Conductivity

The conductivity of hydrolysate and KCl recovery solutions was measured using a YSI conductivity meter (Model 3100) equipped with a YSI immersion probe model 3252, (cell constant K = 1 cm<sup>−</sup>1) (Yellow Springs Instrument Co., Yellow Springs, OH, United States). Similarly to the pH measurements,

the conductivity values were recorded every 5 min during the first hour of the ED run, and then every 15 min. The demineralization rate (DR, in %) of the HMH and the mineralization rate (MR, in %) of the corresponding KCl recovery solution were determined according to Equations (1) and (2), respectively, where *k*t refers to the solution conductivity at time t, and *k*0 refers to the initial solution conductivity [30].

$$DR = \left(1 - \frac{k\_t}{k\_0}\right) \times 100\tag{1}$$

$$MR = \left(1 - \frac{k\_0}{k\_t}\right) \times 100\tag{2}$$

Ash Content

Ash content (in %) was determined using a method adapted from the Association of O fficial Analytical Chemists [31]. Briefly, 10 mL samples of HMH and KCl recovery solutions at initial and final times were weighted before being dried overnight at 105 ◦C in an oven (VWR Gravity Convection Oven, Radnor, PA, USA). Dried samples were then reduced to ashes in a furnace at 550 ◦C until they turned white. Samples were weighted after cooling, and the ash content was determined according to Equation (3), where m refers to the measured weight.

$$Ash\,\text{content} = \left(\frac{m\_{\text{crracible}+\text{ashs}} - m\_{\text{crracible}}}{m\_{\text{crracible}+\text{asymle}} - m\_{\text{crracible}}}\right) \times 100\tag{3}$$

The DR (in %) and the MR (in %) were also determined based on the ash content values according to Equations (4) and (5). 

$$DR = \left(1 - \frac{\text{ash content at time } t}{\text{ash content at initial time}}\right) \times 100\tag{4}$$

$$MR = \left(1 - \frac{\text{ash content at initial time}}{\text{ash content at time 1}}\right) \times 100\tag{5}$$

Global System Resistance

The global system resistance (R, in Ω) was calculated based on the Ohm's law (R = U/I). As voltage (U, in V) value was maintained constant during all of the ED runs, only current intensity (I, in A) values were recorded, which were directly obtained from the power supply.

## Membrane Thickness

Membrane thickness was measured before and after each ED run in 0.5 M NaCl solution, according to the same procedure as Lemay et al. (2019) [32], using an electronic digital micrometer equipped with a 10-mm-diameter flat contact point from Marathon watch company LTD (Richmond Hill, ON, Canada). Six measurements, taking place at di fferent locations on the membrane's surface, were used to obtain the average membrane thickness.

## Membrane Electrical Conductivity

Membrane electrical conductivity was measured before and after each ED run in 0.5 M NaCl solution, according to the procedure described by Lemay et al. (2019) [32], using a YSI conductivity meter model 3100 (Yellow Springs Instrument Co., Yellow Springs, OH, USA) equipped with a specially designed clip from the Laboratoire des Matériaux Echangeurs d'Ions (Université Paris XII, Créteil, Val de Marne, France).

## Volatile Compound Content

Volatile compounds determined as being the most potent odor-active compounds of the studied HMH, according to the GC-O procedure described previously, were used as deodorization indicators. Volatile compounds of HMH and KCl recovery solutions at initial and final times were obtained and

analyzed according to the same extraction and GC-MS procedures described previously. The only difference was that extraction was performed on 45-mL liquid samples instead of solid samples. The area under each peak corresponding to the most potent odor-active volatile compounds was considered to assess their abundance in the different samples.

The TMAO, TMA and DMA contents of HMH and KCl recovery solutions were also determined according to the procedure described previously. The quantification was carried out with a calibration curve of known amounts of TMAO, TMA and DMA standards (from 25 ppm to 755 ppm for TMAO; from 2.5 ppm to 10 ppm for TMA and DMA).

## Statistical Analyses

Analyses of variance (ANOVA) were performed, using SAS software version 9.4 for Windows, on data concerning ED parameters, as well as those regarding volatile compounds (SAS Institute Inc., Cary, NC, USA). A Tukey test (α = 0.05 as probability level) was used to compare the different treatments.

#### **3. Results and Discussion**

#### *3.1. Volatile Compound Analysis*

#### 3.1.1. Overall Content in Volatile Compounds

The overall content of the HMH was determined using GC-MS, while HS-GC-NPD was used to verify the presence of TMAO, TMA and DMA. The GC-MS procedure led to the identification of a total of 86 compounds, as listed in Table 2. Although a minimal match degree of 60% was initially selected, all of the volatile compounds were identified with a score ranging from 72.29% to 98.7%. The identified volatile compounds mainly belonged to eight groups, based on their chemical structures. More specifically, volatile compounds included 27 aldehydes, 17 ketones, 12 alcohols, 8 alkenes, 8 nitrogenous compounds, 5 alkanes, 2 furans and 2 esters, while the 5 other compounds were miscellaneous representatives of other chemical groups. Aldehydes and ketones were the most abundant compounds of the hydrolysate. Aldehyde compounds included saturated and unsaturated compounds. The unsaturated aldehydes involved alkenals, alkadienals and aromatic aldehydes. Aldehyde compounds generally originate from lipid oxidation [11,33]. This would be consistent with the composition of HMH, as it results from a rich source of polyunsaturated fatty acids, as shown in Table 1. Aldehydes can also derive from the Strecker degradation of amino acids [34]. Strecker degradation refers to the reaction involving a dicarbonylated compound and an amino acid [35]. Concerning the ketones content, it may also be the result of lipid and/or amino acid degradation [34]. Alkanes, alkenes and alcohols, also identified in large numbers in the HMH, are mainly known to directly derive from lipid oxidation [35–37]. Then, among the nitrogenous compounds content, 8 compounds were identified by GC-MS, while TMAO, TMA and DMA, also listed in Table 2, were detected by HS-GC-NPD. On the one hand, among the nitrogenous compounds identified by GC-MS, 2 pyrazines and 2 thiazoles were detected. More precisely, concerning the pyrazine content, methylpyrazine and 2,5-dimethylpyrazine were identified. Pyrazines are generally formed during Maillard or pyrolysis reactions in heat-processed foods [36,38]. However, despite the fact that high temperatures promote pyrazine formation, mild temperatures are also sufficient to lead to the formation of these compounds [13]. In the present study, since the HMH was obtained through enzymatic hydrolysis under minor heating, followed by a drying step by atomization involving higher temperatures, the two pyrazines identified might have been formed during one of these steps. Pyrazines can be biosynthesized by microorganisms, as well [34,36]. The thiazole content of HMH involved 2 compounds, 2-acetylthiazole and benzothiazole. Thiazole are sulfur-containing compounds that are supposed to originate from Strecker degradation [35]. Surprisingly, while Cha and Cadwaller (1995a) studied the volatile components of various fish and crustacean pastes, including herring paste and shrimp pastes, they identified 2-acetylthiazole only in shrimp paste, while benzothiazole was only identified in anchovy and hair tailed viscera pastes. Nonetheless, they were not detected in herring paste [39]. Regarding the four other nitrogenous compounds identified by GC-MS, three of them, namely 1-methyl-1H-tetrazole, 3-methyl-butanenitrile and 3-methyl oxime butanal, were not evidenced in fishery materials. These compounds may be characteristic of the volatile content of HMH. However, N-nitrosodimethylamine was already known to be generated from TMA and DMA [40]. On the other hand, the HS-GC-NPD procedure allowed the detection of TMAO, TMA and DMA (Table 2). In general, these compounds play a key role in the volatile content of fishery products. TMAO is a well-known osmoregulator, mainly found in marine fish, which aims to counteract protein destabilization [14]. The transformation of TMAO leads to the formation of TMA and DMA. More specifically, TMA can be formed from TMAO through bacterial degradation, while the formation of DMA is mainly attributed to the action of the endogenous TMAO aldolase enzyme [14,41]. Moreover, while TMAO is in presence of reducing agents, these molecules would be generated as well. Finally, compounds belonging to furan and ester groups were also found in HMH. More specifically, two furan compounds, 2-methylfuran and 5-isopropyl-3,3-dimethyl-2-methylene-2,3-dihydrofuran, were identified. Furans are known to originate from lipid oxidation [35]. Another mechanism of the formation of furan compounds involves Maillard and Strecker reactions [38]. Concerning the ester compounds identified, cis-cyclohexane-1,4-dimethanol diacetate and 2,2,4-trimethyl-1,3-pentanediol diisobutyrate, they may have originated from the esterification reaction occurring between alcohols and carboxylic acids generated through microbial or enzymatic lipid degradation [39]. It is of interest to note that Cha and Cadwallader (1995a) identified several ester compounds in herring paste, but none of them corresponded to those identified in this study. In addition, in herring paste, a high content of esters was observed. This should probably be linked to the fermentation process taking place in this type of products [39]. Also, it is interesting to note that, in this study, no carboxylic acid was identified. Whilst this was consistent with the study of Cha and Cadwaller (1995a) dealing with herring paste, this was not in line with the study of Aro et al. (2002), in which few carboxylic acids were identified in herring raw material [39,42]. Carboxylic acids are known to be lipid oxidation products, or to originate from amino acid degradation [42]. The absence of carboxylic acids might be a property of the volatile content of HMH. Globally, despite the few di fferences noted and specified previously, most of the compounds identified in HMH have already been reported in several marine products of di fferent origins: fishing and aquaculture products [33,36,38,42,43], fish oil [21], fish sauces and fish pastes [25,34,39,44–47], and mollusk hydrolysates [9,48].




**Table 2.** *Cont.*

\* Compounds identified by HS-GC-NPD while all the others were identified by GC-MS.

#### 3.1.2. Most Potent Odor-Active Volatile Compounds

To better understand the contribution of each compound to the overall aroma of the studied HMH, the frequency detection method was used. As shown in Table 3, a total of 15 compounds were perceived by at least 5 of the 13 judges, and would be considered to be the most potent odorants of the HMH. 11 of these 15 odorants were aldehyde compounds. In a general way, aldehydes impact significantly

the overall aroma, due to their lower odor thresholds compared to other chemical groups. This means that only a small concentration of these compounds is necessary to make them perceptible [33]. Among the aldehyde compounds, 3-methylbutanal was perceived by all the panelists, while 2-methylbutanal was perceived by 6 of the 13 judges. 3-methylbutanal and 2-methylbutanal are generally generated from the Strecker reaction of leucine and isoleucine, respectively [21,45]. Both of these saturated aldehydes are characterized by burnt, malted, dark chocolate smells. 3-methylbutanal was identified as being the major aldehyde contributing to the odor of white herring [49]. Pentanal, hexanal, heptanal and octanal were sni ffed by at least 6 of the 13 judges. These compounds are derived from lipid oxidation. On the one hand, pentanal and hexanal are respectively responsible for pungen<sup>t</sup> and green smells, while green, rancid and fishy notes are ascribed to heptanal [21,34,36,38]. On the other hand, octanal is described as having citrus and orange peel smells, but also as having fatty and fishy smells [34,36,38]. The smell of octanal may probably depend on its concentration and its possible interactions with other compounds present in the matrix of interest. An interesting point was noticed by Liu et al. (2017) concerning hexanal. They noted that hexanal boosted the global fishy odor when it acted synergistically with other compounds, while its odor could not really be considered fishy when it stood alone [50]. Also, it is noteworthy that hexanal was already found in herring paste and in white herring [39,49]. (Z)-4-Heptenal was perceived by 7 of the 13 judges. This alkenal would be formed during lipid oxidation, and it is characterized by a fishy o ff-flavor [11,38]. In a general way, alkanal and alkenals were known to contribute to the fatty-oily, slightly rancid odor of marine products [36]. Methional was sni ffed by 11 of the 13 judges, while benzaldehyde was sni ffed by all the panel members. On the one hand, methional is an aldehyde sulfur-containing compound characterized by cooked potato notes, whose formation is associated to the Strecker degradation of methionine. Interestingly, it was the only sulfur-containing compound perceived by the panel. Surprisingly, other sulfur-compounds, such as methanethiol, dimethyl disulfide and dimethyl trisulfide, well-known for the considerable role they play in the odor of fishery products, were not identified in this study, while the two other sulfur-compounds, 2-acetylthiazole and benzothiazole, present in HMH were not sni ffed by the panel [49,51]. Methional was already identified as an odorant compound in white herring [49]. On the other hand, benzaldehyde is an aromatic aldehyde characterized by almond, nutty or fruity aromas [38,46]. The two alkadienals, (E,E)-2,4-heptadienal and (E,Z)-2,6-nonadienal, were perceived by, respectively, five and 12 of the 13 judges. They both originate from lipid oxidation [35]. Aidos et al. (2002) found that (E,E)-2,4-heptadienal was representative of the oxidative status of herring oil [52]. (E,E)-2,4-heptadienal is characterized by fatty and fishy odors, while (E,Z)-2,6-nonadienal is reported to have a cucumber-like smell [21,34,38]. Nevertheless, it was reported that (E,Z)-2,6-nonadienal promotes a fishy o ff-flavor while it stands with other compounds [52]. Another group of chemical compounds that was well perceived by the panel was ketone compounds. More precisely, 2,3-pentadione, (Z)-6-octen-2-one and 2-nonanone were sni ffed by at least nine of the 13 judges. The presence of ketones can, similarly to aldehyde compounds, be attributed to lipid and/or amino acid degradation. Iglesias et al. (2009) mentioned that 2,3-pentanedione could be even used as an indicator of lipid oxidation in chilled fish muscle [51]. 2,3-Pentadione is responsible for butter and fruity notes, while 2-nonanone is characterized by grass and green notes [9,34,49]. However, no data was found in the literature concerning the odor of (Z)-6-octen-2-one, and it seemed that this compound was never identified in marine products. Three hypotheses could be made to explain the perception of (Z)-6-octen-2-one in this study. The first one is that, as the hydrolysate studied in that case is an HMH, this compound may be a specific compound of this type of matrix. Nevertheless, as no study has been carried out so far regarding the volatile compounds of fish milt, to the best of our knowledge, it was not possible to confirm this first hypothesis. A second possible explanation could be that a wrong identification occurred. In fact, there is another ketone with a similar structure, 1-octen-3-one, that is highly reported in the literature for its fishy smell, but it was not identified in the case of this study [11]. Finally, the last explanation could be that the judges continued to perceive the odor of benzaldehyde, which was eluted just before (Z)-6-octen-2-one. Indeed, some studies already highlighted that the use of GC-O was not an exact science, and some mistakes could happen due to the fact that the odor detector of this method is the human nose [38,53]. Since (Z)-6-octen-2-one was identified by GC-MS with a match degree of 92%, and since no information was available in the literature regarding the duration of the smell of benzaldehyde, the hypothesis that seemed to be the most plausible was that (Z)-6-octen-2-one was a specific odor-active compound of HMH. It is of interest to note that (E,E)-3,5-octadien-2-one, a ketone compound identified in this study but not perceived by the panel, is well recognized to contribute to fatty-fishy odors in fishery materials [39]. Finally, 1-Methyl-1H-tetrazole was also a compound detected by at least five of the 13 judges. 1-Methyl-1H-tetrazole is a cyclic nitrogen-containing compound. Similarly to (Z)-6-octen-2-one, no information was found in the literature concerning the odor and the presence of this compound in seafood materials. Moreover, the judge responses did not allow us to clearly distinguish if the smell of 1-methyl-1H-tetrazole was negative or positive. Therefore, it could be supposed that 1-methyl-1H-tetrazole was a specific odor-active compound of HMH. Except for (Z)-6-octen-2-one and 1-Methyl-1H-tetrazole, all the compounds perceived by the panel were already reported for their contribution to the odor of various fishery materials [21,25,34,36,38,44,45,47]. Interestingly, no compound belonging to the alcohol group was detected by the judges, particularly the alcohol compound 1-penten-3-ol. In fact, this compound, well-known to contribute to the fishy smell, was identified in HMH. However, it was not perceived by the judges as a contributor to the overall aroma of this hydrolysate. This could be explained by the fact that alcohols generally do not have a huge contribution to the overall aroma of food products due to their high odor detection thresholds [21,33,36,47]. This also explained the fact that no alkane and alkene compounds were perceived by the panel members.



a Number of judges (out of thirteen) who have perceived an odor. b The average intensity given by the judges (out of thirteen) who perceived an odor. The average intensity was rounded to the nearest whole number i.e., an average intensity between 2.0 and 2.5 was rounded to 2.0, and an average intensity between 2.5 and 3.0 was rounded to 3.0. c The judges were asked to qualify the odor as bad or good. Sometimes the total number was inferior to those related to the number of judges having perceived an odor, since some of them simply considered the odor to be OK. d Odorant properties were indicated only when they were available. They were gathered from the following literature and online database [9,11,16,21,33–36,45–47,54,55]; (http://www.odour.org.uk; http://www.flavornet.org).

#### *3.2. Deodorization by ED*

#### 3.2.1. ED Parameters

pH

The evolution of the pH of both HMH and KCl recovery solutions during ED treatments in the four di fferent conditions is shown in Figure 2. Firstly, independently of the conditions of the current and pH tested, the pH of hydrolysate solution during the four different ED treatment conditions was steady at pH 4 or 7. However, concerning the KCl recovery solution, its pH varied differently according to the current conditions applied (*p* < 0.05), the pH of the hydrolysate (*p* < 0.05), and a combination of current conditions and pH (*p* < 0.05). Hence, for the hydrolysate at pH 4 without current, the pH of the KCl recovery decreased from 7.13 ± 0.06 to 5.60 ± 0.60 (*p* < 0.05), while those with current rapidly decreased from 7.13 ± 0.08 to 3.57 ± 0.08 (*p* < 0.05) during the first twenty minutes, and then continued to steadily drop to reach the final value of 2.42 ± 0.03. In the same time, regarding the treatment of hydrolysate at pH 7 without current, the pH of the KCl recovery solution remained steady at 7.06 ± 0.03 during all of the experiment, while with current, the pH of the KCl recovery increased rapidly from 7.09 ± 0.07 to 8.80 ± 0.65 (*p* < 0.05) during the first twenty minutes, and then continued to steadily rise to reach the plateau value of 9.50 ± 0.38.

**Figure 2.** pH 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 steady value of the pH observed for the hydrolysate solution during the four ED treatments was consistent with the fact that the pH of this solution was constantly adjusted to the desired value. The slight decrease in pH noticed for the KCl recovery solution of the hydrolysate at pH 4 without current would sugges<sup>t</sup> the potential diffusion of acid species through the CEMs. Indeed, as HCl was added to acidify the hydrolysate at pH 4, H<sup>+</sup> coming from the dissociation of HCl molecules diffused to the KCl solution. However, the rapid decrease in pH during the first twenty minutes of the experiment of the KCl recovery solution with current at the same pH could indicate that water dissociation took place at an early stage. Indeed, even if HCl molecules were initially added to lower and then to maintain the pH of the hydrolysate at pH 4, the fact that the pH of the corresponding KCl solution rapidly dropped below that of the hydrolysate implied that the electromigration of H<sup>+</sup> acid species coming from the dissociation of HCl molecules would not have been sufficient to obtain such a decrease. For this reason, the occurrence of water dissociation could be considered. Water dissociation leads to the formation of protons and hydroxyl ions at the IEM (AEM or CEM) diluate interfaces, bringing about a pH variation [56,57]. This phenomenon generally occurs when the electrolyte concentration near the diluate side of the membranes becomes close to zero, and is due to the reaching of the limiting current density (LCD). As a consequence, the usual mass transfer of ionic species is hampered, and water dissociation takes place [27]. In that case, the decrease in pH value observed in the KCl recovery solution could sugges<sup>t</sup> that the dissociation of water was more important at the CEM interfaces than at the AEM ones [56]. Regarding the KCl solution for the hydrolysate at pH 7 without current, where no variation of pH was observed, on the contrary to pH 4, no diffusion of H<sup>+</sup> occurred. However, in that case, the quantity of HCl required to maintain the pH value of the hydrolysate at pH 7 was lower than that which was necessary to maintain the pH value at 4. This might explain the difference observed between pH 4 and 7 conditions. Finally, regarding the KCl recovery solution of hydrolysate at pH 7 with current, the drastic pH increase during the first twenty minutes was due to a rapid migration of basic species. Similarly to the ED experiment conducted on hydrolysate at pH 4 with current, water dissociation took place [32,58]. Nonetheless, in the present case, the increase in pH in the KCl recovery solution was related to the migration of OH– basic species. This implied that the dissociation of water was more intensive at the AEM interfaces than at the CEM ones [32,56,58].
