70
T.J. Givnish
the other Hawaiian lobeliads, Brighamia has white to yellow flowers with straight,
narrow tubular corollas, which are thought to have been pollinated by hawkmoths
based on their nectar chemistry.
Based on morphology, the fleshy-fruited genera are usually seen as the product
of a single colonization event, while the capsular taxa have been thought to be the
products of two to four additional colonizations (Rock 1919; Wimmer 1953;
Mabberley 1974, 1975; Lammers 1989, 1990; but see Givnish et a!. 1995). Here I
will present a molecular phylogeny to show that the Hawaiian lobeliads are instead
the product of a single immigration event from elsewhere in the Pacific basin or the
Southern Hemisphere, and that this ancestor appears to have been woody, winddispersed, bird-pollinated, and adapted to cool, rainy, open, high-elevation habitats.
2.1 Molecular Systematics
Knox et a!. (1993) used restriction-site variation and re-arrangements of the chloroplast genome to infer relationships among Sclerotheca and a sampling of species
in the large, cosmopolitan genus Lobelia. They demonstrated that herbaceous diploids like L. cardinalis are basal within Lobelia, and that woody, montane, mostly
tetraploid species from various parts of the tropics form a monophyletic clade. This
suggested that L. cardinalis would be a suitable outgroup for a cladistic analysis of
relationships among the Hawaiian lobeliads and their putative relatives.
Givnish et a!. (1994, 1995, and unpub!' data) analyzed relationships within and
among the Hawaiian lobeliad genera based on cpDNA restriction site variation,
using global parsimony and five outgroups. Virtually all of the extant Hawaiian
taxa were included in this survey. Our strict consensus tree (Fig. 1) shows that (1)
each endemic genus and each of the endemic sections of Lobelia are monophyletic;
(2) Cyanea is sister to Clermontia; (3) the bizarre cliff succulent Brighamia with
hawkmoth-pollinated flowers and capsular fruits is sister to Delissea, with birdpollinated flowers and fleshy fruits; (4) Cyanea-Clermontia is sister to BrighamiaDelissea; (5) Lobelia sect. Revolutella is sister to the latter group; and (6) Lobelia
sect. Galeatella is sister to Trematolobelia, forming a clade at the base of the Hawaiian lobeliads. This analysis is consistent with the Hawaiian lobeliads being
monophyletic, but lacks many groups of putative relatives.
My colleagues and I therefore sequenced several rapidly evolving spacer regions of the chloroplast genome (atpB-rbcL, trnL-trnF, psbA-trnH), as well as the
internal transcribed spacer (ITS) and 5.8S ribosomal subunit of nuclear ribosomal
DNA for representatives of each of the Hawaiian genera/sections, as well as for
representatives of several groups which have been proposed as potential relatives of
the Hawaiian lobeliads (Table 1). Our cladistic analysis of these data employed
global parsimony and a 1.4:1.0 weighting of transitions to transvcrsions in PAUP*
(kindly provided by D. Swofford), using L. cardinalis as a super-outgroup. The
transition:transversion ratio was estimated as 0.696 (= 1/1.437) using a maximum
likelihood approach; a total of 481 informative characters were detected. We ob-
T.J. Givnish
the other Hawaiian lobeliads, Brighamia has white to yellow flowers with straight,
narrow tubular corollas, which are thought to have been pollinated by hawkmoths
based on their nectar chemistry.
Based on morphology, the fleshy-fruited genera are usually seen as the product
of a single colonization event, while the capsular taxa have been thought to be the
products of two to four additional colonizations (Rock 1919; Wimmer 1953;
Mabberley 1974, 1975; Lammers 1989, 1990; but see Givnish et a!. 1995). Here I
will present a molecular phylogeny to show that the Hawaiian lobeliads are instead
the product of a single immigration event from elsewhere in the Pacific basin or the
Southern Hemisphere, and that this ancestor appears to have been woody, winddispersed, bird-pollinated, and adapted to cool, rainy, open, high-elevation habitats.
2.1 Molecular Systematics
Knox et a!. (1993) used restriction-site variation and re-arrangements of the chloroplast genome to infer relationships among Sclerotheca and a sampling of species
in the large, cosmopolitan genus Lobelia. They demonstrated that herbaceous diploids like L. cardinalis are basal within Lobelia, and that woody, montane, mostly
tetraploid species from various parts of the tropics form a monophyletic clade. This
suggested that L. cardinalis would be a suitable outgroup for a cladistic analysis of
relationships among the Hawaiian lobeliads and their putative relatives.
Givnish et a!. (1994, 1995, and unpub!' data) analyzed relationships within and
among the Hawaiian lobeliad genera based on cpDNA restriction site variation,
using global parsimony and five outgroups. Virtually all of the extant Hawaiian
taxa were included in this survey. Our strict consensus tree (Fig. 1) shows that (1)
each endemic genus and each of the endemic sections of Lobelia are monophyletic;
(2) Cyanea is sister to Clermontia; (3) the bizarre cliff succulent Brighamia with
hawkmoth-pollinated flowers and capsular fruits is sister to Delissea, with birdpollinated flowers and fleshy fruits; (4) Cyanea-Clermontia is sister to BrighamiaDelissea; (5) Lobelia sect. Revolutella is sister to the latter group; and (6) Lobelia
sect. Galeatella is sister to Trematolobelia, forming a clade at the base of the Hawaiian lobeliads. This analysis is consistent with the Hawaiian lobeliads being
monophyletic, but lacks many groups of putative relatives.
My colleagues and I therefore sequenced several rapidly evolving spacer regions of the chloroplast genome (atpB-rbcL, trnL-trnF, psbA-trnH), as well as the
internal transcribed spacer (ITS) and 5.8S ribosomal subunit of nuclear ribosomal
DNA for representatives of each of the Hawaiian genera/sections, as well as for
representatives of several groups which have been proposed as potential relatives of
the Hawaiian lobeliads (Table 1). Our cladistic analysis of these data employed
global parsimony and a 1.4:1.0 weighting of transitions to transvcrsions in PAUP*
(kindly provided by D. Swofford), using L. cardinalis as a super-outgroup. The
transition:transversion ratio was estimated as 0.696 (= 1/1.437) using a maximum
likelihood approach; a total of 481 informative characters were detected. We ob-