Introduction

In recirculating aquaculture systems (RAS), feed—through fish digestion (and potentially feed spill)—is the main source of nutrient input. The ingredients and macronutrient composition of a diet (i.e., the content of proteins, lipids, and carbohydrates), the nutrient digestibility by fish, and the feed conversion ratio (FCR) obtained determine the form, quantity, and composition of the nutrient waste produced (Bureau and Hua 2010; Cho and Bureau 2001). In terms of the form of nutrient waste, indigestible nutrients are excreted as particulate waste (i.e., fecal waste), while digested but non-retained nutrients are excreted as dissolved and suspended waste (Dalsgaard and Pedersen 2011). Correspondingly, fecal waste is characterized by high concentrations of organic carbon (C) and comparatively low concentrations of nitrogen (N) when using high-performance and energy-dense diets. Dissolved and suspended waste, on the other hand, is characterized by high concentrations of N (principally ammonia-N and urea-N derived from protein metabolism) and more moderate concentrations of organic C as compared to fecal waste (Dalsgaard and Pedersen 2011, 2016).

In RAS, nutrients in dissolved and suspended waste as well as unremoved fecal waste will accumulate to equilibrium levels, which are determined by the relative amount of make-up water (i.e., the feed loading) and microbial degradation and assimilation activity (Pedersen et al. 2012). Given the close relationship between diet composition and waste excreted, changes in dietary C:N ratios could be expected to be reflected on C:N ratios in RAS water and/or on physicochemical water quality parameters, such as the concentration of inorganic nitrogen compounds and organic matter. Indeed, the impacts of dietary C:N ratios on water quality and water C:N ratios have been described in biofloc technology systems (Avnimelech 1999; Xu and Pan 2013). However, dietary effects on water quality in RAS are largely unknown, and a better understanding of the extent to which changes in the relative composition of dietary macronutrients and the associated production of nutrient waste affect the physicochemical water quality is important to improve RAS water and waste management.

Diet-induced changes in the availability of C and N in water could also lead to changes in the activity and composition of RAS microbial community. Ecological studies have found the optimum molar C:N ratios for aquatic microbes to vary between 3.6:1 and 12.7:1 (Goldman and Dennett 2000; Vrede et al. 2002). However, microbial communities in RAS are expected to be C-limited rather than N-limited due to an excessive accumulation of nitrate–N (NO3−-N) in the system (Leonard et al. 2002) and the majority of microbes present in water being heterotrophs, which use organic carbon as their energy source (Fossmark et al. 2020). Previously, decreasing the molar water C:N ratios from 0.56 to 0.33 by using membrane filtration to remove organic matter has been found to alter the microbial community composition and decrease the abundance of heterotrophic microbes (Bugten et al. 2022; Fossmark et al. 2020); but there is no knowledge on the effects of directly modifying the dietary C and N content. In addition to potentially affecting C-limited microbial communities in RAS water, diet-induced changes in water C:N ratios could affect biofilter communities. In two studies conducted with laboratory-scale biofilters, nitrification rates were shown to decrease by 50% when increasing molar C:N ratios from 0 to 0.6 (Ling and Chen 2005), and by 70% when increasing the ratio from 0 to 1.2 (Zhu and Chen 2001). This was explained with higher C availability promoting heterotrophic growth that would reduce the activity of autotrophic nitrifiers. However, in real-life RAS conditions, diet-induced changes in water C:N ratios and other physicochemical parameters might be quenched before biofilters, due to the high activity of (heterotrophic) microbes in the water phase.

This study aimed at building a better understanding between dietary C:N ratios and physicochemical water quality and associated microbiology in RAS. We speculated whether changes in dietary C:N ratios would affect elemental C:N ratios in the system and thereby the C:N ratios available to microbes, potentially affecting microbial community structure and/or abundance. This was studied by changing the relative proportions of proteins, lipids, and carbohydrates in four rainbow trout (Oncorhynchus mykiss) diets and feeding them to juvenile rainbow trout in triplicate pilot-scale RAS for 4 weeks. The microbial compositions were examined at both biofilter and water levels.

Materials and Methods

Feed compositions and system setup and operation

Using the same ingredients but adjusting the proportions of protein, fat, and carbohydrate, four diets were formulated to fully embrace the range of macronutrients and C:N ratios found in commercial diets for juvenile rainbow trout. Four experimental diets were named according to their C:N ratio (as per wet weight): CN8.5, CN9.3, CN10.0, and CN12.2 (equivalent to molar C:N ratios of 9.9, 10.8, 11.6, and 14.2, respectively) (Table 1). The diets, extruded as 3-mm pellets, were formulated and produced by Biomar A/S, Denmark.

Table 1 Ingredients, proximate compositions, and C:N ratios of the experimental feeds produced by Biomar A/S
Full size table

The four diets were administered in triplicate to 12 identical pilot scale freshwater RAS at DTU Aqua, Hirtshals, Denmark. System configuration is illustrated in de Jesus Gregersen et al. (2021). Each system was composed of a 500-l fish rearing tank, a 140-l water sump, and a 75-l moving bed biofilm reactor (MBBR) filled with 40 l bioelements (750 m2/m3) (RK BioElements, Denmark). In 3 weeks of acclimation, a mix of the four diets was given to all systems stocked with juvenile rainbow trout. Afterwards, each system (fish biomass 7.31 ± 0.04 kg) was randomly assigned one of the four feed types and 100 g feed/day was fed for 4 weeks. Water exchange rate in each RAS was set at 60 l make-up water/day, equivalent to a feed loading of 1.67 kg feed/m3 make-up water. Dissolved oxygen level was automatically controlled at 85–100% by a monitoring system (Pacific System, OxyGuard International A/S, Denmark). Sodium bicarbonate was added into the sump as required to maintain pH at 7.0–7.4, and settled faeces were removed and discarded every day from the sedimentation columns attached to the bottom of rearing tanks. Water temperature depended on ambient temperatures, averaging 18.3 ± 0.7 °C in the 4 weeks.

Sampling

Prior to daily system maintenance, grab samples (5 l) were collected from the sump in each system and stored at 4 °C until analyzed. Water samples for dissolved nitrogenous compound analysis (nitrate–N (NO3−-N), nitrite-N (NO2−-N), total ammonia nitrogen (TAN), and urea-N), microparticle analysis, and free-living (FL) microbial abundance analysis were collected once a week. Water samples for total and dissolved chemical oxygen demand (COD) analysis, 5-day biochemical oxygen demand (BOD5) analysis, microbial activity analysis, water clarity measurements (i.e., turbidity and UV transmittance), and total N analysis were collected once in weeks 0, 1, 2, and 4. Samples for dissolved nitrogenous compound analysis were filtered at 0.20 μm (Filtropur S, Sarstedt, Germany), while samples for dissolved BOD5 and COD analysis were filtered at 0.45 μm (Adcantec, Toyo Roshi Kaisha, Japan).

For microbial community analysis, water samples were collected in the sump prior to daily system maintenance in week 0, 1, and 4 using syringe filters (0.22-μm Millipore Express® PLUS PES membrane, Merck KGaA, Germany). In addition, seven carrier media from each MBBR were collected in week 4. All microbiological samples were stored at – 20 °C for subsequent DNA extraction.

Physicochemical and microbial activity and community analysis

Table 2 summarizes the different methodologies applied for analyzing water samples, feeds, and fecal samples. Microbial activity in water samples were measured using two different methods: (1) a hydrogen peroxide (H2O2) degradation assay using the rate constant k (1/h) as an expression of microbial catalase activity (Pedersen et al. 2019) and (2) a fluorometric detection method (BactiQuant®) expressing microbial hydrolase activity in standardized BactiQuant® values (BQV) (Pedersen et al. 2017; Robertson et al. 2016). Microparticle size distribution (PSD) was measured by two instruments: (1) an AccuSizer 780 SIS (Particle Sizing Systems, USA) measuring the quantity of microparticles between 10 and 100 μm and (2) a Multisizer 4e Coulter Counter (Beckman Coulter Inc., USA) measuring the quantity of microparticles between 1 and 30 μm.

Table 2 Summary of water and feed sample analysis. Microbial catalase activity was described by H2O2 degradation k-value, and microbial hydrolase activity was described by Bactiquant® value
Full size table

To measure the abundance of free-living (FL) microbes, 10 ml of sampled water from each system was pre-filtered through a cell strainer (40-µm FisherBrand, Thermo Fisher Scientific, USA) and diluted 10 times. Of this, 500 µl of filtrate was labelled with 5 µl of SYBR Green (100x, MilliporeSigma, Germany) and 5 µl of propidium iodide (PI, 600 µM, MilliporeSigma, Germany), and incubated at 37 °C for 10 min. Total cell abundance (cells/ml) was subsequently measured with a BD Accuri C6 Plus flow cytometer (Becton, Dickinson and Company, USA).

The DNeasy PowerLyzer™ PowerSoil DNA Isolation Kit (Qiagen, Germany) was used to extract DNA from samples, and the DNA quantity measured with a Qubit 2.0 Fluorometer (Thermo Fischer Scientific, USA). Microbial community compositions were studied using Ion Torrent PGM next-generation sequencing, targeting the V4 region of the 16S rRNA gene with primers 515F-Y (Parada et al. 2016) and 806R (Caporaso et al. 2011) as in Aalto et al. (2022). The total amount of sequences obtained was 1,140,089. For calculating alpha diversity indices (Shannon diversity, Chao1) and beta diversity (Bray–Curtis dissimilarity index), each sample was subsampled to 13,346 sequences. Sequences have been submitted to NCBI Sequence Read Archive under BioProject PRJNA877850.

Calculations

The content of carbohydrates in feed was calculated as nitrogen-free extract (NFE), i.e., as dry matter subtracted by protein, fat, and ash. Furthermore, macronutrients in the feed were converted to COD based on the oxycalorific coefficient (Dalsgaard and Pedersen 2011; Jobling 1995): 1.77 g O2/g protein, 2.88 g O2/g fat, and 1.16 g O2/g NFE. Feed crude energy was calculated based on the heat of combustion: 23.66 kJ/g protein, 39.57 kJ/g fat, and 17.17 kJ/g NFE. Total organic carbon (TOC) was calculated from COD using a COD:TOC ratio of 2.67 (Khanal and Huang 2003), and C:N ratios in feed and RAS water were subsequently calculated as TOC concentrations divided by N concentrations (total N in feed and total dissolved N in water, respectively). In water samples, particulate COD was calculated as the difference between total and dissolved COD. The same principle was used to calculate particulate BOD5. Particulate N suspended in the water was calculated as the difference between total N (unfiltered) subtracted by total dissolved N (filtered).

To generate a complete PSD between 1 and 100 μm, PSD data were categorized according to size class boundaries as suggested in Patterson et al. (1999). As the two PSD instruments used in this trial had different detection ranges, data in the size class between 12.7 and 32.0 μm were merged and averaged as described by de Jesus Gregersen et al. (2019). Particle β value, extracted from the power-law distribution model by Patterson et al. (1999), was used to indicate the status of particle size distribution.

Statistical analysis

Statistical analysis was performed in R (version 3.6.3; R Core Team, 2020) using the R packages: FSA (Ogle 2018), ggplot2 (Wickham 2016), rstatix (Kassambara 2020), vegan (Oksanen et al. 2019), and phyloseq (McMurdie and Holmes 2013). The effects of dietary C:N ratios on water quality in week 4 were tested by one-way ANOVA followed by Tukey test for pairwise multiple comparisons in case of significant differences. In case data did not meet the normality (Shapiro–Wilk test) and variance homogeneity (Levene test) assumptions, a Kruskal–Wallis test followed by a Dunn test was carried out. To test the effects of dietary C:N ratios on NO3−-N and total dissolved N concentrations along the trial, two-way repeated measures ANOVA was used. Dietary effects at each time point were further tested by one-way ANOVA followed by pairwise t test for multiple comparisons. The relationship between dietary and water C:N ratios was examined with Pearson correlation test. The effect of dietary treatment and/or sample type (water and biofilter biofilm) on microbial community composition was examined with the permutational analysis of variance (PERMANOVA). The differences in alpha-diversity indices (Shannon diversity and Chao1 richness) between diets and sample types were examined with two-way ANOVA and Tukey post-hoc tests. The relationship with the relative abundance of sequences assigned to ammonia or nitrite oxidizing bacteria in the biofilters (AOB and NOB abundance) and dietary C:N ratio was examined with Pearson correlation test. Results are presented as mean ± standard deviation throughout, and the level of significance was set at 0.05.

Results

No feed spill was observed in any system during the 4-week trial period while two mortalities were recorded in two CN12.2 systems and one mortality was recorded in a CN10.0 system during the first 2 weeks.

Nitrogenous compounds

The development of NO3−-N in RAS water was significantly affected by the dietary C:N ratios (two-way ANOVA, F12,32 = 169.0, p < 0.001). NO3−-N concentrations measured in week 4 increased with the increasing dietary protein levels (i.e., the decreasing dietary C:N ratios; Pearson correlation, r =  − 0.97, p < 0.001; Fig. 1). The dietary protein levels also affected the NO2−-N concentrations measured in week 4 (one-way ANOVA, F3,8 = 6.1, p = 0.018), with the average concentration being higher in the CN8.5 group than in the CN12.2 group (Tukey test, p = 0.013) (Table 3). Concomitantly, NO2−-N concentrations decreased with increasing dietary C:N ratios (Pearson correlation, r =  − 0.83, p < 0.001). Urea-N and TAN concentrations, on the other hand, were not affected by dietary protein levels in week 4 (Table 3) (one-way ANOVA, F3,8 = 0.7, p > 0.05; F3,8 = 0.6, p > 0.05, respectively), averaging 30 ± 4 µg urea-N/l and 64 ± 22 µg TAN/l across all dietary treatments. As NO3−-N accounted for approximately 93% of total dissolved N (additionally including TAN, nitrite-N, and non-characterized N), total dissolved N in water was similarly correlated to dietary protein levels (i.e., negatively correlated to dietary C:N ratios; Pearson correlation, r =  − 0.96, p < 0.001).

Fig. 1
figure 1

Weekly development of nitrate–N concentration in systems fed different diets: CN8.5, CN9.3, CN10.0, and CN12.2. Stars indicate the significant levels among treatment groups in a specific week (one-way ANOVA): * stands for p values ≥ 0.01 and ≤ 0.05, ** for p between ≥ 0.001 and < 0.01, while.*** for p < 0.001

Full size image
Table 3 Water quality parameters in week 4 under different dietary C:N ratios (n = 3) (mean ± standard deviation)
Full size table

Organic matter and C:N ratios

Total organic matter (BOD5 and COD) measured in water samples was not significantly affected by the dietary C:N ratios in week 4 (Table 3) (total BOD5: one-way ANOVA, F3,8 = 2.0, p > 0.05; total COD: F3,8 = 1.8, p > 0.05). Only the dissolved COD was found statistically lower in the CN12.2 group than in the CN8.5 group (Tukey test, p = 0.030). Although not significant, numerically the lowest organic matter content was all found in the CN12.2 group both for dissolved and particulate fractions. Molar C:N ratios in water samples did not differ between the treatments in week 4 (one-way ANOVA, F3,8 = 1.3, p > 0.05), ranging between 0.27 and 0.34 (Table 3) and did not correlate with the dietary C:N ratios (Pearson correlation, r =  − 0.42, p > 0.05).

Microparticles and water clarity

Microparticle size distributions (β values) did not differ significantly between the treatments in week 4 (one-way ANOVA, F3,8 = 0.3, p > 0.05), averaging 3.03 ± 0.15 across all treatments by the end of the trial (Table 3). Consistent with this, the abundance of microparticles (both 1–5 µm and 5–100 µm) were not affected either (F3,8 = 2.3, p > 0.05; F3,8 = 1.0, p > 0.05, respectively). Water clarity was significantly affected by the dietary treatments with the highest UVT found in the CN12.2 group (F3,8 = 11.1, p = 0.003), whereas turbidity was not statistically different between the treatments (F3,8 = 1.6, p > 0.05) (Table 3).

Microbial activity and abundance

The dietary treatments significantly affected microbial catalase activities (one-way ANOVA, F3,8 = 4.2, p = 0.047), with the CN12.2 group having a 50% lower average value than the CN9.3 group in week 4 (Tukey test, p = 0.032). In comparison, microbial hydrolase activities measured by BactiQuant® were not significantly affected by the diets (one-way ANOVA, F3,8 = 1.6, p > 0.05). Similarly, different dietary C:N ratios did not affect the abundance of FL microbes (F3,8 = 3.0, p = 0.097), averaging 52 ± 16 million cells/ml across the treatments (Table 3).

Microbial communities in water and biofilms

In water, microbial community composition did not differ in the beginning of the trial (data not shown) but changed significantly from week 1 to week 4 (PERMANOVA, pseudo-F = 14.6, p = 0.001), whereas it was not affected by the dietary C:N ratios (pseudo-F = 1.43, p > 0.05) (Fig. 2). Similarly, the dietary C:N ratios did not cause changes in the microbial community composition in biofilter biofilm (pseudo-F = 1.39, p > 0.05). In contrast, microbial community compositions measured in RAS water samples differed significantly from those in RAS biofilter biofilm independently of the dietary C:N ratios in week 4 (pseudo-F = 30.0, p = 0.001). In water samples, the most abundant bacterial classes (relative abundance of > 20%) were Actinobacteria and Alpha- and Gammaproteobacteria, while in the biofilter biofilm samples the most abundant classes were only the two latter ones (Fig. 3). Neither OTU diversity (Shannon index) nor richness (Chao1) was affected by the diets or the sampling occasions (Shannon: two-way ANOVA, F3,28 = 0.08, p > 0.05, Chao1: F3,28 = 0.08, p > 0.05). Both diversity and richness were lower in the water than in the biofilter biofilm in week 4 across all dietary treatment groups (Shannon: two-way ANOVA, F1,16 = 561.9, p < 0.001, Chao1: F3,16 = 20.5, p < 0.001) (Fig. 4).

Fig. 2
figure 2

Principal coordinate analysis (PCoA) of microbial communities based on Bray–Curtis dissimilarities in water and biofilter biofilm under different dietary C:N ratios. Experimental week is given as number inside the symbols

Full size image
Fig. 3
figure 3

The relative abundances of bacterial classes in A water and B biofilter biofilm between different dietary C:N ratios. Each experimental unit consisted of 3 replicated systems

Full size image
Fig. 4
figure 4

The microbial OTU A diversity (Shannon index) and B richness (Chao1) in biofilter biofilms and water samples in weeks 1 and 4. Different symbol colors refer to different levels of dietary C:N ratios, while different symbol shapes refer to different sample types

Full size image

The relative abundance of ammonia-oxidizing bacteria (AOB abundance) decreased with increasing dietary C:N ratios (Pearson correlation, r =  − 0.61, p = 0.036) (Table 4), while there was no relationship between the dietary C:N ratios and the relative abundance of nitrite oxidizing bacteria (NOB) (Table 4).

Table 4 The relative abundance of ammonia-oxidizing bacteria (% AOB) and nitrite-oxidizing bacteria (% NOB) (number of sequences of all sequences) in biofilter biofilm samples in week 4 under different dietary C:N ratios (n = 3) (mean ± standard deviation)
Full size table

Discussion

In this study, we found that the dietary C:N ratios affected some physicochemical water quality parameters and the microbial activity in RAS water, but not microbial community composition in water or biofilter biofilm. The CN12.2 diet with the highest dietary energy content resulted in the best system performance in terms of low dissolved nitrogenous compounds, low dissolved organic matter, reduced microbial activity, and high water clarity.

In general, partially replacing protein with carbohydrate and fat in a diet based on the protein-sparing principle (Cho and Bureau 2001) generally leads to increases in dietary C:N ratios. Given that diets are optimized to fish requirements, such dietary macronutrient replacements reduce the excretion of ammonia and urea from fish, and subsequently reduce the accumulation of NO3−-N and total dissolved N in RAS (Cho et al. 1994; Godoy-Olmos et al. 2022). This explains why NO3−-N and total dissolved N concentrations in the current study correlated positively with dietary protein levels (i.e., were negatively related to the dietary C:N ratios).

In contrast to dissolved N, organic matter concentrations in RAS water were only moderately affected by the dietary treatments, probably due to the four diets having similar COD concentrations. Organic C in water and dietary COD did not correlate explaining why water C:N ratios and dietary C:N ratios also did not correlate.

Hence, based on the results obtained in this study and in previous RAS studies, there is not a strong relationship between dietary and water C:N ratios in RAS (Fig. 5), acknowledging that the different RAS compared varied in, e.g., system configuration, diet composition, and feed loading. It seems that molar C:N ratios in system water generally vary within a narrow range (0.11–0.63), a factor 10 or more below the ones in the associated fish feeds (8.2–14.2). Given that total BOD5 constitutes approximately 6–30% of total COD in RAS water (Fernandes et al. 2015, 2017; Pedersen et al. 2012; Rojas-Tirado et al. 2018), C:N ratios calculated by using only the readily biodegradable organic carbon (i.e., BOD5) instead of COD could be even below 0.01, indicating highly C-limited conditions for RAS microbes.

Fig. 5
figure 5

Comparison of molar C:N ratios in diets and in water samples derived from different RAS studies. The C:N ratios were calculated as total organic carbon (TOC) or total COD divided by total dissolved nitrogen or nitrate-N, based on the information provided in the different studies

Full size image

In previous ecological studies, the overall bacterial biomass has been found to be quite plastic and quickly adapt its C:nutrient ratios following changes in the environmental ratios (Tezuka 1990), while different specific strains of heterotrophic bacteria have been found to have individual and rather narrow ranges of optimal substrate C:nutrient ratios (Makino et al. 2003; Vrede et al. 2002). This means that increases in C:N ratios would cause a shift in the dominance of different heterotrophic microbial groups. While this is supported by both microbial catalase activity and COD concentrations being low in the CN12.2 group, the microbial community composition in RAS water and biofilter biofilms and the overall microbial abundance in RAS water remained largely unaffected. Previously, decreasing the water molar C:N ratios from 0.56 to 0.33 through particle removed by membrane filtration was found to induce significant changes in RAS water microbiome and decreases in the overall abundance (Bugten et al. 2022; Fossmark et al. 2020). However, this might have been due to the removal of microbial substrate (organic matter) or microbes being directly removed by membrane filtration (20–30 nm) rather than changes in water C:N ratio. As in this study the water molar C:N ratios were substantially lower across all the treatments (ranging from 0.27 to 0.34) than the previously suggested optimal range for exponential growth (C:N ratio of 4–6) (Vrede et al. 2002), it is most likely that the microbial biomass was significantly C-limited even with the highest dietary C:N ratio and that the microbial community in water and/or biofilm already initially consisted of taxa with their optimal C:nutrient ratios. However, we found that microbial community compositions in water differed from those in biofilter biofilm, which is in agreement with previous findings (Almeida et al. 2021; Bakke et al. 2017). This is probably explained by biofilters being an even more C-limited environment due to high microbial biomass receiving only the left-over organic matter from the rearing tanks, which should favour autotrophic nitrifiers but also heterotrophs adapted to C-limited environments.

Even though biofilter biofilm community compositions were generally not affected by the dietary C:N ratios, changes in element concentrations induced by the diets were still able to affect bacterial groups with specialized functions, as we found the relative AOB abundance decreasing with the decreasing dietary protein levels (i.e., the increasing dietary C:N ratios in this study). This was probably caused by the decreasing ammonia excretion from fish under higher dietary C:N ratios, which limited the growth of AOB and also explains the observed lower nitrite concentration in the CN12.2 treatment. The relative NOB abundance, on the other hand, did not differ cross the treatment groups. This is not surprising as NOBs can grow with other substrates than nitrite (Kuypers et al. 2018), e.g., the Nitrospira-group (considered as “NOB”) includes members capable of conducting a complete nitrification process from ammonia to nitrate (Daims et al. 2016), leading the NOB abundance being unrelated to nitrite availability.

Altogether, the results indicate that RAS microbial communities in both water and biofilm are rather insensitive to changes induced by dietary C:N ratios, at least when they span over the values of the normal commercial trout diets. Other measures would thus probably be more effective when aiming to manipulate or control RAS microbiology. A direct removal of organic matter with, e.g., membrane filtration or protein skimming has been shown to be effective in removing and/or suppressing the activity of living microbes (de Jesus Gregersen et al. 2021; Fossmark et al. 2020). Furthermore, manipulating dietary N:P ratios rather than C:N ratios may have a higher potential for influencing RAS microbiomes, as P is an essential component of nucleic acids (DNA and RNA) and lowering P availability (still meeting the minimum P requirement of fish) should favour slow growing bacterial taxa (Hessen et al. 2013; Scott et al. 2012). Recently, we found a positive relationship between dietary and water N:P ratios in RAS (Huang et al. 2023), suggesting an interesting way to influence RAS microbiomes.

Conclusion

In this study, we found that dietary C:N ratios affect some physicochemical water quality parameters and microbial activity in RAS. The highest dietary C:N ratio, achieved by replacing protein mainly with fat, resulted in the best system performance in terms of low concentrations of NO2−-N, NO3−-N, dissolved COD, and microbial activity in addition to high water clarity. In addition, dietary and water COD, and subsequently, C:N ratios were found to be uncorrelated. Water C:N ratios were below the optimum range for aquatic microbes across all the dietary treatments, demonstrating that the microbial biomass was significantly C-limited despite dietary C:N ratio. This phenomenon is likely to be observed in most RAS.