**The Relationship between Glucosinolates and the Sensory Characteristics of Steamed-Pureed Turnip (***Brassica Rapa* **subsp.** *Rapa* **L.)**

**Nurfarhana Diana Mohd Nor 1,2, Stella Lignou 2, Luke Bell 3, Carmel Houston-Price 4, Kate Harvey <sup>4</sup> and Lisa Methven 2,\***


Received: 14 October 2020; Accepted: 19 November 2020; Published: 23 November 2020

**Abstract:** Glucosinolates (GSLs) are phytochemical compounds that can be found in *Brassica* vegetables. Seven separate batches of steamed-pureed turnip were assessed for GSL content using liquid chromatography mass spectrometry (LC-MS) and for sensory attributes by sensory profiling (carried out by a trained sensory panel). Twelve individual GSLs, which included 7 aliphatic, 4 indole and 1 arylaliphatic GSL, were identified across all batches. There were significant differences in individual GSL content between batches, with gluconasturtiin as the most abundant GSL. The total GSL content ranged from 16.07 to 44.74 μmol g−<sup>1</sup> dry weight (DW). Sensory profiling concluded there were positive correlations between GSLs and bitter taste and negative correlations between GSLs (except glucobrassicanapin) and sweet taste. The batches, which had been purchased across different seasons, all led to cooked turnip that contained substantial levels of GSLs which were subsequently all rated as bitter.

**Keywords:** glucosinolates; turnip; *Brassica*; bitter taste; *Brassicaceae*; vegetable

#### **1. Introduction**

*Brassica* vegetables such as turnip, cabbage, broccoli and cauliflower are rich with sulphur-containing glucosinolate compounds (GSLs) [1]. These compounds are water-soluble and have a role in plant defence against pests and diseases [2]. GSLs can be structurally classified into aliphatic, arylaliphatic and indole types [1]. Kim and Park [3] discussed that the degradation products of GSLs possess anticarcinogenic properties, reducing risks of certain cancers in humans. Glucoraphanin, glucobrassicin and gluconasturtiin are among the GSLs that have hydrolysis products shown to have anti-cancer properties, and these are all found in turnip [4].

GSLs are, amongst other compounds, partly responsible for the taste characteristics of *Brassica* vegetables. Individual GSLs such as sinigrin, gluconapin, progoitrin and neoglucobrassicin have been associated with bitter taste [5,6]. Furthermore, Bell et al. [7] reported that GSLs were also correlated with earthy, pepper, mustard flavour and pungency in rocket varieties (*Eruca sativa* Mill.).

GSL contents in *Brassica* vegetables are influenced by many factors, such as environmental conditions and genetic variability between cultivars. The abundance of GSLs in plants is varied, depending on the type of plant species, developmental stage and plant part (root, shoot, seeds and leaves) [8–10]. Concerning cultivars, Kabouw et al. [10] showed that there was a significant difference in GSL content between white cabbage cultivars (*Brassica oleracea* var. *capitata* L.), and Zhu et al. [11] reported significant differences in GSL content between pak choi cultivars. In addition, nutrient supply contributes to the concentration of GSLs in plants. GSL content increases with an adequate supply of sulphur [12], however nitrogen in the absence of sulphur and also selenium supply have been shown to result in a decrease of GSL content [13,14], whereas nitrogen with a sufficient sulphur supply may either increase GSL content or have no effect [14]. Such variations in GSLs can lead to distinctive sensory characteristics of *Brassica* vegetables [15], which are thought to influence their consumption frequency [16].

GSL content in *Brassica* vegetables is also affected when they are handled and prepared before consumption. GSLs undergo hydrolysis to produce breakdown products when the plant cells are wounded [17]. Preparation processes, including cooking and cutting, trigger myrosinase enzymes in plant cells to hydrolyse GSLs and produce isothiocyanates (ITCs) plus other breakdown products; including nitriles, thiocyanates, epithionitriles, oxazolidine-2-thiones and epithioalkanes [15]. A review by Nugrahedi [18] concluded that boiling and blanching significantly reduced GSL content in *Brassica* vegetables due to leaching of compounds. On the other hand, steaming, microwaving and stir-frying may limit GSL loss compared to boiling.

Turnips (*Brassica rapa* subsp. *Rapa* L.) are a traditional vegetable grown in the UK that are no longer frequently consumed by UK consumers in comparison to other *Brassica* vegetables, such as broccoli, cauliflower and cabbage. In 1992, turnip (together with swede) accounted for 5400 hectares of production whereas this had dropped to less than 2700 hectares by 2017. Although the field area for cauliflower fell over the same period, it remained higher than turnip at over 9200 hectares in 2017 [19]. However, turnip could provide a beneficial source of glucosinolates if incorporated more regularly into the diet. As a vegetable that is predominantly consumed cooked, it is the GSL and sensory profile of cooked turnips that are of relevance to the consumer.

Realising that GSL content in commercial turnip may vary between cultivars, growth conditions, seasons and cooking batches, the aim of this study was to evaluate the variability in GSL content and resulting differences in sensory perception, as purchased commercially and as the vegetable would be consumed by consumers. Numerous research papers concerning *Brassica* vegetables focus on the raw vegetable rather than the material as consumed, and where studies focus on cooking, they recommend minimal cooking to preserve GSL content. Minimal processing is not suitable for a hard root vegetable such turnip, and therefore, it is important to establish whether more rigorous cooking and preparation does successfully deliver beneficial GSL to consumers. To achieve our aim, seven batches of steamed-pureed turnip were prepared and subsequently analysed for GSLs (identification and quantification using liquid chromatography mass spectrometry; LC-MS) and sensory profile (trained sensory panel). The hypothesis was that each batch of steamed-pureed turnip would contain substantial amounts of GSLs and have a perceivable bitter taste, regardless of any differences between batches.

#### **2. Materials and Methods**

#### *2.1. Turnip Sample and Preparation*

Seven batches of steamed-pureed turnip were used in this study. Steaming was chosen rather than boiling to reduce loss of GSLs leaching into cooking water. Pureeing was chosen to produce a homogenous sample, and also steam and puree are among many methods used to prepare turnips at home. Turnips (grown in the UK and the Netherlands) were bought from two local stores in Reading (UK), from December 2015 to June 2016, and each batch was cooked on a different day (Table 1).


**Table 1.** Purchase date and the origin of turnips for each batch.

<sup>1</sup> For B3, 75% turnips came from the UK and 25% came from the Netherlands; for B4 24% turnips came from the UK and 76% came from the Netherlands.

The root was used in the preparation of the samples; prior to cooking, turnips were peeled, and stems and tails removed, then washed and sliced to a thickness of approximately 0.5 cm. Between 8.2 and 13.6 kg raw turnips were used to make each batch of cooked turnip. For each cooking cycle, approximately 2.4 kg of sliced turnips were placed into an electric 3-tier steamer (Tefal; 800 g in each tier), with 1 L of water added to the base of the steamer and steamed for 25 min. Sliced turnips from tier 1 were transferred to tier 3 and vice versa (to ensure equal heat circulation), water was added again up to 1 L and steamed for another 25 min to ensure the root was soft enough to be blended. The internal temperature of the steamer was ~64 ◦C. Turnips were then blended using a hand blender (Russell Hobbs) for approximately 5 min until the texture was smooth. All cooked turnips were then placed into plastic containers, labelled, and stored in a freezer at −18 ◦C.

Prior to GSL extraction, samples were frozen (−80 ◦C) and then freeze-dried for 5 days (Stokes freeze dryer, F.J Stokes Corporation, Philadelphia, USA). The dried samples were ground (pestle and mortar) and then sieved (20 mesh) to ensure a fine powder.

#### *2.2. Reagents and Chemicals*

All chemicals used were of LC-MS grade and purchased from Sigma-Aldrich (Poole, UK), unless otherwise stated.

#### *2.3. Glucosinolates Extraction*

The extraction method was adapted from [20]. Three replicates of each batch were prepared as follows: 40 mg of ground steamed-pureed turnip powder was heated in a dry-block at 75 ◦C for 2 min to ensure inactivation of any remaining active myrosinase enzyme. Preheated 70% (*v*/*v*) methanol (1.2 mL; 70 ◦C) was added and the sample placed in a water bath for 20 min at 70 ◦C. Samples were then centrifuged for 10 min (10,000 rpm, 18 ◦C) to collect loose material into a pellet. The supernatant was then filtered through 0.22 μm Acrodisc syringe filters with Supor membrane (hydrophilic polyethersulfone; VWR, Lutterworth, UK) and frozen (−80 ◦C) in Eppendorf tubes until analysis by LC-MS.

#### *2.4. LC-MS Analysis*

LC-MS analysis method was adapted from [21]. Sinigrin hydrate was used as an external reference standard for quantification of GSL compounds. Preparation was as presented by Jin et al. [22]. LC-MS analysis was performed in the negative ion mode on an Agilent 1260 Infinity Series LC system (Stockport, UK) equipped with a binary pump, degasser, autosampler, column heater, diode array detector, coupled to an Agilent 6120 Series single quadrupole mass spectrometer. Separation of compounds was achieved on a Gemini 3 μm C18 110 Å (150 × 4.6 mm) column (with Security Guard column, C18; (4 mm × 3 mm); Phenomenex, Macclesfield, UK). GSLs were separated during a 40 min chromatographic run, with 5 min post-run sequence. Mobile phases consisted of ammonium formate (0.1%; A) and acetonitrile (B) with the following gradient timetable: (i) 0 min (A-B, 95:5, *v*/*v*); (ii) 0–13 min

(A-B, 95:5, *v*/*v*); (iii) 13–18 min (A-B, 40:60, *v*/*v*); (iv) 18–26 min (A-B, 40:60, *v*/*v*), 26–30 min (A-B, 95:5, *v*/*v*); (v) 30–40 min (A-B, 95:5, *v*/*v*). The diode array detector recorded spectra at 229 nm. The flow rate was optimised for the system at 0.4 mLmin<sup>−</sup>1, with a column temperature of 30 ◦C, with 25 μL of sample injected into the system. Quantification was conducted at a wavelength of 229 nm.

MS analysis settings were as follows: API-ES was carried out at atmospheric pressure in negative ion mode (scan range m/z 100–1500 Da). Nebulizer pressure was set at 50 psi, gas-drying temperature at 350 ◦C, and capillary voltage at 2000 V.

Compounds were identified usingMS through both spectra available in the literature [23,24] or from GSL standards in our own laboratory and by comparing relative retention times with those published in the literature [25]. Semi-quantification was carried out using UV absorbance (diode array detector; DAD) peak area data and relating that to the external sinigrin standard (regression: y = 26.7X + 52.6; *r*<sup>2</sup> = 0.998). Relative response factors (RRFs) were used in the calculation of GSL concentrations where available [23]; however, they were assumed to be 1.00 if such data was not available in the literature [25] or from our laboratory standards. All data were analysed using Agilent OpenLAB CDS ChemStation Edition for LC-MS (Agilent, version A.02.10).

#### *2.5. Sensory Analysis*

Sensory analysis was carried out by nine sensory trained panellists, each with a minimum of six months experience, using sensory profiling. The panel developed a consensus vocabulary for the seven batches of steamed-pureed turnip concerning aroma, taste and flavour (Table 2). Spinach, mashed potato, sucrose (granulated sugar) and quinine sulphate solutions were used as references to help the panel to standardise the vocabulary. During duplicate sample evaluations, samples were presented in a balanced sequential order, and each characteristic was scored on a line scales (0–100), using Compusense Cloud Software (Ontario, Canada). Line scales were unstructured except for the sweet and bitter attributes where a structured scale was used. Table 2 shows the levels of reference standards used for these two attributes. The panel tasted and scored the reference standards; their mean values for these standards were used as anchors on the scale. For sweet, the anchor positions for the four standards were 13.8, 29.1, 57.6 and 80.6, respectively. For bitter taste, the anchor positions were 8.1, 23.0, 38.9, 63.2 and 82.6, respectively. Evaluation sessions were conducted in a sensory room within the Sensory Science Centre at the Department of Food and Nutritional Sciences, Reading, UK. Each panellist sat in an individual booth equipped with artificial daylight and with room temperature controlled (approximately 22 ◦C).


**Table 2.** Definition of sensory characteristics associated with 7 batches of steamed-pureed turnip and references used during vocabulary development.

#### *2.6. Statistical Analysis*

The analytical results presented are the mean of three replicates (*n* = 3) for each sample. One-way ANOVA was used for comparison of GSL content between batches of steamed-pureed turnip. A principal component analysis (PCA) was carried out to relate GSLs with sensory characteristics. GSL data were projected onto the PCA with the mean sensory data as supplementary variables; Pearson's correlation was used. These tests were performed by using XLStat (Addinsoft, Paris, France).

Sensory profile data were tested using two-way ANOVA in SENPAQ (Qi Statistics Ltd., Reading, UK) where the main effects (sample and assessor) were tested against the sample by assessor interaction, with sample as fixed effect and assessor as random effect. All significant differences between samples were assessed by using Tukey's HSD post hoc test at a significance level of 5%.

#### **3. Results**

#### *3.1. Identification and Quantification of Glucosinolates*

Twelve individual GSLs were detected across all batches of steamed-pureed turnip (Figure 1), and the concentration of each of GSL varied significantly between batches (Table 3). There were 7 aliphatic GSLs (progoitrin, glucoalyssin, gluconapin, glucobrassicanapin, gluconapoleiferin, glucoerucin, and glucoberteroin), 4 indole GSLs (4-hydroxyglucobrassicin, glucobrassicin, 4-methoxyglucobrassicin, and neoglucobrassicin) and 1 arylaliphatic GSL (gluconasturtiin). Glucoalyssin was only detected in batches B1 and B2, while no glucoerucin was detected in B5. Gluconasturtiin was the most abundant GSL across all batches. Total GSL concentration ranged from 16.07 to 44.74 μmol g−<sup>1</sup> DW.


**Table 3.** Mean

concentration

 of

glucosinolates

 in seven batches of

steamed-pureed

 turnip (B1 to B7). Results are expressed as μmolg-1 DW ±

*Foods* **2020** , *9*, 1719

standard deviation.

#### *3.2. Sensory Characteristics*

Table 4 summarises the mean sensory characteristic scores for the seven batches of steamed-pureed turnip. There was a significant difference in wet aroma where batch B2 had a higher score than B7. No other aroma characteristics were significantly different between batches.

For taste characteristics, there was a significant difference in bitter taste between batches, where batch B2 had the highest intensity for bitter taste, whereas B1 and B4 were significantly less intense. All batches were perceived as bitter with mean ratings varying between 30 and 53 in bitter taste intensity of the 0.0002% and below the 0.0004% quinine standard used. Sweetness did not vary significantly between batches; the range of mean scores were between 26 and 35 on the 100-point scale, being in the region of sweetness of the 1% sucrose standard used.

Significant differences between batches can be found for tannin and apple flavour. B2 was significantly higher than B1, B3, B4 and B5 for tannin flavour. B5 was significantly higher than B2, B6 and B7 in terms of apple flavour. There were no significant differences between batches for other characteristics.


**Table 4.** Mean scores for sensory characteristics for seven batches of steamed-pureed turnip. Different superscript letters indicate significant differences between batches.

#### *3.3. Principal Component Analysis (PCA)*

Principal component analysis (PCA) of the GSL data was carried out to demonstrate the batch separation according to GSLs, and onto this map the sensory data was fitted as supplementary data in order to investigate any correlation of the GSLs with the sensory characteristics (Figure 2). Dimensions 1 and 2 recovered over 78% of the variance in the data. Total GSL and many of the individual GSLs were predominantly located on the right side of PC1, located alongside turnip batches B6 and B7. PC2 was highly correlated with gluoberteroin (*r* = 0.88) and glucoalyssin (*r* = 0.84).

The position for the total GSL content strongly correlated with PC1 (*r* = 0.98) and also to many of the individual GSLs: gluconapin (*r* = 0.99, *p* < 0.001), gluconasturtiin (*r* = 0.98, *p* < 0.001), glucoerucin (*r* = 0.97, *p* < 0.001), 4-hydroxyglucobrassicin (*r* = 0.82, *p* = 0.03), glucobrassicin (*r* = 0.78, *p* = 0.04) and gluconapoleiferin (*r* = 0.77, *p* = 0.04). However, 4 other GSLs strongly correlated with each other and with dimension PC2: glucoberteroin (*r* = 0.88), glucoalyssin (*r* = 0.84), progoitrin (*r* = 0.80) and glucobrassicanapin (*r* = 0.59).

**Figure 2.** PCA biplot of glucosinolate compounds in 7 batches of steamed-pureed turnip (B1 to B7), with mean ratings of sensory attributes fitted onto the plot as supplementary variables. Abbreviations: A, aroma; T, taste; F, flavour.

There was a clear separation of groups of sensory characteristics on the PC biplot. Earthy (aroma and flavour), cooked swede aroma and savoury aroma were positioned to the right of PC1 and negatively correlated with sweet taste. Bitter taste and tannin flavour were positioned in the top right quadrant of the plot and negatively correlated with apple (aroma and flavour).

As expected, many of the GSLs correlated with bitter taste. The total GSL content was positively, but not significantly, correlated with bitter taste (*r* = 0.47, *p* = 0.29). Of the 12 GSLs quantified, one, glucobrassicanapin, had clearly no association with bitter taste (*r* = 0.033, *p* = 0.94) whereas the correlation coefficient between the other GSLs and bitter taste varied between 0.30 and 0.75. The only significant correlation was 4-methoxyglucobrassicin (*r* = 0.82, *p* = 0.02), while glucobrassicin also had a strong positive correlation (*r* = 0.75, *p* = 0.052), despite the levels of these two GSLs not being particularly high in the turnip batches, indeed very low for 4-methoxyglucobrassicin (Table 3). Such correlations cannot prove which of these GSLs have the greatest contribution to bitter taste, but they do support the hypothesis that the GSLs in turnip contribute to bitter taste. Bitter taste will suppress sweet taste, so it was as expected that all GSLs (except glucobrassicanapin) were negatively correlated with sweet taste (*r* = −0.55 to *r* = −0.01).

B1 and B2 were negatively correlated with B6 and B7; B1 and B2 were separated from B3, B4 and B5 along PC2. Moreover, B6 and B7 were separated from the other batches along PC1, and this was driven by the higher level of total GSL and particularly 4-hydroxyglucobrassicin, 4-methoxyglucobrassicin, glucobrassicin, gluconasturtiin, gluconapin and glucoerucin. These 2 batches were indeed the most

bitter tasting, along with B2, which although not as high in total GSL, was highest in glucobrassicanapin. PC2 particularly separated B5 from B2, where B5 was particularly low in all GSLs and higher in apple (aroma and flavour).

#### **4. Discussion**

Twelve individual GSLs were detected across all batches. The total GSL content ranged from 16.07 to 44.74 μmol g−<sup>1</sup> DW with mean value of 27.37 μmol g−<sup>1</sup> DW. The total content is comparable to findings reported by Zhang et al. [26], (16.4 to 31.4 μmol g−<sup>1</sup> DW), but lower than those reported by Lee et al. [4], (117.05 μmol g−<sup>1</sup> DW). Zhang et al. [26] freeze dried the raw turnip roots rather than cooked turnip as in the present study; however, both studies later incubated at 75 ◦C before extraction with methanol. The results remain comparable as the steaming of turnip in the current study would denature the myrosinase enzyme and limit transformation to hydrolysis products. Other reasons might be because of the similarity in environmental factors that both studies have, as the turnips were sown across different seasons, which then would yield similar GSL content. Contradictory to the Lee et al. [4] study, the turnips were sown and grown in a controlled environment (i.e., temperature-controlled) to minimise seasonal differences, hence the large difference in the GSL content.

In the present study, aliphatic GSLs were the most abundant, representing 48.6% of total GSL content, followed by 45.6% of arylaliphatic GSL and 5.8% of indole GSL. These results are in agreement with other studies which confirm that these compounds are common GSLs in turnip varieties [4,26–28]. Gluconasturtiin was the dominant GSL (45.6%), ranging from 8.96 to 19.81 μmol g−<sup>1</sup> DW, with a mean value of 12.48 μmol g−<sup>1</sup> DW. This GSL compound has previously been shown to be the most abundant in turnip greens [28] and turnip roots [26].

There were significant differences in each individual GSL between batches and this is expected as the turnips were bought on different days, across different seasons, and from a variety of suppliers. Although they were all "purple top" turnips, they were potentially of different cultivars. Type of cultivar will affect GSL content; indeed, Kim et al. [29] reported that the GSL content of turnip seeds varied significantly between 12 cultivars. There are many other factors that could also contribute to variability. Kim et al. [30] reported that GSL content in turnip is dependent on harvest times. Subsequent research papers have noted that, in addition to harvest time, growth season could also result in the GSL variation [26,31]. Environmental conditions of different growing sites, such as soil pH, can influence GSL content too [32]. Our PCA plot showed that batches B1 and B2 were similar, as were B4 and B5, and B6 and B7. These similarities can be explained by the month the turnips were purchased. Turnips for batches B1 and B2 were purchased in autumn/winter season, and they were negatively correlated in terms of GSL content and sensory characteristics, with B6 and B7, which were bought in spring/summer season. Although turnips for batch B3 were purchased in a different season from B4 and B5, these three batches were correlated with each other, in terms of GSL content and sensory characteristics. It could be speculated that these three batches may be from the same cultivar of turnip, and the cultivar effect is greater than season effect; however, this cannot be concluded as the turnip cultivar was not controlled for in this study.

In summary, the significant differences in GSL content between cooked turnip batches in this study might be caused by differences in cultivars, seasons or growth conditions. Turnips sold in the UK come from many different countries with different growth conditions. Therefore, variation in GSL content at the point of consumption is expected from turnips purchased in the UK supermarkets at different times of year.

GSLs are among the compounds that are responsible for the sensory characteristics of *Brassica* vegetables. As noted in (Figure 2), most of the GSLs were positively (although not significantly) correlated with bitter taste, the strongest correlations being with 4-methoxyglucobrassicin and glucobrassicin. Although Helland et al. [33] also found 4-methoxyglucobrassicin to be related to the bitterness of swede and turnip, the levels of this compound in the current study were very low. Glucobrassicin was present at higher levels (0.65 to 1.19 μmolg−<sup>1</sup> dry weight) and was clearly correlated with most bitter turnip batches (B2, B6 and B7). Glucobrassicin has previously been reported to cause bitter taste, alongside 4-hydroxyglucobrassicin, progoitrin, gluconapin and neoglucobrassicin, in turnip, swede, rocket, broccoli and cauliflower, [33–35] which is consistent with the current study where all were positively correlated with the bitter taste in turnip (*r* = 0.33 to *r* = 0.75). The GSL in highest abundance in the cooked turnip was gluconasturtiin; this has a positive but relatively weak correlation to bitter taste (*r* = 0.43). Although this finding does not confirm a relationship between gluconasturtiin and bitterness, it was present at high levels in all batches (compared to other GSLs), and all batches were perceived to be bitter. Bladh et al. [36] previously concluded that it was a hydrolysis product of gluconasturtiin, phenethyl isothiocyanate, that had a strong bitter taste. Although GSLs are accepted to impart bitterness, the correlation between GSLs and perception of bitterness does not confirm causality. Bell et al. [37] reviewed the relationship between GSLs and bitterness and noted that there very few studies where GSLs have been isolated and rated by sensory panels; one specific exception being sinigrin where bitter taste thresholds have been reported. Relating bitter taste to specific GSLs in *Brassica* samples is limited by the high correlation between the quantities of individual GSLs. This Bell et al. [37] review also noted that inconsistencies in relating GSLs to bitter taste can also arise from differences in preparation and cooking methods between studies. In addition, hydrolysis product of GSLs are often not quantified, and therefore, their contribution to bitterness is often not accounted for. A further review by Wieczorek et al. [38] concluded that inconsistencies between studies can also result from differences in consumers' sensitivity to GLS-derived bitter taste. Interactions between taste modalities must also be considered; our results showed that all individual GSLs (except glucobrassicanapin) were negatively correlated with sweet taste. This was similarly reported by Francisco et al. [6], suggesting that bitter taste suppressed sweet taste in the perception of turnip.

Bitter taste was positively correlated with tannin flavour, and two individual GSLs were highly correlated with this attribute: 4-methoxyglucobrassicin and glucobrassicin. In our sensory profile data, batches B2, B6 and B7 were rated the highest in tannin flavour and bitter taste. The tannin flavour is likely to originate from phenolic compounds rather than from the GSLs. Such phenolic compounds (flavonoids, quinic acid derivatives, sinapic acids derivatives and tannins) have been found in turnip [39,40] and are also associated with bitter taste [36]. However, phenolic compounds were not measured in the current study, hence the relationship between bitter taste and phenolic compounds could not be determined.

It the present study, it was also observed that gluconapoleiferin, gluconapin, 4-hydroxyglucobrassicin, glucoerucin, glucobrassicin and gluconasturtiin and total GSL were highly correlated with earthy aroma, and gluconapoleiferin, glucobrassicin and 4-methoxyglucobrassicin were highly correlated with earthy flavour. In comparison, Helland et al. [33] observed that gluconapin, glucoerucin and glucobrassicanapin were positively correlated with earthy aroma. However, there are possible compounds other than GSLs that contribute to aroma and flavour of vegetables, such as the breakdown products of GSLs, which were not measured in this study.

#### **5. Conclusions**

The results obtained in this study showed that individual and total GSL varied between different batches of steamed-pureed turnip. The GSL compounds were correlated with aroma, taste and flavour characteristic of turnip. Almost all GSLs positively correlated with bitter taste; however, many GSLs correlated with other GSLs in concentration, which limits interpretation of which have the greatest influence on bitter taste. The strongest correlations with bitter taste were for 4-methoxyglucobrassicin and glucobrassicin; however, these two GSLs were highly correlated and the 4-methoxyglucobrassicin was at particularly low levels, so their individual contribution to bitter taste cannot be confirmed.

Overall, all batches of steamed-pureed turnip demonstrated both bitter and sweet taste, and these two taste characteristics were negatively correlated. It was evident that the bitter taste suppressed the sweet taste of the turnip as the batches containing the least GSL were the sweetest. The impact of this finding is in the conclusion that turnips bought commercially in the UK do provide a substantial amount of GSLs even after rigorous cooking and preparation, and as such cooked turnip could provide the well documented health benefits of GSLs if they were regularly consumed in the diet. However, the cooked product has a consistently bitter taste which may be a barrier to some consumers.

**Author Contributions:** N.D.M.N., C.H.-P., K.H. and L.M.; methodology, S.L., L.B. and L.M.; data analysis, N.D., C.H.-P., K.H. and L.M.; writing—original draft preparation, N.D.M.N. and L.B.; writing—review and editing, S.L., L.B., C.H.-P., K.H. and L.M.; supervision, C.H.-P., K.H. and L.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by The Ministry of Higher Education of Malaysia.

**Acknowledgments:** The authors would like to thank Xirui Zhou and Omobolanle Oloyede of the University of Reading for their guidance and advice.

**Conflicts of Interest:** There are no conflicts of interest to report.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **The Impact of Domestic Cooking Methods on Myrosinase Stability, Glucosinolates and Their Hydrolysis Products in Different Cabbage (***Brassica oleracea***) Accessions**

**Omobolanle O. Oloyede \*, Carol Wagstaff and Lisa Methven**

Department of Food and Nutritional Sciences, Harry Nursten Building, University of Reading, Whiteknights, Reading RG6 6DZ, UK; c.wagstaff@reading.ac.uk (C.W.); l.methven@reading.ac.uk (L.M.) **\*** Correspondence: bola.oloyede@reading.ac.uk; Tel.: +44-(0)118-378-3606

**Abstract:** Glucosinolate hydrolysis products are responsible for the health-promoting properties of *Brassica* vegetables. The impact of domestic cooking on the myrosinase stability, glucosinolates and hydrolysis products in 18 cabbage accession was investigated. Cabbages were steamed, microwaved, and stir-fried before analysis. Cooking significantly affected myrosinase stability and glucosinolate concentrations within and between cabbage morphotypes. Myrosinase was most stable after stir-frying, with up to 65% residual activity. Steaming and microwaving resulted in over 90% loss of myrosinase activity in some accessions. Stir-frying resulted in the greatest decrease in glucosinolate concentration, resulting in up to 70% loss. Steamed cabbages retained the highest glucosinolates after cooking (up to 97%). The profile and abundance of glucosinolate hydrolysis products detected varied across all cooking methods studied. Cooking reduced the amounts of nitriles and epithionitriles formed compared to raw samples. Steaming led to a significant increase in the concentration of beneficial isothiocyanates present in the cabbage and a significantly lower level of nitriles compared to other samples. Microwaving led to a reduction in the concentrations of both nitriles and isothiocyanates when compared to other cooking methods and raw cabbage. The results obtained help provide information on the optimal cooking methods for cabbage, suggesting that steaming may be the best approach to maximising beneficial isothiocyanate production.

**Keywords:** *Brassica oleracea*; cabbage; myrosinase stability: glucosinolates; glucosinolate hydrolysis products; isothiocyanates; epithionitriles; steaming; microwaving; stir-frying

#### **1. Introduction**

Consumption of *Brassica* or cruciferous vegetables such as cabbage (*Brassica oleracea*) is reported to result in chemo-protective effects [1]. This has been attributed to the high amounts of glucosinolates (GSLs) they contain. When plant tissue is damaged as a result of autolysis, plant injury, processing or chewing, GSLs are exposed to, and hydrolysed by, endogenous myrosinase. Upon hydrolysis, glucose and an unstable aglycone (thiohydroxamate-*O*-sulfonate) are produced. The unstable aglycone (thiohydroxamate-*O*-sulfonate) immediately rearranges to form different hydrolysis products such as isothiocyanates (ITCs), thiocyanates, nitriles, epithionitriles (EPTs), and oxazolidine-2-thiones, depending on the conditions of the reaction [2]. Nitriles and EPTs are formed in the presence of epithiospecifier proteins (ESP) instead of the more beneficial ITCs [3]. ITCs and indoles commonly found in cabbages such as sulforaphane (SFP), erucin (ER), allyl ITC (AITC), 2-phenyethyl ITC (PEITC), iberin (IB) and indole-3-carbinol (I3C) are reported to be responsible for some of the health-promoting properties of *Brassicas* [2]. SFP, the most studied of all the ITCs, is reported to possess chemoprotective, antioxidative, antimicrobial and neuroprotective properties [4–9]. AITC has been found to be potent against bladder [10,11], breast [12] and lung [13] cancer cells. I3C is also known to have anti-cancerous activities on reproductive organs, reducing the proliferation of cancer cells in the breast, prostrate,

**Citation:** Oloyede, O.O.; Wagstaff, C.; Methven, L. The Impact of Domestic Cooking Methods on Myrosinase Stability, Glucosinolates and Their Hydrolysis Products in Different Cabbage (*Brassica oleracea*) Accessions. *Foods* **2021**, *10*, 2908. https://doi.org/10.3390/ foods10122908

Academic Editors: Franziska S. Hanschen and Sascha Rohn

Received: 11 October 2021 Accepted: 19 November 2021 Published: 24 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

cervical and colon cell lines and preventing tumour development in rodents [14–16]. GSLs and glucosinolate hydrolysis products (GHPs) are also partly responsible for the bitter taste and pungent flavour and aroma of *Brassica* vegetables, which can reduce consumer acceptability of *Brassicas* [17,18]. Cox et al. [19] reported that *Brassica* acceptance was low in adults due to their sensory characteristics.

Where GSL hydrolysis does not occur in the process of preparing *Brassica* for consumption, it can occur as a result of microbial activities in the gastrointestinal tract of humans. However, despite the ability of microorganisms in the human gut to hydrolyse glucosinolates, it has been reported that the conversion is at least three times less efficient when compared to glucosinolate hydrolysis by myrosinase [20]. In a recent in vivo study conducted by Okunade et al. [21], the addition of exogenous myrosinase from brown mustard powder to cooked and pureed broccoli where myrosinase had been inactivated resulted in a four times increase in sulforaphane bioavailability compared to when the pureed broccoli was consumed alone. It is, therefore, important to ensure that myrosinase enzyme remains active during the consumption of *Brassica* vegetables.

Cabbages, like other *Brassica* vegetables, are mostly subjected to some form of thermal processing or domestic cooking before consumption. Cabbages are commonly boiled, steamed, stir-fried, or microwaved prior to consumption. Thermal cooking processes are considered one of the most important factors affecting the stability of myrosinase enzyme and ESP stability and the profiles, concentrations, and bioavailability of GSLs and their hydrolysis products [22]. This, in turn, can influence the health benefits that can be derived from the consumption of these vegetables making it crucial to determine the effect of cooking processes on myrosinase stability, GSLs and GHP formation within plant tissues.

Cooking cabbage can result in total or partial ESP and myrosinase inactivation, which, in turn, influences the type of GHPs formed. The time and temperature of cooking, vegetable matrix and degree of tissue damage all influence the changes observed during cooking [23]. Several studies have shown that myrosinase is inactivated during the thermal processing or domestic cooking of cabbage, leading to a decreased production of beneficial hydrolytic compounds [24–26]. Most of these studies, however, were based on crude myrosinase extracts or cabbage juice [24,26–28]. Myrosinase enzyme, when present within plant tissue, has been shown to have greater thermal stability than its crude extract, with this stability attributed to the rate at which the core temperature increased [24].

Previous studies on GSL concentrations in cooked cabbage showed conflicting results. Some authors have reported an increase in GSL content after microwaving cabbage [24,28]. Rungapamestry et al. [25], Song and Thornalley [29] and Xu et al. [30] reported minimal losses or no change in GSL concentration after steaming and microwaving cabbage. Xu et al. [30] recorded a 77% loss in GSL concentration after stir-frying. GSL losses during cooking have mostly been linked to leaching into cooking water [24,31]. The variation in myrosinase and GSL stability after processing can be attributed to different cooking conditions and the size of cut cabbage pieces, which, in most cases, do not represent standard domestic ways of cooking cabbage. Some of these studies processed the cabbages under much longer time–temperature combinations compared to what would normally be applicable during the domestic cooking of cabbage [24,26]. Furthermore, most of these studies [24,26,28] have focused on closed heart cabbages (mostly red and white cabbage) with limited data available on open leaf varieties such as black kale. There is insufficient evidence to date linking high myrosinase activity and/or GSL accumulation to high stability after processing/cooking; hence, studies considering both the stability of myrosinase and GSLs within plant tissue for various *B. oleracea* species are needed. Available studies have focused on the effect of cooking on specific GSLs and their GHPs, or just ITCs [25,29,32–37]. Some of these studies have been conducted in model systems [36], which do not consider the various reactions occurring within the plant matrix that can influence the GSL–myrosinase system during domestic cooking processes.

In our earlier paper investigating the effect of accession identity and growing conditions on myrosinase activity, GSL and GHP in 18 cabbage accessions, we discussed in detail

the results of myrosinase activity, as well as the profile and concentration of GSL and GHP, between the raw accessions [38]. In this paper, the focus is on the effects of cooking on these accessions. This study, therefore, examines the effect of steaming, microwaving, and stir-frying on myrosinase stability, GSL and GHP profiles and concentrations in 18 cabbage accessions. Cooking times were chosen to represent standard domestic practices. It was hypothesised that, through controlled domestic cooking processes, myrosinase and GSL stability would be maximised, thereby increasing the production of ITCs, and improving the health benefits associated with cabbage consumption. It was also hypothesised that the genetic background of the cabbage and its morphotype would impact the observed stability of myrosinase and GSLs, and the production of GHPs.

#### **2. Materials and Methods**

#### *2.1. Plant Material*

The seeds of the cabbage accessions used for this study were sourced from the University of Warwick Crop Centre Genetic Resources Unit (Wellesbourne, UK), sown in controlled environments, transplanted to pots (2.5 L) and left to grow in the glasshouse for a short time before transplanting to 7 metre beds on the field. A detailed cultivation protocol can be found in Oloyede et al. [38]. Eighteen cabbage accessions from six different cabbage morphotypes (black kale (BK), wild (WD), tronchuda (TC), savoy (SC), red (RC), and white (WC)) were selected based on their head formation (closed heart or open leaf), geographical location and whether they were of hybrid descent (Supplementary Table S1). Cabbages were grown between 7th March and 25th November in the plant growth facilities, Whiteknights campus of the University of Reading, UK. White cabbage accession WC3, did not germinate under the growing condition.

Upon reaching commercial maturity (based on visual inspection), cabbages were harvested, immediately placed on ice in freezer bags and stored in a cold room (4 ◦C) for 24 h prior to processing. Average weight of each cabbage head per plant was 700 g and 300 g for closed heart and open leaf, respectively. For detailed climatic data and cross section of cultivated cabbages, refer to Supplementary Table S1 in this paper and Figures S1 and S2 in our earlier paper, Oloyede et al. [38].

#### *2.2. Reagents and Chemicals*

Sinigrin standard was obtained from Santa Cruz Biotechnology (Heidelberg, Germany) and D-glucose determination kit from R-Biopharm Rhone (Heidelberg, Germany). All other chemicals used were purchased from Sigma–Aldrich (Dorset, UK).

#### *2.3. Cabbage Thermal Processing*

In order to achieve a representative sample and remove senescent leaves, the outer leaves and central core of 4–5 cabbage heads (biological replicates) were removed and discarded. Cabbages were chopped into pieces of approximately 1 cm in width using a kitchen knife (representing how cabbages would normally be prepared by consumers), mixed together, washed under running tap water, and drained of excess water using a salad spinner (OXO Good Grips Clear Manual Salad Spinner, Chambersburg, PA, USA). Cabbages were subjected to steaming, microwave, or stir-fry cooking. Unprocessed (raw) cabbage samples were used as controls. Cooking methods were chosen to represent common ways of cooking cabbage.

Time and temperature combinations used for each method were based on a preliminary consumer study with 60 participants to determine consumer acceptability of the samples as steamed, microwaved, and stir-fried cabbage (Supplementary Table S2). These conditions were deemed acceptable with a mean score of between 2.7 and 3.8 on a 5-point degree of cooking scale, where "3" represents 'just about right' for extent of cooking (scale from not cooked enough "1", too much too overcooked "5").

#### 2.3.1. Steaming

The method of Rungapamestry et al. [25] was adopted with slight modifications. A total of 120 g cabbage was placed in the topmost layer of a 3-tier, 18 cm stainless steel steamer (Kitchen craft, Birmingham, UK) containing already-boiling water (in the lowest layer) and allowed to steam for 2 min. Core temperature of cabbage during steaming ranged between 75 and 80 ◦C and was measured using a temperature probe.

#### 2.3.2. Microwaving

The method of Rungapamestry et al. [25] was adopted. A total of 120 g of cabbage was put into 1-pint Pyrex glass jug, 16 mL water was added, and the jug was covered with a PVC cooking film pierced with 9 holes. Cabbages were microwaved for 3 min. Microwaving was carried out using a 900 W microwave at 60% power output (SANYO microwave oven EM-S355AW/AS, Osaka, Japan). A microwave thermometer was used to measure the core temperature of the cabbage during processing. Core temperature during processing ranged between 88 and 95 ◦C.

#### 2.3.3. Stir-Frying

Cabbage samples were stir-fried as described by Rungapamestry et al. [39] with modifications. A total of 120 g cabbage was stir-fried in a frying pan for 90 s in 5 mL of preheated olive oil (100 ◦C) (Asda, Reading, UK) with continuous stirring using a wooden spatula. Core temperature of cabbage during stir-frying ranged between 65 and 70 ◦C and was measured using a temperature probe.

Samples were put into sterilin tubes immediately after cooking, placed on ice and transferred to a −80 ◦C freezer. Frozen samples were freeze-dried (Stokes freeze drier, Philadelphia, PA, USA), ground using a tissue grinder (Thomas Wiley® Mini-Mill, Thomas Scientific, Swedesboro, NJ, USA) and stored at −20 ◦C until further analysis.

#### *2.4. Myrosinase Enzyme Extraction and Assay and Protein Content Analysis*

Myrosinase enzyme was extracted using the method described by Ghawi et al. [32] with slight modifications, as described in our previous paper, Oloyede et al. [38]. Myrosinase enzyme was extracted from a 0.1 g sample (at 5 ◦C) using polyvinylpolypyrrolidone (PVPP) and Tris- HCL buffer. Using a D-glucose determination kit, myrosinase activity was determined following the coupled enzyme method as described by Gatfield and Sand [40] and Wilkinson et al. [41] with some modifications, as outlined in our preceding paper [38]. Myrosinase activity of the samples was calculated using a calibration curve prepared from myrosinase enzyme. One unit of myrosinase activity was defined as the amount of enzyme that produces 1 μmol of glucose per minute from sinigrin substrate at pH 7.5.

Protein content from the enzyme extract was determined using the Bradford method [42]. The method is based on protein complex formation with Brilliant Blue G dye with absorbance read 595 nm in a spectrophotometer (Perkin Elmer, Shelton, CT, USA). A standard curve was constructed using Bovine serum albumin (BSA) and used to calculate protein concentration in the enzyme extracts from which myrosinase-enzyme-specific activity (U/mg protein) was determined.

#### *2.5. Glucosinolate and Glucosinolate Hydrolysis Products Analysis*

GSLs and GHPs were extracted following the methods described by Bell et al. [43] and Bell et al. [44], respectively, with modifications as described in our earlier paper Oloyede et al. [38]. GSLs were extracted with 70% methanol, analysed by LC-MS/MS (Agilent, Bracknell, UK), and quantified using sinigrin hydrate standard. Six concentrations of sinigrin hydrate (14–438 μg/mL) were prepared with 70% methanol and used to prepare an external calibration curve (*r*<sup>2</sup> = 0.996). Compounds were identified using their mass parent ion, characteristic ion fragments and through comparing with ion data from literature (Table 1).


**Table 1.** Intact glucosinolates identified in cabbage accessions analysed by LC-MS.

Key: GSL = glucosinolate.

GHPs were extracted using dichloromethane and analysed by GC-MS (Agilent, Manchester, UK). Compounds were identified using the literature on ion data (Table 2; see Supplementary Figure S1 for GC-MS chromatograms) and quantified based on an external standard calibration curve. Five concentrations (0.15–0.5 mg/mL) of sulforaphane standard (Sigma Aldrich, UK) were prepared in DCM (*r*<sup>2</sup> = 0.99). Data analysis was performed using ChemStation for GC-MS (Agilent, Manchester, UK).

#### *2.6. Statistical Analysis*

Results are the average of three biological or processing replicates (each replicate consisting of leaves from 4–5 cabbage heads) and two technical replicates (*n* = 6). All statistical analyses were performed in XLSTAT (version 2019.4.2, Addinsoft, Paris, France). Data obtained were analysed using 2-way ANOVA with both cabbage accession (or morphotype) and processing conditions (raw, steamed, microwaved, and cooked) fitted as treatment effects. Tukey's HSD multiple pairwise comparison test was used to determine significant differences (*p* < 0.05) between samples. Principal component analysis (PCA) and multifactor analysis (MFA) were used to visualise the data in a minimum number of dimensions (two or three). MFA makes it possible to simultaneously analyse several tables of variables, showing relationships and correlations between the observations and variables, which were analysed in such a way that tables that include more variables do not outweigh other tables in the analysis.


**2.** Glucosinolate hydrolysis products identified in cabbage accessions analysed by

 GC-MS.

**Table**  reference spectrum in the NIST/EPA/NIH mass spectra database and those in the literature.

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

#### *3.1. Effect of Domestic Cooking on Residual Myrosinase Enzyme Activity (Relative Activity) across Cabbage Morphotypes and Accessions*

The myrosinase stability of cabbage accessions after domestic cooking was studied and the results are presented in Figure 1, with relative activity results presented in Supplementary Table S3. Relative activity is defined as the ratio of myrosinase activity of processed (cooked) cabbage to unprocessed (raw) cabbage (A/A0). Domestic cooking affected the stability of myrosinase enzyme. Myrosinase stability differed significantly (*p* < 0.05) between domestic cooking processes, where there was no difference between steaming and microwaving (*p* = 0.912), but these processes both differed significantly from stir-frying (*p* < 0.0001). Myrosinase was most stable after stir-frying, retaining up to 65% (i.e., A/A0 = 0.65, Supplementary Table S3) of its activity in some studied accessions. Steaming and microwaving resulted in losses of myrosinase activity of up to 98% and 99%, respectively, with the highest stability of 15% and 13%, respectively. Rungapamestry et al. [39], in their study of broccoli florets, reported that stir-frying retained the highest myrosinase activity (17%) compared to boiling (14%).

**Figure 1.** Comparison of the myrosinase activity of raw versus cooked cabbage morphotypes and accessions (U/g DW). Values are means of three biological (raw samples) or processing (cooked samples) replicates (each replicate comprising 4–5 cabbage heads) and two technical replicates (*n* = 6). Error bars represent standard deviation from mean values. Letters "A–G": bars not sharing a common uppercase letter differ significantly (*p* < 0.0001) between accessions and treatments within a cabbage morphotype. Letters "a–r": bars not sharing a common lowercase letter differ significantly (*p* < 0.0001) between cabbage morphotypes, accessions, and treatments. Key: BK-CNDTP: cavolo nero di toscana o senza palla; BK-CPNT: cavolo palmizio; BK-CNDTT: cavolo nero di toscana o senza testa; WD-8707: wild cabbage 8707; WD-GRU: wild cabbage 7338; WD-8714: wild cabbage 8714; TC-PCM: penca mistura; TC-CPDP: penca povoa; TC-T: tronchuda; SC-HSC: hybrid savoy wirosa; SC-PW: pointed winter; SC-SDG: dark green; RC-RL: red langendijker; RC-RM: rocco marner (hybrid); RC-RD: red Danish; WC-FEM: early market; WC-CRB: couve repolho.

The effect of domestic cooking processes on myrosinase stability varied among cabbage morphotypes and accessions and will be discussed in more detail later. The stability of myrosinase in different *Brassica* vegetables under different processing conditions has been discussed by several authors [24–27,39,55]. Differences in myrosinase stability as a result of cooking can be attributed to the maximum core temperature of the vegetable during heating. In our study, stir-frying had the lowest core temperature (65–70 ◦C) compared to steaming (75–80 ◦C) and microwaving (88–95 ◦C). It has previously been reported that, to prevent myrosinase inactivation, the maximum core temperature that cabbage should reach is between 50 and 60 ◦C, which can be achieved by steaming for 7 min or microwaving (700 W) for 120 s [25]. However, in the study stated, the cabbage samples were cut into wedges, which is not representative of how cabbages are generally prepared before cooking, so the

cooking times to reach the same core temperature in cabbage that is more finely chopped would be shorter.

Verkerk and Dekker [24] reported that inactivation of myrosinase enzyme during microwave cooking is affected by the time–energy output combination. Their study showed that a considerable amount of myrosinase activity was retained when red cabbage was microwaved at 180 W for 24 min and 540 W for 8 min, while microwaving for 4.8 min at 900 W resulted in total loss of myrosinase activity even though the total energy output of all three processes was the same (259.2 KJ). The authors explained the resulting effect as a function of the time it takes for the cabbage to reach its maximum core temperature, with a higher energy output and shorter time reaching a high (100 ◦C) core temperature faster and maintaining that core temperature for the remaining cooking time, while the lower energy output with a longer cooking time resulted in a maximum core temperature of 90 ◦C at a much slower rate.

In the current study, a physical examination of the cooked cabbage samples showed that the stir-fried cabbage looked firmer than the steamed and microwaved cabbage, which can be a helpful way to assess the severity of the thermal process. The intense heat during stir-frying can lead to drying out of the surface area, thereby resulting in a firmer texture, which reduces the rate of heat penetration as a result of less damage to the cell wall [39,56].

#### *3.2. Comparison of the Myrosinase Activity of Raw versus Cooked Cabbage Morphotypes and Accessions*

Figure 1 shows the myrosinase activity and subsequent thermal stability of the 17 studied cabbage accessions. Significant differences (*p* < 0.0001) were observed in the myrosinase activity and stability of cabbages as a result of cabbage morphotype, accession, cooking method, and the interactions between these parameters.

There was no relationship between myrosinase activity in raw cabbage and myrosinase stability post-cooking; indeed, some accessions which had high activity in raw cabbage had the lowest stability. Raw savoy cabbage accessions (SC-HSC, SC-PW, and SC-SDG) had the highest myrosinase activity in all studied accessions (116.3, 142.6 and 154.8 U/g DW, respectively) while raw black kale accessions (BK-CNDTP, BK-CPNT, and BK-CNDTT) had the lowest myrosinase activity (31.5, 36.3 and 44.4 U/g DW, respectively). However, black kale, Tronchuda and red cabbage accessions had the highest enzyme stability, while savoy and white cabbage accessions, which had the highest myrosinase activity, were the least stable after domestic processing (Figure 1). As discussed earlier, steaming and microwaving resulted in lower myrosinase stability overall, with up to 99% inactivation occurring in some cases. However, a critical look at the stability of myrosinase in steamed and microwaved cabbages (Figure 1) shows that some accessions had relatively higher myrosinase stability compared to others. Red cabbage accessions RC-RM and RC-RL were the most stable, retaining up to 15% after steaming (RC-RM) and 13% after microwaving (RC-RL). This result is in agreement with the results of Yen and Wei [27], who stated that red cabbage myrosinase was more stable than white cabbage myrosinase after thermal processing.

A possible reason for the difference in myrosinase stability across accessions might be due to differences in the myrosinase isoenzymes found in each accession, with the red cabbage accessions having more thermally stable myrosinase isoenzyme. Red cabbage contains anthocyanins, which, in addition to being bioactive compounds with health promoting properties, are also pigments that offer effective protection to plants under stress [57,58]. Therefore, red cabbage is more adapted to stressful conditions, and it stands to reason that the myrosinase isoenzyme in red cabbage may also be adapted to operating under heat stress.

Different types of myrosinase isoenzymes have been identified and they vary between *Brassica* vegetables. They can differ to some extent in characteristics and activity, with distribution in plants appearing to be both organ- and species-specific [27,59]. Rask et al. [60] reported that some of the myrosinase isoforms form complexes by interacting with myrosinase-binding proteins, which may enhance their stability during processing.

The myrosinase activity values obtained in this study were higher in most cases than those reported by other authors [25,61], except in the case of white cabbage accessions, where values were similar to those obtained by Penas et al. [62]. This might be because, in most previous studies, cabbages were obtained from supermarkets, while in this study and the study conducted by Penas et al. [62], the cabbages were grown for the experiment and transferred into cold conditions immediately after harvest. These minimal transfer and storage times reduce the postharvest effects experienced by the supermarket samples.

There was no relationship found between accession origin, physical characteristics (open-leaf or heart-forming) and whether cabbages were hybrid on the myrosinase activity and stability of the accessions studied.

#### *3.3. Protein Content and Specific Activity of Raw and Cooked Cabbages*

The protein content and specific activity of cabbage myrosinase before and after cooking is presented in Table 3. There were significant (*p* < 0.05) differences in the protein content and specific activity of all accessions for both raw and cooked samples. Protein content decreased with cooking, with the rate of reduction corresponding to the severity of the cooking process. Stir-fried samples had significantly higher protein contents than steamed and microwave samples. Black kale and red accessions, with the highest protein content, also retained the most protein after stir-frying (up to 87% in BK-CNDTP) but the lowest after steaming and microwaving (up to 67% in steamed BK-CPNT). This can be attributed to the denaturation of protein into free amino acids during cooking.

Cooking led to a significant reduction (*p* < 0.05) in the specific activity of cabbage samples. The specific activity of the cabbages followed a similar trend to myrosinase activity and protein content, where specific myrosinase activity decreased with the severity of the cooking method. The result shows a correlation between myrosinase activity and specific activity, implying that denaturation of the protein is equal to denaturation of the enzyme. Stir-fried cabbages had the most stable specific activity and differed significantly (*p* < 0.001) from steamed and microwaved samples between accessions for all studied morphotypes, with the exception of savoy morphotype, where no significant difference was observed in specific activity between accessions for the studied cooking methods (Table 3). It is worth mentioning that the results obtained for savoy accessions were mostly due to the significantly higher specific activity of the raw samples, instead of a comparable stability across the cooking methods. Similar to our earlier discussion on myrosinase stability, samples with the highest specific activity were not always the most stable after cooking. For example, Savoy cabbage accessions (SC-HSC, SC-PW, and SC-SDG), which had the highest specific activity (4.7, 6.4 and 5.8 U/mg soluble protein, respectively) had the lowest stability after cooking, with an up 97% loss in specific activity after steaming and microwave cooking observed in SC-PW accession. On the other hand, black kale accessions, with some of the lowest specific activity in raw samples, retained the most specific activity after cooking, with up to 80% specific activity observed in BK-CNDTT accession. As expected, the result obtained is in agreement with myrosinase activity results discussed earlier, where black kale accessions with the least myrosinase activity were the most stable after domestic cooking. The differences observed in specific activity can be attributed to variations in myrosinase isoenzyme stability for the different morphotypes and accessions, as discussed in Sections 3.1 and 3.2.


*Foods* **2021** , *10*, 2908

146

pointed winter; SC-SDG: dark green; RC-RL: red langendijker;

 RC-RM: rocco marner (Hybrid); RC-RD: red Danish; WC-FEM: early market; WC-CRB: couve repolho.

#### *3.4. Effect of Domestic Cooking on GSL Profile and Concentration of Cabbage Accessions*

GSL profile and concentrations for all samples before and after cooking are presented in Figure 2, with significant differences within and between cabbage morphotypes presented in Supplementary Table S4.

GSL profile and concentrations varied across accessions within and between cabbage morphotypes. From five to nine individual GSLs were identified within all cabbages studied; seven aliphatic GSLs, namely sinigrin (SIN), gluconapin (GPN) and epi/progoitrin (PROG), Glucoibeverin (GIBVN), glucoerucin (GER), glucoiberin (GIBN) and glucoraphanin (GRPN) and two indole GSLs: glucobrassicin (GBSN) and 4-hydroxyglucobrassicin (4- HOH) (Table 1). Black kale accessions had the lowest number of identified GSLs (five), while nine GSLs were identified in red and white cabbages. GBSN and 4-HOH were the only GSLs identified in all studied accessions. Total GSLs differed significantly between accessions (*p* < 0.0001), post different cooking methods (*p* < 0.0001) and in the interaction between these two factors (*p* < 0.0001). Aliphatic GSLs were the most abundant GSLs in all accessions, making up about 95% of total GSLs.

Cooking significantly reduced GSL concentration in all cabbage samples. GSL stability varied across the accessions and cooking methods studied. GIBN was the least stable GSL resulting in an average loss of 59% across all accessions. However, GIBN loss varied largely between accessions, with tronchuda accession TC-PCM recording a loss of up to 83%, while the loss in savoy SC-HSC was as low as 14%. The obtained results agree with those reported by Oerlemans et al. [28] and Dekker et al. [63], who report variations in GSL stability between GSLs and variations in the stability of the same GSL across different *Brassica* vegetables. In a previous study, concentrations of GIBN (aliphatic GSL) and GBSN (indole GSL) in white cabbage were found to significantly decrease during cooking due to their high potential to leach into the cooking water [64,65].

Total GSLs in steamed cabbage ranged between 16.5 μmol/g DW (BK-CNDTT) and 148.8 μmol/g DW (WD-8714). There was a significant difference in GSL concentrations of steamed cabbages across accessions and between accessions of the same cabbage morphotype. The observed differences were mostly due to the initial GSL concentration of the raw samples rather than the steaming process. In relation to the residual GSL content of cabbage samples after steaming, steamed WC-FEM had the most stable total GSL, retaining up to 97% GSL concentration, while the biggest loss of total GSL was in steamed SC-SDG, where up to 56% loss was recorded. The differences observed in GSL stability may be due to variations in leaf thickness between the accessions, which would impact the rate of heat transfer within the leaves. Thicker leaves would lead to a slower heat transfer rate within the leaves, resulting in reduced GSL degradation and better stability compared to thinner leaves. In some accessions, steaming did not affect the concentrations of some individual GSLs, e.g., SIN and PROG in WD-8714 and WC-FEM, respectively. There was a significant (*p* < 0.0001) reduction in individual and total GSL content for all samples after cooking, except for GPN, which did not differ significantly from raw to cooked samples within each accession for most of the studied accessions. The stability of individual GSLs varied greatly between accessions, within and between cabbage morphotypes. For example, after steaming, in BK samples, the loss of GRPN did not differ between the three accessions (8–10%), while in TC samples, it led to a between 44% (TC-CPDP) and <1% (TC-PCM) loss of GRPN content.

Previous studies reported no loss [25,29,35,66,67] or minimal losses [30,31,35,68] of GSL in broccoli, turnip, and cabbages after domestic processing. Xu et al. [30] reported a loss of about 15% in steamed red cabbage; however, the large sample size (3 cm cubes) may have caused lower losses in comparison to the present study, as this would have had an impact on the core temperature of the samples during cooking. Similar to the current study, Vallejo et al. [31] reported losses in some individual GSL (GRPN) and no loss in others (GIBN) after steaming for 3.5 min. In kale samples steamed for 15 min, SIN degraded more rapidly than GIBN, with more than 80% and 40% loss recorded for SIN and GIBN, respectively [33]. The minimal GSL losses reported in steamed samples were due to the low levels of leaching into cooking water compared to that normally reported under boiling conditions [23].

In microwaved samples, total GSL varied between 11.2 μmol/g DW (BK-CNDTT) and 131.6 μmol/g DW (WD-8714). Microwaving significantly affected the amount of GSLs

in cabbage samples, with reductions up to 76% of GRPN in TC1 and residual total GSL varying between 50% and 93%. Microwaving led to significantly lower GSL concentrations when compared to raw cabbages. As in steamed samples, the effect of microwaving differed between accessions and individual GSLs. Some GSLs were more stable than others in certain accessions within and between cabbage morphotypes. As discussed in Section 3.1, high core temperatures (85–95 ◦C) of microwaved samples led to myrosinase enzyme inactivation, which could have prevented GSL hydrolysis during the microwave process and can account for the high retention of GSL concentrations in some microwaved cabbages.

There are several conflicting reports on the effect of microwaving on GSL content in *Brassica* vegetables. In a recent study, no significant difference was observed in broccoli and red cabbage samples microwaved under different time and power combinations while retaining the same final energy (1080 kJ) [35]. Song and Thornalley [29] and Xu et al. [30] also reported no significant difference in GSL concentration after microwaving green and red cabbage samples for three and five minutes, respectively. The authors stated that the stability of GSL might be due to myrosinase inactivation, and that the absence of water during microwaving prevented GSL leaching into cooking water. The large size of the shredded cabbage pieces in the two studies may also have reduced the loss of GSLs. A study on broccoli resulted in a 74% decrease in total GSL content after microwaving and was attributed to leaching in water and more intense microwave conditions (150 g broccoli to 150 g water and microwaving for 5 min at 1000 W power) [31]. However, a contrary result was observed by Verkerk and Dekker [24] and Oerlemans et al. [28], who reported an increase of up to 78% and 35%, respectively in GSL concentrations after microwaving red cabbage, though the increase was not significant in the Oerlemans et al. [28] study due to the large sample variability. The authors attributed the increase to the enhanced extractability of GSL after microwaving, which could be more of an analytical artefact than an actual increase in GSL concentration.

Stir-frying led to a significant decrease in the total and individual GSL content of cabbages. Total GSL ranged between 10.3 μmol/g DW (BK-CNDTT) and 111.7 μmol/g DW (WD-8714). There was a significant difference in GSL concentrations between accessions, within and between cabbage morphotypes. Residual total GSL varied between 28% (SC-SDG) and 81% (SC-HSC). The highest loss of aliphatic individual GSL concentration was recorded in stir-fried TC-PCM accession, where there was a decrease of between 79 to 83%. Indole GSLs, GBSN and 4-HOH were the most stable GSLs in stir-fried cabbages. The relative thermostability of individual GSLs (if under the same myrosinase level and stability) can be influenced by their chemical structure and has been reported to vary with heating temperature [28,69]. Among all the studied cooking methods, stir-frying resulted in significantly greater losses of GSL than steaming or microwaving, which agrees with previous reports. A study on the effect of different types of cooking oil on GSLs in stir-fried broccoli resulted in up to 49% losses, irrespective of the cooking oil used [70]. Xu et al. [30] also reported a 77% loss in GSL concentration after stir-frying red cabbage while there was no significant loss in GSL content when green cabbage was stir-fried for 5 min [29]. The difference in leaf structure may have influenced GSL stability in green cabbage. Green cabbage can have thicker leaves with a more uneven surface texture, which may create a microclimates around the leaf during the cooking process [29]. It was observed in the present study that green cabbage tended to have thicker leaves than other morphotypes based on visual observation. It is hypothesized that losses due to stir-frying can be attributed to substantial moisture evaporation. During stir-frying, cabbage loses moisture and GSLs are leached into the moisture, which evaporates during the cooking process. A study conducted by Adler-Nissen [56] showed that when carrot cubes were stir-fried, despite temperatures only reaching 70 ◦C, a high evaporation loss was observed. Another possible reason for the lower GSL amounts in stir-fried cabbages can be attributed to GSL hydrolysis by myrosinase and ESP during the cooking process. As mentioned previously, (see Sections 3.1 and 3.2), the low core temperatures (65–70 ◦C) of stir-fried

cabbages resulted in higher myrosinase stability of the samples when compared to steamed and microwaved cabbages. Contrary to the findings of this study, however, some recent studies showed that stir-frying preserved total and most individual GSL contents of various *Brassica* vegetables, with the authors attributing this to quick myrosinase inactivation and no leaching losses into cooking water [33,71,72].

The relative stability of individual and total aliphatic GSLs to indoles varied between accessions and cooking methods, however, generally, indole GSLs were more stable than aliphatics. Previous studies have largely reported indole GSLs to be more heat-labile under domestic cooking conditions than aliphatic GSLs [31,33,64,73–75]. The higher stability of indoles in the present study may be due to the type of *Brassica* and cooking methods that were investigated. The nature of the vegetable matrix and heat treatment have been shown to have an effect on the stability of individual indole GSLs. For example, in the absence of leaching and enzymatic degradation during thermal treatment, 4HOH has been reported to be more thermolabile than GBSN and neoglucobrassicin (NEO) [28,75,76]. NEO has also been found to be more thermostable than GIBVN after roasting broccoli sprouts for 15 min [75]. It is important to mention that the GSL concentrations obtained in the current study are much higher than those reported for mature cabbage in the literature [62,65,77–80]. However, up to 80 μmol/g DW [81] and 111 μmol/g DW [82] have been reported in mature rocket leaves, which is similar to the concentrations reported for most accessions in this study. We hypothesis that the reason for the high GSL accumulation in the accessions studied may be because they are gene bank accessions, which means they have not been thoroughly characterized for their phytochemical content. It is common for wild Brassicas to have much higher concentrations than cultivated varieties, and many of the high GSL genotypes that have been bred came from crosses with gene bank material, an example of which is the "*Beneforte*" broccoli [83].

In summary, WD-8707 accession had the most stable individual and total GSL, while GSLs of SC-SDG were the most thermolabile across all cooking methods, despite having one of the highest GSL concentrations in the raw sample. Different accessions of the same cabbage morphotype can vary in their GSL stability during cooking, resulting in large differences in GSL loss between species. The rate and extent of loss is dependent on the morphotype of the cabbage, sample cut size, cooking time and temperature, amount of moisture, and initial concentration of GSL [65,66]. The variation in residual GSL in the cabbages will have an impact on the amounts of GHPs produced.

#### *3.5. Effect of Domestic Cooking on GHP Profile and Concentration in Cabbage Accessions*

The profile and concentration of GHPs resulting from cooking cabbage are presented in Figure 3, with significant differences between and within the cabbage morphotypes presented in Supplementary Table S5. Twenty-three (23) different GHPs were detected as a result of GSL hydrolysis during cooking. Accession, cooking method and interaction between the two significantly (*p* < 0.0001) influenced GHP profile and concentrations. Total GHPs across all accessions and cooking methods ranged between 0.33 μmol/g DW (microwaved TC-T) and 18.66 μmol/g DW (raw WD-8707). In raw samples, GSL hydrolysis led to the production of majorly nitriles and epithionitriles. Matusheski and Jeffery [84] and Mithen et al. [85], in their studies of fresh and freeze-dried raw broccoli, found that GRPN hydrolysis primarily led to the formation of SFN rather than its ITC, SFP. In most studied accessions, raw and stir-fried cabbages had the highest total GHPs in all samples, apart from red and white cabbage accessions, where the highest total GHPs was recorded in steamed cabbages. Black kale samples had the lowest identified GHPs, which could be related to the lower number of individual GSL present in the accession.

**Figure 3.** Glucosinolate hydrolysis products (GHPs) (μmol/g DW) in different accessions of (**a**) black kale; (**b**) wild cabbage (**c**) tronchuda cabbage; (**d**) Savoy cabbage; (**e**) Red cabbage and (**f**) white cabbage before and after domestic cooking. Results are expressed as sulforaphane equivalents. Error bars represent standard deviation from mean values. Letters above bars refer to differences in total GHP concentration. Letters 'A–F': bars not sharing a common uppercase letter differ significantly (*p* < 0.05) between accessions and cooking methods within a cabbage morphotype (i.e., within each separate graph). For significant differences between cabbage morphotypes, accessions, and cooking methods (i.e., across the separate cabbage morphotype graphs), see Supplementary Table S5. Compounds with similar colour shades are GHPs (with nitriles in pattern fill) of corresponding GSL, presented in Figure 2. Abbreviations: R = raw, ST = steamed; MW = microwaved and SF = stir-fried; BK-CPNT: cavolo palmizio; BK-CNDTT: cavolo nero di toscana o senza testa; WD-8707: wild cabbage 8707; WD-GRU: wild cabbage 7338; WD-8714: wild cabbage 8714; TC-PCM: penca mistura; TC-CPDP: penca povoa; TC-T: tronchuda; SC-HSC: hybrid savoy wirosa; SC-PW: pointed winter; SC-SDG: dark green; RC-RL: red langendijker; RC-RM: rocco marner (Hybrid); RC-RD: red Danish; WC-FEM: early market; WC-CRB: couve repolho. For abbreviations of compounds, see Table 2 (GHPs).

However, some GHPs were identified where intact GSL was not detected, and this occurred across all tested accessions. In black kale and savoy cabbage accessions, 3-butenylITC (3BITC) was detected in cooked samples though intact GPN was not present. A similar trend was noticed by Bell et al. [44], who found 3BITC in rocket samples, in the absence of GPN. The presence of 3BITC might be the result of SFP degradation. A study conducted on broccoli showed that standard SFP solution was degraded to 3BITC under thermal conditions [86]. PEITC and benzenepropanenitrile (BPN), hydrolysis products of gluconasturtiin, were detected in low amounts across all accessions, although intact gluconasturtiin was not detected in samples. The small amounts detected suggest that the GSL was present in low amounts in the sample and may have been hydrolysed during sample preparation, or the amounts present were below the limit of detection for the LC-MS.

Other studies have reported the presence of GHPs where their precursor, GSL, was not detected in different *Brassicas* [82,87,88]. Bell et al. [82] suggested that this discrepancies may be due to the degradation of other hydrolysis products during the analytical process, the inaccurate identification of GSLs and GHPs, very low amounts of GSLs being present in the plant, which were below the limit of detection (LOD) of the analytical method, or a yet-to-be identified mechanism by which GHPs are modified after hydrolysis.

Cooking significantly reduced the nitriles and EPTs that were formed and increased the number of ITCs compared to raw cabbage. Goitrin (GN), iberin (IB) and SFP were the major GHPs in cooked cabbages. Of all the studied cooking methods, microwaved samples had the lowest levels of GHPs; few or no nitriles and EPTs were detected, while very low amounts of ITCs were formed. However, in most cases, more ITCs were formed in microwaved samples when compared to raw samples. The highest concentrations of ITCs were formed in steamed samples across all studied accessions, with up to 23-fold increases in ITCs, compared to those observed in raw samples (SFP in steamed RC-RD), where few or no nitriles were present. In most samples, total and individual GHPs did not significantly differ between stir-fried and raw samples, although higher numbers of ITCs were formed in stir-fried samples. The pattern of GHP formation did not differ across accessions.

Differences in the severity of cooking methods, which may have influenced residual myrosinase activity in relation to ESP activity, can account for the difference in the types and concentration of GHPs present. ESP promotes the formation of nitriles and EPTs from GSL hydrolysis instead of ITCs from myrosinase [3]. The stir-fry cooking temperature was the least severe, leading to the formation of EPTs, nitriles and ITCs, as ESP and myrosinase would have still been active in the samples. The lower amounts of GSL detected in stir-fried cabbages did not seem to affect total GHPs but might have been partly responsible for the higher amounts of nitriles formed, as GSL was hydrolysed by the ESP present in the samples during the stir-frying process. Microwave cooking was the most severe cooking method employed, which was responsible for the negligible amounts of nitriles and low amounts of ITCs. The high core temperatures during microwaving (85–95 ◦C) would have led to the complete denaturation of ESP and almost total myrosinase inactivation, as highlighted earlier (Section 3.1). However, the steaming temperature would have been enough to denature ESP whilst still retaining substantial myrosinase activity (see Sections 3.1 and 3.2). The nitriles detected in both microwaved and steamed samples may have been formed with the residual ESP present during the cooking process, while the ITCs present in microwaved samples could be the result of residual myrosinase activity. In cooked broccoli, ESP was found to be denatured at temperatures above 50 ◦C, with a corresponding reduction in SFN production [3]. Rungapamestry et al. [25], in their study of SIN hydrolysis products in cooked cabbage, found that microwaving for 120 secs resulted in a reduction in nitriles, allyl cyanide and CEP (about 87%), with an increase in AITC formation (about 88%). The authors found that steaming cabbages for seven minutes resulted in an increase in AITC of up to 578%. The authors also found that AITC was formed in cabbages with no residual myrosinase activity and attributed this to formation during the hydrolysis and cooking process, which may have been bound to the cell membranes but released during processing. In a recent study, the steaming and stir-frying of broccolini and kale for

15 min resulted in significantly lower amounts of SFP and IB, respectively, when compared to the uncooked sample. The lower concentrations reported were probably due to the longer cooking time used in the study, which would have resulted in total myrosinase inactivation, preventing the conversion of GSLs to GHPs, although myrosinase activity was not measured in the study [33]. In another study, steaming broccoli resulted in increased SFP content after 5 min with a decrease observed beyond that, while microwaving broccoli and red cabbage for a minute led to a 5-fold increase in SFP with a decline reported beyond this time [35]. This suggests that there is an optimum cooking time to achieve maximum ITC formation, beyond which beneficial ITCs are lost. The low conversion of GSLs to their hydrolysis products in some of the studied samples is underscored. The results show that GHP recovery reduced with increases in the severity of the cooking procedure, with much lower concentrations (about 1% in microwaved TC-T) observed in microwaved samples, which was the most severe treatment employed. The low recovery of hydrolysis products observed in the microwaved samples is not necessarily a surprise, given that most of the ESP and myrosinase enzyme in the samples were already inactivated as a result of the cooking temperature (88–95 ◦C), as discussed in Sections 3.1 and 3.2 (see Figure 1 and Supplementary Table S3), which suggests that most of the GSL present in the sample remained unhydrolysed. However, the GSL concentration of the hydrolysed sample was not measured in the current study. In raw accessions with a low GHP content, we hypothesise that this may be due to the environmental responses of the plant, which, unfortunately, are not very well understood in the context of myrosinase activity and GSL hydrolysis. The low conversion of GSLs to GHPs has also been reported in some other *Brassica* species [44,81].

The results obtained in this study are similar to those observed by several authors during the thermal processing of *Brassica* vegetables [25,29,32,35,55,89]. This study adds to the findings of previous researchers; however, the study is particularly conclusive as it demonstrates similar findings across cabbage morphotypes and accessions. To improve the health benefits derived from cabbage consumption, this paper concludes that steaming is the optimum preparation method, due to the resulting increase in the ITCs formed.

#### *3.6. Principal Component Analysis (PCA) and Multifactor Analysis (MFA) of GSLs and GHPs in Raw and Cooked Cabbage*

To differentiate samples based on their GSLs and GHPs content, PCA analysis was conducted, as shown in Figure 4. Figure 4a shows the biplot for GSL distribution in samples, where dimensions 1 and 2 account for 56.4% of the observed variation. The plot shows TC2 (TC-CPDP), and wild cabbage accessions were characterized by high PROG and GPN contents, while black kale and most red cabbages, except for RC1 (RC-RL), had a higher tendency to accumulate GRPN and GER. Savoy cabbages, RC1 (RC-RL), TC1 (TC-PCM) and TC3 (TC-T) correlated positively with one another and were characterized by the amounts of SIN, GIBVN and GIBN they accumulated. Samples were separated based on cabbage morphotype and accession rather than cooking methods, suggesting that cabbage accession had a higher influence on GSL concentration than the tested cooking methods. However, the PCA biplot for GHPs (Figure 4b) shows differentiations in samples based on cooking. F1 and F2 explain only 39.6% of the variations; however, other dimensions did not provide any new information. Steamed and stir-fried cabbages correlated positively with ITCs, while nitriles mostly correlated with raw cabbages. There was no correlation observed in microwaved samples with GHPs, due to the low amounts of nitriles and ITCs present in the samples. Samples were separated based on their GHP profile and concentrations.

(**b**)

**Figure 4.** (**a**) PCA plot for tested samples and their relative distributions in relation to GSL concentrations. (**b**) PCA plot for tested samples and their relative distributions in relation to GHP concentrations. Abbreviations: R = raw, ST = steamed; MW = microwaved; SF = stir-fried; BK1: cavolo nero di toscana o senza palla (BK-CNDTP); BK2: cavolo palmizio (BK-CPNT); BK3: cavolo nero di toscana o senza testa (BK-CNDTT); WD1: wild cabbage 8707 (WD-8707); WD2: wild cabbage 7338 (WD-GRU); WD3: wild cabbage 8714 (WD-8714); TC1: penca mistura (TC-PCM); TC2: penca povoa (TC-CPDP); TC3 tronchuda (TC-T); SC1: hybrid savoy wirosa (SC-HSC); SC2: pointed winter (SC-PW); SC3: dark green (SC-SDG); RC1: red langendijker (RC-RL); RC2: rocco marner (Hybrid) (RC-RM); RC3: red Danish (RC-RD); WC1: early market (WC-FEM); WC2: couve repolho (WC-CRB). Red coloured compounds = GSLs; Green coloured compounds = GHPs; Blue dots = Samples. For full names of compounds, see Tables 1 and 2.

To better understand the results, MFA was performed on the accessions in relation to their GSL and GHP concentrations, as shown in Figure 5. Dimensions 1 and 2 (F1 and F2) represent only 34.3% of the variations, but other dimensions did not provide additional information. The observed results are similar to those observed in the biplot of GSL. Samples were separated in the same pattern as GSLs, based on cabbage morphotype and accession rather than cooking method. Individual GSLs correlated with their corresponding GHPs. The results observed from the MFA analysis confirm the results obtained from the PCA analysis, confirming the robustness of the findings. The results show that cooking has a greater effect on GHPs than GSLs but, when combined, samples were differentiated on their GSL content and the type of GHP present.

**Figure 5.** MFA map of glucosinolates and glucosinolate hydrolysis products (**a**) distribution of variables and (**b**) sample distribution. Abbreviations: R = raw, ST = steamed; MW = microwaved; SF = stir-fried; BK1: cavolo nero di toscana o senza palla (BK-CNDTP); BK2: cavolo palmizio (BK-CPNT); BK3: cavolo nero di toscana o senza testa (BK-CNDTT); WD1: wild cabbage 8707 (WD-8707); WD2: wild cabbage 7338 (WD-GRU); WD3: wild cabbage 8714 (WD-8714); TC1: penca mistura (TC-PCM); TC2: penca povoa (TC-CPDP); TC3 tronchuda (TC-T); SC1: hybrid savoy wirosa (SC-HSC); SC2: pointed winter (SC-PW); SC3: dark green (SC-SDG); RC1: red langendijker (RC-RL); RC2: rocco marner (Hybrid) (RC-RM); RC3: red Danish (RC-RD); WC1: early market (WC-FEM); WC2: couve repolho (WC-CRB). Colour codes: Pink = Black kale; Brown = Wild cabbage; Dijon yellow = Tronchuda cabbage; Black = Savoy cabbage; Purple = Red cabbage; Blue = White cabbage. Compounds with different shades of the same colour in Figure 5a refer to the GSL and corresponding GHPs. For compound codes on plot, refer to Tables 1 and 2.

This study is not without its limitations. Although precautions were followed to ensure accuracy during sample preparation, the size of cut cabbages, stirring during stirfrying and general reproducibility of the cooking processes across all samples may have slightly differed. In addition, some very volatile GHPs may have been lost during the cooking and analytical processes due to the long extraction method used, which may have affected the results. However, given that standard measures were employed to limit variations due to the above, it is unlikely that possible variations within the samples could have had a significant influence on the results, as only small variations were observed in the biological and technical replicates.

#### **4. Conclusions**

The results of this study confirm that domestic cooking has an effect on myrosinase stability, GSL concentration and GHP profiles, and concentration. Domestic cooking resulted in a significant loss of myrosinase activity, with stir-frying having the highest residual activity compared to the other two cooking methods that were investigated. Microwave cooking was the most severe heat treatment, resulting in the highest loss of myrosinase activity, reaching up to 99% in some cases. The study showed that mild cooking prevents the complete inactivation of myrosinase enzyme. Myrosinase enzyme stability differed significantly between cabbage accessions and morphotypes. Black kale myrosinase was the most stable after stir-frying, while red cabbage accessions were most stable after steaming and microwaving. No correlation was found between myrosinase activity and stability, as the accessions with the highest myrosinase activity did not have the most stable myrosinase after domestic processing.

Cooking led to a reduction in GSL concentrations compared to raw cabbage, with stir-frying leading to the greatest loss compared to the other two cooking methods and mild steaming enabling the greatest retention of GSL compounds. Considering that cabbages are usually consumed cooked, it important for breeders to work alongside nutritionists to select accessions with more thermally stable GSL and myrosinase for breeding, to ensure that the health benefits from cabbage consumption are not lost. The study found a relationship between cabbage core temperature during cooking, myrosinase stability and final GHPs profile. GHPs of raw cabbages were mainly nitriles and EPTs, probably due to the presence of active ESP in the samples. Cooking led to a reduction in the number of nitriles and EPTs formed, with levels differing between cooking methods. Optimal cooking conditions led to the degradation of ESP but retention of active myrosinase. Microwaving resulted in significantly lower amounts of nitriles, EPT and ITC formed, while steaming cabbages led to the production of significantly higher amounts of ITCs. However, the study showed that low residual myrosinase activity can still result in ITC formation.

The study concludes that consumption of raw or severely heat-treated cabbage can reduce possible health benefits, while mild cooking of cabbages, such as mild steaming, maximises beneficial isothiocyanate formation. This was especially true for IB and SFP in the studied cabbages and could provide information to guide consumers on how to improve the possible health benefits derived from cabbage consumption.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/foods10122908/s1, Figure S1: Examples of GC-MS chromatograms for raw and cooked samples for each morphotype of cabbage studied (a) black kale; (b) wild cabbage; (c) tronchuda cabbage; (d) savoy cabbage; (e) red cabbage and (f) white cabbage. Table S1: Origin, botanical and common names of planted cabbage accessions. Table S2. Consumption intent and cooking time scores from preliminary consumer study. Table S3. Relative activity (A/A0 ± SD) of myrosinase enzyme after domestic cooking of cabbage. Table S4: Glucosinolate concentration of raw and cooked cabbage (μmol/g DW). Table S5: Glucosinolate hydrolysis products concentration in raw and cooked cabbage (μmol/g DW sulforaphane equivalent).

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

**Funding:** This research was funded by the Commonwealth Scholarship commission (CSC), UK as part of the doctoral research of the first author (O.O.O.), scholar ID: NGCS-2013-363.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** We would like to specially thank Warwick Genetic Bank for providing the cabbage seeds used for the study, Chelsea Snell for her advice on the cabbage-growing conditions and Valerie A. Jasper, Tobias James Lane and Matthew J. Richardson of the plant-growth unit, the University of Reading for their help with growing the cabbages. A big thank you to Denise Macdonald, Bindukala Radha, Chris Bussey, Josh Stapleford and Charwin Piyapinyo for their help with sample preparation. Our thanks go to Sameer Khalil Ghawi and Olukayode Okunade for support and guidance with myrosinase extraction and assay; Luke Bell, Nicholas Michael, Stella Lignou, Hanis Nadia Yahya and Rashed Alarfaj for support and guidance with glucosinolate extraction and LC-MS analysis and finally, Salah Abukhabta and Stephen Elmore for help and guidance with glucosinolate hydrolysis product extraction and the GC-MS analysis, respectively.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Article* **Determination of Isothiocyanate-Protein Conjugates in a Vegetable-Enriched Bread**

**Mareike Krell 1, Lina Cvancar 1, Michael Poloczek 1, Franziska S. Hanschen <sup>2</sup> and Sascha Rohn 1,3,4,\***


**Abstract:** Vegetables of the plant order Brassicales are believed to have health-promoting properties, as they provide high contents of glucosinolates (GLS) and deriving from these, enzymatically and heat-induced breakdown products, such as isothiocyanates (ITC). Besides their positive physiological effects, ITC are electrophilic and can undergo reactions with food components such as proteins. Following the trend of improving traditional food products with GLS-rich ingredients, interactions of ITC with proteins can diminish the properties of both components—protein's value and functionality as well as ITC's bioactivity. In vegetable-enriched bread, where cresses (*Lepidium sativum* L. or *Tropaeolum majus* L.) are added to the initial dough, together with benzyl cyanide, benzyl isothiocyanate (BITC) is formed during the baking process. The aim of the present study was to investigate the possible migration behavior of the GLS breakdown products and the formation of ITC-wheat protein conjugates. After the baking process, the breads' proteins were enzymatically hydrolyzed, and the ITC-amino acid conjugates analyzed using a LC-ESI-MS/MS methodology. In all samples, BITC-protein conjugates were detected as thiourea derivatives, while formation of dithiocarbamates could not be detected. The study showed that GLS and their breakdown products such as ITC migrate into the surrounding food matrix and undergo reactions with proteins, which could in turn lead to modified protein properties and reduce the bioavailability of ITC and lysine.

**Keywords:** glucosinolates; benzyl isothiocyanate; protein conjugates; functional foods; nasturtium; garden cress; thiourea

#### **1. Introduction**

Plants possess a wide variation of bioactive compounds derived from plant secondary metabolism. When consumed these can provide pharmacological properties and beneficial effects on human health [1,2]. Consequently, the *World Health Organisation* (WHO) recommends an intake of at least 400 g of fruits and vegetables per day, which is still not being reached sufficiently [3,4]. In order to further increase the supply of vegetables with their health-promoting secondary plant metabolites (SPM), different pasta and bread recipes have already been adapted, applying raw materials rich in SPM [5–7]. Bread seems to be a particularly suitable product for a fortification with SPM, as it is consumed quite frequently in Western countries, regardless of age and gender [8].

When vegetables of the plant order Brassicales are used as ingredients in foods, glucosinolates (GLS) are in many cases the dominating SPM [9]. Besides antibacterial and anti-inflammatory effects, some studies even suggest a reduced risk of suffering from certain types of cancer, when vegetables containing GLS are regularly consumed [10,11].

**Citation:** Krell, M.; Cvancar, L.; Poloczek, M.; Hanschen, F.S.; Rohn, S. Determination of Isothiocyanate-Protein Conjugates in a Vegetable-Enriched Bread. *Foods* **2021**, *10*, 1300. https://doi.org/ 10.3390/foods10061300

Academic Editor: Montserrat Dueñas Paton

Received: 17 May 2021 Accepted: 3 June 2021 Published: 5 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

However, the health-beneficial effects seem to be not directly associated with the native GLS, but rather linked to their breakdown products, especially isothiocyanates (ITC) [12]. In addition to ITC, nitriles, thiocyanates, and epithionitriles can be formed after the damage of the plant tissue, as occurring during chewing or cutting. During food processing nitriles are formed via thermally-induced degradation [13,14]. The tissue damage causes the enzyme myrosinase to come into contact with the GLS and hydrolyze them into unstable aglycone intermediates. These are then further converted by a Lossen-like rearrangement or specific proteins into the different breakdown products [15,16]. The profile of the breakdown products formed is dependent on the pH value, the presence of modifier proteins, and the initial composition of the GLS [17].

Due to their electrophilic nature, ITC can react with nucleophiles such as the amino or thiol groups of amino acids, peptides, or proteins. The reaction of ITC with the side chains leads to the formation of thiourea and dithiocarbamate derivatives, respectively, presenting different stabilities and subsequent reaction pathways [18–20]. Conjugates between ITC and different food proteins such as myoglobin, egg white proteins, and whey proteins have already been studied. It was found that the ITC-protein conjugation does not only lead to altered physicochemical properties, but also to a structural change of the proteins [18,21,22]. In food, ITC-protein conjugates were already identified when garden cress (*Lepidium sativum*) was mixed with curd (as in the typical German spread 'Kräuterquark') with the result that 28% of the original GLS were conjugated to lysine and cysteine [23].

In a recent study, bread was enriched with pak choi (*Brassica rapa* subsp. *chinensis*) and kale (*Brassica oleracea* var. *sabellica*) for increasing the intake of GLS and their breakdown products [24]. During breadmaking, the ingredients undergo a heating process, reaching temperatures around 100 ◦C. Consequently, mainly nitriles rather than ITC were detected in those experiments, as higher temperatures conditions preferably lead to the formation of nitriles, which have a higher thermostability [24,25]. The desired health-promoting ITC were only found in bread enriched with pak choi (*Brassica rapa* subsp. *chinensis*) microgreens in a concentration below 0.005 μmol/g fresh weight. Further reasons for the low ITC concentration could have been based on the chemical structure of the GLS, the water content of the bread, and the plant tissue being only marginally destroyed [24]. The behavior of ITC in a processed food with regard to interactions with nucleophilic amino acid side chains was not yet investigated in detail. However, reactions between ITC from the added vegetables and proteins in a more highly processed product such as bread, still need characterization, as fate of the GLS and ITC during breadmaking is not yet completely understood. Such interactions can affect the physiological as well as the technofunctional properties of both components.

It is hypothesized that the thermally-induced breakdown products of GLS can migrate into the bread crumb and form ITC-wheat protein conjugates to a certain extent, away from their place of formation. Consequently, the aim of this study was to investigate the formation of ITC, their migration behavior in a model bread matrix, and the possible conjugation with wheat proteins. Based on a recent experiment, wheat dough was used, mixed with fresh garden cress (*Lepidium sativum* L.) or freeze-dried nasturtium leaf-powder (*Tropaeolum majus* L.) [7]. Microgreens and homogenized material from the different cress genera were used for studying the factors that can influence the formation of ITC conjugates such as different BITC sources and the form of plant-material addition.

#### **2. Materials and Methods**

#### *2.1. Chemicals and Materials*

Boric acid solution (4%), hydrochloric acid (0.1 M), hydrochloric acid (32%), formic acid (FA; 98%), sodium hydroxide (100%), trichloroacetic acid (≥99%), Kjeldahl tablets, sulfuric acid (96–98%, p.a.), methanol (LC-MS grade), water (LC-MS grade), methylene blue, methyl red and methylene chloride (GC Ultra Grade, solvent) were purchased from Carl Roth GmbH & Co. KG (Karlsruhe, Germany). Pronase E (from *Streptomyces griseus*), pepsin (from porcine gastric mucosa), pancreatin (from porcine pancreas), β-glucuronidase (from *Helix pomatia*), DEAE Sephadex® A-25, 4-hydroxybenzyl glucosinolate potassium salt, 2-propenyl glucosinolate potassium salt, benzyl cyanide (≥98%), benzonitrile (≥99.9%), and benzyl isothiocyanate (≥98.5%) were obtained from Merck KGaA (Darmstadt, Germany). 1*H*-Imidazole was purchased from AppliChem GmbH (Darmstadt, Germany), boric acid (powder, pure) from Honeywell International Inc. (Seelze, Germany) and 4- (methylsulfinyl)butyl glucosinolate from PhytoLab GmbH & Co. KG (Vestenbergsgreuth, Germany). Sodium sulfate anhydrous (≥99%) was obtained from VWR International GmbH (Darmstadt, Germany). C18ec solid phase extraction cartridges (3 mL, 200 mg) were obtained from Machery-Nagel GmbH & Co. KG (Düren, Germany). The BITC-lysine and BITC-cysteine standards used were synthesized by Kühn et al. [23].

#### *2.2. Plant Material*

Garden cress seeds (*Lepidium sativum* L.) were purchased from Dürr Samen (Reutlingen, Germany). The seeds were cultivated for seven days at room temperature on cotton wool with natural light on a windowsill and sprayed with water every day. The sprouts were harvested and divided into two parts. One part was used for the analysis of the GLS content of the non-processed material, the other part was directly applied as ingredient in the bread dough.

Freeze-dried nasturtium (*Tropaeolum majus* L.) leaves were obtained from the Leibniz Institute of Vegetable and Ornamental Crops (IGZ) e.V. (Großbeeren, Germany). The cultivation was done according to standard procedures for *Brassica* vegetables, including fertilization, irrigation, and plant protection. After 10 weeks, leaves were harvested and immediately frozen at −50 ◦C, freeze-dried, and milled to a fine powder [26]. A part of the homogenized material was analysed for the GLS content and the other part applied into the bread dough.

#### *2.3. Bread Baking and Material for Analysis*

Three bread types with different cress applications and percentage addition relative to the dough weight were prepared: freeze-dried, powdered nasturtium leaves (*Tropaeolum majus* L., 4%) mixed homogenously into the dough (application No. I); parts of fresh garden cress microgreens (*Lepidium sativum* L., 1.5%) mixed into the dough (application No. II), and freeze-dried nasturtium leaves (*Tropaeolum majus* L., 1.5%) placed centrally in the dough (application No. III).

The different bread types were prepared in triplicate. The basis was a wheat dough with 65% wheat flour (Type 450, EDEKA Zentrale AG & Co. KG, Hamburg, Germany), 33% water, 1% salt, 0.5% sugar, and 1% yeast (EDEKA Zentrale AG & Co. KG). The dough was prepared by adding the water to the dry ingredients and kneaded by hand for 5 min. The dough was divided, and the plain dough was used as control. The second part of the dough was mixed with freeze-dried powdered nasturtium leaves (application No. I) or respectively with parts of fresh garden cress microgreens (application No. II). The resulting levels of benzyl glucosinolate (BG) were around 50 μmol/100 g bread dough for application No. I and 20 μmol/100 g bread dough for application No. II. For the rise and the baking, a commercially available breadmaking device (UNOLD Backmeister® extra Modell 65811, UNOLD AG, Hockenheim, Germany) was used. The program sequence is listed in Table 1 and the temperature profile is given in Figure 1. After cooling down the whole bread, the crust was removed, and the crumb was frozen at −80 ◦C. For the analysis of GLS breakdown products the material was grounded using liquid nitrogen to prevent thawing and stored frozen at −80 ◦C. To investigate ITC-protein conjugates lyophilized powder was used, also stored at −80 ◦C.


**Table 1.** Baking program of the UNOLD Backmeister® extra Modell 65811.

**Figure 1.** Temperature profile of the rise and baking in the bread maker (UNOLD Backmeister® extra Modell 65811, UNOLD AG, Hockenheim, Germany).

For the bread with the freeze-dried nasturtium bolus in the center of the dough (application No. III), the plain dough was treated according to program section number 8. Then, the additional plant material was placed into the dough and the treatment was continued with program sections number 9 and number 10. The BG amount was around 17 μmol/100 g bread dough. After cooling down, the crust and the plant material were removed. The crumb was separated into three further fractions as shown in Figure 2: Crumb material in direct and close contact to the freeze-dried material was the core fraction (fr-1). With increased distance of the original bolus, an intermediate fraction (fr-2), and an outer fraction (fr-3) were prepared.

**Figure 2.** Scheme of the vegetable-enriched model bread application No III (bolus application). The crumb is separated into three fractions: core fraction (fr-1), intermediate fraction (fr-2), and outer fraction (fr-3).

The separated garden cress material from application No. II and nasturtium from application No. III were also used for a GLS analysis to verify the loss of GLS and/or breakdown product formation resulting from the baking process or further transformations of the plant material. This further enables one to evaluate indirectly the leaching and migration into the dough. Additionally, an experiment was performed to investigate a possible heat-induced breakdown of GLS during the baking process of application No. III. For this, homogenized nasturtium was boiled in water for 10 min to deactivate the myrosinase and freeze-dried afterwards. The samples were then parted: one part was used for GLS analysis. The other part was baked into the bread, freeze-dried and also analyzed for the GLS content.

#### *2.4. Protein Content*

For the determination of the protein content the control breads were used and a traditional Kjeldahl protocol was performed. Freeze-dried bread (2.5 g) was mixed with 20 mL sulfuric acid (96–98%, p.a.) and a catalyst tablet. The reaction was heated for 4 to 5.5 h, till the solution was clear. For the distillation, a distillation unit (Büchi Labortechnik GmbH, Essen, Germany) was used. To the cooled down reaction mixture, 75 mL of water and 75 mL of aqueous sodium hydroxide (33%, *w*/*v*) were added, and steam distilled for 5 min at 75% steam pressure into 50 mL aqueous boric acid (4%). Finally, the distillates were titrated with 0.1 M hydrochloric acid until a color change from green to grey/colorless of the Tashiro indicator was observed. The calculation of the protein content was based on the nitrogen content and a protein conversion factor of 5.7 [27].

#### *2.5. Enzymatic Hydrolysis of the Samples*

For the extraction of ITC-amino acid conjugates, a combination of different enzymatic digestion protocols were used. First, the protocol of Pasini et al. was applied, in which an in vitro digestion was imitated to hydrolyze the proteins into peptides [28]. Therefore, 50 mg of freeze-dried bread was mixed with 4 mL of a pepsin solution (0.05 mg/mL in 0.2 M hydrochloric acid, enzyme/protein ratio: 1:30) and incubated for 30 min at 37 ◦C and 400 rpm in a thermoshaker (Hettich Benelux B.V., Geldermalsen, the Netherlands). Afterwards, 1.15 mL of a pancreatin solution (0.25 mg/mL in 1 M boric acid and 0.5 M sodium hydroxide solution, pH 6.8, enzyme/protein ratio 1:21) were added to the samples with a resulting pH value of 7.6. The reaction mixture was incubated again (37 ◦C, 400 rpm) for 150 min and the enzymatic digestion was finally stopped by the addition of 1 mL 20% (*w*/*v*) trichloroacetic acid (TCA). The reaction mixture was allowed to precipitate for 60 min, centrifuged for 10 min at 2576× *g*, and the supernatant was freeze-dried. The latter was diluted in 2 mL PBS-buffer and for further protein digestion, the protocol described by Kühn et al. was used with some modifications [23]. A pronase E solution (60 μL, 7 U/mg, 10 mg/mL in PBS buffer, enzyme/protein ratio 1:100) were added and the mixture was incubated for 18 h at 37 ◦C in a thermoshaker (400 rpm). The enzymatic digestion was stopped by adding 100 μL 20% (*w*/*v*) TCA and the extracts were centrifuged for 10 min at 24,104× *g*.

The supernatants were purified using solid phase extraction (SPE). For the SPE, a C18ec column was used, which was conditioned with methanol (3 mL) and equilibrated with formic acid (FA, 3 mL, 0.1% in water, *v*/*v*). Subsequently, the samples were applied and washed twice with aqueous FA (3 mL, 0.1% in water, *v*/*v*). The amino acid conjugates were eluted with methanolic FA (3 mL, 0.1% in methanol, *v*/*v*). The extracts were evaporated to dryness under a continuous stream of nitrogen and finally concentrated in 100 μL FA (0.1% in methanol/water, 80:20, *v*/*v*).

#### *2.6. HPLC-ESI-MS/MS Analysis of Protein Conjugates*

For LC-ESI-MS/MS analysis of the ITC protein conjugates, the protocol described by Kühn et al. was used with some modifications [20]. An aliquot (4 μL) was injected into the LC-ESI-MS/MS-system. For quantification, an external calibration with BITC-

lysine (BITC-Lys) conjugate synthesized by Kühn et al. was used in a concentration range from 0.01 to 4 μmol/L [23]. The LC-ESI-MS/MS system consisted of a 5500 QTrap triple quadrupole MS/MS system (AB Sciex Germany GmbH, Darmstadt, Germany) combined with an Agilent 1260 Infinity II HPLC-system (Agilent Technologies Deutschland GmbH, Waldbronn, Germany). For data acquisition and analysis, the software Analyst 1.7.0 (AB Sciex Germany GmbH) was used.

The separation of the analytes was performed on a Kinetex® C18 column (5 μm, 100 Å, 150 mm × 2.1 mm; Phenomenex Ltd., Aschaffenburg, Germany). The autosampler temperature was set to 4 ◦C and the column oven to 20 ◦C. The separation was performed with a flow rate of 300 μL/min of the mobile phase, consisting of 0.1% FA in water (eluent A) and 0.1% FA in MeOH (eluent B). At the beginning, the gradient consisted of 90% eluent A for the first minute. Subsequently, the concentration of eluent B increased linearly to 90% within 8 min and remained constant for 1 min. Afterwards, the start composition of eluent A and B was reached within 1 min. Finally, the column was re-equillibrated for 4 min with a composition of 90% eluent A.

The MS system was set to positive ionization mode with an entrance potential of 10 V. The ion spray voltage consisted of 5.5 kV, the desolvation gas temperature was 550 ◦C, the ion source gas pressure 70 psi for gas 1 and 55 psi for gas 2, and the curtain gas pressure was 40 psi.

#### *2.7. HPLC-UV Analysis of Desulfo-GLS*

The extraction of GLS and a conversion into desulfo-GLS were performed as previously described by Wiesner et al. with some modifications [29]. For extracting GLS from 10 mg freeze-dried plant material, 750 μL methanol (70% in water, *v*/*v*, 70 ◦C) were added as well as 25 μL of a *p*-hydroxybenzyl GLS solution (sinalbin, 1 mM in water) as internal standard. The mixture was heated at 70 ◦C for 10 min and afterwards centrifuged at 7748× *g* for 5 min at room temperature. The residue was re-extracted twice with 500 μL methanol (70% in water, *v*/*v*, 70 ◦C) and an incubation time of 5 min. All supernatants were combined.

To clean-up the extracts, small glass columns containing 500 μL of DEAE-Sephadex A-25 were used. The sorbent was activated with 6 M imidazole (2 mL, in FA 30% in water) and washed two times with water (1 mL), before sample-extracts were applied. Afterwards, 75 μL of purified β-glucuronidase (from *Helix pomatia*) were added and incubated overnight. Then, desulfo-GLS were eluted with 2 × <sup>500</sup> <sup>μ</sup>L of water, transferred to Corning® Costar® Spin-X® centrifuge tube filters (Merck KGaA) and centrifuged for 2 min at 5165× *<sup>g</sup>*. The sample solutions passed through the filters were used for HPLC-UV analysis.

An Agilent 1260 Infinity II LC system with an UV detector, equipped with a Poroshell 120 EC-C18 column (2.7 μm, 2.1 mm × 100 mm), was used for the separation. The autosampler temperature was set to 4 ◦C and the column oven to 23 ◦C. The mobile phase consisted of water (eluent A) and acetonitrile (eluent B) with a flow rate of 400 μL/min. In the beginning, eluent A was set to 99.8% for 2 min. Subsequently, the concentration of eluent B was linearly increased to 19.8% within 10 min and hold for 2 min. After that, the concentration of eluent B was further increased to 50% within 1 min and was kept constant for 1 min. Re-equilibration of the column was finally performed for 2 min with 99.8% of eluent A.

The identification of the desulfo-GLS based on a comparison of the retention time with reference GLS such as BG, 4-(methylsulfinyl)butyl glucosinolate and 2-propenyl glucosinolate at 229 nm. The quantification was done with *p*-hydroxybenzyl glucosinolate as internal standard and the specific response factor for each compound [30].

#### *2.8. GC-MS Analysis of GLS Breakdown Products*

Freshly frozen and ground bread (500 mg) was weighed frozen into a solvent resistant vessel. The extraction started with frozen bread powder that was allowed to heat to room temperature after adding the first round of methylene chloride. GLS breakdown products were extracted and analyzed with gas chromatography-mass spectrometry (GC–MS) using the extraction protocol described by Wermter et al. and the GC-MS protocol described by Hanschen et al., except that the transfer line temperature was set to 250 ◦C [31,32].

#### *2.9. Statistical Analysis*

For statistical analysis Statistica 64 (Version 13, Dell Inc., Tulsa, OK, USA) was used. The differences between the results of the two different applications No. I and No. II, and the non-processed and baked plant material was analyzed using the *t*-test. A statistical confidence level of 95% (*p* ≤ 0.05) was defined.

#### **3. Results**

#### *3.1. Quantification of GLS in Nasturtium and Garden Cress before and after the Baking Process*

In this experiment, the thermally-induced degradation of GLS during the baking process was investigated with a main focus on glucotropaeolin (BG) and its breakdown product BITC. For a comparison, the GLS content of the plant materials (*Tropaeolum majus* L., *Lepidium sativum* L.) was analyzed before and after the baking process. The results are shown in Table 2.

**Table 2.** GLS profiles and contents in nasturtium (*Tropaeolum majus* L.) and garden cress (*Lepidium sativum* L.) material, either non-processed (before baking) or after the baking process. Results are presented in μmol/g dry weight (DW) ± standard deviation. Because of the use of fresh garden cress the result of benzyl glucosinolate is additionally expressed in fresh weight (FW) for this material. Abbreviations: n.d.: not detected, BG: benzyl glucosinolate, IMG: indol-3-ylmethyl glucosinolate, 2-PE: 2-phenylethyl glucosinolate, 1-MeO-IMG: 1-methoxy-indol-3-ylmethyl glucosinolate.


<sup>1</sup> Results of three treatment replicates analyzed in technical duplicates.

In the non-processed nasturtium (*Tropaeolum majus* L.), the GLS profile consisted of only BG with a content of 11.5 μmol/g freeze-dried material, which results around 0.5 μmol BG/g bread before the baking process for application No. I and 0.18 μmol BG/g bread for application No. III. In the separated material after the baking process of the nasturtium in application No. III, no BG could be detected, which corresponds to a release and possible degradation of 100% BG during the baking process.

The GLS profile of the fresh garden cress (*Lepidium sativum* L.) microgreens consisted mainly of BG with an amount of 48.80 μmol/g in freeze-dried material, which corresponds to an amount of 12.14 μmol/g in fresh material and around 0.2 μmol BG/g bread in application No. II. It also contained 0.36 μmol/g indol-3-ylmethyl glucosinolate (glucobrassicin, IMG), 0.45 μmol/g 2-phenylethyl glucosinolate (gluconasturtiin, 2-PE), and 0.08 μmol/g 1-methoxy-indol-3-ylmethyl glucosinolate (neoglucobrassicin, 1-MeO-IMG) in freeze-dried material. After the baking process, the BG content in garden cress microgreens decreased to 27.28 μmol/g freeze-dried material, which corresponds to a decline of 44.0%. The reduction of the other GLS in the garden cress differed from 17% to 100% in the following order: 1-MeO-IMG > 2-PE > BG > IMG (Figure 3).

**Figure 3.** Thermal degradation of GLS in nasturtium and garden cress during the baking process. Results are expressed in % of residual glucosinolates of three treatment replicates analyzed in technical duplicates. Abbreviations: BG: benzyl glucosinolate, IMG: indolyl-3-methyl glucosinolate, 2-PE: 2-phenylethyl glucosinolate, 1-MeO-IMG: 1-methoxy-3-indolylmethyl glucosinolate.

With regard to the BG degradation and release of BITC, more BG was degraded when using the homogenized nasturtium leaves instead of the fresh garden cress microgreens, which could lead to a higher concentration of BITC in bread with freeze-dried material.

For the additional experiment where the myrosinase was deactivated by boiling the freeze-dried nasturtium material for 10 min, Figure 4 shows that even by adding the homogenized material to boiling water for 10 min BG is degraded by 75% from 11.51 μmol/g to 2.79 μmol/g in freeze-dried material. After the baking process no BG could be detected. Therefore, GLS in freeze-dried homogenized material can be strongly degraded by heat.

**Figure 4.** Thermal degradation of GLS in nasturtium after boiling in water for 10 min to deactivate the myrosinase and further baked into bread. The experiment and analysis were performed in duplicate. Abbreviations: BG: benzyl glucosinolates, n.d.: not detected.

#### *3.2. ITC-Protein Conjugates in Bread with Different Cress Genera*

To investigate possible reactions between BITC and wheat proteins, homogenized plant material of nasturtium (*Tropaeolum majus* L) was used for creating a large reaction surface and compared to the addition of garden cress (*Lepidium sativum* L.) microgreens, being comparatively more compact in structure.

The baked breads showed a difference in appearance, because of the differently applied plant materials. When adding a freeze-dried, powdered material, a homogenous green color of the crumb was obvious (Figure 5a), while the addition of fresh material

was more heterogeneously distributed. So, individual green particles were noticed and spread all over the crumb (Figure 5b); the reference bread showed no further particularities (Figure 5c).

**Figure 5.** Pictures of the different cress-enriched model breads. (**a**) freeze-dried nasturtium (*Tropaeolum majus* L, 4%) mixed into the dough (application No. I); (**b**) fresh garden cress (*Lepidium sativum* L., 1.5%) mixed into the dough (application No. II); (**c**) reference bread without plant material.

The LC-ESI-MS/MS method was developed for analyzing modifications with amino and thiol groups analyzed as ITC-lysine and ITC-cysteine conjugates. At hand of synthesized standards, the thiourea derivative BITC-lysine (BITC-Lys) and the dithiocarbamate BITC-cystein (BITC-Cys) can be quantified [20].

As shown in Figure 6, in the bread with freeze-dried powdered nasturtium leaves (*Tropaeolum majus* L., application No. I), 20.9 nmol BITC-Lys/g protein corresponding to 3.36 nmol BITC-Lys/g bread were determined, while in the bread with the fresh garden cress (*Lepidium sativum* L.) microgreens (application No. II), 17.2 nmol BITC-Lys/g protein (3.07 nmol BITC-Lys/g bread) were found.

**Figure 6.** Formed BITC-Lys conjugates in bread with freeze-dried nasturtium (application No. I) and garden cress microgreens (application No. II) mixed into the dough. The breads were prepared and analyzed in triplicate. The initial amount of BG in the breads differs around 40%, but the amounts of BITC-Lys conjugates differed in a range of 10% per gram bread. Therefore, the addition of 4% homogenized nasturtium leaves resulted only in a 10% higher concentration of BITC-Lys conjugation compared to the addition of 1.5% fresh garden cress.

In comparison to BITC-Lys, the dithiocarbamate BITC-Cys was not detected in the samples.

#### *3.3. Quantification of BITC and Benzyl Cyanide in Bread with Different Cress Genera*

Next to the identification and quantification of BITC-protein conjugates, the breakdown products of BG, BITC and the corresponding nitrile benzyl cyanide (BC), were investigated. This analysis was performed to obtain an indication of whether free BITC is detectable or whether the GLS has been transformed to the corresponding nitrile during the baking process. This would be visible in a low concentration of BITC and a high concentration of BC. Again, it was suspected that there might be a difference between the

homogenized nasturtium leaves, and the fresh material of garden cress, due to the different degree of destruction of the plant material prior to and during the baking process.

The results in Figure 7 show only a slight difference (ratio of 1:1.38) in the concentration of BITC (5.93 nmol/g bread) and BC (4.30 nmol/g bread) for application No. I. In the bread application No. II, only 1.51 nmol BITC/g bread and 42.74 nmol BC/g bread were analyzed, resulting in a higher ratio of 1:28 BITC to BC. Comparing the different applications, more BITC was detected in application No. I, but 10 times more BC was analyzed in application No. II.

**Figure 7.** Formed benzyl glucosinolate breakdown products BITC and BC in bread with freeze-dried nasturtium (application No. I) and garden cress microgreens (application No. II) mixed into the dough. The breads were prepared in triplicate. Abbreviations: BITC: benzyl isothiocyanate, BC: benzyl cyanide.

#### *3.4. ITC-Protein Conjugates in Bread with a Centrally Placed Nasturtium Bolus*

The migration behavior of ITC into the food matrix and the distribution of reaction products during food processing were further studied in a model bread, where a bolus of freeze-dried nasturtium material was placed centrally (application No. III). Following the baking process, the bolus was removed and the GLS content analyzed. The bread was divided into a core, an intermediate, and an outer fraction (Figure 2). The appearance of the crumb did not show any differences compared to the reference bread (Figure 8a,b).

**Figure 8.** Picture of the cress-enriched model bread. (**a**) with freeze-dried nasturtium (*Tropaeolum majus* L, 1.5%) placed as centrally bolus (application No. III); (**b**) Picture of the reference bread without plant material.

BITC-Lys was detectable in the crumb. As anticipated, the highest concentration with 9.02 nmol BITC-Lys/g protein (1.38 nmol BITC-Lys/g bread) was found in the core fraction, while the analysis of the intermediate fraction showed a lower concentration of only 0.89 nmol BITC-Lys/g protein (0.12 nmol BITC-Lys/g bread) (Figure 9). In the outer fraction BITC-Lys could still be detected with a concentration of 0.26 nmol BITC-Lys/g protein (0.02 nmol BITC-Lys/g bread). Again, the BITC-Cys conjugate could not be detected in any of the fractions.

**Figure 9.** Amounts of the BITC-Lys conjugates in the core, intermediate, and outer fraction of the bread with freeze-dried nasturtium added selectively (application No. III). Three different breads were prepared and analyzed in triplicate.

#### *3.5. Quantification of BITC and Benzyl Cyanide in Bread with a Centrally Placed Nasturtium Bolus*

It was also of interest to investigate the migration behavior of the BG breakdown products BITC and BC in the different fractions of the bread. As shown in Figure 10a, it was found that only in the core fraction BITC could be detected with a concentration of 0.52 nmol/g bread. BC was found in all three fractions with decreasing concentration from the core fraction to the outer fraction. In detail the results in Figure 10b show that in the core fraction 653.02 nmol BC/g bread was found, decreased by 64% to the intermediate fraction (235.25 nmol BC/g bread). The difference from the intermediate to the outer fraction (66.98 nmol BC/g bread) was around 70%.

**Figure 10.** Formed benzyl glucosinolate breakdown products BITC (**a**) and BC (**b**) in bread with freeze-dried nasturtium added as a bolus in the middle of the bread (application No. III). The breads were prepared in triplicate. Abbreviations: BITC: benzyl isothiocyanate, BC: benzyl cyanide.

#### **4. Discussion**

In the present study, the degradation of GLS, the migration of their main breakdown product ITC, the corresponding nitrile, and the formation of ITC-protein conjugates were exemplarily investigated in a vegetable-enriched bread.

For these studies, bread was baked with parts of fresh garden cress (*Lepidium sativum* L.) microgreens and homogenized freeze-dried powdered nasturtium leaves (*Tropaeolum majus* L.). The different cress genera mainly contained the GLS BG, which can be degraded to BITC and the corresponding nitrile. This allowed a direct conclusion on the used material (homogenized or microgreens), independent of different chemical structures and therefore, different degradation rates of GLS during a heat treatment [33]. Because of the different percentage of addition the bread with nasturtium contained 2.5 times more BG before the baking process, therefore it was assumed that there is a more intense release of BITC during the baking process and consequently to an enhanced conjugation with proteins.

With regard to the GLS profile before and after the baking process, 100% BG in the homogenized freeze-dried nasturtium were degraded. In comparison, only 44% BG were released during the process from the garden cress microgreens into the crumb. Therefore, in application No. I around 50 μmol/100 g bread dough were degraded and in application No. II around 9 μmol BG/100 g bread dough. A former study assumed that due to a low destruction of the plant matrix, the GLS remain longer intact in the plant material [24]. The results of the present study underlined this assumption. When using fresh, nonhomogenously material, the migration of released BITC into the surrounding food matrix might be hampered by the intact plant structure and the reduced reaction surface in comparison to a freeze-dried material (Figure 5b). The plant tissue of freeze-dried material is already more or less disintegrated. By adding the material to the moist bread dough, the GLS could already be degraded by the enzyme myrosinase before the baking process, which is a difference to the use of garden cress microgreens. During the baking process, further heat-induced degradation can occur. The more homogeneous, pulverized material also leads to easier migration of the GLS and their breakdown products into the surrounding matrix before and during the baking process because the homogenized material ensures a broader distribution in the crumb (Figure 5a).

For these reasons and the higher BG concentration it was assumed that there would be significantly more protein conjugates in application No. I. The results confirmed that there is indeed a higher concentration of BITC-Lys, but it is only 10% higher than in application No. II even though the BITC levels likely formed were lower, because only 44% BG is released. To explain these results, a correlation can be made to the analyzed breakdown products. The breakdown product results show that there is only a slight difference in the amount of unreacted BITC in the different breads. However, the BC results show a 10-times higher concentration in the bread with fresh garden cress compared to the bread with freeze-dried, homogenized nasturtium. Thus, overall, a higher amount of breakdown products is detectable in application No. II.

The difference can be explained by the different states of the plant materials. In application No. I, BG might be already enzymatically degraded to BITC before the baking process began. This BITC might be degraded by heat action at an early stage of the baking process and thus might be no longer available for follow-up reactions. In addition, there appears to be volatilization of BITC and BC, so that overall fewer breakdown products were detectable in the baked bread of application No. I. When fresh material is used, the degradation of BG starts probably mainly during the baking process, so that initially the released BITC is available for reactions with proteins directly during the baking process. BC very likely is formed in high amounts due to thermal degradation of BG, as nitriles have been found to be the main degradation products from GLS in heated vegetables [13,25]. Therefore, despite the lower initial concentration of BG in application No. II, a larger amount of BITC may have been available for the formation of protein conjugates, as shown in the results of the ITC conjugates.

In the present study, 0.01% of BG in application No. I and 0.02% in application No. II were transformed into BITC-Lys conjugates, resulting in around 0.01% modified lysine in both breads. In an experiment described by Kühn et al., curd was mixed with cress, resulting in a typical German spread ('Kräuterquark') and analyzed for possible ITC protein conjugates. The results showed a higher conjugation of BG to BITC-Lys of up to 2–4% [23]. Considering the pH value, which not only influences the formation of ITC conjugates, but also their stability, more lysine conjugates should be detectable in wheat bread (pH 5.9) than in curd (pH 4.5), since a pH value close to the neutral range leads to an increased formation and stabilization of thioureas as shown by Platz et al. [34]. Therefore, the different results could be explained by the far different processing approaches. Baking produces temperatures of up to 95 ◦C for around 30 min in the bread matrix (Figure 1), whereas curd with herbs is only mixed at room temperature.

In regard to the enzymatic degradation of GLS selected studies showed that a short heat exposure of white cabbage (*Brassica oleracea var. capitata f. alba* cv. Minicole), red cabbage (*Brassica oleracea var. capitata f. rubra* cv. Primero), and kohlrabi (*Brassica oleracea var. gongylodes cv.* Kolibri,) for 10 min at around 60 ◦C led to an increased ITC formation due to inactivation of specifier proteins that enhance enzymatically nitrile and epithionitrile formation. In contrast, longer heat exposure durations (of up to 120 min), especially in the range of 100 ◦C, led to a nitrile formation due to thermal GLS degradation. Nitriles then accumulate in the matrix, because of their heat stability. Compared to the nitriles, the concentration of ITC decreases with prolonged heating, which results in less ITC available for follow-up reactions with proteins [13]. During the baking process, ITC could decompose to a wide range of further breakdown products, depending on their chemical structure. As an example for an aromatic ITC, a study applying phenethyl isothiocyanate (PEITC) in aqueous solutions showed that PEITC can be thermally decomposed to phenethylamine and can further react to *N,N* -diphenethylthiourea [35]. This hydrolysis reaction of ITC results in less protein conjugates in heated food products because less ITC is available for follow-up reactions with proteins.

To understand not only the formation of ITC-protein conjugates, but also the migration of GLS, ITC, and possible reaction products, another bread with a centrally placed freezedried nasturtium bolus was prepared. This experiment showed that BITC migrated up to the outer fraction and the concentration of BITC must have been still high enough to detect BITC-protein conjugates. It could be demonstrated that the concentration of the conjugates was reduced by 90% from the core fraction to the intermediate fraction (Figure 7). However, ITC-protein conjugates are still detectable in the outer fraction reduced by 97% to the core fraction.

The analysis of the BG breakdown products BITC and BC showed, that BITC was only detectable in the core fraction, whereas BC, which was highly abundant, could be quantified in all three fractions reduced by 90% from the core fraction to the outer fraction. This result shows that BITC migrates to the outer fraction during the baking process and reacts with proteins. Furthermore, not only the BITC migrates from the core fraction to the outer fraction, as proven by the protein adducts, but also formed BC.

In comparison to application No. I with homogenized nasturtium leaves mixed into the dough, there might also be a difference of degradation of BG. Whereas BG might be mainly degraded enzymatically in application No. I due to the contact with the dough, the BG in freeze-dried material added as a bolus might be degraded as well by the heat during the baking process, because not all the material is in contact with the moisture of the bread dough. When adding homogenized nasturtium to boiling water and treating it for 10 min at 100 ◦C, already 75% of BG were degraded. Therefore, BG seems to be very heat labile, and this might explain the total degradation of BG that was observed.

In this study, 1.5% to 4% homogenized nasturtium was applied to the bread doughs. In a former study, bread has been enriched with vegetable powder in a range of 1–3%, which lies within the range of quantity used [36]. Studies using fresh diced vegetable ingredients, such as kale, beetroot, and spinach, enriched the breads in a range of 10 to 40% [6,24,37].

Here, the prepared breads with fresh garden cress microgreens were enriched with 1.5%. With regard to a higher enrichment around 10 to 40% the concentration of BG and thus, resulting BITC likely would further increase significantly. This again can lead to a more intense release and migration of ITC and more follow-up reactions with proteins.

In all experiments, only BITC-Lys conjugates were detected, although the calculated difference in the concentrations of lysine (2.3% in wheat proteins) and cysteine (1.6% in wheat proteins) in the initial wheat proteins was only 0.7%. In another study, where garden cress was implemented in curd, higher concentrations of BITC-Cys conjugates were detected, despite the lower concentration of cysteine (0.3%) to lysine (8.2%) of the milk proteins [23]. Because of these results it was assumed that more cysteine conjugates would be formed in the present study, which could not be confirmed. These differences can be explained by the pH value and the heat treatment during the baking process (Figure 1). The pH value of the food matrix not only influences the formation of BITC protein conjugates but also their stability. The pH of the bread (5.9) should lead to a higher formation of BITC-Lys conjugates [34]. Additionally, BITC-Cys conjugation is reversible and shows only little stability with increasing temperature, whereas reaction between the amino group of lysine and ITC are stable and temperature has only a minor influence on BITC-Lys conjugates [20,38,39].

Next to a potentially lower biological value, protein conjugations have further consequences. In former studies, an influence on typical digestive enzymes has already been observed: While trypsin and chymotrypsin were not able to fully hydrolyze ITC-modified proteins, pepsin activity was not hindered. These differences are assigned to the typical specificities of those enzymes to split near selected amino acids. Trypsin, for example, hydrolyzes peptide linkages containing lysine or arginine, whereas pepsin is a non-specific protease [19,40–42]. Therefore, the digestibility by trypsin can be more affected due to the BITC-Lys conjugation. Next to the influence on the digestibility, another experiment showed that ITC modifications also lead to a decrease of enzyme activity as shown with tyrosine phosphatase [43]. In food production, a reduced water solubility of proteins due to a conjugation with the hydrophobic BITC could be relevant, as well as, altered heat aggregation, foaming and emulsifying properties which was demonstrated at hand of AITC modified β-lactoglobulin [41,44].

In addition to GLS, Brassicales vegetables also contain other SPM such as flavonoids and carotenoids. The heat exposure during the baking might also affect these SPM, which could also lead to follow-up reactions [24,45]. It has already been shown that the antioxidant activity of phenolic compounds in vegetables is often not diminished, even when there is a thermally-induced breakdown to smaller compounds, but there is also a formation of complexes with proteins during the baking process [45].

#### **5. Conclusions**

In foods containing proteins and GLS-rich vegetables, reactions between ITC and proteins may occur and the reaction of BITC with lysine can lead to a lower biological value of the protein [46]. However, in comparison to non-heated foods, less protein conjugates are formed in heated food products, because the formation of ITC during the baking process is diminished due to evaporation and hydrolysis reactions. Therefore, a lower bioavailability of lysine due to reactions with ITC is less of a concern in heated processed foods, but there is also no health benefit from more available ITC.

In addition to the ITC-amino acid conjugates with cysteine and lysine investigated herein, there could be further reactions with other amino acids, such as methionine (1.1% in the bread) and tyrosine (1.5% in the bread). Such reactions have not yet been investigated and could lead to a higher ITC-protein conjugation than previously anticipated [47,48]. It could also be interesting to study how the concentration of conjugates change over a longer period of time, because bread is typically consumed over several days.

**Author Contributions:** Conceptualization, M.K. and S.R.; Data curation, M.K.; Formal analysis, M.K., M.P. and F.S.H.; Investigation, M.K. and F.S.H.; Methodology, M.K., L.C. and F.S.H.; Resources, S.R.; Supervision, S.R.; Visualization, M.K.; Writing—original draft, M.K., F.S.H. and S.R.; Writing review & editing, M.K., F.S.H. and S.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** Franziska S. Hanschen is funded by the Leibniz-Association (Leibniz-Junior Research Group OPTIGLUP; J16/2017).

**Data Availability Statement:** The data sets presented in this study are available on request from the corresponding author.

**Acknowledgments:** We would like to thank Maria Riedner and Gaby Graack for providing excellent support with the QTRAP 5500 mass spectrometric measurements and Andrea Jankowsky for her support regarding the quantification of the glucosinolates. We acknowledge support by the German Research Foundation and the Open Access Publication Fund of the TU Berlin.

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

#### **References**


## *Article* **Low pH Enhances the Glucosinolate-Mediated Yellowing of Takuan-zuke under Low Salt Conditions**

#### **Taito Kobayashi 1, Kei Kumakura 1, Asaka Takahashi <sup>2</sup> and Hiroki Matsuoka 1,\***


Received: 9 September 2020; Accepted: 20 October 2020; Published: 23 October 2020

**Abstract:** This study was performed to clarify the enhancement of the 4-methylthio-3-butenyl isothiocyanate induced yellowing of salted radish root (takuan-zuke) by low pH during short-term salt-aging at low temperature and low salinity. We used two different methods to prepare the dehydrated daikon prior to salt-aging: air-drying outdoors (hoshi takuan-zuke) or salting with a stone press (shio-oshi takuan-zuke). Low salt-aging at low temperature was carried out under pH control with citrate-phosphate buffer. The yellowing of both types of takuan-zuke was accelerated below pH 5, and the color of air-dried takuan-zuke was deeper than that of salt-pressed takuan-zuke. To elucidate this phenomenon, several previously reported yellowing-related compounds were analyzed by high-performance liquid chromatography. The result showed that the production of the primary pigment, 2-[3-(2-thioxopyrrolidin-3-ylidene)methyl]-tryptophan, was low compared with that in previous reports. Therefore, we suggest that an unknown pigment was generated through a previously unreported pathway.

**Keywords:** pickled vegetables; yellowing salted radish root; glucosinolate–myrosinase system; tryptophan biosynthesis; isothiocyanates

#### **1. Introduction**

Takuan-zuke (salted radish root) is a popular and traditional Japanese processed food. Japanese radish roots (*Raphanus sativus* L.; daikon) are dehydrated by either air-drying outdoors (hoshi takuan-zuke) or salting with a stone press (shio-oshi takuan-zuke) before pickling. The dehydrated daikon radishes are then pickled in salt or salty rice bran for several months. The color of takuan-zuke is transformed from the white color of daikon to bright yellow during the long salt-aging process at ambient temperature [1]. However, the yellow color of takuan-zuke is easily photobleached by visible light and often fades when it is displayed in stores [2]. Recently, the color of commercial takuan-zuke, which is pre-pickled at low temperature and low salt, has been noted to be a dull yellow. Commercial takuan-zuke is prepared using yellow coloring agents such as tartrazine and gardenia pigment for consistency. However, modern Japanese consumers prefer white takuan-zuke to yellow takuan-zuke, possibly due to the misunderstanding of why white radish is intentionally dyed yellow. The need for adding colorants can be avoided if the yellowing reaction is controlled.

In our previous reports, we elucidated the detailed mechanisms of the yellowing reaction. Figure 1 shows the production process for the yellow pigment contained in takuan-zuke. Briefly, the starting compound for the yellowing reaction is 4-methylthio-3-butenyl isothiocyanate (MTB-ITC; raphasatin), which is the main pungent compound of radish. MTB-ITC is generated enzymatically

from 4-methylthio-3-butenyl glucosinolate (MTB-GLS; glucoraphasatin), which is induced by cell damage during the dehydration process. Because MTB-ITC is readily degraded in the aqueous phase, it plays an essential role in the taste and flavor of processed radish [3]. MTB-ITC is converted to 2-thioxo-3-pyrrolidinecarbaldehyde (TPC) and 3-(methylthio) methylene-2-thioxopyrolidine (MeSTP), gaining a pyrrolidine ring by intramolecular cyclization and elimination of the methylthio group [4,5]. Furthermore, TPC, which has an aldehyde group, reacts with tryptophan via the Pictet–Spengler reaction to form 1-(2-thioxopyrrolidin-3-yl)-1,2,3,4-tetrahydro-β-carboline-3-carboxylic acid (TPCC) as a precursor to the yellow pigment [1,6]. 2-[3-(2-Thioxopyrrolidin-3-ylidene)methyl]-tryptophan (TPMT) from TPCC is the main yellow pigment in long-term salt-aged takuan-zuke [7]. The proportion of geometric isomers in TPMT is an important factor for yellow brightness and intensity, as the color of the (*Z*)-isomer at 100 ppm is significantly yellow with Δ*b*\* 10.5 compared to (*E*)-isomer. The *Z* to *E*-isomerization under visible light irradiation causes a chain reaction, making the preservation of the yellow color difficult without a light-shielding film [2]. We previously reported that TPMT formation from TPCC is a rate-limiting step that is pH- and temperature-dependent [7,8]. Therefore, it is difficult to use the characteristic natural color in the modern manufacturing method.

**Figure 1.** Yellow pigmentation process in takuan-zuke. Abbreviation. 4-methylthio-3-butenyl glucosinolate (MTB-GLS); 4-methylthio-3-butenyl isothiocyanate (MTB-ITC); 2-thioxo-3-pyrrolidinecarbaldehyde (TPC); 3-(methylthio) methylene-2-thioxopyrolidine (MeSTP); 1-(2-thioxopyrrolidin-3-yl)-1,2,3,4-tetrahydro-β-carboline-3-carboxylic acid (TPCC); 2-[3-(2-Thioxopyrrolidin-3-ylidene)methyl]-tryptophan (TPMT).

In our preliminary experiments, takuan-zuke was prepared with 10 mmol/kg buffering agent (pH 4, 5, and 6) at low temperature, to clarify the influence of pH on the yellowing reaction during the salt-aging process. We found that the yellow coloring during short-term salt-aging at low temperature was promoted under acidic conditions. Importantly, it is expected that microorganisms can be suppressed by pickling under acidic conditions [9]. In the present study, we aimed to evaluate the levels of known yellowing-related substances in takuan-zuke, and to understand the influence of acidic pH condition on the yellowing reaction for takuan-zuke.

#### **2. Materials and Methods**

#### *2.1. Preparation of Yellowing-Related Substances from MTB-ITC*

The yellowing-related substances, MeSTP, TPCC, and TPMT, from MTB-ITC were synthesized using previously reported methods [4,7,8,10]. MTB-ITC was synthesized by the reaction of a crude MTB-GLS solution extracted from radish sprouts, with myrosinase extracted from radish. TPC was synthesized by first dissolving MTB-ITC in acetone, adding acetic acid, and then sonicating the mixture. For synthesizing MeSTP, methanol and phosphoric acid were added to the MTB-ITC solution, and the mixture was heated under reflux (70 ◦C) overnight to obtain a precursor fraction. The precursor fraction was trans-solved in acetone, hydrochloric acid was added, and the mixture was synthesized by heating under reflux (60 ◦C). TPCC was synthesized by adding tryptophan and water to MTB-ITC, adjusting the pH to 2 or less with phosphoric acid, and sonicating. TPMT was synthesized by adjusting the crude TPCC solution to pH 7 and heating under reflux (37 ◦C). The synthesized compound was purified using column chromatography.

For preparing the internal standard for TPCC and TPMT analysis, 1-ethyl-1,2,3,4-tetrahydro-β-carboline-3-carboxylic acid (ETCA) was synthesized from L-tryptophan and propionaldehyde [11]. L-tryptophan (612 mg, 3.0 mmol) and propionaldehyde (191 mg, 3.3 mmol) were dissolved in 0.025 mol/L H2SO4 (26 mL) and allowed to react at 40 ◦C overnight. Next, a solid precipitate was collected. Subsequently, (*1S*, *3S*)- and (*1R*, *3S*)-ETCA were separated by preparative Octa Decyl Silyl (ODS) column using middle-pressure liquid chromatography (MPLC) (Smart Flash EPCLC AI-580S, YAMAZEN Co., Yodogawa-ku, Osaka, Japan) equipped with a Biotage® SNAP Ultra C18 column (Filling amount 60 g; Biotage, Sweden).

To prepare the internal standard for MeSTP analysis, (*R*)-3-((*S*)-ethoxy(methylthio)methyl)-2-thioxopyrrolidine (EtOTP) was synthesized from MTB-ITC. MTB-ITC (2.0 mmol) was dissolved in the mixture of ethanol (80 mL) and 1 M phosphoric acid (20 mL). The mixture was concentrated under reduced pressure and extracted with ethyl acetate. The ethyl acetate extract was separated by preparative silica gel-MPLC system equipped with a universal column (Premium 2 L). Subsequently, (*S*)-EtOTP was separated by preparative ODS-MPLC system equipped with a SNAP Ultra C18 column.

#### *2.2. Preparation of Salted Radish Roots (Takuan-zuke)*

Takuan-zuke was prepared according to a previously reported method with slight modification [8]. The pickling schedule is presented in Figure 2. In this experiment, we used the radish cultivar hoshi-riso daikon (Takii & Co., Kyoto, Japan), which was cultivated in a field at Misato-machi (Gunma, Japan) in August–November 2015. We prepared eight types of salted radish under different concentrations of McIlvaine buffering agent: 0 mmol/kg (only NaCl addition) (SR0/DR0), 10 mmol/kg addition (SR10/DR10), 20 mmol/kg addition (SR20/DR20), and 40 mmol/kg addition (SR40/DR40). The buffering agent was prepared using dipotassium hydrogen phosphate and citrate monohydrate (food additive grade, Kanto Chemical Co., Tokyo, Japan).

Hoshi takuan-zuke (abbreviated DR) was prepared according to the following method. After cutting off the root tips (3–4 cm), the whole daikon (269 kg including green leafy top) was hung to dry in a well-ventilated shady area for two weeks until it became dehydrated and flexible. After chopping the tops (3–4 cm) of root and the leaves, the hoshi daikon (78 kg) was pickled in 8 wt% NaCl (wet weight, the total weight of hoshi daikon and its priming water (33 wt% of daikon)) with pH buffering agents (0–40 mmol/kg of dehydrated daikon) for two months. The inner lid was placed on the daikon, and a stone weight (200 wt% of the dried daikon) was placed on the lid. The salting temperature was maintained at 4 ◦C.

Shio-oshi takuan-zuke (shio-oshi takuan-zuke, abbreviated SR) was prepared according to the following method. Fresh daikon (234 kg) was pickled with 8 wt% NaCl and 0−40 mmol/kg buffering agent of daikon and pressed under stone weights (200 wt% of daikon). After two days of salt-pressing, 2 wt% NaCl was added to the daikon, and the daikon was dehydrated for 12 days (total: two weeks). Shio-oshi daikon (27–35 kg) were pickled again in buffered saline (9−12 kg; 6 wt% NaCl, and 0−40 mmol/kg buffering agents). The pickling conditions (stone weight and storage temperature) were the same as that for hoshi takuan-zuke.

**Figure 2.** Time schedule for the pickling of daikon. SR indicates dehydration by salting (shio-oshi takuan-zuke). DR indicates dehydration by air-drying (hoshi takuan-zuke); SS indicates two days of dehydration; S indicates one week of dehydration. "X" is the buffer concentration, 0 mmol/kg (only NaCl addition) (SR0/DR0), 10 mmol/kg addition (SR10/DR10), 20 mmol/kg addition (SR20/DR20), and 40 mmol/kg addition (SR40/DR40).

For tryptophan analysis, three types of dehydrated radishes (DR: air-dried daikon with leaves; nDR: air-dried daikon without leaves; SR: salt-pressed daikon without leaves) were produced by dehydrating them for one week using radishes that were collected in 2016. The dehydration process was conducted using the same procedure that was used for the study conducted in 2015.

For the sample, one barrel was prepared for each condition, and two bodies were collected according to the schedule shown in Figure 2. The division and reduced samples were rapidly frozen with liquid N2 and lyophilized. The lyophilized sample was subjected to freeze grinding and homogenization using a Multi-Beads Shocker (MB901THW(S), Yasui Kikai Co., Osaka, Japan). Lyophilized samples were vacuum packed and then stored at −30 ◦C.

#### *2.3. Determination of Moisture, Salt Content, and pH of Takuan-zuke*

The water content was determined by subtracting the lyophilized weight from the wet weight. The salt content was measured using the coulometric titration method (SAT-210, TOA-DKK Co., Tokyo, Japan). The pH of takuan-zuke was measured by first adding pure water to the lyophilized sample to restore the wet weight, and then using a pH meter (HM-30, TOA-DKK Co., Tokyo, Japan).

#### *2.4. Color Measurement of Takuan-zuke*

Color measurement of takuan-zuke was conducted based on a previous report [8]. The colorimetric change was measured with a Minolta CM-3500d Spectrophotometer, with a D65 illuminant and a 10-degree observer. The Commission Internationale de l'Eclairage (CIE) 1976 *L*\**, a*\*, and *b*\* color scale values of takuan-zuke were obtained by reflectance color measurement (measurement area: 8 mm diameter). Each value was measured in three places on the skin side of the upper, middle, and lower part of the root.

#### *2.5. Quantitative Analysis of MTB-GLS in Takuan-zuke*

Determination of glucosinolate in takuan-zuke was performed as previously reported [12]. The lyophilized sample (ca. 100 mg) was extracted by 1.5 mL of 80% MeOH at 75 ◦C. After the

extraction mixture was heated at 75 ◦C for 10 min, 0.5 μmol sinigrin (internal standard, I.S.) was added. The crude extracts were adsorbed on an anion exchange resin (DEAE Sephadex A-25, GE Healthcare, California, CA, USA) and treated with sulfatase to convert desulfoglucosinolate overnight. The eluate containing the desulfoglucosinolate was quantified by ODS high-performance liquid chromatography (HPLC).

Analytical ODS HPLC was performed with an Agilent 1200-1260 system with a Poroshell 120 EC-C18 (100 × 3.0 mm ø, 2.7 μm; Agilent Technologies, Santa Clara, CA, USA). The flow rate was set at 0.85 mL/min, and the column temperature was set to 35 ◦C. Elution was achieved using a gradient of two eluents: H2O as eluent A and acetonitrile as eluent B. The gradient program was set at 0.2% B for 0.25 min, rising to 19.8% B at 6.00 min, and the remaining at 19.8% B to 7.00 min. Finally, the column was equilibrated using 0.2% B from 7.10 to 9.00 min. The results were detected at a wavelength of 229 nm.

#### *2.6. Quantitative Analysis of TPC in Takuan-zuke*

Quantification of TPC in takuan-zuke was performed by fluorescence derivatization method using 4-(*N,N*-dimethylaminosulfonyl)-7-hydrazino-2,1,3-benzoxadiazole (DBD-H) [13]. The lyophilized sample (10–50 mg) was added to 250 μL of 0.1% DBD-H in acetonitrile, 250 μL of 0.2 mM anisaldehyde in acetonitrile, and 500 μL of 0.5% trifluoroacetic acid in 60% (v/v) acetonitrile. The reaction mixture was shaken and incubated at 25 ◦C for 60 min. To 200 μL of the supernatant after centrifugation, 50 μL of 500 mM McIlvaine buffer (pH 5) and 50 mg of NaCl were added and shaken. The acetonitrile phase was considered the sample for HPLC.

Analytical HPLC was performed with an Agilent 1200–1260 system with a Poroshell HPH-C18 (100 × 3.0 mm ø, 2.7 μm; Agilent Technologies, Santa Clara, CA, USA). The flow rate was set at 0.85 mL/min and the column temperature was set to 40 ◦C. Elution was achieved using a gradient of two eluents: H2O as eluent A and acetonitrile as eluent B. The gradient program was: 25% B rising to 73% B at 5.5 min, further increasing to 100% B at 5.6 min, and remaining at 100% B to 5.99 min. Finally, the separation column was equilibrated using 25% B from 5.99 to 8.0 min. Fluorescence was detected with excitation at 450 nm and emission at 565 nm. Anisaldehyde was used as an internal standard.

#### *2.7. Quantitative Analysis of Yellow Pigment-Related Substances in Takuan-zuke*

The lyophilized sample (ca. 100 mg) was mixed with 250 μL of chloroform, 625 μL of methanol (including 40 nmol/mL EtOTP and 40 nmol/mL ETCA), and 250 μL of H2O (including 2.25% trifluoroacetic acid and 0.45% semicarbazide). The mixture was shaken and incubated at 37 ◦C for 30 min. After cooling on ice for 5 min, the mixture was centrifuged at 20,630× *g* for 1 min. To 1 mL of the supernatant, 500 μL chloroform and 500 μL H2O were added. After cooling on ice for 5 min, the mixture was centrifuged at 20,630× *g* for 1 min. The separated lower layer was obtained as a crude extract for MeSTP analysis, and the upper layer was obtained as a crude extract for L-tryptophan, TPCC, and TPMT analysis. The crude extract for MeSTP was concentrated to dryness using a centrifugal concentrator. The dried sample was dissolved in MeOH (300 μL) for HPLC analysis.

An aliquot of the upper layer extract was diluted to three times the original concentration, with 2% formic acid. A solid phase extraction (SPE) cartridge (Bond Elut Plexa PCX, 30 mg, 1 mL; Agilent Technologies, Santa Clara, CA, USA) was washed with 1 mL each of 1 M NaOH and 1 M HCl and conditioned with 1 mL each of MeOH and 2% formic acid. The diluted samples were loaded onto a PCX cartridge and washed with 1 mL each of 2% formic acid and methanol. The analytes were eluted with 1 mL each of alkaline eluent (30% ammonium hydroxide: 95% methanol = 5:95) in a tube containing 30 μL of concentrated formic acid. The eluate was evaporated to dryness with a centrifugal concentrator. After being dissolved in 200 μL of methanol, the sample solution was irradiated by long-wave ultraviolet (UV) light (UVGL-25, Funakoshi Co., Tokyo, Japan) at 375 nm and analyzed by HPLC.

Analytical HPLC was performed with an Agilent 1200–1260 system with a Poroshell HPH-C18 (100 × 3.0 mm ø, 2.7 μm; Agilent Technologies, Santa Clara, CA, USA). The flow rate was set at 0.7 mL/min, and the column temperature was set to 40 ◦C. Elution was achieved using a gradient of two eluents: 10 mM phosphate borate buffer (pH 8.2) as eluent A and methanol as eluent B. The gradient program was set at 15% B rising to 25% B at 7.00 min, rising to 100% B at 11.00 min. Finally, the column was equilibrated using 15% B from 11.01 to 13.00 min. The detection wavelengths were as follows: 268 nm for ETCA, EtOTP, and TPCC, 320 nm for MeSTP, and 400 nm for TPMT using a diode array detector. Tryptophan was detected by native fluorescence (excitation wavelength 285 nm, emission wavelength 348 nm). Each isomeric mixture of TPCC, TPMT, and MeSTP was separately quantified, and the results were documented as the sum. Internal standards were EtOTP for MeSTP and ETCA for the others.

#### *2.8. Statistical Analysis*

All quantitative data units are expressed in nmol per g of dry weight, and each value is expressed as the mean value ± standard deviation (*n* = 4). Multiple *t*-tests were performed using the Holm–Šidák method (α = 0.05). Significant differences between the treatment groups were determined with a two-way Analysis of variance (ANOVA), followed by a Tukey's multiple comparison test using GraphPad Prism ver. 8 for Macintosh (GraphPad Software, Inc., CA, USA).

#### **3. Results**

#### *3.1. Basic Data and pH Changes in Takuan-zuke Induced by the pH Bu*ff*ering Agent*

Temporal changes in pH in the two types of dehydrated daikon and takuan-zuke samples are shown in Figure 3. The pH of fresh daikon was 6.4. The pH of takuan-zuke without pH buffering agents gradually decreased to 5.6 in SR0 and 5.8 in DR0 after two months of salting. In contrast, the pH changes of SR- and DR-takuan-zuke with the addition of buffering agents were notably lower than in the case of non-buffered takuan-zuke. The pH values of the acidic buffered takuan-zuke samples decreased within one week of salting for SR groups and one month salting for DR groups. The pH lowering effect on DR samples by buffer addition was concentration-dependent, whereas no notable decrease in pH among SR samples was observed. With the addition of 40 mmol/kg buffer, the pH values of the SR40 and DR40 after two months of salting were 4.2 and 4.5, respectively.

**Figure 3.** Time-dependent changes in pH during the dehydration and salting process with different concentrations ofMcIlvaine buffering agent. SR indicates dehydration by salting (shio-oshi takuan-zuke); DR indicates dehydration by sun-drying (hoshi takuan-zuke). The arrows denote the time point of salt addition. F denotes the pH obtained from fresh daikon. S denotes the pH obtained from the dehydrated daikon after one week of salting. "0" denotes the start of salt-aging process. Symbols refer to different concentrations of McIlvaine buffering agent (mmol/kg): •, SR0; , SR10; , SR20; ◆, SR40; -, DR0; , DR10; , DR20; , DR40. Values are mean <sup>±</sup> standard deviation (SD) (*<sup>n</sup>* <sup>=</sup> 3). The error bar cannot be displayed because the standard deviation is small.

#### *3.2. E*ff*ect of Addition of pH Bu*ff*ering Agent on the Color of Takuan-Zuke*

In the fresh daikon, *L*\*, *a*\*, and *b*\* values were 73.1 ± 4.0, −0.6 ± 0.2, and 10.2 ± 1.1, respectively. Although *L*\* values for SR groups during salt-aging treatment fluctuated between 65.0 and 74.0, the difference that depended on buffer concentration was negligible. The *a*\* values during salt-aging treatment changed in the negative direction. Figure 4 shows the *b*\* values changes during the dehydration and salt-aging processes. Although *b*\* value in the SR groups with buffered salting increased more than that in SR0, no buffer concentration-dependent changes in color were observed. However, the yellowing reactions in the DR groups were increased in proportion to the buffer strength. The *b*\* value for DR groups by adding buffer increased significantly compared to that in the case for DR0. The Δ*b*\* value of DR40 after two months salting based on fresh daikon was 2.4-fold that of SR0 and 1.8-fold that of DR0.

**Figure 4.** Time-dependent changes in *b*\* values during the dehydration and salting process with different concentrations of McIlvaine buffering agent. SR indicates dehydration by salting (shio-oshi takuan-zuke); DR indicates dehydration by sun-drying (hoshi takuan-zuke). The arrows denote the time point of salt addition. F denotes the pH obtained from fresh daikon. S denotes the pH obtained from the dehydrated daikon after one week of salting. "0" denotes the start of the salt-aging process. Symbols refer to different concentrations of McIlvaine buffering agent (mmol/kg): •, SR0; , SR10; , SR20; ◆, SR40; -, DR0; , DR10; , DR20; , DR40. Values are mean <sup>±</sup> standard deviation (SD) (*<sup>n</sup>* <sup>=</sup> 3).

#### *3.3. E*ff*ect of pH on the Yellow Pigment Production Pathway*

The temporal changes in the yellowing-related substances in eight takuan-zuke samples are presented in Figure 5. The MTB-GLS level was highest at harvest (51.8 ± 1.0 μmol/g (dry weight; DW)). The levels of MTB-GLS during salt-pressing treatment were decreased and disappeared during the two months of salt-aging treatment. In the buffered SR samples, adding buffering agents slightly suppressed the degradation of MTB-GLS. The degradation rate of MTB-GLS in SR40 was slow compared to that in the other SR groups, and the residue rate in SR40 after two months of salt-aging was 11%. In the DR samples, slight hydrolysis of MTB-GLS was observed during the drying treatment; however, penetration of saline into dried daikon, during the salt-aging process, induced further hydrolysis of MTB-GLS.

**Figure 5.** Change in the amount of each yellow pigment and related substances during the dehydrating and salting process. SR indicates dehydration by salting (shio-oshi takuan-zuke); DR indicates dehydration by sun-drying (hoshi takuan-zuke); F indicates the amount of each pigment obtained from fresh daikon; S indicates the amount of each pigment obtained from the dehydrated daikon after one week of salting. "0" time point indicates the start of salt-aging. Symbols refer to different concentrations of McIlvaine buffering agent (mmol/kg): •, SR0; , SR10; , SR20; ◆, SR40; -, DR0; , DR10; , DR20; , DR40. Values are mean <sup>±</sup> standard deviation (SD) (*n* = 4). Data are analyzed using two-way Analysis of variance (ANOVA), followed by Tukey's multiple comparison test. **Abbreviation.** 4-methylthio-3-butenyl glucosinolate (MTB-GLS); 4-methylthio-3-butenyl isothiocyanate (MTB-ITC); 2-thioxo-3-pyrrolidinecarbaldehyde (TPC); 3-(methylthio) methylene-2-thioxopyrolidine (MeSTP); 1-(2-thioxopyrrolidin-3-yl)-1,2,3,4-tetrahydro-β-carboline-3-carboxylic acid (TPCC); 2-[3-(2-Thioxopyrrolidin-3-ylidene)methyl]-tryptophan (TPMT).

MeSTP, as a degradation product of MTB-ITC, was generated from an early stage of salt and/or saline addition. The production levels were nearly equal between SR and DR takuan-zuke. The effect of buffer on MeSTP production was negligible.

TPC, the primary degradation product of MTB-ITC, was generated after salt and/or saline addition. TPC content in the SR groups reached maximum levels after one month of salt-aging (8.8–18.7 μmol/g). In the DR groups, TPC content increased with the addition of saline and reached a maximum after one month of salt-aging (9.0–13.7 μmol/g). TPC level decreased significantly after two months of salt-aging, depending on pH buffering strength.

The content of tryptophan in the fresh daikon was 255 ± 11 nmol/g. The analysis of tryptophan revealed that air-dried dehydration treatment resulted in a significant increase in its levels after harvest (hoshi processing for one week: 2.9-fold, *p* < 0.001; two weeks: 3.4-fold, *p* < 0.001). In contrast, shio-oshi treatment resulted in a slight change in the tryptophan content. Tryptophan in DR0 showed a maximum content value after one month of salt-aging (988 ± 52 nmol/g), and this value was 2.7 times higher than that in SR0 (*p* < 0.001). With subsequent salt-aging, tryptophan content significantly decreased, depending on pH buffering strength.

TPCC was generated immediately after salt and saline addition and increased with the duration of salt-aging after dehydration. The increased TPCC levels during the salt-aging process were significantly different between SR0 and DR0, and the content of TPCC after two months of salt-aging were 244 ± 8 nmol/g and 392 ± 31 nmol/g, respectively. The effect of pH on TPCC formation increased significantly with salt-aging time in the DR group, but no change was observed in the SR group. The TPCC levels in the buffered DR samples after two months of salt-aging were greater than that in DR0 (DR10: 1.5-fold, *p* < 0.001; DR20: 1.9-fold, *p* < 0.001; DR40: 2.0-fold, *p* < 0.001). Table 1 shows the changes in TPCC and tryptophan in the DR group; tryptophan, which is a substrate for TPCC, showedmaximum content after one month of salting (Table 1). The decrease in tryptophan and the production of TPCC were almost equal in the buffered DR sample (DR10: 96%, DR20: 108%, and DR40: 91%).



The amount of change in tryptophan and 1-(2-thioxopyrrolidin-3-yl)-1,2,3,4-tetrahydro-β-carboline-3-carboxylic acid (TPCC) is shown as an increase or decrease from DR0-1. Concentrationof McIlvaine buffering agent: 0 mmol/kg (DR0), 10 mmol/kg (DR10), 20 mmol/kg (DR20), 40 mmol/kg (DR40). Data are expressed as means ± SD.

The conversion of TPCC to TPMT started immediately after salt-aging and continued to increase after two months of salt-aging. The pH buffering strength significantly increased TPMT in the DR groups, with salt-aging time, whereas no effect was observed in the SR groups. The amount of TPMT was 6.7−19.3 nmol/g in the DR samples after two months of salting, which was 1.7–3.0 times greater than that in the SR samples.

#### *3.4. E*ff*ect of Dehydration Method on L-tryptophan Metabolism*

Temporal change of L-tryptophan in three types of dehydrated daikon using those collected in 2016 (air-dried daikon with leaves; DR, air-dried daikon without leaves; nDR, and salt-pressed daikon without leaves; SR) is shown in Figure 6A. The content of L-tryptophan in DR samples was increased to 1.091 ± 13 nmol/g after seven days of drying, which was 6.7 times higher than that in fresh daikon. In the SR and nDR samples, the extent of increase of L-tryptophan was negligible compared to that in DR samples (*p* < 0.0001).

**Figure 6.** Temporal changes in the quantity of tryptophan obtained from daikon during the dehydrating and salt-aging process (**A**). Effect of dehydration treatment on tryptophan production. F indicates the value derived from fresh daikon. SS indicates the amount of tryptophan obtained from the dehydrated daikon after two days of salting. S indicates the amount of tryptophan collected from the dehydrated daikon after one week of salting. Symbols: -, dehydrated by sun-drying with leaves (DR); , dehydrated by sun-drying without leaves (nDR); •, dehydrated by salt-pressing (SR). The error bar cannot be displayed because the standard deviation is small. (**B**). Localization of tryptophan in dehydrated daikon with leaves after two days of salting (DR-SS) white bar, inside radish; black bar, outside radish (on the skin).

The localization of tryptophan in DR-SS is shown in Figure 6B. Tryptophan level was highest inside the upper part of the root at 304 ± 5.3 nmol/g. Tryptophan level inside the root was significantly higher than outside the root, and the levels reduced significantly toward the root tip.

#### **4. Discussion**

In this study, we analyzed the effect of acidic buffering agents on the yellowing reaction of short-term aged takuan-zuke. Air-dried radish (hoshi takuan-zuke) and salt-pressed radish (shio-oshi takuan-zuke), prepared under acidic conditions, were the brightest yellow at low temperature and low salinity. In particular, the *b*\* value of hoshi takuan-zuke was equivalent to that of takuan-zuke, which was prepared by long-term salt-aging under room temperature and high salinity, as discussed in our previous report [8].

As described in Section 2, the pH buffering agent was added based on the wet weight of the daikon; the amount of buffering agent per radish, excluding water and salt, was 0.7 mmol/g (DW) for shio-oshi daikon, and 0.2 mmol/g (DW) for hoshi daikon. Therefore, the amount of buffering agent per dry weight affected the pH of takuan-zuke. There was no difference in the *b*\* value of shio-oshi

takuan-zuke at pH 5 or lower. Therefore, addition of 40 mmol/kg or more buffering agent into hoshi daikon did not affect yellowing.

To clarify the yellowing effects of acidic pH, the dynamics of the known yellowing-related substances were analyzed. In the takuan-zuke manufacturing process, in the absence of pH buffer, the conversion rate of MTB-GLS into TPC via MTB-ITC in the SR and DR samples was 36.1% and 25.5%, respectively. Since hoshi takuan-zuke showed a slight decrease in MTB-GLS during the drying process, the myrosinase reaction was thought to be suppressed under low water activity. During the salt-aging process, degradation of MTB-GLS was suppressed with increasing buffer strength, and the accumulation level of TPC was decreased. Myrosinase activity has been reported to peak at pH 5.7, whereas at pH 3.9, it decreases to 60% relative to the maximum activity level [14]. In addition, enzymatic reaction by myrosinase, under acidic conditions, has been reported to produce not only isothiocyanates but also nitriles [15–17]. Therefore, we suggested that the amount of MTB-ITC and TPC formed decreased due to the inhibition of enzymatic reaction or induction of nitrile formation.

Tryptophan content was significantly different between the two dehydration processes. Tryptophan levels in hoshi takuan-zuke increased from the air-drying dehydration process to early salt-aging. In hoshi takuan-zuke without leaves, no increase in tryptophan was observed, as with shio-oshi takuan-zuke. Tryptophan was seen to accumulate in the upper-inside part of hoshi daikon with leaves. Since tryptophan is reported to be synthesized by chloroplasts in plants [18,19], it was concluded that tryptophan is synthesized in the chloroplast of the daikon leaf and then transferred to the root through the vascular bundle. This result indicated that the increased tryptophan was synthesized in the chloroplasts of leaves during the drying process, rather than in microbial fermentation.

The content of TPCC, which is a yellow pigment precursor, was significantly increased in hoshi takuan-zuke than in shio-oshi takuan-zuke, similar to the results observed for tryptophan. In SR, the tryptophan production rate was the same as the reaction rate of TPCC synthesis, so it was inferred there was no apparent change in the tryptophan content. TPCC synthesis from tryptophan is considered a stoichiometric reaction when a sufficient amount of substrate is used, and tetrahydro-β-carboline synthesis via the Pictet–Spengler reaction has been reported to show high reaction efficiency under low pH conditions [11,20]. TPCC synthesis in takuan-zuke was revealed to be promoted at acidic pH, although the difference in tryptophan biosynthesis across the dehydration methods was a limiting factor for TPCC synthesis during the salt-aging process.

The content of TPMT in hoshi takuan-zuke was increased to a higher level than in shio-oshi takuan-zuke, similar to the results of tryptophan and TPCC. However, in this study, the conversion of TPCC to TPMT under low temperature and low salt conditions was very negligible. Our previous study had shown the optimal pH for TPMT synthesis from TPCC to be either weakly acidic or neutral pH in vitro [7]. In hoshi takuan-zuke at room temperature, the conversion rate to TPMT was 28.7%, and the color of salt-aged takuan-zuke varied to reddish yellow with prolonged salt-aging [8]. In the present study, there was no increase in the *a*\* value despite the high *b*\* value during the short-termed salt-aged process. These results suggested the contribution of TPMT to the yellowing of takuan-zuke to be low under acidic condition. In addition, HPLC analysis detected two characteristic peaks in takuan-zuke under acidic conditions. This peak had a maximum absorption wavelength of 392 nm and 404 nm, different from that in TPMT. It was suggested that these compounds are unknown yellow pigments that contribute to the yellowing reaction under acidic conditions.

In this report, we clarified that low pH conditions during takuan-zuke processing promotes an unknown bright-yellowing reaction at low temperature. In addition, we found TPC to be the most abundant among the known yellowing-related substances and an essential intermediate for takuan-zuke coloring. Imai et al. have reported that frozen and grated daikon, adjusted to pH 4 or below using acetic acid, turned yellow upon long-term freezing. This yellow pigment is generated by the condensation of two molecules of TPC [21]. Therefore, we proposed that an unknown yellowing reaction with MTB-ITC and TPC occurs during aging of takuan-zuke at low pH and low temperature. Future research would aim to elucidate the structure and reaction mechanism of the unknown pigment.

#### **5. Conclusions**

Temperature and pH conditions during salt-aging are the rate-limiting factors of the yellowing reaction, and we observed that takuan-zuke aged with low salt and at low temperature turns pale yellow. We found that the yellowing reaction was accelerated even at low temperature by the salt-aging of takuan-zuke under acidic conditions. The TPC level, which is one of the important intermediates of the yellow pigment, was highest after one month of salt-aging, regardless of the dehydration treatment. Tryptophan, another important intermediate, was increased only in dried daikon with leaves. The acidified and salt-aged treatment promoted the generation of TPCC, which is a pigment precursor. However, the generation of TPMT, which is a yellow pigment, was marginal compared to that in a previous report. Therefore, it was suggested that the unknown yellow pigment was generated via a pathway different from that described in the previous report regarding the yellow change of takuan-zuke under acidic and low temperature conditions. In future studies, it will be necessary to identify the unknown yellow pigment and the detailed mechanism underlying its generation.

**Author Contributions:** The experimental design was constructed and supervised by T.K., K.K., A.T., and H.M. The samples were produced with the help of T.K., K.K., A.T., and H.T. Instrumental analyses were performed by T.K. and K.K. The manuscript was drafted and written by T.K. and H.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

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

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **A Physiological-Based Model for Simulating the Bioavailability and Kinetics of Sulforaphane from Broccoli Products**

**Quchat Shekarri and Matthijs Dekker \***

Food Quality and Design Group, Wageningen University and Research, P.O. Box 17, 6700 AA Wageningen, The Netherlands; quchat.shekarri@wur.nl

**\*** Correspondence: matthijs.dekker@wur.nl

**Abstract:** There are no known physiological-based digestion models that depict glucoraphanin (GR) to sulforaphane (SR) conversion and subsequent absorption. The aim of this research was to make a physiological-based digestion model that includes SR formation, both by endogenous myrosinase and gut bacterial enzymes, and to simulate the SR bioavailability. An 18-compartment model (mouth, two stomach, seven small intestine, seven large intestine, and blood compartments) describing transit, reactions and absorption was made. The model, consisting of differential equations, was fit to data from a human intervention study using Mathwork's Simulink and Matlab software. SR urine metabolite data from participants who consumed different broccoli products were used to estimate several model parameters and validate the model. The products had high, medium, low, and zero myrosinase content. The model's predicted values fit the experimental values very well. Parity plots showed that the predicted values closely matched experimental values for the high (*r*<sup>2</sup> = 0.95), and low (*r*<sup>2</sup> = 0.93) products, but less so for the medium (*r*<sup>2</sup> = 0.85) and zero (*r*<sup>2</sup> = 0.78) myrosinase products. This is the first physiological-based model to depict the unique bioconversion processes of bioactive SR from broccoli. This model represents a preliminary step in creating a predictive model for the biological effect of SR, which can be used in the growing field of personalized nutrition.

**Keywords:** physiological-based model; sulforaphane; glucoraphanin; compartmental model; broccoli; bioavailability; myrosinase; parameter estimation

#### **1. Introduction**

To determine the bioavailability of bioactive compounds in foods, it is important to know its composition, structure, how it interacts with other food components, and its fate in the human body after being ingested. Isothiocyanates (ITC) are formed from precursors, glucosinolates (GL), which are found in broccoli and other types of brassica vegetables [1]. Numerous studies investigated the health effects of some ITCs. One such ITC is called sulforaphane (SR), derived from the GL glucoraphanin (GR). Sulforaphane is known to reduce the risk of cancer, and has cardiovascular and central nervous system protection benefits [1,2].

Sulforaphane's health benefits have resulted in studies that investigated the physiological mechanisms involved in digesting plants that contain SR, and SR's absorption, metabolism, and excretion [3,4]. Glucoraphanin is converted to SR by plant endogenous myrosinase (MYR), a β-thioglucosidase hydrolase that catalyzes the removal of glucose to form an O-sulfated thiohydroximate intermediate (Figure 1). GLs and MYR are stored in separate compartments in the broccoli plant cells. Cell structure disruption, from processing (chopping, blanching, powdering etc.), mastication, or plant bruising, is required before MYR can bind GL to facilitate ITC formation [5]. Gut bacteria in the colon also have the capability to facilitate this conversion from GR to SR [6–14]. After ingestion, and transfer through the stomach, SR is absorbed in the intestinal tract into the blood, then distributed to various organs before it is eliminated from the body, mainly via renal excretion [15].

**Citation:** Shekarri, Q.; Dekker, M. A Physiological-Based Model for Simulating the Bioavailability and Kinetics of Sulforaphane from Broccoli Products. *Foods* **2021**, *10*, 2761. https://doi.org/10.3390/ foods10112761

Academic Editor: Didier Dupont

Received: 15 October 2021 Accepted: 8 November 2021 Published: 10 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Figure 1.** (**A**) In the mouth, myrosinase (MYR) converts glucoraphanin (GR) to an SR-O-sulfated thiohydroximate intermediate. Depending on pH conditions sulforaphane (SR) forms. SR-nitrile formation is preferred at low pH in the presence of epithiospecifier protein (ESP). (**B**) In the gut GR is converted to SR, SR-nitrile, and erucin by gut bacteria.

The bioavailability of SR is described as the fraction of the amount of SR that is ingested and/or formed in the body that reaches systematic circulation [3]. Related to bioavailability is bioaccessibility, which is the fraction of a compound that is released from the food and that reaches the absorption site. In the context of GLs, MYR and ITCs, the bioaccessible ITC is the fraction of ITCs released from the food matrix [3] or the fraction of GLs transformed to ITCs and released in the body. Bioaccessibility of ITCs is affected by the plant's inherent GL content (which varies from 47 to 806 mg/100 g fresh weight of broccoli [1]), processing, the food matrix [16–18], and the digestion [17,19,20].

The bioaccessibility of ITC could increase or decrease depending on the type of processing. Chopping, blending, powdering are particle size reduction methods that rupture the plant tissue and allow for MYR and GLs to diffuse out and bind to each other [5]. Heating affects epithiospecifier proteins (ESP), MYR, and GL content. ESPs are responsible for the conversion of GLs to nitriles [1] and are less heat stable than MYR. Their inactivation allows for the preferential formation of ITCs [21,22]. Any type of prolonged high temperature heating, however, may cause MYR denaturation [5,19,21] and GL thermal degradation [21]. Freeze drying has been shown to retain MYR and GLs [3]. Prior to freeze drying, microwave cooking at adequate power inputs, inactivates ESP while preserving MYR activity which increases bioaccessibility.

SR formation occurs mostly in two organs: the mouth during mastication and in the gut by the microbiota. Research shows that there are differences between individuals in oral processing of foods [23]. Sarvan et al. [20] investigated, in-vivo, the effect of steaming and chewing time on the bioaccessibility of SR and SR-Nitrile, the GR breakdown products after chewing. Results showed that longer chewing times of broccoli with active MYR led to more GR hydrolysis. Compared to raw broccoli, or broccoli steamed for shorter periods, chewing broccoli steamed for 2 min provided the highest amount of SR. Broccoli steamed for 3 min provided the least amount of SR [20]. The effect of chewing on bioavailability was demonstrated by Shapiro et al. [24] who measured the difference in the amount of ITC metabolites excreted from urine when broccoli sprouts were swallowed whole or chewed thoroughly. They found that chewing increased the amount of urine metabolites by 1.5 times.

In broccoli products with inactivated MYR due to prolonged heating, the gut conversion processes of GR to SR and other degradation products become important. The capability of an individual's gut microbiome to convert GR to SR will depend on the types of microbes, their quantities, and how effective their different mechanisms for bioconversion are. Gut bacteria convert GLs to other compounds besides ITCs (Figure 1). Saha et al. [6] used a batch fermentation model with human gut bacteria to demonstrate that gut bacteria is capable of converting GR to SR, SR-nitrile, erucin, and erucin-nitrile. They also showed that the formation of erucin is preceded by the microbial conversion of GR to glucoerucin [6]. Consequently, the bioconversion of GL to non-ITC breakdown products reduces the bioavailability of ITCs.

Capturing the essence of the physiological processes for SR mathematically so that its biological effect can be simulated and predicted, is known as physiological-based modelling. This is an approach that considers the physiological basis of a bioactive compound's interaction with the human body before mathematical concepts are applied. Physiological-based models vary in terms of the number of physiological aspects (i.e., biological mechanisms, organs) considered. Some models only look at the gastrointestinal tract while others consider the whole body [25].

Various types of compartmental model have been described of which the most basic is the compartmental absorption and transit (CAT) model. In the CAT model, the small intestine is divided into a series of compartments and assumes linear transfer kinetics, passive absorption kinetics and well mixed compartments with uniform concentration [26,27]. The transit and absorption of a drug or food component is depicted by the following equation,

$$\frac{dM\_n}{dt} = K\_l M\_{n-1} - K\_l M\_n - K\_d M\_{n\_l} \text{ } n = 1, 2, \dots \text{7} \tag{1}$$

where *n* is the number of compartments, *M* is the amount or concentration of the component in the *n*th compartment, *Kt* is the transit rate constant between compartments, and *Ka* is the absorption rate constant of the component into the blood.

Based on the CAT model, the advanced compartmental absorption and transit (ACAT) model was developed to include more details. The ACAT compartmentalizes the stomach and large intestine so that gastric emptying and absorption from the large intestine can be considered. In addition to linear kinetics and passive absorption, the model considers non-linear kinetics due to protein binding, liver metabolism, or active transport and physiochemical factors such as particle size, solubility, density, and permeability [28].

Most physiological-based modeling research available are for pharmaceutical drugs. There are few studies that are related to food components and food products, and even fewer studies for modeling broccoli compounds. Punt et al. [29,30] made whole body eight-compartment models to predict the bioactivation and detoxification of herb estragole in humans and rats. Le Feunteun et al. [31] made a five-compartment model that focused on the digestion of mini-pigs to study the effect of product matrices on the digestion of milk proteins. Strathe et al. [32] also made a model with four main compartments and 38 sub-compartments to study the digestion and absorption of macro-nutrients in growing pigs. Moxon et al. [33] made a two-compartment model to investigate the effect of gastric emptying, luminal viscosity and hydrolysis rate on the rate of glucose absorption.

At the time of writing this article one study was found about the physiological-based modeling of SR from broccoli. Li et al. [34] investigated the kinetics and distribution of sulforaphane in the tissues of mice using a physiological-based model, where the whole body was divided into eight compartments. The mice ingested fresh, steamed, and MYR treated steamed broccoli sprout powders. The difference in kinetics and distribution in the tissues between the three different products were compared. The model did not include SR absorption mechanisms, and it did not include GR to SR conversion processes in the mouth, via myrosinase, and in the gut, via microbes. Also, the study did not extrapolate their results from mice to humans. The conversion of GR in the gut and mouth are important process that affect bioavailability. Therefore, a physiological-based model that considers these processes is needed.

The objective of this study was to make a physiological-based model that describes the kinetics and bioavailability of isothiocyanates from broccoli and to evaluate how the derived parameters are impacted by inter-individual variation. The model is validated against urine excretion sulforaphane data from a previous 2014 Wageningen University in-vivo research study by Oliviero et al. [35]. In this study, the effect of residual myrosinase activity on ITC formation, bioavailability, and excretion kinetics was investigated after 15 test subjects (apparently healthy human volunteers, aged 26–50 years, body mass index <sup>21</sup> ± 2 kg/m2, six men and nine women, 13 Caucasian, two Asian, and one Latin American), consumed five broccoli products with different levels of myrosinase activity obtained by different levels of microwave heating.

#### **2. Materials and Methods**

#### *2.1. Pre-Modeling Data Processing*

Participant raw data (measured SR urine conjugate excretion rates) from the Oliviero et al. [35] study was preprocessed for use in Matlab. The data were the time (minutes) and sulforaphane (SR) excretion rate (μmol/min). The technique used to measure SR urine conjugates, solid phase extraction-HPLC-MS/MS [35–37], is associated with experimental error that was quantified by Vermeulen et al. [37]. The relative standard deviation 12, 6, 3% for 1.04, 10.5, and 313 μM SR, respectively, in urine, was used to derive the following exponential equation that helped estimate the experimental error of each data point.

$$y = 0.11505x^{-0.24} \tag{2}$$

The relative error ratio is y, and the concentration of SR is *x*. The experimental errors were plotted as error bars on the data points for the model fittings.

#### *2.2. Model Description and Assumptions*

The model (Figure 2) focuses on the processes involved in the gastro-intestinal transit of glucoraphanin (GR) and sulforaphane (SR). Similar to an advanced compartmental absorption and transit (ACAT) model, it includes the stomach, seven compartments of the small intestine [28], the colon, and a blood compartment for systemic circulation. Unlike an ACAT, the colon, was divided into seven compartments, the stomach into two compartments [31]. A mouth compartment, which is typically not in physiological-based models, was included. As a result of this the full model contains 18 compartments.

The products consumed in the intervention study were portions of 5 g of each broccoli product with 90 mL of water at 40 ◦C, and with 30 g of raisin bun and water ad libitum. During mastication, myrosinase (MYR) and GRs released from the cell structures react to form an O-sulfated thiohydroximate intermediate, which then immediately converts to SR or SR-nitrile. The amount of the intermediate that is converted to SR versus SR-nitrile is a ratio that is subject to change depending on the individual's chewing pattern and broccoli product. In the mouth, it is assumed that SR-nitrile is the only non-ITC compound formed. Mastication time (30 s) and saliva flow rates (0.033 mL/s) [38] are assumed to be the same for all participants. The volume of the mouth compartment is the product plus the saliva excretions, 0.096 L.

**Figure 2.** Schematic diagram of model with one mouth compartment, two stomach compartments, seven small intestine (SI) compartments and seven large intestine (LI) compartments, and one blood compartment. The conversion of glucoraphanin (GR) to sulforaphane (SR) and SR-nitrile and erucin in the mouth and large intestine is depicted. Shown are the transit rate constants between the mouth and stomach (Kmouth), between the two stomach compartments (KSH), between the stomach and duodenum (S), between the small intestines (KtSI), between the large intestines (KtLI). The absorption of SR from the SI and LI into the blood is represented by the absorption rate constant (Ka). Elimination of SR and SR-conjugates from the blood to urine is represented by the rate constant, Ke. During mastication, GR is converted to an intermediate by myrosinase (MYR) and the subsequently converted to SR or SR-nitrile based on a conversion ratio (SRR). The gut microbial conversion of GR to SR and SR-nitrile/erucin is represented by the rate constants Kf and Keni, respectively.

Swallowing transfers the bolus to the stomach. The first stomach compartment accounts for the disintegration of food particles that are too big to pass through the pyloric sphincter valve leading into the small intestine. Food broken down sufficiently, and mixed with gastric fluids in the first compartment, is moved to the second stomach compartment where it mixes with more gastric juices before emptying into the duodenum. Gastric emptying of solid foods has been described as having a biphasic nature due to the time required for enzymatic and mechanical disintegration before emptying into the intestines [39]. Any MYR is assumed to be deactivated irreversibly in the stomach due to the low pH of the gastric fluids [19,40]. It is also assumed that GLs and ITCs are not absorbed in the stomach. Transit from second stomach compartment to the duodenum depends on the stomach emptying time (30 min) which is assumed to be the same for all participants since the size of the meal is the same. Based on the meal size, the volume of both stomach compartments together is 0.2 L, of which 0.05 L is the volume of the first stomach compartment (Table 1).

The chyme is mixed with duodenal secretions in this first compartment of the small intestine (SI). Due to the differences in the intestinal lining of duodenum (less villi/area for absorption) compared to the rest of the small intestine, less nutrients are absorbed in the duodenum; for simplicity, it is assumed there is no ITC absorption. In the remaining six SI compartments, SR is absorbed into the blood as both GR and SR are transferred from one compartment to the next.

The large intestine is divided into seven compartments where the following processes takes place: formation of SR, nitriles, and erucin by gut bacteria, absorption of SR into the blood, and transit of compounds from one compartment to the next. It is assumed that GRs are not absorbed from both the small and large intestines. Movement of chyme in the intestines are in the forward (towards the rectum) direction. Backward movements are known to occur and are represented by the fact that the compartments are assumed to be well mixed.

ITC is absorbed into the blood plasma, which is represented by one compartment that is presumed to be the same volume between all participants (5 L). Absorption of ITCs across the intestinal wall occurs passively by diffusion and is described according to Fick's Law of Diffusion [41]. ITC metabolites are eliminated from systemic blood circulation via glomerular filtration. No other elimination processes (sweat, defecation, respiration) are considered.

All compartment volumes (Table 1) are assumed to stay constant and the concentration of compounds (GR and SR) per compartment is uniform.

**Table 1.** Parameters values used in the model for the broccoli products, oral GR conversion processes, gastro-intestinal processes, gut GR conversion processes. Parameter values for the broccoli products MYR, Cgl0 (initial glucosinolate concentration), and ITC0 (initial sulforaphane concentration) are separated by product (HighBP, HighBF, MedBF, LowBF, NoBF, each of these products have different MYR content due to different levels of microwave heating). Parameters designated as 'Estimated' were used in model fittings.


\* Ranges for the parameters were determined based on literature. \*\* These values were estimated for each individual participant based on the model fit of the experimental values. † ITC0 values used in final fittings for MedBF and LowBF were approximately 3.4 and 9.1%, respectively, of their values in this table due to poor fit results using the original values.

#### *2.3. Compartmental Mathematics*

The enzymatic reaction of GL and MYR to form the O-sulfated thiohydroximate intermediate in the mouth is characterized by a Michaelis–Menten equation that accounts for enzyme inhibition. The intermediate instantly reacts to form ITC or ITC-nitrile, therefore the change in GL concentration is negatively proportional to the Michaelis–Menten equation.

$$\frac{d\mathbb{C}\_{GL}}{dt} = -\left(\frac{V\_{\text{max}} \times MYR \times BR \times \mathbb{C}\_{GL\text{ }Muth}}{K\_{\text{m}} + \mathbb{C}\_{GL\text{ }Muth} + \frac{\mathbb{C}\_{GL\text{ }Muth}^{2}}{K\_{i}}}\right) \tag{3}$$

*MYR* is the estimated mg of myrosinase per one mg of dried broccoli. Details on how *MYR* was estimated is found in Appendix A Part I. *BR* is the amount of dried broccoli (5 g) consumed by the participants. The maximum rate, *Vmax* (μM/min\*mg *MYR*), the Michaelis constant, *Km* (μM), and the inhibition constant, *Ki* (μM) were derived using glucoraphanin data from Roman et al.'s [42] MYR kinetic study. Details on the derivation of these variables can be seen in Appendix A Part II.

The amounts of ITC and nitriles formed is expressed as a fraction of the amount of intermediate (Equations (4) and (5)).

$$\frac{d\mathcal{C}\_{ITC\text{ Math}}}{dt} = \mathcal{S}RR \times \frac{d\mathcal{C}\_{GL}}{dt} \tag{4}$$

and

$$\frac{d\mathbb{C}\_{\text{Nitrille Math}}}{dt} = (1 - SRR) \times \frac{d\mathbb{C}\_{GL}}{dt} \tag{5}$$

*SRR* is the fraction of hydrolyzed *GR* converted to *SR* in the mouth. The remaining, 1 − *SRR*, is converted to nitriles.

Transfer of ITCs and GLs out of the mouth, as well as into and out of the stomach and intestinal compartments are first order rate reactions (Equations (6) and (7)).

$$\frac{d\mathbf{C}\_{ITC\ \dot{i}}}{dt} = k\_{transfer\ (i-1\ \text{to}\ i)} \times \frac{V\_{i-1}}{V\_{\dot{i}}} \times \mathbf{C}\_{ITC\ \dot{i}-1} \tag{6}$$

and

$$\frac{dC\_{GL\ i}}{dt} = k\_{\text{transfer } (i-1 \text{ to } i)} \times \frac{V\_{i-1}}{V\_i} \times C\_{GL\ i-1} \tag{7}$$

The rate constants for transfer from the mouth, *kmouth*, for the stomach compartments, *kSH* and *S*, and intestinal compartments, *ktsi* and *ktli*, are the inverses of the residence time of each compartment (Table 1). Due to differences in some compartment volumes (V), volume ratios are considered.

The gut formation of ITCs, nitriles, and erucin are also represented by the following first order rate reactions,

$$\frac{d\mathbb{C}\_{ITC}}{dt} = k\_f \times \mathbb{C}\_{GL} \tag{8}$$

$$\frac{d\mathbb{C}\_{Nitrille}}{dt} = k\_{cni} \times \mathbb{C}\_{GL} \tag{9}$$

where *kf* is the rate constant of formation for ITC, and *keni* is the rate constant of formation for erucin and nitrile. The change in GL concentration in the gut is proportional to the formation of ITC, nitrile, and erucin.

$$\frac{d\mathbb{C}\_{GL}}{dt} = -(k\_f \times \mathbb{C}\_{GL}) - (k\_{emi} \times \mathbb{C}\_{GL})\tag{10}$$

Absorption into the blood is defined by the following equation,

$$\frac{d\mathbb{C}\_{ITC}}{dt} = k\_4 \times \mathbb{C}\_{ITC} \tag{11}$$

where *ka*, is the rate of absorption (passive diffusion) from the intestines and it is proportional to the effective permeability (*Peff*) of sulforaphane and inversely proportional to the radius, R, of the intestines [49].

$$k\_d = \frac{2P\_{eff}}{R} \tag{12}$$

Each compartment is defined by mass balanced differential equations that contain the processes just described. For example, the full ITC and GL differential equations for the seventh small intestine compartment and first large intestine compartment are shown, Equations (13)–(16),

Seventh small intestine compartment:

$$\begin{array}{ccccc} \frac{dC\_{\text{ITC SID}}}{dt} = \frac{V\_{\text{SI}}}{V\_{\text{SI}T}} & \times \left(k\_{tSI} \times \text{C}\_{\text{ITC SI}}\right) & -\left(k\_{tSI} \times \text{C}\_{\text{ITC SI}T}\right) & -\left(k\_{a} \times \text{C}\_{\text{ITC SI}T}\right) \\ & \text{ITC In} & \text{ITC Out} & \text{ITC Absorption} \\ \end{array} \tag{13}$$

$$\begin{array}{cccc} \frac{d\mathbb{C}\_{\text{GL}\ \text{SLT}}}{dt} = \frac{V\_{\text{S}I\'}}{V\_{\text{S}\!\!T}} & \times \left(k\_{t\!\!SI} \times \mathbb{C}\_{\text{GL}\ \text{S}\!\!f\!\!h}\right) & -\left(k\_{t\!\!SI} \times \mathbb{C}\_{\text{GL}\ \text{S}\!\!T}\right) \\ & \text{GL}\ \text{In} & \text{GL}\ \text{Out} \end{array} \tag{14}$$

First large intestine compartment:

$$\begin{array}{ccccc}\hline \frac{\text{AUC}\_{\text{ITLCLI}}}{\text{d}t} = \frac{\text{V}\_{\text{IT}}}{\text{V}\_{\text{LI}}} & \times \left(k\_{\text{ISI}} \times \text{C}\_{\text{ITC}} \,\text{s} \,\text{f} \,\text{P} \,\right) & + \left(k\_{f} \times \text{C}\_{\text{GI}} \,\text{L} \,\text{f} \,\text{1} \right) & - \left(k\_{\text{LI}} \times \text{C}\_{\text{ITC}} \,\text{L} \,\text{f} \,\text{1} \right) & - \left(k\_{\text{a}} \times \text{C}\_{\text{ITC}} \,\text{L} \,\text{f} \,\text{1} \right) \\ & \text{ITC} \,\text{In} & \text{ITC} \,\text{Formation} & \text{ITC} \,\text{Out} & \text{ITC} \,\text{Out} \\ \hline \end{array} \tag{15}$$

$$\begin{array}{ccccccccc}\hline \frac{\partial \mathbb{C}\_{\text{UL},\text{LII}}}{\partial \mathbf{t}} & -\frac{\text{V}\_{\text{UL}}}{\mathbf{V}\_{\text{LR}}} & \times (\mathbb{k}\_{\text{SII}} \times \mathbb{C}\_{\text{GL }\text{SII}}) & -(\mathbb{k}\_{f} \times \mathbb{C}\_{\text{GL }\text{LII}}) & -(\mathbb{k}\_{\text{LII}} \times \mathbb{C}\_{\text{GL }\text{LII}}) & -(\mathbb{k}\_{\text{wcl}} \times \mathbb{C}\_{\text{GL }\text{LII}})\\ & \text{GL.In} & \text{ITC Formationation} & \text{GL.Out} & \text{Erucic}/\text{Nitrilko Formation} \\ & & & & & \\ \hline \end{array} \tag{16}$$

Full list of differential equations for all compartments are in the Supplementary Files.

#### *2.4. Simulink Model*

The mathematical equations used to represent the different compartmental processes were translated to a block diagram on Matlab's Simulink application (Matlab R2020a). Code scripts written on the MATLAB (MathWorks) interface integrated with the Simulink block diagram model to run sensitivity analyses and to perform fitting on the five different broccoli products.

#### *2.5. Matlab Coding and Fittings*

Sensitivity Analysis. Sensitivity analysis was performed to see the effect of changing parameter values on the model output. A parameter was changed within literature determined ranges while other parameters were kept constant. The analysis was performed for the different broccoli products, NoBF, LowBF, MedBF, HighBF, and HighBP.

Model Fittings. After the sensitivity analysis, parameters with the most influence on the simulation output, were used for fitting the model to each participant's data set. The fitting was done using a Trust-Region-Reflective Least Squares algorithm from the least squares data fitting solver of Matlab's Optimization Toolbox. The simulation period was 1600 min (26 h). Values of parameters used for fitting are in Table 1. This fitting procedure yielded parameter estimates that gave the best fit between the model simulations and the experimental data of the intervention study.

#### *2.6. Statistical and Data Analysis*

Confidence Intervals for each parameter estimate were determined at 90% confidence using Matlab's non-linear parameter confidence interval function. This function also provided covariance matrices that were used to calculate the correlation coefficients.

Using Matlab's trapezoidal numerical integration function, the cumulative SR excreted per participant for the experimental and predicted data sets were calculated.

Full codes are provided in the supplementary files.

#### **3. Results**

#### *3.1. Sensitivity Analysis and Parameters Selection*

The model's sensitivity towards twelve parameters was tested. Based on the analysis, three parameters were used in model fittings for HighBP and HighBF, seven for Med and Low BF, and five for NoBF (Table 2).

**Table 2.** Summary of parameters that influenced the simulation outputs for each broccoli product during the sensitivity analysis. Parameters with a check mark (-) were used for model fittings; parameters with an O, influenced simulation output but were not used in model fittings; parameters with an x did not affect output during the sensitivity analysis and were excluded from fitting. Three parameters were used to fit HighBP and HighBF, seven for MedBF and LowBF, and five for NoBF.


Parameters that did not affect model output were excluded. Gastric emptying time (St), and the rate constant of absorption (ka), were excluded from all model fittings because they had insignificant effects on the output (example in Figure 3A,B). As expected, myrosinase content (MYR), initial GR concentration (Cgl0), initial SR concentration (ITC0), and the ratio of GR converted to SR in the mouth (SRR), caused a direct and proportional upward shift to the HighBP, HighBF, MedBF, and LowBF simulation outputs (example in Figure 3C,D). The MYR and ITC0 contents were low for NoBF, therefore, of the broccoli product related parameters, only Cgl0 affected the output. Furthermore, the conversion ratio of GR to SR in the mouth (SRR), which depends on myrosinase content, did not affect the output of NoBF during the sensitivity analysis since myrosinase was inactive (Figure 3J). The first stomach rate constant (KSH), small intestine transit rate constant (ktSI) and SR elimination from the blood (ke), also caused proportional upward shifts but a narrowing of the curves was observed (example Figure 3L). The gut parameters, ITC formation (kf) and erucin and nitrile formation (keni) rate constants, had opposite effects on outputs. Increases in keni resulted in downward shifts and narrowing of the output curves (Figure 3H), while the curves shifted upwards for kf (Figure 3G). Increasing gut transit rate constant (ktLI), decreased the size of the second peak for MedBF and LowBF, and the single peak for NoBF (example in Figure 3F). Changes in ktLI did not affect outputs of HighBP and HighBF products (Figure 3E).

#### *3.2. Model Fittings*

The model was successfully fit to the data of each participant and for each product. Figure 4 shows the model fits for five of the participants and for all five broccoli products. 75 fittings were possible (15 participants times five products) but only 72 fitting were performed. Three data sets were excluded due to lack of data. While four data sets had to be preprocessed before fitting.

Few data sets had poor fits. Participant *m*, for the MedBF product, was underfitted while participants *q*, MedBF, and *g*, LowBF, were overfitted (Figures 4 and S1: Model Fittings Results for All Participants in the Supplementary Files).

**Figure 3.** HighBP sensitivity analysis for (**A**) gastric emptying time, St; (**B**) the rate constant of absorption, ka; (**C**) myrosinase content, MYR; (**D**) the ratio of GR converted to SR in the mouth, SRR; (**E**) large intestine transit rate constant, ktLI. LowBF sensitivity analysis for (**F**) large intestine transit rate constant, ktLI; (**G**) ITC formation rate constant in the gut, kf; and (**H**) erucin and nitrile formation rate constant in the gut, keni. NoBF sensitivity analysis for (**I**) myrosinase content, MYR; (**J**) the ratio of GR converted to SR in the mouth, SRR; (**K**) and initial ITC concentration (ITC0) and (**L**) small intestine transit rate constant, ktSI. The sensitivity analysis shows how the simulation output changes when all parameters are kept constant while one changes.

**Figure 4.** Model fittings for participants a, b, d, n, q. X and Y axis for each graph is, time (min) and sulforaphane conjugate excretion rate (μmol/min), respectively. Experimental data are the square bullets, and the solid lines are the model fits. Error bars represent potential experimental error from the analytical techniques used by Oliviero et al. to measure the amounts of ITC conjugates in urine. All participant model fittings are in Figure S1: Model Fittings Results for All Participants in the Supplementary Files.

#### *3.3. Bioavailability of Sulforaphane*

Bioavailability was calculated by dividing the cumulative amounts of SR by the amount of GR in the broccoli products. Cumulative amounts of SR for the experimental data and model data were determined using Matlab's trapezoid function to integrate. There were small differences between the predicted bioavailability and experimental bioavailability (Table 3). Average HighBP predictions were 2% larger than the calculated experimental bioavailability. The difference was 1% for MedBF, 0.9% for LowBF, and 0.1% for HighBF and NoBF.


**Table 3.** Average sulforaphane (SR) bioavailability values ±SD for the experimental data and model. There were minor differences between the experimental bioavailability and model bioavailability for each broccoli product type.

#### *3.4. Mouth and Gut Parameter Estimations*

Tables 4–6 shows the distribution (boxplots) of mouth and gut parameter estimates for each product type. Interquartile ranges (IQR) for the parameter estimates varied widely. The distributions of most parameter estimates are positively skewed indicating that 50% of participants are less variable within the first two quartiles than the 50% of participants in quartile three and four.

**Table 4.** SRR estimation results (horizontal red bar: median, blue box second and third quartile, whiskers first and fourth quartiles).


**Table 5.** kf estimation results (horizontal red bar: median, blue box second and third quartile, whiskers first and fourth quartiles, +: outliers).

\* Some or all outliers excluded for better visual presentation and comparison of boxplots. Outliers are values more than 1.5 times the IQR. See Supplementary Files for plotted outliers and for expanded view of the NoBF kf boxplot.

> SRR, the ratio of GR that gets converted to SR in the mouth, was estimated for each participant who consumed the HighBP, HighBF, Med and Low BF products (Table 4). SRR was not fitted for the NoBF product since the amount of MYR is very low. Therefore, any GR converted to SR in the mouth is insignificant for the NoBF product. The medians for HighBF, MedBF, and LowBF mean that half of the participants converted less than 19% GR to SR in the mouth. HighBP distribution, on the other hand, is negatively skewed with half the participants converting 35 to 60% of GR (median = 0.357) to SR.

> Kf and keni represent the rate of formation of sulforaphane and erucin and nitriles, respectively, by gut bacteria in the large intestine. These parameters were estimated for the MedBF, LowBF and NoBF products but not for the High myrosinase products. Kf was fixed at 0.0033 min−<sup>1</sup> and keni was fixed at 0.0015 min−<sup>1</sup> for both products. All distributions are positively skewed indicating that at 50% of the participant metabolize GR to SR slower than 0.042 min−<sup>1</sup> for MedBF, 0.012 min−<sup>1</sup> for LowBF, and 0.003 min−<sup>1</sup> for NoBF (Table 5); and they metabolize GR to erucin and nitrile slower than 0.017 min−<sup>1</sup> for MedBF, 0.033 min−<sup>1</sup> for LowBF, and 0.003 min−<sup>1</sup> for NoBF (Table 6).

The results of other parameters are discussed in Appendix B.


**Table 6.** keni estimation results (horizontal red bar: median, blue box second and third quartile, whiskers first and fourth quartiles, +: outliers).

\* Some or all outliers excluded for better visual presentation and comparison of boxplots. Outliers are values more than 1.5 times the IQR. See Supplementary Files for plotted outliers.

#### *3.5. Certainty of Parameter Estimates*

For each participant and broccoli product, 90% confidence intervals were calculated to determine the range of parameters that are likely to include the parameter estimates. The confidence intervals were very large and are therefore not a good measure of certainty in the parameter estimates. The percentage of participants that had intervals larger than 30% on either side of the estimated parameter, was 62% of participants for HighBP, 53% for HighBF, 96% for MedBF, 98% for LowBF, and 100% for NoBF. The large confidence intervals were due to the limited data of each participant.

#### *3.6. Goodness of Fit*

A comparison between the experimental and the predicted values are shown in Figure 5.

**Figure 5.** Parity plot of High MYR Broccoli Powder (**A**), High MYR Broccoli Florets (**B**), Medium MYR Broccoli Florets (**C**), Low MYR Broccoli Florets (**D**), and No MYR Broccoli Florets (**E**) with lumped participant data. Participants are designated by letters.

The best fits are for the HighBP product (*R*<sup>2</sup> = 0.95), followed by LowBF (*R*<sup>2</sup> = 0.93), HighBF (*R*<sup>2</sup> = 0.92), MedBF (*R*<sup>2</sup> = 0.85), and lastly NoBF (*R*<sup>2</sup> = 0.78). Most predicted values for NoBF are higher than their corresponding experimental values (Figure 5E). Of the five broccoli products NoBF was the most difficult to fit due to the narrow peaks and wider base for most of the excretion curves compared to the other products (see NoBF simulation results in Figure S1: Model Fittings Results for All Participants in the supplementary files.). A noticeable feature of the model for the HighBF, MedBF, LowBF, and NoBF products (Figures 5B–E and S1), is that the model solutions for some participants (e.g., participants e, l, n) approached excretion rates of 0 faster than the data points. The fit on the tail end of the data points, as well as the general fits for NoBF, may be improved by updating the model to include future physiological information on GR gut conversions, ITC absorption, distribution, and elimination.

#### **4. Discussion**

#### *4.1. Sensitivity Analysis and Selection of Parameters*

Sensitivity analysis for each product type was conducted to understand how changing one parameter and keeping the other constant would affect the model simulation output. Parameters to be estimated in model fittings were selected based on the results of the sensitivity analysis and preliminary model fittings. Given the low amount of data (6–15 data points) per participant, this procedure helped reduce the number of parameters fitted to the most necessary, thereby increasing the degrees of freedom.

Parameters that affected model output but were excluded from final fittings were ke, MYR, Cgl0, and ITC0. The model was sensitive to ke for the HighBP, HighBF, MedBF, and LowBF products. However, ke was not included in the fittings for HighBP and BF due to the poor model convergence. For all products, MYR and Cgl0 were fixed at the concentrations calculated from the Oliviero study (Table 1 and Matlab codes in the Supplementary Files). The initial concentration of SR (ITC0) calculated for HighBP, HighBF, and NoBF were kept at the values derived from the Oliviero study, while for MedBF and LowBF, ITC0 was reduced in order to get a good fit (Table 1). The reduction in ITC0 may be explained by the high variability in the measured amounts of SR in the broccoli products. Oliviero et al. [35] measured SR in triplicates and the standard deviation for MedBF was 28.4% of the average, while for LowBF, it was 64.3%; the standard deviations were the two highest out of the five products. ITC0 values used in the model were averages based on a sample size of three. Therefore, it is possible that most participants were consuming less SR initially. With the lower ITC0 values, the fittings for MedBF and LowBF had lower means squared error (MSE) values and better fits visually. As with ke, the model function appeared to be stuck in a local minimum when ITC0 was used at its original values. A better approach for future modeling might be to fit MYR, Cgl0, and ITC0 for all participants simultaneously. With the estimated MYR, Cgl0, and ITC0 kept constant, the remaining parameters sensitive to the model would be fitted.

More data are needed per participant, >15 data points, to get higher degrees of freedom. The number of data points per participant becomes important the greater the number of parameters being estimated. Obtaining a specified number of urine excretion data points from an intervention study is understandably not easy to achieve. Collecting blood samples during the intervention period in addition to urine samples would provide additional data for the model fittings that may improve parameter estimations as well as increase the number of parameters being estimated.

#### *4.2. Model Fittings*

The model succeeded in fitting the data of the various broccoli products well within the experimental error for most of the participants (Figures 4 and S1). It also described the myrosinase mediated conversion of GR in the mouth and the microbial conversion of GR in the gut well. The quick appearance (within 2–3 h after consumption) of the single excretion peaks for HighBP and HighBF products, represents excretion rates of SR that were initially consumed or formed in the mouth, and were absorbed in the small intestine. The NoBF product, which had an insignificant amount of myrosinase, had single peaks that appeared much later (7–8 h after consumption). The NoBF peaks represent excretion rates of SR that was formed by gut bacteria and absorbed only in the large intestine. MedBF and LowBF curves tend to have two peaks. The first peak representing absorption from the small intestine and the second peak from the large intestine. This excretion pattern due to differences in myrosinase content have been observed by other authors [35,36,50,51].

#### *4.3. Bioavailability of Sulforaphane*

This compartment model is a good predictor of bioavailability. Other researchers observed that using a compartmental absorption and transit (CAT) model was better at predicting bioavailability than a single compartment model [52].

#### *4.4. Mouth and Gut Parameter Estimations*

The parameter estimates clearly show the variability between individuals in the amount of GR converted to SR during mastication and in their gut bacteria activity. The distributions (Tables 4–6) also vary across product categories. The median GR conversion ratio, SRR, is higher for HighBP compared to the HighBF, MedBF, and LowBF broccoli products. 50% of the HighBP participants have conversion ratios similar to what Sarvan et al. [20] found in their study for 0.5 min and 1 min cooked broccoli, where after chewing, 41% and 60% GR was converted to SR. The fact that the overall distribution for HighBP is shifted to higher values compared to HighBF demonstrates the importance of the product matrix. Although, both had the same myrosinase content, most of the participants could convert more GR to SR after consuming the powder. General differences in chewing patterns between individuals have been documented [23]. However, more specific studies on the effect of chewing patterns correlated to conversion GR ratios may provide insights into the variations observed between individuals in the parameter distributions.

The variation between products is significant for the gut parameters, kf and keni. The IQR for LowBF and NoBF were expected to be similar since most of the myrosianse was inactivated in both products. However, the NoBF IQR was significantly smaller, which implied that the participants in the Oliviero study had very similar gut bacteria or that their overall bacterial activity was similar. Unfortunately, data on the gut microbial population of the participants was not available to correlate to the parameter estimation results. It is well known that gut bacterial populations differ between individuals and populations [7,53–55]. Forty-seven bacterial species having been identified as having GL metabolizing activities in-vitro [9], but only a few have been investigated for their GL bioconversion mechanisms: *Enterobacter cloacae* [56], *Lactobacillus agilis R16* and *Escherichia coli VL8* [57], *Bacteroides thetaiotaomicorn* [58].

#### **5. Conclusions**

A physiological-based multicompartment digestion and absorption model, was developed to describe the kinetics and bioavailability of sulforaphane (SR) from broccoli, and to evaluate how the derived parameters are impacted by inter-individual variation. The model included reactions during digestion in the mouth and gut. It successfully fit participant data and was able to describe bioavailability of SR very well as there were minimal differences between the predicted and experimentally bioavailability. The parameters estimated during the model fitting represented physiological aspects of the digestion process, which were also sources of inter-individual variability. For the digestion of broccoli, the parameters that represented sources for variation between individuals were SRR, the ratio of GR converted to SR during mastication, and *kf* and *keni*, the conversion rate constants of GR to SR or other break down products. The inter-individual variability between participants was captured in the variability of some these estimates. However, it was not possible to correlate the variability between participants to specific physical

attributes, such as chewing patterns or predominant gut microbes, as that information for the participants was not available.

The model's predicted values fit the experimental values very well, especially for the high and low myrosinase products. The lower quality of the fit for the no myrosinase product, indicates the need to improve the model's representation of microbial gut conversions. The work completed in this study is a preliminary step in creating a validated model, which, in the future, could be a useful tool in being able to predict the biological effects of SR and possibly other bioactive compounds. A future predictive model has the potential to positively influence the growing field of personalized nutrition.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/foods10112761/s1. Table S1: Compartment Equations, Tables S2–S8: Parameter Box Plots, Figure S1: Model Fittings Results for All Participants, Matlab Codes and Simulink Model

**Author Contributions:** Q.S.: Conceptualization, methodology, software, investigation, writing original draft preparation, visualization. M.D.: Conceptualization, methodology, writing—review and editing, supervision. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** Frank Sommerhage for his assistance with Matlab.

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

#### **Appendix A. Myrosinase Calculations**

The amount of myrosinase (mg MYR/mg Broccoli) was calculated based on the methods of Oliviero et al. [35] and the research of Roman et al. [42] Oliviero et al. used a spectrophotometric method to determine the myrosinase activity in the different broccoli products. Roman et al. investigated two mechanisms for substrate inhibition during the conversion of GLs to ITCs by modeling their experimental results (enzyme reaction rate vs. sinigrin concentration) using the modified Michaelis–Menten *Ki* = netics with Substrate Inhibition (MMSI) model (Equation (A1)).

$$w = \frac{V\_{\max} \cdot \left[ S \right]}{K\_m + \left[ S \right] + \frac{\left[ S \right]^2}{K\_i}} \tag{A1}$$

*Vmax* (μM/min\*mg MYR) is the maximum rate of the system, *Km* (μM) is the Michaelis-Menten constant, *Ki* (μM) is the inhibition constant, *S* is the substrate concentration, and *v* is the reaction rate. This physiological-based model assumes substrate inhibition occurs in the catalytic site.

*Part I. Steps to determining myrosinase content in each broccoli product.*

1. Experimental data points were extracted from the reaction rate vs. sinigrin concentration graph [42].


2. The MMSI equation (Equation (A1)) was used to model the data. The parameters, *Vmax*, *Km*, and *Ki*, were solved by minimizing the sum of squares difference using Excel's SUMXMY2 function and solver.


3. The specific enzyme activity of Myrosinase was calculated using information from the materials Oliviero et al. used to determine activity and the parameter estimates from step 2.

Concentration of sinigrin in reaction mixture used by Oliviero et al. to determine myrosinase activity was calculated as follows:

$$\frac{30\text{ mg}}{\text{mL}} \times \frac{\text{mol}}{397.5\text{ g}} \times \frac{\text{g}}{1000\text{ mg}} \times 10^6 = 75.5\,\frac{\text{umol}}{\text{mL}}$$

$$\begin{array}{l} 75.5\,\frac{\upmu\text{mol}}{\text{mL}} \times \frac{0.05\text{ mL}\,\frac{\text{inigrin solution}}{1.105\text{ mL}}}{1.105\text{ mL}} \times 1000\frac{\text{mL}}{\text{L}}\\ = 3415\,\text{uM}\,\text{Sinigrin in reaction mixture} \end{array}$$

Equation (A1) was used to calculate specific enzyme activity

$$\frac{\left(0.96 \frac{\text{\AA M}}{\text{min} \circ \text{mg M} \text{Y} \text{R}}\right) \times 3415 \text{ }\upmu\text{M}}{86.51 \text{ }\upmu\text{M} + 3415 \text{ }\upmu\text{M} + \frac{\left(3415 \text{ }\upmu\text{M}\right)^2}{\angle 80.05 \text{ }\upmu\text{M}}} = 0.178 \frac{\text{umol}}{\text{mg M} \text{Y} \text{R} \text{ }\upmu\text{m}}$$

4. The Oliviero *MYR* activity (column A below) for each product type was divided by 5000 mg to determine the μmol MYR/mg broccoli \* min (column B). Column B was divided by the specific enzyme activity (0.178 μmol/mg MYR\*min) to obtain the mg MYR/mg Broccoli (column C).


*Part II. Estimating Vmax, Km, and Ki, for the myrosinase conversion of GR in the mouth.*

1. Experimental data points were extracted from the reaction rate vs. glucoraphanin concentration graph [42].


2. The MMSI equation (Equation (A1)) was used to model the data. The parameters, *Vmax*, *Km*, and *Ki*, were solved by minimizing the sum of squares difference using Excel's SUMXMY2 function and solver.


#### **Appendix B. Parameter Estimates Results and Discussion for KSH, KtSI, KtLI, and Ke**

Stomach emptying was described as biphasic by having two stomach compartments. The first stomach compartment was for the delay caused by the disintegration of large food particles before it is emptied into the duodenum. Delayed emptying has been observed for wheat products [59] and other solid foods [31,39]. Since participants ate a 30-g raisin bun, right after ingesting 5 g of broccoli florets, a compartment was included to represent disintegration of both food products. KSH, the rate constant from the first stomach compartment to the second, is the inverse of the time required for food disintegration. Parameter estimates for KSH are shown in Table A1. Excluding the outliers, HighBP (IQR = 0.008 min<sup>−</sup>1) and HighBF (IQR = 0.004 min−1), have normal distributions compared to Med, Low and NoBF products. Comparatively, the median values of HighBP (0.009 min−1) and High BF (0.007 min−1) are smaller than the medians of the MedBF (0.017 min−1), LowBF (0.016 min−1), and NoBF (0.014 min−1) products. These values imply that most participants who consumed MedBF, LowBF and NoBF, required less time for stomach disintegration of the broccoli and the raisin bun than HighBF and HighBP. Since all participants consumed the same amounts of florets and raisin bun, differences in food disintegration were only expected between individuals due to physiological differences. However, the median and distribution differences between the High MYR products and the remaining three were not expected. Kong et al. [60] observed longer disintegration times in their stomach model for raw and 2 min cooked carrots versus their 6 min cooked carrots. Since the high MYR broccoli florets, HighBF, were less heat processed, a harder product texture may be the cause of its longer disintegration time. However, this does not explain why the powdered product, HighBP does not have shorter disintegration times compared to the MedBF, LowBF, and NoBF. It is possible that the raisin bun consumed at the time participants consumed HighBP was harder in texture than when the other broccoli products were consumed.

The transit rate constants for each compartment of the small and large intestine are ktSI and ktLI, respectively. The rate constants were determined based on the measured time it takes for food contents to transit through the small and large intestines. Excluding outliers, the distribution of ktSI parameters for MedBF (IQR = 0.014 min−1), LowBF (IQR = 0.013 min−1), and NoBF (IQR = 0.009 min−1) are less variable compared to the High BP (IQR = 0.098 min−1) and High BF (IQR = 0.057 min−1) broccoli products (Table A2). Except for MedBF (outliers excluded), the parameter distributions for the four products are positively skewed. The median values for all broccoli products are similar, around 0.02 min−1, which is within the range of literature cited values—0.01228 min−1– 0.2333 min−<sup>1</sup> [26]. Across all products transit through the small intestine compartments takes 50 min (1/0.02 min−1) or longer for 50% of the participants. For each product, the percentage of ktSI parameter estimates that fall within the literature cited range are: 100% NoBF, 93% LowBF, 71% MedBF, 67% HighBF, and 73% HighBP. KtLI was not fitted for the HighBP and HighBF; it was fixed at 0.003 min−<sup>1</sup> for both products (Table A3) because sensitivity analysis showed this parameter did not have an influence on the model's output. The medians for MedBF (0.280 min<sup>−</sup>1), LowBF (0.220 min−1), and NoBF (0.031 min−1) are larger than literature ranges (0.002 min−1–0.003 min−1) for gut intestinal transit [61]. The parameter distributions are variable especially NoBF in which the interquartile range is 560% of the median. Furthermore, the distribution for NoBF is positively skewed, indicating that 50% of the participants had colon transit times longer than 32 min (1/0.031 min<sup>−</sup>1)

per compartment or 3.8 h for the colon. Based on the skewness, the participants with the long transit times are not as variable as the 50% of participants with transit times shorter than 32 min per compartment. Nine out of the 15 participants in the Oliviero study were women and it is known that women have longer intestinal transit times compared to men [62]. The gender of the participants with their corresponding data were not provided, so it is impossible to correlate transit times to gender.

**Table A1.** KSH estimation results.

\* Some or all outliers excluded for better visual presentation and comparison of boxplots. Outliers (+) are values more than 1.5 times the IQR. See supplementary tables for plotted outliers.

> The rate constant for sulforaphane elimination from the blood, ke, was not fitted for HighBP and HighBF products, but rather fixed at 0.024 min−<sup>1</sup> (Table A4). The median rates of elimination for MedBF, LowBF and NoBF are 0.020 min<sup>−</sup>1, 0.025 min−1, and 0.017 min<sup>−</sup>1, respectively. The IQR is largest for MedBF (0.043 min<sup>−</sup>1), followed by LowBF (0.028 min−1) and NoBF (0.01 min−1). Vermeulen et al. [36] fitted the plasma concentration data to one compartmental model and determined the elimination half-lives (t0.5) of SR for cooked and raw broccoli to be 4.6 h and 3.8 h, respectively. From half-lives (t0.5 = 0.693/ke), the elimination rate constants were calculated to be 0.002511 min−<sup>1</sup> (cooked) and 0.003039 min−<sup>1</sup> (raw) [36]. Vermeulen's values are lower than all estimated parameters of MedBF, LowBF, and NoBF products. The difference is most likely due to the method used in determining ke, because Vermeulen et al. uses a more empirical model than the one described in this thesis.

**Table A2.** ktSI estimation results.

\* Some or all outliers excluded for better visual presentation and comparison of boxplots. Outliers (+) are values more than 1.5 times the IQR. See supplementary tables for plotted outliers.


**Table A3.** ktLI estimation results.

\* Some or all outliers excluded for better visual presentation and comparison of boxplots. Outliers (+) are values more than 1.5 times the IQR. See supplementary tables for plotted outliers.


**Table A4.** ke estimation results.

\* Some or all outliers excluded for better visual presentation and comparison of boxplots. Outliers (+) are values more than 1.5 times the IQR. See supplementary tables for plotted outliers.

#### **References**


## *Article* **Immunomodulating Effect of the Consumption of Watercress** *(Nasturtium officinale)* **on Exercise-Induced Inflammation in Humans**

**Hendrik Schulze 1, Johann Hornbacher 2, Paulina Wasserfurth 1, Thomas Reichel 3, Thorben Günther 4, Ulrich Krings 4, Karsten Krüger 3, Andreas Hahn 1, Jutta Papenbrock <sup>2</sup> and Jan P. Schuchardt 1,\***


**Abstract:** The vegetable watercress (*Nasturtium officinale* R.Br.) is, besides being a generally nutritious food, a rich source of glucosinolates. Gluconasturtiin, the predominant glucosinolate in watercress, has been shown to have several health beneficial properties through its bioactive breakdown product phenethyl isothiocyanate. Little is known about the immunoregulatory effects of watercress. Moreover, anti-inflammatory effects have mostly been shown in in vitro or in animal models. Hence, we conducted a proof-of-concept study to investigate the effects of watercress on the human immune system. In a cross-over intervention study, 19 healthy subjects (26.5 ± 4.3 years; 14 males, 5 females) were given a single dose (85 g) of fresh self-grown watercress or a control meal. Two hours later, a 30 min high-intensity workout was conducted to promote exercise-induced inflammation. Blood samples were drawn before, 5 min after, and 3 h after the exercise unit. Inflammatory blood markers (IL-1β, IL-6, IL-10, TNF-α, MCP-1, MMP-9) were analyzed in whole blood cultures after ex vivo immune cell stimulation via lipopolysaccharides. A mild pro-inflammatory reaction was observed after watercress consumption indicated by an increase in IL-1β, IL-6, and TNF-α, whereas the immune response was more pronounced for both pro-inflammatory and anti-inflammatory markers (IL-1β, IL-6, IL-10, TNF-α) after the exercise unit compared to the control meal. During the recovery phase, watercress consumption led to a stronger anti-inflammatory downregulation of the pro-inflammatory cytokines IL-6 and TNF-α. In conclusion, we propose that watercress causes a stronger pro-inflammatory response and anti-inflammatory counter-regulation during and after exercise. The clinical relevance of these changes should be verified in future studies.

**Keywords:** watercress; cruciferous vegetables; glucosinolates; gluconasturtiin; anti-inflammatory; pro-inflammatory

#### **1. Introduction**

With the revival of domestic greens watercress *(Nasturtium officinale* R.Br.), a member of the Brassicaceae family, gains a growing interest in science. The semi-aquatic plant species native to Europe and Asia is often consumed as a salad or garnish, or as part of a soup, especially in the Mediterranean kitchen. It is valued for its high nutrient density caused by a low energy content and high amounts of vitamins (B1, B2, B3, B6, C, E), minerals (calcium, iron), and phytochemicals (polyphenols, terpenes) [1–3]. Like all members of the

**Citation:** Schulze, H.; Hornbacher, J.; Wasserfurth, P.; Reichel, T.; Günther, T.; Krings, U.; Krüger, K.; Hahn, A.; Papenbrock, J.; Schuchardt, J.P. Immunomodulating Effect of the Consumption of Watercress *(Nasturtium officinale)* on Exercise-Induced Inflammation in Humans. *Foods* **2021**, *10*, 1774. https://doi.org/10.3390/ foods10081774

Academic Editors: Franziska S. Hanschen, Sascha Rohn and Laura Jaime

Received: 21 May 2021 Accepted: 28 July 2021 Published: 30 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Brassicaceae family, watercress contains mustard oil glycosides or glucosinolates (GLS), of which gluconasturtiin is the predominant GLS in watercress. As a precursor, it is converted into the bioactive compound phenethyl isothiocyanate (PEITC) upon tissue disruption due to the action of the thioglucosidase myrosinase.

Several studies investigated the health beneficial effects of watercress and PEITC including antioxidative, anti-inflammatory, antidiabetic, anti-allergic, antibacterial, hypolipemic, cardioprotective, and anticancer effects (reviewed in [2]). While most of these effects have been observed in vitro or in animal studies, only a few human intervention studies with watercress have been carried out. Moreover, human intervention studies that administered watercress have mainly focused on antioxidative [4,5] and anticancer effects [6]. The influence of watercress on the immune system, in particular anti-inflammatory activity, has barely been investigated in human studies thus far. There has yet been no confirmation that watercress and its ingredients gluconasturtiin/PEITC act in a similar way in humans compared to effects observed in vitro, namely, by inhibiting the pro-inflammatory nuclear factor kappa B (NfκB) pathway [7]. Because the NfκB pathway can be stimulated directly by reactive oxygen species (ROS) [8] or indirectly by the ROS-dependent heat shock response [9,10], antioxidants might attenuate the exercise-induced inflammation [11]. As a consequence, it is necessary to determine the levels of antioxidants and their capacity in watercress. Moreover, it remains unknown as to whether other gluconasturtiin metabolites are formed in vivo and how they contribute to an antioxidative effect such as that indicated for benzenepropanenitrile [12].

We performed a pilot study with four subjects to examine the effect of a single dose of fresh watercress on various biomarkers of exercise-induced inflammation [13]. On the basis of the results of that previous study, where we observed indications for anti-inflammatory effects, we conducted this follow-up study with a greater number of subjects to further characterize the inflammatory response of watercress consumption. After consuming 85 g of fresh watercress, untrained subjects had to complete a high-intensity workout to induce a pro-inflammatory condition. Inflammatory blood markers (IL-1β, IL-6, IL-10, TNF-α, MCP-1, MMP-9) were analyzed in whole blood cultures after ex vivo immune cell stimulation via lipopolysaccharides (LPS).

#### **2. Materials and Methods**

#### *2.1. Plant Material*

The administered watercress *(Nasturtium officinale)* was obtained from the Institute of Botany, Leibniz University Hannover. Cuttings were taken for propagation and cultivated in a hydroponic greenhouse system using a Hoagland solution. After 8 weeks, the plant material was harvested freshly on a daily basis. Following a 45 min wet transport, it was cut as little as necessary for consumption.

#### *2.2. Analysis of Glucosinolates by HPLC/LC–MS*

GLS were analyzed by HPLC–UV according to Hornbacher et al. [14]. The GLS content of the watercress samples were measured in triplicate. All standard substances were checked for identity. For the identification of the GSL in *N. officinale*, samples were analyzed by liquid chromatography–mass spectrometry (LC–MS). A volume of 10 μL was injected into the HPLC system (Shimadzu, Darmstadt, Germany) and separated on a Knauer Vertex Plus column (250 × 4 mm, 5 μm particle size, packing material ProntoSIL 120-5 C18-H) equipped with a pre-column (Knauer, Berlin, Germany). A water (solvent A)–methanol (solvent B), both containing 2 mM ammonium acetate, gradient was used with a flow rate of 0.8 mL/min at 30 ◦C. For measuring the samples, the following gradient was used: 10–90% B for 35 min, 90% for 2 min, 90–10% B for 1 min, and 10% B for 2 min. Detection of the spectra in the range 190–800 nm was performed with a diode array detector (SPD-M20A, Shimadzu, Darmstadt, Germany). The HPLC system was coupled to an AB Sciex TripleTOF mass spectrometer (AB Sciex TripleTOF 4600, Canby, OR, USA). At a temperature of 600 ◦C and an ion spray voltage floating of −4500 V, the negative

electrospray ionization (ESI) was performed. For the ion source gas one and two 50 psi were used and for the curtain gas 35 psi. In the range of 100–1500 Da in the TOF range, the mass spectra as well as the MS/MS spectra from 150–1500 Da at a collision energy of −10 eV were recorded. Peaks were identified by analyzing the characteristic mass fragments of ds-4-methoxyglucobrassicin (195, 398, 433, 795) and ds-glucoarabishirsutain (195, 382, 417, 763). Due to lack of standards of the GSLs fractions of the measured samples were collected in a fraction collector (FRC-10A Shimadzu, Darmstadt, Germany), dried in a vacuum centrifuge, and dissolved in 300 μL ultrapure water. The retention time for every GSL was determined by measuring either the collected fraction or the authentic standard (Phytolab, Vestenbergsgreuth, Germany) with the HPLC system, as described above.

#### *2.3. Measurement of Antioxidant Contents and Antioxidant Capacity*

The measurements of carotenoid, total phenol and total flavonoid contents, as well as the measurement of the oxygen radical absorbance capacity (ORAC), were performed according to Boestfleisch et al. [15].

Tocopherol contents were analyzed according to Cruz et al. [16] with modifications. Small portions of finely ground fresh sample (1.0 g) were weighed accurately into amber glass vials containing ascorbic acid (50 mg), butylated hydroxytoluene (1 mg), and internal standard (1 μg *δ*-tocopherol). Samples were homogenized with methanol (2 mL) by vortex mixing for 1 min. Then, dichloromethane (4 mL) was added and vortex-mixed for 1 min. Subsequently, 0.9% (*w*/*v*) NaCl (1 mL) was added, the mixture was homogenized (1 min) and centrifuged (3 min, 14,000× *g*), and the clear lower layer was transferred to an amber flask. Extraction was repeated twice with dichloromethane. The extracts were combined and vacuum-dried in a vacuum centrifuge (Eppendorf, Hamburg, Germany) at 25 ◦C. The extract was recovered with 1 mL of *n*-hexane and anhydrous sodium sulfate was added (around 100 mg). After an additional centrifugation (5 min, 14,000× *g*), the supernatant was analyzed immediately. Analysis was performed with an HPLC system equipped with a Nucleodur C18 column (250 mm × 4.6 mm; Macherey-Nagel, Düren, Germany). Tocopherols were separated with an isocratic gradient consisting of 90% *n*-hexane and 10% diethyl ether at room temperature and a flow rate of 1 mL/min. Analytes were monitored with a fluorescence detector (Shimadzu, Duisburg, Germany). Excitation was performed at 289 nm and fluorescence of analytes was analyzed at 331 nm.

#### *2.4. Human Study Design and Subjects*

An overview of the timeline of the study and the interventions in particular is shown in Figure 1. The inclusion criteria of the cross-over study were age between 18 and 35 years, BMI between 18 and 30 kg/m2, and less than two hours of moderate exercise per week, classifying these participants as untrained. For the questionnaire-based assessment of the training status, we factored in leisure time physical activities such as jogging or weightlifting, as well as daily non-athletic exertions such as movement by foot or bike. The exclusion criteria were cardiovascular or metabolic disease, smoking, pregnancy, drug or alcohol dependency, concurrent participation in another clinical trial or in another study within the last 30 days, and intake of antioxidative or antiphlogistic medicine or dietary supplements.

**Figure 1.** Flow diagram showing the timeline of the study.

As part of a run-in/washout phase, subjects refrained from consuming foods rich in polyphenols, vitamin C and E (mainly berries, nuts, and vegetables of the Brassicaceae family) seven days before each examination. To ensure compliance, participants received written instructions and a list of foods that should not be consumed. On the basis of the consumed amount in similar studies [4,5,13], subjects ate a single dose of 85 g of fresh watercress accompanied by a standard breakfast (two buns, cream cheese or oat spread, yoghurt or balsamic dressing). In the control group, 85 g of iceberg lettuce was administered instead. Two hours after consumption, the first blood sample (t0) was taken. Immediately after blood draw, the subjects completed a 30 min high-intensity endurance workout on echo bikes within a parameter range of 80–92% HRmax, 120–145 W, and a final rating of perceived exertion of 17.8. Throughout the workout, heart rates were recorded using a heart rate monitor watch with a Bluetooth heart rate sensor chest strap (RC 14.11, Sigma-Elektro, Neustadt, Germany). Additional blood samples were taken 5 min (t1) and 3 h (t2) after the end of the exercise. The subjects were served a lunch in the meantime consisting of a potato soup. All blood samples were obtained by venipuncture of an arm vein using Multifly needles (Sarstedt, Nürnbrecht, Germany) into heparin plasma monovettes (Sarstedt, Nürnbrecht, Germany). To assure that subjects did not enter the examination with elevated inflammatory markers due to an infection, we determined Creactive protein in serum using serum monovettes (Sarstedt, Nürnbrecht, Germany). Both interventions (control and watercress) were conducted identically with an intermediary washout phase of 7 days.

The study was carried out following the rules of the Declaration of Helsinki and was approved by the Ethics Committee at the Medical Chamber of Lower Saxony (30/37/2020, Hannover, Germany, 10/2020).

#### *2.5. Measurement of Inflammatory Markers by Bio-Plex Multiplex Immunoassay*

The freshly drawn blood samples were stimulated ex vivo in whole blood cultures via lipopolysaccharides (LPS). Therefore, samples were immediately diluted 1:5 with the cell culture medium RPMI 1640 including 20 mmol HEPES and L-glutamine (Sigma-Aldrich, Hamburg, Germany) and added antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin; Sigma-Aldrich, Hamburg, Germany). The samples were seeded into 12-well

microtiter plates and mixed with 10 ng/mL (final concentration) LPS from *Escherichia coli* (Sigma-Aldrich, Hamburg, Germany). The plates were incubated for 24 h at 37 ◦C without a CO2 application. The used HEPES buffer is able to stabilize the pH over 24 h. The supernatants were frozen at −80 ◦C until analysis.

The levels of the inflammatory markers (IL-1β, IL-6, IL-10, TNF-α, MCP-1, MMP-9) in whole blood culture supernatant were simultaneously determined using a human Magnetic Luminex Assay (Bio-Techne, Abingdon, Oxon, UK) and a Magpix Luminex instrument (Luminex Corp, Austin, TX, USA).

#### *2.6. Data Analysis and Statistical Methods*

Data are presented as the means ± standard deviation. All variables were tested for normal distribution by Shapiro–Wilk test. In the case of not normally distributed data, a suitable transformation was applied, and parametric tests were used. Differences among inflammatory markers were analyzed using ANOVA with repeated measures. In addition, groups were compared using a *t*-test for dependent means. To calculate correlations, we utilized Pearson correlation (parametric data) and Spearman's rho correlation (nonparametric data). Statistical significance was regarded as values of *p* ≤ 0.05. Analyses were conducted using Infostat (version 2012; University of Córdoba, Córdoba, Argentine) and SPSS (version 27; SPSS Inc., Chicago, IL, USA).

#### **3. Results**

Of the 21 recruited subjects, 19 completed the study (Table 1). Incomplete study data were excluded from statistical analyses. One subject failed to participate because of illness. Another subject showed increased serum levels of C-reactive protein (>0.5 mg/L), indicating an elevated systemic inflammation. No health- or workout-related incidents occurred during the study, with the exception of one subject taking a two-minute break from the exercise due to total exhaustion.

**Table 1.** Characterization of the study population.


BMI = body mass index, WHR = waist-to-hip ratio.

#### *3.1. Plant Material*

3.1.1. Levels of Glucosinolates

Gluconasturtiin was recognized as the predominant GLS in watercress with a fraction of 90.5 ± 1.1% of the total GLS content. In addition, minor amounts of glucoarabishirsutain (5.3 ± 0.8%), glucobrassicin (1.4 ± 0.4%), neoglucobrassicin (1.4 ± 0.3%), and 4-methoxyglucobrassicin (1.4 ± 0.4%) were found (Figure 2, Table A1). In the course of the study, GLS contents varied notably with the most impactful difference in gluconasturtiin of 36% at day 3 compared to the previous day. This resulted in a difference of 34% of the total GLS content while maintaining the partial composition of the GLS profile.

**Figure 2.** Mean concentration of different glucosinolates (GLS) (nmol/g DW) in watercress at different sampling times. The standard deviation represents the values for three technical replicates. Analysis of variance (ANOVA) was performed with Infostat. Means with a common letter are not significantly different.

#### 3.1.2. Levels of Antioxidants and Antioxidant Capacity

Contents of flavonoids as well as ascorbic acid in analyzed watercress were similar at all sampling days (Table A2). Contents of total phenols were slightly higher in samples taken at day 3 and day 4, whereas carotenoid contents were slightly higher at day 2 and day 4. Tocopherol contents as well as ORAC were similar at all sampling days.

#### *3.2. Levels of Inflammatory Blood Markers*

Statistical analysis using the ANOVA with repeated measures showed significant differences between the sampling times for all analyzed parameters (*p* ≤ 0.001). Hence, the acute exercise affected concentrations of all measured inflammatory markers. In addition, the same analysis pointed out that the consumption of watercress led to significant differences for the levels of IL-1β, IL-6, and IL-10 across all sampling times (IL-1β (*p* = 0.006), IL-6 (*p* = 0.006), IL-10 (*p* ≤ 0.001)). The remaining parameters showed no significant differences (TNF-α (*p* = 0.382), MCP-1 (*p* = 0.192), MMP-9 (*p* = 0.118)). With regards of the varying GLS levels of the plant material, no correlations with the inflammatory markers were found.

Significantly higher concentrations of the pro-inflammatory cytokines IL-1β (16%), IL-6 (33%), and TNF-α (30%), as well as the enzyme MMP-9 (22%), were observed in the watercress group compared to the control group two hours after the watercress consumption (t0) (Figure 3, Table A3). Upon exercise-stimulation (t1), the control group showed a significant rise in all inflammatory markers, specifically IL-6 (33%), IL-10 (47%), MCP-1 (53%), and MMP-9 (53%) with IL-1β and TNF-α showing no reaction. Compared to the control breakfast, levels of IL-1β, IL-6, IL-10, and TNF-α were significantly higher after the watercress breakfast (31%, 32%, 51%, and 23%, respectively). To determine if the consumption of watercress resulted in a stronger increase of the cytokines regardless of the pre-exercise levels, we compared the differences (t1 − t0). Thereby, a significant stronger upregulation of the anti-inflammatory cytokine IL-10 was found. After the recovery phase (t2), all inflammatory markers except MMP-9 decreased. A comparison of the differences (t2 − t1) between the watercress and the control group revealed a significantly stronger downregulation of the pro-inflammatory cytokines IL-6 and TNF-α and a trend in IL-1β (*p* = 0.062). In the case of TNF-α, the level at t2 was even lower compared to the pre-exercise state of the control group. This post-exercise downregulation can also be observed in the ratio of the anti-inflammatory IL-10 and the pro-inflammatory IL-1β (Figure 4). The watercress consumption influenced the ratio 3 h after exercise towards the anti-inflammatory reaction.

**Figure 3.** Effect of acute watercress consumption on blood markers of inflammation (IL-1β, IL-6, IL-10, TNF-α, MCP-1, MMP-9) in ex vivo LPS-stimulated whole blood cultures after high-intensity workout in untrained subjects. t0, pre-exercise; t1, 5 min post-exercise; t2, 3 h post-exercise. Analysis of variance (ANOVA) was performed with SPSS. \* Significant difference between control and watercress. # Significant difference to previous sampling point.

**Figure 4.** Effect of acute watercress consumption on the ratio of IL-10 (anti-inflammatory) and IL-1β (pro-inflammatory). t0, pre-exercise; t1, 5 min post-exercise; t1, 3 h post-exercise. Analysis of variance (ANOVA) was performed with SPSS. \* Significant difference between control and watercress.

#### **4. Discussion**

Many food compounds are known to assure the maintenance of the immune system or improve its performance [17]. By interacting with ROS or affecting cytokine biology, their intake can modulate immune function. A wide variety of anti-inflammatory nutrients has been investigated thus far. While many act as antioxidants and, hence, indirectly modulate the ROS-dependent NfκB activation [18] such as ascorbic acid, glutathione, or carotenoids, others operate as pro-resolving mediators such as omega-3 fatty acids [19].

A common misconception is that pro-inflammatory processes at all times need to be annihilated to prevent the body from harm. Similar to oxidative stress, pro-inflammatory processes can have both detrimental and beneficial effects to the human body. When inflammatory processes become chronic, they often have negative impacts. Chronic lowgrade inflammation is associated with a wide range of chronic conditions, such as the metabolic syndrome, cardiovascular disease, type 2 diabetes, and non-alcoholic fatty liver disease. However, if the body is able to resolve inflammatory processes, they can provide signals for adaptation in physiological contexts such as sport [20]. In its acute form, it is generally a beneficial procedure, which removes stimuli and initiates the repair system. As the first line of defense, pro-inflammatory cytokines such as IL-1β, IL-6, and TNFα promote the activation and secretion of more cytokines and acute-phase proteins as well as the proliferation and differentiation of T- and B-cells. The time-delayed increase of the cytokine IL-10 has been widely recognized as a suppression of the inflammatory response by its downregulating effects on TNF-α and IL-1 [21–23]. The immune system acts through a fluctuation between a pro-inflammatory response and an anti-inflammatory or inflammation resolving counter-regulation. An increase in the magnitude of the fluctuation could be interpreted as positive for the capacity of the system. Thus, a mild activation of the immune system through dietary components might aid in resolving inflammation by a preceding mobilization. We assume that this immunomodulating effect also applies to watercress and its ingredients.

Upon exercise- and LPS-stimulation, a solid, non-excessive immune reaction was observed in IL-6 and IL-10, with a greater response after the watercress consumption. Interestingly, the subsequent inflammatory counter-regulation of IL-6, TNF-α, and IL-1β was more pronounced in the watercress group, which likely was a result of the higher exercise-induced levels of IL-10. This shift towards an anti-inflammatory response after the watercress consumption was supported by the increased ratio of IL-10 and IL-1β. Our results suggest that watercress intake stimulates the immune system, which might enhance its metabolic capacity. The finding that the consumption of watercress causes an initial proinflammatory reaction followed by a greater exercise-induced response in both pro- and

anti-inflammatory markers is remarkable and has not been observed thus far. We assume that gluconasturtiin, in particular PEITC, is responsible for the observed immunomodulating effect, where PEITC might operate as an activator in T-lymphocytes or macrophages or their receptors and thereby stimulates or sensitizes the cytokine production. It is also conceivable that PEITC activates the heat shock response, which leads to a stimulation of the pro-inflammatory NfκB pathway [24]. Another explanation is that specific components of watercress might interfere with the immunometabolism and thereby stimulate its function. Although the average contents of secondary metabolites besides gluconasturtiin and the ORAC are in the lower range of vegetables, it is likely that the observed effects are mainly caused by the GLS. The extraction procedure for the evaluation of the ORAC uses methanol as organic solvent, which inhibits the hydrolysis of gluconasturtiin to PEITC. However, PEITC would contribute only very little to the overall ORAC, since its capacity to scavenge ROS was reported to be 1.9 μg TE/mg PEITC [25]. The levels of gluconasturtiin in the obtained watercress resemble the results of the pilot study [13] and were in the same range of reported contents for raw watercress [26].

In contrast to our results, PEITC and watercress extracts have thus far been shown to possess only anti-inflammatory and no pro-inflammatory properties [27–31]. The pilot study, on which this investigation is based on, aligns with these observations [13]. In vitro studies presume the mechanism behind this effect of PEITC in the inhibition of the proinflammatory NfκB pathway in macrophages, possibly through the modulation of toll-like receptors [31–33]. Due to the redox-sensitivity of NfκB, the activation of the antioxidative nuclear factor erythroid 2-related factor 2 (Nrf2) pathway by PEITC plays a considerable role in its anti-inflammatory effect [34]. Previous human intervention studies have focused mainly on those antioxidative effects of watercress. They showed that a regular consumption as well as a single dose of watercress reduces the oxidative stress in various biomarkers [4,5]. On the basis of the antioxidative and thereby assumed anti-inflammatory effects, the outcome of this study does not align with the previous literature. Although PEITC is able to promote oxidative stress at very high concentrations, presumably acting as a scavenger for glutathione [35], it remains debatable as to whether a single dose of 85 g of watercress provides the body with the necessary range of concentration. In consequence of the redox-sensitivity of NfκB, an initial pro-oxidative effect after the consumption of watercress could explain the simultaneously increased cytokine levels of IL-1β, IL-6, and TNF-α. The pro-oxidative state would be further enhanced by the exercise, which subsequently results in an even more pronounced pro-inflammatory response in the watercress group and a stronger induction of the counter-regulatory Nrf2-mediated antioxidant response that is observed in the recovery phase.

#### **5. Conclusions**

The course of inflammatory markers after the watercress consumption with initially increasing pro-inflammatory markers and a higher release of an anti-inflammatory marker in the recovery phase has not yet been described in the literature. We interpret this observation with a mild activation of the immune system resulting in a stronger pro-inflammatory reaction, which is more effectively resolved by a powerful anti-inflammatory counterregulation. This thesis and the clinical relevance have to be investigated in future studies. Moreover, cell culture studies are necessary to explore the underlying effects of watercress components on leukocytes. Likewise, it must not be ignored that the immune cells were stimulated ex vivo by LPS. The results and the possible effects on the immune system can therefore not be directly transferred to the situation in vivo. Because watercress is a complex food with considerable amounts of other potentially immunomodulating substances besides PEITC, a causal relationship between the observed effects and gluconasturtiin/PEITC cannot be stated. A clinical trial with isolated gluconasturtiin/PEITC should be conducted in order to confirm the effect and whether its magnitude is influenced by other components. The variable GLS levels of watercress show the need for developing standardized extracts or supplements. Thereby, future clinical trials are provided with a

standardized GLS dosage and can overcome logistic and sensory barriers of administering raw watercress.

**Author Contributions:** Conceptualization, K.K., A.H. and J.P.S.; methodology, J.H., K.K. and J.P.S.; formal analysis, H.S., J.H. and T.R.; investigation, H.S., J.H., P.W., T.R., U.K. and T.G.; resources, J.H., K.K., A.H. and J.P.S.; writing—original draft preparation, H.S. and J.P.S.; writing—review and editing, H.S., J.H., J.P., K.K. and J.P.S.; visualization, H.S.; supervision, J.P. and J.P.S.; project administration, J.P.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee at the Medical Chamber of Lower Saxony (30/37/2020, Hannover, Germany, 10/2020).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Acknowledgments:** We would like to thank Julia Volker for help in the laboratory. The publication of this article was funded by the Open Access Fund of Leibniz University Hannover.

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

#### **Appendix A**

**Table A1.** Mean concentration of different glucosinolates (GLS) (nmol/g DW) in watercress at different sampling times. The standard deviation represents the values for three technical replicates. Analysis of variance (ANOVA) was performed with Infostat. Means with a common letter are not significantly different.


**Table A2.** Mean contents of antioxidative substances (total flavonoids and phenols, ascorbic acid, α- and γ-tocopherol) and antioxidative capacity (ORAC) in watercress at different sampling times. The standard deviation represents the values for three technical replicates. Analysis of variance (ANOVA) was performed with Infostat. Means with a common letter are not significantly different.


CE = catechin equivalents (standard curve was performed with catechin), GAE = gallic acid equivalents (standard curve was performed with gallic acid), ORAC = oxygen radical absorbance capacity, TE = trolox equivalents (standard curve was performed with trolox).


**Table A3.** Effect of acute watercress consumption on blood markers of inflammation (IL-1β, IL-6, IL-10, TNF-α, MCP-1, MMP-9) in ex vivo LPS-stimulated whole blood cultures after high-intensity workout in untrained subjects.

t0, pre-exercise; t1, 5 min post-exercise; t2, 3 h post-exercise. \* Significant difference between control and watercress. # Significant difference to previous sampling point.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Foods* Editorial Office E-mail: foods@mdpi.com www.mdpi.com/journal/foods

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-2976-9