Recovery of Salts from Synthetic Erythritol Culture Broth via Electrodialysis: An Alternative Strategy from the Bin to the Loop

: Sustainability and circularity are currently two relevant drivers in the development and optimisation of industrial processes. This study assessed the use of electrodialysis (ED) to purify synthetic erythritol culture broth and for the recovery of the salts in solution, for minimising the generation of waste by representing an efﬁcient alternative to remove ions, ensuring their recovery process contributing to reaching cleaner standards in erythritol production. Removal and recovery of ions was evaluated for synthetic erythritol culture broth at three different levels of complexity using a stepwise voltage in the experimental settings. ED was demonstrated to be a potential technology removing between 91.7–99.0% of ions from the synthetic culture broth, with 49–54% current efﬁciency. Besides this, further recovery of ions into the concentrated fraction was accomplished. The anions and cations were recovered in a second fraction reaching concentration factors between 1.5 to 2.5 times while observing low level of erythritol losses (<2%), with an energy consumption of 4.10 kWh/m 3 .

The remaining salts from the culture media have to be separated from the culture broth to obtain pure erythritol, and therefore the non-used salts in the culture broth leave the process as a residual stream that has to be removed in the downstream section [34]. In the last decades, the recovery of nutrients has gained interest, becoming a multidisciplinary challenge for industry, water resources management, soil conservation, among other fields [34][35][36][37]. The transition from conventional processes to industrial circular processes represents an opportunity to tackle resource management and climate change concerns. In order to attend the depicted necessities, the European Union has developed projects as Zero Brine [36], INCOVER [37], NITROMAN [38], among other initiatives looking for technological alternatives that are able to reduce the impact caused by the waste and to evaluate the recovery of valuable compounds as phosphorus, sulphate, ammonia, and others. However, there is not enough scientific evidence regarding the consumption and recovery and reutilisation of salts from submerged cultures, especially considering the mandatory demand for salts in these biotechnological processes.
The removal of salts from erythritol culture broth is used as an example of a more general approach for recycling residuals from biotechnological culture and so far has been addressed by adsorption on ion exchange resins in the downstream section [34,[39][40][41]. The principle behind this separation is the removal of salts by exchanging their position with other exchangeable ions (Na + , H + , among others) within cationic and anionic exchange resins. However, despite the high performance and efficiency of this conventional approach, downstream regeneration of ion-exchange resins requires chemical substances. It generates large amounts of alkaline and acid residual waste streams mainly composed of the salts removed from the culture broth. This fact represents a drawback to reach sustainable and cleaner production goals.
Alternatives for removing salts include membrane-based technologies, like forward osmosis, membrane distillation and electrodialysis [42]. Electrodialysis (ED) is an environmentally friendly and low energy demanding membrane-based technology suitable for the recovery of ionisable molecules from spent culture broths [42][43][44][45][46]. An electrical potential applied to the ED cell serves as the driving force for separating ionic species from the uncharged matter in an aqueous solution. Studies in ED application include recovery and reutilisation of alkaline fractions [47], separation and purification of L-phenylalanine [48], lactic acid [49], glycerol [50], glutamine [51], and 1,3-propanediol [52].
During the ED batch process, the feed located in the diluted and concentrated reservoir is recirculated in a closed loop through a stack of membranes in the ED batch mode, driven by a directly applied direct current. Cations and anions migrate from the feed solution to a concentrated fraction by going through cation-and anion-selective membranes, respectively. Two fractions are obtained at the end of the ED treatment: a purified product fraction with a low content of ions and a concentrated fraction rich in ions [50,53]. The main factors affecting the performance of ED treatment include the applied current density, the selectivity of the membranes (monovalent or multivalent), the flow of diluted and concentrated fractions, temperature, and pH [50,51,54].
Limiting current density (LCD) is a critical parameter that rules the overall performance of electrodialysis. The LCD value represents a threshold of a maximum current applied to the ED cell, at and above which adverse effects take place in the ED separation process. If the ED is operated in the limiting (at the threshold) or over-limiting (above the threshold) regime, dissociation of water molecules occurs in a thin layer at the membrane/diluted-solution interface. The H + and OH − ions from water dissociation participate in the transfer of current (through the ED membrane stack), reducing the overall ED process efficiency and increasing the energy consumption. Besides water splitting, fluid leakage, current leakage, and co-ion transfer occur if the threshold current density is exceeded [50].
This paper aims to evaluate the performance of the ED to separate and recover valuable ions from culture-broth process streams. Firstly, remove salts from erythritol culture broth, and then concentrate the same salts in a potentially reusable concentrated fraction. This assessment could become the first stage to evaluate the reutilisation of ions as an approach to treat waste stream from bin to loop.
The experiments were designed in three stages, gradually increasing the complexity of the fed solution. In this sense, the research question was to understand how can the complexity of culture broth solutions affect the performance of ED treatments. For this purpose, three different solutions were tested, and a stepwise voltage approach was used for the experiments. Each stage was judged upon the removal efficiency of ions in the diluted fraction, the concentration factor in the concentrated fraction, and the overall losses of products, by-products, and glucose used as a carbon source. Considering the industrial projection of this ED application, the current efficiency and energy consumption were also determined.

Feed Solutions
The experimental part was developed in three stages, increasing the complexity of the solutions by adding compounds from three different groups present in the erythritol culture media: O), and sodium chloride (NaCl) as the osmotic agent used. These compounds are separated in the concentrated fraction.

Erythritol and By-Products
This fraction included erythritol as the main product, glycerol, and residual glucose as by-products. These compounds are neutral and not dissociable at the operational conditions (erythritol, glycerol, and glucose pKa values 13.9, 14.4, and 12.28, respectively) [56][57][58]. These compounds should remain in the diluted fraction.

Other Compounds from Culture Media
This fraction includes surfactant (tween 80) and peptone used as surfactant and emulsifier for membrane cell protection. These compounds should remain in the diluted fraction. Table 1 presents the concentration levels and compounds added to the experiments. The solution fed to the system increased its complexity in each stage (S, SP, and SCB solutions), while the initial concentrated solution (equivalent to the S solution) remained unchanged during the experiments as the receiving media for the salts. The compounds and concentrations in the feed solutions were defined based on a preliminary characterisation of ions and conductivity of a real erythritol culture broth after the fermentation stage (data not shown). According to these results, the concentrations used in this work corresponded to 50% culture media before fermentation. Considering the lack of studies regarding the consumption of osmotic agents, in this study sodium chloride was added only to adjust the conductivity of feed solutions. The compounds added to the three feed solutions are marked with a cross in Table 1. The first stage considers only the salt content in spent culture media (S solution), where both the diluted and the concentrated solution have the same initial composition. The initial concentrated solution is kept the same (S solution) in further stages, whereas the diluted solution complexity gradually increased. The main products and by-products were added to the diluted stream in the second stage (SP solution). Finally, the third stage studied the treatment of the complete erythritol synthetic culture broth (SCB), including tween 80 and peptone, introduced to the diluted stream.

Experimental Settings
An electrodialysis laboratory unit ED 64004 (PCCell GmbH, Heusweiler, Germany) was used in this work. The ED 64004 consisted of a current generator supplying constant voltage (0-30 V), a control unit, an online measuring system for pH, temperature, conductivity, and voltage (JUMO GmbH & Co. KG, Fulda, Germany), and three independent storage containers for diluted, concentrated and electrode rinsing solution connected in circuits with three magnetically coupled centrifugal pumps NDP 25/4 (ITS-Betzel, Hatterscheim, Germany).
The ED membrane stack was composed of two electrode compartments with 10 identical cells. Each cell unit has an anion exchange membrane (AEM. Ref: PC SA 10x) and a cation exchange membrane (CEM. Ref: PC SK 9x). The effective membrane area was 64 cm 2 per membrane, with 1 mm of spacing between membranes. The anode and cathode materials were Pt/Ir-coated titanium and theV4A steel, respectively. They were placed in the polypropylene electrode housing material.
All the voltages assessed in this approach corresponded to 80% of the limiting current density (LCD) determined using the Cowan and Brown method [59]. A stepwise voltage was the approach used for the application of the direct current in this work. Through this approach, four different voltages (10, 9, 7 and 6 V) were kept constant during a period and changed as soon as the conductivity of the diluted fraction turned from 5.47 mS/cm (initial value) to 3, 2 and 1.4 mS/cm. The four voltage-conductivity pairs were selected from preliminary assays finding the LCD for different conductivity levels (data not shown).
S, SP, and SCB experiments were performed in triplicates. In situ parameters as pH, conductivity, and temperature were recorded at the beginning and during the treatment. 1.5 L of diluted and concentrated solution were pumped and recirculated to the system at a constant flowrate rate of 15 L/h, whereas the electrode-rinsing solution (0.25 M Na 2 SO 4 ) was recirculated at 150 L/h. A total of 12 samples per experiment were taken at different interval times, considering the reduction in the concentration of ions. The first nine samples were taken every 5 min, the 10th and 11th were taken every 10 min, and the last sample was taken after the conductivity of the diluted fraction reached a chosen value of 0.28 mS/cm. This value was selected, having as reference tap water conductivities. Due to the short-term characteristics of the experiments, scaling and fouling phenomena were not considered in the analysis.

Analysis of Erythritol, Glycerol, and Glucose with HPLC
Erythritol, glycerol, and glucose were determined on a High-Performance Liquid Chromatography (HPLC) equipped with a degasser unit DGU-20A3R (Shimadzu, Kyoto, Japan), an autosampler unit SIL-20ACHT (Shimadzu, Kyoto, Japan), a column oven CTO-20A (Shimadzu, Kyoto, Japan), and a refractive index detector RID-20A (Shimadzu, Kyoto, Japan). The mentioned compounds were separated on a Shodex SH1011 column (80 × 300 mm) with a guard column Shodex Sugar SH-G (6.0 × 50 mm). The HPLC was coupled with a refractive index detector RID-20A (Shimadzu, Kyoto, Japan). The separation was performed at an isocratic gradient. The mobile phase used was 5 mM H 2 SO 4 at 0.6 mL/min for 20 min. The oven and RID temperature were 50 • C.

Removal of Ions and Current Efficiency (CE)
The percentage of removal (R j ) of ions was calculated by using Equation (1), where C j (t i ) means an initial or a given time, and C j (t i+1 ) means the final time of the period to be evaluated.
Current efficiency (CE) was calculated based on Faraday's Law following the approach suggested by Wu et al. (2011) [47] (Equation (2)). Where n j i and n j f are the initial and final number of moles of salts in diluted fraction, F is the Faraday constant (96,485.33 sA/mol), I is the current value (A), t is the time (s), N is the number of cell pairs in the stack, and z is the ionic charge (pH-dependent for polyvalent ions). Considering that the energy applied is used to promote the simultaneous migration of cations and anions towards the electrodes, CE was estimated only based on the behaviour of anions in the solution.

Energy Consumption
The energy required for the desalination (E d ) of the feed solution was calculated following Equation (3), where U is the applied voltage (V), I is the applied current (A), t is the desalination duration (h), and V is the treated volume of the feed solution (m 3 ).
Energy consumed by the pumps (E p ) in the diluted, concentrated, and electroderinsing solutions was calculated following Equation (4), where Q is the flow of the diluted, concentrated, or electrode-rinsing solution (m 3 /s), ∆p is the differential pressure (Pa) between the inlet and outlet of the ED stack, η is the pump efficiency (assumed as 75% for all three pumps), and V is the treated volume of the diluted solution (m 3 ). Table 2 presents the characterisation of initial monovalent and bivalent ions in the S, SP, and SCB solutions. The preparation process causes minor deviations in the chemical composition of the solutions. Due to the pH levels and the polyvalent nature, sulphur and phosphorus ions can adopt different dissociation values. Considering the initial pH value (Table 3), the main form of these ions was the monovalent dihydrogen phosphate (H 2 PO 4 − ) and sulphate (SO 4 2− ) with −1 valence.   According to the measured concentrations in Table 2, an average erythritol culture media would need more anions (72% wt.) than cations (28% wt.); and more total monovalent (69% wt.) than bivalent ions (31% wt.). Differences in the amounts of ions added per species are expressed in the concentration gradients. These differences are considered in other sections for the analysis of removal and current efficiency. Table 3 presents the physicochemical parameters measured at the beginning and the end of each experiment. A slight drop between 0.13-0.67 units in pH value was observed in the diluted fraction and between 0.01 and 0.19 in S and SP feed solutions in the concentrated fraction. However, this level of pH variations did not affect the overall removal of cations and anions from the diluted fraction (Table 4). Temperature control plays an important role in avoiding effects caused by temperature depending processes as diffusion and changes in the affinity and selectivity through the cation-exchange and anion-exchange membranes [61]. Finally, the reduction and increase of the conductivity values in the diluted and concentrated fraction, respectively, represent in situ proof of the effectiveness of the ED treatment for the removal of ions from erythritol synthetic culture broth.

Desalination of Product in the Diluted Fraction
A final conductivity value of 0.28 mS/cm was reached after 78 min for S and SP solutions and min 84 for SCB solution, representing an increase of 8% in the treatment time compared to S and SP solution. This 8% increase is related to the differences in the composition of the solutions involved. SCB solution included two substances of different nature: the non-ionic surfactant, Tween 80, and the protein hydrolysate, peptone, widely used in biotechnological cultures. Table 4 shows the removal efficiency of ions in the diluted fraction, the total removal efficiency values were 93%, 93%, and 96% for S, SP, and SCB solutions. In all cases, the removal of individual anions and cations was higher than 86%. Among anions, chloride exhibited the highest and dihydrogen phosphate the lowest removal efficiency. For the cations, the highest removal efficiency was reached for potassium and the lowest for sodium.
The ion removal efficiency over time is presented in Figure 1. Similar removal efficiency was observed within the three solutions, meaning that complex solutions (SCB) presented the same general behaviour as the simplest solution (S). The addition of non-charged compounds to the SP solution did not affect the treatment time and removal of ions kept at the same level as S solutions. Slightly large error bars were observed for some ions in solutions, this is due to the difference in concentration levels and the accuracy of the analytical tools. Besides that, the concentration values exposed in the error bars can be explained by differences during sampling, or the response of the electrodes to the application of current. Notwithstanding, the samples followed a clear trend for ion removal. This consideration is also valid for the concentration of products going to the concentrate fraction ( Figure 2). The ion removal efficiency over time is presented in Figure 1. Similar removal efficiency was observed within the three solutions, meaning that complex solutions (SCB) presented the same general behaviour as the simplest solution (S). The addition of noncharged compounds to the SP solution did not affect the treatment time and removal of ions kept at the same level as S solutions. Slightly large error bars were observed for some ions in solutions, this is due to the difference in concentration levels and the accuracy of the analytical tools. Besides that, the concentration values exposed in the error bars can be explained by differences during sampling, or the response of the electrodes to the application of current. Notwithstanding, the samples followed a clear trend for ion removal. This consideration is also valid for the concentration of products going to the concentrate fraction ( Figure 2 [62] developed a study for the desalination of an erythritol culture broth using ion exclusion resins (Lewatit S3428 and S2568H). Three different fractions were obtained: a saline waste, recycle, and product fraction with a 95% degree of desalination (based on NaCl removal).
The results obtained in this study exhibited a remarkable performance of ions removal (96%), comparable to the reported by Rakicka et al. (2016) [62]. However, an advantage of ED is the absence of a waste stream, compared to large volumes of saline waste, operation recycles and regeneration-wastewater generated for ion-exchange treatments.   [62] developed a study for the desalination of an erythritol culture broth using ion exclusion resins (Lewatit S3428 and S2568H). Three different fractions were obtained: a saline waste, recycle, and product fraction with a 95% degree of desalination (based on NaCl removal).
biofouling, reducing the life of the membranes and the performance of separation.
To illustrate the level of losses in ED, Shen et al. (2005) [44] reported overall losses of about 16% of glutamine (initial concentration glutamine 32 g/L) while removing 95% of ammonium sulphate using ED, which are notably higher than the levels obtained in this work. This fact confirms the suitability of the use of ED in the purification of erythritol [64]. A possible alternative for reducing diffusional effects can be assessed through the decrease of treatment times by adopting a different strategy for the application of voltage.

Energy Consumption
Total energy consumption includes the energy required for the desalination and the energy consumed to pump the water fractions and electrode-rinsing solution through the ED stack. The total average energy consumption for the triplicates of S, SP, and SCB solutions is 3.96 kWh/m 3  The results obtained in this study exhibited a remarkable performance of ions removal (96%), comparable to the reported by Rakicka et al. (2016) [62]. However, an advantage of ED is the absence of a waste stream, compared to large volumes of saline waste, operation recycles and regeneration-wastewater generated for ion-exchange treatments.

Performance of Ion-Removal Rate in S, SP, and SCB Samples
The ion-removal rate is defined as the average percentage of ions removed divided per time (%removed/min); this rate is strongly influenced by the concentration of ions in the diluted fraction and the voltage applied. Hence, reductions in the ion-removal rate within the time depend on the driving force given by the electrical potential, the mobility of the ions, the molecule charges, and concentration gradient ( Table 2). The breakpoint time (BKP) indicates that sudden reductions in the removal rate occur, caused by the downturn in the driving force.
By identifying the BKP it was possible to characterise two periods of ions removal: a fast removal period and a slow removal period. Therefore, the determination of BKP time allows one to understand the processes and determine possible alternatives to improve CE, avoid back diffusion, product losses, and reduce energy consumption. Table 5 presents the breakpoint time (BKP), the reduction in the ion-removal rate after the BKP, the amounts of ions removed (%) until BKP and the average values of the rate of ions removal before and after BKP (% removed/min).
The effect over the mobilities of ions is depicted in the reduction of the ion-removal rate presented at different BKP time as shown in Table 5. In the first place, the ion mobility is decelerated due to the decreasing ionic strength of the diluted fraction among all three solutions (S, SP, and SCB), especially after the BKP time. In the second place, differences in ion mobilities between stages with increasing solution complexity can be observed, as discussed in the following paragraphs.
In S solutions, two BKP times were observed. The trend reflected by the average value of ions-removal rate before and after BKP time allows determining that chlorine was the ion with the highest mobility and overall removal. Notwithstanding the different levels of concentration described in Table 2, phosphate and sodium have a lower average value of ions-removal rate before and after BKP time, demonstrating lower mobilities than the other ions despite the initial concentration. Chloride was the first ion presenting a reduction in the removal rate of 90.3% after 25 min of treatment (BKP time). By that time, approximately 86.4% of the total chloride ions were already removed from the diluted fraction. The second BKP time was observed at 40 min for the other ions in solution with reduction levels in the removal rate between 52.3-81.8% when approximately 58.3-85.1% of the total ions were already removed (SO 4 2− > K + > NH 4 + > Mg 2+ > Na + > H 2 PO 4 − ). SP solutions presented slower removal rates compared to S solutions. In addition, four BKP times at 25, 40, 55, and 65 min were observed indicating a late onset of reduction in the removal rate, while slightly higher overall removal values than S samples were reached.
Despite requiring a longer time to reach the final conductivity value, the SCB solution presented similar levels of the total reduction in the removal rate (74.3-92.6%) compared to SP. In this case, three different BKP times 25, 45, and 55 min were obtained at the moment when approximately 69.5-95.8% of ions were removed. The average value of ions-removal before the BKP point showed an increase compared to SP data. SCB results allow us to make conclusions about three possible phenomena presented. The first regards the effects of non-charged compounds on possible polarisation concentration phenomena presented due to concentration gradients generated between the bulk liquid and the interfacial layer, which increases within the concentration of non-charged compounds. The second regards increases in the membrane resistance due to electrostatic and hydrophobic interactions presented between peptone and the membrane surfaces, specifically on cationic membranes [63,64]. The third phenomena relate to possible effects of polysorbate 80 as a non-charged surfactant on the increase of the average ions-removal rate and the mobility of the ions due to minimisations of the polarisation concentration phenomena, resulting in consequence in slightly higher overall removal values (Figure 1) while demanding longer treatment times. These results are in line with the study developed by Gohil et al. (2005) [65], who described the reduction in the polarisation concentration phenomena caused by the addition of surfactants. The authors proved the formation of a layer of surfactants due to their amphiphilic nature (hydrophobic head and hydrophilic tail). The ions can accumulate temporarily over this layer, reducing the concentration gradients in the membrane-solution interfacial layer, and therefore reducing possible polarisation concentration effects [65].

Current Efficiency (CE)
Current efficiency (CE) is a measure of the current utilisation to transmit ions instead for other uses, such as water splitting. An efficient process can separate the ions while ensuring a continuous movement and transfer of ions between the cathode and anode. Table 6 presents a comparison of the results obtained in this work and other solutions with similar and even higher complexity levels like lignocellulosic effluent, molasses, and brine solutions from downstream processes and waste treatment. This table includes the energy application approach, CE, and removal values. Studies that have reported high CE values (90-220%) indicating the dependence of the energy approach used and the initial levels of ions in diluted fraction by the order of 2-414 g/L [66][67][68][69]. By understanding this dependence, appropriated current densities or current application approaches can be selected, avoiding back diffusion of ions and water splitting, and therefore reach high CE [70]. These conditions can be obtained by applying the appropriate arrangement of the operating parameters such as voltage and feed flow rate with the initial concentration of ions. In this work, CE values between 49% and 54% were obtained for S, SP, and SCB solutions ( Table 6), meaning that around 54% of the energy applied is used to transport the ions, rather lower than most of the studies presented in Table 6. Even though it was not rigorously followed, there were no significant changes in the volume observed. However, water splitting and electro-osmosis transport could be responsible for the remaining 46% of the energy provided.

Recovery of Salts in the Concentrated Fraction
At the end of the treatment, the concentration of ions in the concentrated fraction was 1.5 to 2.5 times higher than the initial concentrations (Table 7). These concentration levels are in agreement with the increases in conductivity presented in Table 3. The results showed different behaviours analogously with the composition of the feed solutions. It can be observed that cation selectivity decreases as the complexity of the solution increases. This fact is analogous to the trend of ions removal presented in the last section and connected to the possible effects of peptone over CEM. With the concentration levels obtained, the number of ions in the concentrated solution is similar to the initial ions used for the culture stage, this represents an attractive alternative to reuse the ions either into erythritol culture media or other processes to reduce waste, close loops, and go deep into circular economy standards.

Losses of Products in the Concentrated Fraction
The concentration of erythritol, glycerol, and glucose in the concentrated fraction ( Figure 2) was measured to determine the losses of product passing the membrane. Concentrations of 0.13 g/L and 0.12 g/L of glycerol and 0.08 g/L of erythritol were determined in the concentrated fractions of SP and SCB solutions after 40 min of treatment. The presence of products in the concentrated fraction was determined after the BKP time when the rate of removal of ions decreased significantly for all the ions. These concentrations represented losses of about 2.31% and 1.96% of glycerol and 1.7% and 1.39% of erythritol in SP and SCB solutions.
Product losses in ED can be related to different factors: the charge of molecules in solution susceptible to pH value and pH changes (e.g., amphoterism), affinity by membrane material, and diffusional effects [48,69,72]. Considering the non-amphoteric nature of the substances and the pH stability of the process (Table 2), a movement of products due to protonation and electric charge is unfeasible. Consequently, losses of product going through the membranes can be mainly associated with diffusional phenomena like diffusion-driven for the difference in concentrations or electro-osmosis caused by water cotransport [69,72]. Besides losses of valuable material, this diffusional phenomenon can cause biofouling, reducing the life of the membranes and the performance of separation.
To illustrate the level of losses in ED, Shen et al. (2005) [44] reported overall losses of about 16% of glutamine (initial concentration glutamine 32 g/L) while removing 95% of ammonium sulphate using ED, which are notably higher than the levels obtained in this work. This fact confirms the suitability of the use of ED in the purification of erythritol [64]. A possible alternative for reducing diffusional effects can be assessed through the decrease of treatment times by adopting a different strategy for the application of voltage.

Energy Consumption
Total energy consumption includes the energy required for the desalination and the energy consumed to pump the water fractions and electrode-rinsing solution through the ED stack. The total average energy consumption for the triplicates of S, SP, and SCB solutions is 3.96 kWh/m 3 , 3.87 kWh/m 3 , and 4.10 kWh/m 3 of the treated solution, respectively. The energy consumed for the treatment of the SCB solution is increased compared to the S and SP solutions due to the increased desalination time required for the same diluted fraction quality. The energy consumed for desalination makes roughly 40%, for pumping the diluted and concentrated fractions 10%, while the pumping of the electrode-rinsing solution used around 50% of the total energy in all tested solutions (S, SP, and SCB).
The energy consumption for culture broth desalination presented in this work is in a similar range to the values reported by Doornbusch et al. (2020) [73]. They attained 4.88-1.72kWh/m 3 for desalting NaCl solution (approx. conductivity 49.3 mS/cm) to drinking water quality, however without accounting for the energy consumed for the pumping of the electrode-rinsing solution.

Discussion
The selectivity for removing ions presented in this study did not follow any pattern based on molar mass, charge, or initial concentration. For the cations, a higher selectivity towards potassium and magnesium than sodium was observed, which is in correspondence to results reported by Luo et al. (2020) [73]. Whereas the anions removal showed significantly higher selectivity for chloride compared to dihydrogen phosphate. Several authors have discussed ion selectivity remarking on some important issues such as the interaction between the ions and the AEM/CEM. These interactions include the groups attached to the membrane skeleton, membrane moisture, mobility in the membranes, rheological conditions [73][74][75][76][77].
The main effect caused by the addition of Tween 80 and peptone to SCB solution was the increase in the treatment time, without reductions in the ions-removal efficiency. (Figure 1). The increase in the treatment time is related to interactions between peptone and the ion exchange membranes. In long-term ED processes, this fact might represent membrane fouling, increasing the energy consumption, and affecting the overall performance of the separation. Ruiz et al. (2007) [62], Mikhaylin and Bazinet (2016) [78], and Nichka et al. (2020) [63], reported fouling on ion-exchange membranes due to the electrostatic and hydrophobic interactions between protein fractions and membranes. These interactions cause material deposition over the membrane surface, producing alterations that increase the membrane electrical resistance, reducing the rate of ion migration through the membranes and the overall performance of the separation. Some alternatives are used to mitigate fouling effects, namely the cleaning-in-place (CIP) routines and the implementation of pulsed electric field routines during treatment [64].
The ED application was capable of removing up to 96% of ions in the complex SCB sample. Chloride showed a separation behaviour significantly better in terms of mobility and removal efficiency in all the feed solutions evaluated, followed by sulphate, potassium, and ammonium. In general terms, the experimental results demonstrated a higher affinity and selectivity for anions than cations, explained by the faster movement of anions than cations at the same conditions. Similar results were reported by Kooistra (1967), who described the polarisation curves for the characterisation of ion-exchange membranes [79].
Considering that ED was demonstrated to be a suitable alternative for removing ions from erythritol culture broth, additional research work is required to improve CE and to increase the knowledge about membrane performance. These efforts could be directed towards the assessment of samples with a higher concentration of ions (concentrated culture broths, increasing the driving force), determining the maintenance requirements for the membrane and increase in the effectivity of the current applied (reduction of the resistance). To address current effectiveness, it is recommendable to explore different current application approaches, including constant current density [66] or new approaches as pulsed electric field routines [64].

Conclusions
Electrodialysis proved to be a suitable alternative to remove and recover salts from synthetic culture broth (SCB) solution, obtaining a remarkable desalination of 96%, comparable with values reported in the literature. The ion-removal rates followed the same trends in the three solutions (S, SP, and SCB), suggesting that ion selectivity and mobility are influenced by interactions between the species in solution and the groups attached to the membrane.
The salts removed from the synthetic culture broth and other solutions were concentrated 1.5 to 2.5 times in a second fraction. Producing a solution with similar characteristics to the initial culture media. Besides, less than 2% of products were found in the concentrate fraction after 40 min of treatment, with average energy consumption up to 4.10 kWh/m 3 .
Current efficiency values of 49-54% were observed during the experiments. These values are still below the average reported in the literature. To improve the CE performance and address product losses, it is necessary to implement strategies to reduce the resistance and improve energy utilisation. The strategies include the concentration of the fed solution, reduction in the total treatment time, and the assessment of different current applications (e.g., constant current treatment or pulsed electric treatment). ED for culture broth desalination represents an attractive option to be implemented in biotechnological downstreaming as an alternative to remove ions, returning them either into culture media or other processes, reducing waste generation, and closing loops moving towards circular economy processes.