A phylogenetic approach to prioritizing crop wild relatives in Brassiceae (Brassicaceae) for breeding applications

Abstract

Background and Aims

Crop wild relatives (CWRs) are key resources for enhancing agricultural resilience, providing genetic traits that can improve pest resistance, abiotic stress tolerance and nutritional composition in domesticated crops. Within the mustard family (Brassicaceae) this is especially significant in the Brassiceae tribe, which includes economically important genera for agriculture such as Brassica and Sinapis. However, while breeding programmes have historically focused on major crops within this tribe, the potential of their wild relatives, particularly for underutilized and minor crops, remains insufficiently explored.

Methods

We sequenced 175 accessions from six genera, Brassica, Crambe, Diplotaxis, Eruca, Raphanus and Sinapis, using herbarium, seed and living collections. We combined those sequences with 30 nuclear internal transcribed spacer (ITS) sequences from GenBank. Libraries were prepared with Angiosperms353 and Brassicaceae bait kits to build a phylogenetic tree to calculate the phylogenetic distances between tips.

Key Results

We studied the ITS region to build a robust phylogeny for 189 accessions spanning 90 taxa, incorporating newly generated sequences, which included a total of 30 taxa not sequenced before, alongside publicly available sequence data. Phylogenetic distances derived from an ultrametric tree were used to infer cross-compatibility and identify 23 candidate CWRs across the six genera. Our results confirm known relationships based on gene pool classifications of CWRs but also highlight previously overlooked or misclassified taxa that may hold value for crop improvement.

Conclusions

This work demonstrates the efficacy of ITS markers for CWR identification and supports their use as a scalable tool for integrating biological collections into modern breeding and conservation strategies. It provides a comprehensive framework for targeting wild genetic diversity in Brassiceae crops and prioritizing species for future pre-breeding research.

INTRODUCTION

The tribe Brassiceae, within the mustard family (Brassicaceae), is globally important for its agricultural and economic significance and is notable for its complex evolutionary history and polyploidy. The tribe possess around 250 taxa across more than 50 genera (Warwick et al., 2009; Kiefer et al., 2014) and includes major crop genera such as Brassica, Raphanus and Sinapis, and several minor, yet increasingly relevant, cultivated genera like Crambe (used as seakale and crambe oil), Diplotaxis and Eruca (both D. tenuifolia and E. sativa are commonly referred to as rocket salad). These crops are growing in interest not only for their culinary and industrial applications, but also for their agronomic traits, prompting a surge in research exploring the potential of their wild relatives.

Crop wild relatives (CWRs) are naturally occurring species that are genetically related to crops. Their importance lies in their high genetic diversity and adaptive traits, making them a critical resource for crop improvement, especially in the face of climate change, emerging pests and declining agrobiodiversity (Maxted et al., 2012; Brozynska et al., 2016).

Major crops within Brassica have received considerable attention to improve agronomic traits (reviewed for biotic stress by Vasquez-Teuber et al. (2024); and for abiotic stress by Quezada-Martinez et al. (2021) and Kashyap et al. (2022)). For example, resistance to Alternaria and heat stress tolerance was successfully transferred from Sinapis alba species as a CWR to Brassica juncea (Kumari et al., 2018). Similarly, resistance to Sclerotinia, found in Brassica fruticulosa, was introduced to B. juncea (Rana et al., 2017), since several Brassica crops exhibited susceptibility to different Sclerotinia isolates (Taylor et al., 2018). Traka et al. (2013) investigated the wild relative Brassica villosa to enhance the concentration of glucosinolates for regulating sulphate assimilation in broccoli. However, there remains a notable gap in the understanding of CWRs for minor crops in the tribe. While studies in rocket salad species (E. sativa and D. tenuifolia) have explored glucosinolate profiles under different temperatures (Jasper et al., 2020) and variation in taste and composition across cultivars (Pasini et al., 2011), the potential of wild relatives remains underexplored.

The Brassiceae tribe has been widely studied for its polyploidy (Lysak et al., 2007; Wang et al., 2011; Hao et al., 2021; Thomas et al., 2025). Major crops such as Brassica oleracea and B. rapa have been extensively investigated to determine their origins and ancestors (Mabry et al., 2021; McAlvay et al., 2021). Despite these efforts, the timing and mechanisms behind these evolutionary changes remain complex and unresolved. A clearer understanding of their phylogenetic relationships is essential to unravel these patterns. Arias et al. (2014) proposed that the centre of diversity of the tribe differs from its centre of origin, suggesting that the Brassiceae evolved in the Irano-Turanian and Saharo-Sindian regions towards the end of the Oligocene.

Limited information is available about their characteristics and how to transfer desirable traits between CWRs and cultivated species (Ford-Lloyd et al., 2011). Knowing the compatibility between crops and CWRs is a crucial step in transferring key traits in conventional breeding. The most-used classification system for grouping compatible CWRs with their respective crop is the gene pool classification, described by Harlan and de Wet (1971). This states that a species listed as part of the primary (same species as the crop) or secondary gene pool is cross-compatible, based on cross-pollination between the species, while the tertiary gene pool represents more distantly related species that may serve as genetic resources under certain conditions. Within the Brassiceae tribe, the Brassica crops are the most studied, with a long list of classified CWRs in the gene pool system. However, information on cross-compatibility is not available for all crops and CWRs and thus another system was described, the taxon group classification, based on taxonomic hierarchy (Maxted et al., 2006). Although this classification is highly valuable given the general lack of information on interspecific crosses, taxonomy remains under constant revision and change. This means the taxonomic hierarchy is sometimes unable to accurately identify compatible species for cross-pollination. Therefore, a third system was more recently proposed, crop wild phylorelatives (CWPs) (Viruel et al., 2021) to complement the existing approaches by using phylogenetic distances obtained from phylogenetic trees. Shorter phylogenetic distances between species (crop and CWRs) suggests a higher potential for cross-compatibility. Together, these tools can help to identify new CWRs, especially for minor crops where limited information is available (Castillo-Lorenzo et al., 2024). Sequencing CWRs can provide useful information for identifying target genes linked to traits of agronomic interest that could be used in breeding programmes. It can also enhance the value and sustainable use of populations conserved ex situ (Brozynska et al., 2016) while highlighting long-preserved herbarium vouchers as reliable sources of DNA (Staats et al., 2011). Although sequencing techniques have become faster and more cost-effective in recent years, the sequences of 70 % of wild Brassicaceae species have yet to be published or characterized for global accessibility (Warwick et al., 2010; Graper et al., 2021; Castillo-Lorenzo et al., 2024). In the present article we generate sequences of some of these missing species for the tribe Brassiceae, and include cultivated samples for comparison with wild species.

Nuclear markers, such as the internal transcribed spacer (ITS) region of ribosomal DNA, provide an effective means of assessing phylogenetic relationships, and have been widely used in the past (Warwick et al., 2010; Kiefer et al., 2014; Graper et al., 2021). In this study, we analysed newly generated data focused on cultivated species of the Brassiceae tribe from the genera Brassica, Crambe, Diplotaxis, Eruca, Raphanus and Sinapis and their wild relatives, combined with DNA sequences available from the GenBank database (NCBI 2022). We used the ITS region to generate the most complete Brassiceae phylogenetic tree to date, allowing the identification of potential cross-compatible CWRs, based on phylogenetic distances. Our results should be useful in future breeding programmes.

MATERIALS AND METHODS

Species sampling

We sequenced from different collections for six crop genera of the Brassiceae tribe (Brassica, Crambe, Diplotaxis, Eruca, Raphanus and Sinapis) and their wild relatives (from the same genera) held by the Royal Botanic Gardens, Kew (UK), Naturalis Biodiversity Centre, Leiden (The Netherlands) (herbarium samples in both and seed samples and living collections in the former). Local Brassica cultivars were obtained from Misión Biológica de Galicia (MBG-CSIC) in Spain. These newly generated sequences were combined with 30 sequences from GenBank (NCBI, 2022) from the same genera, previously published with the ITS marker (Castillo-Lorenzo et al., 2024). We gathered data from a total of 188 accessions (90 different species) and extracted DNA from 157 accessions (Supplementary Data Table S1). The accepted scientific names were consulted and matched from the World Checklist of Vascular Plants (Govaerts, 2022) and BrassiBase (Kiefer et al., 2014).

DNA extractions

For DNA extraction, we used the CTAB protocol (Doyle, 1990) except for fresh seeds, where the tissue was firstly frozen with liquid nitrogen followed by grinding with a pestle and mortar. The dust was then processed following the CTAB protocol used for the other samples. DNA concentration was measured using a Quantus™ fluorometer (Promega Corporation, Madison, WI, USA) and an Agilent 4200 TapeStation (Agilent Technologies, Santa Clara, CA, USA) was used to assess fragment length.

Library preparations and sequencing

We used 200 ng of DNA to construct genomic libraries following the methods described in Viruel et al. (2019) and Hendriks et al. (2023). They were hybridized prior to sequencing using the Angiosperms353 (Johnson et al., 2019) and Brassicaceae target capture bait kits (Nikolov et al., 2019). The sequencing was done by the BaseClear laboratory (Netherlands) for the samples held in Naturalis and by Macrogen Inc. (The Netherlands) for the sequences held in Kew Gardens and MBG-CSIC.

Raw sequencing data were quality-checked using FastQC software (Andrews, 2010) and MultiQC (Ewels et al., 2016) followed by Trimmomatic (Bolger et al., 2014) for all the samples to remove the adapters of Illumina and the samples with low quality (using Trailing:30 to exclude samples with a Phred value <30). Paired reads were analysed with HybPiper (Johnson et al., 2016). We took advantage of the off-target reads from target capture to recover the ITS marker (ITS1 and ITS2) using as a reference Brassica carinata (GenBank: KX709380).

We aligned the sequences (newly generated and the ones from the GenBank using MAFFT v.7.505; Katoh and Standley, 2013). The alignment was cleaned with trimAl (Capella-Gutiérrez et al., 2009) applying the parameters -resoverlap 0.75 -seqoverlap 70, and the result was then manually analysed with MEGA X (Kumar et al., 2018) to remove sequences with large gaps.

The cleaned and curated alignment was concatenated with AMAS (Borowiec, 2016) and a phylogenetic tree was constructed in IQ-TREE v.2.0.6 (Minh et al., 2020) using the maximum likelihood approach. We applied the GTR + I + G model, identified among the best substitution models by ModelFinder Plus. We estimated two support values, UFBoot and SH-aLRT support (both parameters set at 1000). The tree was rooted to the outgroup species Sisymbrium orientale (tribe Sisymbrieae; Brassicaceae).

We used the statistics obtained from HybPiper and AMAS to select the best sample. In cases when there was more than one sample per species, we selected the longest sequence with fewer missing data. The only exceptions were cases in which both cultivated and wild samples belonged to the same species. In such cases, both samples were retained to better understand crop–wild relationships in the phylogeny. Pairwise phylogenetic distances were calculated with the ‘patristic’ function (from the adephylo R package; Jombart and Dray, 2010) on the phylogenetic tree containing only the best sequence (longest and fewer gaps) per species (and crops). These calculations were made on an ultrametric tree generated with treePL (Smith and O’Meara, 2012) by setting the longest distance between terminals to 0.5.

We used the information available on cross-compatibility between crops and CWRs (from the gene pool classification (FitzJohn et al., 2007; Warwick et al., 2009)) and Vincent et al. (2013) (summarized in Supplementary Data Table S2) to estimate the compatibility thresholds for each crop (i.e. phylogenetic distance (PD)), following the CWP system (Viruel et al., 2021). In general, the most distantly related (but known) cross-compatible CWR was used as the maximum PD for potential compatibility, and the lowest PD for a known incompatible CWR (unsuccessful crosses) was also considered to estimate the compatibility thresholds. However, there were some exceptions where there were contradictory data, or when such information was unavailable, in which case the wild species were ranked by increasing phylogenetic distance from the crop.

RESULTS

Sampling and sequencing results

All samples contained enough good-quality DNA (>200 ng) for target capture sequencing, except for 20 samples that failed DNA extraction and had high amounts of missing data. Despite the potential for DNA degradation in herbarium samples, DNA concentrations were not significantly different between fresh and herbarium material (Supplementary Data Fig. S1). Similarly, the ITS marker was fully recovered in >75 % of all samples and >95 % of the length was recovered in >90 % of the samples independently of the sample type (herbarium, seed or silica-dried fresh material; Supplementary Data Fig. S2).

Updated Brassiceae phylogeny

The phylogenetic tree contained 204 tips (Supplementary Data Fig. S3), after including the alternative sequences of Mutarda carinata (also known as Brassica carinata), B. juncea and B. napus (represented with ID_NODE1). The phylogeny used to estimate the PDs had a total of 99 tips (Fig. 1) with one sample per species, including the cultivated samples. The PD in the ultrametric tree ranged between 0 and 1.

Fig. 1.

Maximum likelihood phylogenetic tree of the Brassiceae tribe for the ITS marker, showing the distribution of Brassica genus tips in green and Raphanus in cyan (sections A–D), Eruca in ochre (section E), Crambe in blue (sections F, G), Diplotaxis in purple and Sinapis in red. The numbers in the nodes are the UFBoot and SH-aLRT percentage values. For detailed phylogeny see Supplementary Data Fig. S3.

The genus Brassica clustered most of the species in one clade, which was divided into four subclades (Fig. 1). Three of the main crop taxa, B. rapa, B. oleracea and B. napus, were recovered in a single subclade, together with some closely related CWRs (Fig. 1, section A). One of the samples in this subclade, B. oleracea sampled from the Canary Islands, was very distant in the phylogenetic tree from the other B. oleracea accessions. The second clade consisted of other wild relatives of the Brassica and Raphanus genera (Fig. 1B) and the third clade had a mix of wild Brassica species and three wild Diplotaxis species (Fig. 1C). Finally, the other small clade was formed by the new Mutarda species (Fig. 1D; bootstrap value of 77), which included M. nigra and M. arvensis (also known as Brassica nigra and Sinapis arvensis, respectively), and separated in another group we had M. carinata and a wild accession of M. nigra (bootstrap value of 93).

In a different clade we recovered other Brassica taxa sisters to Eruca and Diplotaxis (Fig. 1E). Two Brassica species were not recovered in the main clades, B. repanda subsp. glabrescens and B. rupestris. In both cases, we only had one sequence per species available.

The genus Crambe was almost completely sampled (34 out of 37 accepted species), and most of the accessions clustered in one clade, which in turn was divided into two main clades (Fig. 1F, G, with 96 % bootstrap). One clade could be divided in two, one subclade included the edible C. maritima and the other the industrial oilseed C. hispanica subsp. abyssinica. The clade represented in clade G (Fig. 1) contains several wild Crambe species. Some exceptions were C. hispanica and C. gordjagini, which were not recovered within the Crambe genus.

Representatives of the genus Diplotaxis were distributed across the phylogeny but clearly divided into two clades. One large clade was formed by several wild species of Diplotaxis and the edible species of rocket salad, D. tenuifolia and Eruca sativa (Fig. 1E). The second clade was smaller and included D. erucoides, D. villosa and D. tenuisiliqua (Fig. 1C). Finally, D. siifolia was sister to species of the genus Mutarda and was closer to the smaller group of D. erucoides.

We included four species of Eruca (E. sativa, E. vesicaria, E. foleyi and E. pinnatifida), where all the samples grouped together in the phylogeny, supported by 92 % bootstrap.

Finally, the small genus of the white mustard, Sinapis (with only three accepted species), was represented in the tree. Only S. alba and S. flexuosa grouped together and distantly from other Brassiceae species. Sinapis pubescens fell in the clade of wild Brassica species (section B of Fig. 1), sister to B. procumbens and B. cadmea (bootstrap 97 and 86 % respectively).

New Brassiceae CWRs

We identified 23 new CWRs based on phylogenetic distances and the cross-compatibility information available (Supplementary Data Table S3), which was available for cultivated species, except for seakale (Crambe maritima). Most of the new predicted CWRs correspond to minor crops of the genera Crambe, Diplotaxis and Eruca, of which 11 had not been sequenced before.

In general, the cultivated Brassica shared a similar PD threshold, ranging from 0.000 to 0.2265 (Table 1), indicating that species within this range may be genetically compatible with the corresponding crops and suitable for breeding. For the Ethiopian mustard, rapeseed and black mustard, the range for cross-compatibility was up to 0.5348, since within the next PD range (>0.5348) there were some incompatible species mixed with cross-compatible species (Supplementary Data Table S3). In the case of the minor crops, crambe oil (C. hispanica subsp. abyssinica) was the only one with three wild species classified as secondary and tertiary gene pool (but there was no information about successful conventional crosses; Supplementary Data Table S3), for which PD ranged between 0.0000 and 0.1339. A similar threshold was use for seakale, C. maritima, since there was no information about this minor crop (up to 0.1738 PD). The thresholds for potentially compatible CWRs for rocket salad ranged from 0.0000 to 0.2742. Radish (Raphanus raphanistrum subsp. sativus) had limited information, with only one CWR reporting low-success conventional crosses, Brassica oxyrrhina (Table 1, PD value 0.2271). Unfortunately, for white mustard, Sinapis alba, there was only one species, S. flexuosa, already classified as compatible, but without information about successful crosses (Supplementary Data Table S3).

Table 1.

Identification of CWRs based on PD from an ultrametric tree for the Brassiceae tribe using the ITS nuclear marker. The secondary CWRs are obtained from the gene pool classification (Harlan and de Wet, 1971; USDA, 2025) and from the conventional crosses described in FitzJohn et al. (2007) and Warwick et al. (2009) Part III (more details are given in Supplementary Data Table S3).

Common name	Scientific name	Potential CWRs (ITS)	Phylogenetic distance
Ethiopian mustard	Brassica carinata/Mutarda carinata	Brassica nigra (Y)	0.1791
		Diplotaxis griffithii	0.5348
		Diplotaxis acris
		Eruca sativa
Rapeseed	Brassica napus	Brassica rapa (Y)	0.169
		Brassica cretica (low success)
		Brassica oleracea (Y)	0.2265
		Brassica incana
		Brassica tineoi *
		Brassica bourgeaui
		Brassica montana
		Brassica villosa *
		Brassica insularis
		Brassica macrocarpa *
		Brassica hilarionis
		Diplotaxis griffithii *	0.4276
		Diplotaxis acris
		Eruca sativa
		Eruca vesicaria
		Eruca foleyi *
		Eruca pinnatifida *
		Brassica repanda *
		Eruca setulosa
Black mustard	Brassica nigra/Mutarda nigra	Brassica carinata (low success)	0.1791
		Diplotaxis griffithii *	0.5348
		Diplotaxis acris *
		Eruca sativa *
		Eruca vesicaria *
		Eruca foleyi *
		Eruca pinnatifida *
		Brassica repanda *
		Eruca setulosa *
		Brassica rapa (low success)
Cabbage	Brassica oleracea	Brassica bourgeaui (Y)	0.0891
		Brassica montana (Y)	0.1236
		Brassica incana (Y)
		Brassica tineoi *
		Brassica villosa (Y)	0.1371
		Brassica insularis (Y)
		Brassica macrocarpa (low success)
		Brassica hilarionis (NI)
		Brassica rapa (Y)	0.2265
		Brassica cretica (Y)
		Brassica napus (low success)
Turnip	Brassica rapa	Brassica cretica (N)	0.1475
		Brassica napus (Y)	0.169
		Brassica oleracea (Y)
		Brassica incana
		Brassica tineoi *
Crambe or Abyssinian mustard	Crambe hispanica subsp. abyssinica	Crambe filiformis	0.0685
		Crambe hispanica subsp. glabrata	0.1339
		Crambe kralikii
Seakale	Crambe maritima	Crambe steveniana *	0.1238
		Crambe tataria var. aspera*	0.1456
		Crambe pinnatifida *
		Crambe tataria *
		Crambe orientalis *	0.1738
		Crambe cordifolia *
		Crambe juncea *
		Crambe grossheimii *
		Crambe koktebelica *
Perennial wall rocket	Diplotaxis tenuifolia	Diplotaxis simplex (Y)	0.0376
		Diplotaxis griffithii *	0.2742
		Diplotaxis acris *
		Eruca sativa (NI)
		Eruca vesicaria (NI)
		Eruca foleyi *
		Eruca pinnatifida *
		Brassica repanda *
		Brassica setulosa *
		Brassica desnottesii *
		Brassica baldensis *
Rocket salad	Eruca sativa/Eruca vesicaria subsp. sativa	Eruca vesicaria	0.0542
		Eruca foleyi *	0.0854
		Eruca pinnatifida (NI)	0.1328
		Diplotaxis griffithii *	0.1949
		Diplotaxis acris *
		Brassica repanda (NI)	0.2534
		Brassica desnottesii *
		Brassica baldensis *
		Brassica setulosa *
		Diplotaxis simplex (NI)	0.2742
		Diplotaxis tenuifolia (Y)
Radish	Raphanus raphanistrum subsp. sativus	Raphanus raphanistrum	0.0422
		Brassica barrelieri *	0.2271
		Brassica oxyrrhina (low success)
		Brassica fruticulosa
White mustard	Sinapis alba	Sinapis flexuosa (NI)	0.3518

Y, successful conventional crosses documented; N, unsuccessful; NI, no information found despite being classified as CWRs.

^*New potential CWRs compatible with their respective crops identified in this study.

In general, the majority of secondary CWRs were close in the phylogeny (Table 1), confirming their compatibility with existing crops, with some exceptions where tertiary CWRs or wild taxa with low or no successful crosses were also close in the phylogeny.

DISCUSSION

The approach described in this study is particularly valuable for minor crops and edible species with limited information on cross-compatible CWRs. By exploring phylogenetic analyses, especially for underutilized taxa, we can better prioritize wild species with potential for crop improvement. Although PD is a complex and sometimes inconsistent proxy, especially for well-studied crops, our findings reaffirm known relationships and highlight overlooked CWRs in species such as C. maritima, E. sativa and D. tenuifolia. Expanding taxon sampling within genera to improve the resolution of the phylogeny in the Brassiceae tribe will enhance the accuracy of the PD approach, especially where PD values overlap.

Integrating ITS-based phylogenetics to support CWR identification

Our study reinforces the versatility of target capture approaches, which are not only provided with hundreds of nuclear genes (not used in this study) but also allow the recovery of molecular markers from the off-target molecular markers compatible with traditional phylogenetic studies. We demonstrate the compatibility of high-throughput sequencing with GenBank data to produce the largest sampling of Brassiceae in a phylogenetic study. Using both new and publicly available ITS sequences, we constructed a robust phylogeny encompassing 185 accessions from 88 taxa, including 31 previously unsequenced species. This allowed us to estimate phylogenetic distances and identify candidate CWRs with potential for crop improvement across six genera: Brassica, Crambe, Diplotaxis, Eruca, Raphanus and Sinapis. Our results confirm known compatibility relationships and highlight overlooked CWRs that are phylogenetically close to crops, even in cases where empirical crossing data are lacking.

Despite the increasing use of genomic data in plant systematics, the ITS region remains a valuable marker due to its high variability, broad representation in public databases, and ability to be recovered from degraded DNA such as herbarium material. Previous phylogenies of the Brassiceae tribe have been investigated using plastid markers or partial taxon sampling (Warwick and Sauder, 2005; Arias and Pires, 2012). All the markers used in the high-throughput sequencing method for this study are nuclear. Unfortunately, we were unable to recover the plastid marker matK for most of the sequences (<40 %, Supplementary Data Table S1), preventing comparison with the nuclear marker ITS (matK phylogenetic tree for 72 sequences in Supplementary Data Fig. S4) as published in previous publications (Castillo-Lorenzo et al., 2024; Dominicus et al., 2025). However, in the present study we have generated new sequences showing a more comprehensive ITS phylogenetic tree, expanding species coverage for this tribe.

Our approach is particularly valuable for underutilized crops, where crossing experiments or genomic resources may be limited or absent. Phylogenetic distances derived from the ITS-based ultrametric tree provided a complementary tool to existing CWR classification systems based on gene pool (Harlan and de Wet, 1971) or taxonomic groupings (Maxted et al., 2006). Our findings support the use of PD as a practical proxy for cross-compatibility, especially where empirical data are lacking or contradictory. This is especially relevant for breeding programmes targeting minor crops such as seakale (C. maritima) and rocket salad (E. sativa, D. tenuifolia), which are gaining recognition for their nutritional and ecological value but remain genetically underexplored.

In this study we focused on identifying cross-compatible CWRs (through conventional crosses) to prioritize their use without relying on advanced biotechnological techniques (e.g. embryo rescue, ovary rescue, somatic hybridization or double haploid production). This approach aims to facilitate the use of these materials by a wider range of stakeholders. However, it is important to note that the use of biotechnological techniques could further expand the potential for introgression or increase the success of crosses between the species (Inomata, 1993; Gupta and Pratap, 2010; Fu et al., 2018; Treccarichi et al., 2022). Most of the wild species examined in this study are likely to be cross-compatible with cultivated species at varying levels when such techniques are employed. This will offer additional opportunities for trait introgression and genetic improvement, especially for overcoming incompatibility barriers.

Phylogenetic resolution and CWR identification in the Brassica and Raphanus genera

The genus Brassica remains a focal point of research due to its global agricultural importance and complex evolutionary history. Numerous studies have characterized both cultivated and wild Brassica taxa for morphological, genetic and biochemical traits (Branca and Cartea, 2011; Warwick, 2011). Several compatible CWRs have been identified and used to improve crops through hybridization, although interspecific crosses are often challenged by reproductive barriers (FitzJohn et al., 2007; Kaneko et al., 2009; Warwick et al., 2009). In our study, Brassica species were grouped into a large clade, separating cultivated and close relatives of other edible (Raphanus and Mutarda) and wild species. The large gap between the B. oleracea accession from the Canary Islands and the other B. oleracea accessions suggests possible ongoing introgression or hybridization. Further genomic analyses incorporating more genes are required to understand the cause of this phylogenetic displacement.

Phylogenetic distances calculated from the ITS ultrametric tree confirm known relationships within the Triangle of U model. For instance, B. juncea, B. napus and M. carinata, all amphiploids derived from Brassica diploid progenitors, cluster closely with their parental lineages and compatible CWRs. Specifically, two alternative ITS sequences for B. juncea were found in our analysis, and both were clustered near the B genome (B. nigra; Supplementary Data Fig. S3). The cultivar Brassica napus grouped closer to B. rapa than to B. oleracea, suggesting stronger ITS affinity to the A genome, consistent with previous findings using nuclear markers (Allender and King, 2010). This supports B. rapa (A genome) as the likely maternal donor for most of the local cultivars of B. napus (23Q01 and 23Q02). In contrast, another local cultivar of B. napus (23Q18) showed most TS sequences clustering near the Oleracea clade with only one sequence aligning with Rapa (NODE 8). This suggests a greater representation of the C genome for this cultivar.

Interestingly, the phylogenetic tree supports recent taxonomic revisions that reclassify B. carinata, B. nigra and S. arvensis as species of the genus Mutarda, all of which formed a coherent clade (German, 2022; Al-Shehbaz, 2025; Supplementary Data Fig. S3). This consistency with Viruel et al. (2021) and Castillo-Lorenzo et al. (2024) reinforces the need to align phylogenetic results with ongoing taxonomic frameworks to improve clarity in breeding applications. Our PD analysis largely supports known gene pool classifications and hybridization outcomes. For instance, the presence of Raphanus samples within the Brassica clade reflects their ability to cross and produce hybrids as documented previously (Warwick et al., 2009). Importantly, our data also reveal several wild Brassica species with unexpectedly close PDs to key crops despite limited or unsuccessful crossing records, such as B. cretica, which appears as a sister in the phylogeny to the cultivated Brassica crops B. rapa and B. napus. Indeed, there have been several successful attempts at cross-pollination between some wild Brassica species and major crops (FitzJohn et al., 2007) and transfer of resilience and tolerance to pests and diseases (Katche et al., 2019).

Local cultivars sequenced in this study also revealed strong clustering: one group was composed mainly of B. rapa landraces with their wild relatives (including one sample of B. cretica), while the second included B. oleracea landraces and a larger list of wild relatives (Fig. 1A). This offers insight into local landraces retaining close genetic relationships with wild taxa, emphasizing their conservation value for crop improvement and resilience. However, the use of landraces in breeding, as well as efforts to develop them into new commercial cultivars, remains limited despite the fact that some have been characterized and proved to carry valuable agronomic traits (Monteiro and Dias, 1994; Ciancaleoni et al., 2014; Rodríguez et al., 2014, 2015; Tortosa et al., 2017).

Phylogenetic resolution and CWR identification in the Crambe genus

The genus Crambe is well recognized as phylogenetically distinct within the Brassiceae tribe, often forming a separate clade in most published phylogenetic studies (Warwick and Sauder, 2005; Kiefer et al., 2014). Previous classifications have split Crambe into three sections (Dendrocrambe, Leptocrambe and Sarcocrambe) based on morphological traits and geographical distribution (Rudloff and Wang, 2011). Nearly half of its species are endemic to the Macaronesian archipelagos, especially the Canary Islands, a known hotspot for Brassicaceae diversity and endemism (Francisco-Ortega et al., 2002; Hooft van Huysduynen et al., 2021; Zizka et al., 2022).

We sequenced nearly the entire genus (34 out of 37 accepted species, Kiefer et al, 2014), providing one of the most comprehensive ITS-based phylogenetic trees for Crambe to date. Most accessions formed a coherent group subdivided into three sub-clades, generally congruent with the historically defined taxonomic sections. These results align with patterns observed in earlier molecular work focused on Macaronesian taxa, which revealed three main lineages (Francisco-Ortega et al., 2002). Some exceptions in our phylogenetic tree merit further investigation. For instance, C. hispanica, sampled from Jordan, clustered ambiguously in relation to other subspecies (abyssinica and glabrata), potentially indicating misidentification or unresolved taxonomy. Similarly, C. fruticosa, endemic to Madeira, was placed outside the main Crambe group, potentially reflecting island-driven genomic divergence, which is consistent with prior findings of distinct evolutionary trajectories among Macaronesian species (Francisco-Ortega et al., 2002). These cases highlight the need for deeper genomic resolution to clarify their evolutionary relationships.

Within Crambe, C. maritima (seakale) and C. hispanica subsp. abyssinica represent contrasting but complementary domestication trajectories, one for edible shoots in coastal Europe, the other for industrial seed oil. Though once widely cultivated in France and the UK (Briard et al., 2002), C. maritima remains a neglected crop despite its promising traits, including salinity tolerance up to 100 mm NaCl (De Vos et al., 2010). The lack of knowledge relating to its CWRs shows the unexplored potential for this genus. In contrast, C. hispanica subsp. abyssinica has gained attention for its erucic acid content, used in bio-lubricants and industrial oils (Qi et al., 2018). The known CWRs for crambe oil, C. hispanica subsp. hispanica and C. filiformis, were classified based on their proximity in the phylogeny (USDA, https://npgsweb.ars-grin.gov/gringlobal/taxon/taxonomysearchcwr, accessed 10 February 2025), which corresponds to our findings. Our study reveals that both of these crops cluster closely with other wild Crambe species, many of which are also edible or exhibit useful agronomic traits. Notably, species such as C. tatarica, C. orientalis and C. cordifolia, known for their edible roots (Rudloff and Wang, 2011), could serve as novel CWRs for C. maritima or C. hispanica subsp. abyssinica breeding programmes.

Despite this diversity, Crambe remains underutilized and under-characterized genetically, especially regarding its response to biotic and abiotic stresses. Our data fill a critical gap by identifying phylogenetically close CWRs to both major and minor cultivated Crambe taxa. This sets a strong foundation for further investigation and pre-breeding work, particularly for minor crops with potential for saline or marginal lands.

Phylogenetic resolution and CWR identification in the Diplotaxis and Eruca genera

The Diplotaxis and Eruca genera represent relatively recent additions to mainstream cultivation and breeding efforts. Consequently, they have received comparatively limited attention in terms of genetic characterization and crop improvement programmes. Despite their growing nutritional and commercial relevance, few studies have explored their full potential, particularly in relation to their CWRs.

Most of the studies included only one or two Eruca species in the phylogeny of the tribe (Warwick and Sauder, 2005; Arias and Pires, 2012) or multiple accessions from the cultivar E. sativa and E. vesicaria were mapped (Warwick et al., 2007; Zhu et al., 2021). In our phylogenetic results, Eruca species formed a strongly supported (bootstrap value 92) monophyletic group. This pattern corroborates earlier studies (Perfectti et al., 2017) where Eruca was placed separately from other species of the Brassiceae tribe. In contrast, Diplotaxis species were distributed across multiple clades, consistent with previous findings (Warwick and Sauder, 2005) and reflective of the complexity of the evolutionary patterns in the genus.

Diplotaxis species are mostly diploid with varied chromosome numbers (disploidy); thus, species differ in chromosome numbers, and some of them have their own cytodeme (Pignone and Martínez-Laborde, 2011). The genus exhibits considerable morphological and ecological diversity, with many species adapted to Mediterranean environments. However, there are limited genomic or phenotypic data available across the genus. In our analysis, the distribution of Diplotaxis species alongside Eruca is in agreement with the one published by Warwick and Sauder (2005). We observed D. siifolia unexpectedly grouping closer to the Mutarda clade, suggesting possible interspecific and intergeneric introgression. However, more phylogenetic studies are needed to clarify and resolve these unexpected relationships.

Eruca sativa, also known as E. vesicaria subsp. sativa, is the most commonly cultivated rocket species, often described as further domesticated than D. tenuifolia. However, conflicting evidence exists, with Bell and Wagstaff (2014) questioning this assumption based on the use of D. tenuifolia as a commercial product and the lack of supporting evidence. Eruca sativa is highly adaptable, including tolerance to heat, drought and salinity, and has demonstrated resistance to certain pests and diseases (Warwick et al., 2007). Nonetheless, the yield performance can decline under heat stress in both rockets, E. sativa and D. tenuifolia (Jasper et al., 2020). These findings underline the importance of identifying CWRs with traits for greater environmental resilience. Our phylogenetic distance analysis provides a framework for identifying candidate CWRs closely related to cultivated rocket salad species. For example, D. simplex, D. acris, D. griffithii and E. foleyi were found to be the closest relatives to D. tenuifolia and E. sativa, respectively, based on phylogenetic distance (PD = 0.2741 and 0.1949, respectively, Table 1), confirming their potential for genetic introgression.

Despite recent advances (e.g. Bell and Wagstaff, 2019), comprehensive evaluations of the full potential of wild Eruca and Diplotaxis species under diverse agronomic conditions are still rare. Their evaluation under stress conditions and the integration of these findings with phylogenetic data, as shown in our study, will be essential in developing improved cultivars suited to future climate and food security challenges.

Phylogenetic resolution and CWR identification in the Sinapis genus

The Sinapis genus, though small, holds significant agronomic value due to its role in mustard (S. alba) production and its close relationship with several cultivated Brassica species. Previous studies using ITS sequence data (Agerbirk et al., 2008) resolved the three currently accepted Sinapis species (S. alba, S. flexuosa and S. pubescens) into separate clades. Our phylogeny broadly corroborates these earlier findings and adds valuable data by integrating both wild and cultivated accessions from herbaria and seed bank collections.

In our analysis, S. alba and S. flexuosa formed a strongly supported clade (bootstrap value of 99.7), consistent with their close taxonomic relationship. However, S. pubescens was recovered close to certain wild Brassica species, including B. procumbens and B. cadmea (Fig. 1B), which was previously observed in phylogenetic studies (Warwick and Sauder, 2005; Arias and Pires, 2012). This suggests that S. pubescens may occupy a more intermediate phylogenetic position than previously assumed, potentially informing its use in intergeneric breeding. One accession of S. alba in our dataset was recovered with S. arvensis (Supplementary Data Fig. S3), now reclassified as Mutarda arvensis, which may indicate sample misidentification, underscoring the importance of rigorous taxonomic verification in phylogenetic studies and using GenBank accessions.

Sinapis species possess valuable biotic stress resistance; both S. alba and S. arvensis demonstrated resistance to flea beetles (Phyllotreta spp.) and blackleg disease caused by Leptosphaeria maculans (summarized in Winter, 2010). Particularly, S. alba has contributed to breeding programmes for its resistance to Alternaria black spot and tolerance to heat (Kumari et al., 2018).

In the context of CWR identification, S. alba is classified as a tertiary gene pool for several Brassica crops, with compatibility often requiring non-conventional crossing techniques such as tissue culture (Quezada-Martinez et al., 2021). Our phylogenetic distance analysis supports this: S. alba and Mutarda carinata are distant in our tree (PD 0.5348). However, M. arvensis was recovered next to M. nigra and M. carinata, of which M. arvensis is classified as secondary gene pool for the former (CWR inventory 2010; Vincent et al., 2013), confirming their relevance as potential gene donors. Meanwhile, S. pubescens, despite being more distantly related to the cultivated Sinapis, still offers useful traits for other cultivated species within the tribe. Overall, this genus exemplifies the potential of small but agriculturally relevant groups to contribute to crop improvement, particularly when paired with phylogenetic support.

Advancing CWR identification through phylogenetic tools

Our work underscores the utility of ITS as an informative marker in phylogenetic CWR assessment and demonstrates the feasibility of using herbarium material as sources of high-quality genetic data. These legacy collections offer critical insights for conservation genomics and breeding strategies, bridging historical biodiversity with modern applications.

These results demonstrate that while hybridization research and compatibility (especially focused on Brassica crops; Mabry et al., 2021; McAlvay et al., 2021; Cai et al., 2022; Saban et al., 2023; Wang et al., 2023; Maggioni et al., 2024) is advancing, our understanding of reproductive compatibility between different species within the tribe remains incomplete. For example, for Crambe genus most of the studies have been carried out only in the industrial oil crop C. abyssinica (Rudloff and Wang, 2011). Interspecific crosses have been tried between C. abyssinica, C. hispanica and C. kalikii (Du et al., 2014). Similarly, for rocket salad most of the breeding efforts are within cultivars of the species but not much with wild relatives (Padulosi, 1995). There are many potential species combinations yet to be tested for compatibility. Expanding our knowledge in this area is essential to unlock the full potential of CWRs and support the development of new crops.

Moreover, this study highlights the value of integrating historical collections, both herbaria and seed accessions, into modern molecular analyses. The successful recovery of full-length ITS sequence from preserved material validates not only the feasibility of using herbarium specimens for DNA extraction but also reinforces the utility of ITS as an informative marker for phylogenetic assessment of CWRs. These resources are essential components for bridging the historical biodiversity data with present-day conservation and breeding efforts. As sequencing costs continue to drop and techniques for working with degraded DNA improve, this integrative strategy will become increasingly important for leveraging existing ex situ collections for genetic resource development.

SUPPLEMENTARY MATERIAL

Supplementary data are available at Annals of Botany online and consist of the following. Table S1: details of the samples and species used in the article with the recovery percentage of ITS and matK markers, DNA concentration, concentration after library preparation and new sequences published with the GenBank ID are in Table S1.1. Table S2: list of crops and edible species with the corresponding wild species present in the phylogeny. New predicted CWRs based on phylogenetic distance (PD) are identified with the PD thresholds for each crop. Figure S1: DNA concentration of the different extracted plant materials from fresh silica-dried leaves, herbarium samples and fresh seeds. Figure S2: percentage of ITS recovery for the sequences from different types of material we used to extract DNA. Figure S3: PDF file of a maximum likelihood phylogenetic tree of the Brassiceae tribe for the ITS marker, showing the distribution of Brassica genus in green, Raphanus in cyan, Crambe in blue, Diplotaxis in purple, Eruca in ochre and Sinapis in red. The numbers in the nodes are the UFBoot and SH-aLRT percentage values. These are 158 newly generated ITS sequences and 30 sequences extracted from GenBank (204 tips, excluding the outgroup). Figure S4: PDF file of a maximum likelihood phylogenetic tree of the Brassiceae tribe for the matK marker showing the distribution of Brassica genus in green, Crambe in blue, Diplotaxis in purple, Eruca in ochre and Sinapis in red. The numbers in the nodes are the UFBoot and SH-aLRT percentage values. The tree is formed of 72 tips excluding the outgroup.

FUNDING

This project was internally funded by the Royal Botanic Gardens, Kew.