Conductivity

The evolution of the conductivity of both HMH and KCl recovery solutions during the ED treatments in the four different conditions is presented in Figure 3. Firstly, the conductivity of the hydrolysate solution was mainly impacted by the current conditions (*p* < 0.05). The same trend was observed for the KCl recovery solution. Hence, when a current was applied on hydrolysate at pH 4, a decrease in the hydrolysate conductivity, from 4.57 ± 0.20 mS/cm to 3.40 ± 0.13 mS/cm, and an increase in the KCl solution conductivity, from 3.63 ± 0.07 mS/cm to 6.95 ± 0.13 mS/cm, were observed (*p* < 0.05). However, when the hydrolysate at pH 4 was treated without current, no change in conductivity of both the hydrolysate and the KCl recovery solutions was observed (*p* > 0.05). Regarding the ED treatment with current on hydrolysate at pH 7, a decrease in the conductivity of the hydrolysate, from 2.79 ± 0.06 mS/cm to 1.91 ± 0.11 mS/cm, and an increase in the conductivity of the KCl solution, from 3.69 ± 0.06 mS/cm to 5.10 ± 0.15 mS/cm, were also observed (*p* < 0.05), as for pH 4, but in a lower way. Finally, without current on hydrolysate at pH 7, no change was observed in the conductivity of both the hydrolysate and KCl recovery solutions, as for pH 4 (*p* > 0.05).

**Figure 3.** Conductivity evolution in HMH solutions at pH 4 and 7, and in the corresponding KCl recovery solutions, treated with and without current during ED treatments of 240 min.

The decrease in conductivity observed for hydrolysate at pH 4 and 7 during the ED treatments conducted with current was representative of their demineralization, resulting in final respective DRs of 25.62 ± 0.92% and 31.45 ± 2.49%. Compared to other demineralizations performed by ED using a similar configuration, these two DRs were quite low. Indeed, in their studies, Dufton et al. (2018) and Lemay et al. (2019) reached a final DR of almost 70% for acid and sweet whey, respectively [58,59]. However, as it was already mentioned, the fact that, in this present study, the pH of the hydrolysate was constantly adjusted to the desired pH value with HCl hindered the demineralization process. In addition, in accordance with the pH evolution observed, some water dissociation took place rapidly after the first twenty minutes of both of these treatments, thus counteracting the efficiency of the demineralization. Nevertheless, the conductivity of hydrolysate at pH 4 was surprisingly quite high, to totally justify the possible occurrence of water dissociation suggested by the pH evolution. Indeed, Dufton et al. (2018) noted the occurrence of water dissociation at an acid whey conductivity close to 3.0 mS/cm [58], whereas, in the case of HMH at pH 4, the water dissociation would have begun at a conductivity close to 4.0 mS/cm. On the contrary, the fact that the conductivity of the hydrolysate at pH 7 was lower than 2.5 mS/cm made the occurrence of water dissociation due to the limited availability of ionic species for electric current transport even more plausible [27,60]. Regarding the two KCl solutions of the ED treatments performed with current, the increase in their conductivity values was correlated to the demineralization of the corresponding hydrolysate solutions, and thus to

their mineralization, resulting in final MRs of 48.67 ± 0.50% and 27.65 ± 1.74% for the KCl recovery solution corresponding to the treatment of hydrolysate at pH 4 and 7, respectively. As a comparison, Dufton et al. (2018) obtained an MR of 74% for the recovery compartment of their study dealing with the demineralization of acid whey [58]. The lower MRs obtained in the case of this study were in line with the lower DRs of the hydrolysate discussed previously. Moreover, the conductivity of the KCl solution for hydrolysate at pH 4 with current increased linearly. This suggests the continuous migration of ionic species, among them H<sup>+</sup> species, into the KCl compartment. However, it was less obvious to qualify the evolution of KCl conductivity as linear for the ED treatment conducted on hydrolysate at pH 7 with current, indicating that the H<sup>+</sup> migration would not be as continuous in that case. Finally, the absence of changes in the conductivity of the hydrolysate solutions at pH 4 and 7, and the corresponding KCl recovery solutions for the ED treatments conducted without current, showed that no demineralization and no mineralization occurred. This was in line with the fact that no current was applied. Concerning ED on the hydrolysate at pH 4 without current, the di ffusion of H<sup>+</sup> suggested by the pH evolution was not perceived in terms of conductivity in that case. This observation was not consistent with the fact that H<sup>+</sup> species are known to impact conductivity [61]. However, the observed drop of pH was approximately of one unit, meaning that this variation was caused by around 0.000001 M of H<sup>+</sup> and, normally, the conductivity contribution of this ionic species would not have been perceived in 0.02678 M of 2 g/<sup>L</sup> KCl. The fact that such pH variation was visible suggests that the membrane integrity could have been altered by a potential fouling.

## Ash Content

The ash content of HMH and KCl recovery solutions was analyzed, and is presented in Tables 4 and 5. The ash content of the hydrolysate solution was only impacted by the current conditions applied (*p* < 0.05). The same trend was observed for the KCl recovery solution. At the initial time, the ash content of hydrolysate at pH 4 was 0.478 ± 0.005%. Hence, after ED with current, the ash content of the hydrolysate at pH 4 decreased to 0.400 ± 0.048% (*p* < 0.05), while no change was observed regarding the ash content of hydrolysate at pH 4 after ED without current (*p* > 0.05). Concerning the hydrolysate at pH 7, its ash content was 0.484 ± 0.007% at the initial time. This value was similar to those obtained for hydrolysate at pH 4 at the initial time (*p* > 0.05). After ED of the hydrolysate at pH 7 with current, the ash content decreased to 0.427 ± 0.020% (*p* < 0.05). The final ash contents of both hydrolysates at pH 4 and 7 after ED with current applied were similar (*p* > 0.05). However, ED carried out without current did not lead to any change in the ash content of the hydrolysate at pH 7 at the final time compared to the initial time (*p* > 0.05). With regard to the KCl recovery solution, its initial ash content was 0.166 ± 0.007%. After ED with current on the hydrolysate at pH 4, the ash content of the KCl recovery solution rose to 0.240 ± 0.022% (*p* < 0.05). At the same time, the ash content of the KCl recovery solution for ED with current on hydrolysate at pH 7 increased as well, compared to the initial time, since the final value obtained in that case was 0.230 ± 0.005% (*p* < 0.05). The ash content of both KCl recovery solutions after ED with current on the hydrolysate at pH 4 and pH 7 was similar (*p* > 0.05). Finally, no change was observed in the ash content of KCl recovery solutions after ED of the hydrolysate at pH 4 and 7 without current, compared to initial time (*p* > 0.05).

**Table 4.** Ash content of the HMH at the initial time and after the di fferent ED treatments (mean ± standard deviation).


Values within the same row with different letters (a–c) are significantly different *p* < 0.05 (Tukey test).


**Table 5.** Ash content of the KCl recovery solution at initial time and after the di fferent ED treatments (mean ± standard deviation).

Values within the same row with different letters (a–b) are significantly different, *p* < 0.05 (Tukey test).

The ash content of the hydrolysate at pH 4 and the hydrolysate at pH 7 at the initial time was surprisingly similar. Indeed, as the acidification of the hydrolysate to pH 4 required more HCl than the acidification to pH 7, it would have been more logical to obtain a higher value of ash content for the hydrolysate at pH 4 than for the hydrolysate at pH 7, due to the contribution of Cl– species formed by the dissociation of HCl molecules. Nevertheless, the fact that the di fference in HCl volume added to the initial hydrolysate to reach these two pH values was only in the order of a few milliliters might explain this result. The lower ash content of the hydrolysate at pH 4 and 7 after ED with current was related to their demineralization. This observation was in line with the decrease in the conductivity of these two solutions, as described previously. Based on the initial and final ash content values, the DRs obtained for the hydrolysate at pH 4 and for the hydrolysate at pH 7 was 16.71 ± 4.59% and 11.62 ± 4.71%, respectively. However, these two DR values were not consistent with those of 25.62 ± 0.92% for the hydrolysate at pH 4 and those of 31.45 ± 2.49% for the hydrolysate at pH 7, determined by means of the conductivity measurements. This discrepancy between the DR values would confirm the occurrence of water dissociation during these two ED treatments. Indeed, it was already observed that the calculation of DR based on conductivity values could be biased in the case of water dissociation, since this phenomenon leads to the formation of ionic species, namely H<sup>+</sup> and OH–, impacting the conductivity but not the ash content [61]. Regarding the corresponding KCl solutions, their increase in ash content was correlated to their mineralization. This was consistent with the increase in their conductivity mentioned before. For the KCl recovery solutions, based on their ash content values, the corresponding MRs were 29.27 ± 3.40% and 28.87 ± 3.24% for the hydrolysate at pH 4 and 7, respectively. However, the MRs obtained in the case of the conductivity measurements were 48.67 ± 0.50% and 27.65 ± 1.74% for the KCl recovery solutions after ED of the hydrolysate at pH 4 and 7, respectively. If the MRs of the KCl recovery solution of the hydrolysate at pH 7 were similar in both cases, at pH 4 they were highly di fferent depending on the equation used to calculate them. This could sugges<sup>t</sup> that a higher migration of H<sup>+</sup> species took place in the KCl recovery solution during ED of the hydrolysate at pH 4 with current. As explained previously, the MR of the KCl recovery solution was overestimated, based on conductivity values [61]. This could also explain why the conductivity evolved linearly for the KCl solution during ED with current of the hydrolysate at pH 4, but not for the KCl with current at pH 7. Finally, the fact that the ash content of the hydrolysate at pH 4 and 7 after the ED treatments without current remained unchanged compared to initial time was in line with the absence of demineralization, since no current was applied. This also explained the fact that no change was observed in the ash content of the corresponding KCl solutions.

## Global System Resistance

The global system resistance evolutions of ED treatments conducted at pH 4 and 7 with current are shown in Figure 4. Regarding the treatment conducted at pH 4, the global system resistance increased significantly from 25.00 ± 0.00 Ω to 88.89 ± 9.62 Ω (*p* < 0.05), while those concerning the treatment carried out at pH 7 increased significantly from 33.33 ± 0.00 Ω to 125.00 ± 0.00 Ω (*p* < 0.05). Moreover, for both treatments, the increase in global system resistance was even more visible after the first twenty minutes of treatments.

**Figure 4.** Global system resistance evolution during ED treatments of 240 min at pH 4 and pH 7, with current.

The ED treatments performed on the hydrolysate at pH 4 and 7 presented a 3.5-fold increase and a 3.75-fold increase in global system resistance, respectively. Dufton et al. (2018) experienced a similar increase in global system resistance during the demineralization of acid whey by ED [58]. Such an increase in global system resistance could not only be due to the demineralization process but also to the occurrence of water dissociation as a cause or as a consequence of potential membrane mineral and/or protein fouling, as previously observed by Dufton et al. (2018) [58]. This is corroborated by the fact that this increase in global system resistance for both ED treatments took place at the same time as the changes in pH evolution mentioned previously. Another interesting point to mention was that, during ED performed on hydrolysate at pH 7, the intensity dropped considerably after the first twenty minutes (data not shown). The decrease in intensity was representative of a lack of ionic species to carry the electric current, and was thus consistent with the increase in global system resistance that was even more visible after the first twenty minutes of treatment. This observation was in line as well with the lower conductivity of the hydrolysate at pH 7 noted rapidly after the beginning of the treatment, suggesting that the LCD was reached. This could explain why the global system resistance was, in that case, so high. However, the fact that the global system resistance of ED conducted on the hydrolysate at pH 4 was lower could indicate that, at that stage, it was still not possible to clearly identify whether the LCD was reached and thus explain the water dissociation.

## Membrane Thickness

The evolution of the membrane thickness over the different ED treatments is shown in Figures 5 and 6. None of the membranes evidenced an increase in thickness over ED, with and without current, performed on the hydrolysate at pH 4 (*p* > 0.05) (Figure 5). Regarding the ED treatments carried out on the hydrolysate at pH 7, both AEM1 and CEM3 showed an increase in thickness (*p* < 0.05) for the experiment without current, while only AEM4 presented an increase in thickness (*p* < 0.05) for the experiment with current (Figure 6).

Membrane thickness is an indicator of membrane integrity, and more particularly of membrane fouling [58]. This could indicate that no fouling occurred during ED treatments on the hydrolysate at pH 4, showing the efficiency of the NaCl rinsing after each run. This was consistent with the visual observations of the membranes after each run as well. However, even if a few membranes evidenced an increase in thickness after the ED treatments on the hydrolysate at pH 7, the final membrane thickness

values were still representative of those reported in the literature. Indeed, Lemay et al. (2019) reported average values of 0.142 ± 0.006 mm and 0.141 ± 0.005 mm for AEMs and CEMs, respectively [32]. Therefore, based on membrane thickness evolution, no fouling phenomena seemed to happen at that stage regarding the treatments on the hydrolysate at pH 7.

**Figure 5.** Membrane thickness before and after each 4-h ED treatment conducted at pH 4, (**a**) without current and (**b**) with current. Values with different letters corresponding to the same membranes are significantly different *p* < 0.05 (Tukey test).

**Figure 6.** Membrane thickness before and after each 4-h ED treatment conducted at pH 7, (**a**) without current and (**b**) with current. Values with different letters corresponding to the same membranes are significantly different, *p* < 0.05 (Tukey test).

## Membrane Conductivity

The membrane conductivity evolution over the different ED treatments is presented in Figures 7 and 8. Regarding ED conducted on the hydrolysate at pH 4, AEM2 and all the CEMs for the experiment conducted without current, and all the CEMs only for the experiment conducted with current, evidenced a decrease in their conductivity over time (*p* < 0.05) (Figure 7). Concerning ED on the hydrolysate at pH 7, AEM3, AEM4 and all the CEMs for the experiment without current, and all the AEMs except AEM4 and all the CEMs for the experiment using current, experienced a decrease in their conductivity over time (*p* < 0.05) (Figure 8).

**Figure 7.** Membrane conductivity before and after each 4-h ED treatments conducted at pH 4, (**a**) without current and (**b**) with current. Values with different letters corresponding to the same membranes are significantly different *p* < 0.05 (Tukey test).

(**b**)

**Figure 8.** Membrane conductivity before and after each 4-h ED treatments conducted at pH 7, (**a**) without current and (**b**) with current. Values with different letters corresponding to the same membranes are significantly different *p* < 0.05 (Tukey test).

Similarly to membrane thickness evolution, the evolution of membrane conductivity can be considered to be an indicator of membrane integrity [58]. Firstly, the conductivity of both CEMs and AEMs before any run was comparable to the following values reported in the literature: 5.197 ± 0.257 ms/cm and 8.960 ± 0.442 mS/cm for AEMs and CEMs, respectively [32]. Then, all the four ED conditions evidenced a change in membrane conductivity as a function of time. In their study, Lemay et al. (2019) also noted a decrease in membrane conductivity after sweet whey demineralization [32]. In that case, the observed drop was attributed to the substitution of the counterions present in the initial membranes by divalent ionic species of the sweet whey having lower conductivity values, resulting in a decrease in membrane conductivity [32]. Nonetheless, in the present study, the fact that membranes evidenced a decrease in conductivity not only after the ED treatments conducted with current but also after those performed without current may sugges<sup>t</sup> that another explanation could be involved. More specifically, regarding the ED treatments carried out on hydrolysate at pH 4, a decrease in membrane conductivity was already noted after the first run. This could indicate that, independently of the current conditions, ionic compounds present in the hydrolysate solution interacted with the membranes. Indeed, previous works have already showed that charged compounds such as peptides and amino acids could interact with the boundary layers of membranes [56]. As the decrease in membrane conductivity was mainly observed for CEMs, this means that the compounds involved were cationic and interacted with the negatively charged sulfonic groups present in the CEMs. More specifically, Persico et al. (2017) showed in their study the ability of peptides containing histidine (pKa of 6.0), lysine (pKa ~10.5) and arginine (pKa ~12.5) residues, in addition to their amine group at the N-term (pKa ~9.8), to interact electrostatically with the negatively charged sulfonic groups of CEMs due to the positive charges they carry, even when no current was applied [62]. Free arginine could even be considered to be a major agen<sup>t</sup> responsible for the fouling of CEMs [63]. Since HMH is mainly composed of amino acids containing materials (Table 1) including a high amount of arginine in both bound and free forms [1], the important decrease observed in the CEM conductivity could be explained by the neutralization of the fixed membrane charges by these molecules, among them arginine amino acid. At that stage, it is worth it to mention that, after each run, NaCl rinsing was performed. The aim of such rinsing was to reduce the electrostatic interactions occurring between the ionogenic constituents of the membranes and amino acids containing materials from the hydrolysate due to the high ionic strength of the salt rinsing solution [64]. However, the fact that fouling was observed even after this rinsing indicates that other interactions than electrostatic ones could be involved between the ionogenic groups of membranes and components from the hydrolysate. This suggests that hydrophobic interactions could be involved between volatile compounds and membrane constituents as well, as that type of interaction was not impacted by NaCl rinsing [64]. This hypothesis was already formulated in the studies of Cros et al. (2005) and Chindapan et al. (2011), in which they ascribed the decrease in the abundance of certain volatile compounds to hydrophobic interactions occurring with membrane components [25,26]. In addition, the observed decrease in membrane conductivity could be explained by another phenomenon. Indeed, it was already shown that, depending on the nature of the groups from both the membranes and the matrix present at the membrane interface, the catalysis of water dissociation during an ED treatment could take place even if the LCD was not reached [56,65]. More precisely, at the CEM interface, the catalytic action of carboxylic acid present at the C-term of the peptides was already proven [56]. Therefore, the decrease in pH observed in the KCl recovery solution corresponding to the ED treatment of the hydrolysate at pH 4 with current, and the increase in its conductivity, could be e ffectively due to the occurrence of water dissociation resulting from the action of cationic catalysts compounds, such as peptides or free amino acids involving a carboxylic group, at the boundary layers of the CEMs. On the contrary, for ED of the hydrolysate at pH 4 without current, since a decrease in the CEM conductivity was evidenced, but with no huge pH change of the KCl solution, this could sugges<sup>t</sup> that the interaction of cationic compounds took place, resulting in membrane fouling without bringing about water dissociation due to the absence of current. Regarding ED on the hydrolysate at pH 7, since both AEMs and CEMs presented a decrease in conductivity for

both current conditions applied, this may sugges<sup>t</sup> not only that cationic compounds interacted with CEMs but also anionic ones with ammonium groups of AEMs. This observation was consistent with the study of Persico et al. (2016), in which it was observed that fouling of AEMs was absent at acidic pH, while it tended to be more important at a pH close to neutral [64]. This was due to the fact that, at a pH close to neutral, negative residues of carboxylic acid present at the C-term (pKa ~2.1) of peptides, or at the side-chains of aspartic and glutamic acids (pKa ~4.0), were able to interact electrostatically with the positively charged ammonium groups of AEMs [64]. As both aspartic and glutamic acids are present in HMH, in both bound and free forms [1], they could have been responsible for the fouling observed on the AEMs due to their carboxylic acid residues leading to the observed decrease in conductivity. Nevertheless, as noted for CEMs, the fact that fouling was evidenced even after the NaCl rinsing could sugges<sup>t</sup> that other interactions than electrostatic ones could be involved. Moreover, similarly to what was mentioned previously as well, water dissociation phenomena could have taken place at the interface of AEMs, due to the catalytic action of both compounds present in the AEMs and in the hydrolysate [56,65,66]. More precisely, the conversion of the initial quaternary ammonium groups (-N+(CH3)3) present in the AEMs into tertiary (=N(CH3)2) or secondary amine (≡N(CH3)) is a factor accelerating water dissociation due to the lone electron pair carried by the newly formed groups [56,66]. Weak-acid anions [65], such as glutamic and aspartic residues present in the studied hydrolysate, can catalytically accelerate the water dissociation as well. Furthermore, this hydrolysate also contains nucleic acids that carry a negative charge due to the presence of phosphoric acid in its anion form. Therefore, nucleic acids could have also established interactions with AEMs, or could have played a role in the catalysis of water dissociation. At this stage, it is worth to mention that the occurrence of water dissociation due to the generation of H<sup>+</sup> and OH– could be a factor promoting membrane fouling by increasing the interaction with amino-containing materials at the membrane interface [67]. In addition, the fact that both types of IEM were fouled for the condition at pH 7 could explain why the global system resistance increased more significantly than those during the ED treatment conducted on the hydrolysate at pH 4. Another interesting point to note is that the standard deviations corresponding to the global system resistance of the ED treatment of the hydrolysate at pH 7 were quite important compared to those displayed for the ED treatment of the hydrolysate at pH 4. This would probably be due to the progressive decrease in conductivity observed for both types of IEM. It is worth to mention that at that stage, the potential interactions occurring between constituents of the studied hydrolysate, including volatile compounds and the IEM, seemed to be the most plausible explanation. Therefore, the occurrence of water dissociation observed during the ED treatment of the hydrolysate at pH 4 with current could be due more to the action of catalysts than to the reaching of the LCD.

#### 3.2.2. Hydrolysate and Recovery Solutions Analyses

## Volatile Compound Analysis

#### Most Potent Odor-Active Compounds

The abundance of compounds determined to be the most potent odorants of the HMH over the di fferent ED and deaerator treatments is shown in Table 6. Firstly, concerning the composition of the hydrolysate at the initial time, the results indicate that pH had an important impact on the volatile compounds' abundance. It appeared that the abundance of the majority of the most potent odor-active molecules significantly decreased while pH increased from 4 to 10. For example, the abundance of 3-methylbutanal, 2,3-pentanedione, pentanal, hexanal, (Z)-4-heptenal, heptanal, methional, (Z)-6-octen-2-one, (E,E)-2,4-heptadienal, octanal, 2-nonanone and (E,Z)-2,6-nonadienal dropped by at least 50% between the hydrolysate at pH 4 and the hydrolysate at pH 10 at the initial time (*p* < 0.05). 1-methyl-1H-tetrazole was the only compound that did not follow this trend. In fact, it was identified in the hydrolysate at pH 7, but it was not present in the hydrolysate at pH 4 and pH 10. In general, pH is known to be a major factor influencing the content of volatile compounds. Indeed, volatile compounds are able to interact with molecules like lipids through hydrophobic interactions, and amino acids constituents such as proteins, peptides and free amino acids through covalent irreversible bindings, in addition to hydrophobic and ionic interactions [68–70]. Among these di fferent interactions, those taking place between volatile compounds and amino-acid containing compounds are the most impacted by pH, as this factor modifies the conformation and charge of proteins, peptides and free amino acids, and thus the ability of binding of volatile compounds [69]. Based on this fact, two hypotheses can be made concerning the general decrease in volatile compounds observed while pH increased. The first could be that alkaline pH might be responsible for the breaking of interactions taking place between volatile compounds and amino groups. As the targeted compounds are volatile, breaking these interactions could promote their loss. On the contrary, the second could be that the lower abundance of volatile compounds observed in the HMH at pH 10 at the initial time may be representative of a higher interaction with amino acid materials. As no study has been carried out regarding the impact of pH on the retention of volatile compounds by amino acid constituents from HMH materials so far, it was not obvious to clearly validate one hypothesis rather than the other. Nevertheless, some studies with similar purposes were already conducted on milk proteins [71,72], as well as on animal tissues proteins and peptides [69,73]. These studies showed that there was a general trend of amino acid containing molecules, such as peptides, to retain volatile compounds to a higher extent while pH increased. Several explanations are involved, depending on the proteins. For example, the milk protein β-lactoglobulin is reported to bind a larger proportion of volatile compounds at pH 9 than at pH 3. The increase in retention ability is, in this case, explained by better access to the hydrophobic amino acid residues of β-lactoglobulin due to conformation changes occurring under alkaline conditions [70,72]. Interestingly, leucine, a hydrophobic amino acid, is present in both β-lactoglobulin and HMH. Leucine is even present in high proportion in the latter (Table 1). It could be suggested that leucine might participate in the retention of volatile compounds, and while pH increases, the loss of the proton H<sup>+</sup> on the amine group could promote this phenomenon. Independently of pH, Meynier et al. (2004) observed the unavailability of lysine and histidine of milk proteins in the presence of aldehydes, suggesting a potential interaction occurring between these amino acids and volatile compounds. It was proposed to explain this loss that the carbonyl group of aldehydes could react with the primary amine of lysine either by Michael addition or by Schi ff base formation. Concerning histidine, it was suggested that aldehyde and, preferentially, alkenal could react with the imidazole ring of histidine [74]. Since among the 15 compounds identified as being the most potent odor-active of HMH, 11 are aldehydes and one of them is an alkenal, namely (Z)-4-heptenal, and since lysine and histidine are both present in this product (Table 1), it could be possible that these interactions occur between these volatile compounds and these amino acids. Histidine is also a constituent of carnosine, a dipeptide found in animal tissues, and whose ability to retain volatile compounds is also known to increase while pH increases [69]. In that case, as pH a ffects the retention of volatile compounds, it could be proposed, similarly to leucine, that the loss of the proton H<sup>+</sup> on the imidazole ring occurring under alkaline conditions could promote the interaction between histidine and aldehydes. Interestingly, HMH contains a high amount of arginine (Table 1), which is an amino acid with an amine-containing side-chain similar to lysine and histidine. Based on this, it should be proposed that arginine was also involved in the interactions, explaining partially the decrease in volatile compound abundance observed. Therefore, the hypothesis that seemed to be the most plausible regarding the decrease in the abundance of volatile compounds at pH 10 compared to pH 4 and pH 7 would be those implying a higher degree of interaction occurring between volatile compounds and amino acid containing compounds, such as peptides present in the hydrolysate. In that case, the fact that 1-methyl-1H-tetrazole was not detected in the hydrolysate at pH 4 and 10 might sugges<sup>t</sup> that this compound could have more interactions at these pH values than at pH 7, allowing its detection at this pH value only. Also, it is of interest to mention that, while volatile compounds are bound to other components, both their release and perception are hindered [69]. This means that HMH should be globally less odorous at pH 10 than at pH 4 and pH 7.


*Membranes* **2020**, *10*,

**Table 6.** Abundance of the most-potent odor active compounds of HMH (×107 Arbitrary Unit (A.U)) at the initial time and after the different treatments (mean ±

127

Then, regarding the ED treatments, no significant di fference was globally observed between the hydrolysate at a given pH at the initial time and the hydrolysate treated with or without current at the final time. More precisely, if attention is paid to the ED treatments conducted without current, the fact that no decrease in the content of the targeted volatile compounds was observed would sugges<sup>t</sup> that a simple circulation of the hydrolysate solution for 240 min, independently of its pH, was not su fficient enough to allow a loss of these molecules due to their volatile state. Concerning ED conducted with current, no change in the content of volatile compounds occurred except for (Z)-6-octen-2-one, (E,E)-2,4-heptadienal and (E,Z)-2,6-nonadienal whose abundance was inferior at final time for ED on the hydrolysate at pH 4 (*p* < 0.05). These results could be representative of the non-migration of volatile compounds. Since ED is a process based on the migration of charged compounds, and since the targeted compounds were not supposed to be charged under the conditions tested, it was not surprising, at the first glance, to obtain such results. However, in their studies, Cros et al. (2005) and Chindapan et al. (2011) observed that an ED treatment could lead to a drop of volatile compounds even if they are not charged [25,26]. Several points could explain the discrepancy between these two studies and the present one. The first could be that the compounds whose abundance dropped during ED treatments in the studies of Cros et al. (2005) and Chindapan et al. (2011) were not the same compounds targeted in the present study. Indeed, Cros et al. (2005) observed a significant decrease in the non ionizable (Z)-4-heptenal, 2,3-butanedione, 3-octen-2-one and limonene compounds. The only compound that this study and the present one had in common was (Z)-4-heptenal. Nevertheless, it is noteworthy to mention that Cros et al. (2005) noticed the important decrease of the compounds listed before only while the LCD was reached. Di fferent hypotheses were formulated to explain such a decrease under this specific condition in this study. The first was that the formation of protons H<sup>+</sup> and hydroxyls OH– resulting from water dissociation under this critical condition could have altered volatile compounds, explaining their decrease. Another was that the LCD could have also brought about a local membrane heating, potentially leading to a thermal degradation of volatile compounds. Or, simply, the volatile compounds could have been adsorbed on the membranes through hydrophobic and ionic interactions [26]. The migration of these molecules was not considered to be a potential explanation, as none of them were found in the recovery solution. Regarding (Z)-4-heptenal only, Cros et al. (2005) hypothesized its hydrogenation in heptanal as a possible explanation for its decrease [26]. Since not all the ED treatments conducted in the present study seemed to have evidenced reaching LCD, it was not possible to totally verify all the hypotheses formulated by Cros et al. (2005). However, as suggested by the analyses of the parameters of the ED treatment conducted on the hydrolysate at pH 7, it seems that this condition experienced the reaching of LCD. As no change regarding the volatile compound content was observed in that case, it may indicate that the LCD was not a su fficient condition to lead to a decrease in the abundance of these compounds. In addition, the fact that both ED treatments conducted with current presented water dissociation in the present study could show that the generation of H<sup>+</sup> and OH– species could not be e ffectively responsible for the alteration of the volatile compounds. On the contrary to Cros et al. (2005), Chindapan et al. (2011) did not reach the LCD condition in their study. Despite this fact, they observed a significant decrease in 2,6-dimethylpyrazine, phenol and carboxylic acids (acetic acid, butanoic acid, 2-methylbutanoic acid, pentanoic acid, 4-methylpentanoic acid) while the ED treatment was performed, to reach a salt concentration of 2% in the treated fish sauce. Chindapan et al. (2011) gave two main reasons for the loss of these compounds: either their adsorption on the membranes or their transport through the membranes occurring at the same time as electroosmosis. Nevertheless, as no mention concerning the composition of the recovery solution was made, it was not possible to know if the latter reason was plausible in that case [25]. Concerning the decreased in (Z)-6-octen-2-one, (E,E)-2,4-heptadienal and (E,Z)-2,6-nonadienal observed in the hydrolysate at pH 4 treated with current at final time, three hypotheses could be made, based on those previously mentioned. The first one would be that a slight loss of these molecules due to their volatile state happened during the ED process. However, as no decrease of these compounds was observed for other conditions, this hypothesis does not seem highly plausible. The second hypothesis would be that preferential interactions occurred at pH 4 between these three volatile compounds and other constituents of the HMH. Finally, the last hypothesis would be that these compounds preferentially adsorb on the membranes due to hydrophobic interactions. This last hypothesis appears to be the most probable, based on the membrane conductivity evolution discussed previously.

The content in volatile compounds of the KCl recovery solutions was analyzed for each condition, and is listed in Table 7. The results show the unchanged presence of 3-methylbutanal, hexanal, heptanal, benzaldehyde, octanal and 2-nonanone in KCl solution at final time independently of the ED treatment. It is worthwhile to mention that none of these compounds were detected in the KCl solution at the initial time. Except for benzaldehyde and hexanal, in all cases, the presence of volatile compounds in the recovery solution can be considered to be trace. This should probably be due to a punctual contamination of these compounds due to their volatility from the hydrolysate to the KCl solution. The fact that this phenomenon could be considered to be punctual was accredited by the generally high values of standard deviations proportionally to those of means, and even sometimes the higher values of standard deviations compared to the corresponding means. However, another explanation could be involved for hexanal and benzaldehyde. Regarding hexanal, its presence in the KCl solution could be due to its di ffusion or migration. However, based on the membrane conductivity analysis discussed previously, it seemed that some interactions with membrane components also occurred during the di fferent ED treatments. Therefore, another explanation could be that hexanal may have interacted with the sulfonic groups present in the CEMs, resulting in its release into the KCl compartment thereafter. Interestingly, the same trend was not found for compounds similar to hexanal, such as pentanal and heptanal. In that case, the di fferences observed should probably be due to the presence of hexanal in higher quantity in HMH, compared to pentanal and heptanal. Regarding benzaldehyde, the same explanations as those mentioned for hexanal could be involved. Nevertheless, on the contrary to hexanal, the fact that benzaldehyde was found in higher abundance only in KCl solution of the hydrolysate at pH 4 treated with current may indicate that a special mechanism was involved in that case. Initially, as benzaldehyde is not charged, it was not supposed to migrate. However, its recovery in the KCl solution might sugges<sup>t</sup> that benzaldehyde could have either established interactions with another positively charged constituent that migrated into the CEMs, or that benzaldehyde established an interaction with the sulfonic groups of the CEMs, resulting in its release into the KCl compartment thereafter. Nonetheless, assuming that an interaction with another constituent could explain the presence of benzaldehyde in that case, this interaction could have been broken once this compound was finally in the KCl solution, as its detection was still allowed. Indeed, as mentioned previously in this study, while volatile compounds interact with other constituents, it hinders their detection [69]. Moreover, the results could show that this potential interaction occurred only at pH 4, as a similar trend was not found at pH 7. The charged compounds in that case could be histidine, present mainly in its free form in the HMH [1] as, at pH 4, its side-chain was totally protonated (pKa ~6.0), allowing its migration to the cathode through CEMs, while at pH 7 this latter was in its non-charged form. Interestingly, the presence of (Z)-6-octen-2-one, (E,E)-2,4-heptadienal and (E,Z)-2,6-nonadienal, whose abundance was lower in the hydrolysate at pH 4 treated with current at the final time, was not found in the corresponding KCl solution. Therefore, supposing that the hypothesis formulated before aiming that these compounds could have established interactions with membranes, this could indicate that none of these three compounds were released into the KCl compartment thereafter.

Finally, the performance of ED to decrease the abundance of the most potent odor-active compounds of the HMH was compared to that of a deaerator (Table 4). In this case, in addition to pH 4 and 7, the treatment was also conducted on the hydrolysate at pH 10. Compared to pH 4-hydrolysate at the initial time, the deaerator allowed the decrease in seven compounds (*p* < 0.05), namely 3-methylbutanal, 2-methylbutanal, 2,3-pentanedione, pentanal, hexanal, 2-nonanone and (E,Z)-2,6-nonadienal. A similar trend was observed for pH 7-hydrolysate, for which the deaerator allowed a drop in the abundance of the seven following compounds (*p* < 0.05): 3-methylbutanal, 2-methylbutanal, 1-methyl-1H-tetrazole, pentanal, hexanal, (Z)-6-octen-2-one and 2-nonanone. It is of interest to note that, in this case, the deaerator conducted the total loss of 1-methyl-1H-tetrazole, pentanal and (Z)-6-octen-2-one. That could sugges<sup>t</sup> that the ability of such device to remove volatile compounds was better at pH 7 than at pH 4. As it was mentioned previously, the hypothesis that looked more plausible to explain the di fference in volatile compound content between the initial hydrolysate at di fferent pH values was the following: while pH increased, higher interactions between the volatiles and other constituents occurred. This could be in line with the results of the deaerator for the hydrolysate at pH 4 and pH 7. Indeed, it seemed that this device could break weak interactions occurring between volatile compounds and other compounds present in the hydrolysate, resulting in a better decrease rate at pH 7 than at pH 4. Nevertheless, the deaerator did not lead to a decrease in the volatile content while the hydrolysate was treated at pH 10. This could indicate that, at pH 10, the chosen hypothesis was not enough to totally explain the mechanisms involved. It could be supposed that, at pH 10, a certain proportion of volatile compounds could take part in strong interactions, such as covalent bonds, but at the same time, some of them could have been lost due to their promoted passage in the headspace of the hydrolysate solution as well, or simply altered, hindering their detection.


**Table 7.** Abundance of the most-potent odor active compounds of HMH recovered in the KCl solution (×10<sup>7</sup> A.U) after the di fferent ED treatments (mean ± standard deviation).

Values within the same row with different letters (a–c) are significantly different, *p* < 0.05 (Tukey test).

#### TMAO, TMA and DMA

The TMAO, TMA and DMA contents of HMH are shown in Table 8. Firstly, concerning the hydrolysate at the initial time for the three tested pH values, their concentration in TMAO, TMA and DMA was similar (*p* > 0.05). The only di fference observed was related to the content of TMAO of the hydrolysate at pH 10, which was 20 times lower (*p* < 0.05) than those of the hydrolysate at pH 4 and 7 at the initial time. In this context, it is worth to mention that the procedures used for the analysis of TMAO, TMA and DMA recommend to alkalize samples of interest, to allow a better detection of these molecules based on their higher release into the sample headspace [41]. Therefore, the huge decrease in TMAO content observed in the hydrolysate at pH 10 could be related to a loss following their release into the headspace of the hydrolysate solution due to its high volatility.

Then, concerning the content in TMAO, TMA and DMA after the four ED treatments, no di fference was observed between the hydrolysate at initial and final times. ED treatments were especially designed to assess whether TMA and DMA, two positively charged compounds at pH 4 and 7, were able to migrate. However, the results indicated that no migration happened while experiments were conducted with current. As suggested by the ED parameter analyses, some water dissociation took place during the treatments conducted at pH 4 and 7 with current. Therefore, it could be hypothesized that TMA and DMA had been in competition with the generated H<sup>+</sup> to migrate into the CEMs, and that H<sup>+</sup> could have prevailed over TMA and DMA. Another explanation could be that fouling occurring on CEMs, as suggested by the membrane conductivity analysis, hindering the migration of TMA and DMA, thus explaining such results. Chindapan et al. (2011) experienced, in their study, a decrease in TMA, and explained this result by its loss occurring during ED due to its high volatility [25]. However, the results obtained in the present study may indicate that TMA could not be lost as easily, since ED

treatments carried out without current did not evidence any change in the content of this compound. Moreover, the fact that no change in the concentration of TMAO was observed between the hydrolysate at the initial and final times treated with current was more expected. Indeed, this molecule is a zwitterion, and the absence of global charge makes it less likely to migrate during an ED process.

The contents in TMAO, TMA and DMA of KCl recovery solution were analyzed (Table 9). The results show that the initial solution was free of these compounds, while the KCl solution at the final time of all the tested conditions only evidenced the presence of TMAO. The presence of TMAO in the recovery solution of treatments conducted with current was not expected, as the global charge of this compound was neutral. However, the fact that TMAO was present in the recovery solution of treatments carried out without current as well could sugges<sup>t</sup> that another mechanism than electromigration could be involved. In addition, the concentration of TMAO in the di fferent KCl recovery solutions was surprisingly as important as those of the corresponding hydrolysate at the initial time, and since the concentration of this compound in the hydrolysate did not evolve during the di fferent ED treatments, this could sugges<sup>t</sup> that new TMAO was generated over the time. The most logical explanation at the first glance could have been that some TMA evidenced oxidation, resulting in the formation of much more TMAO. However, this was not possible in the case of this study, as the initial hydrolysate, independently of its pH, had too low a content of TMA. This means that more complex mechanisms occurred. TMAO is traditionally produced from nitrogenous compounds, such as choline, betaine or carnitine, through metabolism pathways involving enzymes and gu<sup>t</sup> microbiota [75]. Interestingly, HMH contains phospholipids whose choline can be a constituent and carnitine as well (Table 1). Even if metabolism pathways could not be involved in that case, it could be supposed that some TMAO was generated from the choline of phospholipids and carnitine through other reactions, such as oxidation. Nevertheless, this could only explain the occurrence of much more TMAO compared to the initial time, and not its recovery in the di fferent KCl solutions. Another hypothesis could be that reactions between constituents of HMH, such as the choline of phospholipids or carnitine, as mentioned before, and those of CEMs could have taken place. This latter hypothesis seems to be even more plausible, as the analysis of ED parameters, and more specifically those regarding membrane conductivity, revealed that some interactions happened between hydrolysate constituents and membranes. However, at this stage, it is not possible to e ffectively favor one hypothesis rather than another one. A last point that is worth mentioning is that the absence of TMA and DMA in the KCl recovery solutions of treatments conducted with current was e ffectively representative of their non-migration.

Finally, the comparison of the performance of ED with those of a deaerator was assessed. The results are presented in Table 8. They indicate that the deaerator was only e ffective in decreasing the concentration of TMAO (*p* < 0.05) of the hydrolysate at pH 4 and 7. Regarding the hydrolysate at pH 10, this device had no e ffect on its composition. These results were consistent with those obtained for the most potent odor-active compounds, as discussed before. However, in that case, the fact that no impact regarding the TMAO content of the hydrolysate was observed gave credit to its loss following its release into the headspace of the hydrolysate sample, promoted by alkaline conditions and occurring before the deaerator treatment, as mentioned previously.


**Table 8.** TMA, DMA and TMAO content of HMH at the initial time and after the different treatments (mean ± standard deviation).

**Table 9.** TMA, DMA and TMAO content of KCl recovery solution at initial time and after the different ED treatments (mean ± standard deviation).


Values within the same row with different letters (a–b) are significantly different *p* < 0.05 (Tukey test).

## **4. Conclusions**

GC-MS analysis allowed the identification of a total of 86 volatile compounds in the HMH. Among these 86, the following 15 were determined to be the most potent odor-active compounds of this hydrolysate by GC-O, combined with the detection frequency method: 3-methylbutanal, 2-methylbutanal, 1-methyl-1H-tetrazole, 2,3-pentanedione, pentanal, hexanal, (Z)-4-heptenal, heptanal, methional, benzaldehyde, (Z)-6-octen-2-one, (E,E)-2,4-heptadienal, octanal, 2-nonanone and (E,Z)-2,6-nonadienal. In addition, the HS-GC-NPD analysis revealed the presence of TMAO, TMA and DMA in the HMH. Furthermore, the performance of ED as a deodorization method was compared to that of a deaerator device. The results showed that pH had a huge impact on the volatile compound contents of the hydrolysate at the initial time. In fact, the abundance of the targeted molecules was lower at pH 10 than at pH 4, and intermediate at pH 7. While the pH increased from pH 4 to 7, volatile compounds were more involved in interactions with amino-acid-containing materials, explaining their lower availability and thus their lower abundance. However, at pH 10, more than one mechanism could be involved. Indeed, part of the targeted odor-active compounds should have been lost due to their volatility, while another part participated in irreversible bonds or was altered, hampering their detection. Regarding TMAO specifically, its lower content in the hydrolysate at pH 10 could be related to its loss resulting from a greater release into the headspace solution. On the other hand, ED did not a ffect the volatile compound contents of HMH. Concerning ED treatments conducted with current, no migration of volatile compounds, and more precisely no migration of TMA and DMA, occurred. Two phenomena were considered to be the main possible limiting process conditions regarding the removal of the targeted compounds. The first was the occurrence of fouling on IEM due to both electrostatic and hydrophobic interactions between IEM and HMH constituents, including volatile compounds. The second was the occurrence of water dissociation on the IEM interfaces due to the reaching of LCD, as well as the presence of water dissociation catalyzers involved in both IEM and HMH constituents. Moreover, the fact that ED treatments without current did not impact the volatile compound contents implied that no loss of these molecules due to their volatile nature happened during the circulation of the hydrolysate solution in an ED system. Interestingly, independently of the pH and current conditions of the ED treatments, it appeared that new TMAO was generated over the time. Two hypotheses were considered. The first would be that TMAO was generated directly in the hydrolysate solution from its precursors, while the second could involve its precursors, as well as the constituents of CEMs. On the contrary to ED, treatments conducted by deaerator significantly decreased the abundance of the targeted compounds at pH 4 and 7, but had no effect at pH 10. Therefore, the conditions leading to the best removal levels of the targeted volatile compounds were the deaerator treatment performed on the hydrolysate at pH 7, and the alkalization of this latter until pH 10. Despite the fact that the relevance of ED to be used as a deodorization method of HMH was not proven at that stage, it appeared that the establishment of strategies to avoid both fouling and water dissociation phenomena could lead to a better process e fficiency. However, this supposes to deepen the knowledge regarding, especially, the fouling resulting from interactions between IEM and HMH constituents, with particular interest in those involving volatile compounds, by performing Attenuated Total Reflection–Fourier Transform Infrared (ATR-FTIR), as well as identifying the mechanisms leading to TMAO formation during ED. This is currently under investigation. Finally, the use of electromembrane processes, other than conventional electrodialysis, could be another promising solution that is worth further investigation. Therefore, electromembrane processes have a chance to become an e ffective deodorization method in the future.

**Author Contributions:** Conceptualization, S.T. and L.B.; methodology, S.T. and L.B.; software, S.T.; validation, S.T., V.P., M.B., P.D., C.G., Y.B. and L.B.; formal analysis, S.T.; investigation, S.T.; resources, S.T., V.P., M.B., P.D., C.G. and L.B.; data curation, S.T.; writing—original draft preparation, S.T.; writing—review and editing, S.T., V.P., M.B., P.D., C.G., Y.B. and L.B.; visualization, S.T. and L.B.; supervision, L.B.; project administration, L.B.; funding acquisition, L.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) Industrial Research Chair on ElectroMembrane processes aimed at the ecoefficiency improvement of biofood production lines.

**Acknowledgments:** The NSERC Industrial Research Chair on ElectroMembrane processes aimed at the ecoefficiency improvement of biofood production lines is acknowledged. The authors are grateful to Ocean NutraSciences for providing the herring milt hydrolysate powder. The authors thank Diane GAGNON and Jacinthe THIBODEAU, professional researchers at Université Laval, for their involvement in this project.

**Conflicts of Interest:** The authors declare no conflict of interest.
