Allyl ITC (Allyl-ITC; ≥99%), benzonitrile (phenyl-CN, ≥99.9%), benzyl cyanide (98%), benzyl isothiocyanate (98%), 3-butenenitrile (allyl-CN, ≥98%), Coomassie brilliant blue R staining solution, D/L-dithiothreitol (DTT), iron (II) sulphate heptahydrate (FeSO₄(H₂O)₇, ≥99%), isopropyl-β-d-thiogalactopyranoside (IPTG, ≥ 99%), kanamycin sulfate, myrosinase (thioglucosidase from Sinapis alba seeds, ≥100 units/g) and sodium carbonate (Na₂CO₃, ≥ 99%) were obtained from Sigma-Aldrich Chemie GmbH (Steinheim, Germany); acetic acid (supra quality, 100%), allyl GLS (sinigrin monohydrate, ≥99%), ampicillin sodium salt (≥99%), benzyl GLS (>99%), vitamin C (l-(+)-ascorbic acid, ≥99%), methylene chloride (GC Ultra Grade), NaOH (≥98%), N-2-hydroxyethylpiperazine-N’-2-ethane sulphonic acid (HEPES, ≥ 99.5%), 2-(N-morpholino)-ethane sulphonic acid (MES, ≥ 99%), 3-(N-morpholino)-propane sulphonic acid (MOPS, ≥ 99.5%), N,N-bis-(2-hydroxyethyl)-glycine (BICINE, ≥ 99%), sodium hydrogen carbonate (NaHCO₃, ≥ 99.5%) were purchased from Carl Roth GmbH + Co. KG (Karlsruhe, Germany); Na₂SO₄ anhydrous (≥99%) was obtained from VWR International GmbH (Darmstadt, Germany); acetonitrile (LC-MS grade) was purchased from Th. Geyer GmbH & Co. KG (Renningen, Germany). 4-(Methylthio)butyl GLS (4MTB, ≥ 90%), 4-(methylsulfinyl)butyl GLS (4MSOB, ≥ 95.0%) and indol-3-ylmethyl GLS (I3M, ≥ 98%) were obtained from Phytolab GmbH and Co. KG, Vestenbergsgreuth, Germany. 4-(Methylthio)butyl ITC (4MTB-ITC, ≥ 98%) was purchased from Santa Cruz Biotechnology (Heidelberg, Germany). 5-(Methylthio)pentanenitrile (4MTB-CN) and 5-(methylsulfinyl)pentanenitrile (4MSOB-CN) (all ≥ 95% purity) were purchased from Enamine (SIA Enamine, Riga, Latvia). 4-(Methylsulfinyl)butyl ITC (4MSOB-ITC) was from Toronto Research Chemicals (Toronto, Canada). Gateway™ LR clonase™ II enzyme mix, GeneJET gel extraction kit, pENTR-D/TOPO, Pierce™ Bradford protein assay kit and indole-3-carbonitrile (I3N, 98%) were obtained from Thermo Fisher Scientific (Germany). The Gateway compatible pMAL-c2 was from New England Biolabs.

MS grade solvents and ultrapure water were utilized in all experiments.

The A. thaliana NSP1 (At3g16400.1), NSP2 (At2g33070.3), NSP3 (At3g16390.1), NSP4 (At3g16410.1) and NSP5 (At5g48180.1) nucleotide and amino acid sequences were retrieved from TAIR (https://www.arabidopsis.org/) and used as a query against the B. oleracea var. oleracea genome at NCBI using the BLASTn and BLASTp search tools (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The BLASTp searches were done against the non-redundant protein sequences database (nr). The protein sequences of the putative BoNSPs identified at NCBI were used as a query against the B. oleracea genome at Ensembl Plants with the BLASTX search tool (https://plants.ensembl.org/Brassica_oleracea/Tools/Blast). The Kelch domains in the putative BoNSPs were predicted manually based on multiple sequence alignment with the full amino acid sequence of TaTFP (Gumz et al., 2015; Kuchernig et al., 2011). The amino acid sequences upstream of the predicted Kelch domain regions were extracted and the jacalin domains were predicted by InterPro 106.0 (Blum et al., 2024) (released on 19 June 2025). The Kelch and jacalin domains in the putative BoNSPs were schematically depicted using IBS2.0 (Xie et al., 2022). The amino acid sequence identity between the putative BoNSP and AtNSP isoforms was assessed by multiple sequence alignment of their predicted Kelch domain sequences using ClustalW in Clustal Omega (Madeira et al., 2024) and visualized in Jalview version 2.11.4.1 (Waterhouse et al., 2009). The molecular weight of XP_013583566.1 and XP_013589780.1 not detected in Ensembl Plants was calculated using ExPASy ProtParam (Gasteiger et al., 2005).

To investigate the evolutionary history of the Kelch domain region, the amino acid sequences of the putative BoNSPs and AtNSPs were trimmed manually to obtain the sequence parts representing only the Kelch domain region. The Kelch domain regions of these specifier proteins as well as the putative ancestral proteins (At3g07720 and B. oleracea homolog) were aligned using the L-INS-i iterative refinement method in MAFFT version 7 (Katoh et al., 2019; Kuraku et al., 2013) and the phylogenetic tree was constructed using the Maximum likelihood algorithm and JTT matrix-based model with 1000 bootstrap replicates in MEGA11 (Tamura et al., 2021) using Vitis vinifera XP_002267128.1 as an outgroup. The Maximum Parsimony method was automatically applied to generate the initial tree(s) for the heuristic search. The differences in evolutionary rates among different sites were modelled using a discrete Gamma distribution (5 categories (+G, parameter = 1.3173)). Twenty sequences were analyzed and the complete deletion option was applied to remove all positions with gaps and missing data leaving a final dataset with 304 positions.

The jacalin domains were used separately for phylogenetic analysis. The jacalin domains, predicted by InterPro (Blum et al., 2024), were manually extracted, supported by the Group Protein tool in Sequence Manipulation Suite (Stothard, 2000) to locate the jacalin regions. For BoNSP candidates with more than one jacalin domain, the domains were assigned single-letter codes as done previously (Burow et al., 2009; Kuchernig et al., 2012), but starting from the C-terminus and analyzed separately. A total of twenty-one jacalin domains from the putative BoNSPs were identified. A significantly shorter jacalin domain (BoNSP6c_(Bo3g181230.1), 94 amino acids) (Figure 1A and Supplementary Table S4) was excluded from the phylogenetic analysis. To select representative jacalin domains for further analysis, the jacalin domains were first aligned as was done for the Kelch domains. The alignment was then used to construct a phylogenetic tree using the Neighbor-joining algorithm with 1000 bootstrap replicates in MEGA11 (Tamura et al., 2021). To search for related myrosinase binding proteins encoded in the B. oleracea genome, six jacalin domains ((BoNSP1b_(XP_013588183.1), BoNSP4b_(XP_013627380.1), BoNSP5b_(XP_013627381.1), BoNSP5a_(XP_013627381.1), BoNSP12_(XP_013587058.1) and BoNSP15b_(XP_013589780.1)) (Supplementary Table S4) were selected as representatives of the twenty jacalin domains based on their separate grouping in an initial Neighbour-joining tree. BLASTp searches using the six jacalin domain sequences identified three putative BoMBP2 isoforms (designated as BoMBP2-1, XP_013609657.1; BoMBP2-2, XP_013609679.1 and BoMBP2-3, XP_013624973.1) in the B. oleracea var. oleracea genome at NCBI (Supplementary Table S4). Using the three putative BoMBP2 sequences, three BoMBP orthologues were identified in the Oryza sativa Japonica Group (rice) genome and two in the Populus trichocarpa (poplar) genome. In all the BLAST searches, the nr database was searched and only the protein that appeared as the top hit was retrieved, except for rice, where a hypothetical protein (GenBank: EEE64302.1), which appeared as a top hit for the putative BoMBP2-3 (XP_013624973.1), was excluded, and the jacalin-related lectin isoform X1 protein (XP_066166746.1), which emerged as the second hit, was retrieved. The jacalin domain sequences from the candidate BoNSPs and AtNSPs (analyzed separately if the protein had more than one jacalin domain), the full-length protein sequences of the putative BoMBP isoforms, the BoMBP orthologues from rice and poplar, AtMBP1 (At1g52040.1), AtMBP2 (At1g52030.1), and Oryza sativa Japonica Group NP_001396161.1, used as an outgroup, were aligned using the iterative refinement method L-INS-i in MAFFT (Katoh et al., 2019; Kuraku et al., 2013). The phylogenetic tree was constructed using the Maximum likelihood algorithm and WAG + G (Whelan and Goldman, 2001) model with 1000 bootstraps in MEGA11 (Tamura et al., 2021). The Maximum Parsimony method was automatically applied to generate the initial tree(s) for the heuristic search. Moreover, the differences in evolutionary rates among different sites were modelled using a discrete Gamma distribution (5 categories (+G, parameter = 2.7148)). Thirty-six sequences were analyzed and the partial deletion option was applied to remove all positions with less than 95% site coverage, leaving a final dataset with 127 positions.

The BoNSP2 (XP_013609641.1) and BoNSP11 (XP_013587057.1) cDNAs were commercially synthesized (Eurofins Genomics GmbH, Cologne, Germany). The cDNAs were amplified using gene-specific primers (listed in Supplementary Table S1), cloned into the pENTR-D/TOPO Gateway entry vector and confirmed by whole plasmid sequencing (Eurofins Genomics GmbH, Cologne, Germany). For protein expression, the coding sequences were then recombined into the Gateway-compatible version of pMAL-c2 with an N-terminal maltose-binding protein-tag. The recombinant BoNSP2 and BoNSP11 were expressed in Escherichia coli BL21 cells by overnight induction in LB medium with 0.5 mM isopropyl β-d-1-thiogalactopyranoside at 220 rpm and 16 °C. The bacterial cultures were then centrifuged for 15 min at 4000 g and 4 °C, the supernatant discarded, the bacteria pellet lysed by sonication and the heterologously expressed proteins purified using amylose resin following the manufacturer’s instructions. Recombinant proteins were eluted with maltose binding protein elution buffer composed of 20 mM Tris-HCl (pH 8), 150 mM NaCl, 1 mM dithiothreitol (DTT) and 20 mM maltose. Protein expression and purity of the eluted proteins were assessed using SDS-PAGE followed by Coomassie blue staining. The Bradford protein assay (Bradford, 1976), using bovine serum albumin as a standard, was used to determine the concentration of the eluted proteins. After adjustment of the protein concentration to 0.5 µg/µl in deionized water and glycerol added to a final concentration of 10% (v/v), the proteins were flash-frozen in liquid nitrogen and stored at – 80 °C.

The protocol used previously (Witzel et al., 2019) to determine the BoESP activity was slightly adapted for NSP activity. NSP activity was assessed in assays containing 50 µl of purified recombinant BoNSP2 or BoNSP11 (25 µg) mixed with 10 µl of 25 mM vitamin C solution, 50 µl of 10 mM allyl GLS solution and 350 µl of 50 mM sodium acetate (NaAc) buffer (pH 5.5) containing 1 mM DTT and 50 µM Fe²⁺ (added as iron (II) sulphate heptahydrate). All reactions with BoNSP2 were started by addition of 50 µl of 0.5 U/ml Sinapis alba myrosinase. Similarly, all reactions with BoNSP11 were started by the addition of 50 µl of 0.28 U/ml Sinapis alba myrosinase. According to the manufacturer, 1 U was defined as the quantity of Sinapis alba myrosinase which catalyzes the production of 1 µmol of glucose per minute from allyl GLS at pH 6 and 25 °C. The units of myrosinase used for the assays with BoNSP2 and BoNSP11 were calculated based on the enzymatic activity specified for each batch by the manufacturer (Batch BCBQ2804V – 495 U/g for BoNSP2 and Batch BCCG4678 – 279.6 U/g for BoNSP11). After 1 h of incubation at 25 °C, internal standard (0.2 µmol phenyl-CN) was added and the GLS hydrolysis products were extracted thrice using methylene chloride and analyzed by a gas chromatography - mass spectrometry (GC-MS) system (7890 A GC with 5975C Inert XL MSD, Agilent Technologies Deutschland GmbH, Waldbronn, Germany) using an HP-5MS Ultra Inert column (30 m x 0.25 mm x 0.25 μm; Agilent Technologies, Waldbronn, Germany) as described previously (Hanschen, 2024). Since CNs are also formed in the absence of NSPs, controls containing all reaction components except BoNSPs were conducted. NSP activity was assessed as the % allyl-CN relative to all detected allyl GLS hydrolysis products (Allyl-CN, Allyl-ITC) and calculated as follows: (% Allyl-CN = ([Allyl-CN]/([Allyl-ITC] + [Allyl-CN]) * 100%). This protocol was modified for the different analyses carried out in this study, and the specific details are outlined in the following sections. Each BoNSP was expressed, purified and characterized three times independently.

To assess the effect of ferrous ion (Fe²⁺ - added as iron (II) sulphate heptahydrate) concentration on BoNSP activity, the NSP assay was performed as described above with some modifications. For BoNSP2, the assays were performed without and with Fe²⁺ supplementation to a final concentration of 6.86 µM, 13.7 µM, 34.3 µM, 68.6 µM, 137 µM and 343 µM. To determine whether BoNSP2 activity is strictly dependent on the added iron, EDTA was added to a final concentration of 250 µM and 500 µM in assays containing 34.3 µM Fe²⁺. The experiment was also conducted for BoNSP11 but using only 34.3 µM Fe²⁺ and 343 µM Fe²⁺. The effect of EDTA was tested at 250 µM in assays containing 34.3 µM Fe²⁺.

NSP activity was investigated as described above using allyl GLS, 4-(methylthio)butyl GLS (4MTB GLS), 4-(methylsulfinyl)butyl GLS (4MSOB GLS), benzyl GLS and indol-3-ylmethyl GLS (I3M GLS) as substrates for myrosinase. The fold-increase in CN formation was calculated by dividing the amount of CNs (in µmols) recovered in the assay with BoNSP divided by the CNs (in µmols) recovered in the control assays with all reaction components except the recombinant BoNSP.

To investigate the influence on pH on BoNSP2 activity, the NSP assay was performed as described above but with six buffers, NaAc, 2-(N-morpholino)-ethane sulphonic acid (MES), 3-(N-morpholino)-propane sulphonic acid (MOPS), N-2-hydroxyethylpiperazine-N’-2-ethane sulphonic acid (HEPES), N,N-bis-(2-hydroxyethyl)-glycine (BICINE) and carbonate-bicarbonate buffer (Na₂CO₃/NaHCO₃), with overlapping pH ranges. The final concentration of all six buffers in each sample was 34.3 mM and the pH values used for each buffer were as follows: NaAc buffer pH 5.5; MES buffer pH 5.5, pH 6 and pH 6.5; MOPS buffer pH 6.5, pH 7 and pH 7.5; HEPES buffer pH 7.5 and pH 8; BICINE buffer pH 8, pH 8.5 and pH 9, and Na₂CO₃/NaHCO₃ buffer pH 9, pH 9.5, pH 10 and pH 10.5.

Although NaAc has an effective pH range of 3.7 to 5.6, the sensitivity of BoNSP2 and BoNSP11 to changes in pH was also assessed in NSP assays set up as outlined above but with 50 µl of a 12 mM allyl GLS solution and NaAc solution set from pH 4 to pH 12 (for BoNSP2) and pH 4 to pH 9 (for BoNSP11), by the addition of acetic acid or NaOH.

Statistical analysis was performed using SigmaPlot for Windows Version 14.0, Build 14.0.3.192 (Systat Software, Inc., San Jose, CA, USA). Normal distribution was assessed using the Shapiro-Wilk test and the homogeneity of variance by the Brown-Forsythe test. If normality and homogeneity of variance were confirmed, one-way ANOVA followed by Tukey’s posthoc test (p≤0.05) was used. If normality or homogeneity of variance, or both, were not confirmed, the non-parametric Kruskal−Wallis ANOVA followed by Dunn’s post-hoc test was used instead. Homogeneous groups were assigned using the multcompLetters() function in R (v4.4.3) (Posit Software, PBC, Boston, MA, USA).

In a first approach, a search for putative BoNSPs was performed in the B. oleracea var. oleracea genome at NCBI using nucleotide and amino acid sequences of AtNSP1–AtNSP5 as a query. The retrieved amino acid sequences of putative BoNSPs were then used for a search in the Ensembl Plants database. From the database search, a total of seventeen BoNSP candidates were identified. The number was revised to sixteen (Table 1) after identification and exclusion of the likely ancestral protein of the BoNSP family through phylogenetic analysis, which will be detailed in the following section. Five of the sixteen BoNSP isoforms were previously found to be expressed in mature kohlrabi (Mbudu et al., 2025) (Table 1). Two proteins encoded in the B. oleracea var. oleracea genome and previously annotated as myrosinase-binding protein 2-like (isoforms XP_013627380.1 and XP_013627381.1) were also designated as putative BoNSPs as they possess N-terminal jacalin-like lectin domains and Kelch domains, which is characteristic of NSPs (Burow et al., 2009). Two candidates (XP_013583566.1 and XP_013589780.1) were only found at NCBI (Table 1). From Ensembl Plants, two additional proteins (Bo3g181230.1 and Bo5g126040.1) with significant sequence similarity to the BoNSP candidates from NCBI were identified. Protein Bo5g125020.1 in Ensembl Plants, which has a length of 542 amino acids, starts with a proline suggesting an inaccurate gene model. Therefore, Bo5g125020.1 was not included in further analyses. However, protein XP_013583566.1 in NCBI was 100% identical to the 495 C-terminal amino acids of Bo5g125020.1, indicating that this sequence is derived from the same genomic position (Table 1). Similarly, the genomic position of XP_013589780.1 was determined based on the genomic position of proteins with partial sequence identity found in Ensembl Plants: (1) Bo6g027570.1, 100% identical to the C-terminal 287 amino acids of XP_013589780.1, and (2) Bo6g027580.1, 100% identical to the N-terminal 478 amino acids of XP_013589780.1 (Table 1). Thirteen putative BoNSPs (XP_013588183.1, XP_013609641.1, XP_013620038.1, XP_013627380.1, XP_013627381.1, Bo3g181230.1, XP_013631262.1, XP_013583566.1, XP_013587057.1, XP_013587058.1, XP_013585314.1, XP_013589780.1 and XP_013593056.1) for which complete protein sequences were obtained based on seemingly correct gene models and whose genomic position could be assigned with sufficient certainty (Table 1) were analyzed in more detail.

Based on the information in Ensembl Plants and NCBI, the thirteen putative BoNSPs selected for further analyses, consist of 322 to 1064 amino acids (Table 1) and have a molecular weight of 34.8 kDa to 116.4 kDa. Manual prediction of the Kelch domains based on the sequence of TaTFP (GenBank: AEL16674.1) (Gumz et al., 2015; Kuchernig et al., 2011) (Supplementary Figure S2) and prediction of the jacalin domains by InterPro (Blum et al., 2024) revealed that all candidate BoNSPs were composed of a Kelch domain and nine of them possessed up to five jacalin-like lectin domains at the N-terminus of their Kelch domain (Figure 1A).

To elucidate the evolutionary relationships among the putative BoNSPs, a phylogenetic tree based on the Maximum likelihood algorithm was generated from the Kelch domains of the thirteen putative BoNSPs, five AtNSPs (AtNSP1–AtNSP5), At3g07720.1, the ancestor of the AtNSPs (Burow et al., 2009), and Vitis vinifera XP_002267128.1 as an outgroup (Figure 1B). At the base of the tree, XP_013585314.1 grouped together with At3g07720.1 (Figure 1B). As At3g07720.1 does not seem to possess specifier protein activity (Burow et al., 2009), this suggests that XP_013585314.1 represents the corresponding ancestral protein in B. oleracea and may have its primary role outside of the GLS-myrosinase system. XP_013585314.1 was therefore not designated as a putative BoNSP. Both At3g07720.1 and XP_013585314.1 have only the Kelch domain, supporting the hypothesis that the presence of jacalin domains in NSPs may be a derived state (Burow et al., 2009; Kuchernig et al., 2012). Having excluded XP_013585314.1 as a candidate BoNSP, the rest of the putative BoNSPs were assigned consecutive numbers (BoNSP1–BoNSP16) (Table 1) based on their genomic position inferred from their Ensembl Plants IDs (Table 1). This resulted in a different numbering scheme than previously used (Mbudu et al., 2025). The new and the previous names are indicated in Table 1. Multiple sequence alignment of the Kelch domains from the thirteen putative BoNSPs included in the phylogenetic analysis (BoNSP1–BoNSP7, BoNSP9, BoNSP11, BoNSP12, BoNSP15 and BoNSP16) (Table 1) and AtNSPs using Clustal Omega, revealed 40.1% to 89.4% amino acid sequence identity among putative BoNSPs and 40.2% to 86.5% sequence identity with the AtNSPs (Supplementary Table S3).

The other branch of the phylogenetic tree had two major clades: one with AtNSP5 as well as BoNSP3 and BoNSP16 and another one with two subclades. All proteins in the AtNSP5 clade lack a jacalin domain. In spite of the low branch support for AtNSP5, the grouping of BoNSP3 and BoNSP16 with AtNSP5 might indicate that proteins in this clade are the oldest functional NSPs (Kuchernig et al., 2012). In contrast, all proteins in the other major clade (AtNSPs and putative BoNSPs, except BoNSP7) possess one or several jacalin domains (Figure 1A). Although the phylogenetic tree was built using only the Kelch domain regions, the BoNSP candidates in the two subclades of this major clade group according to their numbers of jacalin domains. One subclade contains the four BoNSP candidates with three to five jacalin domains (BoNSP4–BoNSP6 and BoNSP15) and no AtNSPs. The other subclade comprises six BoNSP candidates (BoNSP1, BoNSP2, BoNSP7, BoNSP9, BoNSP11 and BoNSP12) with a maximum of two jacalin domains (Figure 1A) and AtNSP1–AtNSP4 with one or two jacalin domains. Although the two subclades have a high bootstrap support, the positioning of BoNSP7 in the tree is surprising as this is the only BoNSP in the clade that lacks a jacalin domain.

Past reports suggest that the jacalin domains in AtNSPs may derive from MBPs and may be from different sources (Burow et al., 2009; Kuchernig et al., 2012). To test whether this is likely for the putative BoNSPs, phylogenetic analysis of the jacalin domains was performed similarly to what has been described before (Burow et al., 2009; Kuchernig et al., 2012). A total of twenty-one jacalin domains from nine candidate BoNSPs (BoNSP1, BoNSP2, BoNSP4–BoNSP6, BoNSP9, BoNSP11, BoNSP12, and BoNSP15) (Figure 1A and Supplementary Table S4) were identified. If a BoNSP candidate had more than one jacalin domain, domains were assigned single-letter codes (a–e), starting at the C-terminus of the protein. One jacalin domain (BoNSP6c_(Bo3g181230.1), 94 amino acids), which was significantly shorter than the average length of the twenty-one jacalin domains (139 amino acids) (Figure 1A and Supplementary Table S4), was excluded from the phylogenetic analysis. A phylogenetic tree was generated from the twenty putative BoNSP and five AtNSP jacalin domain sequences, and the full-length protein sequences of the three putative BoMBP2 isoforms (designated as BoMBP2-1, XP_013609657.1; BoMBP2-2, XP_013609679.1 and BoMBP2-3, XP_013624973.1), three BoMBP2 orthologues in rice and two in poplar (Supplementary Table S4), AtMBP1 (At1g52040.1) and AtMBP2 (At1g52030.1) (Supplementary Figure S5). The tree topology strongly supported the branching off of the B. oleracea and A. thaliana sequences from the rice and poplar sequences. The B. oleracea and A. thaliana sequences formed two well supported clusters. One cluster (cluster A) was composed of sequences from both species and contained the jacalin domains located directly upstream of the Kelch domain region of all included putative BoNSPs and AtNSPs. In addition, several further jacalin domains (second, third or fourth jacalin domains upstream of the BoNSP Kelch domain region and jacalin domain “a” of AtNSP4) as well as AtMBP1, AtMBP2 and two BoMBP2 isoforms (BoMBP2-2 (XP_013609679.1) and BoMBP2-3 (XP_013624973.1)) belonged to this cluster. The other cluster (cluster B) was confined to B. oleracea sequences and contained only BoNSP jacalin domains located not directly adjacent to the Kelch domain region as well as BoMBP2 isoform BoMBP2-1 (XP_013609657.1). Again, this included second, third, fourth or fifth jacalin domains upstream of the BoNSP Kelch domain region. Thus, the jacalin domains directly adjacent to the Kelch domain of all included NSP sequences appear to share a common origin and are closely related to AtMBP1, AtMBP2, and BoMBP2 isoforms BoMBP2-2 (XP_013609679.1) and BoMBP2-3 (XP_013624973.1). The additional jacalin domains are either closely related to the first ones (same cluster, A) or the result of an early gene duplication in B. oleracea (or the genus Brassica) leading to cluster B with BoMBP2 isoform BoMBP2-1 (XP_013609657.1). Jacalin domains of NSPs with only one or two jacalin domains of both B. oleracea and A. thaliana were present only in cluster A while jacalin domains of NSPs with more than two jacalin domains were present across clusters A and B (Supplementary Figure S5). Bootstrap support within the clusters was too low to draw further conclusions. Adjusting the parameters used for the multiple sequence alignment in MAFFT (Katoh et al., 2019; Kuraku et al., 2013) and for constructing the phylogenetic tree in MEGA (Tamura et al., 2021) did not make the phylogenetic tree more robust.

BoNSP2 (XP_013609641.1) and BoNSP11 (XP_013587057.1) with contrasting abundance patterns in kohlrabi parts (Mbudu et al., 2025) were further characterized. BoNSP2 (XP_013609641.1) was detected in all mature kohlrabi parts analyzed excluding the leaf lamina whereas BoNSP11 (XP_013587057.1) was solely detected in the root (Mbudu et al., 2025). The alignment of the peptides identified in our previous study (Mbudu et al., 2025) with the complete protein sequence revealed a sequence coverage of 48.2% (BoNSP2, 20 peptides) and of 10.4% (BoNSP11, 6 peptides) (Supplementary Figures S3, S4 and Supplementary Table S2). To determine whether the BoNSPs are functional, their NSP activity was investigated in vitro using purified recombinant proteins. The non-enzymatic formation of CNs from GLS hydrolysis in vitro has been reported before (Burow et al., 2009; Kong et al., 2012). As expected, some allyl cyanide (allyl-CN) was formed in assays with all the reaction components except BoNSP2 (Figure 2A) or BoNSP11 (Figure 2C). The proportion of allyl-CN formed upon GLS hydrolysis increased in the presence of BoNSPs as depicted here for assays containing 6.86 µM Fe²⁺ and BoNSP2 (Figure 2B), and 34.3 µM Fe²⁺ and BoNSP11 (Figure 2D), confirming the NSP activity of BoNSP2 and BoNSP11 in vitro.

Next, we tested, at which concentration of added Fe²⁺ the highest proportion of CN is formed from allyl GLS in BoNSP-myrosinase reaction mixtures. Addition of Fe²⁺ to a final concentration of up to 343 µM led to an increase of the proportion of allyl-CN formed from allyl GLS in both the control assays as well as in BoNSP2 (Figure 3A) or BoNSP11 (Figure 3C) containing assays. Net CN formation was obtained by subtracting CN proportion in controls from that in reactions containing BoNSPs. Net CN formation increased from 11.6% in assays with no Fe²⁺ supplementation to 40.5% at 34.3 µM Fe²⁺ in assays with BoNSP2 and from 12.3% to 24.1% in assays with BoNSP11 (Figures 3B, D and Supplementary Table S5). Highest net CN formation was observed in assays containing 34.3 µM Fe²⁺ (Figures 3B, D and Supplementary Table S5). When EDTA at 250 µM or 500 µM was added to the reactions containing 34.3 µM Fe²⁺, net CN formation in reactions with or without BoNSP2 and BoNSP11 was significantly reduced (Figures 3B, D), but not completely abolished. Thus, both the background CN formation by myrosinase and CN formation by BoNSPs did not strictly depend on added iron.

Next, we compared the activity of BoNSP2 and BoNSP11 upon myrosinase-catalyzed hydrolysis of allyl GLS, 4MTB GLS, 4MSOB GLS, benzyl GLS and I3M GLS in vitro. In comparison to the control reactions without NSPs, BoNSP2 and BoNSP11 significantly increased CN formation from all five GLSs (Figures 4A, B; Supplementary Table S6), with increases ranging from 2-fold to 7.1-fold (Figures 4C, D; Supplementary Table S6). The effect of BoNSP2 on CN formation from allyl GLS, 4MTB GLS, 4MSOB GLS and benzyl GLS did not differ significantly, while it was slightly higher for I3M GLS compared to 4MSOB GLS (Figure 4C). The strongest increase in CN formation by BoNSP2 was observed for I3M GLS where the corresponding IACN (indole-3-acetonitrile) increased 7.1-fold (Figure 4C; Supplementary Table S6). Regarding BoNSP11, the increase of CN formation was similar upon hydrolysis of allyl GLS, 4MSOB GLS, benzyl GLS and I3M GLS, but significantly lower upon 4MSOB GLS hydrolysis compared to 4MTB GLS (Figure 4D).

The optimal pH for BoNSP2 activity was assessed in the pH range 5.5 to pH 10.5 using NaAc solution and five biological buffers, MES, MOPS, HEPES, BICINE and Na₂CO₃/NaHCO₃, within their effective pH ranges. BoNSP2 activity assessed as the net CN formation (%) increased with increasing pH value till pH 8 (Figure 5) and activity was optimal at pH 7.5 to pH 8 (HEPES buffer) with net CN formation of 86% for allyl-CN (Supplementary Table S7). While buffer compounds used for the pH range from 5.5 to pH 8 did not seem to affect BoNSP activity, this was not the case for buffer compounds used for higher pH values. In assays with HEPES buffer at pH 8, the net CN formation rate was 86% whereas it was only 14% with BICINE buffer (Figure 5, Supplementary Table S7). Barely any BoNSP2 activity was detected from pH 9 to pH 10.5 with the Na₂CO₃/NaHCO₃ buffer (Figure 5).

Further tests assessed the sensitivity of BoNSP2 and BoNSP11 to changes in pH using NaAc beyond its effective pH range (pH 4 to pH 12 for BoNSP2 and pH 4 to pH 9 for BoNSP11). Similar to the NSP assays with the six buffers with overlapping pH ranges (Figure 5), BoNSP2 activity (assessed as the net CN formation) was highest at pH 8 (83%) (Supplementary Figure S6B; Supplementary Table S8). However, with NaAc, BoNSP2 activity did not drop between pH 8 and pH 9 but stayed at the same level until pH 11 (Figure 5; Supplementary Figure S6B). Similar to BoNSP2, the highest BoNSP11 activity was at pH 7 and above reaching 73% of net CN formation (Supplementary Figure S6D; Supplementary Table S8).