampling events, with highest values in March 2015 and lowest values in August 2013, but differences were not statistically significant (p=0.125). Nitrite concentrations were not different between seasons, although they were significantly higher at station 10 than other stations (p< 0.001). Nitrate concentrations followed the pattern of NH4+ concentrations and were highest in March 2015 and lowest in August 2013 (p< 0.001). Orthophosphate concentrations followed a seasonal pattern with lowest concentrations in March and highest in August (p< 0.005), and o-PO43- concentrations at station 10 were significantly higher than at any other station (p< 0.001).

In August 2013, light uptake rates (all NH4+ uptake are potential rates) were uniform across sites (mean =0.40±0.04 µmolL-1h-1) and did not vary between surface and bottom waters (Fig. 2a). In June 2014, light uptake rates in surface waters at stations 1, 7, and 10 (mean =0.80±0.06 µmolL-1h-1) were significantly higher than deep rates (mean =0.31±0.08 µmolL-1h-1; p< 0.001). However, light uptake rates at station 3 did not differ from zero at either depth (Fig. 2a). Mean surface and deep uptake rates in the dark in August 2013 (0.25±0.01 µmolL-1h-1) and June 2014 (0.13±0.05 µmolL-1h-1) were significantly lower than light uptake rates (Fig. 2b; p< 0.05). In March 2015, light uptake rates at stations 1–7 (mean =0.12±0.04 µmolL-1h-1) were lower than those during August 2013 and June 2014 (mean =0.43±0.41 µmolL-1h-1) except for station 10, where the rates were significantly higher (mean =1.36±0.20 µmolL-1h-1; p< 0.001). In contrast to summer, dark uptake rates in March 2015 were not significantly different than light rates (Fig. 2b). In July 2016, light uptake rates were highest at stations 1, 7, and 10 (1.31–6.82 µmolL-1h-1). Stations 3 and 7 rates were highest in bottom waters (0.80±0.16 and 2.55±0.14 µmolL-1h-1, respectively). In July 2016, light and dark uptake rates did not differ significantly (p=0.15); highest dark uptake rates were observed at station 1 in surface water (3.33±0.67 µmolL-1h-1). Light uptake rates, across all stations and seasons, correlated positively with TDSs and NH4+ : NO3- and negatively with pH, while dark uptake rates correlated positively with TDSs, NH4+, and NH4+ : NO3-, and negatively with pH (Table 2).

Regeneration rates in the light and dark (all NH4+ regeneration rates are actual rates, not potential) were not significantly different from each other across all years and seasons; therefore, light and dark rates were averaged together (Fig. 2c). Regeneration rates did not differ significantly between the summer bloom sampling events in August 2013 and June 2014 (mean =0.22±0.03 µmolL-1h-1), but July 2016 regeneration rates (mean =0.75±0.16 µmolL-1h-1) were significantly higher than in August and June (p=0.004), with exceptionally high regeneration rates occurring in surface waters in July at station 1 (mean =2.37±0.16 µmolL-1h-1). In March 2015, mean surface and deep regeneration rates decreased from the river mouth (station 10; 0.88±0.15 µmolL-1h-1) towards the center of the lake, with significantly higher regeneration rates at station 10 than stations 1–7 (mean =0.10±0.03 µmolL-1h-1; p< 0.01). Regeneration rates were positively correlated with TDSs, NH4+, and o-PO43- concentrations, and NH4+ : NO3- (Table 2).

Note that nitrification rates are presented in units of nanomoles per liter per day (nmol L-1 d-1) for consistency with literature-reported values (Fig. 3). At stations 1, 3, and 7 15NH4+ additions, 91.8 % of the label was detected as 15NO3- and only 8.2 % as 15NO2- (Fig. 3a). Total nitrification rates at station 3 did not vary across seasons. At station 7 in the central lake, highest total nitrification rates were observed in March 2015 (mean =663±69.4 nmol L-1 d-1) in both surface and deep waters compared to the lowest rates in July 2016 (mean =1.58±0.78 nmol L-1 d-1). At station 1, the highest rates were measured in surface waters in July 2016 (mean =773±50.7 nmol L-1 d-1), but the rates at depth followed a seasonal pattern from high in the spring (mean =646±158 nmol L-1 d-1) to an order of magnitude lower in the summer (mean =9.86±3.28 nmol L-1 d-1).

Total nitrification rates at station 10 were significantly higher than other stations (Fig. 3b; p< 0.001). Rates were, at times, orders of magnitude higher, and total nitrification ranged 148–3750 nmol L-1 d-1 (mean =1590±1390 nmol L-1 d-1), compared to stations 1–7 ranging 2.00–771 nmol L-1 d-1 (mean =270±277 nmol L-1 d-1). At station 10 in July 2016, 80 % of the 15NH4- addition was detected as 15NO2-.

Abundance of the bacterial amoA gene for all years (2014–2016) varied from undetectable to 2.85×105±5.20×104 copies mL-1. Archaeal amoA abundance ranged from undetectable to 1.03×107±3.37×106 copies mL-1 (Fig. 4a). Neither AOB nor AOA amoA gene copy abundances were statistically different between the three seasons. The highest ratio of AOB : AOA gene abundance (1.81) was reported at station 3 in Meiliang Bay (Fig. 4b), and the lowest ratio (0.01) was observed at station 7. AOB gene abundance was positively correlated with NH4+, NO2-, and o-PO43- concentrations, and NH4+ : NO3-, while AOA gene abundance was not significantly correlated with any environmental variable (Table 2).

Ammonium uptake rates (0.02–6.82 µmolL-1h-1) reported here were within the range of or slightly higher than rates reported in other studies (Table 3). Rates were higher than uptake rates reported previously in Meiliang Bay (0.11–1.54 µmolL-1h-1) and the central lake (0.03–0.32 µmolL-1h-1) but within the range of rates reported in the Liangxihe River (0.70–4.19 µmolL-1h-1; McCarthy et al., 2007). Light uptake rates in March, June, and August resembled rates in eutrophic Lake Okeechobee but were higher than rates in Missisquoi Bay (Lake Champlain), Lake Michigan, and eutrophic New Zealand lakes Rotorua and Rotoiti (Table 3 and references therein). Higher light uptake rates were reported only in hypereutrophic Lake Maracaibo, Venezuela (Table 3), and in Maumee Bay, Lake Erie, during a summer cyanoHAB bloom (Gardner et al., 2017). Potential NH4+ uptake rates in these systems, evaluated using the same methods, increase with chlorophyll a (p< 0.05), but the proportion of community uptake that can be supported by regeneration remains relatively consistent (Table 3).

Light uptake rates in Taihu Lake were marginally higher (p=0.08) than dark uptake rates, presumably due to reduced photosynthetic phytoplankton activity. Photoautotrophs may continue to assimilate nutrients in the dark under nutrient limitation (Cochlan et al., 1991), but Taihu Lake is generally nutrient replete, so we assume that dark uptake rates can be attributed mostly to heterotrophic or chemolithoautotrophic organisms. Uptake rates were significantly higher in July 2016 than at other times, which may have been due to higher precipitation and subsequent runoff; during summer 2016, average rainfall in June and July was about 305 mm compared to 106 mm in June 2014, 105 mm in August 2013, and 54 mm in March 2015 (https://www.worldweatheronline.com/; accessed on 8 February 2017); however, it is within the range of typical summer rainfall (185–320 mm; https://www.worldweatheronline.com/). Dark uptake rates in Taihu Lake exceeded dark rates reported in Lake Okeechobee (0.02–0.04 µmolL-1h-1; James et al., 2011), Missisquoi Bay, Lake Champlain (0.10 µmolL-1h-1; McCarthy et al., 2013), and Lake Michigan (7 nmol L-1 h-1; Gardner et al., 2004), suggesting high activity of both heterotrophs and chemolithoautotrophs in Taihu Lake. A previous metagenomics study of the bloom composition in Taihu Lake revealed an overlooked contribution of heterotrophic bacteria to N assimilation processes by Microcystis, which could be important in driving toxic blooms (Steffen et al., 2012).

Internal NH4+ cycling via regeneration is important in Taihu Lake and varies seasonally (McCarthy et al., 2007; Paerl et al., 2011). In March 2015, about 38 % of light uptake for all sites and depths was supported by regeneration (Fig. 2d). This proportion increased in June 2014 and July 2016 to 58 and 42 %, respectively, and was highest in August 2013 (109 %). The importance of regeneration corresponded to decreasing in situ NH4+ concentrations (Fig. 2d). These results suggest that, in March and June, regeneration supplemented ambient NH4+ in the water column to support algal production, whereas cyanobacteria relied more heavily on NH4+ from regeneration to sustain blooms in July and August. Water column regeneration may supply more NH4+ for blooms than sediment NH4+ regeneration in Taihu Lake due to combined spatial, temperature, and biogeochemical factors (McCarthy et al., 2007; Gardner et al., 2017). Rapid decomposition of cyanoHAB biomass may provide NH4+ for nitrification, which provides substrate for denitrification. High rates of sediment denitrification (McCarthy et al., 2007) also may drive N limitation in late summer and fall (Paerl et al., 2011; Xu et al., 2010)

To calculate whole-lake water column NH4+ regeneration and uptake rates, we divided the lake (2338 km2; Qin et al., 2007) into four different sections based on geochemical and ecological properties (Qin, 2008): (1) three northern bays (361.8 km2; depth =1.9 m) most affected by the blooms; (2) the main lake (1523.9 km2; depth =1.9 m); (3) the east Taihu Lake region, dominated by rooted and floating macrophytes (357.5 km2; depth =1.4 m); and (4) shorelines < 1 m deep (94.8 km2). We considered regeneration and uptake rates from stations 1 and 3 to represent the northern bays area, station 7 as the main lake, station 10 as shoreline, and regeneration rates previously reported for east Taihu Lake (McCarthy et al., 2007; Paerl et al., 2011). When extrapolated to the volume of these four zones in Taihu Lake, regeneration returned about 3.04×107 kg of NH4+ annually in the three northern bays, 6.71×107 kg of NH4+ in the main lake, 8.87×106 kg of NH4+ along the shorelines, and 2.88×106 kg of NH4+ in east Taihu Lake. These values sum to 1.09×108 kg of NH4+ recycled in the water column, approximately 2 times higher than reported external N loadings, which range from 5.11×107 to 7.00×107 kg annually (Chen et al., 2012; Yan et al., 2011). The same procedure for extrapolation of whole-lake uptake rates yields 3.5×108 kg of NH4+, which is 4–6 times higher than external N loads. The combination of external loads and regeneration cannot support the demand for NH4+, suggesting that the remaining NH4+ demand must be satisfied by internal loads from sediments or some other unknown source, or that reported total nitrogen (TN) loads are underestimated. These rough estimates of lake-wide regeneration and uptake are based on rates measured at specific stations at discreet times; improved spatial and temporal resolution of measurements is needed to improve these estimates. Additionally, these calculated values are probably an overestimate given that most of the rates measured and reported in this study are during spring and summer months, not fall and winter, when we might expect lower rates. Taihu Lake is a complex ecosystem with 172 rivers and channels connected to the lake (Qin et al., 2007), making any estimations of total N loadings challenging. As such, we believe that the reported total N loads to Taihu Lake are likely an underestimate. However, our results show that these external N loads lead to higher biomass and fuel high regeneration rates. Combined with high ambient nutrient concentrations, these data suggest that microbial denitrification cannot remove N fast enough to keep pace with external N loading. Increasing nutrient loads can result in decreasing efficiency of denitrification (Gardner and McCarthy, 2009; Mulholland et al., 2008), which will limit the ability of a system to self-mitigate excess N loads.

Total nitrification rates reported in this study exceeded previously reported rates in most oligotrophic and mesotrophic freshwater systems. Published nitrification rates in lakes include the water columns of saline Lake Mono, CA, USA (60–480 nmol L-1 d-1; Carini and Joye, 2008), and Lake Superior, USA (0–51 nmol L-1 d-1; Small et al., 2013), both measured via 15NH4+ tracer additions, and Lake Okeechobee, FL, USA (67–97 nmol L-1 h-1; James et al., 2011), measured via the 15NO3- pool dilution method (Carini et al., 2010). Rates on this scale were previously reported only in eutrophic Lake Mendota (WI, USA; 1700–26 000 nmol L-1 h-1; Hall, 1986) and the Pearl River estuary (China; 2100–65 100 nmol L-1 d-1; Dai et al., 2008). However, these rates were measured from accumulation of NO2- and NO3-, not stable-isotope additions. High total nitrification rates in Taihu Lake can be attributed to high ambient NH4+ concentrations, up to 40 µM at station 1 in 2016 and 135 µM at station 10 in 2014. These high concentrations of NH4+ are due to high external N loadings, including N in organic matter, into the lake, of which ∼ 1.32×107 kg was loaded as NH4+ in 2009 (Yan et al., 2011). The significant relationships between nitrification and NH4+, NO2-, and NO3- concentrations (p< 0.05; Table 2) support these observations.

Substrate concentrations drive NH4+ oxidation rates and, therefore, end-product pools, since it is the rate-limiting step of nitrification (i.e., completion of nitrification is dependent on the first step). Accumulation of 15NO3- exceeded accumulation of 15NO2- by a factor of 9 at stations 1, 3, and 7 across all sampling events (Fig. 3a), indicating that NO2- oxidation is keeping pace with or exceeding NH4+ oxidation. Higher accumulation of 15NO3- was expected, since NO3- is the final product of total nitrification.

At station 10, accumulation of 15NO3- exceeded 15NO2- in March 2015 and June 2014. In July 2016, however, accumulation of 15NO2- was 3 times higher in surface water and comparable at depth (Fig. 3b). Ambient NO2- concentration at station 10 in July 2016 was 9.6 µM in surface water and 8.4 µM at depth (Table 1). This accumulation of NO2- suggests that NO2- oxidizers were saturated, consistent with Km values reported for NO2- oxidation in the oligotrophic open ocean of 0.25±0.16 µM (Sun et al., 2017). However, culture experiments report Km values ranging from 6 to 544 µM for Nitrospira, Nitrobacter, and Nitrotoga spp. (Blackburne et al., 2007; Nowka et al., 2015; Ushiki et al., 2017).

At most stations, nitrification rates in Taihu Lake were highest in March, lower in June, and lowest in July. During the spring sampling, nitrification accounted for about 8 % of light uptake and 15 % of dark uptake at stations 1–7. In June, nitrification accounted for 2.6 % of light uptake and 9.6 % of dark uptake, and in July only 0.2 and 0.3 % of light and dark uptake, respectively. These results show a seasonal trend of decreasing contribution of nitrification to total uptake rates and higher contribution of nitrifiers to dark uptake. As stated above, chemolithoautotrophs (including nitrifiers) do not rely on light for energy and continue to assimilate NH4+ in dark conditions, while photoautotrophic cyanobacteria generally assimilate NH4+ in the dark only when nutrient limited (Cochlan et al., 1991). However, the presence of high dissolved-inorganic-N concentrations in ambient water samples suggests that the observed dark uptake was likely performed primarily by non-photoautotrophs, including nitrifiers.

We observed no significant seasonal change in nitrification across all stations and no consistent pattern between temperature and nitrification. While the lack of relationship of nitrification with temperature agrees with nitrification studies in the ocean (Ward, 2008), other studies have reported temperature as a potential driver of nitrification in coastal waters (Heiss and Fulweiler, 2016). Although not statistically linked to changes in temperature, the contribution of nitrification to total uptake rates decreased in summer months, likely as a result of competition with the Microcystis bloom and associated heterotrophic bacteria. Non-N2-fixing cyanobacteria, including Microcystis, are exceptional competitors for NH4+ in high-nutrient environments (Blomqvist et al., 1994). With a high saturation threshold and reported Km values from 26.5 to 37 µM (Baldia et al., 2007; Nicklisch and Kohl 1983) in culture, and up to 112.9 µM in Taihu Lake populations (Yang et al., 2017), Microcystis should be able to outcompete nitrifiers at the high ambient NH4+ concentrations in Taihu Lake as nitrifiers may become saturated at much lower concentrations. Additionally, Microcystis can regulate its buoyancy and scavenge nutrients throughout the water column to effectively compete for light with other phytoplankton (Brookes and Ganf, 2001).

Nitrification at station 10 differed dramatically from other stations. Total nitrification rates were, at times, orders of magnitude higher than at other stations. Also, station 10 did not follow the trend of decreasing nitrification contribution with the bloom. Nitrification accounted for 19 % of light uptake and 64.8 % of dark uptake in June and only 1.7 and 2 %, respectively, in March. We speculate that station 10 differs from other stations because of the large nutrient and suspended-particle loads from the Dapugang River, the second largest inflow into the lake (Yan et al., 2011). Suspended particles from sediments could trigger heterotrophic and anaerobic processes at station 10, including reduction of NO3- to NO2- (Krausfeldt et al., 2017; Yao et al. 2016). In fact, denitrification and anammox gene transcripts were observed recently in the water column at station 10 (Krausfeldt et al., 2017). These authors also speculated that the discharge of suspended sediments from the river might play a role in coupling anaerobic and aerobic processes in the turbid water column, resulting in rapid cycling of reduced and oxidized forms of N. Nitrification is the link between introduction of reduced N into the system and the removal of N through denitrification. Therefore, the efficiency of nitrification is crucial to the removal of N from this hypereutrophic lake.

AOB and AOA coexist in the environment, and environmental variables shape the community structure. AOA often dominate in environments with low substrate concentrations, such as the open ocean or oligotrophic lakes (Beman et al., 2008; Bollmann et al., 2014; Newell et al., 2011), while AOB are often more abundant in nutrient rich waters and soils (Hou et al., 2013; Jia and Conrad, 2009; Kowalchuk and Stephen, 2001; Verhamme et al., 2011). This substrate concentration adaptation is dictated by different physiological abilities to assimilate NH4+. Culture studies show that AOA have a very high affinity (low half-saturation constant; Km) for NH4+ and in general are saturated faster than AOB (Martens-Habbena et al., 2009). The low half-saturation constant (Km=0.132 µM; Martens-Habbena et al., 2009) of AOA gives them a competitive advantage in low-NH4+ conditions. In contrast, the high Km of AOB (10–1000 µM) allows them to assimilate more NH4+ before becoming fully saturated, an advantage for higher-NH4+-concentration conditions. Although oligotrophic AOA appear to proliferate in the environment (Francis et al., 2005), some species adapt to higher substrate concentrations (Jung et al., 2011; Tourna et al., 2011).

Results from the amoA gene copy abundance analysis show that AOA were more abundant than AOB across all stations and seasons in Taihu Lake. Although this result does not support our original hypothesis, the results agree with previous studies in the water column and sediments in Taihu Lake (Zeng et al., 2012), which reported higher AOA abundance (4.91×105–8.65×106 copies g-1 sediment) than AOB (3.74×104–3.86×105 copies g-1 sediment) in Meiliang Bay. Similarly, another Taihu Lake sediment study showed more AOA than AOB in sediments at all 20 investigated stations (Wu et al., 2010).

The differences in abundance of AOOs between stations, represented as AOB : AOA, show spatial variability between the more nearshore and central lake stations (Fig. 4b). In this study, AOA were more abundant in the central lake (station 7), whereas AOB were more abundant closer to shore. Due to a higher affinity for substrate (lower Km), AOA are likely more competitive when nutrient concentrations are lower, such as in the open lake (mean offshore NH4+ concentration =3.69 µM). In contrast, AOB, with higher Km, thrive at higher NH4+ concentrations at nearshore locations (mean nearshore NH4+ concentration =31.3 µM). These results agree with previous research in Taihu Lake, where AOA outnumbered AOB in sediments at mesotrophic sites and AOB were more abundant at hypereutrophic locations (Hou et al., 2013). Another study on Taihu Lake sediments also reported that both AOA abundance and AOA : AOB were negatively correlated with ambient NH4+ concentration (Wu et al., 2010). However, the data reported in this study show no significant relationship between AOA abundance and NH4+, NO2-, and NO3- concentrations (Table 2).

Despite AOA outnumbering AOB, AOB abundance was correlated with total nitrification rates for all stations and all seasons (p< 0.005), but AOA abundance was not. This result agrees with a previous study on Taihu Lake sediments, where AOA were negatively correlated (r=0.53, p< 0.05) with potential nitrification rates (0–3.0 µgNO3-Ng-1 dry sediment; Hou et al., 2013). We speculate that AOA oxidized NH4+ at lower rates due to oversaturation and inhibition and may not have contributed as much as AOB to nitrification rates in our study. This conclusion was also reached in Plum Island Sound (MA, USA), where abundance of archaeal amoA was higher than bacterial amoA, but potential nitrification rates did not correlate with AOA (Bernhard et al., 2010). The authors hypothesized various scenarios, including inhibition of AOA due to high substrate concentrations, competition for NH4+ with AOB, or AOA using an alternative energy source (Bernhard et al., 2010). Our results support the interpretation that AOA are at a disadvantage when competing with AOB for NH4+ in a hypereutrophic system and most likely did not play a major role in observed nitrification in Taihu Lake. Recent studies show that AOA can oxidize cyanate (Palatinszky et al., 2015) and urea (Tolar et al., 2016), although growth and oxidation rates may be slow. Therefore, AOA may play an expanded role in Taihu Lake, beyond just NH4+ oxidation.

The best-fitting multiple-regression models for N dynamics in Taihu Lake (Table 4) supported the Kendall non-parametric analysis (Table 2). Ammonium uptake and regeneration rates and nitrification were correlated with ambient NH4+, NO2-, and NO3- concentrations. Additionally, the best-fitting models revealed that variables changing with season had major influences on the models (Table 4). For example, uptake in the light and dark and regeneration rates were positively influenced by temperature and DO and negatively by pH. However, the model for nitrification rates did not reveal that the seasonal variables, such as temperature, played a major role in the model.