Critical analysis of the topology and rooting of the parabasalian 16S rRNA tree.

Abstract

The morphological classification of the protozoan phylum Parabasala is not in absolute agreement with the 16S rRNA phylogeny. However, there are strong indications that tree-construction artifacts play a considerable role in the shaping of the 16S rRNA tree. We have performed rigorous analyses designed to minimize such artifacts using the slow-fast and taxa-exclusion methods. The analyses, which included new sequences from the genera Monocercomonas and Hexamastix, in most respects confirmed the previously suggested tree topology and polyphyly of Hypermastigida and Monocercomonadidae but detected one artificial cluster of long branches (Trichonymphidae, Pseudotrichonymphidae, Hexamastix, and Tricercomitus). They also indicated that the rooting of the phylum on the trichonymphid branch is probably wrong and that reliable rooting on the basis of current data is likely impossible. We discuss the tree topology in the view of anagenesis of cytoskeletal and motility organelles and suggest that a robust taxonomic revision requires extensive analysis of other gene sequences.

Full text

Page 1

Critical analysis of the topology and rooting of the parabasalian
16S rRNA treeq
Vladim? ır Hampl,*Ivan Cepicka, Jaroslav Flegr, Jan Tachezy, and Jaroslav Kulda
Department of Parasitology, Faculty of Science, Charles University, Prague, Czech Republic
Received 13 August 2003; revised 1 March 2004
Available online
Abstract
The morphological classification of the protozoan phylum Parabasala is not in absolute agreement with the 16S rRNA phy-
logeny. However, there are strong indications that tree-construction artifacts play a considerable role in the shaping of the 16S
rRNA tree. We have performed rigorous analyses designed to minimize such artifacts using the slow–fast and taxa-exclusion
methods. The analyses, which included new sequences from the genera Monocercomonas and Hexamastix, in most respects con-
firmed the previously suggested tree topology and polyphyly of Hypermastigida and Monocercomonadidae but detected one
artificial cluster of long branches (Trichonymphidae, Pseudotrichonymphidae, Hexamastix, and Tricercomitus). They also indicated
that the rooting of the phylum on the trichonymphid branch is probably wrong and that reliable rooting on the basis of current data
is likely impossible. We discuss the tree topology in the view of anagenesis of cytoskeletal and motility organelles and suggest that a
robust taxonomic revision requires extensive analysis of other gene sequences.
? 2004 Elsevier Inc. All rights reserved.
Keywords: Parabasala; Phylogeny; 16S rRNA; Long-branch attraction; Slow–fast method; Taxa-exclusion method; Classification; Anagenesis;
Hypermastigida; Trichomonadida; Monocercomonadidae; Monocercomonas; Hexamastix
1. Introduction
The phylum Parabasala is comprised of anaerobic
amitochondriate flagellates. Characteristic features of
the phylum are: parabasal apparatus (Golgi complex
associated with parabasal fibers), presence of a double
membrane bounded organelle named the hydrogeno-
some, and cell division by semiopen pleuromitosis with
extranuclear spindle. The vast majority of parabasalid
species live endobiotically either as harmless intestinal
commensals of various animal hosts, or as intestinal
symbionts (mutualists) in termites and wood-eating
cockroaches. The pathogenic parasites represent a tiny
part of the parabasalian species diversity, however, some
of them such as—Trichomonas vaginalis, Tritrichomonas
foetus, and Histomonas meleagridis—are of considerable
medical or veterinary importance. There are only four
known free-living species in this phylum—Pseudotricho-
monas keilini, Ditrichomonas honigbergii, Monotricho-
monas carabina, and Monotrichomonas sp. These species
live in anoxic habitats in salt or fresh water.
Traditionally, the phylum is divided into two orders,
Trichomonadida and Hypermastigida (Corliss, 1994).
The order Hypermastigida typically comprises large
forms (hundreds of micrometers long) equipped with
many flagella. Although this order covers a significant
part of the morphological and species diversity of Pa-
rabasala, all its members live exclusively as intestinal
symbionts of insects. The typical representatives of the
second order Trichomonadida have smaller cells (not
longer than 20lm) with up to six flagella, excepting the
polymonad family Calonymphidae. The order Tricho-
monadida encompasses the whole ecological diversity of
Parabasala, including free-living, endosymbiotic, com-
mensal, and pathogenic species.
Typical members of the order Trichomonadida are
classified into the family Trichomonadidae. Their char-
qSupplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.ympev.2004.03.005.
*Corresponding author.
E-mail address: vlada@natur.cuni.cz (V. Hampl).
1055-7903/$ - see front matter ? 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2004.03.005
Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
MOLECULAR
PHYLOGENETICS
AND
EVOLUTION
www.elsevier.com/locate/ympev
ARTICLE IN PRESS

Page 2

acteristic feature is the presence of an undulating
membrane and costa. The undulating membrane, an
important motility organelle of trichomonads, is formed
by recurrent flagellum adhering to a fold of cytoplasmic
membrane. The costa is a striated root fiber attached to
the basal bodies complex that underlies the undulating
membrane, presumably providing its mechanical sup-
port (Kulda et al., 1988). There are two types of costae
(A, B) that differ in pattern of their striation, resulting
from different organization of the fiber substructures.
The A type banding pattern is also found in the ubiq-
uitous parabasalian striated roots—parabasal fibers that
maintain Golgi dictyosomes in perinuclear position
(Brugerolle and Viscogliosi, 1994). Despite the differ-
ences between costa A and B, Viscogliosi and Brugerolle
(1994) demonstrated that the main protein compounds
of both types are related. Their results also suggest the
presence of some common epitopes between the costae
and parabasal fibers. In contrast, absence of immuno-
logical cross-reactivity and different molecular mass
range indicates lack of relatedness between the proteins
of costae and constituents of striated roots such as as-
semblin, centrin or the kinetodesmal protein, known to
occur in other protists. Many species of the order
Trichomonadida lack either the costa, or both costa and
undulating membrane. These two structures are absent
or modified in the families Devescovinidae and Cal-
onymphidae which differ substantially from the basic
morphological pattern of the Trichomonadidae. How-
ever, the costa is also absent and the undulating mem-
brane reduced, or absent, in some species that otherwise
conform to the basic Trichomonadidae morphology.
The lack of costa and undulating membrane in these
species was regarded to be a taxonomically important
character substantiating their classification into the
separate family Monocercomonadidae. The four-family
classification of Trichomonadida (Trichomonadidae,
Monocercomonadidae, Calonymphidae, and Devesco-
vinidae) proposed by Honigberg (1963) with the addi-
tion of the fifth single-genus family Cochlosomatidae
(Pecka et al., 1996) is still widely used.
A scheme of evolution in the phylum Parabasala
based on light microscopic morphology was proposed
by Honigberg (1963) and later amended by inclusion of
ultrastructural data (see Brugerolle, 1976). In this pro-
posal, the family Monocercomonadidae was regarded as
ancestral to the whole phylum as it is morphologically
simple. The type genus Monocercomonas was considered
to represent the ancestral form, from which radiated
other genera of Monocercomonadidae. The Trichomo-
nadidae were also thought to have arisen from one such
lineage by stepwise development of undulating mem-
brane and costa. The current genera Hypotrichomonas
and Pseudotrichomonas, possessing undulating mem-
brane but lacking costa, were regarded as descendants of
an intermediate form in this transformation. According
to this scenario the Trichomonadidae lineage split to
yield the trichomonad and tritrichomonad branches and
from tritrichomonads arose the families Devescovinidae
and Calonymphidae. These were regarded as the ances-
tors of morphologically very complex Hypermastigida.
The advent of molecular techniques brought new in-
sight into the relationships between the lineages in the
phylum Parabasala. Phylogenetic trees based on gene
sequences of 16S rRNA and those based on morphology
have contradicted one another in many respects (Dacks
and Redfield, 1998; Delgado-Viscogliosi et al., 2000;
Edgcomb et al., 1998; Gerbod et al., 2000, 2001; Keeling
et al., 1998; Ohkuma et al., 1998, 2000; Viscogliosi et al.,
1999). First, the expected origin of Hypermastigida
within the Trichomonadida clade was challenged by a
16S rRNA phylogeny that suggested an opposing sce-
nario, in which the Trichomonadida diverge after sev-
eral Hypermastigida branches. Furthermore, members
of the family Monocercomonadidae form neither
monophyletic nor basal branches of the Trichomona-
dida subtree. This suggests that their simplicity is not a
primitive state and that the costa and undulating
membrane probably disappeared secondarily in these
species. Moreover, this loss probably happened several
times in unrelated branches. Molecular analyses also
indicated the polyphyletic or paraphyletic nature of the
families Calonymphidae and Devescovinidae (Gerbod et
al., 2002) and order Hypermastigida (Brugerolle and
Patterson, 2001; Delgado-Viscogliosi et al., 2000;
Frohlich and Konig, 1999; Gerbod et al., 2001, 2002;
Keeling et al., 1998; Ohkuma et al., 2000).
This molecular-based phylogeny not only contradicts
the morphology-based concept of parabasalid evolution
but it is also incongruent with the current classification
of the phylum Parabasala, because it suggests the
polyphyly of several taxa. Therefore, many authors have
pointed out the need to revise the classification in order
to make it consistent with molecular data (Delgado-
Viscogliosi et al., 2000; Gerbod et al., 2001; Keeling,
2002; Keeling et al., 1998; Viscogliosi et al., 1999).
Brugerolle and Patterson (2001) have made the first step
towards revision in proposing the division of the phylum
into three orders: Trichomonadida, Cristamonadida,
and Trichonymphida instead of the current two
Trichomonadida and Hypermastigida.
Molecular phylogenetic trees are prone to several
types of artifacts that may result in a potentially mis-
leading topology. The long-branch attraction (LBA)
artifact results in the artificial clustering of taxa whose
branch lengths noticeably exceed those of other organ-
isms (Felsenstein, 1978). Such long-branch taxa may not
be closely related and their increased length can be
caused, for example, by higher mutational rates. The
parabasalian 16S rRNA tree includes many such long
branches indicating the possible influence of LBA on the
tree topology.
2
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
ARTICLE IN PRESS

Page 3

In this paper, we present rigorous molecular phylo-
genetic analyses of the phylum Parabasala based on
currently available 16S rRNA sequences and 10 new
sequencesfromthegenera
Hexamastix (Hampl et al., Manuscript in preparation).
The analysis is aimed at critical assessment of tree to-
pologies produced from 16S rRNA alignments by in-
vestigating different subsets of characters and taxa
sampling.
Monocercomonas
and
2. Materials and methods
The sequences of 16S rRNA were determined for the
strain of Monocercomonas spp. (PYR-1-1, EUMM,
HAD, VAR-1, R208, TSC, and R183 isolated from
various reptile species), Monocercomonas ruminantium
(strains HER5 and KOJ 14 isolated from cattle), and
Hexamastix spp. (strains CYCL and T isolated from
reptiles). Accession numbers of the sequences are
AY319267–AY319280 and AY321149, the detailed in-
formation on the origin of the strains is given in Hampl
et al. (Manuscript in preparation). The secondary
structure-based alignment of 16S rRNA sequences of
parabasalids and outgroups was downloaded from the
ribosomal RNA database (http://rrna.uia.ac.be). The
sequences of Dientamoeba fragilis and H. meleagridis
and the new sequences were appended to this alignment
and realigned using the function ‘‘realign selected se-
quences’’ in ClustalX 1.81 (Thompson et al., 1997). The
alignment was then manually refined in the Bioedit se-
quence alignment editor (Hall, 1999). This full set of
taxa was used only in the first complete analysis (Fig. 1).
As isolates of Monocercomonas spp. PYR-1-1, EUMM,
HAD, VAR-1, R208, TSC, and R183 were very similar
to the database sequence of Monocercomonas sp. Ns-
1PRR (ATCC 50210) and isolate KOJ14 identical to
isolate HER5, only strains Ns-1PRR and HER5 were
included to all further analyses to save computational
effort.
Phylogenetic trees were constructed using the maxi-
mum likelihood (ML), maximum parsimony (MP), and
Fitch–Margoliash with Logdet distance (LD) methods
implemented in PAUP* 4.0 (Swofford, 1998) and by the
Bayesian method implemented in the program MrBayes
(Huelsenbeck and Ronquist, 2001).
Maximum likelihood trees in PAUP* were con-
structed using the best substitution models as deter-
mined by hierarchical nested likelihood ratio test
implemented in Modeltest version 3.06 (Posada and
Crandall, 1998). Except for some slow–fast analyses the
trees were searched using heuristic method with 10
replicates each. The starting tree was constructed by
random taxa addition. To save computer time, in some
slow–fast analyses only one replicate of heuristic sear-
ches was used with ‘‘as-is’’ addition of sequences. For
the maximum parsimony tree construction, the heuristic
searches with 10 replicates and random taxa addition
were used. One replicate of heuristic searches was used
for the LD method and the starting tree was constructed
by neighbor-joining. In order to lower the violation of
the rate homogenity across sites assumption, constant
positions were excluded from the alignment before
performing the LD analyses (Waddell and Steel, 1997).
In MrBayes, base frequencies, rates for six different
types of substitutions, number of invariant sites, and
shape parameter of the gamma correction for rate het-
erogeneity with four discreet categories were allowed to
vary. Usually 200,000 generations of the Markov Chain
Monte Carlo were run by using the default setting (four
simultaneous chains, heating temperature 0.2). The first
250–1000 trees were discarded as the ‘‘burnin.’’
The v2test of deviation from the expected nucleotide
composition for outgroup sequences was performed
using Treepuzzle (Strimmer and vonHaeseler, 1996).
Testing of phylogenetic hypotheses was performed by
using program Consel v0.1f (Shimodaira and Hasegawa,
2001) that includes, among others, KH and SH tests,
weighted KH and SH tests, and approximately unbiased
test (Shimodaira, 2002).
Rooting was performed in PAUP* 4.0 (Swofford,
1998) by the outgroup and midpoint methods, and by
using maximum likelihood with an enforced molecular
clock.
The slow–fast (S–F) method (Brinkmann and Phi-
lippe, 1999) was used to estimate, and lower the effect of
long-branch attraction (LBA) artifact by sequential re-
moval of positions with higher mutational rate that are
supposedly responsible for stochastic information noise.
In the S–F method the positions in the alignment were
divided into 11 (0–10) classes according to their in-
creasing mutational rate. The mutational rate was esti-
mated in the following way: the sum of the number of
changes for each position within the well-supported,
distinct clades comprising higher number of OTUs (in
our case clades 1, 2, 7, and 12 in Fig. 1) was calculated
by using maximum parsimony in PAUP*. The new
alignments (s0, s1, s2, ..., s9) were created from the
complete alignment, in which only positions with 0, up
to 1, up to 2, ..., up to 9 changes, respectively, were
included. All gaps were removed from the alignments
and new phylogenetic trees were constructed from these
alignments.
The exclusion of the positions with high mutational
rate lowers LBA artifact, but at the same time increases
the influence of stochastic effects at the tree topology as
a result of the decreasing amount of data. At a certain
point the stochastic effects overweight the information
contained in the sequences and the topology collapses.
To be able to consider the influence of both conflicting
effects, the tree robustness and number of included po-
sitions for each level of S–F was graphically visualized
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
3
ARTICLE IN PRESS

Page 4

(supplementary material Fig. s1). The tree robustness
was expressed as the average of the posterior probabil-
ities calculated in MrBayes (APP) or average of boot-
strap values (ABV). The APP was calculated as the
average of posterior probability values in the tree. To
lower the effect of random fluctuation, the Markov
Chain Monte Carlo analysis was run three times for
each alignment and for each run the APP was calcu-
lated. Because different runs in some cases produced
slightly different topologies, all three replicates were
Fig. 1. Phylogenetic tree of Parabasala based on 16S rRNA gene sequences, the tree was constructed by the ML method using TrN+I+G model of
substitution. The values at the nodes indicate statistical support for node estimated by three methods (LD bootstrap/MP bootstrap/MrBayes
posterior probability). Bar indicates the branch length corresponding to 10% of sites that underwent substitution event. The shaded boxes and
numbers indicate the well-supported clades: (1) Spirotrichonymphidae, Holomastigotoididae, and termite symbionts; (2) Calonymphidae and
Devescovinidae and termite symbionts; (3) Tritrichomonas foetus; (4) Dientamoeba fragilis and Histomonas meleagridis; (5) Gf10 termite symbiont; (6)
eight reptile isolates of Monocercomonas sp.—Ns-1PRR (ATCC 50210), PYR-1-1, EUMM, HAD, VAR-1, R208, TSC, and R183; (7) Trich-
onymphidae and termite symbionts; (8) Eucomonymphidae and termite symbionts; (9) termite symbionts (Gf8, Cd5, and Cb symbiont 1) indirectly
assigned to genus Tricercomitus or Hexamastix (Keeling et al., 1998; Ohkuma et al., 2000); (10) isolates of Hexamastix sp.—T and CYCL; (11) free-
living trichomonads (Ditrichomonas, Monotrichomonas, and Pseudotrichomonas) and isolates of M. ruminantium—HER5, KOJ14; (12) Trichomo-
nadina, Trichomitopsiinae, Pentatrichomonoidinae, and termite symbionts; (13) Hypotrichomonas acosta; and (14) Trichomitus batrachorum. New
sequences are printed in bold. The arrows indicate the position of the root as inferred by the outgroup method, midpoint method, and maximum
likelihood with molecular clock.
4
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
ARTICLE IN PRESS

Page 5

plotted in the graph. ABV was calculated as the average
of MP and LD bootstrap values in the tree (300 boot-
strap replicates). Because the S–F method is based on
the gradual exclusion of exactly those positions that
carry information on the internal topology of clades 1,
2, 7, and 12, and hence the bootstraps of these nodes
decrease and reach zero at the s0 level, these nodes were
not included in calculating ABV and APP.
3. Results and their interpretation
3.1. Phylogenetic relationships among parabasalids based
on 16S rRNA gene sequences
A phylogenetic tree of Parabasala based on 16S
rRNA gene sequences was constructed by using the
maximum likelihood method (ML) with the TrN+I+G
model of nucleotide change. The Parabasala formed 14
distinct clades (Fig. 1). These 14 clades were revealed in
all analyses despite using different tree-construction
methods. The new sequences, from Monocercomonas
and Hexamastix, are printed in bold.
The relationship between the 14 parabasalid clades
varied with the tree-construction method used. The tree
constructed from the same data set by Bayesian method
had essentially the same topology as that shown in Fig. 1.
In the phylogenetic tree constructed by maximum par-
simony (MP) the clades 7, 8, 9, and 10 (trichonymphids,
eucomonymphids, Hexamastix, and putative Tricer-
comitus) clustered with the clade 1(spirotrichonymphids)
and formed a sister branch to Trichomitus batrachorum
and Hypotrichomonas acosta. In Fitch–Margoliash
Logdet distance tree (LD) the clades 7, 8, 9, and 10
clustered with clades 1 and 4 (spirotrichonymphids and
dienthamoebids) and formed the sister branch to clades
2, 3, 5, and 6 (devescovinids, Tritrichomonas, reptile
monocercomonads, and termite symbiont Gf10).
Because the clades 1, 4, 7, 8, 9, and 10 represented the
longest branches in the tree by far, their clustering may
be a result of the long-branch attraction (LBA) artifact
(Felsenstein, 1978). To investigate the influence of LBA
on the position of these branches and on the topology of
parabasalian tree in general, we used the S–F method
(Brinkmann and Philippe, 1999) and the taxa-exclusion
method, an approach, in which we performed separate
analyses for each single long branch.
3.2. Slow–fast analysis
The most remarkable topological change induced by
gradual removal of fast evolving sites was the splitting of
the long-branch cluster 7–8–9–10 in Bayesian and ML
trees, and similarly, cluster 1–7–8–9–10 in MP trees. The
topology of the Bayesian trees down to s6 level did not
differ in important features from the topology of the tree
constructed from the total data. However, the number
of excluded positions down to s6 was rather low. At the
s5–s2 levels, the number of excluded positions consid-
erably increased and the support for the clustering of
clades 7, 8, 9, and 10 decreased concomitantly. The
position of clades 9 and 10 remained the same, while
clades 7 and 8 moved to the position of sister group to
T. batrachorum and H. acosta. Corresponding to the
topological changes, the average tree robustness de-
creased at s5 level, but then began to rise and at s3 level
reached a value equal to or even a little higher than the
value with all sites included. Prominent clustering of
almost all long branches was observed in the s0 and s1
tree, but at these levels the tree robustness considerably
decreased—probably because of the low amount of
data—and the tree topologies are therefore unreliable.
Trees inferred by maximum likelihood were similar to
the Bayesian topologies. However, the 7–8–9–10 cluster
was split only in the s3 tree. The splitting of the 1–7–8–
9–10 cluster in maximum parsimony trees followed, al-
most exactly, the pattern observed in the Bayesian trees.
The topology at the s3 level was virtually identical to the
Bayesian tree. In the LD trees, the cluster of long
branches 1–4–7–8–9–10 was not split at any level. The
tree topologies and robustness for all levels using the S–
F method are summarized in supplementary material
(Figs. s1 and s2).
The Bayesian, ML, and MP methods converged to the
same topology at the s3 level after removal of fast
evolvingsites(Fig.2).Becauseitwasinferredfromslowly
evolving sites, that tree may more correctly reflect true
phylogenetic relationships among parabasalian clades
than the tree inferred by using the complete alignment.
3.3. Taxa-exclusion analysis
The second approach that was used to reduce the
effect of the LBA artifact, was to first exclude all un-
stable long-branch-forming clades (1, 4, 7, 8, 9, and 10)
from the data set and then perform a series of analyses
that re-included only one of these clades at a time. This
taxa-exclusion method allowed us to infer the relation-
ship of the long branch to other OTUs without any
artificial attraction to other long branches. In each case
we also performed S–F analyses from s6 to s1 level.
The position of the D. fragilis–H. meleagridis clade 4
and putative Tricercomitus clade 9 in the reduced trees
was the same as that in Fig. 1. The statistical support for
dientamoebids as a sister branch to T. foetus obtained in
trees using all nucleotide positions was 70, 65, and 100 in
MP (bootstrap), LD (bootstrap), and MrBayes (pos-
terior probability) analyses, respectively. Similarly, the
statistical support for Tricercomitus as a sister branch to
the free-living trichomonad branch was 81, 71, and 100,
respectively. These results strongly support the position
of these clades as shown in Fig. 1.
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
5
ARTICLE IN PRESS

Page 6

The situation for the other four long-branching
clades was more complicated. Clade 1 (Spirotrich-
onymphidae and Holomastigotoididae) appeared, ex-
cept for in the s2 tree for MP, within or at the base of the
Tritrichomonas–Monocercomonas–Calonymphidae–
Devescovinidae branch. The statistical support for this
position in the tree (using all nucleotide positions) was
80, 62, and 99 in MP (bootstrap), LD (bootstrap), and
MrBayes (posterior probability) analyses, respectively.
These results indicated that spirotrichonymphids and
holomastigotoidids belong to the Tritrichomonas–Mon-
ocercomonas–Calonymphidae–Devescovinidae
as revealed in Fig. 1. However, they failed to determine
the exact branching point of this clade.
Clades 7 (Trichonymphidae) and 8 (Eucomonym-
phidae) did not branch as a sister group to the free-living
trichomonads and M. ruminantium, i.e., at position
congruent to the ML tree in Fig. 1. In most analyses,
they appeared as a branch related to T. batrachorum and
H. acosta. The statistical support of the T. batrachorum–
subtree
H. acosta–Trichonymphidae cluster in the tree obtained
by using all nucleotide positions was 80, 52, and 96 in
MP (bootstrap), LD (bootstrap), and MrBayes (pos-
terior probability) analyses, respectively. For the T.
batrachorum–H. acosta–Eucomonymphidae cluster the
statistical support was 59, less than 50 and 99, respec-
tively. These results contradicted the tree in Fig. 1 and
suggest that the position of these clades in Fig. 1 may be
the result of LBA. Because both clades were placed in a
similar position in the reduced trees, they may form a
monophyletic group.
The position of clade 10 (Hexamastix) was rather
unstable, but its placement next to the free-living
trichomonad branch as in the complete ML tree (Fig. 1)
was favored by most analyses, especially those with
more strict S–F levels (including the most consistent s3
level). These observations suggest to us that the sister
position of Hexamastix to free-living trichomonads re-
vealed in the tree in Fig. 1 may be correct. Because the
placement of clade 10 corresponds to the placement of
putative Tricercomitus clade 9, the two clades are
probably related as shown in Fig. 1. The results of taxa-
exclusion analyses of clades 1, 7, 8, and 10 are illustrated
in detail in the supplementary material (Fig. s3).
Fig. 3 shows a scheme of relationships among groups
of Parabasala that summarizes the results of taxa-ex-
clusion analyses. The scheme is congruent with tree from
the s3 level of S–F analysis (Fig. 2), and differs from the
ML tree in Fig. 1 in the position of the hypermastigid
clades 7 (Trichonymphidae) and 8 (Eucomonymphidae).
The color of each clade in the figure reflects their family
or order classification.
3.4. Rooting of the parabasalian tree
To investigate the position of the root of the para-
basalian tree, we chose 12 sequences from various eu-
karyotic groups, that did not significantly differ in their
base composition from the parabasalian sequences and
used them as outgroups—Cryptomonas sp., Fucus gard-
neri, Chlorarachnion reptans, Oryza sativa, Phreat-
amoeba balamuthi, Naegleria gruberi, Ceramium rubrum,
Prymnesium parvum, Saccharomyces cerevisiae, Euglena
gracilis, Hexamita inflata, and Eimeria necatrix. The
relationship among parabasalid groups in the rooted
treeconstructedby maximum
(TrN+I+G model of nucleotide change) was generally
in agreement with the unrooted tree in Fig. 1. The root
was situated on the Trichonymphidae branch inside the
cluster 7–8–9–10 (Fig. 1, arrow). A similar topology was
also recovered by maximum parsimony. In the tree in-
ferred by the LD method, clade 1 (Spirotrichonymphi-
dae and Holomastigotoididae) joined the 7–8–9–10
cluster at the root.
The affinity of particular long branches to the out-
groups could be caused by LBA artifact. To investigate
likelihoodmethod
Fig. 2. Result of slow–fast analyses. The tree based on the most con-
sistent s3 level of slow–fast was constructed by MrBayes. Identical
topology was revealed also by MP and ML methods. The numbering
of clades corresponds to Fig. 1.
6
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
ARTICLE IN PRESS

Page 7

this, we used four methods: the S–F method, the ex-
clusion of long parabasalian branches, the use of single
outgroups, and testing of constrained trees.
The removal of fast evolving sites during the S–F
analysis had no significant influence on the root posi-
tion. The root was situated either inside the 7–8–9–10
cluster, at its base or, in case of the s4 analysis, on the
Eucomonymphidae branch (clade 8) separate from the
7–9–10 cluster.
As the next approach, we excluded all the long-
branch taxa (clades 1, 4, 7, 8, 9, and 10) from the tree so
they could not attract outgroups by LBA. If the root
was truly located at Trichonymphidae or another ex-
cluded clade, it should appear in the position, where this
clade emerged in the complete tree (i.e., in the T. ba-
trachorum–H. acosta clade in case of Trichonymphidae).
We used four tree-construction methods (ML, Bayesian
method, MP, and LD) and performed S–F analyses for
each of them. Results are summarized in Fig. 4. The
position of the root varied with the tree-construction
method and in most analyses was inconsistent with the
position in the complete tree.
To investigate the influence of each outgroup se-
quence on the position of the root, we performed sep-
arate analyses with each single outgroup. The results are
summarized in Fig. 5. The position of the root varied
considerably with the different outgroups and with tree-
construction method used.
The topology of the ML tree obtained by using the
complete set of taxa (Fig. 1) is probably wrong, and this
can affect the rooting. Thus, in the next analysis, we
used the complete set of taxa but constrained the to-
pology to the topology of Parabasala as it is predicted
in Fig. 3. Then we tested likelihood differences between
the five most probable positions of the root (Fig. 3,
arrows). We performed this test on the complete
alignment, as well as with the s3 alignment. We used
either the set of 12 outgroups, or Hexamita as a single
outgroup. Neither an approximately unbiased test, or
other tests implemented in Consel showed statistically
significant differences between likelihoods of the posi-
tions tested.
We also used rooting methods that do not require
outgroups—the midpoint method, and ML with an en-
forced molecular clock. Using these methods we esti-
mated the root position in the complete tree and the tree
with long branches excluded. The root inferred by these
methods appeared, in both cases, between Monocerco-
monas sp. and the H. acosta–T. batrachorum branch
(Fig. 1 arrow and Fig. 4 shade box).
Although some analyses placed the root at the base of
the Trichonymphidae or at a position congruent with
such rooting, most of them favored other positions. Our
analyses, thus, challenged the Trichonymphidae rooting
butdidnotsuggestanyotherrobustpositionfortheroot.
3.5. Testing of polyphyly of the order Hypermastigida and
family Monocercomonadidae
The representatives of the orders Hypermastigida and
Trichomonadida, and families Monocercomonadidae
and Trichomonadidae appeared to be polyphyletic in
the trees. However, the evidence on the polyphyly of
these groups was not strong.
Fig. 3. Result of taxa-exclusion analyses. The tree is a composition based on results of taxa-exclusion analyses. The colors of clades indicate their
taxonomic classification: red, family Monocercomonadidae; blue, family Trichomonadidae; green, family Devescovinidae; brown, family Cal-
onymphidae; and yellow, order Hypermastigida. The thick lines indicate the branches connecting the Trichomonadidae. The forms at these branches
probably possessed costa (see Section 4). The arrows indicate the root positions that were tested (see Section 3).
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
7
ARTICLE IN PRESS

Page 8

Because all taxa in the tree are members of either Hy-
permastigida or Trichomonadida, the problem of poly-
phyly of Hypermastigida and Trichomonadida is
interconnected; either Hypermastigida or Trichomona-
dida (or both) are polyphyletic. To test the significance of
polyphyletic nature of Hypermastigida/Trichomonadida
we constructed a phylogenetic tree based on 16S rRNA
gene sequences in which monophyly of Hypermastigida/
Trichomonadida was constrained, and then used tests
implemented in the program Consel v0.1f to test the sig-
nificance of the difference between the likelihood value of
the constrained tree and that of the best tree. All tests
showedthatthisdifferenceisnotsignificantatthe5%level
(approximately unbiased test p ¼ 0:122, SE ¼ 0:006).
Analogously, we tested the polyphyly of the family
Monocercomonadidae. Because the classification of H.
acosta to the family Monocercomonadidae can be
wrong (Kulda, 1965), we performed two separate tests.
In one, we regarded H. acosta as a member of Monoc-
ercomonadidae and in the other we tested the mono-
phyly of Monocercomonadidae without this species. In
both cases all the tests implemented in Consel v0.1f
showed that the overall best tree is significantly better
than the tree with monophyletic family Monocercomo-
nadidae at the 0.01 level (approximately unbiased test
p ¼ 0:01, SE60:002 for both constrained topologies).
We tested the polyphyly of the family Trichomo-
nadidae in the same way. Again we performed two
separate tests with H. acosta either included or excluded
from the family Trichomonadidae. In both cases all the
tests showed that the overall best tree is significantly
better than the tree with monophyletic family Tricho-
monadidae at 0.02 level (approximately unbiased test,
Hypotrichomonas
included:
Hypotrichomonas excluded: p ¼ 0:005, SE ¼ 0:001).
Because the alignment at the s3 level of S–F probably
contained a less biased phylogenetic signal, we per-
formed all these tests again using this alignment. The
results were identical and p values were similar to those
based on the non-reduced alignment.
p ¼ 0:003,
SE ¼ 0:001;
4. General discussion
4.1. Phylogenetic relationships in the phylum Parabasala
Results of our analyses concerning the relationship
among parabasalid taxa are generally consistent with
Fig. 4. Rooting of the tree of Parabasala after exclusion of long branches. The tree of Parabasala without long branches was rooted using 12 eu-
karyotic outgroups. The position of the root for various methods and S–F analyses is indicated. The methods and levels of S–F are listed in boxes.
‘‘Total’’ designates the analysis based on all nucleotide positions (not S–F). The diameter of the dot corresponds to the number of methods that
placed the root in the particular position. If the topology of the reduced tree slightly changed, and it was not possible to place the branch in the figure
exactly, the region where the branch should belong was marked by an ellipse.
8
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
ARTICLE IN PRESS

Page 9

the previous results (Delgado-Viscogliosi et al., 2000;
Edgcomb et al., 1998; Gerbod et al., 2000, 2001, 2004;
Keeling et al., 1998; Ohkuma et al., 2000; Viscogliosi et
al., 1999). However, the positions of certain clades in the
comprehensive analysis (Fig. 1) were unstable and
method-dependent. All clades in question formed long
branches resulting from a high divergence of their 16S
rRNA gene sequences. The position of these branches
could, therefore, be influenced by stochastic effects and
artifacts of the tree-construction methods.
We reanalyzed the position of these branches using
the S–F (Brinkmann and Philippe, 1999) and taxa-ex-
clusion methods, both designed to minimize the attrac-
tion between long branches in the tree. As expected,
results from both methods led to similar conclusions
(Figs. 2 and 3).
Both analyses cast doubt on the phylogenetic rela-
tionship of Trichonymphidae and Eucomonymphidae
(clades 7 and 8) to the 9–10–11 cluster. The affinity of
these clades to 9–10–11 cluster in Fig. 1 probably results
from LBA between clades 7, 8 and 9, 10. Because only
few published analyses have included the Tricercomitus
clade 9 (Delgado-Viscogliosi et al., 2000; Keeling, 2002;
Keeling et al., 1998), the affinity of spirotrichonymphids
and eucomonymphids to Tricercomitus has not previ-
ously attracted much attention. Moreover, two of these
analyses (Delgado-Viscogliosi et al., 2000; Keeling et al.,
1998) used distance or quartet puzzling methods that in
our opinion may have introduced even more serious
biases that obscured the relationships of parabasalian
clades (e.g., clustering of Dientamoeba with hyperm-
astigids), so this artifact remained hidden. But the recent
ML tree of Keeling (2002) is virtually identical with our
ML tree in Fig. 1 and also includes this artificial cluster.
4.2. The root of Parabasala
Previous molecular analyses that attempted to de-
termine the root of Parabasala (Delgado-Viscogliosi
et al., 2000; Keeling et al., 1998; Ohkuma et al., 2000)
used the outgroup method with several eukaryotic taxa.
In detailed analyses focused on the rooting of Paraba-
sala (Keeling et al., 1998; Ohkuma et al., 2000) the au-
thors used the KH test for testing of the various
hypotheses of the root position. Most of the analyses
favored rooting at the Trichonymphidae branch.
Fig. 5. Rooting of the tree of Parabasala after exclusion of long branches with each single outgroup independently. The diameter of the dot cor-
responds to the number of outgroups that placed the root in the particular position. *indicate the cases, in which the topology of the reduced tree was
changed but it was still possible to place the root in the figure.
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
9
ARTICLE IN PRESS

Page 10

Our results strongly challenged this hypothesis but
did not suggest any robust alternative position of the
root. A major problem of rooting for the Parabasala is
probably the lack of any close outgroup species. The
phylum stays as a separate branch with no clear and
close relationship to any other eukaryote group in vir-
tually all analyses concerning eukaryotic phylogeny
(e.g., Edgcomb et al., 2001; Sogin and Silberman, 1998).
Some evidence suggests (Baldauf et al., 2000; Henze
et al., 2001) that diplomonads may represent the nearest,
but still very distant, sister taxon to Parabasala. The
large genetic distances between ingroup and outgroup
complicate the identification of the root position
(Huelsenbeck et al., 2002). It has been shown that as the
length of the branch leading to outgroup sequence in-
creases, the ability of the outgroup method to determine
the root decreases. As the length approaches infinity, the
position of the root becomes essentially random (Huel-
senbeck et al., 2002). Additionally, several groups of
Parabasala have highly divergent 16S rRNA sequences
and form long branches in the parabasalian tree. The
presence of long branches probably enhances the influ-
ence of artifacts of the tree-construction methods such
as LBA. Long ingroup branches can be strongly at-
tracted to the long branch of outgroup in the absence of
phylogenetic signal, as probably happened in case of the
Trichonymphidae. LBA also affects statistical tests
based on comparing of likelihoods.
The basal position of multiflagellate parabasalids like
the Trichonymphidae is difficult to explain from a
morphological point of view. Polymastigont organiza-
tion is not very common in other taxa of flagellates and
all potential relatives of Parabasala typically possess
four-kinetosome mastigont (doubled in Diplomona-
dida). The four-kinetosome mastigont is also regarded
as plesiomorphic for Trichomonadida. The basic set of
four privileged kinetosomes with characteristic cyto-
skeletal appendages can be identified even in genera with
complex mastigonts supplemented by numerous addi-
tional kinetosomes (Brugerolle, 1991). The most parsi-
monious scenario would therefore predict that the
Parabasala evolved from four-kinetosome flagellates
rather than trichonymphid-like ones.
4.3. Polyphyly of the orders Hypermastigida and Tricho-
monadida, families Monocercomonadidae and Trichomo-
nadidae, and anagenesis of cytoskeletal and motility
organelles
Family Trichomonadidae was split into three distinct
clades in our analyses (trichomonads, tritrichomonads,
and Trichomitus) and its polyphyly was statistically
significant (p < 0:02). This fact could, however, reflect
the polyphyly of Monocercomonadidae or Hyperm-
astigida or both (see below). In other words, the com-
mon ancestor of all parabasalids could theoretically be a
Trichomonadidae-like protozoan. In this case, Tricho-
monadidae would be paraphyletic rather than poly-
phyletic. To distinguish between para- and polyphyly it
would be necessary to know the morphotype of the last
common ancestor.
Morphologically the Trichomonadidae family is
characterized by the presence of undulating membrane
and costa. The independent origin of the undulating
membrane in three separate Trichomonadidae branches
is relatively plausible, because the membranes of each
group differ in their ultrastructure (Brugerolle, 1976).
Moreover, analogous undulating membranes are also
present in unrelated protozoa (e.g., trypanosomes). In
contrast, the triple independent origin of the costa ap-
pears to be less probable. The costa, a dominant striated
root fiber of the Trichomonadidae, might have been
evolved from parabasal fibers that are present in virtu-
ally all representatives of Parabasala. Indeed, there is a
close similarity in ultrastructure between the parabasal
fiber and the costa of Trichomitus and tritrichomonads,
both possessing the A type pattern of banding. The B
type costa of Trichomonadinae and Trichomitopsis
shares, with the A type, a 42nm periodicity of the major
repetitive bands, but contains additional longitudinal
filaments providing its characteristic lattice appearance
in longitudinal sections. There are also minor differences
in the topology of attachment to kinetosomes. Despite
these differences, comparative morphological studies on
trichomonad mastigonts (Brugerolle, 1976) revealed a
common basic pattern of organization of kinetosome
associated fibers in Trichomonadidae, and substantiated
homology of the main components of the mastigont,
including the costa. Moreover, the available immuno-
cytochemical and protein analyses (Viscogliosi and
Brugerolle, 1994) suggest that the major proteins of both
types of costa belong to a common protein family. These
results favor the hypothesis that costa originated only
once in the common ancestor of Trichomonadidae, thus
implying that the family is paraphyletic rather than
polyphyletic.
The polyphyletic nature of family Monocercomona-
didae has been proposed in several studies. In our
analyses (Fig. 3), the representatives of this family
formed four groups. Although the bootstrap support for
nodes separating Monocercomonadidae is very low, all
tests support the hypothesis of Monocercomonadidae
polyphyly.
The polyphyly of Monocercomonadidae implies mul-
tiple origin of forms without costa and with a reduced, or
absent, undulating membrane. This scenario seems
plausible, because multiple losses, or reductions, of
functionalstructurescanbeeasilyexplained,forexample,
by the loss of selectable advantages that they bring to the
organisms experiencing new ecological conditions.
The orderHypermastigidaappearedinouranalysesas
diphyletic. Families Trichonymphidae+Eucomonym-
10
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
ARTICLE IN PRESS

Page 11

phidae represented the first, though weakly supported,
clade and Spirotrichonymphidae+Holomastigotoididae
the second well-supported clade, as previously suggested
by Gerbod et al. (2001). Although the hypothesis of di-
phyletic Hypermastigida was not statistically significant
andwasnotsupportedbyhighbootstraps,itwasthemost
probable hypothesis. It is important to mention that se-
quences for Koruga bonita and Joenina sp. are missing in
our analyses. We did not include them, because the 16S
rRNA gene sequence of K. bonita is only partial and the
other sequence was only putatively ascribed to the genus
Joenina. However, it has been shown by other authors
that both sequences branch inside the Calonymphidae
and Devescovinidae clade with high bootstrap support
(Frohlich and Konig, 1999; Gerbod et al., 2002), which is
in accordance with morphological observations (Bruge-
rolle and Patterson, 2001). These results clearly show the
polyphyly of the order.
The polyphyly of Hypermastigida implies multiple
origin of a complex Hypermastigida morphology. Other
than the large cell size and higher cell complexity in
general, the hypermastigid clades differ in the organi-
zation of the mastigont and cytoskeleton (Brugerolle
and Lee, 2000). Moreover, the trend in Trichomonadida
to multiply flagella or mastigonts in certain cases is well
known, so it is not difficult to imagine multiple origins of
polymastigont or multiflagellated forms in the evolution
of this order.
The second order, Trichomonadida, is also polyphy-
letic in our analyses (Fig. 3). Because all taxa in the tree
belong either to order Trichomonadida or Hyperm-
astigida, the polyphyly of both orders has the same
statistical significance and has probably the same cause.
In our opinion, Trichomonadida are paraphyletic rather
than polyphyletic. As we concluded in the previous
section, their common ancestor probably morphologi-
cally resembled the family Trichomonadidae. Therefore,
we favor the hypothesis that the polyphyletic distribu-
tion of Hypermastigida/Trichomonadida is due to mul-
tiple origins of the hypermastigote morphology, which is
easier to explain.
The previous discussion is summarized in Fig. 3, in
which a thick line between Trichomonadidae branches
indicates the presence of a costa and (supposedly) un-
dulating membrane. The monocercomonad- and hy-
permastigid-like morphology originated several times
from these costa bearing parabasalids. Because we do
not know the position of the root, we can only speculate
whether the Trichomonadidae morphology is plesio-
morphic for the whole order, or whether it originated
from simpler (Monocercomonadidae) or more complex
(Hypermastigida, Calonymphidae, and Devescovinidae)
morphotypes.
Unlike the costa and undulating membrane charac-
ters, the morphology of the pelta–axostylar complex is
reflected in some respects by the topology of the tree in
Fig. 3. Species in the right half of the tree (Tritricho-
monas, Monocercomonas spp., Devescovinidae, Cal-
onymphidae, and Joeniidae) possess a stout hyaline
axostyle. The trunk of their axostyle has more or less
uniform diameter along its whole length and tapers
abruptly at the posterior end. The microtubular sheet
forming this axostyle is rolled up in a tube-like fashion,
not cone-like as it is in most of other Trichomonadida.
In Calonymphidae, Devescovinidae, and Joeniidae
(Brugerolle and Patterson, 2001) the microtubular sheet
is more spiralized and fills the internal space of the
axostylar trunk. In Calonymphidae, the multiple axo-
styles either form one central bundle (Calonympha,
Snyderella), or stretch separately in the cytoplasm (Co-
ronympha, Metacoronympha). These two types of orga-
nization correspond to the existence of two unrelated
clades of Calonymphidae in the tree (Gerbod et al.,
2002). Exceptions to the rule in this part of the tree are
axostyles of dientamoebids. Dientamoeba has lost the
pelta and axostyle completely and the axostyle in His-
tomonas is reduced.
In the left part of the tree, the branch of Trichomo-
nadidae, Hexamastix, free-living trichomonads, and M.
ruminantium shares, with few exceptions, a relatively
slender type of axostyle. The axostyles of H. acosta and
T. batrachorum are stouter but, similar to the previous
group, they are formed by cone-like rather than tube-
like coiling of the microtubular sheet.
An exception to the scheme is the presence of multiple
axostyles formed by microtubular bands in two unre-
lated groups—Trichonymphidae/Eucomonymphidae
and Spirotrichonymphidae/Holomastigotoididae.
Molecular phylogeny sheds new light on the poly-
morphism of the shape of the axostyle in the genera
Monocercomonas and Hexamastix. Honigberg (1963)
suggested that the wide polymorphism of these usually
conservative structures is the result of a primitive evo-
lutionary status of this genus. However, the results of
molecular analyses indicate that it is an artifact of the
polyphyly of the genus. For example, in the set of cur-
rently available taxa, the representatives of the genus
Monocercomonas split into two unrelated groups in
agreement with the morphology of their axostyles:
Monocercomonas spp. from reptilian hosts on one side
and M. ruminantium on the other. A similar situation is,
or could be, possible for the genus Hexamastix.
4.4. Taxonomy of the phylum Parabasala
Several authors have pointed out the need to revise the
classification of the phylum Parabasala on the basis of
phylogenetic analyses of molecular data (Delgado-Vi-
scogliosi et al., 2000; Gerbod et al., 2001; Keeling et al.,
1998; Ohkuma et al., 2000; Viscogliosi et al., 1999). To
reconcile the classification to the current knowledge of
parabasalian phylogeny, Brugerolle andPatterson (2001)
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
11
ARTICLE IN PRESS

Page 12

proposedanewclassificationofParabasalaattheordinal
level. They divided the phylum into three orders (Tri-
chonymphida, Cristamonadida, and Trichomonadida)
instead of the current two (Hypermastigida and Tricho-
monadida). The newly created order Cristamonadida
comprises the families Devescovinidae and Calonym-
phidae—currently classified to the order Trichomona-
dida—and the families Joenidae, Lophomonadidae,
Deltotrichonymphidae, Rhizonymphidae, and Kofoidi-
dae—currently classified under the order Hypermastig-
ida. For the remaining hypermastigid families they
created the order Trichonymphida, that they regarded as
basal within the phylum Parabasala. This modification
reflected the growing molecular and ultrastructural evi-
dence that some representatives of Hypermastigida are
related to Calonymphidae and Devescovinidae (Frohlich
and Konig, 1999; Gerbod et al., 2002).
However, this proposed classification is still incon-
gruent with molecular and morphological data in several
respects. First, the monophyletic nature of the order
Trichonymphida is doubtful. Our analyses suggest that
the families Spirotrichonymphidae and Holomastigoto-
ididae do not form a clade with the families Eucom-
onymphidaeandTrichonymphidae.
cannot exclude the possibility that the proposed order
Trichonymphida is monophyletic, to establish this taxon
at the current stage of knowledge is, in our opinion,
premature. Second, the designation of Cristamonadida
and Trichomonadida as sister orders does not corre-
spond with current views, or with the results of our
study. Although current data fully support the mono-
phyly of the order Cristamonadida, creation of this or-
der causes the paraphyly or polyphyly of the order
Trichomonadida. Both molecular and morphological
data indicate that the clade Cristamonadida arose from
one lineage of the order Trichomonadida. The closest
relatives to Cristamonadida are probably the subfamily
Tritrichomonadinae and genera Dientamoeba, Histo-
monas, and Monocercomonas. To accommodate the
classification to the phylogeny either the aforementioned
taxa must be included within the order Cristamonadida,
or the group Cristamonadida must be reclassified as a
member (perhaps family or suborder) of the order
Trichomonadida with its current species composition.
The classification of an organismal group should re-
flect the phylogenetic relationships among the species.
Although the available data are clearly in conflict with
the current classification, they are still not sufficient to
understand the phylogenetic relationships among the
species. Based on analyses of 16S rRNA gene sequences,
we were able to identify 14 robust clades and to recon-
struct the possible relationship among them (Fig. 3).
However, this tree is based on the single well-sampled
gene, has low support for deep nodes, and some key
information, for example, the root position, cannot be
inferred reliably. In our opinion, any taxonomic revision
Although we
may be premature and risky at this stage. We suggest
that future work in this field should be focused on ver-
ification of the relationships among the robust clades as
deduced from the 16S rRNA by gathering and analyzing
sequences of another independent gene. The first serious
move in this direction was made by Gerbod et al. (2004).
Until a robust parabasalian phylogeny is recovered we
suggest the retention of the current classification system
for the orders Hypermastigida (revised by Hollande and
Carruette-Valentin, 1971), and Trichomonadida (revised
by Honigberg, 1963, and modified by Brugerolle, 1976,
1980; Camp et al., 1974; Honigberg and Kuldov? a, 1969;
Pecka et al., 1996).
Acknowledgments
We thank David S. Horner and Joel B. Dacks for
critical reading of the manuscript and helpful comments.
The work was supported by the Grants GAUK 264/
1999, GA?CR 204/03/1243, and MSM 113100004.
References
Baldauf, S.L., Roger, A.J., Wenk-Siefert, I., Doolittle, W.F., 2000. A
kingdom-level phylogeny of eukaryotes based on combined protein
data. Science 290, 972–977.
Brinkmann, H., Philippe, H., 1999. Archaea sister group of bacteria?
Indications from tree reconstruction artifacts in ancient phyloge-
nies. Molecular Biology and Evolution 16, 817–825.
Brugerolle, G., 1976. Cytologie ultrastructurale, systematique et
evolution des Trichomonadida. Annales de la Station Biologique
Besse-en-Chandesse 10, 1–57.
Brugerolle, G., 1980. Etude ultrastructurale du flagell? e Protrichomonas
legeri (L? eger 1905), parasite de l?estomac des bogues (Box boops).
Protistologica 16, 353–358.
Brugerolle, G., 1991. Cell organization in free-living amitochondriate
heterotrophic flagellates. In: Patterson, D., Larsen, J. (Eds.), The
Biology of Free-living Heterotrophic Flagellates. Clarendon Press,
Oxford, pp. 133–148.
Brugerolle, G., Lee, J.J., 2000. Phylum Parabasala. In: Lee, J.J.,
Leedale, G.F., Bradbury, P. (Eds.), The Illustrated Guide to the
Protozoa, second ed. Allen Press, Lawrence, pp. 1196–1250.
Brugerolle, G., Patterson, D.J., 2001. Ultrastructure of Joenina
pulchella Grassi, 1917 (Protista, Parabasalia), a reassessment of
evolutionary trends in the parabasalids, and a new order Crista-
monadida for devescovinid, calonymphid and lophomonad flagel-
lates. Organisms Diversity and Evolution 1, 147–160.
Brugerolle, G., Viscogliosi, E., 1994. Organization and composition of
the striated roots supporting the Golgi-apparatus, the so-called
parabasal apparatus, in parabasalid flagellates. Biology of the Cell
81, 277–285.
Camp, R.R., Mattern, C.F., Honigberg, B.M., 1974. Study of
Dientamoeba fragilis Jepps & Dobell. I. Electron microscopic
observations of the binucleate stages. II. Taxonomic position and
revision of the genus. Journal of Protozoology 21, 69–82.
Corliss, J.O., 1994. An interim utilitarian (users-friendly) hierarchical-
classification and characterization of the protists. Acta Protozoo-
logica 33, 1–51.
Dacks, J.B., Redfield, R.J., 1998. Phylogenetic placement of Trich-
onympha. Journal of Eukaryotic Microbiology 45, 445–447.
12
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
ARTICLE IN PRESS

Page 13

Delgado-Viscogliosi, P., Viscogliosi, E., Gerbod, D., Kulda, J., Sogin,
M.L., Edgcomb, V.P., 2000. Molecular phylogeny of parabasalids
based on small subunit rRNA sequences, with emphasis on the
Trichomonadinae subfamily. Journal of Eukaryotic Microbiology
47, 70–75.
Edgcomb, V.P., Roger, A.J., Simpson, A.G.B., Kysela, D.T., Sogin,
M.L., 2001. Evolutionary relationships among ‘‘jakobid’’ flagel-
lates as indicated by alpha- and beta-tubulin phylogenies. Molec-
ular Biology and Evolution 18, 514–522.
Edgcomb, V., Viscogliosi, E., Simpson, A.G.B., Delgado-Viscogliosi,
P., Roger, A.J., Sogin, M.L., 1998. New insights into the phylogeny
of trichomonads inferred from small subunit rRNA sequences.
Protist 149, 359–366.
Felsenstein, J., 1978. Cases in which parsimony or compatibility
methods will be positively misleading. Systematic Zoology 27, 401–
410.
Frohlich, J., Konig, H., 1999. Rapid isolation of single microbial cells
from mixed natural and laboratory populations with the aid of a
micromanipulator. Systematic and Applied Microbiology 22, 249–
257.
Gerbod, D., Edgcomb, V.P., No€ el, C., Delgado-Viscogliosi, P.,
Viscogliosi, E., 2000. Phylogenetic position of parabasalid symbi-
onts from the termite Calotermes flavicollis based on small subunit
rRNA sequences. International Microbiology 3, 165–172.
Gerbod, D., Edgcomb, V.P., No€ el, C., Zenner, L., Wintjens, R.,
Delgado-Viscogliosi, P., Holder, M.E., Sogin, M.L., Viscogliosi,
E., 2001. Phylogenetic position of the trichomonad parasite of
turkeys, Histomonas meleagridis (Smith) Tyzzer, inferred from
small subunit rRNA sequence. Journal of Eukaryotic Microbiol-
ogy 48, 498–504.
Gerbod, D., No€ el, C., Dolan, M.F., Edgcomb, V.P., Kitade, O., Noda,
S., Dufernez, F., Ohkuma, M., Kudo, T., Capron, M., Sogin,
M.L., Viscogliosi, E., 2002. Molecular phylogeny of parabasalids
inferred from small subunit rRNA sequences, with emphasis on the
Devescovinidae and Calonymphidae (Trichomonadea). Molecular
Phylogenetics and Evolution 25, 545–556.
Gerbod, D., Sanders, E., Moriya, S., No€ el, C., Takasu, H., Fast, N.M.,
Delgado-Viscogliosi, P., Ohkuma, M., Kudo, T., Capron, M.,
Palmer, J.D., Keeling, P.J., Viscogliosi, E., 2004. Molecular phylog-
eniesofParabasaliainferredfromfourproteingenesandcomparison
with rRNA trees. Molecular Phylogenetics and Evolution 31, 572–
580.
Hall, T.A., 1999. BioEdit: a user-friendly biological sequence align-
ment editor and analysis program for Windows 95/98/NT. Nucleic
Acids Symposium Series 41, 95–98.
Hampl, V., Cepicka, I., Flegr, J., Tachezy, J., Kulda, J., Phylogenetic
and morphological analysis of genera Monocercomonas, Pseudo-
trichomonas and Hexamastix (Monocercomonadidae, Parabasala).
Manuscript in preparation.
Henze, K., Horner, D.S., Suguri, S., Moore, D.V., Sanchez, L.B.,
Muller, M., Embley, T.M., 2001. Unique phylogenetic relation-
ships of glucokinase and glucosephosphate isomerase of the
amitochondriateeukaryotes
barkhanus and Trichomonas vaginalis. Gene 281, 123–131.
Hollande, A., Carruette-Valentin, J., 1971. Les atractophores, l?induc-
tion du fuseau, et la division cellulaire chez les Hypermastigines,
? etude infrastructurale et r? evision syst? ematique des Trichonym-
phines et des Spirotrichonymphines. Protistologica 7, 5–100.
Honigberg, B.M., 1963. Evolutionary and systematic relationships in
the flagellate order Trichomonadida Kirby. Journal of Protozool-
ogy 10, 20–63.
Honigberg, B.M., Kuldov? a, J., 1969. Structure of a nonpathogenic
histomonad from the cecum of galliform birds and revision of the
Giardiaintestinalis,
Spironucleus
family Monocercomonadidae Kirby. Journal of Protozoology 16,
526–535.
Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference
of phylogenetic trees. Bioinformatics 17, 754–755.
Huelsenbeck, J.P., Bollback, J.P., Levine, A.M., 2002. Inferring the
root of a phylogenetic tree. Systematic Biology 51, 32–43.
Keeling, P.J., 2002. Molecular phylogenetic position of Trichomitopsis
termopsidis (Parabasalia) and evidence for the Trichomitopsiinae.
European Journal of Protistology 38, 279–286.
Keeling, P.J., Poulsen, N., McFadden, G.I., 1998. Phylogenetic
diversity of parabasalian symbionts from termites, including the
phylogenetic position of Pseudotrypanosoma and Trichonympha.
Journal of Eukaryotic Microbiology 45, 643–650.
Kulda,J.,1965.PhDthesis,CharlesUniversityinPrague,Unpublished.
Kulda, J., Noh? ynkova, E., Ludvı ´k, J., 1988. Basic structure and
function of the trichomonad cell. Acta Universitatis Carolinae 30,
181–198.
Ohkuma, M., Ohtoko, K., Grunau, C., Moriya, S., Kudo, T., 1998.
Phylogenetic identification of the symbiotic hypermastigote Trich-
onympha agilis in the hindgut of the termite Reticulitermes speratus
based on small-subunit rRNA sequence. Journal of Eukaryotic
Microbiology 45, 439–444.
Ohkuma, M., Ohtoko, K., Iida, T., Tokura, M., Moriya, S., Usami,
R., Horikoshi, K., Kudo, T., 2000. Phylogenetic identification of
hypermastigotes, Pseudotrichonympha, Spirotrichonympha, Holom-
astigotoides, and parabasalian symbionts in the hindgut of termites.
Journal of Eukaryotic Microbiology 47, 249–259.
Pecka, Z., Noh? ynkov? a, E., Kulda, J., 1996. Ultrastructure of Cochlo-
soma anatis Kotl? an, 1923 and taxonomic position of the family
Cochlosomatidae (Parabasala: Trichomonadida). European Jour-
nal of Protistology 32, 190–201.
Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of
DNA substitution. Bioinformatics 14, 817–818.
Shimodaira, H., 2002. An approximately unbiased test of phylogenetic
tree selection. Systematic Biology 51, 492–508.
Shimodaira, H., Hasegawa, M., 2001. CONSEL: for assessing the
confidence of phylogenetic tree selection. Bioinformatics 17, 1246–
1247.
Sogin, M.L., Silberman, J.D., 1998. Evolution of the protists and
protistan parasites from the perspective of molecular systematics.
International Journal for Parasitology 28, 11–20.
Strimmer, K., vonHaeseler, A., 1996. Quartet puzzling: a quartet
maximum-likelihood method for reconstructing tree topologies.
Molecular Biology and Evolution 13, 964–969.
Swofford, D.L., 1998. PAUP*. Phylogenetic Analysis Using Parsi-
mony (*and Other Methods). Version 4. Sinauer Associates,
Sunderland, MA.
Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins,
D.G., 1997. The ClustalX windows interface: flexible strategies for
multiple sequence alignment aided by quality analysis tools.
Nucleic Acids Research 24, 4876–4882.
Viscogliosi, E., Brugerolle, G., 1994. Striated fibers in trichomonads—
costa proteins represent a new class of proteins forming striated
roots. Cell Motility and the Cytoskeleton 29, 82–93.
Viscogliosi, E., Edgcomb, V.P., Gerbod, D., No€ el, C., Delgado-
Viscogliosi, P., 1999. Molecular evolution inferred from small
subunit rRNA sequences: what does it tell us about phylogenetic
relationships and taxonomy of the parabasalids? Parasite 6, 279–
291.
Waddell, P., Steel, A.M., 1997. General time-reversible distances with
unequal rates across sites: mixing gamma and inverse Gaussian
distributions with invariant sites. Molecular Phylogenetics and
Evolution 8, 398–414.
V. Hampl et al. / Molecular Phylogenetics and Evolution xxx (2004) xxx–xxx
13
ARTICLE IN PRESS