LANKESTERIANA


The subtribe Zygopetalinae comprises about 418
species (Royal Botanic Gardens, Kew, 2003) of
Neotropical orchids with diverse vegetative and floral
morphologies. Zygopetalinae possess four superposed
pollinia; in most taxa, the pollinia are flattened and the
stigma is transversely narrow and slit-like. Traditionally,
Zygopetalinae have been placed in tribe Maxillarieae;
however, as molecular data indicate that Maxillarieae
sensu Whitten et al. (2000) is sister to a paraphyletic
grade of cymbidioid taxa, Chase et al. (2003) lumped
Maxillarieae together with Cymbidiinae, Eulophiinae,
Bromheadiinae, and Catasetinae to create a broader and
monophyletic Cymbidieae. Regardless of taxonomic
rank, the generic relationships within Cymbidieae are
becoming clarified by molecular systematic studies.

Several classifications of Maxillarieae were produced in
the past decade: Senghas and Dietrich (1992), Dressler
(1993), Szlachetko (1995), Whitten et al. (2000), and
Chase et al. (2003). The three earlier classifications
(based on morphology) disagree on circumscriptions
of Zygopetalinae; Dressler (1993) proposed a broad
Zygopetalinae containing several informal alliances,
whereas Szlachetko (1995) divided these taxa among
six subtribes. The combined molecular analysis of
Maxillarieae (Whitten et al. 2000) indicated high
bootstrap support for a monophyletic Zygopetalinae
and supported the inclusion of two morphologically
anomalous genera within Zygopetalinae: Cryptarrhena
(4 species) andDichaea(ca. 111 species). Zygopetalinae
sensu Dressler(1993)has been further divided by various
authors, formally or informally, into groups based upon
several characters, especially: 1) the presence/absence,
size, and number of intemodes of pseudobulbs; 2)
the number of flowers per inflorescence; and 3) leaf
vemation revolutee or conduplicate).

We examine relationships within Zygopetalinae using
cladistic and Bayesian analyses of combined molecular
data sets of internal transcribed spacers 1 and 2 (nuclear
ribosomal DNA; hereafter referred to as ITS), of the
plastid trnL intron and trnL-F spacer (hereafter referred
to as trnL-F), and of the plastid gene matK. Our
sampling of taxa is more complete for the Huntleya
clade (one flower/inflorescence; pseudobulbs small or
lacking; conduplicate leaves), and our discussion will
focus on this clade.


Rudolfiella (Maxillariinae) were used as outgroups based
on the combined analyses of Maxillarieae (Whitten
et al. 2000). Protocols for extraction, amplification,
primers used, and sequencing are presented in Whitten
et al. (2000). Sequences were aligned manually
using Se-Al (Rambaut 1996). The aligned data
matrices are available from the authors (WMW) and
as a PopSet in GenBank. All cladistic analyses were
performed using PAUP* version 4.0b (Swofford 1999).
Bayesian analyses were performed using MrBayes
3.0 (Huelsenbeck & Ronquist 2003). The data matrix
consisted of 105 individuals (two outgroups; 99 species
plus six duplicates).

Search strategies Each matrix (ITS, trnL-F, matK,
and the combined ITS/trnL-F/matK) was subjected
to 1000 replicates of random taxon entry additions,
MULTREES on, using sub-tree pruning and re-grafting
(SPR) swapping, but saving only five trees per replicate
to minimize time spent swapping on suboptimal islands.
The resulting shortest trees were swapped to completion
or until 20 000 trees were saved. Confidence limits for
trees were assessed by performing 1000 replicates of
bootstrapping (Felsenstein 1985) using equal weighting,
SPR swapping, MULTREES on, and holding only five
trees per replicate. We assessed congruence of the
separate data sets by visual inspection of the individual
bootstrap consensus trees. We considered the bootstrap
trees to be incongruent only if they displayed "hard"
(i.e., highly supported) incongruence, rather than
"soft" (poorly supported) incongruence (Seelanan et al.
1997, Wiens 1998). We use the following descriptions
for categories of bootstrap support: weak, 50-74%;
moderate, 75-84%; strong 85-100%. We consider
percentages less than 50% to be unsupported because
such groups do not occur in the majority of the trees.
Bayesian analyses were performed on the combined
data set only using MrBayes 3.0 (Huelsenbeck &
Ronquist 2003). The parameters for the Bayesian
analysis were as follows: Iset nst=2; rates=gamma; set
autoclose=yes; mcmcp ngen=2,000,000; printfreq=l 00;
samplefreq=10; nchains=4; savebrlens=yes; mcmc;
sumt; bumin=200,000 contype=halfcompat. The
first 10000 trees were omitted and the majority rule
consensus tree was obtained in PAUP* from the
remaining trees.


RESULTS


MATERIALS AND METHODS


Species examined, voucher information, and GenBank
accession numbers are listed in Table 1. Maxillaria and


Table 2 presents the number of included aligned
positions in the matrix, the number of variable sites, the
number and percentage of phylogenetically informative


Vol. 5, N 2