**Seasonal Variation of Glucosinolate Hydrolysis Products in Commercial White and Red Cabbages (***Brassica oleracea* **var.** *capitata***)**

#### **Nicole S. Wermter 1,2, Sascha Rohn 2,3 and Franziska S. Hanschen 1,\***


Received: 26 October 2020; Accepted: 12 November 2020; Published: 17 November 2020

**Abstract:** *Brassica* vegetables contain glucosinolates, which are well-known for their potential to form health-promoting isothiocyanates. Among those crucifers, white and red cabbage are commonly consumed vegetables, exhibiting different glucosinolate and hydrolysis profiles thereof. Regarding the health beneficial effects from these vegetables, more information, especially concerning the seasonal variation of glucosinolate profiles and the formation of their bioactive hydrolysis products in commercial cabbages, is needed. In this study, glucosinolates and glucosinolate hydrolysis product profiles in red and white cabbages from three different food retailers were monitored over six different sampling dates across the selling season in autumn. For the first time, it was shown that, while glucosinolate profiles were similar in each cabbage variety, glucosinolate hydrolysis product profiles and hydrolysis behavior varied considerably over the season. The highest total isothiocyanate concentrations were observed in conventional red (1.66 μmol/g FW) and organic white (0.93 μmol/g FW) cabbages purchased at the first sampling date in September. Here, red cabbage was with up to 1.06 μmol/g FW of 4-(methylsulfinyl)butyl isothiocyanate (sulforaphane), an excellent source for this health-promoting isothiocyanate. Cabbages purchased 11 weeks later in autumn released lower levels of isothiocyanates, but mainly nitriles and epithionitriles. The results indicate that commercial cabbages purchased in early autumn could be healthier options than those purchased later in the year.

**Keywords:** glucosinolate; cabbage; isothiocyanate; epithionitrile; nitrile; *Brassica*; seasonal variation; food retailer

#### **1. Introduction**

With a consumption of 5.2 kg/person and year in 2017/2018, red cabbage (*Brassica oleracea* var. *capitata* f. *rubra*) and white cabbage (*Brassica oleracea* var. *capitata* f. *alba*) are the most consumed Brassicaceae vegetables in Germany, contributing to 5% of total vegetable intake [1]. These vegetables are rich in glucosinolates (GLSs), secondary, sulfurous plant constituents, which are particularly present in vacuoles of plant cells of the Brassicaceae family. The chemical GLS structure is determined by the nature of the side chain, depending on the amino acid inserted during biosynthesis [2,3] and GLSs typically contain a glucose unit, bound with a central carbon atom with nitrogen grouping via a thioether bridge. The carbon atom, in turn, is linked to a sulfate group and an organic aglycone residue, possessing an alkyl, alkenyl, aryl, or indole group [2]. This organic residue is inherent to the GLS, its chemical properties and its flavor, respectively [4,5]. Upon attack by herbivores or

due to cutting or chopping of vegetables rich in GLSs, GLSs, which were previously present in spatially separated cell vacuoles, are hydrolyzed by myrosinase to various herbivore-toxic degradation products [6,7]. Hydrolysis by myrosinase occurs due to enzymatic cleavage of the thioglycoside bond, first resulting in an unstable aglucone (thiohydroximate-*O*-sulfate). The aglucone can then undergo a lossen-like rearrangement to form isothiocyanates (ITCs) or decompose into nitriles and molecular sulfur. Moreover, in the presence of certain proteins such as the epithiospecifier protein (ESP), an aglycone with a terminal double bond favors epithionitrile (ETN) formation [7].

Consumption of ITCs can positively affect human health as they have antimicrobial, antidiabetogenic, chemopreventive, and anticarcinogenic properties [8,9]. Previous studies have shown a positive correlation between ITC uptake and cancer prevention [8,10,11], and especially 4-(methylsulfinyl)butyl ITC (sulforaphane; 4MSOB-ITC) is valued for its anticarcinogenic potential [8]. Simple nitriles and ETNs on the other hand, seem to have less health beneficial effects [12,13]. Studies have elucidated that *Brassica* vegetables, not only ITCs, but also nitriles and ETNs can be the most predominant degradation products [14,15]. Consequently, in order to estimate the health beneficial potential of *Brassica* vegetables, it is of great importance to not only analyze intact GLSs, but also their behavior during hydrolysis.

The natural GLS content in vegetables varies in accordance with genotype, plant developmental step, soil and cultivation conditions, and other ecophysiological influences, but it is also affected by storage [14,16,17]. Moreover, GLS levels vary over the growth season and several studies reported higher GLS levels in *B. oleracea* plants grown in spring compared to plants grown in autumn [17], while in broccoli (*B. oleracea* var. *italica*), GLS levels were higher when grown in the summer season compared to the spring season [18]. Similarly, Nuñez-Gómez et al. (2020) recently reported higher GLS levels in broccoli grown in autumn compared to broccoli grown in the spring season. Moreover, when comparing two spring seasons, the one with less rainy days and higher temperature resulted in higher GLS levels [19]. Experiments performed under controlled conditions indicate that temperature, as well as day length affect GLS biosynthesis in *B. oleracea* in a structure-dependent way [20–22].

In order to predict GLS-based health beneficial properties, more information especially concerning GLS profiles and especially on their hydrolysis behavior in *Brassica* foods available for the consumer, is needed. Possessing a long harvesting period—typically ranging from June to November—red and white cabbages are facing seasonal changes and different ecophysiological influences, which can have a great impact on the GLS profile [16,20] and might also affect the potential to form health-preventive ITC. To date, little is known about how GLS hydrolysis products are affected by different cultivation conditions. No revealing insight has been given on how GLS, and even more importantly, the formation behavior of their hydrolysis products varies in commercial cabbages across the whole season. Therefore, the objective of this study was to monitor the variation of GLS levels and the formation of their bioactive hydrolysis products in commercial red and white cabbage heads obtained from three local retailers in Germany and to link the data with the cultivation practices and post-harvest storage conditions.

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

#### *2.1. Chemicals*

Allyl GLS (Allyl; ≥99%, reference compound), 4-hydroxybenzyl GLS (4OHbenzyl; purity ≥99%, internal standard), and methylene chloride (GC Ultra Grade, solvent) were purchased from Carl Roth GmbH and Co. KG (Karlsruhe, Germany). Allyl ITC (Allyl-ITC; ≥99%, reference compound), benzonitrile (≥99.9%, internal standard), DEAE-Sephadex A-25 (anion exchanger), and the reference compounds 3-butenenitrile (Allyl-CN; ≥98%), 4-pentenenitrile (3But-CN; ≥97%), 3-phenylpropanenitrile (≥99%), were obtained from Sigma-Aldrich Chemie GmbH (Steinheim, Germany). The reference compounds 3-butenyl ITC (3But-ITC; ≥95%) and 4-pentenyl ITC (≥95%) were purchased from TCI Deutschland GmbH (Eschborn, Germany). 3-(Methylsulfinyl)propyl ITC (3MSOP-ITC) and 4-(methylsulfanyl)butyl ITC (4MTB-ITC; ≥98%)

were purchased from Santa Cruz Biotechnology (Heidelberg, Germany). 4-(Methylsulfinyl)butyl ITC (4MSOB-ITC) was purchased from Enzo Life Sciences GmbH (Lörrach, Germany). The ETN 1-cyano-2,3-epithiopropane (CETP; ≥95%) was synthesized by Taros Chemicals GmbH and Co. KG (Dortmund, Germany) and 1-cyano-3,4-epithiobutane (CETB; ≥95%) was synthetized by ASCA GmbH Angewandte Synthesechemie Adlershof (Berlin, Germany). 5-(Methylsulfanyl)pentanenitrile (4MTB-CN) and 5-(methylsulfinyl)pentanenitrile (4MSOB-CN) were purchased from Enamine (SIA Enamine, Latvia, Riga). 3-Butenyl GLS (3But; ≥95%), 2(*R*)-hydroxy-3-butenyl GLS (2OH3But; ≥98%), 4-(methylsulfanyl)butyl GLS (4MTB; ≥98%), 4-(methylsulfinyl)butyl GLS (4MSOB; ≥98%), 3-(methylsulfinyl)propyl GLS (3MSOP; ≥98%), and 2-phenylethyl GLS (2PE; ≥98%) were acquired from Phytolab GmbH and Co. KG, Vestenbergsgreuth, Germany. The solvents methanol (≥99.95%), acetonitrile (LC-MS grade), and arylsulfatase (enzyme) were purchased from Th. Geyer GmbH and Co. KG (Renningen, Germany).

#### *2.2. Plant Material*

Three red cabbages (*Brassica oleracea* convar. *capitata* var. *rubra* L.) and three white cabbages (*Brassica oleracea* convar. *capitata* var. *alba)* were purchased at regular intervals from the same two local conventional supermarkets (CON1, CON2) and from the same organic supermarket (ORG1) in Brandenburg, Germany over a period of 3 months (September–November 2019). The supermarkets were selected with regard to different German food trading companies. The two conventional supermarkets selected belong to the two biggest food trading companies in Germany and the organic supermarket also belongs to a big organic food supermarket chain. The exact sampling dates (S1–S6) are given in Table 1. Additionally, red and white cabbage heads, grown in the field at the Leibniz-Institute of Vegetable and Ornamental Crops (IGZ) in Grossbeeren, Germany, were harvested freshly in order to compare GLSs and their hydrolysis products with commercial cabbages. Therefore, red (cultivar 'Redma RZ F1 ) and white (cultivar 'Dottenfelder Dauer') cabbage seeds were sown (13.06.2019 and 20.06.2019) on loamy soil (pH 7.3) and then grown for 3 months at the IGZ (52◦20 59.0 N 13◦18 57.5 E). The red cabbage was cultivated with 100% of the required nitrogen level and fertilized using calcium ammonium nitrate and Patentkali® (419 kg N/ha). Before cultivation, the field was fertilized once with calcium ammonium nitrate (CAN) and Patentkali® and later fertilized a second time, with CAN only during the cultivation period. In total, 377.78 kg CAN/ha and 40.74 kg Patentkali® were applied for fertilization.

For white cabbage, Aminofert® Vinasse fertilizer (BayWa AG, Munich, Germany) was applied (60 kg N/ha) with 30% of the required nitrogen level. Here, fertilization occurred before the sowing of the seeds on 06 June 2019 and a second time on 25 June 2019 for head formation. The origin and cultivation background of commercial cabbages were investigated by interrogating service staff at supermarkets and by contacting growers. Cabbages sold at CON1 and ORG1 were procured from different farming areas in northern Germany (CON1: Neuenkirchen and Helse (Schleswig-Holstein), Germany; ORG1: Blankensee (Mecklenburg-Western Pomerania), Hedwigenkoog (Schleswig-Holstein), Vierlinden and Seeblick (Brandenburg, Germany), whereas cabbages purchased from CON2 could continuously be procured from the same farming area, but on different fields within a 30 km perimeter in Neuenkirchen, Schleswig-Holstein, Germany) (Table 2). With regard to the cultivars, mainly white cabbage varieties such as "Storema", "Lennox", "Marcello", and "Impala" and red cabbage varieties, especially "Futurima", "Rodima", "Bandolero", and "Klimaro" were cultivated in the Dithmarschen region (Helse, Neuenkirchen) for CON1 and CON2. Cultivars "Rodynda"(red cabbage) and "Dowinda" (white cabbage) were mainly cultivated in Hohennauen, Germany for ORG1. Early cabbage cultivars ("Marcello", "Bandolero") were likely purchased between 04 and 05 September 2019, whereas later cabbage cultivars ("Storema", "Lennox", "Impala", "Futurima", "Klimaro", "Rodynda", "Dowinda") could be purchased between 09 September 2019 and 06 November 2019.


**Table 1.** Overview of the dates of the purchased and harvested red and white cabbage heads.


**Table 2.** Total overview of the German origin, cultivation, and storage conditions of commercial red and white cabbages and red and white cabbages cold storage warehouse. (3) Long-term stored (stored). (iv) 2–6 ◦C for 1–2 days after harvest, then storage at 4–7 ◦C for 1–3 days in cold storage warehouse (max 48 h). (v) 0.1–0.3 cold storage warehouse.

#### *Foods* **2020**, *9*, 1682

 from the field

Long-term stored white cabbages ('Storema", "Lennox", "Impala", "Dowinda") were purchased as of 18 November 2019 and stored at 0.1–0.3 ◦C in warehouses at the wholesaler. Cabbages from CON1 and CON2 were fertilized using a combination of urea, calcium ammonium nitrate, phosphate and potassium, whilst organically cultivated cabbage ORG1 was fertilized using *hair-meal pellets* (200 kg N/ha) or compost (Table 2). According to growers in Blankensee, Neuenkirchen, Vierlinden, and Helse, only hybrid cultivars such as "Storema", "Impala", "Lennox", and "Bandolero" were grown, harvested, and later sold as ripe cabbages to CON1 and CON2, whereas non-hybrid cultivars such as "Rodynda" and "Dowinda" (grown for ORG1) were generally grown in Hedwigenkoog and Hohennauen. According to growers in Helse, seeds were sown from the 16th to the 20th calendar week of 2019, and cabbages were harvested from the 23rd to the 46th calendar week on heavy, sea marsh soil. Similar sowing and harvesting dates also applied for cabbages cultivated in Neuenkirchen, which were harvested between weeks 23 to 45 and also grown on heavy, sea marsh soil. According to growers in Helse, which supplied CON1 with cabbage, cabbage heads grown for the conventional market generally grew slower in the Dithmarschen region and were grown for 150 d in Helse on sea marsh soil (late cabbage). The growth of early cultivars (S1 sampling of CON1 and CON2) was accelerated with non-woven fibre barriers. The alleged storage conditions, according to all growers before the selling period and conditions in the supermarket during selling time, according to salespersons in CON1, CON2, and ORG1 are listed in Table 2, while in Supplemental Table S1 all information collected during this study on the analyzed cabbages are given for each cabbage separately and in more detail.

In order to give a better understanding of how GLSs and their hydrolysis products can differ in commercial cabbages and how they might change within the season, GLSs and their hydrolysis products were additionally monitored by comparing cabbage heads from an organic and two conventional supermarkets (ORG1, CON1, CON2) with cabbage heads grown at the Leibniz-Institute of Vegetable and Ornamental Crops (IGZ), Grossbeeren, Germany. A further aim of this work was to link the results with the common cultivation practices and the alleged storage conditions (Table 2).

#### *2.3. Sample Preparation*

Fresh cabbage heads were chopped in half. Of the two obtained halves, one of the halves was halved again, and two quarters were obtained. Afterwards, one of the obtained quarters was divided into 2–3 strips (weight: 70–180 g, width: 1–1.5 cm) along the middle, and the strips were frozen at −20 ◦C overnight before lyophilization (11 d) and were later ground. The remaining plant material of the same quarter, from which the strips were obtained, was then cut into small pieces of 1 cm width. The chopped plant material was thoroughly mixed by hand, and 15–20 g fresh cabbage was then given into a round bottom glass vessel for homogenization. Afterwards, 15–20 mL of distilled water was added, in order to obtain a 1:1 ratio of plant material and water. Then, samples were homogenized for 1 min at a rate of 20,000 rpm using a mixer (H04, Edmund Bühler GmbH, Tübingen, Germany) and incubated for 1 h at room temperature (22 ◦C).

#### *2.4. Analysis of Glucosinolates*

To determine the profiles and concentrations of GLS in red and white cabbages, 10 mg of lyophilized powder was extracted and GLS was analyzed as their desulfo-form [23]. Briefly, 10 mg of dry plant powder was extracted thrice using 70% of hot methanol in the presence of 0.025 μmol 4-hydroxybenzyl GLS as an internal standard. The extracts were combined and desulfated on a DEAE-Sephadex A-25 ion-exchanger column using aryl sulfatase. Afterwards, desulfo-GLSs were eluted with 1 mL of water and analyzed using an Agilent UHPLC-DAD-ToF-MS system equipped with a Poroshell 120 EC-C18 column (100 × 2.1 mm, 2.7 μm; Agilent Technologies), a gradient of water, and 40% acetonitrile, as described previously [23]. Desulfo-GLSs were quantified at 229 nm via the internal standard and the calibration factor reported in the DIN EN ISO 9167-1 and calculated on this basis for 4-hydroxybenzyl GLS.

#### *2.5. Determination of Glucosinolate Breakdown Products*

For the analysis of GLS hydrolysis products released from red and white cabbage tissue, the protocol described by Hanschen and Schreiner (2017) was followed with small modifications [14]: Briefly, 500 mg of the homogenized fresh samples (containing 50% of water) were weighed into solvent resistant centrifuge tubes. During the first two samplings, 1 g of sample homogenate was used, which might have led to reduced recoveries for nitriles and epithionitriles, due to higher water to solvent ratio. Then, the internal standard benzonitrile (0.2 μmol) was added and GLS hydrolysis products were extracted 3 times using methylene chloride: 2 mL during the first extraction and 1.5 mL of methylene chloride in the second and third extraction. Then, samples were analyzed as described previously [14], except that in the present study an Agilent J&W VF-5ms GC-MS column (30 m × 0.25 mm × 0.25 μm) coupled to a 10 m EZ-Guard P/N:CP9013 column was used for analyte separation.

#### *2.6. Statistical Analysis*

To investigate differences between different sampling dates (S1–S6), means were compared using ANOVA and Tukey's HSD test and STATISTICA version 13.5.0.17 software (TIBCO Software Inc., Palo Alto, CA, USA) with a significance level of *p* ≤ 0.05. All analyses were carried out in triplicate by analyzing three biological replicates.

#### **3. Results**

Samples (three cabbage heads) from each of the three retailers (conventional supermarkets CON1, CON2, and organic supermarket ORG1) were collected between September and November 2019 every 2 to 3 weeks, summing up to a total of six sampling dates (S1–S6). GLS-profiles and the corresponding GLS hydrolysis products were monitored over the six sampling periods and additionally compared to fresh samples (four red and white cabbage heads) harvested from a field at the IGZ in Grossbeeren, Germany. Most analyzed cabbage heads differed in their regional origin and harvest time (Table 2). The dates of the purchased or harvested cabbages over the six sampling periods (S1–S6) are listed in Table 1.

#### *3.1. Glucosinolates in White and Red Cabbage from Local Food Retailers*

The GLS profile of the most abundant GLS of white cabbage purchased from the different food retailers over the 3-month period is given in Figure 1A–C, whilst the GLS profile for red cabbage is displayed in Figure 1D–F. Table 3 shows the chemical structures of the most abundant cabbage GLS, as well as their GLS hydrolysis product names including the abbreviations. In the heads of the analyzed white and red cabbage cultivars, a total of 12 chemically different GLSs were detected (Table S2). The main GLSs were allyl GLS (Allyl), 3-butenyl GLS (3But), 2-hydroxy-3-butenyl GLS (2OH3But), 3-(methylsulfinyl)propyl GLS (3MSOP), 4-(methylsulfinyl)butyl GLS (4MSOB), and indol-3-ylmethyl GLS (I3M) (Figure 1). The GLS profile in white and red cabbage cultivars between different supermarkets was often found to be similar within the same cabbage variety. In that way, Allyl, 3MSOP, and I3M were found to be most dominant in white cabbage, with maximum Allyl concentrations reaching 0.71 ± 0.02 μmol/g FW in S5 (Figure 1A), up to 0.64 ± 0.17 μmol/g FW 3MSOP in S1 of ORG1, and up to 0.49 ± 0.12 μmol/g FW I3M in S4 of ORG1 (Figure 1C). Red cabbage was often the richest in 2OH3But and 4MSOB (Figure 1D–F), with up to 0.99 ± 0.19 μmol/g FW 2OH3But in S1 of CON1 (Figure 1D) and 0.93 ± 0.05 μmol/g FW 4MSOB in S1 of CON1 (Figure 1D), respectively. However, Allyl, 3But, 3MSOP, and I3M were also formed in considerable amounts in red cabbage (Figure 1D–F).


*Foods* **2020** , *9*, 1682

In general, red cabbage heads produced higher levels of GLSs (Figure 1D–F) compared to white cabbage heads (Figure 1A–C). The highest total GLS levels in red cabbage heads were detected in S1 from CON1 (3.12 ± 0.27 μmol/g FW) (Figure 1D), whereas the highest total GLS content for white cabbage was observed in S5 from CON1 (1.61 ± 0.11 μmol/g FW) (Figure 1A).

Whilst general increases in total GLS concentration were found from sampling S1 to S6 (especially due to Allyl) in white cabbages from CON2 (Figure 1B), a general decreasing trend in total GLS concentration was detected in white cabbages procured from ORG1 (Figure 1C). In red cabbage, total GLSs varied for the individual samples from the same food retailers in all purchased red cabbage heads (Figure 1D–F), and no specific trend was noted between purchase dates from the different retailers. With regard to individual GLSs, many GLSs in cabbages from the three supermarkets did not significantly change over the different sampling dates, such as GLSs in red cabbage from CON2 (Figure 1E), while others were affected (Figure 1).

The two major GLSs detected in white cabbage heads were 3MSOP and Allyl. In white cabbages procured from CON1 (S1-S6, Figure 1A), Allyl increased by 2.1-fold from S1 to S5 and then decreased in S6 to levels similar to S1. In white cabbages, which were purchased from CON2, a general increase in Allyl content was observed from S1 to S6 (0.17 ± 0.03 to 0.61 ± 0.07 μmol/g FW) (Figure 1B), while in white cabbages purchased from ORG1 (Figure 1C), Allyl did not significantly change over time. In red cabbage, no significant changes were observable for Allyl (Figure 1D–F). 3MSOP as the other main GLS of white cabbage stayed the same over the sampling period in cabbages from CON2 and ORG1, but displayed increased levels in S4 and S5 of cabbages from CON1 compared to the S1-S3 samples (Figure 1A). In red cabbage, 3MSOP did not change across the consecutive sampling dates (Figure 1D–F). 4MSOB as a major GLS in red cabbage displayed reduced content in S2-S4 compared to S1 in CON1-cabbage (Figure 1D), but did not change in cabbages from CON2 and ORG1. Likewise, in white cabbage of ORG1, 4MSOB was not affected. In white cabbage from CON1, 4MSOB slightly decreased from S1 to S5 but was highest in S6 (Figure 1A), and in CON2-cabbages, this GLS generally displayed similar levels over the sampling dates (Figure 1B). Similarly, 2OH3But varied significantly only in white and red cabbage procured from CON1 (Figure 1A,D), but not in cabbages from the other supermarkets. In CON1-cabbages, 2OH3But was highest in S1 (red cabbage) or S2 (white cabbage) samples, then decreased until S4 (red cabbage) or S5 (white cabbage) and then again increased in the last samples (by tendency in red cabbage; significantly in white cabbage). The indole GLS I3M generally had similar levels over the different sampling dates in white and red cabbages, as well (Figure 1).

**Figure 1.** Glucosinolate (GLS) profile in white (**A**–**C**) and red cabbages (**D**–**F**) from conventional (**A**,**B**,**D**,**E**) and organic supermarkets (**C**,**F**) on different sampling dates (S1–S6). Exact sampling dates can be found in Table 1. A) and D) Represent conventional supermarket 1 (CON1), B) and E) stand for conventional supermarket (CON2), and C) and F) show results from organic supermarket 1 (ORG1). Each color in the bar of the given bar chart represents the mean plus standard deviation (SD) of the GLSs from three cabbage heads from the same supermarket (*<sup>n</sup>* = 3). Lower case letters indicate significant differences in means between the levels of a GLS on different sampling dates, as tested by ANOVA and Tukey HSD test at the *p* ≤ 0.05 level. Abbreviations: FW: Fresh weight; abbreviations of compounds as listedinTable3.

#### *3.2. Glucosinolate Hydrolysis Product Formation in White and Red Cabbages from Local Food Retailers*

Resulting from the homogenization of the fresh cabbage material, GLSs in cabbages from the different food retailers were degraded. The most pronounced GLS hydrolysis products released from white and red cabbages included the ITCs (or follow-up products from ITC) 3-(methylsulfinyl)propyl ITC (3MSOP-ITC), 4-(methylsulfinyl)butyl ITC (4MSOB-ITC), 3-butenyl ITC (3But-ITC), and 5-vinyloxazolidine-2-thione (OZT), the nitriles 5-(methylsulfinyl)pentanenitrile (4MSOB-CN) and 4-(methylsulfinyl)butanenitrile (3MSOP-CN) and the ETNs 1-cyano-2,3-epthiopropane (CETP), 1-cyano-3,4-epithiobutane (CETB), and isomers of 1-cyano-2-hydroxy-3,4-epithiobutane (CHETB A, CHETB B) (Table S3). Overall, the main GLS hydrolysis products in white cabbages were 3-MSOP-CN and 3MSOP-ITC, which were formed from 3MSOP and CETP, originating from the GLS Allyl (Figure 2A–C).

Red cabbages released mainly 4MSOB-CN and 4MSOB-ITC, originating from 4MSOB and CETP, but also 3MSOP-products and CETB and CHETB were often released in higher amounts (Figure 2D–F). Usually, homogenized red cabbages released more GLS-hydrolysis products compared to white cabbages (Figures 2 and 3). The formation of the cancer-preventive ITC 4MSOB-ITC was highest in red cabbages procured from CON1 in the S1 sample (1.06 ± 0.25 μmol/g FW), where it was also the main GLS-hydrolysis product. Although still being the main GLS-hydrolysis product in some samples, less 4MSOB-ITC was released in CON2- and ORG1- cabbages (up to 0.33 ± 0.18 μmol/g FW in S3 from CON2 and up to 0.44 ± 0.16 μmol/g FW in S2 from ORG1) (Figure 2D–F). The formation of GLS hydrolysis products in cabbages purchased from the three different supermarkets generally varied over the course of the sampling period (Figures 2 and 3). Overall, the S5 sample of red cabbages with 2.68 ± 0.57 μmol/g FW displayed the highest total level of released GLS-hydrolysis products.

Regarding white cabbages, total ITC concentrations were highest with up to 0.50 ± 0.19 μmol/g FW in the later samples S4 and S5 from CON1, where they were also the main GLS hydrolysis product type. In addition, total nitrile levels were higher in these later samples. Total ETN levels were also highest in S5 (Figure 3A). In white cabbages from CON2, total ITC levels did not change during the sampling season, while ETN levels increased in later samples (S3–S6) and nitriles only displayed increased levels in S4 (Figure 3B). In organic white cabbages from ORG1, cabbages from the first sampling date (S1) released with up to 0.96 ± 0.14 μmol/g FW mainly ITCs, which was generally the highest observed ITC level in white cabbages. On the other hand, samples S3, S5, and S6 only showed low levels of ITC formation (0.02–0.06 μmol/g FW) with nitriles and ETNs as the most dominant GLS-hydrolysis products (Figure 3C).

Red cabbages from the first samples (S1–S3) usually mainly released ITCs, while later samples (S3–S6) mainly released nitriles or ETNs (Figure 3D–F). More specifically, the total ITC formation was highest in the first samples and peaked in cabbages from S1 (CON1 and CON2) or S2 (ORG1), while later samples released much lower total ITCs (Figure 3D–E). In red cabbages from CON1, ETNs and nitriles displayed increased levels in S5 and S6 cabbages compared to S1–S4 (Figure 3D). Comparably, in CON2, the red cabbages total ETN formation was also higher in S5, but nitriles were not affected (Figure 3E). In organic red cabbages from ORG1, total nitriles were also highest in S5, while ETN-formation fluctuated and displayed increased levels in S2 and S5 compared to S3 and S6 (Figure 3F).

*p*

Regarding the ratios (%) of total EPTs, nitriles, and ITCs relative to the total amount of formed GLS hydrolysis products, usually the relative ITC formation was higher in the first samples of cabbages, where they were often the main GLS hydrolysis products, but ITC formation decreased towards later sampling dates. Relative nitrile formation often behaved the other way around and was higher in later samples compared to early samples (Figure 4). More specifically, in white cabbages from CON1, the % of ITCs more than halved, while relative nitrile levels more than tripled from S1 to S6 and relative ETN formation was slightly increased in S2 compared to the other samples (Figure 4A). In white cabbages from CON2, the relative ITC formation was highest in S2 with 63 ± 11% and then decreased to S5 by 56% to 28 ± 9% of ITC formation. While relative nitrile formation was hardly affected, relative ETN formation increased from S1 to S5 to 46 ± 10% of ETN formation in S5 (Figure 4B). In organic white cabbages from ORG1, GLSs also mainly released ITCs with 75 ± 3% in S1, while cabbages from later samples (S3, S5, and S6) mainly released nitriles with up to 65 ± 7% (in S5) (Figure 4C). Red cabbages showed a very similar GLS hydrolysis behavior: With up to 75 ± 7% of ITC formation (S1 from CON1) the first samples (S1-S3) released mainly ITCs and the formation decreased towards later sampling dates, while nitrile formation increased in reverse with later samples to up to 64 ± 8% (S5 from ORG1).

The relative ETN release was not affected in red cabbages from CON2 and ORG1, but increased with later samples in CON1 up to 45 ± 9% of all GLS products (in S5) (Figure 4D–F). As an indicator of ESP-activity, the relative formation of CETP, Allyl-ITC, and Allyl-CN were monitored, as well. In red cabbages from CON1 and ORG1, the relative release of CEPT increased from first to last samples, while in white cabbages, an increase from S1 to S2–S5 was found (Supplemental Figure S1). The relative CETP formation was unaffected in red cabbages from CON2 and conventional white cabbages.

#### *3.3. Glucosinolates and Glucosinolate Hydrolysis Products Formation in Freshly Harvested White and Red Cabbages*

The GLS profile of freshly harvested white and red cabbages was similar to the commercial ones, and 3MSOP and Allyl were most dominant in white cabbage (Figure 5A), whilst 2OH3But and 4MSOB contributed the most towards the total GLS content of freshly harvested red cabbage (Figure 5D). The total GLS and GLS-hydrolysis product level of red cabbages was higher compared to the freshly harvested white cabbages. With regard to individual GLS hydrolysis product formation, the main GLS hydrolysis products released from homogenized freshly harvested white cabbages were CETP and 3MSOP-CN (Figure 5B) and from freshly harvested red cabbages 4MSOB-CN and 4MSOB-ITC (Figure 5E). Of the detected GLS hydrolysis products from freshly harvested white cabbages, 36 ± 13% were nitriles (mainly 3MSOP-CN), 34 ± 7% were ETNs, and 30 ± 9% were ITCs (Figure 5C), while in red cabbages GLSs were degraded to 52 ± 3% nitriles (mainly 4MSOB-CN), 28 ± 2% ETNs, and 20 ± 4% ITCs (Figure 5F).

**Figure 5.** Glucosinolate (GLS) (**A**,**D**) and their absolute (**B**,**E**) and relative (**C**,**F**) hydrolysis product formation in white (**A**–**C**) and red cabbages (**D**–**F**) harvested freshly from the field. Each color in the bar of the given bar chart represents the mean plus standard deviation (SD) of the GLS (**A**,**D**), their respective hydrolysis products (**B**,**E**), or the ratio of relative isothiocyanates (ITCs), nitriles (CNs), and epithionitriles (ETNs) (**C**,**F**) from three cabbage heads freshly harvested from the field (*n* = 3). Abbreviations: FW: Fresh weight; abbreviations of compounds are listed in Table 3.

#### **4. Discussion**

In this study, the GLS content and formation of GLS hydrolysis products was evaluated in commercial white and red cabbages purchased from two conventional and one organic supermarket in Germany over a period of 3 months and compared to freshly harvested cabbages. In general, the composition of individual GLSs in red and white cabbages among different food retailers displayed only slight fluctuations and the GLS profile and levels were similar over the six sampling periods (Figure 1). With Allyl and 3MSOP being the main GLS in commercial and freshly harvested white cabbages and red cabbages being rich in 4MSOB and 2OH3But, (but also of Allyl, 3But, 3MSOP, and I3M), the GLS profiles and levels were in accordance to previous reports [14,24]. The small

variability in GLS levels is an unexpected observation, as the analyzed white and red cabbages differed in genotype, came from different regions, were cultivated on different soil types using different fertilizers and storage practices, and were also purchased from different food retailers that belonged to different food trading companies (Table 2, Supplemental Table S1). Previous studies showed that GLS levels in *Brassica oleracea* vegetables are affected by cultivar (genotype) [14,25], nutrient supply [26,27], climatic conditions [17,20,28], as well as storage conditions [29,30]. As the variability of the GLSs was relatively low, it is suspected that genotypes were similar in their initial GLS concentrations and that also cultivation practices and storage conditions had no major effect on the GLS content of the cabbages, when they were finally sold in the supermarket. On the other hand, long-term storage (2 ◦C, 95% of relative humidity, up to 100 days) was shown to decrease the GLS content in Chinese cabbage (*Brassica rap*a L. spp. *pekinensis*), with GLSs being more stable in cabbages stored under a controlled atmosphere (CA) (2% O2 and 2% CO2) [31]. Accordingly, Osher et al. (2018) even reported an increase for aliphatic ITC-formation in cabbage (*B. oleracea*) stored at 1 ◦C under CA for 60 days (CA: 2% O2, 5% CO2), while ITC formation declined when stored under normal atmosphere (up to 45 and 72% decline in Allyl-ITC after 60 and 90 days, respectively) [32]. In the present study, I3M levels of organic white and red cabbages from ORG1 were often higher, compared to the cabbages from the conventional food retailers CON1 and CON2 (Figure 1C). Likewise, using NMR spectroscopy, Lucarini et al. (2020) recently also found nearly 3 times as much I3M in organic broccoli compared to conventionally grown ones [33]. In that study, the main difference in both farming practices was the fertilization practice, as no pesticides were used. While the same amount of nitrogen was supplied to the soil, for conventional broccoli with 0.2 t/ha urea and 15 t/ha bovine manure were applied, while organic grown broccoli was fertilized with 28 t/ha [33]. Nevertheless, increased I3M biosynthesis could be also explained due to the absence of chemical pesticides in organic cultivation practices, resulting in the stimulation of indole GLS biosynthesis upon herbivory damage via the methyl jasmonate signalling pathway [34].

Upon homogenization, GLSs were hydrolyzed in cabbages, yielding nitriles, ITC (or breakdown product thereof), and ETNs. Usually, the recovery of aliphatic GLS hydrolysis products was good with recoveries in a range of 60–130%. However, in some samples, low recoveries of aliphatic GLS hydrolysis samples were also observed (for example, S3 and S6 white cabbage samples and the S3 red cabbage sample from ORG1) (Figures 1 and 2). Regarding this observation, myrosinase activity in these samples was probably low, therefore, resulting in a low recovery of hydrolysis products. In pre-experiments performed for the current study, the recovery of GLS hydrolysis products did not benefit from longer incubation times. Probably due to chemical instabilities of the products [35,36], it is likely that the myrosinase activity decreases with incubation time and that the initial myrosinase activity might be a major factor for the recovery of products. This hypothesis is further supported by the observation that sulfate, which is released during GLS hydrolysis, is a competitive inhibitor of myrosinase activity [37].

With regard to the differences in GLS hydrolysis products in the different samples, in contrast to GLS, the formation of GLS hydrolysis products varied strongly between the purchase dates (S1–S6; Figure 2). Especially during the first samples (S1–S3) mainly ITCs were released from red cabbages, while in later samples nitriles were preferentially formed (S5, S6) (Figure 3). Likewise, the relative ITC formation in white and red cabbages generally decreased until the last sample, while nitriles were usually the major hydrolysis products (Figure 4). These results show that in contrast to some previous reports [14,38], ITCs can be the main hydrolysis products in cabbages, as this was the case for cabbages purchased in early autumn. Here, red cabbage could be an excellent source for cancer-preventive 4MSOB-ITC (sulforaphane) (up to 1.06 μmol/g FW in red cabbage from CON1 at S1), releasing levels, which were 6-times higher compared to mature broccoli and comparable to the 4MSOB-ITC release from broccoli sprouts [14].

With regard to the high nitrile and ETN formation at later purchase dates, the ESP protein activity is made responsible for ETN-release from alkenyl GLS and for increased formation in simple nitriles

from non-alkenyl GLS [39,40]. Therefore, it was suspected that the ESP-activity increased towards later samples, while it remained low in the first samples. When regarding the relative formation of Allyl hydrolysis products, as an indicator of ESP activity, it can be further supported that the ESP activity increased with later purchase dates in red cabbage from CON1 and organic cabbage, while it was not considerably affected in conventional white cabbage (Figure S1). As the relative nitrile formation significantly increased in white cabbage from CON1 in later samples (Figure 4A) (but relative CETP release not, Figure S1), it is suspected, that next to the ESP activity, also other factors influence GLS hydrolysis, which could have resulted in changes in hydrolysis product behavior from S1 to S6 due to their variation. As especially nitriles increased (Figure 4A), it is suspected that white cabbage contains nitrile specifier proteins, which are involved in nitrile formation in *Arabidopsis thaliana* [41]. This suspicion is strengthened by the observation that *Brassica oleracea* contains a gene with a homology of 80% compared to the nitrile specifier protein 1 of *A. thaliana* [42]. Further, this hypothesis is supported by a recent study, which could neither prove the nitrile specifier protein activity for three *B. oleracea* ESP isoforms in vitro nor in vivo, but nitrile formation from alkyl GLS was observed [43].

To date, there is little data how pre- and postharvest factors affect glucosinolate hydrolysis. Freshly harvested white and red cabbages showed a similar GLS hydrolysis behavior compared to the supermarket cabbages purchased during similar dates (S5 white cabbage, S4 red cabbage) (Figure 5; Figures 2–4). Due to this finding and due to the different storage conditions of growers and retailers, in the present study storage does not seem to be the factor that caused a reduced ITC release in cabbages purchased in later autumn. With regard to preharvest factors, nitrogen and sulfur supply affected the release of GLS hydrolysis products in the ETN-producer Chinese cabbage (*Brassica rapa* L. ssp. *pekinensis*) and ITCs were reduced in response to the increasing N and decreasing S supply [44]. In pak choi (*B. rapa* subsp. *chinensis* (L.) Hanelt), which also mainly released ETN, the ITC/CN and ITC/ETN ratio increased with the increasing sulfur supply [45]. With regard to the present study, it is unlikely that differences in fertilization were responsible for the observed changes in the GLS hydrolysis behavior as also the red cabbage from CON2 which originated from the same grower (Table 2, Table S1) showed reduced % ITC release at later purchase dates (Figure 4E). Moreover, herbivore feeding can affect the GLS hydrolysis behavior, and simple nitrile formation was shown to increase in response to the specialist insect feeding (*Pieris rapae*) in *Arabidopsis thaliana* Col-0 [46], while ITC-emitting plants appear to be better defended against generalist herbivores [47]. However, as organic and conventional cabbages displayed a similar hydrolysis behavior (Figure 4D–F), it is suspected that climatic conditions such as reduced radiation or decreasing temperatures across the autumn season might be responsible for the observed shifts. Moreover, all of the conventional cabbages originated from the Dithmarschen region (Schleswig-Holstein, Germany; Table S1), which is characterized by a coastal climate and marsh soil (being ideal conditions for growing cabbages). It is the largest coherent cabbage-growing area in Europe. In Table S4, the climatic data of the presumable growing season in 2019 is presented. Consequently, it is likely that temperature and radiation interact with regard to the GLS hydrolysis behavior. Recently, Jasper et al. (2020) showed that at higher temperatures during growth, more GLS hydrolysis products were formed from rocket (*Eruca sativa*), while GLS levels were less affected [48]. Likewise, Ku et al. (2013) reported different ITC conversion rates in broccoli grown in 2 different years and linked this to different climatic conditions. Unfortunately, in that study, nitriles were not analyzed and therefore, conversion rates could have been also affected by changes of the myrosinase activity [49]. As organic cabbages which originated from different regions in Germany (Brandenburg, Mecklenburg-Western Pomerania, Schleswig-Holstein; Table 2 and Table S1) also showed similar changes in the GLS hydrolysis behavior compared to the conventional cabbages originating from the Dithmarschen region, it can be assumed that the results obtained in this study might be also valid for other countries and regions with similar climatic conditions. The specific role of climatic growth conditions on GLS hydrolysis in *B. oleracea* vegetables will need to be evaluated in future studies.

#### **5. Conclusions**

Current findings in this study have highlighted the great diversity, particularly of the GLS hydrolysis behavior in white and red cabbages between the different supermarkets over the six sampling periods: Whilst the GLS composition and content remained similar between the different food retailers, the composition and content of the individual hydrolysis products formed varied across the season and high ITC levels were generally noted in early sampling periods (early September) and decreased, particularly in red cabbages over time. Here, the increased specifier protein activity is made responsible for the reduced ITC-release.

In conclusion, with regard to their potential to release more ITCs, consumption of commercial cabbages purchased in early autumn could be healthier options than those purchased in later autumn months. The fact that ITCs can be preferentially formed in earlier autumn months, but hardly towards the end of autumn, underlines the need to unravel the factors that affect the GLS-hydrolysis outcome. The results of this study might also help growers and food companies produce cabbages and products with more pungency due to a higher ITC formation. Due to the potential of red cabbage to form high rates of health-promoting 4MSOB-ITC in cabbages purchased in early autumn, red cabbage consumption could be an alternative for people who dislike broccoli.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/9/11/1682/ s1, Figure S1: Allyl glucosinolate hydrolysis product ratios [percentage of allyl isothiocyanate (Allyl-ITC), 3-butenenitrile (Allyl-CN) and 1-cyano-2,3-epithiopropane (CETP)] from white (A–C) and red cabbages (D–F) from conventional (A, B, D, E) and organic supermarkets (C, F) on different sampling dates (S1–S6). Table S1: Overview of white and red cabbages purchased from three different food retailers or harvested freshly from the field (Großbeeren, D). Table S2: Glucosinolates (GLSs) content in commercial and cabbages harvested freshly from the field in μmol/g fresh weight. Table S3: Glucosinolate (GLS) hydrolysis product formation in commercial and cabbages harvested freshly from the field in μmol/g fresh weight. Table S4: Climatic data of the Dithmarschen region in northern Germany in 2019.

**Author Contributions:** Conceptualization, N.S.W., F.S.H., and S.R.; methodology, N.S.W. and F.S.H.; validation, F.S.H.; investigation, N.S.W. and F.S.H.; resources, F.S.H.; writing—original draft preparation, N.S.W. and F.S.H.; writing—review and editing, F.S.H. and S.R.; visualization, N.S.W. and F.S.H.; supervision, F.S.H. and S.R.; project administration, F.S.H.; funding acquisition, F.S.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** Franziska Hanschen and this research study are part of the Leibniz-Junior Research Group OPTIGLUP, funded by the Leibniz-Association (J16/2017).

**Acknowledgments:** The excellent technical assistance of Andrea Jankowsky, Andrea Maikath, and Jessica Eichhorn is gratefully acknowledged.

**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* **Influence of Cabbage (***Brassica oleracea***) Accession and Growing Conditions on Myrosinase Activity, Glucosinolates and Their Hydrolysis Products**

**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:** Glucosinolates are secondary plant metabolites present in *Brassica* vegetables. The endogenous enzyme myrosinase is responsible for the hydrolysis of glucosinolates, yielding a variety of compounds, including health-promoting isothiocyanates. The influence of cabbage accession and growing conditions on myrosinase activity, glucosinolates (GSL) and their hydrolysis products (GHPs) of 18 gene-bank cabbage accessions was studied. Growing conditions, cabbage morphotype and accession all significantly affected myrosinase activity and concentration of glucosinolates and their hydrolysis products. In general, cabbages grown in the field with lower growth temperatures had significantly higher myrosinase activity than glasshouse samples. Profile and concentration of glucosinolates and their hydrolysis products differed across the accessions studied. Aliphatic glucosinolates accounted for more than 60 % of total glucosinolates in most of the samples assessed. Nitriles and epithionitriles were the most abundant hydrolysis products formed. The results obtained showed that consumption of raw cabbages might reduce the amount of beneficial hydrolysis products available to the consumer, as more nitriles were produced from hydrolysis compared to beneficial isothiocyanates. However, red and white cabbages contained high concentrations of glucoraphanin and its isothiocyanate, sulforaphane. This implies that careful selection of accessions with ample concentrations of certain glucosinolates can improve the health benefits derived from raw cabbage consumption.

**Keywords:** *Brassica oleracea*; cabbage; growing condition; myrosinase activity; glucosinolates; glucosinolate hydrolysis products; isothiocyanates; nitriles; epithionitriles

#### **1. Introduction**

Cabbage (*Brassica oleracea*) belongs to the *Brassicaceae* family and comprises eight distinct cultivar groups, all descended from wild cabbage (*B. oleracea* var. *oleracea*) [1]. Epidemiological studies have shown that the consumption of *Brassica* vegetables reduces the risks of cardiovascular diseases and cancer [2] and is reported to have a cytoprotective effect against tissue damage associated with oxidative stress as well as antimicrobial activity against bacterial and fungal pathogens [3,4].

*Brassica* vegetables are unique in comparison to other vegetables because they contain the enzyme myrosinase and a group of thioglucosides called glucosinolates (GSLs). GSLs are sulphur and nitrogen containing biologically active secondary metabolites found in plants of the order Capparales, which includes the *Brassicaceae* family and other economically important agricultural crops [5–7]. In plants, GSLs act as plant defense mechanisms against stress, insect, and pest attack [8]. GSLs have been grouped into three main classes based on the structure of their different amino acid precursors; these groups are aliphatic, aromatic and indole GSLs. Aliphatic GSLs are derived from alanine, leucine, isoleucine, methionine or valine; aromatic GSLs are from phenylalanine or tyrosine, while tryptophan-

**Citation:** Oloyede, O.O.; Wagstaff, C.; Methven, L. Influence of Cabbage (*Brassica oleracea*) Accession and Growing Conditions on Myrosinase Activity, Glucosinolates and Their Hydrolysis Products. *Foods* **2021**, *10*, 2903. https://doi.org/10.3390/ foods10122903

Academic Editors: Franziska S. Hanschen and Sascha Rohn

Received: 11 October 2021 Accepted: 18 November 2021 Published: 23 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/).

derived GSLs are called indole GSLs [9,10]. A recent review by Blaževi´c et al. [11] stated that between 88–137 glucosinolates (GSLs) have been characterised in plants to date.

GSLs and myrosinase enzymes coexist in separate compartments in the plants; while glucosinolates exists in the vacuoles of various cells [6], myrosinase enzymes are localised inside the myrosin cells. When plant tissue is disrupted, GSLs are hydrolysed by plant myrosinase enzymes, resulting in the formation of various hydrolysis products such as isothiocyanates (ITCs), thiocyanates, nitriles and epithionitriles [5]. The extent of glucosinolate hydrolysis and the type of hydrolysis compound produced is dependent on a number of factors, which include coexisting myrosinase enzyme, presence of epithiospecifier protein (ESP), ascorbic acid, Fe2+ and MgCl2, structure of the glucosinolate side chain, the plant species, as well as reaction conditions such as pH and temperature [9,12,13].

ITCs, the primary products of GSL hydrolysis from myrosinase, are responsible for the well-documented health-promoting properties of *Brassica* vegetables, such as reduced risk of cardiovascular diseases (CVD) and cancer [2,5]. For example, sulforaphane (SFP), the hydrolysis product of glucoraphanin present in high concentrations in broccoli and cabbage, has been reported to possess chemoprotective, antimicrobial, anti-inflammatory, and antithrombotic properties [14,15]. Allyl isothiocyanate (AITC), another common ITC present in cabbages and produced upon myrosinase hydrolyses of the glucosinolate sinigrin (SIN), was reported to be potent against human breast cancer cells [16], human erythroleukemic K562 cells [17], and more potent on human A549 and H1299 non-small cell lung cancer cells in vitro than 2-phenylethyl-ITC (PEITC; ITC from gluconasturtin) [18]. However, in the presence of epithiospecifier proteins (ESPs), nitriles and epithionitriles (EPTs), which have not been shown to proffer any beneficial characteristics for health, are formed [19]. GSLs and ITCs are also partly responsible for the bitter taste and pungent aromas of *Brassica* vegetables, which limits consumer acceptance and liking of *Brassica* vegetables [20–23].

There are several factors that affect the GSL-myrosinase system in *Brassicas*; these factors include climatic factors, location, and growing conditions [24–27], morphotype and the variety of plant [28,29]; with the impact of these factors varying between studies. For example, while some authors have suggested that the effect of plant genotype on GSL concentrations is greater than that of environmental factors [30,31], others have reported higher variations in GSL concentrations as a result of environmental conditions than genetic factors [32,33].

To date, most studies on myrosinase activity have focused on single cultivars of B. *oleracea* species [28,34–37], with studies on the myrosinase activity of different varieties within a species limited [29,38]. Variations in myrosinase activities were reported in different varieties of Brussels sprouts, broccoli, cauliflower, Chinese cabbage, and white cabbage [38]. The authors found a two-fold difference in the myrosinase activities of five broccoli varieties as well as two cauliflower and Chinese cabbage varieties. Low temperature conditions are reported to increase the myrosinase activity of *B. oleracea* species (Brussels sprout, broccoli, cauliflower, cabbage, and kale) grown in the autumn season [25].

Several studies have been undertaken on the formation of GSLs in cabbage varieties, some of which have investigated GSL concentrations in cabbages grown under different conditions; with , most focused on GSL concentrations of cabbages grown in different locations or different seasons [29,32,39–41]. However, none of the studies analysed the glucosinolate hydrolysis products (GHPs) of cabbages under different plant growth conditions and instead made suggestions on potential GHP concentrations of the samples based on the concentrations of GSLs observed. These suggestions may be problematic, as studies have shown that GSL concentration is not necessarily correlated with the abundance of GHPs formed [42,43].

Little is known of the GHPs in cabbages, as most studies have focused on a specific cabbage variety [44] or ITCs in other *B. oleracea* such as broccoli [45]. A recent study analysed the GSLs and GHPs of cabbages with a focus on red, white and savoy cabbages, but the samples were grown under the same conditions [43]. To fully understand the

health benefits that can be derived from cabbage consumption, however, there is a need to characterize the GHPs produced from GSL hydrolysis and understand the factors affecting the type and concentrations of GHPs formed. With growing health campaigns promoting the consumption of more fruits and vegetables, and consumers wanting to include more fresh vegetables like cabbage in their diet, many people now grow their own cabbages at home in pots, either in green/glasshouses or in the garden [46,47]. It is therefore important to investigate the effect of these plant growth conditions on the GSL-myrosinase system to ensure that the health benefits desired from their consumption are not lost.

In light of this gap in present knowledge, the purpose of this study was to investigate the influence of growing conditions and accession identity on myrosinase activity as well as the GSL and GHP content of cabbage. A total of 18 cabbage accessions across six different cabbage morphotypes were selected from a genetic resources unit and grown under two different conditions. In addition to wild cabbage, this study used red, white, and green (savoy) cabbage (*B. oleracea* var. *capitata*), kale (*B. oleracea* var. *acephala*) and sea kale (*B. oleracea* var. *tronchuda*). The primary hypothesis of the study was that cabbage growth conditions will affect myrosinase activity as well as the GSL and GHP contents of cabbage. The secondary hypothesis was that while cabbage morphotype and accession would affect myrosinase activity, cabbage morphotype rather than accession will affect the profile and concentrations of the GSLs and GHPs formed. The results of myrosinase activity and variations in the amount and profile of GSLs and GHPs in cabbage accessions across both plant conditions studied are presented.

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

#### *2.1. Plant Material*

Cabbage accessions were selected from the University of Warwick Crop Centre Genetic Resources Unit (Wellesbourne, UK). Eighteen cabbage accessions comprising six cabbage morphotypes (wild (*B. oleracea* var. *oleracea*), black kale (*B. oleracea* var. *acephala*), tronchuda (*B. oleracea* var. *tronchuda*), savoy, red and white (*B. oleracea* var. *capitata*)) were used for the experiment. Cabbages were selected based on their geographical origin, whether or not they were of hybrid descent, and morphology of head formation (closed heart or open leaf), as shown in Table 1 and Supplementary Figure S1. Seeds of one white cabbage accession (WC-DLI) did not germinate when sown and thus will not be discussed further. Out of the remaining 17 accessions planted, RC-RM (red cabbage) and SC-SDG (savoy cabbage) did not survive in the glasshouse.

A total of 15 biological replicates of each accession were germinated in seedling trays using potting compost under controlled environmental conditions (Saxcil cabinets). A 16 h photo period was used (16 h light, 8 h dark); relative humidity was set to 60%, with day and night temperatures of 22 ◦C and 16 ◦C, respectively. Seedlings were allowed to grow in seedling trays until the appearance of 3–4 true leaves, before being transplanted to individual 2.5 L pots containing loam-based compost (7–8 May 2014) and left to grow in the glasshouse (minimum night temperature 13 ◦C). After 50 days (26–27 June 2014), five replicates of each accession were transplanted to larger pots (10 L) containing loam-based compost and allowed to grow until commercial maturity in the glasshouse, while seven replicates of each accession were transplanted to the field and allowed to grow to commercial maturity. On the field, each accession was planted on 7 metre beds with 0.6 metres between plants and rows. Both glasshouse and field cabbages were fertilized twice weekly with nitrogen phosphate potassium (NPK) (100 kg/ha N, 100 kg/ha P and 200 kg/ha K) fertilizer. Standard agricultural practices were employed in the cultivation of the cabbages, including a programme of pest management using insecticides and fungicides. Cabbages were grown between 7 March–25 November 2014 in the plant growth facilities, Whiteknights campus of the University of Reading, UK (Supplementary Figure S2).


Names in bold refer to cabbage morphotype

a

**Table 1.** Origin and botanical and common names of cabbage accessions planted between May and November 2015.

Cabbages were harvested over a period of two days upon reaching commercial maturity based on visual inspection. Though some accessions attained commercial maturity earlier than others, they had sufficiently good field holding capacity to be left until all accessions were mature before harvesting, so that all plants experienced equivalent environmental conditions. Harvested plants were placed on ice in freezer bags and immediately stored in a cold room at 4 ◦C for 24 h before processing. The average weight of each field cabbage head per plant was 700 g (closed heart) and 300 g (open leaf), while the glasshouse cabbages were smaller (400 g for closed heart and 250 g for open leaf cabbages) (Supplementary Figure S1). Climatic data for both growing conditions are presented in Supplementary Table S1.

#### *2.2. Reagents and Chemicals*

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

#### *2.3. Sample Preparation*

The outer leaves and central core of 4–5 cabbage heads (biological replicates) were removed and discarded in order to remove senescent leaves and achieve a representative sample spanning similar leaf ages for each morphotype. Cabbages were chopped into pieces of approximately 1 cm in width using a kitchen knife (representing how cabbages would normally be sliced by consumers), mixed together, and washed under running tap water; excess water was drained using a salad spinner (OXO Good Grips Clear Manual Salad Spinner, Chambersburg, PA, USA). A total of 120 g of cabbage samples was put into sterile sterilin tubes, immediately 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*

Myrosinase enzyme was extracted using the method described by Ghawi et al. [48]. A sample of 0.1 g was suspended in 0.15 g polyvinylpolypyrrolidone (PVPP) and 10 mL of Tris-HCL buffer (200 mM, pH 7.5) containing 0.5 mM ethylenediaminetetracetic acid (EDTA) and 1.5 mM dithiothreitol (DTT). The mixture was stirred for 15 min at 5 ◦C and centrifuged (11,738× *g*) for 15 min at 5 ◦C. The final volume of supernatant was made up to 10 mL using the Tris-HCL buffer. Then, 6.2 g ammonium sulphate was added to the supernatant to achieve 90% saturation and stirred at 5 ◦C for 30 min. The samples were then centrifuged (13,694× *g*) for 15 min at 5 ◦C. The resulting pellet was suspended in 2 mL Tris-HCl buffer (10 mM, pH 7.5) and assayed for myrosinase activity.

Myrosinase activity was measured using the coupled enzyme method described by Gatfield and Sand [49] and Wilkinson et al. [50], with slight modifications. The procedure depends on the glucose released from the reaction between myrosinase enzymes and the substrate (sinigrin). The mixture for the reaction consisted of 0.9 mL of 5 mM ascorbic acid, 0.5 mL ATP/NADP+ solution, 10 μL hexokinase/glucose-6-phosphate dehydrogenase and 50 μL crude enzyme extract. The mixture was homogenized and allowed to stand for 3 min, and then 50 μL sinigrin substrate (0.6 M) added. The change in absorbance due to NADPH formation was read on a spectrophotometer at 340 nm. Myrosinase enzyme activity was determined by taking the slope of the linear part of the curve of absorbance versus the time of reaction. One unit of myrosinase activity is defined as the amount of enzyme that produces 1 μmol of glucose from sinigrin substrate per minute at pH 7.5.

#### *2.5. Protein Assay*

Protein content was measured using the Bradford method [51]. The procedure is based on formation of a complex between dye (brilliant Blue G, Sigma-Aldrich, Dorset, UK) and the protein present in the sample, and absorbance is read at 595 nm using a spectrophotometer. 50 μL filtered crude enzyme extract was added to 1.5 mL of concentrated dye reagent, vortexed and allowed to stand for 20 min before taking the absorbance reading. Bovine serum albumin (BSA) (Sigma-Aldrich, Dorset, UK) was used to construct a standard curve, and the protein concentration of sample was calculated from the standard curve obtained. Protein content was used to calculate the specific activity of myrosinase enzymes (U/mg protein).

#### *2.6. Glucosinolate Extraction and LC-MS2 Analysis*

The method used for GSL extraction is as described by Bell et al. [52], with modifications. Briefly, 40 mg ground cabbage powder was heated in a heat block at 75 ◦C for two minutes. Then, 1 mL 70% (*v*/*v*) methanol preheated to 70 ◦C was added to each sample, vortexed and placed in a preheated (70 ◦C) water bath for 20 min. Samples were centrifuged at full speed for five minutes (18 ◦C), and supernatant was collected in fresh Eppendorf tubes. The volume was adjusted to 1 mL with 70% (*v*/*v*) methanol and frozen at −80 ◦C until further analysis.

Samples were filtered using 0.22 μm Millex syringe filters with a low protein binding Durapore polyvinylidene fluoride (PVDF) membrane (Fisher scientific, Loughborough, UK) and diluted with 9 mL HPLC-grade water. LC-MS analysis of GSL extracts was performed in negative ion mode on an Agilent 1200 Series LC system (Agilent, Stockport, UK) equipped with a variable wavelength detector and coupled to a Bruker HCT ion trap (Bruker, Coventry, UK). Sample separation 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 a 5 min post-run sequence. Mobile phases consisted of 95% of 0.1% ammonium formate solution and 5% acetonitrile. The flow rate was optimised for the system at 0.4 mL/min, with a column temperature of 30 ◦C and with 5 μl of sample injected into the system. GSLs were quantified at a wavelength of 229 nm.

MS analysis settings were as follows: electrospray ionization (ESI) 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. GSLs were quantified using sinigrin hydrate standard. Five 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 parent mass ion and characteristic ion fragments as well as comparing with literature ion data (Table 2). Compounds were quantified using Bruker Daltonics HyStar software (Bruker, Coventry, UK). Relative response factors (RRFs) were used in the calculation of GSL concentrations where available [53]. Where such data could not be found for intact GSLs, RRFs were assumed to be 1.0.


**Table 2.** Intact glucosinolates identified in cabbage accessions analysed by LC-MS.

Key: GSL = glucosinolate; <sup>a</sup> Base ion highlighted in bold

#### *2.7. Extraction of Glucosinolate Hydrolysis Products*

GHPs were extracted and analysed following the method described by Bell et al. [57]. A total of 0.5 g of lyophilized cabbage was mixed with 10 mL deionized water, vortexed and allowed to incubate for three hours at 30 ◦C. The mixture was then centrifuged at 5000× *g* (18 ◦C) for ten minutes, and the supernatant collected. The pellet was extracted two more times with 10 mL deionized water, and the supernatants were combined and filtered (0.45 μm syringe filters, Epsom, UK) into glass centrifuge tubes. GHPs were extracted by adding an equal volume of dichloromethane (DCM) to the supernatant, vortexed for one minute and centrifuged at 3000× *g* for ten minutes. After centrifugation, the organic phase was collected, and the extraction step repeated twice. The organic phase collected was combined, 2 g sodium sulphate salt was added to remove any excess liquid present, and the mixture was filtered into a round-bottom flask. The filtrate was dried using a rotatory evaporator (37 ◦C), re-dissolved in 1 mL DCM, and filtered (0.22 μm filter; Fisher scientific, Loughborough, UK) in GC-MS glass vials (VWR, Lutterworth, UK) for GC-MC analysis.

#### *2.8. GC-MS Analysis*

GC–MS analysis was performed on an Agilent 7693/5975 GC–MS autosampler system (Agilent, Manchester, UK). The sample was injected onto a HP-5MS 15 m non-polar column DB-5MS (J and W scientific, Santa Clara, CA, USA) (0.25-μm film thickness, 0.25 mm I.D.). The injection temperature was 250 ◦C in split mode (1:20). The oven temperature was programmed from 40 to 320 ◦C at a rate of 5 ◦C/min until 250 ◦C. The carrier gas was helium, with flow rate of 1.1 mL/min and pressure of 7.1 psi. Mass spectra were obtained by electron ionization at 70 eV, and mass scan from 35 to 500 amu. A total of 1 μL of the sample was injected, and compounds were separated during a 42 min run. Compounds were identified using the National Institute of Standards and Technology (NIST) library and literature ion data (Table 3; see Figure S3 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, Dorset, UK) were prepared in DCM (*r*<sup>2</sup> = 0.99). Data analysis was performed using ChemStation for GC-MS (Agilent, Manchester, UK).


 **3.** Glucosinolate hydrolysis products identified in cabbage accessions analysed by GC-MS.

**Table**

*Foods* **2021**, *10*, 2903

reference spectrum in the

NIST/EPA/NIH

 mass spectra database and that in the literature. c Base ion highlighted in bold.

#### *2.9. Statistical Analysis*

The results are the average of three biological replicates (each replicate consists of leaves from 4–5 cabbage heads) and two technical replicates (*n* = 6). Data obtained were analysed using 2-way ANOVA, with both cabbage accession (or morphotype) and growing condition (glasshouse and field) fitted as treatment effects, and Tukey's HSD multiple pairwise comparison test used to determine significant differences (*p* < 0.05) between samples. Multifactor analysis (MFA) was used to visualise the GSL and GHP data in a minimum number of dimensions (two or three). All statistical analyses were performed in XLSTAT (version 2019.4.2, Addinsoft, Paris, France).

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

#### *3.1. Effect of Growing Conditions, Cabbage Morphotype and Accession on Myrosinase Activity*

The myrosinase activity of cabbages grown on the field and in the glasshouse is shown in Figure 1. Myrosinase activity ranged from 12.2 U/g DW (BK-CPNT) to 127.4 U/g DW (SC-PW) in glasshouse samples and from 31.5 U/g DW (BK-CPNT and RC-RL) to 154.8 U/g DW (SC-PW) in field samples. Growing condition (glasshouse versus field), cabbage morphotype, cabbage accession and the interactions between these parameters significantly (*p* < 0.0001) affected myrosinase activity. The myrosinase activity of cabbage accessions within a cabbage morphotype differed significantly for all cabbage morphotypes studied. This agrees with previous reports that myrosinase activity varies within varieties and plant species [65]. Singh et al. [38] and Penas et al. [29] also reported variations in the myrosinase activity of different cabbage varieties within and between cabbage morphotypes. There were significant differences in the myrosinase activity of field and glasshouse grown cabbages across most of the accessions studied. Field grown cabbages had significantly higher myrosinase activity than glasshouse cabbages, except for WC-FEM, where the myrosinase activity of the glasshouse sample was significantly (*p* < 0.003) higher than that of the field grown counterpart.

**Figure 1.** Myrosinase activity of field and glasshouse grown cabbages. Values are means of three biological replicates (each replicate comprising 4–5 cabbage heads) and two separately extracted technical replicates (*n* = 6). Error bars represent standard deviation from mean values. Missing data points implies cabbage accession did not survive under glasshouse growing conditions. Letters "A-D": bars not sharing a common uppercase letter indicates significant differences (*p* < 0.0001) between accessions and growing conditions within a cabbage morphotype. Letters "a-k": bars not sharing a common lowercase letter indicates significant differences (*p* < 0.0001) between accessions and growing conditions between cabbage morphotypes. See Table 1 for full names of cabbage accessions.

The myrosinase activity of TC-PCM, RC-RL and WC-CRB accessions did not differ significantly between field and glasshouse grown cabbages. Authors have previously reported that growing/environmental conditions affect myrosinase activity in *B. oleracea* species [24–26,29], and the results obtained from this study agree with their reports. The lower myrosinase activity of glasshouse cabbages might have been due to higher growth temperatures than those grown in the field. Minimum and maximum glasshouse temperatures were 14 and 43 ◦C, respectively, while minimum and maximum field temperatures were 6 and 24 ◦C, respectively (Supplementary Table S1). There are several possible reasons for the differences observed. One hypothesis could be that high temperatures reduced myrosinase enzyme synthesis or led to its more rapid denaturation. Another possible reason may have been that the process of synthesis and degradation of the enzyme (turn-over rate) was occurring faster at the higher growth temperatures, meaning that the plant did not accumulate a pool of enzymes at any one time. However, given that we can only see a snapshot in time when plants are sampled for enzyme assays, and each accession was harvested just once at a consistent time of day, it is not possible to infer the kinetics of these reactions occurring within the plant from the data in the present study. The kinetics of myrosinase synthesis and degradation within the plant is an area that warrants further study. Penas et al. [29], in their study of cabbages grown in different parts of Spain, reported that myrosinase activity was lower in cabbages grown in eastern Spain that were exposed to a higher growing temperature when compared to those grown in northern Spain with lower growing temperatures. It is, however impossible to say unequivocally that the lower myrosinase activity observed in the glasshouse samples is as a result of higher growth temperatures and not due to other stress factors, as we were unable to grow the plants in the glasshouse under lower temperatures similar to those observed in the field due to unavailability of cooling facilities within the glasshouse used in the study.

Another possible reason for the significantly lower enzyme activity in glasshouse cabbages could be due to stress factors during growth. Glasshouse cabbages were grown in pots, which may have led to stress from restricted root volume and reduced the amounts of nutrients (sulphur and nitrogen) available, potentially resulting in fewer enzymes and substrates being synthesized. Cabbage grown in the glasshouse achieved a lower above ground biomass than the field grown ones, indicating some form of stress. This was also evident in the differences in size of the closed heart cabbage heads, with the glasshouse plants having smaller heads than the field plants, as reported in Section 2.1. Their leaves appeared to be thinner and less robust than the field cabbages, as is often found in plants grown in protected environments that are not exposed to stimuli, such as wind, which for decades has been known to encourage the formation of thicker cell walls and smaller cells [66]. Hirai et al. [67] found that under nitrogen and/or sulfur limiting growth conditions, genes encoding myrosinase enzyme synthesis were down-regulated in Arabidopsis in order to facilitate storage of these elements in the form of glucosinolates in the leaf tissue. Yuan et al. [68] and Rodríguez-Hernández et al. [69] showed that salt stress reduced myrosinase activity in radish sprouts and broccoli, respectively. Pests and insect attack in field cabbages may have also led to higher myrosinase synthesis and/or accumulation in the cabbages. Accessions that did not show significantly different myrosinase activities between the two growing environments, or in the case of WC-FEM, higher myrosinase activity in glasshouse samples, might have been able to tolerate the glasshouse conditions and may have found it conducive for growth, while accessions that did not survive in the glasshouse may have found the conditions too harsh. Increased myrosinase activity as a result of abiotic stress, such as salt, temperature and drought, has been reported in various *Brassicaceae* species [70–72]. Increased myrosinase activity would result in enhanced glucosinolate hydrolysis to beneficial isothiocyanates, which would not only be beneficial to consumers but would also serve as defence compounds for the plants, thereby protecting them against insect and pest attacks.

#### *3.2. Protein Content and Specific Myrosinase Activity of Glasshouse and Field Grown Cabbages*

The protein content and specific activity of myrosinase for all accessions and growing conditions studied are presented in Table 4. The protein content and specific activity of samples studied were significantly (*p* < 0.05) affected by growing conditions and cabbage accession. Protein content did not correlate with myrosinase activity.

**Table 4.** Protein content ((mg/g ± SD) DW) and specific activity ((U/mg soluble protein ± SD) DW) of cabbage accessions grown in the glasshouse and on the field.


Values are means of three processing replicates and two technical replicates (*n* = 6 ± SD). SD: standard deviation from mean; dng: did not grow. Letters "A-E": mean values not sharing a common uppercase letter differ significantly (*p* < 0.05) between accessions and growing condition within a cabbage type for each parameter (i.e., protein content and specific activity). Letters "a-q": mean values not sharing a common lowercase letter differ significantly (*p* < 0.05) between cabbage types, accessions, and growing condition for each parameter (i.e., protein content and specific activity). See Table 1 for full names of cabbage accessions.

> Savoy and white cabbage accessions, which had the highest myrosinase activity, had the lowest protein contents. Just like myrosinase activity, the protein content of glasshouse samples was significantly lower than the field samples. This might be as a result of plant stress during growth, which prevents the plant from producing more nutrients than required or using up its stored nutrients in order to survive, as previously discussed in Section 3.1. Plant proteins have been reported to react negatively to environmental stress [26]. The results obtained are in agreement with Rosa and Heaney [73], who reported higher protein contents in Portuguese cabbage grown in lower environmental temperatures compared to those grown in higher temperatures.

> Specific activity of the cabbages was similar to the myrosinase activity and protein content, with field grown cabbages generally having higher specific activity than the glasshouse cabbages. Savoy and white cabbage accessions had significantly higher specific activities than other cabbage morphotypes, as indeed both were found to have significantly higher total myrosinase activity (Figure 1). White cabbage has previously been reported to have higher specific activity than red cabbage [28], which is in agreement with the results of this study. However, a study conducted by Singh et al. [38] showed red cabbage with a higher specific activity than white and savoy cabbage. This might have been due to the

differences in varieties studied or protein content of the cabbages, which was not reported in their study.

#### *3.3. Effect of Cabbage Morphotype and Accession on GSL Profile and Concentration of Field Grown Cabbages*

GSL profiles across cabbage accessions are presented in Figure 1; the statistical output of significant differences within and between cabbage morphotypes is documented in Supplementary Table S2. In total, nine different GSLs were identified across all accessions tested (Table 2): 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). PROG, GIBN and GRPN were the most abundant GSLs across all accessions studied, with 4-HOH, GIBVN and GER being the least abundant. 4-HOH was present in negligible amounts (<1.0 μmol/g DW) in all accessions, contributing not more than 1% to the total GSL content of the cabbages. When considering the ratio of total aliphatic to indole GSL concentrations in the accessions, over 60% of total GSL concentration was made up of aliphatic GSLs, with less than 30% from indole GSLs, with the exception of savoy SC-PW accession, where indole GSL comprised 36% of total GSL concentration (Supplementary Table S2).

GSL profiles and concentrations varied across cabbage accessions and differed significantly (*p* < 0.05) in some cases between and within cabbage morphotypes and accessions. Only five of the nine individual GSLs identified in the cabbages studied were found in black kale accessions:—GIBN, GRPN, GBSN, 4-HOH and GER—the last of which was present in BK-CNDTT alone. GRPN was the major GSL present in black kale accessions, consisting of over 50% on average of the total GSL content of black kale. The proportion of GRPN is similar to those previously reported by Kushad et al. [74] but much higher than those reported by Cartea et al. [40]. Previous studies detected SIN and PROG in kale and reported SIN as the main GSL in kale varieties [32,40,74]; however, SIN and PROG were not detected in the current study. There was a significant difference in total and individual GSL concentrations within black kale accessions, except for 4-HOH, which did not differ significantly (*p* = 0.401). BK-CPNT had the highest total GSL content (47.5 μmol/g DW).

GIBVN and GER were not identified in any of the wild and tronchuda cabbage accessions studied, while GIBN and GRPN were identified in all accessions except for WD-8707 accession. The concentration of individual GSLs differed significantly (*p* < 0.0001) across all wild and tronchuda cabbages. PROG and GPN were the most abundant GSLs in WD-8707 and WD-8714, while PROG and GRPN were the most abundant in WD-GRU. In tronchuda cabbages, SIN, GIBN and GBSN were at the highest concentrations, with SIN comprising up to 42% in TC-T.

A previous study [40] on GSL profile and concentrations in tronchuda cabbage identified 14 GSLs, compared to seven found in this study. However, GER was not identified in both studies, and proportions of the individual GSLs identified in both studies were similar.

The total GSL content of wild and tronchuda accessions differed significantly (*p* < 0.01 and *p* < 0.0001, respectively) between accessions within each cabbage morphotype. The most abundant GSLs in savoy cabbages were GIBN, SIN and GBSN, with GIBN concentrations as high as 61.3 μmol/g DW (57% of the total GSLs) in SC-SDG. GER was not identified in savoy accessions, and GPN was present in very low amounts in SC-SDG only. Similar proportions of savoy GSLs were reported by Ciska et al. [41] and Hanschen and Schreiner [43], but in both studies more individual GSLs were identified in the savoy varieties investigated than those reported in this study. For example, both studies identified GER in savoy cabbages, although present in trace amounts in the Ciska et al. [41] study. The total GSL content of savoy cabbages ranged from 47.6 μmol/g DW to 108.5 μmol/g DW. SC-SDG accession had significantly higher (*p* < 0.0001) total GSLs than SC-HSC and SC-PW, with SC-HSC having significantly lower total GSLs than the other two accessions.

In red and white cabbages, PROG, GIBN and GRPN were the most abundant GSLs. GBSN was also abundant in WC-CRB and RC-RL accessions, while GER was not identified

in either accession. The concentrations of GRPN, GIBVN and GER did not differ significantly between red cabbage accessions. WC-CRB had significantly higher amounts of SIN, GIBN, GBSN and total GSL compared to WC-FEM, but differences in PROG and GRPN content were not significant. The total GSL content of RC-RL was significantly (*p* < 0.0001) higher than the other two red cabbage accessions. The results obtained agree with those previously reported [22,41,43]. However, a few studies disagree with the findings of this study; a previous study conducted by Park et al. [75] quantifying red cabbage GSL reported SIN absent in red cabbage, while Zabaras et al. [76] found GPN as the most abundant GSL in red cabbage.

Individual GSLs and total average GSL concentrations differed significantly (*p* < 0.0001) across all accessions, irrespective of cabbage morphotype. Total average GSL concentrations of accessions studied ranged from 18.9 μmol/g DW (BK-CNDTT) to 163.1 μmol/g DW (WD-8714). These differences were due to variations in GSL profiles and concentrations of individual GSLs. Wild cabbages generally had higher total GSL concentrations (Figure 2b) than other cabbage morphotypes, and these high concentrations were driven by significantly higher amounts of PROG in wild cabbages. Lower concentrations of total GSL observed in black kale accessions (18.9 μmol/g DW to 47.5 μmol/g DW) were due to lower numbers and concentrations of individual GSLs compared to the other cabbage morphotypes studied (Figure 2a). The variability in GSL concentrations between and within cabbage morphotypes and accessions is in agreement with previous reports that GSL profiles and concentrations vary between *Brassica* species and varieties [5,29,39,40,43,52,77]. The difference in GSL profiles of *Brassica* vegetables has been linked to genetic factors, while interactions between environmental and genetic factors are largely responsible for differences in GSL concentrations [8]. In general, concentrations of individual and total GSL of the gene bank cabbages reported in this study are much higher than those reported for commercial and gene bank cabbage varieties/accessions in the literature [29,40,41,43,74]. One reason for this may be due to the different varieties/accessions studied, implying that gene banks may indeed be a useful source from which to select accessions with higher GSL concentrations.

Differences in postharvest handling/time could have also contributed to the higher abundance of GSLs observed in the current study. Most varieties used in the literature were obtained from the supermarket and would have gone through a standard commercial supply chain upon harvest, unlike the samples used in this study, which were transferred to the laboratory immediately after harvest. The absence of commercial postharvest storage and handling processes in the current study could account for the differences observed between the samples and those reported in the literature. Total GSL abundance has been shown to decrease in *Brassica* vegetables stored for 7 days at 4–8 ◦C [78]. Lastly, differences in the conditions under which the plants were grown and/or harvested could also be responsible for the variations in GSL concentrations observed. This suggests that it is not only important that the right accession/variety is selected, but it must also be grown under optimal conditions and given as short a supply chain as possible to achieve optimum GSL abundance in the plants. The higher GSL concentrations in the present study can enhance the potential health benefits that may be derived from their consumption.

The differences in GSL profiles and concentrations of the accessions studied can potentially influence the sensory and health properties of the cabbages. For example, the absence of SIN and PROG in black kale accessions and higher concentrations of PROG reported in wild cabbage accessions may potentially influence the sensory characteristics of these cabbages, given SIN and PROG have been linked with bitter taste in *Brassica* vegetables [22,79]. On the other hand, higher amounts of GRPN (the precursor GSL for SFP formation linked to several health promoting properties of *Brassicas*) in kale, red and white cabbages could enhance the potential health benefits derived from their consumption [80]. The differences in cabbage accessions, growing conditions and geographical location, as well as environmental factors during cabbage cultivation, all play a vital role in GSL profile and concentration and therefore make comparing results between different studies difficult.

**Figure 2.** Glucosinolate concentrations (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 grown in the field and glasshouse. Error bars represent standard deviation from mean values. Letters above bars refer to differences in total GSL concentration. Letters "A-D": bars not sharing a common uppercase letter differ significantly (*p* < 0.05) between accession and growing conditions within a cabbage morphotype (i.e., within each separate graph). Letters "a-q": bars not sharing a common lowercase letter differ significantly (*p* < 0.0001) between cabbage morphotypes, accessions, and growing conditions (i.e., between the separate cabbage morphotype graphs). Abbreviations: F = Field, G = glasshouse; dns = did not survive. For abbreviations of accessions and compounds see Table 1 (cabbage accessions) and Table 2 (GSLs).

#### *3.4. Effect of Growing Conditions on GSL Concentrations in Cabbage Accessions*

The effect of growing conditions on GSL concentration is presented in Figure 1, with significant differences within and between cabbage morphotypes presented in Supplementary Table S2. While the GSL profile of cabbage accessions studied did not differ between growing conditions, there was a difference in GSL abundance between glasshouse and field grown cabbages. Total GSL concentrations in field grown samples ranged from 18.9 μmol/g DW (BK-CNDTT) to 163.1 μmol/g DW (WD-8714) and glasshouse samples from 8.81 μmol/g DW (BK-CNDTP) to 105.5 μmol/g DW (WD-8707). WD-8714 had significantly (*p* < 0.0001) higher concentrations of total GSLs compared to all other accessions, and this was largely due to the abundance of PROG and GPN, making up 83% and 69% of total GSLs in field and glasshouse samples, respectively.

Cabbages grown in the field had higher total GSL concentrations than glasshouse samples across most accessions studied, with a few exceptions (BK-CNDTT, TC-T, SC-HSC, and RC-RD), where total GSL concentrations were higher in glasshouse samples. These differences were significant in some but not all cases. Growing conditions significantly affected individual GSL concentrations between and within cabbage morphotypes and accessions. With the exception of black kale accessions, both field and glasshouse cabbages were predominantly abundant in aliphatic GSLs, with averages of 82 and 78%, respectively across all accessions, while indole GSLs comprised only 18 and 22% of total GSLs in field and glasshouse samples, respectively. In black kale accessions, however, growing conditions seemed to influence the ratio of aliphatic to indole GSL present in the samples. All black kale accessions grown in the glasshouse had much higher total indole GSL, with up to seven-fold differences reported in BK-CNDTP samples (Supplementary Table S2). The differences observed are mainly due to differences in the ratio of individual aliphatic to indole GSL present in the samples and not higher concentrations of indole GSLs in the glasshouse samples, as there was no significant difference observed in the concentrations of the most abundant indole GSL, GBSN, present in the samples between growing conditions (except for BK-CNDTT).

There was no clear pattern for the abundance of individual GSLs, as some GSLs were significantly higher in glasshouse samples for some accessions, but lower or not significantly different in others. PROG and GRPN were either significantly higher in field samples or did not significantly differ from glasshouse samples within accessions, except for RC-RD accession, where GRPN was significantly higher (*p* < 1.0001) when grown in the glasshouse. GRPN abundance in BK-CNDTP and BK-CPNT field grown accessions was up to 90% more than the corresponding glasshouse grown cabbages. GBSN was the most stable GSL across growing conditions, as there was no significant difference (*p* = 0.101) in GBSN between field and glasshouse cabbages.

Growing conditions such as growth temperature and photoperiod have been shown to influence the abundance of GSLs. There are several possible reasons for the differences observed in GSL concentrations in the different growing conditions. The higher total GSL content reported in most field samples could be due to production of higher amounts of GSLs by the plant in response to insect and pest attack on the field when compared to glasshouse samples. GSL compounds are plant metabolites produced by plants for defence against stress and attack from insect and pests [8,81]. In addition, the higher amount of GSLs in field samples could also be due to the lower average temperatures during growth (6 to 24 ◦C) compared to the higher temperatures in the glasshouse (14 and 43 ◦C) (Supplementary Table S1). Growth temperatures have been reported to influence GSL concentrations in *Brassica* vegetables. *Brassica* vegetables are generally thought to be cool weather crops, with average growing temperatures between 4–30 ◦C [82]. The optimum temperature for growth varies between different types of *Brassicas* and going below or above that temperature could affect GSL concentrations. The exact mechanism of GSL biosynthesis under different temperature conditions is unclear because of several interacting factors, such as drought and photoperiod, but it has been reported that plant stress due to high or low growing temperatures may enhance activities of transcription factors

such as *MYC2* and *MYB28*, which promote GSL biosynthesis [42,83]. Literature studies have, however, generally reported higher GSLs at higher growing temperatures; Rosa and Rodrigues [27] reported a higher GSL content in young cabbage plants when grown at 30 ◦C compared to 20 ◦C. Lower GSL concentrations was reported in kale grown at lower temperatures compared to those grown at higher temperatures [32,33]. In addition, several authors have reported higher GSL concentrations in spring/summer grown cabbages (average temperatures between 25–30 ◦C) compared to autumn grown plants (temperatures < 20 ◦C) [29,39–41]. The lower amounts of GSL accumulated in glasshouse plants could also be the result of plant growing conditions. Glasshouse samples were grown in pots with drainage holes to allow excess water to seep out. However, this could have also led to sulphur leaching, leading to sulphur deficiency in the soil, and plants were not fed with sulphur fertilizers. Sulphur is a major precursor for GSL biosynthesis, and its deficiency has been reported to reduce GSL concentrations in *Brassica* plants, especially aliphatic GSLs, as sulphur deficiency limits methionine synthesis (basic substrate for aliphatic GSL biosynthesis) as opposed to tryptophan, a non-sulphur amino acid and precursor for indole GSL biosynthesis [84]. On average, reduced amounts of aliphatic GSLs were accumulated in glasshouse plants compared to field plants, while glasshouse samples accumulated higher amounts of indole GSLs than field samples. Sulphur was reported to influence the aliphatic GSL concentrations in rapeseed more than indole GSL [84]. However, glasshouse plants, which had significantly higher GSL concentrations compared to their field counterparts, may have found the glasshouse conditions more favourable than other accessions, which resulted in enhanced GSL production.

The results of this study show that cabbages differ in their requirements for growth, and it is important to plant cabbage accessions in growing conditions that are best suited for their maximum development, as individual plants respond differently under different environmental conditions. Optimizing agronomy practices and applying limited abiotic stress in a controlled manner could be a way of increasing myrosinase activity and GSL production in some *Brassica* species.

#### *3.5. Effect of of Cabbage Morphotype and Accession on Glucosinolate Hydrolysis Products (GHPs) of Field Grown Cabbages*

A total of 22 GHPs were identified and quantified from the cabbage accessions studied, comprising 11 ITCs and 11 nitriles/epithionitriles (Table 3). Concentrations of GHPs are presented in Figure 2, with significant differences between and within cabbage morphotypes and accessions presented in Supplementary Table S3. Results are expressed as sulforaphane equivalents.

The type and concentration of GHPs formed differed between cabbage accessions. Predominant GHPs did not differentiate between accessions within a cabbage morphotype but varied across cabbage morphotypes. There was a significant difference in the concentrations of individual and total GHPs formed within and between cabbage morphotypes and accessions (Figure 3 and Supplementary Table S3). Wild cabbage accessions had the highest levels of GHPs formed (8.79 μmol/g DW—8.6 μmol/g DW; Figure 2b) and tronchuda accessions the lowest (0.95 μmol/g DW—3.27 μmol/g DW; Figure 2c).

GHPs of GRPN and GBRN were the main GHPs detected in black kale accessions, with nitrile concentrations accounting for 74–89% of the total GHPs. BK-CPNT accessions had significantly lower total GHPs than BK-CNDTP. Isomers of CHETB, nitriles of PROG hydrolysis, were the most abundant GHPs formed in wild cabbages, except for WD-GRU, which had higher amounts of GN (PROG ITC) compared to the nitriles formed. This was unexpected, and it is unclear why this happened, because more nitriles than ITCs were formed for other GSLs present in the same sample. A possible explanation for this could be the activity of epithiospecifier modifier proteins (ESMs), enhancing the activity of specific myrosinase isoenzymes, which hydrolyse PROG present in the samples. ESM inhibits the activity of ESP, preventing the formation of nitriles and epithionitriles, and instead promotes ITC formation [15,85,86]. GN have been associated with bitter taste [87] and adverse effects on thyroid metabolism, leading to goitre formation. The reports on goitre

formation are limited and based on animal studies, which show that average daily intake is not enough to produce adverse effects in humans [8]. However, to limit the health risks, genetic manipulation and selective breeding methods used to increase GRPN contents by threefold in '*Beneforte*' broccoli [88] could be employed to reduce PROG contents in the wild accessions. The main GHPs of tronchuda accessions were CETP and IBN, nitriles of SIN and GIBN, respectively. Total GHPs of TC-CPDP were significantly higher than TC-T. IBN and IB (GIBN hydrolysis products) were the most abundant GHPs in savoy cabbages, and SFP and SFN (hydrolysis products of GRPN) the most abundant in red and white cabbages.

In savoy, SC-HSC varied significantly from SC-PW and SC-SDG accessions, containing up to 60% more GHPs than the other two accessions. The much lower concentrations of GHPs in SC-PW compared to SC-HSC were unexpected due to similar concentrations of GSLs in both accessions. A similar trend was noticed between WC-CRB and WC-FEM accessions, where much lower GHPs were formed in WC-CRB accession, with significantly higher GSLs than WC-FEM. This might be related to variation in myrosinase and ESP activities within the samples. As previously discussed in Section 3.1, WC-FEM had significantly higher myrosinase activity than WC-CRB (Figure 1), which may explain the higher concentrations of GHPs formed. However, this is not the case in savoy cabbages, as SC-PW had the highest myrosinase activity (see Figure 1). It is hypothesized that myrosinase isoenzymes and ESP of SC-PW accession may be less stable than the other accessions and was, therefore, denatured before permitting full hydrolysis. As previously discussed, ESM activities promoting ITC formation may also be responsible for the higher GHP concentrations observed. For example, although GIBN concentration in WC-FEM was significantly (*p* < 0.0001) lower than that of WC-CRB, the amount of IB, the ITC formed from GIBN, was significantly (*p* < 0.0001) higher in WC-FEM than in WC-CRB. Another possible reason for the variation in GHP concentrations could be due to the type of myrosinase isoenzyme present within the samples. It has been reported that myrosinase isoenzymes differ in the rate at which they hydrolyse individual GSLs, though little is known of their substrate specificity. James and Rossiter [89] found that in the presence of ascorbic acid, two myrosinase isoenzymes identified in *Brassica napus* L. differed in the way they degraded SIN and neoglucobrassicin (NEO), with SIN being degraded more rapidly than NEO by both isoenzymes under the same conditions. While there are limited studies on the conversion ratio of GSLs to GHPs, studies on GHP formation in *Brassica oleracea* [43] and rocket salad [42] have shown that conversion of GSLs to GHPs is not always a linear reaction and GHP concentrations are generally much lower than the precursor GSL concentrations.

Several GHPs were identified in cabbage accessions where their GSLs were not detected: tiny amounts of 3BITC (GPN hydrolysis product) were formed in BK-CNDTT; 4MBN (nitrile of GIBVN) in tronchuda; EVN (GPN nitrile) in savoy cabbages; and ER and ERN (GER GHPs) in red and white cabbages. PEITC and BPN (GHPs of gluconasturtiin), PITC and BAN were also formed in most accessions. This could be due to concentration of the respective GSLs being below the limits of detection of the LC-MS2 instrument used. A previous study of turnips detected GHPs of glucoberteroin, though the intact GSL was not detected [90]. A recent study on horseradish, wasabi, watercress, and rocket also detected GHPs, where their intact glucosinolates were not identified [91]. The profile of GHPs in this study is in agreement with the study of Hanschen and Schreiner [43]. However, in their study, they found CETP (nitrile from SIN hydrolysis) as the main GHP in savoy, red and white cabbages, which is inconsistent with this study, where GIBN GHPs (IB and IBN) and GRPN GHPs (SFP and SFN) were the main compounds detected. This difference can be attributed to the different varieties/accessions studied.

In general, the relationship between individual GSLs and their corresponding GHPs within an accession was as expected, where the dominant GSL resulted in their corresponding dominant GHPs, which is helpful in confirming the efficiency and accuracy of the GHP extraction method. Overall, nitriles and epithionitriles were the major hydrolysis products formed across all cabbage accessions, as has been reported previously in raw

cabbage [62,92]. This is due to the activity of ESP and other nitrile forming proteins present in the samples, which hydrolyse GSLs to epithionitriles and nitriles instead of the more beneficial ITCs [92].

**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 grown in the field and glasshouse. Error bars represent standard deviation from mean values. Letters above bars refer to differences in total GHP concentration. Letters "A-D": bars not sharing a common uppercase letter differ significantly (*p* < 0.05) between accessions and growing conditions within a cabbage morphotype (i.e., within each separate graph). Letters "a-l": bars not sharing a common lowercase letter differ significantly (*p* < 0.0001) between cabbage morphotypes, accessions, and growing conditions (i.e., between the separate graphs). Compounds with colour shades similar to one another are GHPs of corresponding GSLs presented in Figure 2. Abbreviations: F = Field; G = glasshouse; dns = did not survive. For abbreviations of accessions and compounds see Table 1 (cabbage accessions) and Table 3 (GHPs).

#### *3.6. Effect of Growing Condition on GHP Concentrations*

GHP profile and concentration in the two different growing conditions studied is presented in Figure 3, with the significant differences between growing conditions reported in Supplementary Table S3. The profile of the GHPs detected were similar between growing conditions, with a few exceptions. For example, BPN was identified in black kale field samples but not detected in glasshouse samples. GHP concentrations in field and glasshouse ranged from 0.95 μmol/g DW (TC-T) to 18.6 μmol/g DW (WD-8707) and 0.59 μmol/g DW (BK-CNDTP) to 15.9 μmol/g DW (WD-8707), respectively. Within accessions, total GHP accumulation was significantly higher in field plants than glasshouse, except for wild cabbage accessions, TC-PCM and WC-CRB, where total GHPs were higher in glasshouse samples; however, the differences were not significant, except in WC-CRB, where a significant difference was observed. Generally, total GHP concentrations followed a similar pattern to total GSLs, with a few exceptions. For example, the BK-CNDTT glasshouse sample had significantly lower total GHPs compared to the field sample (Figure 3a), despite the significantly higher total GSL in the glasshouse sample (Figure 2a). Significantly higher myrosinase activity, and possibly ESP activity, in the BK-CNDTT field compared to glasshouse sample may have led to the formation of more GHPs (Figure 1). A similar trend was observed in savoy accessions, where an abundance of GSL under one growing condition did not necessarily result in higher amounts of GHP formed. The results obtained in our study are in agreement with those reported by Jasper et al. [42], where growth temperatures had different effects on the amount of GSL and GHPs formed in rocket salads.

In summary, the results of this study show the importance of having both high myrosinase activity and GSL accumulation in plants, as they have a direct impact on the amount of hydrolysis compounds formed. It is therefore important to ensure that cabbages are cultivated under optimised growing conditions (such as temperature, available sulphur/nitrogen and controlled biotic stress) that favour both high myrosinase and GSL accumulation and not only one or the other.

#### *3.7. Multifactor Analysis (MFA) of GSLs and GHPs Identified in Cabbage Accessions Grown under Two Different Conditions*

To investigate the underlying structure of the results, MFA was performed on the GSL and GHP data from the cabbage accessions. Figure 4 shows distribution of the cabbage accessions as well as the scores and loadings of MFA performed on the mean data of GSLs and GHPs. Dimensions 1 and 2 (F1, F2) explained 42% of the variance in the data, but other dimensions did not provide any new information; therefore, only F1 and F2 are presented and discussed. The plot demonstrates that individual GSLs were positively correlated with their corresponding GHPs. From the plot, it is clear that cabbages were mostly distinguished based on morphotype rather than accessions or growing conditions, except for wild cabbage accessions, where there was a clear separation of WD2 (WD-GRU) from WD1 (WD-8707) and WD3 (WD-8714).

Based on the MFA, samples were grouped into three distinct clusters: one cluster comprised of black kale, red cabbage, white cabbage and WD-GRU accessions, another tronchuda and savoy cabbage accessions, and the final cluster WD-8707 and WD-8714 accessions. Black kale, red cabbage, white cabbage and WD-GRU correlated positively with GRPN, GER, 4-HOH and their hydrolysis products. Tronchuda and savoy cabbage samples correlated positively with GIBN, GIBVN, SIN and their hydrolysis products. WD1 and WD2 correlated positively with GPN and PROG and their nitriles, as well as total GSLs and GHPs, but was negatively correlated with black kale, red cabbage, white cabbage and WD-GRU accessions. An additional Pearson correlation demonstrating significant correlations (*p* < 0.05) between various GSLs and GHPs is presented in Supplementary Table S4. GIBN correlated negatively (r2 <sup>&</sup>gt; −0.3; *<sup>p</sup>* < 0.01) with PROG and its hydrolysis products, GPN and its hydrolysis products, and PITC. On the contrary, GPN was strongly positively correlated (r<sup>2</sup> > 0.6; *p* < 0.0001) with PROG and its hydrolysis products, EVN, PITC, total GSL and total GHPS. Total GSLs were significantly positively correlated (r2 = 0.5; *p* < 0.01) with total GHPs. Strong significant positive correlations (r<sup>2</sup> > 0.5; *p* < 0.05) were observed between individual GSLs and their corresponding GHPs. For example, GRPN was positively correlated with SFP and SFN (r2 > 0.5 and 0.8; *p* < 0.01 and *p* < 0.0001 respectively).

**Figure 4.** MFA map of glucosinolates and glucosinolate hydrolysis products (**a**) distribution of variables and (**b**) sample distribution. For codes and distribution on plot, refer to Table 1 (cabbage accessions) and Tables 2 and 3 (compounds). Compounds with different shades of the same colour in Figure 3a refer to the GSL and corresponding GHP. Key: F = Field ; G = Glasshouse; • GSL = Glucosinolates; • GHPs = Glucosinolate hydrolysis products; • BK= Black kale; • WD = Wild cabbage; • TC= Tronchuda cabbage; • SC = Savoy cabbage; • RC = Red cabbage; • WC = White cabbage.

It is obvious that the separations observed between samples are mainly driven by differences in GSLs and GHPs most accumulated in the samples: GN, GRPN, GER, 4-HOH and their GHPs in black kale, red cabbage, white cabbage and WD-GRU accessions; GIBN, SIN, GIBVN and their GHPs in tronchuda and savoy cabbage accessions; and lastly, PROG, GPN and their GHPs in WD-8707 and WD-8714 accessions. WD-8707 and WD-8714 had the highest concentration of total GSLs and GHPs, and this was responsible for the positive correlation of these accessions to total GSLs and GHPs observed. It is worth mentioning that PROG and CHETB, which were largely responsible for the high concentrations of total GSLs and GHPs in these accessions, correlated positively with total GSLs and GHPs. The result obtained provides a clear picture of the similarities and differences in GSL and GHP profile and concentrations of the different cabbage morphotypes and accessions studied.

Like any other study, some limitations were encountered in this study. First, the cabbage seeds used in the study were obtained from a gene bank. This means they have not been bred for uniformity in terms of plant characteristics and abundance of phytochemical compounds. Breeding programmes to date have mostly focused on developing disease-resistant and environmentally resilient crops, with less emphasis on the content of phytochemical compounds. This implies that there may be large variations in phytochemical compounds between cabbage heads/plants of the same accession, as has been observed in *Marathon* broccoli heads [93], and this may have influenced the results obtained in the present study. To reduce the effects of possible variation between plant heads, four to five heads were mixed together to obtain a representative sample. However, considering the amounts of heads used during the study, some variations may still have existed within the samples.

Second, the GC-MS method used for GHP analysis was long and required several steps to ensure that all GHPs present in the sample could be identified. However, some GHPs may have been lost or converted into other compounds in the process due to their very volatile and unstable nature. Though care was taken during the analysis to prevent losses, the rigorous analytical method may have led to some losses of the more volatile compounds.

#### **4. Conclusions**

In line with the primary hypothesis of the study, the results demonstrated that myrosinase activity as well as profiles and concentrations of GSLs and GHPs were all influenced by growing conditions, cabbage morphotypes and accession. However, in agreement with our secondary hypothesis, the profile and concentration of GSLs and GHPs formed were substantially more influenced by cabbage morphotype than accession. The study showed that planting cabbages in high growth temperatures and stressful conditions resulted in lower myrosinase activity. Myrosinase activity differed between accessions and cabbage morphotypes, although morphotype tended to have the more significant impact. Savoy cabbage accessions had the highest myrosinase activity, while black kale accessions had the lowest myrosinase activity.

The concentration and profile of GSL and GHP compounds accumulated differed between growing conditions and accessions, within and across cabbage morphotypes. While genetic factors had more influence on the GSL profile of the cabbages, differences in the GSL concentration were more affected by environmental factors during growth, which agrees with previous studies [8]. Growing conditions and cabbage accessions seem to have different effects on GSL and GHP formation, with higher GSL concentrations observed within a growing condition or accession not always resulting in a corresponding greater accumulation of GHPs and vice versa. Results obtained from the study showed that a possible reason for the higher GHP concentrations could be higher myrosinase activities in accessions with lower GSLs, as was observed in white cabbage and black kale accessions. However, this this was not the case in all accessions, suggesting there may be other reasons for the differences obtained. The results obtained therefore suggest that it would be incorrect to assume that higher myrosinase activity and/or GSL accumulation would automatically always result in high concentrations of GHPs.

Variations in the GSL and GHP contents imply differences in the potential healthpromoting and sensory characteristics of the cabbages studied. For example, the high amounts of SFP present in red and white cabbages could potentially provide more health benefits on consumption when compared to other accessions. Conversely, high concentrations of PROG and GN (compounds linked to bitter taste) in wild accessions may reduce consumer acceptance and liking. However, the contents of GSLs and ITCs in *B. oleracea* vegetables alone does not provide a clear picture of the sensory characteristics of *B. oleracea* vegetables, as other compounds in the plant matrix, such as sugars and sweet tasting amino acids, can influence and modulate the sensory perception of these vegetables, as has been shown in previous studies on *Brassicas* and other crops such as lettuce [22,94–96].

Field grown cabbages had much higher GSLs and GHPs than glasshouse plants, with a few exceptions (SC-HSC and RC-RD). However, the biggest differences observed were between cabbage morphotypes, irrespective of the conditions under which they were grown. The result of this study suggests that cabbage morphotype and accession might be more important factors for GSL and GHP profiles of plants than the conditions under which they are grown. All individual GSLs and their corresponding GHPs were identified in the accessions studied, and a correlation between GSLs and GHPs was found. The difference in myrosinase activity and GSL and GHP concentrations could not be linked to morphology of head formation (closed heart or open leaf). The influence of growing conditions on cabbage biochemistry will be an important consideration, as the use of highly protected environments for crop production becomes more prevalent through indoor farming, which will also lead to breeding of cabbages with more compact morphology. Our data indicate that protected conditions need to be optimised, possibly by inclusion of controlled abiotic stress, in order to generate the GSL abundance that is observed in field grown crops.

Aliphatic GSLs, nitriles and epithionitriles were the most abundant compounds identified. The results suggest that consumption of raw cabbage may provide limited health

benefits, as more nitriles and epithionitriles are formed than the more beneficial ITCs. It is therefore recommended to process the cabbages in ways that ensure hydrolysis of GSL to ITCs rather than nitriles. Despite the high amounts of nitriles and epithionitriles formed overall, high amounts of health beneficial SFP were detected in some red and white cabbage accessions. The result suggests that some gene bank accessions can be a good source of beneficial compounds and could be used in breeding programmes to introgress areas of the genome that regulate these compounds from the gene bank accessions into elite commercial cultivars. This can also be helpful for selection of more beneficial accessions for commercial cultivation and production. Given that accessions with lower GSL concentrations and higher myrosinase resulted in high GHP concentrations for some of the accessions studied, breeding programmes should not on only focus on selection of accessions with high GSL concentration but should also consider accessions that have high myrosinase activity and ESM, if maximum conversion of GSLs to ITCs is to be achieved.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/foods10122903/s1, Figure S1: Cross-section of planted cabbage morphotypes (**a**) Black kale (**b**) Wild cabbage (**c**) Tronchuda cabbage (**d**) Savoy cabbage (Field grown) (**e**) Savoy cabbage (Glasshouse grown) (**f**) Red cabbage (Field grown) (**g**) Red cabbage (Glasshouse grown) (**h**) White cabbage (Field grown) (**i**) White cabbage (Glasshouse grown). Figure S2: Cross-section of cabbages grown under (**a**) Controlled environment and (**b**) Glasshouse. Figure S3: Examples of GC-MS chromatograms for field and glasshouse grown 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: Climatic data of field and glasshouse cabbages. Table S2: Glucosinolate concentration in cabbages grown under different conditions (mg/g DW). Table S3: Glucosinolate hydrolysis products concentration in cabbages grown under different conditions (μg/g DW sulforaphane equivalent). Table S4: Pearson correlation matrix table showing correlations between glucosinolates and glucosinolate hydrolysis products identified in cabbage grown under two different conditions (a) correlation coefficients (r) and (b) significance of the correlation (p value).

**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, 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 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* **Important Odorants of Four Brassicaceae Species, and Discrepancies between Glucosinolate Profiles and Observed Hydrolysis Products**

**Luke Bell 1, Eva Kitsopanou 2, Omobolanle O. Oloyede <sup>2</sup> and Stella Lignou 2,\***


**Abstract:** It is widely accepted that the distinctive aroma and flavour traits of Brassicaceae crops are produced by glucosinolate (GSL) hydrolysis products (GHPs) with other non-GSL derived compounds also reported to contribute significantly to their aromas. This study investigated the flavour profile and glucosinolate content of four Brassicaceae species (salad rocket, horseradish, wasabi, and watercress). Solid-phase microextraction followed by gas chromatography-mass spectrometry and gas chromatography-olfactometry were used to determine the volatile compounds and odorants present in the four species. Liquid chromatography-mass spectrometry was used to determine the glucosinolate composition, respectively. A total of 113 compounds and 107 odour-active components were identified in the headspace of the four species. Of the compounds identified, 19 are newly reported for 'salad' rocket, 26 for watercress, 30 for wasabi, and 38 for horseradish, marking a significant step forward in understanding and characterising aroma generation in these species. There were several non-glucosinolate derived compounds contributing to the 'pungent' aroma profile of the species, indicating that the glucosinolate-derived compounds are not the only source of these sensations in Brassicaceae species. Several discrepancies between observed glucosinolates and hydrolysis products were observed, and we discuss the implications of this for future studies.

**Keywords:** volatile compounds; odorants; glucosinolate; Brassicaceae; 'salad' rocket; wasabi; horseradish; watercress

#### **1. Introduction**

Crops of the Brassicaceae family are grown all over the world, and they form an important part of many different cuisines and cultures [1]. Some species are noted for their distinctive, and often very strong, tastes and flavours. *Armoracia rusticana* (horseradish), *Eruca sativa* ('salad' rocket), *Eutrema japonicum* (wasabi), and *Nasturtium officinale* (watercress) are four such examples, which are noted for their pungent, peppery, and aromatic organoleptic properties [2–4].

Horseradish and wasabi produce large roots that are grated and used as a condiment in many cultures across the world, most notably in Eastern Europe and the United Kingdom (horseradish) and Japan (wasabi; [5]). Horseradish is a vegetative perennial that grows widely in temperate regions [6], whereas wasabi can only be grown in a very few locations, owing to its sensitivity to temperature change and root oxygen availability [7]. Wasabi is traditionally cultivated in damp river valleys of Japan, although commercial operations have been established elsewhere, such as in the UK.

Salad rocket originates from the Middle East and has spread throughout the Mediterranean basin [8]. It has become naturalised on every inhabited continent and is considered

**Citation:** Bell, L.; Kitsopanou, E.; Oloyede, O.O.; Lignou, S. Important Odorants of Four Brassicaceae Species, and Discrepancies between Glucosinolate Profiles and Observed Hydrolysis Products. *Foods* **2021**, *10*, 1055. https://doi.org/10.3390/ foods10051055

Academic Editors: Franziska S. Hanschen and Sascha Rohn

Received: 8 April 2021 Accepted: 7 May 2021 Published: 11 May 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/).

an invasive weed in some regions. Watercress has similarly become naturalised (for example in North America) and grows in shallow rivers and streams. It can be cultivated commercially on large scales using artificial growing 'pools' flooded with stream water. Its leaves and shoots are a popular salad and sandwich garnish, and they can also be made into soups [9].

It is widely accepted that the distinctive aroma and flavour traits are produced by glucosinolate (GSL) hydrolysis products (GHPs [10]). GSLs are sulphur-containing secondary metabolites produced by Brassicales plants in response to biotic and abiotic stress [11]. Myrosinase enzymes are responsible for the hydrolysis of GSLs in water to form a plethora of GHPs, the conformation of which can be determined by the presence of enzyme cofactors, pH level, metallic ion concentration, and the precursor GSLs side-chain structure [8]. These products include: isothiocyanates (ITCs), nitriles, epithionitriles, indoles, oxazolidine-2 thiones, and other diverse products that result from tautomeric rearrangements (Figure 1).

**Figure 1.** The glucosinolate–myrosinase reaction and hydrolysis products. Abbreviations: ESM1, epithiospecifier modifier protein 1; TFP, thiocyanate forming protein; NSP, nitrile specifier forming protein; ESP, epithiospecifier protein.

Previous work has reported olfactometry data for each of these crops; however, data are generally very scarce. Only five studies of *E. sativa* volatile compounds have been published in the last twenty years [3,12–15]. Very little information regarding wasabi root volatile composition and aroma is available outside of Japanese language journals [7,16]. Similarly, very little information is available for watercress, with only four papers published in the last 40 years [17–20]. Horseradish is the most well characterised of these four species, but still, only six studies of note have been published in the last 50 years [6,21–25].

There is also an 'elephant in the room' regarding previous reports of GHPs present in aroma profiles of these Brassicaceae vegetables. Many of the reported GHP compounds are derived from GSL precursors that are not regularly reported as part of the profile for each respective species. In *E. sativa*, for example, GHPs such as methyl ITC, 3-butenyl ITC, 1-isothiocyanato-4-methylpentane, and 5-(methylsulfanyl)pentanenitrile have all been previously reported [13]. The GSL precursors to these compounds (glucocapparin, GCP; gluconapin, GNP; 4-methylpentyl GSL, 4MP; and glucoberteroin, GBT; respectively) have never been reliably or consistently reported as being part of the GSL profile in this species [26,27]. This begs the question whether these identifications are correct or

if the reports of GSL components in these species are incomplete. This may be due to a lack of sensitivity in reported mass spectrometry methods or because there is a lack of analytical standards to confirm compound identities. Another possibility is that there is a post-hydrolysis modification of GHPs, either by enzymatic means or through reactions with other phytochemical components, or as part of thermolytic reactions during gas chromatography. Very few examples of such modifications have been reported within the literature [28], but this may explain the presence of some GHPs within profiles of species where the GSL precursor is absent.

Other non-GSL derived compounds have also been reported to contribute significantly to the aromas of Brassicaceae crops. 2-Isopropyl-3-methoxypyrazine has been found to produce a strong pea, or green, pepper-like aroma in horseradish, for example ([6] Figure 2). In rocket, 'green-leaf' volatiles such as 3-hexenal and 1-penten-3-one have also been highlighted as having high odour potency [12]. The role of these compounds in aroma generation in GSL-containing crops is often not fully appreciated, and there are many diverse compounds with equally high odour intensities to GHPs present within the volatile bouquet.

**Figure 2.** Chemical structures of volatile compounds found in horseradish, rocket, wasabi, and watercress samples (numbers in parentheses refer to compound codes in Table 2); \* two separate peaks present for this compound; \$ tentatively identified.

The aims of this study were to (i) identify and describe in detail the key odorants of four Brassicaceae species (salad rocket, horseradish, wasabi, and watercress) by gas chromatography-olfactometry (GC-O) and gas chromatography-mass spectrometry (GC-MS), and (ii) associate the observed GHPs with their respective GSL profiles by liquid chromatography-mass spectrometry (UPLC-MS/MS). Our goal was to improve upon existing compound characterisation and odour descriptors for compounds in salad rocket, horseradish, wasabi, and watercress. The contribution of pungency by GHPs to aroma profiles is well studied; however, other aroma traits generated by non-GHPs are not well described for these species, and they likely create distinctive and subtle sensory characteristics. Additionally, detailed GSL compositions and MS/MS spectra for rocket, wasabi, and watercress is presented, highlighting discrepancies with observed hydrolysis products.

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

#### *2.1. Samples*

Individual horseradish and wasabi roots were purchased from Morrison's supermarket (Reading, UK) and The Wasabi Company (Dorchester, UK) respectively. Watercress was purchased as whole bags of leaves from ASDA supermarket (Reading, UK). Salad rocket was grown in controlled environment conditions at the University of Reading using seeds donated by Elsoms Seeds Ltd. (Spalding, UK) and designated RS4 and RS8. Seeds of each cultivar were sown into module trays containing peat-based seedling compost and germinated at 30 ◦C (daytime; 25 ◦C night). Lighting conditions were set to a long-day cycle (16 h light, 8 h dark). Light intensity was set to 380 μmol m−<sup>2</sup> s<sup>−</sup>1. Humidity was ambient. Plants were considered mature upon the development of 10 to 15 leaves. Harvested leaves were taken intact and placed inside a Zip-loc bag.

Roots and leaves were placed in a fridge upon either purchase or harvest (4 ◦C) until further analysis was performed.

#### *2.2. Chemicals*

For headspace solid-phase-microextraction (HS-SPME), calcium chloride and the alkane standards C6-C25 (100 μg/mL) in diethyl ether were obtained from Sigma-Aldrich (now Merck; Poole, UK). For ultra-high performance liquid chromatography mass spectrometry (UPLC-MS), authentic compounds of glucoiberin (GIB; 99.61%, HPLC), progoitrin (PRO; 99.07%, HPLC), sinigrin (SIN; 99%, HPLC), glucoraphanin (GRA; 99.86%, HPLC), glucoalyssin (GAL, 98.8%, HPLC), gluconapin (GNP, 98.66%, HPLC), 4-hydroxyglucobrassicin (4HGB; 96.19%, HPLC), glucobrassicanapin (GBN; 99.22%, HPLC), glucotropaeolin (GTP; 99.61%, HPLC), glucoerucin (GER; 99.68%, HPLC), glucobrassicin (GBR; 99.38%, HPLC), and gluconasturtiin (GNT; 98.38%, HPLC) were purchased from PhytoPlan (Heidelberg, Germany). Methanol (HPLC grade), formic acid (LC-MS grade), and acetonitrile (LC-MS grade) were purchased from VWR (Leicestershire, UK).

#### *2.3. Volatile Compounds*

#### 2.3.1. Headspace Solid Phase Microextraction (SPME)

Samples of respective leaf and root tissues were homogenised by means of a commercial blender for 30 s, and 2 g of each was weighed into a SPME vial of 15 mL fitted with a screw cap. Samples were left aside for 10 min for the enzymatic hydrolysis of GSLs to take place. After exactly 10 min, 2 mL of saturated CaCl2 was added in order to cease the enzymatic reactions. After equilibration at 40 ◦C for 10 min, a 50/30 μm DVB/CAR/PDMS fibre was exposed to the headspace above the sample for 20 min. Three biological replicates were prepared for GC-MS analysis, and two replicates for each of the three assessors were prepared for the GC-O analysis.

#### 2.3.2. GC-MS Analysis of SPME Extracts

After extraction, the SPME device was inserted into the injection port of an Agilent 7890A gas chromatography system coupled to an Agilent 5975C detection system

equipped with an automated injection system (CTC-CombiPAL). A capillary column HP-5MS (30 m × 0.25 mm × 0.25 μm film thickness) (Agilent, Santa Clara, CA, USA) coated with (5% Phenyl Methyl Silox) was used for the chromatographic separation of volatile compounds. The oven temperature program used was 2 min at 40 ◦C isothermal and an increase of 4 ◦C/min to 250 ◦C. Helium was used at 3 mL/min as carrier gas. The sample injection mode was splitless. Mass spectra were measured in electron ionisation mode with an ionisation energy of 70 eV, the scan range from 20 to 280 *m/z* and the scan rate of 5.3 scans/s. The data were controlled and stored by the HP G1034 Chemstation system. Identities were confirmed by running the samples on a Stabilwax-DA (30 m × 0.25 mm × 0.25 μm film thickness) polar column from Restek (Bellefonte, PA, USA). Volatile compounds were identified or tentatively identified by comparison of each mass spectrum with spectra from authentic compounds analysed in our laboratory, or from the NIST mass spectral database (NIST/EPA/NIH Mass Spectral database, 2014), or spectra published elsewhere (see Supplementary Data S1 for GC-MS chromatograms and compound fragmentation spectra). A spectral quality value >80 was used alongside linear retention index to support the identification of compounds where no authentic standards were available. LRI was calculated for each volatile compound using the retention times of a homologous series of C6-C25 *n*-alkanes and by comparing the LRI with those of authentic compounds analysed under similar conditions. The compound peak areas were normalised and converted to the relevant abundance of each component as a percentage of the total peak area.

#### 2.3.3. GC-O Analysis of SPME Extracts

After extraction, the SPME device was inserted into the injection port of an Agilent 7890B Series ODO 2 (SGE) GC-O system equipped with a non-polar HP-5MS column (30 m × 0.25 mm × 0.25 μm film thickness). The outlet was split between a flame ionisation detector and a sniffing port. The contents of the SPME fibre were desorbed for 3 min in a split/splitless injection port, in splitless mode, onto five small loops of the column in a coil, which were cooled in solid carbon dioxide and contained within a 250 mL beaker. The injector and detector temperatures were maintained at 280 ◦C and 250 ◦C, respectively. During desorption, the oven was held at 40 ◦C. After desorption, the solid carbon dioxide was removed from the oven. The oven was maintained at 40 ◦C for a further 2 min and then, the temperature was raised at 4 ◦C/min to 200 ◦C and at 8 ◦C/min to 300 ◦C. Helium was the carrier gas, and the flow rate was 2.0 mL/min. Three assessors were used for the detection and verbal description of the odour active components of the SPME extracts. Each assessor participated in three training sessions for each sample species prior to scoring sessions. Each assessor evaluated by sniffing each sample in duplicate and documented the odour description, retention time, and odour intensity (OI) on a seven-point scale (2–8), where <3 = weak, 5 = medium, and 7 = strong. Only those odours that were detected by all three assessors were recorded in the results. *n*-Alkanes C6-C25 were analysed under the same conditions to obtain LRI values for comparison with the GC-MS data.

#### *2.4. Non-Volatile Compounds*

#### 2.4.1. Glucosinolate (GSL) Extraction

GSL extraction was performed as per the protocol presented by [29] with modifications. Briefly, 40 mg of dried leaf powder was placed into Eppendorf tubes and put into a heat block (80 ◦C for ten minutes). Afterwards, 1 mL of preheated methanol water (70% *v/v*) was added to dried powder, vortexed vigorously, and placed in a water bath (75 ◦C) for 20 min. Samples were cooled and centrifuged at full speed for five minutes at room temperature (≈22 ◦C); the supernatant was collected and filtered (0.22 μm PVDF Acrodisc syringe filters; VWR, Lutterworth, UK). Crude extracts were dried using a centrifugal evaporator and resuspended in 1 mL of LC-MS-grade H2O and stored at −80 ◦C until analysis. Immediately before analysis by UPLC-MS, samples were diluted five-fold with LC-MS-grade H2O.

#### 2.4.2. UPLC-MS Analysis of GSL Extracts

UPLC-MS was performed on a Shimadzu Nexera X2 series UHPLC, coupled with an 8050 triple quadrupole mass spectrometer system (Shimadzu UK Ltd., Milton Keynes, UK). Separation of standards and samples was achieved using a Waters BEH C18 Acquity column (100 × 2.1 mm, 1.7 μm; Waters Corp., Wilmslow, UK) with an Acquity in-line filter. Mobile phases consisted of 0.1% formic acid in LC-MS grade H2O (A), and 0.1% formic acid in LC-MS grade acetonitrile (B) and GSLs were separated during a five minute run with the following gradient timetable: (i) 0–50 s (A-B, 98:2, *v/v*), (ii) 50 s–3 min (A-B, 70:30, *v/v*), (iii) 3–3 min 10 s (A-B, 5:95, *v/v*), (iv) 3 min 10 s–4 min (A-B, 5:95, *v/v*), (v) 4–4 min 10 s (A-B, 98:2, *v/v*), (vi) 4 min 10 s–5 min (A-B, 98:2, *v/v*). The flow rate was 0.4 mL per min and the column oven temperature was 35 ◦C.

Two MS methods were used for the identification and quantification of GSLs. First, a Product Ion Scan (PIS) method was established to identify GSLs based on known primary ion masses ([M-H]-) characteristic fragment ions (357, 258, and 97 *m/z*; Table 1). Then, MS/MS spectra were compared to authentic standards and available literature sources [30–38]. Pentyl GSL (PEN), isobutyl GSL (ISO), glucoputranjivin (GPJ), and butyl GSL (BUT) were tentatively identified due to the possible presence of isomers [38], and/or no reliable reference MS spectra could be found in the literature. Total ion chromatograms of glucosinolates identified were included in the Supplementary Data (S2).

MS/MS settings for the PIS method were as follows: samples were analysed in the negative ion mode with a scan range of 70–820 *m/z*. A collision energy of 25 eV and a scan speed of 30,000 u per s−1. For the quantification of GSLs, a Multiple Reaction Monitoring (MRM) method was established. Based on the fragmentation observed in the PIS method, confirmation and quantification transitions were established (Table 1). Dwell times for each precursor and product ion were set to 5 s.

Authentic GSL compounds were run as external standards. Limits of detection (LOD) and limits of quantification (LOQ) were established for each and are presented in Table 1. As standard compounds are not available for all GSLs, SIN was used to semi-quantify glucorucolamine (GRM), glucoputranjivin (GPJ), diglucothiobeinin (DGTB), glucoberteroin (GBT), glucocochlearin (GCL), glucosativin (GSV), dimeric 4-mercaptobutyl GSL (DMB), glucobarbarin (GBB), and tentatively identified GSL compounds. Similarly, GBR was used to semi-quantify the indolic GSLs 4-methoxyglucobrassicin (4MGB) and neoglucobrassicin (NGB).

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

#### *3.1. Volatile Compounds*

The volatile compounds identified in the headspace of the four Brassicales species are listed in Table 2, detailing their PubChem compound identification (PubChem CID) as well as their linear retention indices (LRI) in a polar and non-polar column. Semiquantitative characterisation results are also shown in Table 2 as relative area. ITCs and alcohols were the chemical classes of compounds dominating the volatile profile of the samples with other compounds such as aldehydes, esters, and terpenes also present.

GC-olfactometry analysis of the samples yielded a total of 107 odorants across the four species, which are presented in Table 3. Qualitative differences were observed between the samples with horseradish and watercress yielding a total of 52 and 51 odorants, respectively. Green/grassy, radish, sulphury, and horseradish were some of the terms that were mostly used by the assessors to describe the odours. Additionally, a total of 46 odorants of unknown identity were detected within the headspace of the four Brassicales analysed that may contribute to the odour profiles of these crops. These compounds matched no corresponding peaks and LRI values within the GC-MS data. This suggests that the compounds responsible for generating the perceived aromas were present at levels below the detection threshold of the instrumentation used. The number and diversity of unidentified compounds and aromas is indicative of the fact that characterisation of the species' volatile profiles is far from complete, and it is likely that many more will be discovered in future studies.


**Table 1.** Glucosinolate compounds identified in salad rocket, wasabi, and horseradish by UPLC-MS/MS.


**Table 1.** *Cont.*

80

glucobrassicin;

 tentative identification.


*Foods* **2021**, *10*, 1055


#### *Foods* **2021** , *10*, 1055

**Table 2.** *Cont.*



#### *Foods* **2021** , *10*, 1055



present; nd, not detected; \* newly reported for species.









#### 3.1.1. 'Salad' Rocket

A total of 57 volatile compounds were identified or tentatively identified in the headspace of *E. sativa* leaf samples. Nineteen compounds are newly reported for this species, some of which make up relatively large portions of the total volatile compounds' bouquet (Table 2). Compounds with the greatest relative abundances were (*Z*)-3-hexen-1-ol (**31**, 40.9%), (*E*)-2-hexenal (**44**, 14.3%), hexanal (**43**, 8.6%), erucin (**23**, 7.2%), and 1 isothiocyanato-4-methylpentane (**16**, 5%; Figure 2). These observations are broadly in agreement with previous studies of 'salad' rocket [3,12].

Despite its high relative abundance, (*Z*)-3-hexen-1-ol produced only a weak, green, radishy aroma (Table 3) in rocket leaves. (*E*)-2-Hexenal by comparison was 2.9-fold less abundant in relative terms but produced a slightly stronger aroma, described by assessors as green, and apple-like. Hexanal (**43**) produced a pungent, green, grassy aroma of relatively high intensity, which has not been previously described in rocket to our knowledge.

2-Isopropyl-3-methoxypyrazine (**102**) by comparison was of low relative abundance in the rocket headspace (0.3%; Table 2) but was found to have one of the strongest aromas in rocket (rotten, potato-like, vegetative; Table 3). 1-Isothiocyanato-4-methylpentane had a weak aroma and was given a tentative new description of 'musty', as no previous studies have reported an odour for this compound.

The ITC erucin (**23**) is known for its anticarcinogenic properties, but its aroma was only recently described [12]. In agreement with a previous report, this compound produced a radishy aroma of weak intensity. Interestingly, the compound previously associated with characteristic "rocket-like" aroma (sativin, **20**) [12] was relatively weak-smelling. In this study, the compound was found to have a burnt, rubbery, and soil-like aroma. This suggests that sativin is not the main driver of pungency or aroma in 'salad' rocket. Other compounds such as ethyl vinyl ketone (**65**; [41]), hexanal, 3-butenyl ITC (**11**) [48,54], and several unknown compounds (see final paragraph in this section) all had descriptions of pungency at higher intensities than sativin (Table 3). It may also be likely that no single compound is responsible for this attribute of rocket aroma but rather several.

Pentyl ITC (**15**, 0.6%) not previously identified in 'salad' rocket produced a strong odour, which was characterised as cabbage-like, green, and rotten (Table 3, Figure 2). As will be discussed in Section 3.2, we have tentatively identified pentyl GSL (Table 1) as a significant and previously unreported component of the GSL profile of 'salad' rocket, which gives rise to this ITC compound.

Other compounds not previously identified in 'salad' rocket included 2-phenylethanol, 4-heptenal and 2-*sec*-butyl-3-methoxypyrazine. 2-Phenylethanol (**37**, 0.2%) was noted to impart a floral aroma at a medium-weak intensity (Table 3). This compound is derived from phenylalanine and has been found to contribute to aroma and flavour in many foods, such as tomatoes [66]. 4-Heptenal (**45**) occurred in rocket leaves with a relative abundance of 0.1% (Table 2). Despite this low amount in terms of the overall volatile profile, the compound was perceived at a medium intensity by the assessors (Table 3) and described as grassy and green. This compound has been variously described as mushroom-like [50], fatty and fishy [67], and potatoey [68]. The variation in these descriptions may be associated with the isomerisation of the compound, which could not be resolved in this study. 2-*sec*-Butyl-3-methoxypyrazine (**106**, 0.2%) has been reported in several Brassicales species, such as white mustard, rapeseed [60], and horseradish [25], but not in *E. sativa* (Figure 2). It has been variously described as having a pea-like, musty, green, and bell pepper-like aroma. Assessors described the compound as earthy, similar to rotten potatoes, and vegetablelike. Despite its relatively low abundance, it was perceived as a medium-weak smelling compound in the sample headspace.

A total of 13 unidentified odorants were also detected in the headspace of rocket by GC-O (Table 3). These varied in intensity but all were distinguished and reported by assessors. Odour descriptions for these compounds were buttery (**116**), sulphury (**117**, **122**, **124**, **141**, **145**), horseradish-like, rancid (**117**), cabbage-like (**122**), rotten onion (**124**), mustard, pungent (**128**), oniony (**128**, **138**), green, sour apples (**131**), soily, earthy (**139**), gas (**141**, **145**), burnt (**141**), roasted (**141**, **142**), smoky (**142**), spicy, chemical (**144**), fresh cucumber, rotten, and vegetable-like (**150**).

#### 3.1.2. Wasabi

A total of 43 compounds were identified or tentatively identified in the headspace of wasabi roots (Table 2) with 30 compounds newly described for the species, making this a significant step forward in the understanding of aroma composition in wasabi roots. Two peaks of near-identical spectra were observed and identified as allyl ITC (**5/6**, 8.6%/52.1%) and were the most abundant compounds, which agrees with previous observations [7]. It is unknown why two distinct peaks were formed in this manner, and further investigation may be required to determine the isomeric differences responsible for the separation. 4-Pentenyl ITC (**14**, 12.4%) was also found to be high in terms of overall relative abundance.

Despite having near identical spectra, compounds **5** and **6** presented distinct differences in aroma and intensity. Peak 6 was characterised as being very pungent (Table 3) and having garlic, mustard, and horseradish-like qualities. Allyl ITC is one of the most well characterised ITCs and is well known for these properties [1,48,63] (Figure 2). Peak **5** by contrast had no discernible aroma in wasabi but was apparent in horseradish (see Section 3.1.3). 4-Pentenyl ITC likewise exhibited a pungent aroma and strong odour intensity but also had peppery, sulphurous, and musty notes. This compound is commonly reported in Brassicaceae crops [69]. 3-Butenyl ITC (**11**, 4.2%) was scored as a high odour intensity compound, despite its much lower relative abundance and was described as having a pungent, green, and aromatic odour.

Several other GHP odorants are newly reported in wasabi including cyclopropane ITC, isoamyl ITC, pentyl ITC, 1-isothiocyanato-4-methylpentane, and benzyl ITC. Cyclopropane ITC (**7**, 0.1%) is likely to be a cyclic reaction product of allyl ITC and has been previously reported in brown mustard [49] (Figure 2). To our knowledge, no odour description of this compound has been previously made, but assessors described it as sulphurous, horseradish, garlic, and onion-like (Table 3). The high odour intensity score indicates that it is a significant component of wasabi aroma. Isoamyl ITC (**13**, 0.4%) was described as having a pungent grassy aroma and being of high odour intensity. This compound is not commonly reported in Brassicales species, but it is used as a food additive [70]. Most ITC compounds are noted for their sulphurous and mustard-like potency in Brassicales; however, the contribution of grassy aroma ITCs to volatile compound bouquets has not been previously appreciated or fully understood. Pentyl ITC (**15**, 0.1%) and 1-isothiocyanato-4 methylpentane (**16**, <0.1%) were observed and shared the same odour characteristics as in 'salad' rocket (see Section 3.1.1.). Benzyl ITC (**22**, <0.1%) was reported to have a rotten grass and cooked aroma of medium-weak intensity. Similar to isoamyl ITC, this compound is not regularly reported as a constituent of Brassicales headspace, but these data indicate that even in very low relative abundance, it is odour active.

Another interesting compound was also found in wasabi headspace: cyclohexyl isocyanate (**101**, <0.1%; Figure 2). This has been previously reported in black mustard [53], though it is unclear if it is related to or derived from GHPs. Assessors perceived this odour having a medium-strong intensity and described it as peppery, cooked, and potato-like (Table 3). This is a tentative new odour description for this compound, and our data suggest it to be an important constituent of wasabi aroma.

Two additional compounds not previously identified in wasabi were 1-octen-3-ol and eucalyptol. 1-Octen-3-ol (**34**, <0.1%), despite its very low relative intensity in root tissue headspace (Table 2, Figure 2), exhibited a high odour intensity imparting a mushroom-like odour in agreement with previous descriptions [50]. Eucalyptol (**97**, <0.1%), previously observed in Brassicales crops [18] but not in wasabi (Figure 2), was found to have a mediumstrong, characteristic eucalyptus, and mint aroma, and it is likely an important component of the overall volatile bouquet.

Similar to rocket, 17 unidentified odorants were detected in wasabi root samples (Table 3). Reported aromas were sulphury (**114**, **117**, **123**, **146**), buttery (**116**), horseradishlike (**117**, **119**), rancid (**117**), mustard (**119**, **128**), onion (**121**, **128**), garlic (**123**, **146**), pungent (**128**), cooked, roasted chicken, chicken soup (**129**), potato (**133**), spicy, cinnamon-like, nutty (**149**), fresh cucumber, rotten, vegetable-like (**150**), radish (**79**), green (**79**, **153**), sweet (**152**), floral (**152**, **158**), violets, perfume (**152**), peppery, earthy (**153**), liquorice, medicinal (**154**), soapy, and grassy (**158**), confirming our statement that wasabi's volatile profile is far from complete.

#### 3.1.3. Horseradish

A total of 75 compounds were identified or tentatively identified in the headspace of horseradish roots, 38 of which are newly reported (Table 2). As with wasabi, the peaks with the highest relative abundances were dominated by GHPs. Compounds with the highest relative abundances were allyl ITC (**5/6**, 7.4%/39.3%), phenethyl ITC (**24**, 32.5%), thiiraneacetonitrile (**77**, 3.6%), 4-pentenyl ITC (**14**, 2.3%), (*Z*)-3-hexen-1-ol (**31**, 1.9%), and *sec*-butyl ITC (**9**, 1.8%; Figure 2).

As stated in Section 3.1.2, allyl ITC was identified as two distinct peaks (**5** and **6**). As in wasabi, **6** was of the greatest abundance and odour intensity, producing a very strong, pungent, garlic, mustard, and horseradish-like aroma (Table 3), whereas **5** produced a medium intensity, pungent horseradish smell. Thus far, the presence of two peaks has not been addressed or explained satisfactorily within the literature, with only one previous paper reporting the same phenomenon of separate and distinct allyl ITC peaks [71]. *Sec*butyl ITC (**9**) produced a medium-weak intensity aroma that was vegetative and radish-like (Table 3). The compound was also present in wasabi at a medium intensity. The compound has been previously reported in horseradish as having green, chemical, and mustard like aromas [21], and it is known to activate the human Transient Receptor Potential Ankyrin 1 (TRPA1). This receptor is known to act in response to environmental irritants, and several ITCs identified in this study are known to activate it to varying degrees (isopropyl ITC, **3**; isobutyl ITC, **10**; allyl ITC, **5/6**; 3-butenyl ITC, **11**; 4-pentenyl ITC, **14**; benzyl ITC, **22**; phenylethyl ITC, **24**; [72]). Phenylethyl ITC (**24**) is known to be a key constituent of horseradish aroma, and our data are in agreement with previous reports [73]. Assessors described the compound as radish and gooseberry-like, with a sweet note. It had a high odour intensity and contributed significantly to the odour profile of roots. Likewise, 4 pentenyl ITC (**14**) was observed to have the same odour attributes as previous reports [1] and those found for wasabi in this study, but at a lower intensity. By contrast, pentyl ITC (**15**) was present at much lower relative intensities to other GHPs (0.2%, Table 2) but produced a strong, green, rotten, and cabbage-like aroma.

Thiiraneacetonitrile (**77**) has been previously reported in horseradish [22] and is an epithionitrile hydrolysis product of sinigrin. To our knowledge, no previous studies have described the odour of this compound. We found it to have a sweaty, gas-like aroma of medium intensity (Table 3).

2-Isopropyl-3-methoxypyrazine (**102**, 0.1%) and 2-*sec*-butyl-3-methoxypyrazine (**106**, 0.4%) have been previously described and characterised in horseradish roots [25] as having green and pepper-like aromas. Our data agree with previous reports but found the compounds to be of very high aroma intensity, despite relatively low abundances within the headspace (Table 3). Assessors described the compounds as rotten, earthy, potato-like, and vegetative.

We report several compounds previously unidentified in horseradish including GHPs, isocyanates, alcohols, aldehydes, and a ketone and ester. As in wasabi root, cyclopropane ITC (**7**, 0.1%) produced an intense aroma containing horseradish, garlic, onion, and sulphur notes (Table 3). Therefore, it is likely to be a significant contributor to root odour and the volatile profile, despite its very low abundance, which may be a reason why it has not been previously detected and/or reported.

As discussed in Section 3.1.2, it is unknown if the presence of isocyanates is linked with GSLs and their hydrolysis products. Allyl isocyanate (**99**, <0.1%) and phenethyl isocyanate (**107**, 0.4%) were both observed for the first time in horseradish roots. Given that high abundances of allyl ITCs (**5/6**) and phenethyl ITC (**24**) were observed, it seems likely that isocyanates may be derived from them and/or directly from parent GSLs. Isocyanates are not commonly reported in the literature, and their formation may be because of asyet-unstudied enzymatic or post-hydrolysis modification processes. Allyl isocyanate was described as having a weak musty and burnt plastic aroma; and phenethyl isocyanate was described as being pungent, with ground pepper and horseradish-like quality at a medium intensity. We are not aware of any previous odour descriptions for these compounds, so these are tentative new characterisations.

1-Octen-3-ol (**34**, <0.1%) was identified, and as in wasabi root, it produced a mushroomlike aroma of medium-weak intensity (Table 3). Benzyl alcohol (**36**, 0.1%) has been previously reported in *Brassica oleracea* [74] and rapeseed [58], but not in horseradish. It was characterised as having a medium intensity aroma, described as fruity, medicinal, and wine-like. Its low abundance but relatively high odour intensity may make it a subtle but key constituent of the root aroma profile. Two aldehyde compounds were also found to contribute to odour within the headspace. 2-Pentenal (**41**, <0.1%) produced a green, apple-like aroma [43], and 2,4-hexadienal (**47**, <0.1%) produced a green, rotten smell, both of medium-weak intensity (Table 3; [51]). A ketone, 3,5-octadien-2-one (**71**, 0.1%) was also tentatively identified (Table 2) and described as having a pungent green aroma of medium intensity (Table 3).

In the esters group, methyl salicylate (**60**, <0.1%), a common compound throughout the plant kingdom, has previously been identified in Brassicales as part of systemic acquired resistance response to herbivory [75], and it was identified for the first time as a volatile constituent of horseradish headspace (Table 2). Its aroma is characteristic of, and present in, plants such as wintergreen. Assessors described its odour as medicinal and camphorous at a medium intensity (Table 3). Ethyl decanoate (**61**, <0.1%), known to be present in *B. oleracea* [59], was described by the assessors as green and waxy (Table 3; [65]). Benzyl tiglate (**113**, <0.1%) was reported at low relative abundance, but a perceptible musty aroma was apparent for this compound.

The presence of *D*-limonene (**96**, 0.1%) is reported for the first time in horseradish root. It is known to be a constituent of *B*. *oleracea* headspace [76], but its sensory contribution to Brassicales is not well defined. Assessors found this compound to have a medium-weak intensity aroma of lemon and being vegetable-like (Table 3 [57]). This agrees with previous descriptions of the odour properties of the compound.

Finally, 6-methylquinoline (**111**, <0.1%), an aromatic compound, produced an unpleasant hydrogen sulphide and egg-like aroma of medium-weak intensity, and it has not been previously reported.

Similar to the other species tested, 15 unidentified odorants were detected in horseradish root samples (Table 3) and were of weak to medium intensity. Reported aromas were sulphury (**114**, **117**, **123**, **141**, **145**, **146**, **155**), buttery (**116**), horseradish-like (**117**, **118**), rancid (**117**), rotten (**120**), cabbage-like (**120**, **155**), garlic, (**123**, **146**), mustard, pungent, oniony (**128**), cooked (**129**, **155**), roasted chicken, chicken soup (**129**), nutty, spicy (**130**), gas (**141**, **145**), burnt, roasted (**141**), sweet, floral, violets, perfume (**152**), minty, cooling, and fresh (**160**).

#### 3.1.4. Watercress

A total of 42 compounds were identified or tentatively identified in the headspace of watercress, 26 of which are newly reported (Table 2). The headspace profile was dominated by alcohol and ITC compounds: (*Z*)-3-hexen-1-ol (**31**, 36%), phenethyl ITC (**24**, 30.7%), 1-penten-3-ol (**25**, 8.1%), and (*Z*)-2-penten-1-ol (**28**, 4.8%; Figure 2).

1-Penten-3-ol is a compound present widely in Brassicales species [77]. It exhibited a high intensity in watercress leaves, producing a sulphurous and cabbage-like aroma. These attributes are often attributed to ITCs and other sulphur-containing compounds; however, our data suggest that some of these characteristics in watercress could be attributed to this alcohol. (*Z*)-3-Hexen-1-ol by comparison was much higher in relative abundance but produced a medium intensity aroma that was green and radishy (Table 3).

Phenethyl ITC (aroma attributes described in Section 3.1.2.) had one of the highest intensity aromas in watercress, along with phenylacetaldehyde (**52**, 0.2%). The latter, similar to phenethyl ITC, is derived from phenylalanine, but occurs in much lower abundance (Table 2, Figure 2). Its aroma was described as honey-sweet (Table 3) and is likely a significant contributor to watercress odour that has previously gone unrecognised.

Other compounds contributing high odour intensities despite low relative abundances were 3-pentanone (**66**, 1.9%) and β-ionone (**98**, 0.4%; Figure 2). 3-Pentanone is regularly reported in Brassicales [3] and was described as high intensity, green, grassy, and floral smelling (Table 3, Figure 2). β-Ionone is common to many plant species as a degradation product of carotenoids [78], and it was described as soapy and fusty by assessors, with a high intensity.

Several sulphur-containing compounds, aldehydes, alcohols, ketones, not previously identified in watercress were identified. Methyl thiocyanate (**2**, 0.5%) imparted a sulphury and oniony note, with a medium intensity. This compound is known to be a GSL hydrolysis product of GCP (methyl GSL), but as will be discussed in Section 3.2, this compound was not detected in the UPLC-MS/MS analysis. Therefore, we suggest that it is not directly derived from this GSL and may be a degradation product of other GSL hydrolysis products within the tissues and headspace of the tested Brassicaceae.

Cyclopentyl-1-thiaethane (**8**, 0.5%) was detected, and uniquely present in watercress compared with the other three species tested (Table 2). Little is known about this compound in a biological context. It produced a sweaty, sulphury, medium-weak intensity aroma that is tentatively described in this species for the first time (Table 3).

Five aldehydes, two alcohol and two ketones, are newly reported for watercress, which produced perceptible odours within the headspace bouquet: 2-pentenal (**41**, 0.5%), hexanal (**43**, 1.2%), 4-heptenal (**45**, 0.5%), heptenal (**46**, 0.1%), and nonanal (**54**, 0.3%). Both heptenal and nonanal exhibited fatty, green aromas of medium intensity (Table 3) and are common to other Brassicaceae species [77]. 1-Octen-3-ol (**34**, 0.1%) produced a medium strength mushroom-like aroma (as described in Section 3.1.3.), and 2-phenylethanol (**37**, 2.6%) produced a floral scent. 6-Methyl-5-hepten-2-one (**68**, 0.2%) has been previously observed in 'wild' rocket and described as having a citrus aroma [55,79]. In this study, it was also identified with this characteristic, but also as floral and perfume-like, in both watercress and 'salad' rocket. It produced a medium-strong aroma in watercress. 2,2,6- Trimethylcyclohexanone (**69**, 0.1%) has been variously described as thujonic, menthol-like, and camphorous [80]. Here, it was described by assessors as imparting floral and green odours, with citrus notes.

One ester, (*Z*)-pent-2-en-1-yl acetate (**57**, 0.1%), was only detected in watercress, and it produced a medium-weak aroma. It was described as sulphury and rotten, and we are not aware of any previous odour attributes associated with this compound. As such, this is a tentative first description.

Octyl ITC (**21**, 0.2%) has been previously identified in horseradish [64] but not watercress to our knowledge. Its exact derivation and parent GSL are unclear in the literature, though watercress has been reported to contain glucohirsutin (GHS, (*RS*)- 8methylsulfinyl)octyl GSL; [38]). This will be discussed further in Sections 3.2 and 3.3. Assessors found the compound to be of medium aroma strength having a green and vegetative character. Again, we are unaware of previous odour descriptions for this compound.

Finally, there were 29 unidentified odorants detected by GC-O, making this the highest number of the four species analysed (Table 3). Aromas described by assessors were sulphury (**114**, **124**, **125**, **126**, **141**), cooked onions (**115**, **125**), buttery (**116**), rotten, cabbagelike (**120**), rotten onion (**124**), pungent (**125**, **128**, **157**), oniony (**126**, **128**), green (**127**, **131**, **137**, **153**, **156**), parsley (**127**), mustard (**128**), cooked, roasted chicken, chicken soup (**129**), sour apples (**131**), apples (**132**), grass (**132**, **134**, **151**, **158**), potato (**133**, **156**), earthy (**135**, **153**), musty (**135**), petrol, aromatic (**135**, **147**), bread-like (**136**), medicinal (**140**, **159**), floral (**140**, **143**, **158**), gas, burnt, roasted (**141**, **142**), smoky (**142**), cucumber (**143**, **148**), flowers (**143**), fruity, chemical, dried fruit (**151**), peppery (**153**), radish (**156**), and soapy (**157**, **158**,

**159**). This indicates that the volatile profile of watercress is far from complete, and further research is required to elucidate these compounds.

#### *3.2. Non-Volatile Compounds (Glucosinolates)*

GSL composition and concentrations for 'salad' rocket, wasabi, and watercress are presented in Table 4. Due to an unforeseen termination of supply, it was not possible to include horseradish roots in this analysis.

**Table 4.** Glucosinolate concentrations of 'salad' rocket, wasabi, and watercress determined by UPLC-MS/MS.


<sup>a</sup> GSL: glucosinolate; \$ = tentative identification. <sup>b</sup> Concentration in μmol g−<sup>1</sup> dry weight; means are from six replicates for salad rocket, five replicates for wasabi and eight replicates for watercress; nd, not detected.

#### 3.2.1. 'Salad' Rocket

'Salad' rocket contained the highest dry weight concentrations of GSLs (111.1 ± 14.6 <sup>μ</sup>mol g−<sup>1</sup> dw). This was predominantly due to high amounts of DMB (78.9 ± 8.4 <sup>μ</sup>mol <sup>g</sup>−<sup>1</sup> dw). Other GSLs of note included GSV (6.6 ± 2.4 <sup>μ</sup>mol g−<sup>1</sup> dw), GRM (7.6 ± 1.2 <sup>μ</sup>mol <sup>g</sup>−<sup>1</sup> dw), and DGTB (4.6 ± 1.1 <sup>μ</sup>mol g−<sup>1</sup> dw), which are unique to the genera *Eruca* and *Diplotaxis*. Other routinely reported GSLs for this species were GRA (1.8 ± 0.5 <sup>μ</sup>mol g−<sup>1</sup> dw) and NGB (2.7 ± 0.4 <sup>μ</sup>mol g−<sup>1</sup> dw).

Interestingly, several other GSLs that have not been, or are rarely reported for the species, were also detected; some in relatively high concentrations: GIB (<0.1 ± <0.1 μmol

<sup>g</sup>−<sup>1</sup> dw), PEN (4.9 ± 1.6 <sup>μ</sup>mol g−<sup>1</sup> dw), GPJ (<0.1 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw), GBT (0.4 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw), GTP (<0.1 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw), 4MP (1.3 ± 0.8 <sup>μ</sup>mol g−<sup>1</sup> dw), HEX (0.2 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw), and BUT (0.3 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw).

Of note is the high abundance of PEN (*m/z* 388). It seems unlikely that a GSL of such relatively high concentration has gone undetected in previous analyses. Therefore, we postulate that previous studies may have attributed the negative ion mass incorrectly to that of PRO, which is also *m/z* 388 (Table 1). The authentic standard of PRO did not match the retention time or MS/MS spectra of PEN, and it was found in only very low concentrations by comparison. While PEN is only a tentative identification (due to the possibility of other isomeric GSLs such as glucojiaputin and 3-methylbutyl GSL), the presence of pentyl ITC (**17**) within the headspace of rocket makes this the most likely identification. See Section 3.3 for further discussion.

#### 3.2.2. Wasabi

Sixteen GSL compounds were identified in wasabi roots, totaling 27.5 ± 2.8 <sup>μ</sup>mol g−<sup>1</sup> dw (Table 4). The most abundant compound was SIN (11.1 ± 0.2 <sup>μ</sup>mol g−<sup>1</sup> dw), which agrees with previous studies [81]. Wasabi is known to have a diverse GSL profile, and we observed relatively high abundances for ISO (3.4 ± 0.8 <sup>μ</sup>mol g−<sup>1</sup> dw), GAL (2.1 ± 0.5 <sup>μ</sup>mol g−<sup>1</sup> dw), GPJ (2.3 ± 0.5 <sup>μ</sup>mol g−<sup>1</sup> dw), GCL (2.2 ± 0.2 <sup>μ</sup>mol g−<sup>1</sup> dw), 7MSH (4.6 ± 0.4 <sup>μ</sup>mol g−<sup>1</sup> dw), and NGB (1.5 ± 0.2 <sup>μ</sup>mol g−<sup>1</sup> dw). Other compounds occurring in low abundance that are not frequently reported were GIB and GRA.

#### 3.2.3. Watercress

Fourteen GSLs were found in watercress leaves, amounting to 31.6 ± 1.4 <sup>μ</sup>mol g−<sup>1</sup> dw (Table 4). In most previous studies of this species, GNT (1.5 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw) has been found to have the greatest abundance [4]; however, our analysis revealed that 7MSH had the highest total concentration (21.5 ± 1.2 <sup>μ</sup>mol g−<sup>1</sup> dw), dominating the overall profile in these samples. There were also relatively high concentrations of 7MTH (2.7 ± 0.1 μmol <sup>g</sup>−<sup>1</sup> dw), and the indolic GSLs 4MGB (2 ± 0.1 <sup>μ</sup>mol g−<sup>1</sup> dw), and NGB (1.6 ± <0.1 <sup>μ</sup>mol g−<sup>1</sup> dw). Minor amounts of SIN, GRA, and GTP were also observed, and they are not frequently reported in this species.

#### *3.3. Discrepancies between Identified Glucosinolate Hydrolysis Products and Glucosinolate Profile Precursors*

There is often an 'elephant in the room' regarding volatile GSL hydrolysis products and reported GSL profiles in Brassicales crops: there are often GSLs found with no corresponding hydrolysis products, or more troublingly, hydrolysis products observed but no GSL precursor. Table 5 presents a list of the GSL-derived compounds identified within the headspace of 'salad' rocket, wasabi, and watercress, alongside their expected GSL precursors. It is apparent from our data that the present study is no exception when it comes to discrepancies of this nature, and there is a need to find a robust solution to prevent the inaccurate reporting of both GSLs and their volatile hydrolysis products.

As discussed in previous sections, compounds such as methyl thiocyanate (**2**) may be produced from degradation of other hydrolysis products. Others such as the presence of *sec*-butyl ITC (**9**) in rocket, butyl ITC (**12**) in wasabi, and octyl ITC (**21**) in watercress cannot be so easily explained. There are several explanations with varying levels of likelihood: firstly, the most likely is that the ITCs and other GSL hydrolysis products have been identified incorrectly, and that they belong to other parent GSLs present within the analysed tissues. Despite researchers' best efforts to obtain authentic standards to test and match MS profiles, this is not always possible or affordable, and so there is a heavy reliance upon reference libraries. These are often incomplete and not always accurate. It is also possible that the high temperatures utilised in GC-MS cause thermolytic reactions to occur in GHPs, thus yielding compounds not 'naturally' produced by tissues. The next possibility is that the GSLs responsible for producing hydrolysis products are below the LOD of the MS/MS method. In this case, this is unlikely, as the LOD for each standard

GSL was low, and it is likely much more sensitive and accurate than volatile compound measurements by GC-MS. Thirdly and perhaps least likely is that there is some as-yetunknown mechanism(s) by which GSL hydrolysis products are modified post-hydrolysis. This is speculation, but there have been recent reports of previously unknown tautomeric rearrangements [82] and enzymatic actions [28] upon hydrolysis products that do not preclude this as an impossibility. Indeed, the dynamics of reactions occurring within the headspace of Brassicales is virtually unstudied, and so some may be produced through the degradation or rearrangement of structurally similar compounds. This is an area that requires much more detailed scrutiny.

**Table 5.** Identified volatile glucosinolate hydrolysis products in the headspace of 'salad' rocket, wasabi, and watercress, and the presence/absence of their respective glucosinolate precursor.


\* Hydrolysis product observed but not glucosinolate precursor; \$ tentatively identified.

#### **4. Conclusions**

This study has highlighted numerous newly identified and tentatively identified volatile compounds present in the headspace of 'salad' rocket, wasabi, horseradish, and watercress. Many of these appear to contribute strongly to the aroma profiles of each respective crop. We have also found 46 aroma traits present in the headspace of the samples that have no association with the identified compounds. This suggests that there are many as-yet undiscovered odour-active compounds present within Brassicales headspace.

Our data also highlight the need for more detailed studies on the volatilome of Brassicales species, and that the sole focus should not be upon GSL hydrolysis products. While accounting for a large proportion of the respective volatile profiles and odour active components of species, we have identified numerous instances where non-GSL derived compounds have odour intensities greater than those that are GSL-derived, such as 1-penten-3-ol (**25**), phenylacetaldehyde (**52**), 2-*sec*-butyl-3-methoxypyrazine (**106**), and

β-ionone (**98**). Several non-GSL derived compounds also share similar pungent odour characteristics with GHPs, indicating that the latter may not be the only source of these sensations in Brassicales crops. There is also a clear need for the improvement of mass spectral libraries and the availability of GSL and GHP standards in order to overcome discrepancies between GSL profiles and the reported volatiles derived therefrom.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/foods10051055/s1, Sup-plementary Data S1. Compound fragmentation spectra with corresponding library matches (where available) and GC-MS chromatograms.

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

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

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

**Acknowledgments:** The authors would like to thank Richard Tudor of Elsoms Seeds Ltd. (Spalding, UK) for providing *Eruca sativa* seeds.

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

#### **References**


#### *Article*

## **High Glucosinolate Content in Rocket Leaves (***Diplotaxis tenuifolia* **and** *Eruca sativa***) after Multiple Harvests Is Associated with Increased Bitterness, Pungency, and Reduced Consumer Liking**

#### **Luke Bell 1,\*, Stella Lignou <sup>2</sup> and Carol Wagsta**ff **<sup>2</sup>**


Received: 28 October 2020; Accepted: 1 December 2020; Published: 3 December 2020

**Abstract:** Rocket (*Diplotaxis tenuifolia* and *Eruca sativa*) leaves delivered to the UK market are variable in appearance, taste, and flavour over the growing season. This study presents sensory and consumer analyses of rocket produce delivered to the UK over the course of one year, and evaluated the contribution of environmental and cultivation factors upon quality traits and phytochemicals called glucosinolates (GSLs). GSL abundance was positively correlated with higher average growth temperatures during the crop cycle, and perceptions of pepperiness, bitterness, and hotness. This in turn was associated with reduced liking, and corresponded to low consumer acceptance. Conversely, leaves with greater sugar content were perceived as more sweet, and had a higher correlation with consumer acceptance of the test panel. First cut leaves of rocket were favoured more by consumers, with multiple leaf cuts associated with low acceptance and higher glucosinolate concentrations. Our data suggest that the practice of harvesting rocket crops multiple times reduces consumer acceptability due to increases in GSLs, and the associated bitter, hot, and peppery perceptions some of their hydrolysis products produce. This may have significant implications for cultivation practices during seasonal transitions, where leaves typically receive multiple harvests and longer growth cycles.

**Keywords:** glucosinolates; rucola; arugula; *Diplotaxis*; *Eruca*; bitter taste; flavour; postharvest

#### **1. Introduction**

Rocket (also known as arugula and rucola) salad species such as *Diplotaxis tenuifolia* and *Eruca sativa* are leafy vegetables of the order Brassicales, and are popular throughout the world [1]. They are commonly sold in bags of loose leaves, or as part of a leafy salad mixture with other crops, such as lettuce, spinach, and watercress [2]. Previous studies have evaluated sensory properties of rocket leaves [3–8] in conjunction with phytochemical compositions, and in one instance, consumer preferences according to human taste receptor genotype [9]. One factor not accounted for in any of these studies is the temporal variability of rocket produce over the course of a growing season, and the inherent environmental variability associated with this.

Rocket salad is in demand year-round in the UK; however, no British region is suitable for its continuous cultivation. As such, produce is typically sourced from several different countries throughout a year, according to the season [10]. In the UK, the vast majority of rocket is imported from Italy, with only seasonal summer rocket production possible in the south of England, as winters are too cold, wet, and humid for viable winter growth. In Italy, rocket is grown in the north and east during summer months; typically, in regions such as the Veneto, Lazio, and Emilia-Romagna.

In winter, there is a shift in cultivation towards the south where temperatures typically remain higher, and humidity lower; for example to Campania and Apulia. In times of high demand, rocket is sourced for the UK market from other European countries (e.g., Spain), Northern Africa (e.g., Morocco), or even as far as the United States and India.

It has been well documented in the literature that crop quality and respiration rates are influenced by seasonality [11]. This is intrinsically linked to growing temperature, as metabolic rates tend to be higher under warmer conditions. Growth temperature therefore plays a distinct role in determining shelf life longevity and visual acceptability of leaves [12]. There is also evidence suggesting that pungency is increased during spring and summer months [13], though this has not been shown quantitatively in rocket. Conversely, sugars have been shown to be reduced in some Brassicales species under "high" growth temperatures (>20 ◦C) [14], which has implications for sensory traits. Anecdotal evidence suggests that source and season significantly affects the quality and consistency of rocket leaves. This presents a problem for producers and supermarkets as leaves are typically marketed as generic products, but the quality of the produce is not consistent.

A previous study by Bell et al. [9] highlighted that pungency (or hotness) of rocket leaves is the main driver of rocket liking. Excessive pungency is typically rejected by consumers, and most people in that study preferred milder and sweeter leaves. There was a significant plant genotypic component determining the pungency of leaves, but it remains unstudied how pungency and consumer preference may be affected by changes in climate and seasonal growth. The pungency of rocket leaves is due to the presence of glucosinolates (GSLs) within tissues, which are hydrolysed by myrosinase enzymes to produce isothiocyanates (ITCs) and numerous other products. One compound present in both *E. sativa* and *D. tenuifolia* is glucosativin (4-mercaptobutyl GSL; GSV) and its dimer (dimeric 4-mercaptobutyl GSL; DMB). These produce the ITC, 4-mercaptobutyl ITC, which undergoes spontaneous tautomeric rearrangement post-hydrolysis to form 1,3-thiazepane-2-thione (sativin; SAT) [15]. It is unknown if it is the ITC or tautomer that is responsible for pungency and flavour, but olfactometry has shown a distinctive rocket-like aroma associated with the hydrolysis product of GSV [16]. The inherent variability of GSL biosynthesis in response to growth environment affects the presence and abundance of GSV and health related GSLs (such as glucoraphanin, GRA; and glucoerucin, GER) [17], and therefore may impact upon sensory properties of leaves and consumer acceptance.

An often-neglected component of taste and flavour perception in vegetables is sugar content and composition. Sugar content of Brassicales leaves is known to be variable according to growth conditions [18]. In combination with effects on GSL composition, sugars may therefore have a strong influence upon sensory traits and consumer preference throughout a growing season.

We present phytochemical, sensory analysis, and consumer preference and perception data from a year-long study of rocket produce sourced from commercial farms delivered to UK-based processors and supermarkets. We hypothesised that GSL and sugar content would significantly affect sensory properties of rocket leaves at different times of the year. This in turn would affect consumer perceptions of leaves and their preference of rocket would change seasonally. We also present the effects of climatic factors (such as temperature) and cultivation practices imposed on rocket crops (such as multiple harvests/cuts of the same crop).

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

#### *2.1. Plant Material*

Rocket leaf material was sourced and delivered to the University of Reading Sensory Science Centre monthly (with the exceptions of May and December) for one year (2014) by Bakkavor Ltd. (Spalding, UK). This material corresponded to leaf material delivered to a processing facility in England, and the material that would be used in products destined for UK supermarkets and consumers. Each sample batch was harvested, washed, and processed in accordance with industry practice. For reasons of commercial sensitivity, the exact locations of growers that supplied material for this study will not be named. Only the country of origin will be detailed, along with the growth environment (polytunnel, glasshouse, or open field) and cultivar used. It should be noted that as with many other crops, cultivars are selected by growers according to season and performance under specific conditions and climates. As such, the cultivars tested each month were not always the same. This was intentional to properly evaluate the level of consistency of produce as it exists within the UK supply chain, not to evaluate the changes of any single cultivar over the year or between different countries. Data were provided by growers relating to the length of each crop cycle (i.e., how long each of the samples had been grown for), the number of cuts each sample had undergone, and the percentage of dry matter of each batch at intake. See Table 1 for a summary of samples tested in each respective month.

#### *2.2. Temperature Data*

Temperature data for each of the rocket growing sites were supplied by Bakkavor Ltd. Data were logged throughout the cropping cycle of plants via on-farm weather stations. Four measurements were provided: the average daily temperature for the entire cropping cycle (referred to as "avg. temp"), the maximum temperature of the week preceding harvest (referred to as "max. temp. week"), the minimum temperature of the week preceding harvest (referred to as "min. temp. week"), and the average temperature in the week preceding harvest (referred to as "avg. temp. week").

#### *2.3. Sensory Analysis*

A panel of 12 previously trained and experienced assessors (ten female, two male) evaluated the samples at the Sensory Science Centre (University of Reading, Reading, UK). All training and monitoring of the sensory panel was in accordance with ISO 8586:2012 and ISO 11132:2012 standards. The panellists used the consensus vocabulary developed by Bell et al. [4] to describe the samples; the developed terms included appearance, odour, taste, flavour, mouthfeel, and after-effects attributes. Panellists were familiarized with the evaluated sensory attributes prior to each monthly scoring session, with reference standards when required.

Each month the panellists rated the samples individually in isolated well ventilated, temperature-controlled booths (22 ◦C), under artificial daylight. Attribute intensity was scored on 15 cm unstructured line scales (data scaled 0–100). Samples were presented monadically in a balanced presentation order, coded with three-digit random codes. Cold water and natural unflavoured yoghurt were provided for palate cleansing, and warm water for washing fingers between samples. Data were collected in duplicate using Compusense (version 5.5, Guelph, ON, Canada).

#### *2.4. Consumer Analysis*

Consumer recruitment and assessments were conducted as per the protocols of Bell et al. [9]. Evaluations were held on a bimonthly basis. Briefly, consumers were recruited from in and around the University of Reading and asked to attend as many of the evaluation sessions as possible over the course of the year-long study. A total of 55 consumers (out of 101) attended every session of the study.

Volunteers were presented with three leaves and asked to score their liking of leaf appearance, taste, and "overall" liking of each sample on a line scale (0–10), and based on their own individual experience of rocket sensory attributes. Volunteers were also asked to score their perceptions of bitterness, hotness (pungency), sweetness, and pepperiness. Scores were entered into a general labelled magnitude scale (gLMS) ranging from "not detectable", "weak", "moderate", "strong", "very strong", to "strongest imaginable". Data were subsequently converted to antilog values and normalized for statistical analysis [19]. Samples were presented monadically in a randomized, balanced presentation order, with three-digit random codes in duplicate. Data were collected using Compusense (version 5.5, Guelph, ON, Canada).




**Table 1.** *Cont.*

**34.2 7.4** = 1st cut; - = second cut; = 2nd+ cut; = <30 day crop cycle; = 31–60 day crop cycle; - = 61–90 day crop cycle; = >91 day crop cycle. Symbol codes denoting cut numbercrop cycle length are also utilised in subsequent text tables and figures. Numbers in bold are monthly averages. N.B. Hyphens (-) indicate data was not supplied from the grower.

 and

The demographics and characteristics of each respective panel are presented in Supplementary Table S1. The average number of recruits for bimonthly evaluations was 87, with an average age of 35 years old, ranging from 18 to 70. Volunteers were predominantly female (70.7%, on average), which is partly due to the gender balance present within the School of Chemistry, Food and Pharmacy at the University of Reading. We acknowledge that the sample population of consumers may not be representative of the "typical" UK consumer, however it does incorporate a broad range of culturally and ethnically diverse individuals that encompass a wide diversity of potential sensory genotypes.

On average, 50.9% were employed, and 46.9% were students. Of these, 27.2% were food and nutrition students from within the school. The multicultural nature of the staff and student body produced a diverse cohort, with 48.2% identifying as "white", 27.7% as "other" (i.e., non-white and/or European), and 6.7% of Chinese nationality. The remainder were composed of those regarding themselves as African (4.8%), Caribbean (2.3%), Indian (2.7%), or of mixed race (1%). Of the volunteers 3.3% declined to provide a response.

#### *2.5. Phytochemical Analyses*

#### 2.5.1. Preparation of Samples

Upon receipt of samples at the University of Reading, a subset of leaves (50 g) was taken for analysis, frozen at −80 ◦C, and lyophilized prior to extraction. Tissues were then milled into a fine powder using a Wiley Mini Mill (Thomas Scientific, Swedesboro, NJ, USA).

#### 2.5.2. Glucosinolate Analysis

DMB, GER, GRA, and GSV concentrations were determined by Liquid Chromatography Mass Spectrometry (LC–MS) as per the methodology presented by Bell et al. [20]. Three separate biological replicate extractions were performed on each sample, with three technical replicates analysed by LC–MS (*n* = 9). The data for each sample were then averaged to give a representative concentration of each assessment month. Individual sample averages were retained and used for subsequent PCA. Individual cultivar results for each respective month of the study can be found in Supplementary Data File S1.

#### 2.5.3. Sugar Analysis

Concentrations of fructose, galactose, and glucose were determined by extraction and analysis by capillary electrophoresis (CE) according to the methodology presented by Bell et al. [4]. The same level of replication as for the analysis of GSLs was used for each sample (*n* = 9) and averaged to produce a representative monthly concentration for presentation. As above, individual sample averages were retained and used for subsequent PCA. Individual cultivar results for each respective month of the study can be found in Supplementary Data File S1.

#### *2.6. Statistical Analysis*

#### 2.6.1. Panellist Performance

Data were collated and panellist performance evaluated using SenPAQ (v5.01; Qi Statistics, Reading, UK). For each monthly assessment, scores were averaged and used for further statistical analysis. Discrimination, repeatability, and consistency were checked for all assessors.

#### 2.6.2. Analysis of Variance

Shapiro–Wilk normality tests were conducted for all sensory and consumer variables. All of which were concluded to fit with a normal distribution and allow for statistical comparison using a parametric test. Analysis of variance (ANOVA) was performed on each data set (sensory, consumer, and phytochemical) and supplied temperature data. Each test was performed using XLSTAT (Addinsoft, Paris, France) with a protected post-hoc Tukey's honest significant difference (HSD) test (*p*-values ≤ 0.05). Only attributes with statistically significant differences were selected for presentation.

#### 2.6.3. Agglomerative Hierarchical Clustering

Agglomerative hierarchical clustering (AHC) was conducted on the consumer liking data using XLSTAT. This approach was used to cluster consumers who had similar liking patterns (for taste and overall liking) for each of the bimonthly panels. Dissimilarity of responses was determined by Euclidean distance, and agglomeration using Ward's method (set to automatic truncation).

#### 2.6.4. Principal Component and Correlation Analysis

Consumer liking and perception response data were used to extract principal components (PCs; with Varimax rotation) and we performed correlation analyses (Pearson, *n* − 1). Phytochemical, temperature, and agronomic data for each sample were regressed as supplementary variables within the PCA model. Variables such as month, cultivar (variety), cut, and country of origin were regressed as qualitative variables to generate categorical centroids within the model. Seven PCs were extracted with the first four components containing a cumulative 98.3% of variability. PCs 1 and 4 had eigenvalues of 4.1 and 0.3, respectively) and were selected for presentation after Varimax rotation. Correlation matrices of all attributes used in the analysis were produced at the 5%, 1%, and 0.1% significance levels, and are summarized in Supplementary Data File S1.

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

#### *3.1. Monthly Di*ff*erences in Rocket Agronomic Practices*

The majority of rocket supplied to the UK market is *D. tenuifolia*, with *E. sativa* making up a small amount. The latter is usually supplied in winter months due to its faster establishment, early vigour, and cold tolerance [10]. In this study *E. sativa* was only supplied in November (Table 1).

Cultivation practices varied distinctly between countries, and indeed between individual growers, based on local cultural practices and individual experience. In Italy, produce destined for the UK market is typically cultivated under polytunnel or glass, year-round; whereas UK grown material is either grown in open field or under glass (Table 1). One dominant reason for this difference is that the wetter and more humid climate of the UK can cause severe fungal pathogen outbreaks. The reduced airflow within polytunnels typically exacerbates this problem, and so open field is preferred to minimize losses.

The length of crop cycles depends on the season, though there are large differences between individual growers and countries (Table 1). Cycle length is longer in the winter and spring months, with much faster growth and regrowth in summer and autumn. The shortest average crop cycle in this study was 27 days (August), and the longest 96 (March). The extremes of the overall range (Table 1) can vary from 23 (June) to 180 days (April).

As establishment of rocket crops is more difficult in winter months, Italian growers favour repeated harvests until warmer weather arrives. It is not unusual for >5 cuts to be taken from a single sowing. During the experiment, sourced material came not only from Italy, but the USA and India (Table 1) in order to meet shortfalls in demand. During the summer season, UK rocket enters the market and typically has short growth cycles and receives only one cut. The humid climate does not favour regrowth, as damaged leaves become infected with fungal pathogens and are unsaleable.

The length of crop cycle and cut number have important implications for rocket taste, flavour, and acceptability. It is widely acknowledged that the more harvests a rocket crop undergoes, the more pungent and aromatic it becomes, due to the initiation of wound response and increases in secondary metabolites, such as GSLs [21]. However, no quantitative research has been conducted to evaluate consumer preferences for first, second, or multiple cut leaf material of rocket. As will be discussed in the following sections, cut number is a key determinant of taste and flavour perception, and liking of leaves at different times of the year.

#### *3.2. Monthly Variation in Rocket Growth Temperature*

Due to the seasonal distribution of rocket production geographically throughout a growing season, crops may be exposed to a range of temperature maxima and minima. Supplementary Figure S1 presents an average of the temperatures recorded at each farm location, giving a representative value of all growing sites for each month.

The highest average temperature across the growing season was 21 ◦C in August, with the highest average temperature in the week preceding harvest being 21.6 ◦C. The highest average maximum and minimum temperatures in the week preceding harvest were also in August; 27 ◦C and 15.6 ◦C, respectively. Lowest average temperatures were observed in January (avg. temp. 10 ◦C, max. temp. week 14.8 ◦C, min. temp. week 4.6 ◦C, and avg. temp week 9.8 ◦C). The significant differences observed in monthly temperatures correspond to distinct changes in phytochemical content, sensory perceptions, and consumer acceptance.

#### *3.3. Phytochemical Composition and Monthly Variability*

#### 3.3.1. Glucosinolates

The monthly average GSL concentrations of rocket leaves are presented in Figure 1a. For individual cultivar concentrations, see Supplementary Data File S1. The data show a very large amount of variability over the course of the year. This lack of consistency likely plays a significant role in the perceived quality changes in rocket produce by processors, supermarkets and consumers.

**Figure 1.** Average glucosinolate concentrations (**a**), sugar concentrations (**b**), and analysis of variance (ANOVA) pairwise comparisons (post-hoc Tukey's honest significant difference; (**c**) of rocket leaves observed on a monthly basis. Significant differences of glucosinolate concentrations between each sampling month are indicated by differing lower case letters within each column. Concentrations are expressed as mg·g−<sup>1</sup> of dry weight. Error bars represent standard error of the mean of each compound. See insets for compound colour coding (**a**,**b**). For individual cultivar composition data of samples received each month see Supplementary Data File S1.

Figure 1c details the significant differences between each growing month for GSL composition. Total concentrations of the four major GSLs of rocket were highest in October (17.6 <sup>±</sup> 0.6 mg g−<sup>1</sup> dw) and lowest in July (5.8 <sup>±</sup> 0.7 mg g−<sup>1</sup> dw). GRA concentrations were significantly higher in October (4.2 <sup>±</sup> 0.1 mg g−<sup>1</sup> dw) and February (3.3 <sup>±</sup> 1.3 mg g−<sup>1</sup> dw) compared with the months from March to September. GER concentrations were significantly higher in January (2.3 <sup>±</sup> 0.1 mg g−<sup>1</sup> dw) than at any other time of the study year. Previous studies in broccoli sprouts [22] have shown that cooler temperatures (<16 ◦C) increase the concentrations of methylthioalkyl GSLs, such as GRA and GER. This may be due to a abiotic stress response and upregulation of secondary metabolite biosynthesis, causing greater concentrations of these health related GSLs. GER, GRA, and their respective hydrolysis products are not known to have any significant odour or flavour, but the elevations observed in winter months suggests that cultivation in lower-temperature climates may improve rocket nutritional potential. Both sulforaphane (SF) and erucin (ERU; isothiocyanate hydrolysis products of GRA and GER, respectively) are known to be effective against some forms of cancer [23,24].

Concentrations of DMB were also significantly higher in January (2.6 <sup>±</sup> 0.2 mg g−<sup>1</sup> dw), October (4.2 <sup>±</sup> 0.2 mg g−<sup>1</sup> dw), and November (3.4 <sup>±</sup> 0.4 mg g−<sup>1</sup> dw), whereas amounts of the monomer GSV were highest in September (10.4±1.0 mg g−<sup>1</sup> dw). The relationship between GSV, DMB, and sensory properties is not understood, but previous studies have noted associations between GSV content and pungency (likely due to hydrolysis producing SAT), but not DMB [4]. The significant variations in monomer and dimer forms across the year suggest that there is some as-yet-unknown mechanism by which the two are interconverted [15]; possibly on a genetic and enzymatic level. This process may dictate the levels of pungency found in leaves.

#### 3.3.2. Sugars

The pattern of sugar accumulation in rocket leaves was much more distinct than for GSLs. Total concentrations were significantly higher from June to September (Figure 1b,c) indicating a strong relationship with seasonal climate. This could conceivably be linked to temperature and light intensity duration and quality during summer months. A study on broccoli [25] previously observed that glucose and fructose concentrations were significantly elevated under higher temperature conditions, for example.

Glucose was the dominant monosaccharide in rocket leaves, and concentrations were significantly higher from June to September (peaking in July, 93.9 <sup>±</sup> 3.0 mg g−<sup>1</sup> dw). Fructose accumulations also followed this pattern, with 27.5 <sup>±</sup> 0.9 mg g−<sup>1</sup> dw in July, compared to only 2.5 <sup>±</sup> 0.2 mg g−<sup>1</sup> dw in January. These data are strong evidence for the role of season and climate in the generation of sugars in rocket leaves; and as will be discussed, this has implications for preference and quality of leaves.

#### *3.4. Sensory Profiling Monthly Variability*

#### 3.4.1. Appearance Traits

Leaf size and uniformity of size were the only two appearance attributes tested that varied significantly between monthly assessments of rocket produce (*p* = 0.005 and <0.0001, respectively; Supplementary Figure S2 and Table 2). Leaf size was significantly smaller in January compared with April, June, August, and October. Similarly uniformity of size was significantly lower in January than any other month (with the exception of February). Combined with the low average temperatures (Supplementary Figure S1) at this time of year, it is likely that the colder temperatures and reduced light levels (short days) in Italy at this time of year result in slower, and more uneven growth rates [26] compared to other times of the year.



#### *Foods* **2020** , *9*, 1799

S2 for data and standard errors.

#### 3.4.2. Odour Traits

The odour attributes of rocket leaves defined as green, stalky, earthy, peppery, sweet, and mustard were all found to vary significantly between assessment months (Supplementary Figure S2 and Table 2). Green, peppery, and earthy odours were observed to be elevated, on average, in January, whereas stalky and sweet odours were scored higher in July, August, and September. Volatile profiles are known to be influenced by seasonal variations, and storage conditions [27], and so differences between the UK and Italian climates likely play a role in determining the intensity of these odours.

#### 3.4.3. Taste and Flavour Traits

Sour taste, savoury taste, stalky flavour, peppery flavour, and earthy flavour of rocket leaves were found to vary significantly between months (Supplementary Figure S2 and Table 2). Sour and savoury taste scores were highly variable between months, with no distinct pattern emerging according to seasonality. As with aroma attributes, stalky, peppery, and earthy flavours were each scored highest in September and January.

#### 3.4.4. Mouthfeel Traits

Significant variation was observed between monthly assessments of rocket for crisp and drying mouthfeels. Leaves tested in January were significantly less crispy than those received from March to October (Supplementary Figure S2 and Table 2). Soluble sugars are known to help maintain turgidity of leaves [28], and the low concentrations accumulated at this time of year may therefore be related to mouthfeel quality.

Drying sensation was perceived as significantly more in September and October than the months from March to June, and November. Little is known about the cause of drying sensation caused by rocket leaves, but one possible explanation is the presence of polyphenols [29], which have been observed to increase significantly under heat stress conditions [30].

#### 3.4.5. Aftereffect Traits

Aftereffect attributes with significant monthly variation are presented in Supplementary Figure S2 and Table 2. Of note are sweet and peppery aftereffects, which have previously been associated with improved consumer acceptance [9]. Sweet aftereffects were significantly higher in July, corresponding to the peak of glucose and fructose concentrations within leaves.

Peppery aftereffects were significantly higher in January, in agreement with the aroma and flavour scores for this attribute. Some GSL hydrolysis products are known to have different aromas at different concentrations [31], and the low abundances of GSV in January (Figure 1a) would suggest that SAT production may also be reduced, and correspond to reduced pungency and increased pepperiness.

#### *3.5. Correlation Analysis of Sensory Attributes*

#### 3.5.1. Growing Temperature

Correlation analyses and significances are presented in Supplementary Data File S1. Average crop cycle temperature, the minimum, and average temperatures in the week preceding harvest were significantly correlated with sweet odour of leaves (all *r* = >0.462; *p* = <0.0001). Abiotic stress is known to promote formation of secondary metabolites in many plant species [32] and so higher growth temperatures may promote the synthesis of aldehydes that impart sweet odour, as have been identified in other Brassicales species [33].

All temperature data were also significantly correlated with crisp mouthfeel (*r* = 0.527; *p* = <0.0001). Previous research and modelling of rapeseed plants has shown that growth temperature significantly impacts leaf morphology; particularly leaf length and thickness [34]. This may partly explain why

rocket leaves are perceived as crispier in summer and autumn months compared with winter (Supplementary Figure S2).

#### 3.5.2. Cultivation Practice

One of the largest differences observed between months was the length of the crop cycle (Table 1). Correlation analysis (Supplementary Data File S1) found that the length of the crop cycle was significantly associated with key sensory traits potentially linked with consumer acceptance. These were: bitter taste (*r* = 0.3; *p* = 0.043) and bitter aftereffects (*r* = 0.325; *p* = 0.027). Sweet aftereffects were also significantly and negatively correlated with the length of crop cycle (*r* = −0.316; *p* = 0.032). The low sugar:GSL ratio in samples with longer crop cycles might explain some of these correlations. With lower sugar concentrations in the winter/early spring months (Figure 1b), GSLs and their hydrolysis products may be perceived more strongly with the masking effect of sugars reduced.

#### 3.5.3. Glucosinolates

Individual GSL concentrations are known to be associated with sensory attributes of rocket species [3]. GRA is a compound not known to impart any taste or flavour [31], but correlation analysis revealed significant negative associations with sweet odour (*r* = −0.543; *p* = <0.0001), taste (*r* = −0.402; *p* = 0.005), and aftereffects (*r* = −0.304; *p* = 0.035). The abundance of GRA was negatively correlated with the average growth temperature (*r* = −0.344; *p* = 0.017) and max. temperature in the week preceding harvest (*r* = −0.306; *p* = 0.035). These two points indicate that GRA biosynthesis is lower in samples grown in months with higher temperatures, which also corresponds to increased sugar concentrations (Figure 1).

GER is similar to GRA in the respect that it is not known to impart taste [31], however its hydrolysis product erucin (ERU) has been described as having a "radish-like" aroma [16]. In this study, several previously unobserved associations were found. GER itself is negatively correlated with pungent odour (*r* = −0.299; *p* = 0.039), but positively with green and peppery odours (*r* = 0.459; *p* = 0.001, and *r* = 0.364; *p* = 0.011, respectively) and flavours (*r* = 0.367; *p* = 0.01, and *r* = 0.337; *p* = 0.019, respectively). While these data are not conclusive of a causative relationship with these attributes, it does suggest that occurrence of GER in high concentrations may elicit, or be associated with, perceptions of pepperiness and green attributes, and is worth studying in greater detail in future studies.

GSV exists in a monomer and dimer form (DMB), and typically makes up the largest proportion of the GSL profile of rocket [20]. A previous study found that its hydrolysis product SAT has a "rocket-like" aroma [16]. While this may be considered a somewhat subjective description, it is speculated that SAT is responsible for the perceived pungency of rocket leaves. The data in this study agreed with this hypothesis, as GSV concentrations were significantly correlated with pungent aroma (*r* = 0.393; *p* = 0.006). It was however also negatively correlated with peppery odour (*r* = −0.295; *p* = 0.042), suggesting that the two attributes are separate, with only GSV being indirectly responsible for pungency.

Correlations of DMB with sensory attributes were distinct and separate from the monomer, suggesting that concentrations of the two forms are influenced by the environment and as-yet-unknown genetic regulation, possibly in response to abiotic stress. It is unknown if DMB itself imparts taste or flavour, but its abundance was positively correlated with savoury taste (*r* = 0.323; *p* = 0.025) and aftereffects (*r* = 0.391; *p* = 0.006). This is in agreement with previous sensory and consumer studies of rocket [4,9]. It was also observed that GSV was significantly correlated with each of the four temperature measurements used in the analysis (Supplementary Data File S1; Supplementary Figure S1) whereas DMB was negatively correlated with the max. temperature in the week preceding harvest (*r* = −0.311; *p* = 0.031). This suggests that the relative abundances of the monomer and dimer forms of GSV had an environmental component, with greater concentrations of GSV present in hotter months.

#### 3.5.4. Sugars

Total sugars, fructose, and glucose concentrations were significantly correlated with dry matter percentage (*r* = 0.494; *p* = 0.001, *r* = 0.622; *p* = <0.0001, and *r* = 0.439; *p* = 0.003, respectively). This suggests that this physical property of leaves may be indicative of a dry matter concentration effect. This is reflected in several negative correlations with moistness mouthfeel (*r* = −0.385; *p* = 0.007, *r* = −0.515; *p* = 0.000, and *r* = −0.332; *p* = 0.021, respectively). Only galactose concentrations were significantly correlated with sweet taste (*r* = 0.39; *p* = 0.006), and fructose and galactose with sweet aftereffects (*r* = 0.303; *p* = 0.036, and *r* = 0.308; *p* = 0.033, respectively). Despite the significantly higher sugar concentrations in summer months (Figure 1b,c) there were no significant correlations with growth temperature.

The sugar:GSL ratio was also similarly correlated with the aforementioned mouthfeel effects (Supplementary Data File S1) and dry matter content (*r* = 0.571; *p* = <0.0001); but only sweet aftereffects (*r* = 0.393; *p* = 0.006) and not sweet taste. This association is not as strong as found in previous studies of rocket [4].

#### *3.6. Consumer Acceptability and Perception*

#### 3.6.1. Liking of Taste

Consumer liking of taste and the results of AHC are presented in Supplementary Table S2. Liking of taste is defined as liking associated with taste and flavour attributes alone (bitterness, sweetness, pepperiness, and hotness), irrespective of appearance traits. Three clusters were identified in each respective month, except for March, where four clusters were observed. The largest clusters in each month consistently scored cultivars higher for taste liking than the overall cohort and monthly averages. This indicates that for most consumers, the taste of rocket is acceptable year-round, with average scores consistently >6.0.

March and April/May had significantly lower taste liking scores than any of the other months. Highest average taste liking was in January, which is contrary to our hypothesis that rocket liking would be greater during summer months. Average scores for July, September, and November were also relatively high (6.0, 6.2, and 6.2, respectively), indicating that in terms of consumer taste liking, spring months show a distinct reduction in acceptability.

Crop cycles of rocket in spring are also typically very long (96.3 days, average) with successive cuts (>2), potentially producing very pungent and bitter leaves. Figure 2 presents consumer perception data of bitterness, hotness, and pepperiness. All these attributes were scored highest in March, with bitterness being a dominant attribute until July. Sweetness perception by comparison remained relatively unchanged, peaking in July. The reason for increased taste liking in January may therefore be explained by the significantly lower perception of hotness of leaves relative to spring and summer months.

**Figure 2.** Consumer perceptions of bitter, hotness, sweetness, and pepper attributes of rocket leaves on a bimonthly basis over a growing season. Inset table presents the results of analysis of variance (ANOVA) pairwise comparisons (post-hoc Tukey's honest significant difference). Significant differences for each perception attribute are indicated by differing lower case letters within rows. See inset for colour coding of attributes. Values are presented as normalized averages of each respective consumer assessment. See Supplementary Table S1 for the numbers of participants in each respective consumer panel.

#### 3.6.2. Overall Liking

Table 3 presents AHC data and average monthly scores for overall rocket leaf liking. Overall liking encompasses liking of both taste and appearance attributes. Analysis identified three groups for each respective month, except for January, where five clusters were observed. Appearance of leafy salads is known to be a significant factor in consumer purchase intent and liking [2] and the sensory panels determined significantly smaller leaf size and uniformity of shape in January (Supplementary Figure S2). This disparity between cultivars seems to be compensated for by higher taste liking (Supplementary Table S2), suggesting that appearance liking may be secondary to taste liking for some consumers; for example in cluster 4 January (*n* = 43) where scores were all consistently higher than the total cohort average.

The consistency of cultivars was extremely variable during the March and April/May panels. "Fast Grow" (-; 6.8) was scored significantly higher on average than all the other samples in March and preferred by all three cluster groups. A similar pattern of inconsistency was observed in the autumn months of September and November. This further suggests that seasonal transitions result in more variable rocket produce.



significant difference. N.B. Hyphens (-) indicate data was not supplied from the grower.

#### *3.7. Principal Component Analysis*

#### 3.7.1. Relationships between Consumer Liking and Perceptions

Despite sensory panels not detecting significant differences in sweet or bitter tastes, consumers were able to do so, and this significantly affected their liking for rocket throughout the growing season. This is likely due to the increased diversity of taste receptor profiles present within the population compared with the sensory panel [35].

PCA of the data sets from each consumer panel month revealed a distinct separation between sweetness perception and hotness, bitterness, and pepperiness perceptions along PC1 (Figure 3a). Taste and overall liking are in turn more positively associated with sweetness perception in the upper left quadrant along the PC4 axis, which is in agreement with previous observations in rocket [9]. These data are therefore strong evidence that most consumers are likely to reject rocket if it is too pungent and bitter, as is found in samples received in March and April/May.

#### 3.7.2. Influence of Growing Temperatures on Consumer Preference and Perceptions

Despite the higher sugar concentrations in July, this does not colocalize with overall/taste liking within the PCA (Figure 3b). The months of March and April/May are in fact most negatively associated with consumer liking, and January and November positively associated with these. Therefore, rocket produced at cooler temperatures is more likely to be preferred by consumers, as bitterness and hotness perceptions are likely to be lower in these months (Figure 3a).

#### 3.7.3. Influence of Cultivation Practice on Consumer Liking and Perceptions

Figure 3b shows that cut number also explains some separation for taste and overall liking. Rocket leaves that were of first cut are generally more common in the upper left quadrant of the PCA plot, and more closely associated with taste liking and sweetness perception (Figure 3). Leaves with more than two cuts separate in the opposite direction towards the lower right quadrant, in the direction of bitter/hotness/pepperiness perception. Anecdotal evidence of traditional cultivation practices by growers has suggested that second cuts (and above) are preferred, because leaves are more uniform, more greatly serrated in shape, and have a more intense flavour. These assertions are in agreement with this study; however none are associated with positive consumer liking or taste liking of rocket leaves. While a subset of consumers may prefer cultivars with increased hotness (as seen in AHC analysis; Supplementary Table S2 and Table 3) consumers generally do not like this attribute, and prefer milder, sweeter leaves. Thus, conventional agronomic practices of harvesting multiple cuts of rocket may be detrimental to consumer acceptance; particularly in spring months when crop cycles are longer, and the growing season is transitioning.

There is also a significant gap in research more generally about the response of rocket species and cultivars to differences in growth environments and cultivation practices. In this study cultivars were supplied from various growers, and this variable was not controlled so as to assess "real-world" differences in rocket consistency as supplied to consumers. Future studies should aim to assess the variability of multiple cultivars across growing regions, and sample multiple cuts. Such studies are logistically difficult to organise, however it would provide valuable information on how environment influences GSL and hydrolysis product formation on a genotypic basis and how performance of cultivars varies according to the environment. Controlled environment studies have begun to explore these effects [17], but none have to date accounted for variances in soil composition or climatic conditions in the field.

**Figure 3.** Principal component analysis of consumer liking and perception data of rocket cultivars over the course of one growing season. Biplots display principal components (PCs) 1 and 4, which represent 68.1% of variation within the data; (**a**) factor loadings plot and (**b**) factor scores plot. In (**a**) red circles represent consumer liking and perception attributes, blue circles average monthly samples, and teal squares supplementary phytochemical, and cultivation temperature. Variable label symbols (refer to Table 1): = 1st cut; - = second cut; = 2nd + cut;= <30 day crop cycle; = 31–60 day crop cycle; - = 61–90 day crop cycle; = >91 day crop cycle. In (**b**), see inset for monthly colour coding of samples. Black and red diamonds indicate supplementary variable centroids.

#### 3.7.4. Influence of Glucosinolate Contents on Consumer Liking and Perceptions

Higher concentrations of GER are associated with higher taste and overall liking of rocket leaves (Figure 3a). This trend is in opposition to the abundance of GSV, which is in the bottom right quadrant with bitter/hotness/pepper perceptions. DMB is also separated from GSV and negatively associated with hotness, bitterness, and pepperiness perceptions. This is new evidence that suggests the ratio between monomer and dimer forms of GSV may play a significant role in determining consumer acceptability of rocket leaves. Nothing is known of the genetic mechanisms responsible for the biosynthesis of GSV and DMB, or the mechanisms responsible for determining their relative abundances; but it is generally accepted that SAT (derived from GSV and responsible for pungency) is produced from the monomer form [15]. This may therefore explain the association between GSV and perception traits and indicates that DMB has no objectionable taste of its own.

#### **4. Conclusions**

This study has found evidence for significant sensorial variability in rocket leaves produced over the course of a growing season as a result of varied cultivation practices and growing locations. This in turn results in variations in consumer liking, which may influence purchase intents and repurchase of rocket leaf products. Seasonal practices, such as growth temperature and the number of cuts crops received, underlie changes in phytochemical composition and may result in the production of overly pungent leaves that consumers are likely to reject. To produce more consistent and acceptable rocket leaves, the practice of multiple harvests should be reserved for developing products targeted at those consumers who like high pungency leaves. First cuts tend to be milder and could therefore be marketed to a wider set of consumers that prefer sweeter leaves and low levels of pungency and bitterness.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/9/12/1799/s1. Supplementary data file S1: Pearson *n* − 1 correlation tests at three levels of significance (*p* = ≤0.05, ≤0.01, and ≤0.001) and month-by-month individual cultivar glucosinolate and sugar concentrations of rocket samples used in sensory and consumer analyses. Figure S1: Monthly average temperature data of rocket crop cultivation areas over the course of one year, and analysis of variance (ANOVA) pairwise comparisons (post-hoc Tukey's honest significant difference). Averages represent temperatures at multiple growing locations in a given month. Significant differences are indicated by differing lower case letters. Max. temp. week, min. temp. week, and avg. temp. week refer to average temperature values in the week preceding harvest of each rocket crop. Figure S2: Sensory analysis results for appearance (a), odour (b), taste (c), flavour (d), mouthfeel (e), and aftereffect (f) traits of rocket leaves analysed monthly over the course of a growing season. See insets for individual line chart colour coding of attributes. Values are presented as averages of each respective monthly sensory assessment (*n* = 12 panellists). Table S1: Consumer demographics and characteristics for each of the rocket study panel months. Table S2: Taste liking of rocket cultivars for the clusters of consumers obtained from agglomerative hierarchical clustering.

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

**Funding:** This research was funded by Bakkavor Group Ltd. (Spalding, UK).

**Acknowledgments:** The authors would like to thank Lorraine Shaw and Chris Jeffes of Bakkavor Ltd. (Spalding, UK) for arranging supply and delivery of rocket leaves. Special thanks to Suzanne Bird for performing glucosinolate and sugar analyses, and help running sensory and consumer panels.

**Conflicts of Interest:** This work was conducted with funds provided by Bakkavor Ltd. (Spalding, UK). The authors declare that the research was conducted in the absence of any financial relationships that could be construed as a potential conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; or in the writing of the manuscript.

#### **Abbreviations**


#### **References**


35. Reed, D.R.; Tanaka, T.; McDaniel, A.H. Diverse tastes: Genetics of sweet and bitter perception. *Physiol. Behav.* **2006**, *88*, 215–226. [CrossRef]

**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*
