The Brassicaceae then and now: advancements in the past three decades, a review

Abstract

Background and Aims

The Brassicaceae and many species of its various genera are currently recognized as models in various fields of biology. The impact of the major advancements in molecular phylogeny, development, genomics and related fields on the systematics of the family led to the recognition of the first phylogenetic tribal classification of the family and recognition of monophyletic genera. The present review is aimed to cover almost all advancements in the family systematics for the first time.

Methods

The comprehensive literature on the molecular phylogeny, development, cytology, genomics and related fields was assembled relative to its systematics value to Brassicaceae, especially for tribal classification, generic delimitation, origin, hybridization and migration.

Key Results

Adjustments to the limits of problematic tribes Brassiceae, Coluteocarpeae and Thelypodieae, in addition to various genera (e.g. Atacama, Delpinophyton, Menonvillea, Noccaea and Pseudoarabidopsis) are discussed and likely solutions suggested. A complete list of apomorphic characters in the family is assembled for the first time, and all species or genera in which they occur are listed.

Conclusions

This comprehensive review is the first historical advancement of the Brassicaceae systematics during the past three decades, and it aims to help current and future students of the family to understand the background of such developments.

INTRODUCTION

The Brassicaceae (or Cruciferae) are a large family represented by native species on all continents except Antarctica. This family is economically important for numerous vegetable crops, ornamentals and a source of vegetable oils and condiments, in addition to including ~130 species of widespread or noxious cosmopolitan weeds. It also includes Arabidopsis thaliana, the model organism in modern biology whose genome was sequenced at the turn of the present millennium.

Plants of the family grow in a diverse range of habitats, of which some are rather extreme, and they range from fully submersed inland aquatics, as in the American Nasturtium floridanum and Rorippa aquatica and sometimes the Eurasian Rorippa amphibia, or lakeshores Subularia aquatica, in addition to those of intertidal zones, such as the American Cochlearia sessilifolia and Cardamine longii. At the highest elevation (6400 m) reached by any vascular plant on Mt. Everest are the Brassicaceae Lepidostemon everestianus and the Asteraceae Saussurea gnaphalodes (Dentant, 2018), followed by a few species at 6300 m, including the mustard Solms-laubachia himalaica. Most plants of the highest elevations or those of high southern latitudes (~50°S), such as Pringlea antiscorbutica, flower when they are basically covered with snow. With the exclusion of widespread or cosmopolitan weeds of the family, the vast majority of species have rather restricted distribution. The exceptions are Anastatica hierochuntica (desert annual of western Mauritania eastwards into North Africa, Arabian Peninsula, into Pakistan; a range of ~8000 km), and its plants tolerate the highest desert heat and drought. Several species of Brassicaceae are known to tolerate and accumulate high levels of heavy metals (e.g. cadmium, nickel, zinc) or salt tolerance, variously addressed by Koch and German (2013) in the genera Eutrema, Arabidopsis, Noccaea and Schrenkiella, in Noccaea caerulescens by Dinh et al. (2018), selenium and sulphur accumulation in Stanleya (Cappa et al., 2014), and nickel accumulation in Odontarrhena (Cecchi et al., 2020) and Streptanthus (Boyd et al., 2009).

The family includes many shrubby plants, lianas to 9 m tall (Polypsecadium adscendens) and subtrees or small trees (Farsetia somalensis, South African Heliophila brachycarpa and H. florulenta, Hawaiian Lepidum serra, and Canarian Parolinia spp.).

The Brassicaceae include some 4158 species in 357 genera (estimated in BrassiBase: https://brassibase.cos.uni-heidelberg.de/; D. A. German, pers. comm.), more than half of which (2275, or ~55 %) belong to the largest 14 genera, including Draba (~410 spp.), Cardamine (284), Erysimum (274), Lepidium (272), Alyssum (114), Physaria (108), Heliophila (106), Boechera (~100), Arabis (~100), Odontarrhena (91), Rorippa (90), Isatis (85), Noccaea (~85) and Aethionema (~72). The family also includes 132 each of monotypic and oligospecific (two to five species) genera, together totalling 546 species. The remaining 1332 species belong to ≥78 genera.

Since its first recognition as Cruciferae by de Jussieu in 1789, the Brassicaceae has been accepted as a well-defined family for 235 years. Its delimitations was briefly expanded by Judd et al. (1994) to include the Capparaceae and Cleomaceae based on a cladistic study of few genera and limited morphological characters. The phylogenetic study presented by Hall et al. (2002) strongly supported the recognition of three families and the sister relationship of the Brassicaceae to the Cleomaceae.

Subsequent advancements on the relationship of all families of Brassicales have been updated continuously on the Angiosperm Phylogeny website (http://www.mobot.org/MOBOT/researh//APweb) to the present. The interested reader on the order and its 17 families should consult the above website, in addition to the most recent findings of Edger et al. (2018), Mabry et al. (2023), Zuntini et al. (2024), and references therein.

The modern advancement in biology started when an unassuming, non-aggressive European weedy mustard (Arabidopsis thaliana) took the stage to become the model organism of experimental biology a few decades before its entire genome was sequenced (Arabidopsis Genome Initiative, 2000). Since then, numerous species of many genera have been sequenced, and the Brassicaceae became the forerunner model family among all plants.

The beginning of molecular systematics of the family was initiated when a handful of European and American geneticists and population biologists, in addition to their students and postdocs, focused from the early to middle 1990s on the study of variation of a single chloroplast gene (e.g. rbcL, ndhF, trnLF region), restriction-site variation of the entire plastid DNA and (or in combination) nuclear encoded Internal Transcribed Spacer (ITS) in several species of a given genus or several genera and assessed the phylogenetic implications of their data on the systematic relationships and character evolution. Several laboratories in Canada, Germany and the USA laid the foundation, on which advance every aspect of the family systematics from the 1970s through well after the sequencing of A. thaliana genome. The fields influenced include genetics, development, reproductive biology, cytology, phylogenetics, genomics and systematics.

This review focuses primarily on the tremendous advancements in the systematic and phylogenetic relationships of the Brassicaceae, starting initially from early 1980s to the present.

Pioneering studies

The founding fathers of modern Brassicaceae research were plant geneticists and population biologists, and the first was the Austrian-born Hungarian geneticist George Rédei (14 June 1921–10 November 2008). Rédei moved to the University of Missouri (Columbia, MO, USA) in 1957 and brought with him seeds of Arabidopsis thaliana, a species on which he worked for well over 20 years. Unfortunately, he did not get funded from the US National Science Foundation (NSF) and was told to quit this species and work on other plants in seeking funds for future research. He was the only botanist in the USA working on A. thaliana for two decades. He made significant contributions, notably on the biology of this species (Rédei, 1969) and a two-volume encyclopedia of genetics. There were only a few others interested in this species, notably the Dutch botanist Maarten Koornneef, who ultimately constructed a detailed genetic map of A. thaliana. Rédei lived to see the publication of entire genome sequence of this species (Arabidopsis Genome Initiative, 2000), a project ironically funded from 1995 to 2000 by 100 million dollars by the NSF. When he died in 2008, there were ~16 000 laboratories worldwide focused on this little mustard using his research methods. By early 1980s, well-established molecular biologists, such as Chris Sommerville and Elliot Meyerowitz, specialists on Escherichia coli and Drosophila genetics, respectively, had already switched to Arabidopsis studies (see Pennisi, 2000). Since then, numerous publications have covered the value of this model mustard in basically most fields of biological research. More recently, species of other mustard genera (e.g. Arabis, Boechera, Brassica, Capsella, Eutrema, Lepidiume) have also become model organisms for various research fields.

Herbert Hurka (5 May 1940–), professor emeritus (Osnabrück University, Germany), worked initially (1965–1974) on fossil algae, then (1976 onwards) on the Brassicaceae (mainly Capsella). His research, in addition to collaboration with 15 graduate students (e.g. M. A. Koch, K. Mummenhoff, A. Franzke, B. Neuffer), focused initially on genera such as Arabidopsis, Arabis, Capsella, Cardamine, Cochlearia, Diplotaxis, Lepidium, Microthlaspi, Rorippa and Thlaspi. In collaboration with hundreds of botanists worldwide, they addressed numerous phylogenetic, genomics and systematic aspects of family.

Jeffrey D. Palmer (professor emeritus, Indiana University) was among the first who worked on the plastid phylogeny of three ‘diploid’ species and their amphiploid derivatives commonly known as the ‘triangle of U’. Thomas C. Osborn (formerly in Wisconsin University, now Bayer Company, St. Louis) was among the earliest to focus on the economically important crops of the genus Brassica. Osborn was the pioneer on RFLP (restriction fragment length polymorphism) mapping in Brassicaceae and has been a leading scientist in developing molecular breeding in Brassica in Bayer Company. Notable among his former students are J. Chris Pires and M. Eric Schranz, both of whom continue to work on various evolutionary aspects of the family.

IMPACT OF ARABIDOPSIS ON BRASSICACEAE SYSTEMATICS

Prior to the modern era of molecular phylogenetics, 62 species were assigned to Arabidopsis, with a Himalayan centre of highest diversity (O’Kane and Al-Shehbaz, 1997). The genus currently includes 18 species in the unigeneric tribe Arabidopsideae, and the remaining species were assigned to ten other tribes (German, 2019; German et al., 2023).

Developmental studies on Arabidopsis and the role of a few mutations in creating substantial morphological differences have altered the thinking of systematic botanists in assessing relationships and delimitation of taxa at the tribal, generic and species levels (see review by Bowman, 2006). A handful of examples demonstrate this. First, bracteate vs. ebracteate racemes have long been considered an essential key character taxonomically. However, Dinneny et al. (2004) showed that misexpression of the JAGGED gene in A. thaliana causes the formation of bracts, whereas its suppression prevents formation of bracts. The presence of bracts has traditionally been used to separate several related genera (e.g. Erucastrum from Brassica). Second, the work of Melzer et al. (2008) demonstrated in A. thaliana the role of expression of the MADS-box genes Soc1-3 and FUL-2 in the apical meristems for inducing flowering, but their mutants fail to do so, and the plants continue to grow vegetatively for a few years, the stems become woody, and the plants become perennial. Most recently, Zhai et al. (2024) found that a single MADS-box gene is capable of converting an annual to a perennial.

Fruit dehiscence vs. indehiscence were emphasized by Schulz (1936), who placed the 14 genera with indehiscent fruits in the Euclidieae, and they are currently assigned to eight tribes. A case in point, Schulz (p. 475) recognized two species in Euclidium (E. syriacum and E. tenuisissimum), but phylogenetic studies retained the first in the Euclidieae and the latter, as Litwinowia, in the Chorisporeae. The molecular background that regulates fruit dehiscence in A. thaliana was well studied for more than a decade, starting with Liljegren et al. (2000) and discussed well by Mühlhausen et al. (2012). As shown by these authors, expression of the developmental genes ALCATRAZ, INDEHISCENT and SHATTERPROOF (ALC, IND, SHP1 and SHP2, respectively) in A. thaliana, which controls the valve-margin dehiscence zone of parenchymatous cells, leads to dehiscent fruits, and their failure in expression leads to formation of sclerenchyma that results in indehiscent fruits. Mühlhausen and colleagues demonstrated that orthologues of these genes resulted in dehiscent fruits in Lepidium campestre and indehiscent ones in L. appelianum. Their cladogram shows that indehiscent fruits evolved independently at least nine times within Lepidium. A survey of the family for indehiscent fruits shows that it occurs in 26 of the currently recognized 58 tribes. Over the years, taxonomists have placed morphologically related species with dehiscent and indehiscent fruits in separate genera, and some rejected placing them together based on molecular data. The classic examples include indehiscent silicles of monospecific Twisselmannia, Ochthodium, Tchihatchewia and Brachcarpaea, Cycloptychis, Silicularia, Schlechteria and Thalspeocarpa (two species), all of which nested within and have already been united with Tropidocarpum (Al-Shehbaz, 2003b), Sisymbrium (Warwick et al., 2010; Žerdoner Čalasan et al., 2021) and Hesperis (German and Al-Shehbaz, 2018) and (last five genera) with Heliophila (Al-Shehbaz and Mummenhoff, 2005), respectively. Based on molecular studies, Warwick and Al-Shehbaz (1998) transferred monospecific Boleum (indehiscent silicles) and Euzumodendron (dehiscent siliques) to Vella (dehiscent silicles). The opposite scenario is when almost all morphologies lead to the placement of unrelated plants in the same genus. A case in point is the genus Conringia, which used to include six species, but has recently been split into three genera of three tribes: unigeneric Conringieae includes only three species, C. planisiliqua is transferred to monotypic Iljinskaea of Isatidieae (Al-Shehbaz et al., 2021), and the remaining two species plus three added from Zuvanda and a novelty form Plagioloba in tribe Plagiolobeae (German, 2021; Khosravi and Eslami-Farouji, 2022). A far more classic example is the Coluteocarpeae, where some molecular studies support the recognition of 14 genera whereas others support a unigeneric tribe (see below in the section on "Adjustments on problematic tribes and genera").

Rosette flowering, which is the development of solitary flowers (or fruit) on long pedicels originating from the basal rosette, is well known among some genera closely related to others with perfectly developed racemes. Yoon and Baum (2004) found that changes at the LFY locus were involved in the multiple independent origin of rosette flowering and failure to produce racemes. This feature characterizes all species of the previously recognized South American Brayopsis that Salariato et al. (2022) found nested within Eudema. Their findings were the basis for making several generic adjustments, including the recognition of Ancashia and Borealandea (Al-Shehbaz et al., 2023). In fact, sometimes both rosette flowering and the production of racemes can be found within a single species, as in the eastern North American Leavenworthia and Selenia (Urban and Bailey, 2013).

Kempin et al. (1995) demonstrated the effect of APETALA1 and CAULIFLOWER genes, both of which are expressed in floral primordia and perhaps function as transcription factors. Their study of cauliflower (Brassica oleracea var. botrytis) showed that the CAULIFLOWER gene homologue in this plant is non-functional, and it presents the molecular background for production of the cauliflower head, a feature that puzzled botanists for hundreds of years.

Lee et al. (2019) showed some unusual, although very interesting, mutations in A. thaliana, especially in the transcription factors CRABS CRAW (CRC) and SUPERMAN (SUP) and their major roles in the floral meristem determinacy. The CRC mutation leads to apocarpous gynoecium, in which the two carpels of the pistil remain free, whereas the SUP gene mutants, first discovered by Bowman et al. (1992), produce flowers with numerous stamens. It is yet to be determined whether Megacarpaea bifida and M. polyandra, the two species in the family with polyandrous stamens (Al-Shehbaz, 2015), have a mutant SUP gene as does Arabidopsis. The earlier literature on the molecular developmental biology on the latter genus is substantial, and Bowman et al. (1992) should be consulted for leads.

The above six classic examples amply demonstrate the caution systematists ought to take in relying heavily on a single morphological feature in delimiting taxonomic ranks above the species and without knowledge of the processes controlling the developmental aspect of a given feature.

Early molecular studies and their systematic implications

Numerous botanists contributed >200 publications from the first on Brassica in 1978 to 2005, prior to the establishment of monophyletic tribes (Al-Shehbaz et al., 2006). More than 60 % of these tribes are well covered by Koch et al. (2003), who listed all used markers and studies of taxa, in addition to Arabidopsis and its relatives, migration and phylogeography, and molecular clock estimates and age of Brassicaceae. Most studies focused on Arabidopsis, followed by Brassica and Brassiceae, Thlaspi s.l. complex, Lepidium, with fewer on Cardamine, Cochlearia, Draba, Boechera, etc. covering an additional 24 genera. The studies used one to few markers, including RFLP, chloroplast Rubisco, alcohol dehydrogenase (rbcL), nuclear chalcone synthase and arginine decarboxylase, plastid ndhF and mat-K, nuclear gene pistillata and, subsequently and most frequently, the nuclear ITS and/or plastid trnL-F spacer.

Because of space limitations, only four genera are dealt with here in some detail, plus a few miscellaneous ones. Examples of all genera covered in this section amply demonstrate the tremendous value of molecular phylogenies in the evaluation of various morphological markers, in the delimitation of genera and tribes, and in assessing relationships among or within genera.

Brassica and Brassiceae: example from a clade comprising important crop species

Eight of the earliest molecular phylogenetic studies in the family (Uchimiya and Wildman, 1978; Erickson et al., 1983; Palmer et al., 1983; Yanagino et al., 1987; Figdore et al., 1988; Song et al., 1988a, 1988b, 1990) focused on Brassica species, especially the six economically important ones commonly known as the U triangle (U, 1935). The species involved are three ‘diploid’ B. rapa (n = 10; genome A), B. oleracea (n = 9; genome C) and B. nigra (n = 8; genome B) and their allopolyploid hybrids B. napus (n = 19; AC), B. juncea (n = 18; AB) and B. carinata (n = 17; BC). Although the earlier literature treated the former three Brassica species as ‘diploids’, it is generally accepted that they, and the entire tribe Brassiceae, have triplicated genomes, whereas most of the family have whole genome duplication (WGD) (Lysak et al., 2005).

In the first tribal molecular phylogeny in the Brassiceae, Warwick and Black (1991) included 28 species in nine genera and were the earliest to show that the traditional limits of Brassica, Diplotaxis and Sinapis are not monophyletic, that B. nigra is sister to Sinapis arvensis but not related to the core members of the genus, and that the genera sampled fell into the rapa-oleracea and the nigra lineages.

In 1800 the German botanist Johann J. Bernhardi placed B. nigra in its own genus, Mutarda, but subsequent botanists did not follow that. Mutarda is currently accepted and includes B. carinata, B. nigra, S. allionii and S. arvensis (German, 2022). In a more comprehensive study, Warwick and Sauder (2005) addressed the evolutionary relationship within the tribe using ITS sequence data from nuclear ITS and chloroplast trnL intron and restriction-site polymorphisms. They included 85 species in 29 genera, although they placed within the tribe the controversial Calepina and Conringia perfoliata (currently assigned to the Calepineae and Plagiolobeae, respectively), in addition to Orychophragmus (see the section on "Adjustments to problematic tribes and genera"). Their data show several genera that are polyphyletic or paraphyletic, and these are addressed in the section on tribal and generic adjustments.

The controversial case of Vella, one versus three genera, was addressed molecularly by Warwick and Black (1994), who later (Warwick and Al-Shehbaz, 1998) united the genus with Boleum and Euzomodendron based on molecular, cytological (all have x = 17) and morphological data. Crespo et al. (2000) found the last genus is nested within Vella in the ITS analysis but was sister to Vella in the combined molecular and morphological data. More recently, Simón-Porcar et al. (2015) found both Boleum and Euzomodendron nested within Vella in all nuclear and plastid phylogenies. The combined genus is a unique basal group in the tribe Brassiceae by having x = 17 chromosomes, a number found only in the unrelated allotetraploid Brassica carinata discussed above.

Thlaspi: past influence of molecular phylogeny in advancing its limits

Three early studies (Mummenhoff and Zunk, 1991; Koch et al., 1993; Mummenhoff and Koch, 1994) focused on the Thlaspi s.l., a taxonomically difficult and heterogeneous genus that Meyer (1973, 1979) divided into 12 segregate genera based primarily on differences in seed-coat anatomy, and several taxonomists (e.g. Al-Shehbaz, 1986) then did not follow that narrow concept. The first two studies used isoelectric focusing analysis of Rubisco, and the third explored the plastid DNA restriction site variation. They demonstrated amply that the Thlaspidinae sensu  Schulz (1936) is not monophyletic, that some of its component genera ought to be excluded and that Thlaspi segregates do not fall together with this genus. Subsequent family-wide phylogenetic studies placed Thlaspi and few related genera in the tribe Thlaspideae, whereas all of Meyer’s 11 segregates plus an additional pair described by Ali et al. (2016) were assigned to the Coluteocarpeae, a tribe with controversial generic boundaries discussed further in the section on "tribe Coluteocarpeae".

Arabidopsis as a model

Mummenhoff and Hurka (1994) used isoelectric focusing analysis to study the polypeptide composition of Rubisco subunits in three species of Arabidopsis, and they were the first to support Hylander’s (1957) hypothesis of the hybrid origin of A. suecica from A. arenosa (both as Cardaminopsis) and A. thaliana. The nuclear-encoded small subunits of A. arenosa and the chloroplast-encoded large subunits of A. thaliana provided the evidence supporting paternal and maternal origin of A. suecica from these two species, respectively. Their findings agreed fully with the study by Price et al. (1994), which was probably published simultaneously (see text on "Family-wide monophyleti studies"), and with the studies by Mummenhoff and Hurka (1995), Kamm et al. (1995) and O’Kane et al. (1996) using different molecular markers.

Prior to molecular phylogenies and as indicated above, the limits of Arabidopsis were so broadly circumscribed that it included 69 species centred in the Himalayan region. Koch et al. (1999) presented strong evidence that this genus is paraphyletic and, together with early studies, many species were excluded and assigned to other tribes, including Alyssopsideae, Arabideae, Camelineae, Crucihimalayeae, Descurainieae, Erysimeae, Euclidieae, Eutremeae, Oreophtoneae and Schrenkielleae (see O’Kane et al., 1996; O’Kane and Al-Shehbaz, 1997, 2003; Al-Shehbaz et al., 1999; Al-Shehbaz and Yunnusov, 2003; Al-Shehbaz and Warwick, 2004; Beck et al., 2007; German, 2019). The initial tribal placement of Arabidopsis in the tribe Sisymbrieae remained since 1821, then it was assigned to the Camelineae (Al-Shehbaz et al., 2006) and eventually to its own tribe, Arabidopsideae (German et al., 2023). The genus currently includes 18 species centred primarily in Europe and Asian Far East.

The literature on phylogenetic relationships within Arabidopsis is extensive, and the interested reader should consult Schmickel and Koch (2011, and references therein), who demonstrated that hybridization, introgression and polyploidy during the glaciation–deglaciation cycles played a major role in the evolution of the genus in the amphi-Beringian region and Fennoscandinavia.

Lepidium and expansion of its traditional limits

The earliest molecular studies on Lepidium, which used isoelectric focusing analysis of Rubisco (Mummenhoff and Hurka, 1991; Mummenhoff et al., 1992) and chloroplast DNA restriction-site variation (Mummenhoff et al., 1995), were found to be useful at the sectional classification of the genus. Mummenhoff (1995) found that Cardaria draba should be retained in Lepidium, thus supporting Linnaeus in his original generic placement of the species in 1753.

Lepidium flowers vary in the development of petals and the number of stamens, and the basal group of species have well-developed petals and six stamens. I have been working for more than three decades on a monograph of the genus, and examination of the flowers of 262 species shows that 180 have fully developed petals, 59 have reduced ones, and 23 are apetalous. Furthermore, 122 species have six stamens, 13 have four median stamens, 28 have two lateral and two median stamens, and 99 have only two median stamens. Study of the floral developments in Lepidium by Bowman and Smyth (1988) and Bowman et al. (1999) showed that petal primordia are always formed, although they are suppressed in species with reduced or deficient petals. In contrast, the reduction in the number of stamens results from the loss of lateral stamens (stamens become four median), fusion of primordia of median pairs (two median + two lateral stamens) or loss of laterals and primordial fusion of medians (two median stamens). In their study of 24 species, Bowman et al. (1999) showed that the independent reduction to two stamens happened at least twice, that the reduction to four stamens also evolved independently, and that one reversal from two stamens led to four (2 + 2) in the New Zealand-endemic Lepidium oleraceum. This species is nested within a clade in which its two sampled species (African Lepidium africanum and Australian Lepidium hyssopifolium) and its assumed ancestors have two stamens.

Lepidium species with dehiscent fruits usually exude copious mucilage upon wetting, which enables them to disperse by sticking to the body of migratory animals. Many species are autogamous, and this breeding system facilitates establishment in new habitats. Lee et al. (2002), who used the first intron of the pistillata single-copy nuclear gene, showed that a number of New World and Australian species have two or three copies of that gene from different origins and concluded that these species are allopolyploids.

Equally interesting is solving the origin of Australian/New Zealand species of Lepidium. The Australian species exhibit the most diversified floral and fruit morphology anywhere in the genus, whereas New Zealand species are the least diversified in most morphological aspects. Using ITS and chloroplast DNA phylogenies, Mummenhoff et al. (2004) studied 18 native species of both countries and showed that all shared Californian chloroplast DNA (trnT-L spacer, trnL intron and trnL-F spacer), of which 11 also shared Californian ITS and seven shared South African ITS. They suggested an origin of Australian/New Zealand species during the Pliocene/Pleistocene, which were the product of two trans-oceanic bicontinental dispersal events, followed by hybridization and subsequent homogenization of ITS to either the South African or Californian types. However, their study did not account for the ploidy level in the species involved.

Dierschke et al. (2009) used genomic in situ hybridization and florescence in situ hybridization to test the above proposed migration, hybridization and ITS homogenization. Their genomic in situ hybridization findings confirmed the hybrid origin of seven species but did not support such an origin in nine allopolyploid species and assumed that the Californian ancestors radiated rapidly in Australia and New Zealand, followed by hybridization with African species. The chromosome numbers ranged from octoploids (2n = 64) in the New Zealand Lepidium sisymbrioides and Lepidium tenuicaule (Soza et al., 2014) to 18-ploid (2n = 144) in the Australian Lepidium naufragorum (De Lange and Murray, 2002), the highest ever recorded chromosome number in Lepidium, although octoploids are uncommon, because they are found in few species in Asia and the Americas (Warwick and Al-Shehbaz, 2006).

Lepidium sisymbrioides, an endemic of the South Island of New Zealand (Otago region) and the sole dioecious species in the entire Brassicaceae, was subjected to detailed taxonomic (Heenan et al., 2007) and developmental, genomic and molecular phylogenetic study (Soza et al., 2014). The latter authors used the first intron of the gene pistillata and found that L. sisymbrioides and L. tenuicaule originated from hybridization and polyploidization of four distinct diploid genomes. Their developmental studies demonstrated that dioecism in L. sisymbrioides evolved by the selective reduction of reproductive organs at their intermediate developmental stage, during which cell death occurred in stamens of pistillate flowers and by cell arrest in ovaries of staminate flowers.

The generic limits of Lepidium expanded a great deal based on molecular studies (Mühlhausen et al., 2012) to include species previously assigned to Cardaria, Coronopus, Cyphocardamum, Lithodraba, Stroganowia, Stubendorffia and Winklera (see Al-Shehbaz et al., 2002; Al-Shehbaz and Mummenhoff, 2011; Al-Shehbaz, 2021).

The combination of phylogenetic and floral developmental studies in Lepidium provided the solid background on which to base a sound taxonomy. As a result, Lepidium became a more inclusive monophyletic genus, instead of paraphyletic, within which some segregates were polyphyletic.

Arabis and Boechera: past and present delimitation and tribal affiliation

Species of the almost exclusively North American Boechera were previously assigned to Arabis, although the former was first recognized in 1976 but ignored by subsequent botanists (e.g. Rollins, 1993). Using ITS, Koch et al. (1999) were the first to point out that Arabis is paraphyletic and that its 15 sampled species belong to five genera, including six species in Boechera. A year later, Koch et al. (2000) studied both genera using chalcone synthase and alcohol dehydrogenase and reached the same conclusions.

These pioneering studies prompted Al-Shehbaz (2003a) to transfer most of the American Arabis to Boechera. Both hybridization and apomixis were firmly established in Boechera (see Koch et al., 2003; Schranz et al., 2005), and apomixis (via apospory or diplospory) is currently known in five of the nine genera of tribe Boechereae, including individual species in Borodinia, Phoenicaulus, Polyctenium and Sandbergia, but more so in >20 species of Boechera (Mandáková et al., 2020a). Systematics of the tribe Boechereae has been expanded substantially in the past decade (Alexander et al., 2013; Hay et al., 2023, and references therein) to include >120 species, all except two of which are distributed primarily in the USA and less so in Canada. Both Boechera furcata and Borodinia macrophylla are endemic to Siberia and the Russian Far East.

Arabis was very broadly delimited to include 180 species (see Al-Shehbaz, 1988), but the painstaking and continuous molecular studies from the late 1990s to the present in Marcus Koch’s laboratory (Heidelberg University) have resulted in substantial progress in the past 25 years. Following the transfer or recognition of >100 species from Arabis to 25 genera of 14 tribes, including Acirostrum, Boechera, Borodinia, Catolobus, Crucihimalaya, Dendroarabis, Fourraea, Hurkaea, Parryodes, Pennellia, Pseudoturritis, Scapiarabis, Sinoarabis and Turritis, to name some, Arabis s.s. remains large, with ~100 species, and still represents a paraphyletic taxon. However, on species, Arabis verna should be excluded from Arabis and united with Aubrieta (Karl and Koch, 2013; Koch et al., 2017). In fact, the last reference indicated (p. 19) for Aubrieta that ‘Its closest relative and sister taxon is Arabis verna, from which Aubrieta diverged ~2.7 Mya’. Placing this single species in a monospecific genus is unrealistic, because the only differences would be the annual (vs. perennial) habit and minor differences in fruit, both of which can be unreliable in generic delimitation.

Genomically speaking, Arabis retains the ancestral crucifer karyotype of n = 8, though it differs from the vast majority of genera of Brassicaceae by having frequent centromere repositioning found thus far in ≥26 species, in comparison to three species from the rest of the family and found in Camelina, Cardamine and Neslia (Mandáková et al., 2020b). Despite the tremendous progress made on Arabis thus far, more work in needed because the conventional few markers (in comparison to the entire genome) are insufficient, as evidenced from study of the Arabis scabra group (Koch et al., 2020), where the generic type, Aabis alpina is basal to the rest of the genus that included Draba hispanica, Dendroarabis frutescens and Pachyneurum grandiflorum in both ITS and plastid trnLF. Should whole genomic studies covering all or majority of species in the genus hold for the position of these three species and those of smaller genera, placing all these taxa within a more inclusive Arabis would require only minimal nomenclatural changes, because the vast majority of their species were transferred most recently to Arabis. The literature on Arabis and Arabideae is substantial, and the interested reader should consult Karl and Koch (2013, 2014), Koch et al. (2020), Wötzel et al. (2021), and references therein. The most recent contribution on the evolution of tribe Arabideae is based on a 1000-gene dataset and demonstrates, while carefully studying polyploidy and the evolution of ortho- and paralogue genes, reticulate evolutionary processes among most lineages and clades (Walden et al., unpublished)).

Cardamine

The earliest molecular study on the genus, by Franzke et al. (1998), used ITS and the plastid trnT-L spacer of ~20 species from Australia, Eurasia and New Guinea, and they found that Dentaria is nested within Cardamine, polytomy of Eurasian species, and some of the early recognized sections were monophyletic. Sweeny and Price (2000) used the plastid ndhF and trnL intron of 15 species and agreed with the first study and demonstrated that Dentaria has multiple origins, with monophyletic Cardamine sister to Nasturtium. Bleeker et al. (2002) used ITS and the plastid trnL intron and trnL-F spacer to study the phylogeny/biogeography of 16 species from three continents and found low sequence divergence between South American Cardamine glacialis and Australian species, suggesting long-distance dispersal between the two continents ~0.5 Mya. They also studied Cardamine hirsuta from three continents and indicated a polyphyletic origin of the species because of wide differences in the positions of plastid and nuclear markers on the tree. Finally, using the same markers as the last study, Carlsen et al. (2009) studied ~80 species from all continents and agreed with earlier studies of rapid radiation and speciation, rather long-distance dispersals, distinct lineages and widespread polyploidization. More comprehensive studies covering nearly all species and their genomes are needed to address the phylogeny of one of the most complex and morphologically diverse second largest genus in the family.

Cardamine irsute has become a model for studies of leaf-shape morphology. Hay and Tsiantis (2006) used the species and Arabidopsis thaliana to find that the genetic basis for differences behind the formation is pinnately divided leaves in the former and entire leaves in the latter species. They found that in A. thaliana the exclusion homeobox KNOX protein from the leaves led to the production of entire, undivided leaves, whereas its expression in C. irsute delays the cellular differentiation and leads to the formation of pinnately lobed leaves. In contrast, Vlad et al. (2014) studied in both species the effect of REDUCED COMPLEXITY (RCO) homeodomain protein, because its expression in the leaves is required for the development of pinnately divided leaves in C. irsute and its loss leads to the formation of entire leaves in A. thaliana.

The South American Cardamine chenopodiifolia is the only species in the family that exhibits amphicarpy (production of subterranean indehiscent silicles and aerial dehiscent siliques). Emonet et al. (2024) found that the morphological differences between these fruit types are reflected in ‘a large number of differentially regulated genes’ involved in various functions, such as cell-wall formation, photosynthesis and defence responses. The interested reader is advised to consult their fascinating findings.

Draba

This complex and largest genus in the family lacks comprehensive, worldwide systematic and phylogenetic studies, and only scattered contributions have been made. Among the earliest studies are those by Widmer and Baltisberger (1999) and Scheen et al. (2002), which dealt with the polyploid origin of central European Draba ladina and polar Draba lactea, respectively, and some other circumpolar Draba were covered by Grundt et al. (2004). The first large-scale study was by Koch and Al-Shehbaz (2002), who used ITS and the trnL intron and spacer to study the phylogenetic relationships of 72 American and six European species. They showed that the existing sectional classification of the genus is artificial, the European plastome is widely distributed among the New World species, Erophila should be merged in Draba, and that both Draba and Arabis should be maintained in the same tribe. Several American annuals formed a clade distinct from the rest of the genus and are currently placed in Tomostima (six species), a North–South American genus first recognized in 1825. Beilstein and Windham (2003) studied 17 North American species using ITS, and the species sampled fell in a group with x = 8 and another with deviating aneuploid numbers. Their data also show Erophila and Drabopsis nuda nested within Draba, thus supporting prior findings by Koch and Al-Shehbaz.

Finally, Jordon-Thaden et al. (2010) sampled 169 species and analysed their relationships using ITS and trnL-F markers. Their data supported the recognition of the American Tomostima and Abdra and European monospecific Drabella (D. muralis) as separate genera sister to Draba, in addition to the merger of Arabis rimarum, Drabopsis, Erophila, Graellsia hederifolia (presently recognized as Helicodraba hederifolia) and Schivereckia in a more inclusive Draba. Fortunately, the entire genus and Arabideae are undergoing extensive phylogenetic studies in Marcus Koch’s laboratory.

Halimolobine taxa

This New World alliance was studied by Bailey et al. (2002) using nuclear ITS and pistillata and first intron and plastid trnL-F region, and they concluded that the core group forms a monophyletic clade of well-supported species subgroups corresponding to five genera. Following recognition of the alliance as the new tribe Halimolobeae (Al-Shehbaz et al., 2006), Bailey et al. (2009) delimited it to include 44 species in five genera and make the tribe the first to be well studied phylogenetically and systematically.

Heliophila

This southern African genus is the most diversified in habit, flowers and fruits in the entire Brassicaceae, and all 106 of its species are native to South Africa (92 endemic), with the ranges of ten extending into neighbouring Namibia, three in Lesotho, and one in both countries and Swaziland (author's unpublished compilation). Prior to the first molecular study by Mummenhoff et al. (2005), Brachycarpaea (one species), Chamira (one), Cycloptychis (two), Schlechteria (one), Silicularia (one) and Thlaspeocarpa (one) were recognized as distinct genera. Their ITS sequence data of 46 species of the above seven genera found that Chamira (currently in the unigeneric tribe Chamireae) is a sister clade to the rest, that the other five other genera are nested within Heliophila, and that the ingroup falls in three clades, of which two are annuals and one includes perennials and subshrubs. As a result, Heliophila is currently a more inclusive genus (Al-Shehbaz and Mummenhoff, 2005). Mandáková et al. (2012) supported the above study and division of Heliophila into three clades, and found that the annual clade mostly has 2n = 20 and fewer 2n = 22, and the perennials have 2n = 16 and 22. Their chromosome painting study indicated that 90 % of the painting probes show three homeologous regions in their haploid chromosome complement, thus suggesting that the entire Heliophileae have whole genome triplication (WGT) following hybridization between genomes resembling those of the ancestral crucifer karyotype, and that those ancestors underwent genome-wide diploidization. For further and more detailed updates on the above two tribes and the study of random repeats and Heliophila cladogenesis, the reader is encouraged to consult Dogan et al. (2021).

Sisymbrium

By the beginning of this millennium, Sisymbrium included 94 species, of which 53 are restricted to the New World and 41 to Eurasia and Africa. Warwick et al. (2002) presented the first molecular phylogenetic study on the genus using ITS of 33 species from all previously recognized sections and from 30 additional species previously placed in the genus or related genera. They found out that the Old World species Neotorularia aculeolata and N. afghanica, and S. linifolium (Canada and USA; as Schoenocrambe) fall in one clade, that all of the New World species form another clade with members of the Thelypodieae, and that S. supinum (Europe) and S. thellungii (South Africa) form a clade with the Brassiceaen genus Erucastrum (see Warwick and Al-Shehbaz, 2003). In a subsequent study, Warwick et al. (2010) found that both Ochthodium aegyptiacum and Pseudofortuynia esfandiari (later P. leucoclada) form a sister clade with Sisymbrium. Finally, Žerdoner Čalasan et al. (2021) carried out a thorough phylogenetic study using five nuclear- and three chloroplast-encoded loci in 43 of the currently recognized 46 Sisymbrium species. Their results agree fully with the above findings of Warwick et al. (2002, 2010), especially with the species formerly placed in Neotorularia, Ochthodium, Pseudofortuynia (see also German and Al-Shehbaz, 2018) and Schoenocrambe.

Family-wide monophyletic classification

Until early 1990s, the taxonomy, tribal classification and floristics of the Brassicaceae basically followed the German botanist Otto Eugen Schulz (31 October 1874–17 February 1936), who (Schulz, 1936) divided the family into 19 tribes (six of which were published invalidly) and 351 genera, of which 103 were subsequently synonymized by various authors. Except for unigeneric tribes, his other tribes were delimited artificially on a few characters that were shown to be highly plastic owing to convergence, an evolutionary event not recognized by him. A case in point is the Lepidieae, a tribe delimited solely on angustiseptate fruits (flattened at a right angle to the replum), a fruit type currently known to have evolved independently in ≥29 of the presently recognized 58 tribes. Four of his tribes (Pringleeae, Romanschulzieae, Stanleyeae and Streptantheae) were reduced by Al-Shehbaz (1973) to synonymy of Thelypodieae. This almost exclusively New World tribe was first considered by Hayek (1911), followed by Schulz (1936),  Rollins (1939, 1956), Al-Shehbaz (1973, 1984), Cronquist (1981) and Takhtajan (1997), as basal in the family, because the flowers and fruits of some of its American genera (e.g. Stanleya, Warea) were superficially similar to those of some Cleome species (Cleomaceae) by having six stamens of equal length and fruits borne on long gynophores. In contrast, Dvořák (1973) indicated an Old World relationship to the Cleomaceae through the Brassicaceae tribe Hesperideae because of the presence in both families of multicellular glands, whereas Hedge (1976) believed that the Brassicaceae originated in Eurasia, the centre of its greatest diversity.

In addressing the phylogenetic relationships of Arabidopsis thaliana, Price et al. (1994) used the plastid rbcL gene of eight species in seven genera. Despite its small sampling size, their study was among the first to include members of multiple tribes of the family. It showed that the basal genus is the primarily SW Asian Aethionema, not the North American Thelypodieae. All subsequent family-wide molecular phylogenetic studies to the present have fully supported that finding and firmly established that the origin of the family was SW Asia. They also used the plastid restrictions-site variation and demonstrated that the tribes Arabideae and Sisymbrieae are polyphyletic, that the closest relatives of Arabidopsis are European species, that Arabis lyrata is closer to Arabidopsis than to other congeners, and that A. thaliana is the maternal parent of A. suecica (see also above).

Koch et al. (2001) addressed the molecular systematics using chs and matK of 38 species in 23 genera of five tribes and found that the Arabideae, Lepidieae and Sisymbrieae are not monophyletic. Perhaps one of the most significant and often overlooked contributions of the early 2000s is by Koch (2003). This classic work surveyed all the molecular markers used for all taxa to that date and presented the first comprehensive Brassicaceae phylogeny based on nuclear chs, adh and ITS and plastid matK of 167 species in 86 genera. Most the 32 currently recognized tribes received >50 % bootstrap support in that study, which recommended the merger of Drabopsis, Erophila and Schievereckia with Draba; Cheesemania and Ischnocarpus with Pachycladon; Iti and Dentaria with Cardamine; Schoenocrambe with Sisymbrium; and Gagria with Pachyphragma. It also supported the recognition of Nasturtium as distinct from Rorippa; Arabis, Fourraea and Boechera as distinct genera; and division of Thlaspi into the segregate genera first recognized by Mayer (1973). These generic alignments and several others were addressed by Koch et al. (2003), who presented a comprehensive survey of the literature, critically discussed the traditional versus molecular phylogenies, and demonstrated the widespread homoplasy in almost all morphological characters.

In synthesizing all of the ~250 molecular phylogenetic studies on Brassicaceae prior to the first formal recognition of monophyletic tribes (Al-Shehbaz et al., 2006), several accomplishments have been made. First, Arabidopsis thaliana became the model organism in modern biology, and its entire genome was sequenced before that of humans. Brassica and Capsella soon followed to include model species. Second, it was shown that A. thaliana chromosome number, the second lowest in the family, represents a reduction from ancestors with WGT, and that all members of tribe Brassiceae are derived from hexaploid ancestors. Third, the primarily SW Asian Aethionema is sister to the rest of the family, and these findings support the Old World rather than New World origin of the family, a concept followed for >80 years. Fourth, homoplasy is predominant in almost every morphological feature and that all prior tribal classifications are artificial. Fifth, the Brassicaceae, Cleomaceae and Capparaceae are distinct families, rather than one, and that the first two are sister families that underwent independent WGD. Sixth, early molecular phylogenies demonstrated the polyphyly of tribes Arabideae, Lepidieae and Sisymbrieae and genera Arabidopsis, Arabis, Sisymbrium and Thlaspi, all of which required splitting to establish monophyly. Seventh, early age estimation of the Arabidopsis- Cardaminopsis- and Aethionema-core Brassicaceae splits were established. Eighth, developmental–genetic studies of various characters in Arabidopsis demonstrated that many features, such as dehiscent vs. indehiscent, bracteate vs. ebracteate, herbaceous annual vs. woody perennial, rosette vs. raceme flowering, to name a few, are controlled by changes in the function of a few genes, and that they should not be overemphasized in the delimitation of taxa because of their occurrence within individual genera (e.g. Lepidium). Finally, the family genome consists of well-defined chromosome blocks, and their study provides a powerful tool in establishing the phylogenomics within and among tribes.

The first large-scale, single-gene family phylogeny was introduced by Beilstein et al. (2006) using the plastid ndhF gene sequences of 113 species in 101 genera. The study provided monophyly of 20 clades with bootstrap support ~85 % or higher, and most clades were placed in three lineages (I–III), although several other clades were not assigned to lineages. This study, together with the wealth of >250 molecular studies pre-dating it, were synthesized by Al-Shehbaz et al. (2006) to provide the basis of the first molecular phylogeny that formally recognized 25 monophyletic tribes, to which were assigned ~77 % of the genera. However, the assignments of some genera to tribes with no molecular data (e.g. Acanthocardamum in Lepidieae) or based on medium to low support (e.g. Malcolmia and Neotorularia in Euclidieae) were subsequently shown to result in polyphyletic tribes.

More than 30 family-wide phylogenies have since been published, and this section focuses only on the major advancements made in some of them. Notable among these was the global ITS phylogeny of Bailey et al. (2006) that included 461 species in 146 genera, along with supermatrix analysis of ten markers for 65 taxa. Their study demonstrated that several tribes (Alysseae, Anchonieae, Arabideae, Camelineae, Euclidieae, Lepidieae and Schizopetaleae) sensu  Al-Shehbaz et al. (2006) were not monophyletic. They also indicated that family-wide phylogenies based on single or few genes might be inadequate to resolve the deeper nodes among lineages owing to the extensive rapid radiation in their early evolution. Beilstein et al. (2008) were among the first to suggest the Brassicaceae as a model family ideal for developmental and evolutionary biology and the first to use nuclear PHYA in family-wide phylogenies. Their findings fully supported the study by Bailey et al. (2006) regarding polyphyly of the above tribes.

The ITS phylogeny of German et al. (2009), who sampled 184 species in 121 genera, recognized 35 monophyletic tribes, dismantled the Camelineae s.l. into additional tribes (Camelineae s.s., Erysimeae, Microlepidieae and Turritideae), revised the assignments of several genera to tribes (e.g. Borodinia in Boechereae, Pseudoclausia in Chorisporeae, Atelanthera and Streptoloma in Euclideae, and Pugionium in Megacarpaeeae), excluded a number of others from previous tribal assignments (e.g. Asperuginoides, Didymophysa and Ptilotrichum from Alysseae, Macropodium and Stevenia from Arabideae) and reduced a number of genera to synonymy (e.g. Berteroella and Ptilotricuum in Stevenia, Anchonium and Oreoloma in Sterigmostemum, and Transbergia in Crucihimalaya). A year later, Warwick et al. (2010) sampled species of 308 of the then recognized 336 genera (or 92 % of the family), added eight new tribes, excluded Thelypodieae from Schizopetaleae, and recognized a total of 44 tribes. They suggested the reduction of several genera by synonymy and assignment of others to tribes, for which further supportive evidence was not published until a decade or so later. These include synonymizing Tchihatchewia with Hesperis; Ochthodium and Pseudofortuynia with Sisymbrium; Chalcanthus and Pegaeophytum with Eutrema; and Lithodraba, Physocardamum and Stubendorffia with Lepidium; in addition to excluding Delpinophytum from the Lepidieae and the placement of Subularia and Idahoa in the Subularieae and Bivonaea and Horwoodia in the Brassiceae.

The ITS study of Khosravi et al. (2009) included 155 species in 72 genera, of which 13 were assigned to tribes for the first time. Among these were Acanthocardamum in the Aethionemeae, Brossardia in the Coluteocarpeae (as Noccaeeae), Pseudofortuynia in the Sisymbrieae, and Camelinopsis and Didymophysa in the Thlaspideae.

Couvreur et al. (2010) took a supermatrix approach using eight nuclear, mitochondrial and plastid markers for species of 230 genera in 33 tribes. They interpreted the largely unresolved backbone of the Brassicaceae as a product of the rapid radiation of the family. Two years later, the number of tribes jumped to 49 (Al-Shehbaz, 2012).

The recognition of a three-lineage division of the family continued from Beilstein et al. (2006) for >16 years, although Lineage II was substantially expanded to include almost all except those of lineages I and III (Franzke et al., 2011). Huang et al. (2016) based their study of 133 nuclear genes on only 55 species, and they divided the family into six clades, from basal F (Aethionemeae), followed by clade E (= Lineage III), clades B–D (Lineage II) and terminal clade A (Lineage I). Their phylogeny corresponds well to the later-published accounts by Guo et al. (2017), Nikolov et al. (2019) and Hendriks et al. (2023), except for the position of the terminal clade A.

The family nuclear phylogeny of Nikolov et al. (2019) dealt with 50 of 52 tribes and was based on targeted enrichment sequences of >940 000 bases (1827 exons) of 63 species plus sequence data of 16 species. They recognized five lineages (plus the Aethionemeae). As shown later, their entire phylogeny completely agrees with the latest phylogeny of Hendriks et al. (2023) as far as the major six clades of the family, including the Aethionema. Furthermore, they resolved the position of 16 previously unassigned species and proposed the recognition of Subularieae and new tribes each for Schrenkiella, Arabidopsis, Hemilophia and Dipoma, and for Arabis josiae and Fourraea.

The plastome phylogeny of Walden et al. (2020) was based on 250 species in 153 genera of 51 tribes, and the outgroup included members of 15 families of the order Brassicales. They recognized three lineages and noted conflicting cytonuclear phylogenies that they attributed to intertribal hybridization, rapid evolutionary radiation and nuclear introgression. Furthermore, they assembled a morphomatrix of 37 characters (including 111 character states) scored for the 351 genera. Such a pioneering detailed study led to the conclusion that the increased morphological disparity happened together in tribes with WGD or diversification rate shift. Finally, they estimated that Aethionemeae diverged from the rest of the family ~32 Mya and that the other lineages started to diverge about the boundary of Oligocene and Miocene some 23 Mya. Their supplementary notes dealt with many taxonomic discrepancies, no assignments, and problems that were mostly resolved subsequently.

With the publication of phylogenetically based intrafamilial classification (German et al., 2023), the recognition of two subfamilies (Aethionemoideae and Brassicoideae) and five supertribes (Arabodae, Brassicodae, Camelinodae, Heliophilodae and Hesperodae), the controversial recognition and number of lineages versus clades of previous studies was put to rest.

Finally, the most comprehensive suprageneric Brassicaceae phylogeny to date (Hendriks et al., 2023) was based on species of all but one of the 58 currently recognized tribes, and they sampled 319 of the 349 recognized genera using 1081 nuclear genes and 265 genera sampling 60 plastome genes. As in previous studies that included both nuclear and plastid markers, the findings showed discordance in the cytonuclear phylogenies largely attributed to hybridization and introgression within and between lineages. In fact, the plastome phylogeny showed that both supertribes Arabodae and Brassicodae were polyphyletic. Their overall findings most closely resemble those of Nikolov et al. (2019), although it supported the recognition of four of their five major clades. Despite the use of >1000 genes, the deeper nodes representing the supertribes remain incompletely resolved. Furthermore, Hendriks and colleagues estimated the divergence of Brassicaceae from its ancestor with sister family Cleomaceae to be 38.8 Mya (mid to late Eocene), late Oligocene (24.5 Mya) of family crown age, and the median crown age of the subfamilies and five supertribes to be early to middle Miocene (19.9–14.4 Mya). These age determinations are basically compatible with some of the earlier estimation (see below). Three tribes (Camelineae, Cochlearieae and Subularieae) were polyphyletic in the nuclear phylogeny, and the first was also in the plastome phylogeny.

Undoubtedly, with expansion of the study by Hendriks et al. (2023) to include all genera and almost all species of the family, their findings will probably lead to some adjustments by the recognition of additional smaller or monospecific tribes (e.g. intertribal hybrids) and resolve the status of controversial tribes, such as the Camelineae and Cochlearieae.

Estimates of Brassicaceae divergence dates

The divergence age of the Brassicaceae from the sister family Cleomaceae, as well as the split of the basal tribe Aethionemeae from the core tribes, varied substantially depending on the use of macrofossils versus the molecular dating using nuclear versus plastome data. The first such dating was done by Koch et al. (2000, 2001) using the nuclear alcohol dehydrogenase and chalcone synthase and plastid matK genes. They determined the split of Arabidopsis thaliana from its ‘generic’ relatives at 5.8 Mya and the split age of Aethionemeae from the rest of the family at ~40 Mya. Sharp contrasts to those estimates were determined by (Franzke et al., 2009) at mid-Miocene (15 Mya) and Beilstein et al. (2010) at early Eocene (54 Mya). The estimates of the last authors are the highest to date, and they also gave a higher split for A. thaliana from its congeners at 13 Mya. They used what subsequent workers (e.g. Franzke et al., 2009) considered as a controversial fruit fossil of Thlaspi primaevum as a reference point for age determination.

Using a combination of the mitochondrial nad4 intron 1 and a supermatrix approach of sequences of eight markers, Couvreur et al. (2010) determined the Brassicaceae age at 37.6 (24.2–49.4) Mya, and this comes rather closer to the latest determination by Hendriks et al. (2023) at 38.8 (36.9–40.5) Mya (mid-Eocene) using nuclear and plastome datasets and the Brassicales Turonian macrofossil Dressiantha bicarpellata (93.6–89.3 Mya). Using plastome data, Hohmann et al. (2015) determined the age of the Brassicaceae crown group radiation at 32.4 Mya and the A. thaliana split at 6 Mya, and their dates were basically similar to those determined by Koch et al. (2001) more than two decades earlier.

Regardless to the above differences in age determination of the split of Brassicaceae from sister family Cleomaceae or the age of core tribes of the family, most recent studies agree that the crown age of the family, at which the Aethionemeae diverged from core tribes of the family, was ~24.4 (23.1–25.7) Mya (Hendriks et al., 2023), at the transition between the Oligocene and Miocene. The family went through highest radiation about mid Miocene, as new ecological niches, such as deserts and scrublands, became available.

Phylogenomics

Parkin et al. (2005) demonstrated that Brassica napus (n = 19) shared 21 chromosome segments with Arabidopsis thaliana. A year later, Lysak et al. (2006) used comparative chromosome painting of A. thaliana (n = 5), A. lyrata and Capsella rubella (both with n = 8), Neslia paniculata (n = 7), and Hornungia alpina and Turritis glabra (both with n = 6) to study the mechanisms of chromosome reduction in A. thaliana. These two brilliant studies, together with the assumption that the ancestral crucifer karyotype had eight chromosomes (as in the above two species), were integrated by Schranz et al. (2006) to introduce an initial crucifer karyotype system of 24 chromosomal blocks that were adjusted a decade later to 22 (Lysak et al., 2016). Tremendous advancements in the phylogenomics of the family have since been made by Martin Lysak and Terezie Mandáková (Masaryk University, Czech Republic), who focused on >160 species of ~93 genera and some 41 tribes. Furthermore, it is well known that Aethionema and the rest (core) of Brassicaceae share the alpha (α) WGD after the split from sister family Cleomaceae, and that ancestors of both families and other of Brassicales preceding the split from Setchelanthaceae or possibly Limnanthaceae share the beta (β) WGD (Lysak, 2018; Mabry et al., 2023). The ages of WGD were estimated by Edger et al. (2018) to be 31.8–42.8 and 85–92.2 Mya for the α and β duplications, respectively. The Cleomaceae had independent WGD subsequently after split from sister Brassicaceae (Schranz and Mitchell-Olds, 2006). In addition to the WGD in 6 of the 17 Brassicales families, Mabry et al. (2023) addressed the comparative distribution of C₄ photosynthesis, glucosinolates (all except Koeberliniaceae), ~100 reversals from herbaceous to woodiness, and adaptations to extreme environment (see introduction of this paper).

The are too many studies demonstrating the value of phylogenomics in systematics to deal with here, and only exemplary cases are briefly discussed below. A classic early example involved the tribe Brassiceae. Previous delimitation of the tribe included the genera Calepina, Conringia and Orychophragmus. Lagercantz (1998) and Parkin et al. (2005) showed that Brassica nigra and B. napus have genome duplication or triplications, respectively. However, Lysak et al. (2005, 2007) demonstrated through comparative chromosome painting that the first three genera do not have the WGT that the rest of tribe has. The tribal placements of Calepina and Conringia were solved subsequently, but Orychophragmus remained unassigned to tribe (see the following section on Orychophrgmus).

At the tribal grouping, Mandánková et al. (2017b) showed that a clade of seven tribes, which were previously recognized as Lineage III (Beilstein et al., 2006) or clade E (Huang et al., 2016) and currently supertribe Hesperodae (German et al., 2023), forms a well-supported group characterized by having the largest genome size in the family and low chromosome number of n = 5–7. The ancestral karyotype remained conserved in Chorisporeae or has undergone reorganization via translocations (Euclidieae).

Karyotype evolution by descending dysploidy was observed in about half of the species of the tribes Anchonieae and unigeneric Hesperideae. They studied three species of the genera Farsetia, Lobularia and Morettia of tribe Anastaticeae and concluded that it lacks three genome block associations (A–B, M–Jb and U) characteristic of the ‘Hesperodae’.

Another early classic study involved the karyotype evolution of six tribes [Calepineae, Conringieae, Eutremeae, Isatideae, Coluteocarpeae (as Noccaeeae) and Sisymbrieae] with x = 7. Mandáková and Lysak (2008) indicated that the ancestor of this group, the Proto-Calenpineae karyotype shared five chromosomes and conserved genome blocks with ancestral crucifer karyotype (n = 8), that it was ancestor to the tribe Brassiceae prior to its WGT, and that the entire group forms a well-supported monophyletic clade. All seven tribes fell in Lineage II (Beilstein et al., 2006), clade B (Huang et al., 2016) and currently in supertribe Brassicodae (German et al., 2023). As currently delimited, this supertribe includes 14 tribes, all of which are based on x = 7 or rarely x = 6. Although no genomic studies are available for seven of these, they are most likely to follow that evolutionary pattern included in the above study. Orychophragmus and Sinalliaria do not belong to the Brassiceae (see the following section on the genus). As discussed below, the generic boundaries in Thelypodieae will definitely need major adjustments, and Pringlea (x = 6) do not fall morphologically and geographically with the rest of the tribe and are likely to be excluded, although further studies are needed.

The Australian/New Zealand tribe Microlepidieae (16 genera, 60 species) was studied critically by Mandáková et al. (2017c), who indicated that its allopolyploid (n = 15) ancestor resulted from hybridization between Crucihimalayeae (n = 8, maternal) and Smelowskieae (n = 7, paternal) that migrated through long-distance dispersal. The migrants subsequently went through diversification and diploidization that resulted in perennials with n = 10 (Pachycladon) or n = 12, then by autopolyploidy to n = 24 (Arabidella). However, the rest of Microlepidieae went through fast diploidization and produced annuals with n = 6 (Menkea) or the remaining genera (n = 7→4). The Australian Stenopetalum nutans (Mandáková et al., 2010), in addition to S. decipiens and S. velutinum (Shaw, 1972), are the only species in the Southern Hemisphere with n = 4, the lowest number of chromosomes in the family. It is shared by ten species of Physaria of the USA (P. bellii, P. brassicoides, P. chambersii, P. floribunda, P. geyeri, P. lepidota, P. oregona, P. rollinsii, P. thamnophila and P. vitulifera). However, several of these species are unrelated, and it is most likely that n = 4 evolved independently more than once within Physaria, a genus with many species having n = 4→7 to as high as n = 30 in P. arctica. Independent mesohexaploidy is known in Physaria, Cochlearia and Leavenworthia alabamica (Mandáková et al., 2017a), Brassiceae (Lysak et al., 2005, 2007) and Heliophileae (Mandáková et al., 2012). Unfortunately, many of the Physaria species are highly specialized edaphic endemics that are hard to grow in the greenhouse. The highest number of chromosomes in the family (2n = 256) was counted once, in Cardamine concatenata of the eastern to central USA (see Warwick and Al-Shehbaz, 2006).

The first phylogenetic study of tribe Thlaspideae (Esmailbegi et al., 2018) showed that the currently recognized Alliaria petiolata consists of diploid populations (n = 7; Caucasus and SW Asia) that forms a sister clade to Lysakia (n = 7) and Sobolewskia, whereas the hexaploid populations (n = 21; Europe, naturalized in the New World, Australia and New Zealand) are sister to Parlatoria cakiloidea (n = 7). In their study of the genome structure of the tribe, Bayat et al. (2021) demonstrated substantial chromosome rearrangements via pericentric inversions. Their comparative chromosome painting also showed that chromosome 4 of hexaploid A. petiolata consists of three identical copy pairs, one of each is homologue to ancestral structure, one identical to that of P. cakiloidea and one to diploid A. petiolata, and that there are 24 repeat clusters found only in the three taxa. Based on these findings they proposed that European A. petiolata is the allohexaploid product of interspecific hybrid genomes of species related to the present P. cakiloidea and diploid A. petiolata. The non-weedy and restricted diploids and aggressively weedy and widespread hexaploid populations are indistinguishable in basically every aspect of morphology except the pollen size and shape (see Esmailbegi et al., 2018: 327). Based on these findings, nomenclatural adjustments are underway.

The most recent comparative phylogenomic study of the Brassicaceae by Walden and Schranz (2023) included entire genome sequences of 11 annual diploids representing all six major clades of the family. They concluded that genomic synteny reliably identifies orthologues for comparative phylogenomics, and they were the first to construct the ancestral genome of core Brassicaceae pre-dating its lineage diversification some 25 Mya. Liu et al. (2024) focused on the genome architecture of the family during its early diversification by the analyses of chromosome structure of 14 annual species of 11 tribes representing the Aethionemeae and members of the five supertribes sensu  German et al. (2023). They added one species each of Alysseae, Anastaticeae, Erysimeae, Megacarpaeeae and Thlaspideae, and they identified >9700 conserved genes and constructed the genome of ancestral core Brassicaceae based on nine ‘pseudochromosomes’ with 65 conserved genome blocks. They placed Megadenia pygmaea (Biscutelleae) in the supertribe Camelinodae, thus agreeing with the phylogenies of Kiefer et al. (2019) and Guo et al. (2021) and conflicting with tribal assignment of Hendriks et al. (2023) in the supertribe Heliophilodae. The placement of Megadenia in the Biscutelleae was first elucidated by German et al. (2009) and poses no problem, but it is a controversial position of the tribe relative to the supertribes.

The genomes of ≥49 Brassicaceae species are known (Kiefer et al., 2019; Liu et al., 2024), although ~90 are listed in GenBank, of which the smallest (120 Mbp) in the family is found in Arabidopsis thaliana and Schrenkiella parvula (Dassnayke et al., 2011, Lamesh et al., 2012; respectively), although a more recent publication (Hou et al., 2022) reported 133.9 Mbp for A. thaliana. The largest genome in the family thus far reported is 8117 Mbp in tetraploid Hesperis matronalis (Hloušková et al., 2019). In contrast, the smallest genome size, only 61 Mbp, in flowering plants is found in Brazilian carnivorous plants Genlisea tuberosa (Lentibulariaceae) (Fleischmann et al., 2014), whereas the largest, found in the Japanese Paris japonica (Melanthiaceae), is 148.9 (billion) Gbp (Pellicer et al., 2010). The highest genome size of any vascular plant is that of the New Caledonian fern Tmesipteris oblanceolata (Psilotaceae), with a genome size of 160 Gbp or ~50-fold that of the human genome (Fernández et al., 2024).

All five annual species of Aethionema form a terminal clade in the unpublished phylogeny by Mark Menke, and the remaining species of the genus are perennials with the most common n = x = 12. It would be interesting to have a complete chromosome number and phylogeny of the genus to pave the way for the phylogenomic of the entire genus.

There are several other important studies relevant to the systematics of the family but because of space limitations, the interested reader is advised to consult Dogan et al. (2022) on sister relationships of Idahoa and Subularia, in addition to Megadenia (Yang et al., 2020) and Eutrema (Wang et al., 2022).

ADJUSTMENTS TO PROBLEMATIC GENERA AND TRIBES

There are at least three major tribes that need substantial generic re-adjustments, all of which belong to supertribe Brassicodae. These including the Brassiceae (53 genera, 243 species), Thelypodieae (34 genera, 249 species) and Coluteocarpeae (1–14 genera, ~130 species). The Brassiceae, and less so Thelypodieae, have tremendous diversity in fruit, and Al-Shehbaz et al. (2006: 103) suggested that ‘… rapid evolutionary bursts of fruit morphology, which are likely controlled by a relatively few genes, have most likely occurred independently of other aspects of morphology, and therefore obstructed the true relationships within the tribe and led to inadequate taxonomy’. The remaining tribes are reasonably well delimited and hardly need any recircumscriptions. Although a handful of genera (e.g. Atacama, Delpinophytum, Menonvillea, Pseudoarabidopsis) were previously assigned to tribes, some of the latest thorough molecular studies (e.g. Walden et al., 2020) raised concerns about their tribal status, and these will be discussed in the end section below on Miscellaneous genera.

Tribe Brassiceae

Orychophragmus

Since its establishment in 1821, the primarily the western Mediterranean Brassiceae have been recognized as a natural tribe characterized by the conduplicate cotyledons and/or segmented fruits, and almost all members that lack fruit segmentation evolved from ancestors with segmented fruits.

The tribal placement of Orychophragmus has been controversial, and the genus was first included in the tribe (Schulz, 1923) and subsequently maintained by some authors (e.g. Gómez-Campo, 1999; Warwick and Sauder, 2005) or excluded from it (Gómez-Campo, 1980), although the controversy remained to the present. Warwick and Sauder (2005) placed it as sister to the eight species of Vella. However, Lysak et al. (2007) were the first to point out that Orychophragmus violaceus has WGD rather than WGT, which is characteristic of the rest of tribe Brassiceae (Lysak et al., 2005), and their results provided solid evidence supporting the exclusion of the genus from the tribe.

Hendriks et al. (2023) showed that Orychophragmus and sister genus Sinalliaria form a clade sister to the tribe Brassiceae. Both genera lack the segmented fruits and are restricted to eastern Asia, where no native species of the tribe grow. The morphological, geographical and molecular findings support the exclusion of this generic pair from the tribe, preferably in their own. In fact, Khosravi et al. (2009) first suggested that Orychophragmus and Conringia planisiliqua (currently in the Isatideae) should be placed in one tribe, although they did not make any nomenclatural adjustments (see Thomas et al. paper in this journal issue).

Bivonaea and Horwoodia

These monospecific genera were first included in the ITS phylogeny of the family by Warwick et al. (2010), and they showed that they form a monophyletic clade sister to the Brassiceae. The position of both genera was resolved almost two decades later. Bivonaea lutea (Algeria, Italy, Sardinia and Tunisia) was placed in its own tribe, Bivonaeeae (Koch, 2012), and was shown by Nikolov et al. (2019) and Hendriks et al. (2023) to fall in a clade sister to the Brassiceae. In the absence of genomic data showing whether it has WGD or WGT, it appears that Bivonaea, which lacks conduplicate cotyledons and segmented fruits, is not a member of the Brassiceae, thus agreeing with Koch (2012).

Horwoodia was shown in plastome phylogeny of Walden et al. (2020) to form a clade with Crambe, and both are sister to the rest of the family. The plastome phylogeny of Hendriks et al. (2023) showed Horwoodia sister to a clade consisting of a North African subclade of Schouwia and Quezeliantha and another of Zilla plus the SW Asian Physorhynchus and Douepea. Their nuclear phylogeny shows Horwoodia sister to a clade of Carrichtera plus Vella within the Brassiceae. Shawn Thomas (unpublished data) indicates that Horwoodia dicksoniae has WGT and, therefore, it is a member of the tribe. It has longitudinally folded incumbent cotyledons that are best described as subconduplicate, although it does not have segmented fruits.

Tribe Coluteocarpeae

Prior to the recognition of Coluteocarpeae as a distinct tribe, most of its species were placed primarily in Thlaspi (~75 species). The German botanist Friedrich Karl Meyer (29 December 1926–22 December 2012) depended heavily on seed-coat anatomy, and he (Meyer, 1973, 1979) split that genus into 12 segregates. He retained six species in Thlaspi, of which one is synonymized and two were transferred to Mummenhoffia (Esmailbegi et al., 2018), whereas the remaining 11 genera and 98 species currently constitute the Coluteocarpeae. Subsequent taxonomists did not accept this concept for the following two decades, but the earliest molecular studies (Mummenhoff and Zunk, 1991; Koch et al., 1993; Mummenhoff and Koch, 1994; Zunk et al., 1996), followed by other studies, supported the recognition of Meyer segregates.

Unconvinced by the above trivial features relied on to distinguish Meyer’s 11 segregates currently placed in Coluteocarpeae, Al-Shehbaz (2014) reduced them to synonymy of Noccaea as an inclusive genus of 128 species. However, there is no doubt that Meyer’s taxonomic contribution to the entire Thlaspi complex was highly significant.

Ali et al. (2016) used plastid matK and trnL-F and nuclear ITS and sampled only 21 species (11 % of the tribe), including 13 Noccaea, to study the phylogenetic relationships in the Coluteocarpeae, and recognized Microthlaspi (five species), new monospecific Friedrichkarlmeyeria and Ihsanalshehbazia, Neurotropis, Vania and Noccaea (including Callothlaspi, Masmenia, Raparia, Pseudosempervivum and Thlaspisceras). However, Microthlaspi and its two segregates cannot reliably be separated morphologically.

Özüdoğru et al. (2019) also used the ITS and trnL-F markers, and they sampled 92 species, or 66 % of the Coluteocarpeae. Their phylogenetic study showed Noccaea in nine positions, plus paraphyletic Callothlaspi (three positons), Eunomia (three), Kotschyella (two), Masmenia (two) and Vania (two), all nested among Noccaea species. and they leaned towards a more inclusive Noccaea.

None of the previous studies is adequate, because of the limited sampling, use of a handful markers and reliance on dubious characters for generic delimitations. Without an almost complete sampling of Coluteocarpeae, its generic delimitation remains a problem. The comprehensive phylogeny of the entire family, including almost all species of the Coluteocarpeae, is being conducted by Hendriks and colleagues, and it appears that most of the above segregate genera will be accepted, with Noccaea being the largest. The recognition of Noccaea as the sole genus in the tribe (Al-Shehbaz, 2014) should perhaps be abandoned.

Tribe Thelypodieae

Warwick et al. (2009) used the ITS spacer region and plastid gene ndhF to focus on the evolutionary relationships within the tribe and demonstrated that its generic resolution was highly limited in both phylogenies, and almost all genera with more than one species, especially the small and medium-sized ones, were para- or polyphyletic, especially ten North American (Chaunanthus, Caulanthus, Dryopetalon, Hesperidanthus, Sibara, Stanleya, Streptanthus, Thelypodiopsis, Thelypodium and Warea) and six South American (Dictyophragmus, Mostacillastrum, Neuontobotrys, Polypsecadium, Weberbauera and Werdermannia). Cacho et al. (2014) used three single-copy nuclear genes, three other nuclear genes (ITS, phyA and PEPC) and two chloroplast regions (trnL and trnK-psbA) to study the phylogenetic relationships among 50 species of six North American genera. Their study showed Streptanthus to form five distinct, well-defined clades, of which one included generic types of Caulanthus, Steptanthella, Streptanthus and Stanleya. Caulanthus also formed six clades, of which three species are currently placed in Guillenia and two fell within the Thelypodium clade, including its type (Thelypodium laciniatum).

The above studies clearly demonstrate that the current generic delimitation of Thelypodieae is the least satisfactory of all large tribes, and without a comprehensive study of almost all its species the existing generic delimitation will remain artificial and unstable. With the exception of monotypic or a few oligospecific genera, many genera ought to be split or combined, because the present findings show that their generic types do not fall with most species of a given genus. Undoubtedly, future comprehensive molecular studies would lead to substantial nomenclatural adjustment, especially for the genera listed above.

MISCELLANEOUS GENERA

The problematic tribal affiliations of the following genera were first compiled by Walden et al. (2020) and are discussed further here.

Atacama

The type of this monotypic genus, Atacama nivea, was previously recognized in Mathewsia (Schizopetaleae), but Toro-Núñez et al. (2015) provided evidence that it falls outside latter genus. Although Salariato et al. (2016) placed the genus in the Schizopetaleae, it formed a distinct clade sister to the two genera of the tribe, Mathewsia and Schizopetalon. In contrast, Walden et al. (2020) found it sister to the Cremolobeae rather than the Schizopetaleae. A comprehensive study involving all taxa of this group is needed to establish a tribal home for Atacama firmly.

Delpinophytum

This monotypic genus, which is endemic to Patagonian Argentina (Chubut and Santa Cruz provinces), morphologically resembles Xerodraba, another Patagonian endemic, that was placed by Al-Shehbaz (2012) in the South-American endemic tribe Eudemeae largely on morphological grounds. Both genera are pulvinate perennials, somewhat woody basally and with imbricate, scale-like leaves, solitary terminal or subterminal flowers, four lateral nectar glands, and two- or occasionally few-seeded silicles. They are indistinguishable in all but fruit characters (angustiseptate in Delpinophytum and latiseptate in Xerodraba). Molecular phylogenetic studies (Salariato et al., 2015) using ITS and three plastid regions showed that Delpinophyton did not fall within the Eudemeae, but the plastid data showed it to be sister to but not included to the tribe Lepidieae, whereas in the ITS data it formed a clade within a polytomy including Smelowskieae, Cardamine and a larger polytomous clade including 11 tribes, one of which is the Lepidieae. Although Salariato et al. (2015) suggested that Delpinophytum belongs to the Lepidieae, the discrepancy of plastid and nuclear findings does not support that. As suggested by Walden et al. (2020), the genus is the product of ancient intertribal reticulate evolution involving the Lepidieae as the maternal clade and most probably ought to be placed in its own tribe.

Menonvillea

Menonvillea (24 species) formed a monophyletic clade sister to the rest of the tribe Cemolobeae (Salariato et al., 2016), but Walden et al. (2020) clearly demonstrated that the genus is sister to the Eudemeae. As in Atacama, further studies are needed to resolve the tribal assignments, and it is possible that former reticulate evolution is behind the discrepancies of both genera.

Pseudoarabidopsis

Pseudoarabidopsis was initially assigned to the tribe Camelineae by German et al. (2009) and Warwick et al. (2010) based on ITS data, whereas the plastome data of Walden et al. (2020) assigned it to the Turritideae, and Huang et al. (2020) placed it between these two tribes. The genus has remained unassigned to the present following German et al. (2023), and little can be said without a thorough family-wide phylogeny.

UNIQUE (APOMORPHIC) CHARACTERS

As indicated above, almost all morphological characters in the family exhibit substantial homoplasy. A compilation of apomorphic characters, largely unpublished data, is presented here. Within the largest genera listed above, substantial independent character evolution is observed repeatedly. A case in point is development of winged seeds in Draba, a feature extremely rare and known thus far from only 6 of the ~410 species that fall in different clades (M. A. Koch, pers. comm.). These include the North American D. asterophora, D. carnosula and D. pterosperma, the Argentinean–Chilean D. gilliesii and the Himalayan D. radicans and D. aubrietioides.

Among the rather rare cases of homoplasy of occurrence involve the production of four-valved fruits in Tropidocarpum capparideum (Descurainieae) and four to six valves in Rorippa barbaraefolia (Cardamineae). Also, the production of pseudoflowers, where two sepals, two petals, one lateral filament and two median filaments on each side form a pseudoflower, rather than the typical cruciform flower. This floral design is known thus far only in Sterigmostemum anchonioides (Anchonieae) of SW Iran and the widespread (NW Africa into Pakistan) Anastatica hierochuntica (Anastaticeae).

There are, however, <20 features that evolved only once in the Brassicaceae. The well-known apomorphic features include dioecism in the New Zealand Lepidium sisymbrioides of the Lepidieae (Soza et al., 2014), in addition to monoecism only in Megacarpaea megalocarpa (Megacarpaeeae), a species of W China, Kazakhstan, Kyrgyzstan, Mongolia, Uzbekistan and adjacent Russia. In the latter species, the lowermost two to five flowers of the raceme are pistillate, naked (no perianth) and without stamens, whereas the remaining flowers are staminate, with perianth and sterile pistil.

Other apomorphic characters include the wind pollination in Pringlea antiscorbutica of some of the South Indian Ocean islands (Crozet and Kerguelen islands, France; Marion and Prince Edward islands, South Africa; and Heard and McDonald islands, Australia); one-lobed median anthers in Himalayan and adjacent central Asian monospecific Atelanthera (Euclideae); deciduous styles in the central Asian Litwinowia tenuisissima (Chorisporeae); unicellular glandular papilla in Descurainia; and appendaged petals in Heliophila, although they are confined to clades A and C sensu  Mummenhoff et al. (2005). The South African Chamira circaeoides (Chamireae) is unique in the family for having spurred sepals and longitudinally and marginally deeply folded persistent cotyledons that are the main photosynthetic organs in the plant.

All eight species Leavenworthia (Cardamineae) are winter annuals that grow on thin soil on outcrops of dolomitic limestone primarily in the southeastern USA. Three species have accumbent cotyledons with a very short, slightly curved radicle, whereas the remaining five have straight embryos, a feature not known elsewhere in the family. The five species form a monophyletic terminal clade in the phylogeny of Urban and Bailey (2013).

The primarily Chilean Schizopetalon (Schizopetaleae) consists of ten species, of which S. corymbosum, S. maritimum and S. walkeri have globose seeds with deeply bifid, spirally twisted cotyledons. Although globose seeds evolved independently and are common in the Brassiceae, no other species in the family have the deeply bifid and twisted cotyledons. The cladistic study of the genus (Al-Shehbaz, 1989) demonstrated that the above three species occupy a terminal clade in the genus, and subsequent molecular phylogenies (Toro-Núñez et al., 2013; Salariato et al., 2016) have fully supported that.

The genus Cardamine is well known for having elastically dehiscent fruits, and this feature is unique in the entire family. However, fruits of New Zealand endemic Cardamine lacustris lack such dehiscence (Heenan, 2002). The species has only two stamens, a feature that evolved independently only in Lepidium.

Cardamine is the only genus with palmately and bipinnately compound leaves, and Heliophila is unique in the family for having intercalary racemes restricted to only seven species (H. cedarbergensis, C. dregeana, C. esterhuyseniae, H. rimicola, H. scoparia, H. tulbagensis and H. xylopoda) of the 106 species of the genus. These seven species form a monophyletic clade in the study by Mummenhoff et al. (2005), but a comprehensive phylogeny of the genus is needed to assess the apomorphy of this character.

Megacarpaea includes nine species, of which six are central Asian and three are Himalayan. As in the vast majority of the Brassicaceae, seven species of this genus have six stamens, but Himalayan M. bifida has 7–11 stamens, and M. polyandra has (8–)12–16(–24) stamens (Al-Shehbaz, 2015). The genus has not yet been studied phylogenetically, and it is likely that this pair forms a terminal clade.

There are several examples where a feature evolved uniquely in a given genus or tribe but was subsequently lost, and a few cases are presented here. Hesperis (Hesperideae) is unique in the family for having uniseriate, multicellular glands, a feature recognized as ancestral in the genus (Eslami-Farouji et al., 2021), but they were lost independently more than once (e.g. H. matronalis, H. isatidea). Another case is the occurrence of polycolpate (4–11 colpi) pollen unique to the Physarieae, whereas the rest of the family have tricolpate pollen. The tribe consists of seven genera and 136 species, and all except six are North American (six Physaria species are disjunct in Argentina and Bolivia). Two species of Lyrocarpa (L. coulteri and L. linearifolia) have tricolpate pollen, although the former species sometimes have tetracolpate pollen. Lyrocarpa  forms a monophyletic clade in a three-genus terminal trichotomy in the tribal phylogeny (Fuentes-Soriano and Al-Shehbaz, 2013).

Geocarpy (the production of subterranean fruits) evolved independently in only six tribes and nine species, namely Aphragmeae (Aphragmus hinkuensis), Brassiceae (Morisia monanthos, Raffenaldia platycarpa and R. primuloides), Cardamineae (Cardamine chenopodiifolia), Eutremeae (Eutrema angustiseptatum and E. watsonii), Euclideae (Pycnoplinthopsis bhutanica) and Microlepidieae (Geococcus pusillus). However, amphicarpy (production of two fruit types on the same plant) evolved only once, in the South American Cardamine chenopodiifolia (Al-Shehbaz and Marhold, 2024). The species produces areal, linear, many-seeded siliques and hypogeal, obovoid-fusiform, one- to four-seeded fruits with larger seeds. The evolution of this unique feature has recently been studied by Emonet et al. (2024) and discussed earlier under Cardamine. Dimorphic production of seeds and fruits on the same plant is known also in several Aethionema species, and it has been studied recently in Aethionema arabicum by Chandler et al. (2024), an important reference that should be consulted.

Finally, 40 species of Thelypodieae are the only mustards with three stamens of unequal length, where the adaxial (dorsal) filament pair is the longest, followed by the abaxial pair, then the laterals. This feature characterizes 32 of 37 Streptanthus species, 6 of 14 Caulanthus species, and monospecific Sibaropsis and Streptanthella. Filaments of the adaxial stamen pair are united in some of the species, and the anthers are often sterile. The lack of a comprehensive phylogeny of the tribe prevents any sensible conclusions about the evolution of this unique feature.

Conflicts of interest

There are no conflicts of interest to disclose by this author.