Shrimp (Penaeus vannamei) post-larvae, acquired from the commercial laboratory AQUATEC Ltda., located in Rio Grande do Norte, were used. The shrimp were kept in a biofloc system in the Marine Shrimp Laboratory until the beginning of the experiment.

Fish of the species Mugil liza, from the Marine Fish Farming Laboratory of UFSC, located in Barra da Lagoa, in Florianópolis, Santa Catarina, were used. The fish were kept in a biofloc system in the Marine Shrimp Laboratory and, for the initial stocking in the experimental units, a pre-selection of fish size was made.

Four experimental groups consisting of different fish feeding rates were evaluated with four replicates in a completely randomized design: 0%, 1%, 2% and 3% of fish biomass. The shrimp feeding was unchanged in between treatments.

The experimental units were composed of a tank with a useful volume of 800 L, in which the shrimps were stocked, connected by recirculation to a tank with a useful volume of 90 L, used for the stocking of the mullets (Fig. 1). The water was pumped from the shrimp tank to the fish tank by a Sarlo-Better 650 L hour^-1 pump, returning to the shrimp tank by gravity, keeping the system in recirculation 24 hours a day. Each 800 L tank was equipped with a central aeration ring, while the 90 L tanks each contained four porous stones connected to an aeration system, in order to keep dissolved oxygen levels within the appropriate range for the species and prevent sedimentation of solids in the tanks. The tanks were located in a greenhouse with natural photoperiod.

Shrimp (2.36 ± 0.02 g) were stocked under a density of 375 animals m^-3 (300 animals tank^-1), whereas fish (7.06 ± 0.65 g) were stocked under a density of 167 animals m^-3 (15 animals tank^-1).

In each experimental unit, 800 L of mature biofloc water (0.47 mg L^-1 TAN, 0.16 mg L^-1 N-NO₂-, alkalinity 226 mg CaCO₃ L^-1, 9.44 mg L^-1 N-NO₃-, 9.6 mg L^-1 P-PO₄ ^3-) from a shrimp culture of the Marine Shrimp Laboratory, and 90 L of seawater, pumped from Barra da Lagoa beach, totalling 89.89% of inoculum in each experimental unit. In addition, the 800 L tanks were equipped with four artificial substrates composed of Needlona®, oriented vertically, corresponding to 80% of the tank surface area. To maintain the adequate water temperature for the culture of both species, each experimental unit was equipped with an 800 W titanium heater, located in the shrimp tanks.

The shrimp were fed four times a day (8:30 a.m., 11:30 a.m., 2:00 p.m. and 5:30 p.m.), initially with the Guabitech feed (1 mm), with 45% crude protein, until the morning of the seventh day of experiment, when they started to be fed with the Guabi Poti Guaçu feed (1.6 mm), which contains 35% crude protein. In treatments 1%, 2% and 3%, the mullets were fed twice a day (11:30 a.m. and 4:00 p.m.) with Nutripiscis starter 0.8 mm, which contains 40% crude protein. In these treatments, 1, 2 and 3% of the fish biomass of feed were offered. Fish from treatment 0% were not fed. The adjustment of the amount of shrimp feed was made according to the feeding table developed by Van Wyk (1999) through weekly biometrics, and fortnightly for fish.

Alkalinity was corrected with calcium hydroxide when the measured values were below 120 mg L^-1 (Furtado et al., 2011). The amount of calcium hydroxide was calculated using 5 to 20% of the amount of feed supplied to the experimental unit on the previous day.

Temperature and dissolved oxygen were measured twice a day (morning and afternoon), with a YSI pro20 oximeter both in the shrimp tanks as well as in the fish tanks, while the remaining variables were measured in the shrimp tanks only. Twice a week, pH was measured with a Thermo Scientific Orion Star A211 pHmeter. Salinity was measured twice a week with the Ecosense EC300A digital conductivity meter. Alkalinity was measured twice a week using the titrimetric method. Total ammonia nitrogen (Grasshoff et al., 1983) and nitrite-N (Strickland & Parsons, 1972) were measured twice a week. Nitrate-N was measured at the beginning, middle and end of the experiment, according to the method of Strickland & Parsons (1972). Total suspended solids (TSS) were measured twice a week (APHA et al., 2005).

Dissolved oxygen (shrimp tank: 5.34±0.47 mg L^-1; fish tank: 5.05±0.60 mg L^-1), temperature (shrimp tank: 29.5±1.3°C; fish tank: 29.6±1.4°C) and salinity (33.9±1.4 gL-¹) remained within ranges commonly used to successfully rear whiteleg shrimp (Van Wyk & Scarpa, 1999) and Mugil spp. (Legarda et al., 2019, 2021).

For the quantification of total heterotrophic bacteria and Vibrio spp., 10-mL water samples from each tank were collected at the beginning and end of the experiment using sterile assay tubes with screw caps. A serial dilution using 1-mL samples was performed using tubes filled with 9 mL of sterile saline solution (3% NaCl), after which 100 µL aliquots of the homogenized contents were seeded in Petri dishes containing trypticase soy agar (TSA) and thiosulfate-citrate-bile salts-sucrose agar (TCBS) for viable heterotrophic bacteria and Vibrio spp., respectively. The seeded Petri dishes were incubated in a bacteriological stove at 30^oC for 24 h and the colonies counted through colony forming units (CFU mL^-1) plate counting.

Data on water quality and zootechnical performance of organisms were tested for homoscedasticity and normality through Levene and Shapiro-Wilk tests, respectively, for subsequent application of ANOVA. When significant differences were found, the Tukey test was used to compare the means between treatments. When the data did not meet the assumptions for application of ANOVA, the Kruskal-Wallis nonparametric test was used. Linear regressions were also performed on the productive performance variables. The analyses were carried out with the software Statistica (TIBCO) and jamovi (The jamovi project, 2022) at a significance level of 5%.

Water quality variables are presented in Table 1, with no significant differences found between the experimental groups (p>0.05). The treatments had pH values ranging from 7.3 to 8.5, with a mean value higher than 7.9 in all treatments. Mean total ammonia nitrogen values remained below 0.1 mg L^-1, and mean nitrite-N values below 1 mg L^-1. Alkalinity remained around 150 mg CaCO₃ L^-1 (92‒200 mg CaCO₃ L^-1) in all treatments. The mean values of total suspended solids remained between 500 mg L^-1 and 600 mg L^-1 in all treatments. The mean nitrate-N values remained between 16 mg L^-1 and 20 mg L^-1, increasing from an overall mean of 11.61 ± 3.22 mg L^-1 at the beginning to 23.34 ± 4.33 mg L^-1 at the end of the experiment.

Table 2 shows the productive performance of both species and the system as a whole. For the shrimp performance, no significant differences between treatments were found (p>0.05). The mean final weight of the shrimp was above 12.0 g. The final shrimp biomass presented mean values above 3.0 kg. Shrimp grew an average of around 1.5 g per week in all treatments. Mean daily growth coefficient values around 2.1 % day^-1 were observed in all cases. The feed conversion ratio had a mean value close to 1.6. All treatments showed mean survival above 82% and mean yield around 3.9 kg m^-3.

For the fish, significant differences were found for final mean weight, final biomass, daily growth coefficient and yield (p>0.05), with each increase in the fish feeding rate resulting in a significant increase in the assessed variable (Table 2). Survival was above 95% in all cases and mean feed conversion ratio mean values in the treatments that received feed ranged between 1.34 and 1.40.

As regards the system performance, similar to the case with the fish, significant increase in final biomass and yield were observed as the fish feeding rate increased (p<0.05) (Table 2), with the 3% feeding rate treatment presenting the highest value in comparison with both the 0% and 1% treatments. The system mean feed conversion ratio remained around 1.6 in all experimental groups.

Significant linear regression models were found for some of the fish and system performance variables (p<0.05) (Table 3), i.e. fish final mean weight, final biomass, daily growth coefficient and yield, in addition to system final biomass and yield. Fish final mean weight, daily growth coefficient and yield regression models presented R² values higher than 0.9 (Fig. 2).

There were no significant differences between treatments for the counts of total heterotrophic bacteria and Vibrio spp. (p>0.05) (Fig. 3). The overall values were: total heterotrophic bacteria 5.0 ± 0.7 CFU mL^-1 (log) and Vibrio spp. 3.2 ± 0.6 CFU mL^-1 (log).

The increasing input of feed did not have deleterious effects on water quality, with the values of pH, TAN, nitrite-N, nitrate-N and alkalinity remaining within appropriate ranges for shrimp (Van Wyk & Scarpa, 1999; Lin & Chen, 2001) and within values previously employed to successfully rear mullets in BFT (Legarda et al., 2019; Chamorro-Legarda et al., 2020; Holanda et al., 2020; Legarda et al., 2021).

There are studies evaluating the optimal TSS concentration for different species in BFT, such as shrimp P. vannamei (400 to 600 mg L-1; Schveitzer et al., 2013), South American catfish (Rhamdia quelen) larvae (200 mg L^-1; Poli et al., 2015) and pacu (Piaractus mesopotamicus) juveniles (>250 mg L^-1; Pellegrin et al., 2022). For lebranche mullets (M. liza), no such study has been published, but previous studies have successfully cultivated Mugil spp. under mean TSS concentrations varying from values as low as 89.79 ± 18.32 mg L^-1 (Holanda et al., 2020), intermediate values between 300 mg L1 and 600 mg L^-1 (Legarda et al., 2019; Chamorro-Legarda et al., 2020; Legarda et al., 2021) and as high as 706.52 ± 252.13 mg L^-1 (Borges et al., 2020). In the current study, the fish exhibited high survival (above 96% in all treatments) in a situation where the mean TSS concentration varied from 521.98 ± 171.05 mg L^-1 to 575.98 ± 189.49 mg L^-1, suggesting the feasibility of cultivating mullets under these concentrations.

Strategies employed to manage biofloc systems vary, from using different organic carbon sources to stimulate the immobilization of ammonia, to using tactics to hasten the establishment of a nitrifying community (Ebeling et al., 2006). In the case of this study, the use of an inoculum of mature bioflocs, i.e. water from a tank in which nitrification is already occurring, allowed for the maintenance of low concentrations of the most toxic nitrogenous compounds, TAN and nitrite-N, throughout the entire experiment. It can also be highlighted that the nitrate-N concentration increased at the end of the experiment in comparison with the inoculum (from 11.61 ± 3.22 mg L^-1 to 23.34 ± 4.33 mg L^-1), although remaining in concentrations that are not harmful to the cultured animals, where concentrations as high as 177 mg L^-1 having been reported as acceptable for rearing P. vannamei under a salinity of 23 g L^-1 (Furtado et al., 2015). These two observations, the low concentrations of TAN and nitrite-N, along with the increasing nitrate-N concentration, suggest that the system was characterized as being a mature biofloc system, as expected from the management employed.

Regarding the shrimp, the lack of significant effects of the fish feeding rates on its performance was likely a result of the increased rates not deteriorating water quality. A similar phenomenon was also observed when P. vannamei was cultured in integration with Nile tilapia in a study evaluating different feeding rates for the fish (Silva et al., 2022). The authors also observed no significant effects of the fish feeding rates on the shrimp performance.

In contrast, the significant effects observed on the fish performance variables indicate that mullet growth and productivity were influenced by the feeding rates. As the feeding rate increased, there was an incremental improvement in fish final mean weight, final biomass, daily growth coefficient, and yield. This suggests that a higher feeding rate provides a greater availability of feed resources for the fish, leading to enhanced growth and productivity. The positive relationship between feeding rate and fish performance aligns with previous studies that have reported similar findings for aquatic animals reared in BFT, such as Nile tilapia (Oliveira et al., (Oliveira et al., 2021; Silva et al., 2022) and whiteleg shrimp (Khanjani et al., 2016) under different feeding levels.

The observed significant effects on the overall system productive performance further support the influence of feeding rates on the integrated culture system. Specifically, there were significant effects on the final biomass and yield of the system. As the feeding rate increased, both the final biomass and yield of the system increased accordingly. This indicates that a higher feeding rate promotes a higher overall production output, highlighting the importance of feed availability for system productivity. Silva et al. (2022) also reported higher system productive performance when Nile tilapia were fed at 1% of their biomass in the integrated culture with whiteleg shrimp, when compared to a treatment in which the fish were not fed at all.

The results of this study have practical implications for the optimization of feeding strategies in the integrated culture of marine shrimp and mullets using biofloc technology. The findings suggest that increasing the fish feeding rate can positively impact fish growth and system productivity without negatively affecting shrimp performance. However, it is important to note that there may be limits to the positive effects of increasing feeding rates. Beyond a certain point, excessive feeding rates may lead to diminishing returns or potential negative impacts, such as deteriorating water quality and poor growth performance (Oliveira et al., 2021).

The counts of total heterotrophic bacteria and Vibrio spp. in the water were measured to assess the impact of different fish feeding rates on water microbiology in the integrated culture of whiteleg shrimp and lebranche mullet using biofloc technology. The results showed no statistically significant effects of the fish feeding rate on water microbiology. This suggests that the microbial communities associated with biofloc formation and maintenance remained resilient and stable, unaffected by the feeding rates tested. This was also observed in a study evaluating the effect of fish feeding rate on the integrated culture of Nile tilapia and whiteleg shrimp, in which no statistical significance was found for Vibrio spp. and total heterotrophic bacteria when different fish feeding rates were used (Silva et al., 2022)

The lack of significant effects could be related to the fact that the fish feeding represented a small percentage of the overall feed input, due to the greater shrimp biomass in the system. A support for this hypothesis is a study that evaluated the effects of different feeding rates and species configurations in the integrated culture of Tilapia hornorum and P. monodon (Tendencia et al., 2006). The system that employed shrimp and fish with both being fed resulted in statistically higher values of total bacterial count when compared with a system in which the fish were not fed. However, the proportion of biomasses differed from the ones used in the current study. For example, Tendencia et al. (2006) used 0.5 kg m^-3 of fish and only 0.080 kg m^-3 of shrimp, compared to the current study where the mean final biomasses were at most 0.64 kg of fish m^-3 and 3.54 kg of shrimp m^-3 when considering the whole volume of the system.

These findings indicate the robustness of the microbial ecosystem in biofloc-based systems and highlight the potential for adjusting feeding rates within a reasonable range without significant disruptions to the microbial balance. Future research should investigate the long-term effects of feeding rates on water microbiology and explore the functional roles of specific microbial taxa in biofloc systems for a comprehensive understanding of their contributions to system performance and sustainability.