As of early January, 2002, we have extracted
DNA from 87 specimens representing seven outgroup species
and 78 Croton species. They cover about 22 of the
40 sections treated by Webster (1993) and most of the large
sections. Taxa were selected based on several criteria:
a) a broad sampling of sections and geographical regions
of the globe, using herbarium material at Missouri Botanical
Garden, and b) material that we were able to collect fresh
or in silica gel ourselves or that we obtained from collaborators.
We successfully sequenced 64 taxa for the ITS region (Fig.
1) and 65 taxa for trnL-F (Fig. 2). Both regions are sufficiently
variable to provide high statistical support for many clades
found in the tree. ITS, with 39% of sites potentially informative,
appears better suited to resolve relationships among very
closely related species, whereas the more slowly evolving
trnL-F region, with 12% of sites potentially informative,
may be better at resolving early branching clades (Figs.
1 and 2).
The ITS and trnL-F data
sets had 58 taxa in common, and these are congruent with
each other according to the incongruence length difference
test (p=0.143; Farris et al. 1994). The combined data set
resolves many clades in the tree with bootstrap supports
well above 50% (Fig. 3). The main differences between the
ITS and cpDNA data sets are the positions of Croton setigerus
(sect. Eremocarpus) and C. michauxii (sect. Crotonopsis),
both of which fall in different positions in the cpDNA from
the ITS tree and the combined tree, where they are near
the base. It is possible that they are being forced to the
base of the tree due to long branch attraction (Felsenstein
1978), but we should be able to resolve the positions of
these morphologically distinct, small sections by adding
a more slowly evolving cpDNA region such as ndhF.
It was difficult to align outgroup taxa with ingroups using
trnL-F and ITS, both of which are composed largely of noncoding
sequences, and there were also some early branching nodes
in the single and combined trees with low bootstrap support.
This suggests that more slowly evolving coding regions will
be useful for examining relationships among major clades
within Croton and its outgroups. We have begun to
sequence the chloroplast ndhF region, and using a subset
of five divergent species from the combined data set, we
found 14% of nucleotide positions to be parsimony-informative,
with the same relationships as those obtained using trnL-F
data (not shown).
Prior
to the results reported here, we found no published results
of Croton molecular sequences, and as of Jan. 2002,
GenBank listed only two pharmaceutically related sequences
for the genus. Therefore, our initial results fill a complete
void, and they point to several interesting possibilities
that had never been suggested in previous papers on Croton
phylogeny. Some of the more interesting findings are:
1. Monophyly
of Croton. With our broad sampling of the genus to
date, our data are consistent with a large, monophyletic Croton
clade. To substantiate this, however, we will need a much
broader sampling of ingroups and outgroups. Particularly critical
are the Cuban endemic genera Moacroton and Cubacroton,
for which we have not so far been able to extract DNA from
older herbarium specimens.
2. In
both ITS
and trnL-F
trees, Croton lobatus forms the earliest branch in
Croton and is sister to all other crotons sampled.
It belongs to sect. Astraea, a very distinctive assemblage
of 10 species from southern Brazil and Paraguay, with C.
lobatus the sole widespread neotropical species. Characters
possibly unique to this group include a mixture of simple
and stellate hairs; highly divided, fan-shaped styles; a deep
palmate lobing of the leaves; cylindric-tetragonous seeds;
the staminate floral receptacle glabrate or very shortly pubescent;
and the only chromosome count in the genus so far of n = 9
(just C. lobatus examined in the section to date).
If the position of this section is confirmed with denser sampling,
it points to a New World origin for Croton.
3. Croton
alabamensis is the next lineage branching off within Croton,
which is highly supported in both data sets. This is a rare
and enigmatic species, at one time placed both in sect. Eluteria
(for the well developed female petals) and Lamprocroton (lepidote
hairs, eglandular leaves), but anomalous in both. It is also
the only species with chromosome numbers of n=16 and 32.
4. Eremocarpus
and Crotonopsis are incongruent in their positions
in the two data sets, but basal in the combined
tree (discussed above). It will be important
to resolve the position of these morphologically distinct
sections using additional sequence data (both have unilocular
fruits, the first with the perianth completely absent in both
sexes).
5. Support
for a series of New World sections. Pairs of species sampled
from six different New World sections were strongly supported
as being monophyletic, as was the sectional assemblage of
Pilinophytum + Julocroton + Argyroglossum + Lasiogyne. Some
sections show strong morphological and chromosomal synapomorphies
and geographically defined ranges, such as Drepadenium, which
has ten species centered in the southern U.S., Caribbean,
and Mexico, petals absent in both staminate and pistillate
flowers, and chromosome numbers of n = 14 and 28. Eluteria
is similarly well-defined (though no chromosome counts yet),
and is the only section with axillary inflorescences.
6. A
single Old World clade. This is probably the most significant
finding if borne out by further sampling. In separate and
combined
analyses, the 15 OW species always group together,
despite divergent sectional assignments. A late addition after
our figures were completed is C. gratissimus, an African
species which nests well within the OW clade in the trnL-F
tree. These results contradict the bi- or tricontinental circumscriptions
of several of Webster?s sections, and suggests the evolutionary
scenario that all or most Old World Crotons are nested together
within a New World grade. This is where Webster's sections
are least informative, and further work will likely produce
a completely different taxonomic arrangement for the Old World
taxa.
7. A
robust core Cyclostigma clade. There is very strong support
for a core Cyclostigma clade. These are New World, mostly
tree species, including the medicinal "sangre de dragos."
The few members of this section that have been studied cytologically
show uniquely high ploidy levels in the genus (n=32, 64).
8. A
'New World Cascarilla' clade that includes some members previously
assigned to section Cyclostigma. The bulk of New World species
are assigned to a weakly defined section Cascarilla. There
is strong support here for at least a core Cascarilla clade,
and this would include at least some members previously assigned
to section Cyclostigma. We plan a denser sampling and reevaluation
of the morphological features of these sections.
9. Herbarium
material amplifies well for recent collections. Of the 86
specimens extracted, 53 came from leaves collected fresh or
in silica gel, and 33 from herbarium specimens. The data presented
reflect a successful amplification rate of 85% for fresh &
silica-dried material, and 61% for herbarium material. Additional
extractions and amplification protocols are being tried for
samples that did not amplify. We were not able to amplify
DNA from the available specimens of Moacroton, which
were >50 years old.
|
