Systematics and phylogeny of the Brassicaceae
(Cruciferae): an overview
I. A. Al-Shehbaz
1
, M. A. Beilstein
2
, and E. A. Kellogg
2
1
Missouri Botanical Garden, St. Louis, Missouri, USA
2
Department of Biology, University of Missouri-St. Louis, St. Louis, Missouri, USA
Received November 9, 2005; accepted December 13, 2005
Published online: June 19, 2006
Springer-Verlag 2006
Abstract. A critical review of characters used in the
systematics of the Brassicaceae is given, and aspects
of the origin, classification, and generi c delimita-
tion of the family discussed. Molecular phyloge-
netic studies of the family were reviewed, and major
clades identified. Based on molecular studies, espe-
cially from the ndhF chloroplast gene, and careful
evaluation of morphology and generic circumscrip-
tions, a new tribal alignment of the Brassicaceae is
proposed. In all, 25 tribes are recognized, of which
seven (Aethionemeae, Boec hereae, Descurainieae,
Eutremeae, Halimolobeae, Noccaeeae, and Sme-
lowskieae) are described as new. For each tribe, the
center(s) of distribution, morphology, and number
of taxa are given. Of the 338 genera currently
recognized in the Brassicaceae, about 260 genera
(or about 77%) were either assigned or tentatively
assigned to the 25 tribes. Some problems relating to
various genera and tribes are discussed, and future
research developments are briefly covered.
Key words: Brassicaceae, characters, origin,
classification, generic circumscription, molecular
data, major clades, new tribal alignments.
The Brassicaceae (Cruciferae), or mustard
family, is a monophyletic group of about 338
genera and some 3709 species distributed
worldwide. It includes many economically
important ornamental and crop species (veg-
etables or sources of industrial and cooking
oils, forage, and condiments), but is perhaps
best known for thale cress, Arabidopsis
thaliana (L.) Heynh., the model organism of
flowering plants. The latter species is cur-
rently used in almost every discipline of
experimental biology and its completely
sequenced genome (The Arabidopsis Genome
Initiative, 2000) paved the way to a better
understanding of every aspect of plant
biology.
The morphological diversity, systems of
classification, earlier literature, endemism, and
distribution of the family are discussed in
Hedge (1976), Al-Shehbaz (1984), and Appel
and Al-Shehbaz (2003). These aspects will not
be repeated here, and the interested reader
should consult those works and the references
cited therein. For extensive reviews on the
molecular phylogenetic studies of the family,
Koch (2003), Koch et al. (2003a), Mitchell-
Olds et al. (2005), and Beilstein et al. (2006)
should be checked.
The present paper addresses the following
major aspects of the family: the evaluation of
characters and their utilization in infrafamilial
classifications, delimitation of genera, molecu-
lar data and major subdivisions of the family,
Pl. Syst. Evol. 259: 89–120 (2006)
DOI 10.1007/s00606-006-0415-z
problematic taxa, and future challenges of
research.
Characters
Numerous studies (e.g. Al-Shehbaz 1984, Price
et al. 1994, Appel and Al-Shehbaz 2003, Koch
et al. 2003a, Mitchell-Olds et al. 2005) have
amply demonstrated that morphological char-
acters in the Brassicaceae are highly homoplas-
ious, making it virtually impossible to utilize
them alone in establishing phylogenetic rela-
tionships on a family-wide basis or sometimes
even within genera (Mummenhoff et al. 1997a).
The lack of a robust phylogeny of the family led
some recent authors (e.g. Rollins 1993, Appel
and Al-Shehbaz 2003) to adopt an alphabetical
system in their enumeration of taxa.
Fruit morphology and seed embryo type
(position of the radicle in relation to cotyle-
dons) have been used almost exclusively in the
delimitation of taxa at all taxonomic levels, but
particularly at generic and tribal ranks, while
floral, vegetative, and trichome characters
have often been considered far less significant.
However, as shown below, fruit and embryo
features can be subject to considerable conver-
gence and therefore are sometimes taxonomi-
cally unreliable. For example, diplecolobal
cotyledons (i.e. with incumbent cotyledons
folded two or more times), thought to be
unique to the tribe Heliophileae (Schulz 1936),
evolved independently in Lepidium L. s.l.
(Hewson 1981, Mummenhoff et al. 2001a).
From that type, spiral cotyledons probably
evolved in Brachycarpaea DC. (Appel and Al-
Shehbaz 1997, Mummenhoff et al. 2005), a
genus recently reduced to synonymy of Helio-
phila L. (Al-Shehbaz and Mummenhoff 2005).
Incumbent and accumbent cotyledons, the
most common cotyledonary types in the fam-
ily, are the least reliable because they occur
within numerous genera, including Arabidopsis
(DC.) Heynh. and Erysimum L. Unfortu-
nately, little is known about the genetic control
of cotyledonary position, and A. thaliana
would be the ideal species to study the evolu-
tion of this character.
Although the Brassicaceae was once
thought to be exclusively stenopalynous (i.e.
with uniform pollen) with only tricolpate
pollen (Erdtman 1971), preliminary surveys
(e.g. Rollins and Banerjee 1979) demonstrated
the existence in the New World of several
genera with 4–11-colpate pollen. This group
with ‘‘polycolpate’’ pollen has subsequently
been shown by O’Kane and Al-Shehbaz (2003)
to form a monophyletic clade. However, a
more comprehensive palynological survey of
the family is needed to determine the utility of
pollen in taxonomic and phylogenetic studies.
In fact, pollen data were shown to be useful in
the separation of putatively closely related
genera (Rollins 1979, Al-Shehbaz 1989).
Despite the conservative floral architecture
of the family, there can be enormous variation
among related groups or even within genera.
For example, Lepidium, Heliophila, Alyssum
L., and Streptanthus Nutt. all exhibit tremen-
dous floral diversity that is quite useful in
defining lineages and assessing relationships
(Mummenhoff et al. 2001a, 2005). Many other
genera (e.g. Stenopetalum R.Br., Schizopetalon
Sims, Ornithocarpa Rose, Stanleya Nutt.,
Warea Nutt., Iberis L., to name a few) are
readily recognized by their flowers and evi-
dently are monophyletic. However, little atten-
tion has been given to the value of floral
morphology in establishing monophyletic
groups in the vast majority of the family.
Finally, trichome morphology, first
emphasized by Prantl (1891) but utilized only
a little in subsequent studies (e.g. Rollins and
Banerjee 1975, 1976), appears to be far more
useful in the separation of closely related
genera (e.g. Al-Shehbaz et al. 1999) and holds
significant promise in the delimitation of
monophyletic groups. Although both simple
and branched trichomes can be found in most
major clades of the family (Beilstein et al.
2006, Bailey pers. com.), the trichome subtypes
can be far more valuable. The extensive studies
on trichome development in Arabidopsis thali-
ana (e.g. Schwab et al. 2000, Hu
¨
lskamp 2000,
Szymanski et al. 2000, Beilstein and Szyman-
ski 2004) have identified a significant number
90 I. A. Al-Shehbaz et al.: Systematics and phylogeny of the Brassicaceae
of genes (e.g. ANGUSTIFOLIA, STICHEL,
ZWICHEL) responsible for the genetic path-
ways that control trichome morphology. How-
ever, sequence comparisons from such genes
are not available for other genera of the
family. It is not yet known if sequence data
from such genes are useful in phylogenetic
studies in the family.
Origin and classification
Hayek (1911), followed by Schulz (1936) and
Janchen (1942), provided the first theory on
the origin of the Brassicaceae. They believed in
a New World origin of the family from the
capparaceous subfamily Cleomoideae through
the ‘‘basal’’ mustard tribe Theylopodieae
(Stanleyeae). Views of this German school
were followed by Al-Shehbaz (1973, 1985a),
Hauser and Crovello (1982), and Takhtajan
(1997). Indeed, Nuttall (1834) proposed the
name Stanleyeae as a distinct family interme-
diate between the Capparaceae and Cruciferae.
It included Stanleya Nutt. and Warea Nutt. By
contrast, Dvora
´
k (1973) proposed an Old
World origin from the Cleomaceae via the
tribe Hesperideae, but his views were not
subsequently followed.
Molecular studies (Hall et al. 2002, Koch
et al. 2003a and references therein, Mitchell-
Olds et al. 2005, Beilstein et al. 2006) have
clearly demonstrated that the Brassicaceae
evolved in the Old World and is sister to the
Cleomaceae, that Aethionema R.Br. is the most
‘‘basal’’ genus in the family, that the Thel-
ypodieae (hereafter Schizopetaleae; see below)
is rather advanced, and that the remarkable
superficial floral and fruit similarities (e.g.
exserted stamens equal in length, linear anthers
coiled at dehiscence, dense racemes, linear
fruits, sessile stigmas, long gynophores) be-
tween members of this tribe, especially Stan-
leya and Warea (first discovered species of
each was originally described as Cleome L.)
and the Cleomaceae evolved through conver-
gence. A closer comparison of Aethionema
with some members of the Cleomaceae is given
in the section on major clades of the family.
Although Schulz’s (1936) classification
of the Brassicaceae has been modified and
criticized (e.g. Janchen 1942; Al-Shehbaz 1973,
1984), it continued to be the most widely used
to the present. However, all of his new
suprageneric taxa that first appeared in that
work are invalidly published (see Greuter et al.
2000). Regardless of the number of infrafamil-
ial taxa recognized, his and the earlier systems
of Prantl (1891) and Hayek (1911) utilized a
limited number of characters, many of which
have recently been shown to be subject to
convergence. As a result, almost all of their
major subdivisions of the family have been
shown by molecular studies to be polyphyletic
and artificial (Price et al. 1994; Warwick and
Black 1994, 1997a, 1997b; Koch et al. 1999a,
2000, 2001, 2003a; Zunk et al. 1999; Bailey
et al. 2002; O’Kane and Al-Shehbaz 2003;
Mitchell-Olds et al. 2005; Warwick and Sauder
2005; Beilstein et al. 2006). A classic example
of the artificiality of previous classifications is
illustrated by Arabidopsis, Capsella Medik.,
Neslia Desv., and Arabis pendula L. (now
Catolobus (C. A. Mey.) Al-Shehbaz). Schulz
(1936) placed these taxa in the tribes Sisym-
brieae, Lepidieae, Euclidieae, and Arabideae,
respectively, but molecular data (Koch et al.
2001, O’Kane and Al-Shehbaz 2003, Beilstein
et al. 2006) clearly demonstrated that the four
genera are very closely related and therefore
should be placed in the same tribe/clade (see
below).
Generic delimitations
There is a considerable lack of agreement
among authors of the past century regarding
the number of genera in the family (Table 1).
Of those currently recognized, 225 genera
(66% of the total) are either monotypic or
oligotypic (with 2–4 spp.). The vast majority of
these monotypic and oligotypic genera are
nested within, and should be united with, the
larger ones (see below).
Comparative sequence data of rapidly
evolving DNA regions of the chloroplast
(e.g. ndhF gene) and nucleus (e.g. ITS)
I. A. Al-Shehbaz et al.: Systematics and phylogeny of the Brassicaceae 91
indicate that many taxa show remarkable
sequence similarities but drastically different
fruit morphologies and embryo positions
(Warwick and Black 1994; Crespo et al.
2000; Beilstein et al. 2006; Mummenhoff
et al. 2001a, 2005). Such data suggest that
major changes in fruit morphology can occur
rather rapidly and independent of other
characters. As shown in Arabidopsis (see
below), a relatively small number of genes
are responsible for significant alterations in
fruit shape, and it is quite likely that the
same holds for the rest of the family.
Therefore, drastic bursts of fruit evolution
can rapidly take place that are independent
of molecular markers or other aspects of
morphology. In cases like these, differences
in fruit morphology would result in errone-
ous classifications or generic delimitations
and would obscure phylogenetic relation-
ships.
Fruit development in Arabidopsis thaliana
has been reasonably well studied, and a few
genes (e.g. FRUITFUL, MADS-box, SHAT-
TERPROOF) are known to alter fruit shape
(i.e. length/width ratio; silique vs. silicle)
and dehiscence (Ferrandiz et al. 1999, 2000;
Ferrandiz 2002; Dinneny and Yanofsky 2004).
At least six genes have been identified that
control fruit dehiscence in this species, and as
few as one or double mutant genes can be the
difference between dehiscent and indehiscent
fruits (Liljegren et al. 2000, 2004; Dinneny and
Yanofsky 2004; Polster 2005). These findings
should caution the use of fruit dehiscence vs.
indehiscence as the main criterion for the
delimitation of genera. In fact, the only
difference that distinguished Cardaria Desv.
from Lepidium and Boleum Desv. from Vella
L. is having dehiscent instead of indehiscent
fruits. Cardaria is nested within Lepidium
(Mummenhoff 1995, Mummenhoff et al.
2001a) and Boleum within Vella (Warwick
and Black 1994, Crespo et al. 2000). There-
fore, the reduction of Cardaria to synonymy of
Lepidium (Al-Shehbaz et al. 2002) and Boleum
to that of Vella (Warwick and Al-Shehbaz
1998) are fully justified.
Most of the smaller genera can easily be
accommodated within larger ones if the mor-
phological variation of their fruit characters
are neither overemphasized nor used as the
sole basis for generic delimitation. In fact,
molecular studies provide ample support to
that view. A classic example is the reduction of
the South African Brachycarpaea (1 sp.),
Cycloptychis E. Mey. (2 spp.), Schlechteria
Bolus (1 sp.), Silicularia Compton (1 sp.), and
Thlaspeocarpa C. A. Sm. (2 spp.) to synonymy
of the larger Heliophila (previously 73 spp.), a
genus within which all these smaller genera are
nested (Al-Shehbaz and Mummenhoff 2005,
Mummenhoff et al. 2005). Other examples are
the reduction of Twisselmannia Al-Shehbaz
and Agallis Phil. to synonymy of Tropidocar-
pum Hook. (Al-Shehbaz and Price 2001, Al-
Shehbaz 2003a) and Euzomodendron Coss. and
Boleum to synonymy of Vella (Warwick and
Al-Shehbaz 1998).
Prior to the utilization of molecular data in
phylogenetic studies, some of the larger genera
(e.g. Arabidopsis, Arabis L., Sisymbrium L.,
Thlaspi L.) were considered to be natural,
based on their fruit morphologies. However, it
soon became evident that none of them are,
and each had to be split into several segregates.
For example, Arabis once included about 180
species worldwide (Al-Shehbaz 1988a), of
Table 1. Enumeration of genera recognized by various authors
Hayek Schulz Authors Appel and Al-Shehbaz Warwick et al.
[1911] [1936] To [1984] [2003] [2006]
Total genera 231 351 369 337 338
Synonyms 11 21 88 59 20
Added genera 34 141 106 27 21
92 I. A. Al-Shehbaz et al.: Systematics and phylogeny of the Brassicaceae
which 80 occur in North America (Rollins
1993). The genus was delimited solely on the
presence of linear latiseptate fruits (flattened
parallel to the septum), accumbent cotyledons,
and branched trichomes. As shown by Al-
Shehbaz (2003b), this character combination
evolved independently in at least 25 genera of
the family. Molecular data (Koch et al. 1999a,
2000; O’Kane and Al-Shehbaz 2003; Beilstein
et al. 2006) have clearly shown that the ten
segregates of Arabis currently recognized (Al-
Shehbaz 2003b, 2005) are unrelated to each
other and to Arabis s.str., though they are
remarkably similar in fruit morphology and
cotyledonary position.
As for Arabidopsis, the approximately 60
species previously assigned to the genus are
now placed in several segregate genera (Al-
Shehbaz et al. 1999, Al-Shehbaz and O’Kane
2002a), and as presently delimited, the genus
consists of only 11 species (O’Kane and Al-
Shehbaz 1997, Warwick et al. 2005b). Simi-
larly, Sisymbrium was once thought to consist
of 90 species distributed both in the Old and
New Worlds (Al-Shehbaz 1988b), but molec-
ular studies (Warwick et al. 2002, 2005a) have
shown that the genus consists of about 40
species, all except one of which (S. linifolium
Nutt.) are restricted to the Old World. The
New World taxa previously assigned to Sis-
ymbrium belong to an entirely different and
morphologically distinct clade recognized by
Warwick and Al-Shehbaz (2003) as the Thel-
ypodieae alliance and herein at the tribal rank.
Further molecular studies are needed to estab-
lish generic reassignments of the New World
species of Sisymbrium. For these reasons,
Sisymbrium still includes about 90 species in
the checklist accompanying this issue (War-
wick et al. 2006).
Earlier studies maintain Thlaspi as a large
genus of over 80 species, and Meyer’s (1973,
1979) 12 segregates, which were based largely
on seed-coat anatomy, were not recognized
(Hedge 1976, Al-Shehbaz 1986). The genus
was delimited solely on its angustiseptate fruits
(flattened at a right angle to the septum) with
four or more seeds. However, molecular data
(Mummenhoff and Koch 1994; Zunk et al.
1996; Mummenhoff et al. 1997a, b; Koch and
Mummenhoff 2001) strongly support the
recognition of a few (Callothlaspi F.K. Mey.,
Microthlaspi F.K. Mey., Noccaea Moench,
Noccidium F.K. Mey., and Vania F.K. Mey.)
of Meyer’s segregates. The data also show that
the apparent similarities in fruit morphology
are the result of convergence. As delimited by
Meyer (2001)Thlaspi s.str. consists of only six
species, and it is most closely related to Alliaria
Heist. ex Fabr. than to the remaining species
previously assigned to it.
To conclude, three important points need
emphasis regarding the delimitation of genera.
First, monotypic or oligotypic genera should
not be established without prior molecular
studies and critical evaluation of morphology.
Second, because of widespread convergence in
most morphological characters, especially fruit
types and embryo position, these characters
should be used with extreme care in establishing
generic boundaries. Finally, major differences in
fruit morphology can be misleading, and the
examples of Heliophila, Tropidocarpum, and
Vella should be a constant reminder about the
dangers of making erroneous taxonomic con-
clusions by overemphasizing fruit morphology
at the expense of other, potentially very useful
vegetative and floral characters.
Molecular data
Numerous phylogenetic or genetic studies of
the Brassicaceae have used chloroplast restric-
tion fragment length polymorphisms (RFLPs),
restriction site variation of cpDNA, or ampli-
fied fragment length polymorphisms (AFLP)
fingerprinting. Although such studies provided
a wealth of information and were consulted,
they were not included in the present survey.
However, the interested reader should consult
Koch et al. (2003a) and Koch (2003) for a
complete coverage on those and other markers.
Sequence comparisons of the internal tran-
scribed spacers of the nuclear ribosomal DNA
and the 5.8S rRNA gene (collectively, the ITS
region) have been the most frequently used
I. A. Al-Shehbaz et al.: Systematics and phylogeny of the Brassicaceae 93
