The experiment (Aug. 14 to Oct. 8, 2020) was conducted in 6 glass aquaria (45 cm × 30 cm × 45 cm), using seawater and sediment from Xiangshan Bay of ECS (N 29°35´28.55″, E 121°54´54.55″). Sediment (TN_Sed, 0.66 ± 0.04 mg g⁻¹ dw; TP_Sed, 0.59 ± 0.00 mg g⁻¹ dw; OM_Sed, 32.37 ± 2.68 mg g⁻¹ dw) was collected from a depth of 0–10 cm below the water–sediment interface and subsequently mixed and added to the 6 aquaria to obtain a sediment layer of 10 cm. The remaining 35 cm was filled with well-mixed seawater (TN, 0.5 mg L⁻¹; TP, 0.04 mg L⁻¹; salinity, 23.4‰) using an immersion pump. Extra seawater was stored synchronously in black tanks to compensate for the water loss during the experiment due to evaporation and sampling. The experiment was run in the glasshouse with a light intensity of 1,528–2,094 lux and an air temperature of 26°C–32°C. Circulating water pumps (power, 15 W) were used to simulate the natural mixing regime according to the method by Ma et al. (2021).

Two treatments with three replicates were established: Control without sodium acetate [Na(CH₃COO)] addition (coded as Control) and Na(CH₃COO) addition treatment (+Carbon, equivalent to 5 mg C L⁻¹). DOC < 5 mg L⁻¹ is one of the threshold parameters for evaluating aquatic ecosystem quality in the US National Lakes Assessment (NLA) program (NLA, 2012). The first Na(CH₃COO) addition (0.75 g) was carried out on day 3 (Aug. 16) to obtain the target C concentration. The Na(CH₃COO) was dissolved in water from the aquarium before being injected evenly into the middle water layers by a sprayer (0.5 L). To maintain the target level of C concentration, C source was further added once a week on day 11 (Aug. 24), day 18 (Aug. 31), day 25 (Sep. 7), day 33 (Sep. 15), day 39 (Sep. 21), day 46 (Sep. 28), and day 52 (Oct. 4). Total doses of 1.7 g of Na(CH₃COO) were applied in +Carbon treatment. The amount of C needed (F, mg) was calculated based on the difference between the target concentrations and the mean measured concentrations of the three replicates in each treatment before C addition: F = ( T − M   ) × V where T is the target C concentration (mg L⁻¹), M is the mean measured C concentration of three replicates (mg L⁻¹), and V is the volume of aquarium seawater (L).

Water temperature (WT), DO, redox potential (Eh), and pH at middle water depth were measured 7 times (1–2 weeks’ interval) with a Horiba (U-52, Kyoto, Japan). Water samples were taken 7 times at the middle water layer (1–2 weeks’ interval) at three randomly chosen locations within each aquarium using a 0.5-L syringe. In total, 0.5 L of well-mixed water was collected for chemical and biological analyses. All the chemical parameters were determined according to standard methods (Administration of Quality Supervision, Inspection and Quarantine (AQSIQ), 2007). Chl a and APA in the water samples were determined by applying the method of Ma et al. (2018), in which total APA (APA_Tot) was separated into algal APA (APA_Alg) and bacterial APA (APA_Bac).

Labile-P distribution (termed as easily changeable or mobile P fractions, including phosphate and forms of P loosely adsorbed to sediment solids) in sediment profiles was determined using Zr-oxide diffusive gradients in thin films (DGT-labile P). Zr-oxide DGT probes (Easysensor Co. Ltd, Nanjing, China) assembled with standard DGT holders were inserted through the sediment–water interface (SWI) on day 1 (pre-C addition, Aug. 14) and day 56 (post-C addition, Oct. 8) by hand. The probes were forced 10 cm into the sediment and kept 4 cm above the water surface. After 24 h, the probes were retrieved and brought to the laboratory for analyses. Further details on the data processing are given by Ding et al. (2015). The flux of P (Flux-P) across the SWI was calculated as follows: F = F w + F s = − D w ( δ C D G T δ χ w ) ( χ = 0 ) − φ D s ( δ C D G T δ χ s ) ( χ = 0 ) where F _w and F _s are the fluxes of P from the overlying water to the SWI and from the sediment to the SWI, respectively. F is the flux across the SWI. (δC_DGT/δχ_w)_{(χ= 0)} and (δC_DGT/δχ_w)_{(χ= 0)} are the DGT-labile P concentration gradients (i.e., the slopes) in the overlying water and sediment, respectively. φ is porosity in sediment and was estimated at 0.9 in the top 5-mm layer. D _w and D _s are the diffusion coefficients of H₂PO₄ ⁻ in water and sediment, respectively. D _s was calculated from the diffusion coefficient of H₂PO₄ ⁻ in water using φ ³ for φ ≥ 0.7 (Ding et al., 2015).

The top 5-cm sediment pH (pH_Sed) was measured with a soil pH meter (pH400 and pH600, USA). Sediment samples of the top 5 cm from three random locations in each aquarium were collected twice on day 1 (Aug. 14) and day 56 (Oct. 8) using a core sampler (XDB0204, New Landmark, Beijing, China). The sediment samples were mixed and partitioned into three subsamples: one was stored at −80°C for total microbial DNA extractions and bacterial community composition analyses; the second was partially air-dried and passed through a 2-mm sieve for total nitrogen (TN_Sed), total phosphorus (TP_Sed), and organic matter (OM_Sed) analyses; and the third was stored at 4°C for analyses of chlorophyll a of benthic algae (Chl a _Ben) within 24 h. TN_Sed, TP_Sed, and OM_Sed were measured following the standard methods (Lu, 1999). Chl a _Ben was analyzed according to Boer et al. (2009).

DNA in water and sediment was extracted using the FastDNA spin kit for soil (Q-BIOgene, Carlsbad, CA, USA). The genomic DNA concentration and purity were measured using an Eppendorf Biophotometer plus (Eppendorf, Hamburg, Germany). The bacterial community composition was assessed by sequencing the V3–V4 region of the 16S rRNA gene using the PCR primers 338F (5′-ACT​CCT​ACG​GGA​GGC​AGC​A-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). High-throughput sequencing was carried out on the Illumina MiSeq platform (BioMarker Technologies Co. Ltd, Chuangyeyuan, China). The results were processed according to established methods (Han et al., 2019).

Origin 9.0 and SPSS 20 software were used for data processing, and the results were expressed as means and standard errors. Repeated-measures ANOVA was used for statistical comparisons, and statistical significance was accepted at p < 0.05.

The initial conditions before C addition were comparable. After C addition, total carbon concentration (TC) was significantly higher in the +Carbon treatment than the Control (p < 0.001) (Figure 1A). Similarly, the total inorganic carbon (TIC) levels were significantly higher in the +Carbon treatment than in the Control (p < 0.001). However, no significant difference in total organic carbon (TOC) was observed between treatments (p = 0.06), indicating a rapid conversion of TOC to TIC (Figures 1B,C).

After C addition, TP was significantly higher in the +Carbon treatment than the Control (p = 0.03) (Figure 1D). TDP% did not differ significantly between treatments (p = 0.06), although it tended to be decreased in the +Carbon treatment (Figure 1E). For DGT-labile P in 7-cm-deep sediment, there were notable increases after C addition relative to the initial condition in the +Carbon treatment (p = 0.01), while minor increases were observed after C addition in the Control (p = 0.22) (Figure 2). The DGT-labile P values recorded for overlying water 3 cm above the sediment decreased clearly after C addition in both the Control (p < 0.001) and +Carbon treatments (p < 0.001) (Figure 2). In terms of Flux-P across the SWI, no obvious difference appeared between pre-C (0.01 μg cm² s⁻¹) and post-C addition (0.01 μg cm² s⁻¹) for the Control, while a noticeable increase from 0.01 to 0.02 μg cm² s⁻¹ was traced for + Carbon treatment (Figure 2).

DO was significantly lower in the +Carbon treatment than in the Control (p = 0.006) (Figure 3A). Similarly, the Eh levels were significantly lower in the +Carbon treatment than in the Control (p = 0.006) (Figure 3B). No significant difference was found between treatments for pH (p = 0.06), although it tended to be higher with C addition (Figure 3C).

APA_Tot tended to increase with C addition and reached its peak (1.9 μg P L⁻¹ h⁻¹), although there was no statistical difference between treatments (p = 0.07) (Figure 3D). There was no significant difference in Chl a between treatments (p = 0.70), although Chl a tended to be higher in the +Carbon treatment (on average 1.65 times higher than the Control) (Figure 3E). No significant differences in APA_Alg, APA_Bac, and DO_SWI were observed between before or after C addition (p = 0.06–0.88), although the values tended to be higher (APA_Alg and APA_Bac) or lower (DO_SWI) in the +Carbon treatment (Figures 4A–C). pH_Sed was significantly lower in the +Carbon group than the Control after C addition (p < 0.001) (Figure 4D).

The most abundant classes of the bacteria in water were Oxyphotobacteria (22%–43%), Alphaproteobacteria (17%–31%), Gammaproteobacteria (9%–18%), Bacteroidia (7%–12%), and Acidimicrobiia (0.7%–15%) (Figure 5A). At the genus level, 7%–30% of the bacteria in water were unculturable (Figure 5A), while 45%–59% were classified as “others” (<1% in relative abundance). The most abundant genera of bacteria shifted from Cyanobium (22%) to Leptolyngbya (14%) after C addition (Figure 5B).

The most abundant classes of bacteria in sediment were Gammaproteobacteria (16%–25%), Bacteroidia (9%–16%), Actinobacteria (8%–14%), Alphaproteobacteria (8%–13%), and Clostridia (4%–15%) (Figure 5C). At the genus level, three distinct genera (with a relative abundance >1% in at least one sample) were detected before C addition, Candidatus sulcia sp. (6.7%), Candidatus vidania sp. (5.4%), and Woeseia sp. (2.5%). After C addition, C. sulcia sp. (6.3%), C. vidania sp. (5.9%), and Bacteroides sp. (3.2%) were detected. A noticeable increase in abundance of 2.3% was observed from pre- to post-C addition for Bacteroides sp. (increased by 2.3% in +Carbon) (Figure 5D).