Molecular Phylogenetics and Evolution 61 (2011) 351–362

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier .com/ locate /ympev

A multi-locus analysis of phylogenetic relationships within cheilostomebryozoans supports multiple origins of ascophoran frontal shields

Sarah Knight a, Dennis P. Gordon b, Shane D. Lavery a,c,⇑a School of Biological Sciences, University of Auckland, Private Bag 92019, Auckland Mail Centre, Auckland 1142, New Zealandb National Institute of Water and Atmospheric Research, Private Bag 14901, Kilbirnie, Wellington, New Zealandc Leigh Marine Laboratory, University of Auckland, Box 349, Warkworth, Northland 0941, New Zealand

a r t i c l e i n f o

Article history:Received 24 December 2010Revised 4 July 2011Accepted 7 July 2011Available online 21 July 2011

Keywords:BryozoaCheilostomataPhylogeneticsEvolutionMulti-locus

1055-7903/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.ympev.2011.07.005

⇑ Corresponding author at: School of Biological SciePrivate Bag 92019, Auckland Mail Centre, Auckland 11373 7417.

E-mail addresses: s.knight@auckland.ac.nz (S. K(D.P. Gordon), s.lavery@auckland.ac.nz (S.D. Lavery).

a b s t r a c t

Phylogenetic relationships within the bryozoan order Cheilostomata are currently uncertain, with manymorphological hypotheses proposed but scarcely tested by independent means of molecular analysis.This research uses DNA sequence data across five loci of both mitochondrial and nuclear origin from91 species of cheilostome Bryozoa (34 species newly sequenced). This vastly improved the taxonomiccoverage and number of loci used in a molecular analysis of this order and allowed a more in-depth lookinto the evolutionary history of Cheilostomata. Maximum likelihood and Bayesian analyses of individualloci were carried out along with a partitioned multi-locus approach, plus a range of topology tests basedon morphological hypotheses. Together, these provide a comprehensive set of phylogenetic analyses ofthe order Cheilostomata. From these results inferences are made about the evolutionary history of thisorder and proposed morphological hypotheses are discussed in light of the independent evidence gainedfrom the molecular data.

Infraorder Ascophorina was demonstrated to be non-monophyletic, and there appears to be multipleorigins of the ascus and associated structures involved in lophophore extension. This was further sup-ported by the lack of monophyly within each of the four ascophoran grades (acanthostegomorph/spino-cystal, hippothoomorph/gymnocystal, umbonulomorph/umbonuloid, lepraliomorph/lepralioid) definedby frontal-shield morphology. Chorizopora, currently classified in the ascophoran grade Hippothoomor-pha, is phylogenetically distinct from Hippothoidae, providing strong evidence for multiple origins ofthe gymnocystal frontal shield type. Further evidence is produced to support the morphological hypoth-esis of multiple umbonuloid origins of lepralioid frontal shields, using a step-wise set of topologicalhypothesis tests combined with examination of multi-locus phylogenies.

� 2011 Elsevier Inc. All rights reserved.

1. Introduction

The phylum Bryozoa consists of aquatic, colonial invertebratespredominantly found in the marine environment. They have a glo-bal distribution with roughly 6000 known living species and a fur-ther 15,000 fossilised species, currently classified entirely onmorphological characteristics. Despite non-skeletal characteristicssuch as tentacle number, behavioural and reproductive attributesand colony colour being reported to vary among species, theseare often poorly described and skeletal characteristics have be-come the focus for classification (McKinney and Jackson, 1989).This is partly because soft-part characters in cheilostomes are bestdetermined in living or histologically prepared specimens, and are

ll rights reserved.

nces, University of Auckland,42, New Zealand. Fax: + 64 9

night), d.gordon@niwa.co.nz

typically poorly preserved in ship-collected material that is bulk-fixed or dried. Also skeletal characteristics are the only reliableones detectable in fossil specimens and are thus useful for secularclade comparisons (McKinney and Jackson, 1989). Further, thereare generally enough skeletal characters in most cheilostomeclades to allow cladistic analysis (e.g. Gordon et al., 2002); howeverdetermining which morphological characteristics are taxonomi-cally informative is often difficult. Variation in zooidal morphologyhas been observed in single colonies, which, in some species, maybe induced by different environmental conditions or threats ofpredation, adding to the difficulties of classification via traditionaltaxonomy (Jackson and Cheetham, 1990; Schwaninger, 1999; Yag-unova and Ostrovsky, 2008). In a recent study on the taxonomy ofthe cheilostome genus Macropora, difficulties were encounteredfrom having only a few differences in morphological characteristicscompared to species number, making it difficult to obtain a phylog-eny that was not susceptible to small changes in the characteristicsused (Gordon and Taylor, 2008). This raises questions about howmuch variation and what kinds of variation are important in

352 S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362

species identification and inference of evolutionary relationships,and indicates a clear need for analysis independent of morpholog-ical characteristics.

One approach is to use DNA sequencing technologies to analysethe molecular characteristics of different species. While analyses ofthis type would be greatly beneficial in untangling the evolution-ary relationships within this phylum, problems are also encoun-tered using these techniques. The cheilostome radiation wasrapid, beginning in the mid-Cretaceous and resulting in this ordernow being the most abundant and diverse of the bryozoan orders(Jablonski et al., 1997). When radiations occur within a short timeperiod, there is little time for lineage sorting to take place before asubsequent divergence. What mutations do occur along the shortinternodes may then be masked by substitution saturation alongthe long terminal branches, making it difficult to untangle the trueorder of divergences leading to the extant lineages (Whitfield andLockhart, 2007). By combining the strengths from both morpholog-ical and molecular analyses, a clearer picture of cheilostome evolu-tion can be uncovered.

Phylum Bryozoa nominally occurs in the late Cambrian (Land-ing et al. 2010), but the evidence presented is equivocal. It isknown with certainty from the earliest Ordovician (several orders).There are three recognised classes – Phylactolaemata, Stenolae-mata and Gymnolaemata – with the focus of our study being onthe Gymnolaemata (see Fig. 1 for the current levels of classifica-tion). Stenolaemata, with all members calcified, is the earliest re-corded class of bryozoans and dominated the bryozoan faunauntil the Permian–Triassic boundary when almost all Stenolaematagroups were rendered extinct; survivors of four orders continuedinto the Triassic (Schäfer and Fois, 1987), with only the Cyclosto-mata persisting to the present-day (Ryland, 1970; McKinney andJackson, 1989; Fuchs et al., 2009). Gymnolaemates dominate mod-ern bryozoan faunas and mostly comprise marine, skeletizedforms. The class is currently generally divided into two orders –the Ctenostomata, comprising uncalcified species, and Cheilosto-mata, with calcified species. This arrangement has been questionedon both morphological (Silén, 1942; Jebram, 1992) and moleculargrounds (Jebram, 1992; Fuchs et al., 2009). The Cheilostomata ap-pears nested in Ctenostomata, with the first cheilostomes appear-ing in the latest Jurassic, having ctenostome ancestry (Taylor, 1990,1994; Todd, 2000).

Early cheilostomes were of anascan grade (i.e. lacking an as-cus) with a non-calcified frontal membrane (McKinney and Jack-son, 1989; Jablonski et al., 1997; Dick et al., 2009). Emergingsubsequently, ascophoran bryozoans are defined by the presenceof a flexible sac called an ascus which is located beneath the

Fig. 1. A simplified representation of the hypothesised r

calcareous frontal shield of the zooid and fills and empties withwater, acting as a hydrostatic system for extending the lopho-phore (Dick et al., 2009). The development of the ascus allowedfor constraints on the calcification of the frontal wall to be over-come, resulting in the development of (in order of appearance inthe fossil record) spinocystal, umbonuloid, gymnocystal and lep-ralioid frontal shields (Gordon and Voigt, 1996; Dick et al.,2009). These frontal shield types are what define the four gradeswithin Ascophorina.

It is now thought that frontal shields observed within Ascopho-rina may have arisen on multiple occasions from different ances-tors, suggesting that these groups, including Ascophorina itself,are not monophyletic (Gordon and Voigt, 1996; Gordon, 2000).Multiple models of frontal wall or shield evolution in cheilostomebryozoans have been inferred from morphological investigations(Gordon, 2000); however these are yet to be suitably tested byindependent means of molecular analysis. One of these models in-cludes the hypothesis that lepralioid frontal shields arose multipletimes from umbonuloid-shielded ancestors (Gordon, 2000); umbo-nuloid and lepralioid shields have a similar frontal appearance infully differentiated zooids and are often confused on initial exam-ination (Gordon and Voigt, 1996). This hypothesis suggests thatreduction and subsequent loss of the umbonuloid-shield compo-nent and invagination of the ascus could allow lepralioid shieldsto be relatively easily derived (Gordon, 2000). There is preliminarymolecular evidence supporting this hypothesis, based on the 18SrRNA gene, with the placement of the lepraliomorphs Celleporinaand Schizomavella with the umbonulomorphs Escharoides andUmbonula (Tsyganov-Bodounov et al., 2009), although furtherinvestigation with larger taxon sampling and across a wider rangeof loci is required to verify these findings.

Molecular analyses investigating the internal relationshipswithin Bryozoa are only in their preliminary stages but are becom-ing more common (e.g. Fuchs et al., 2009; Tsyganov-Bodounovet al., 2009); however, these are limited in their taxonomic cover-age when the age and diversity of Bryozoa are taken into account,and few studies have focussed specifically on diversification withinthe order Cheilostomata. A study of the Cheilostomata using 16SrRNA gene sequences (Dick et al., 2000) revealed that the groupAscophora (infraorder Ascophorina) appears to be polyphyletic,as suggested by Gordon (2000) from morphological data, withthe genus Celleporella being placed amongst anascans. Subsequentphylogenies constructed from the 16S rRNA gene for 40 cheilos-tome species from Australia, China and the United States of Amer-ica have indicated discrepancies with the morphologicalclassifications at higher taxonomic levels (Hao et al., 2005).

elationships within Bryozoa (Hausdorf et al., 2010).

S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362 353

More recently, two independent molecular studies of the high-er-order relationships among bryozoans have been published.Fuchs et al. (2009) used a combined analysis of cytochrome oxi-dase subunit I (COI), 18S rRNA and 28S rRNA genes from 32 speciesacross 23 families within Bryozoa. These findings revealed that thegymnolaemate orders Cheilostomata and Ctenostomata do not ap-pear to be monophyletic groups (Fuchs et al., 2009). This has pre-viously been suggested from the analysis of morphologicalcharacteristics such as the structure and development of theparietal muscles and related features as well as larval morphology(Jebram, 1992; Santagata, 2008). However, another recent molecu-lar study based on a structural alignment of the 18S rRNA gene byTsyganov-Bodounov et al. (2009) showed Cheilostomata to bemonophyletic within the paraphyletic Ctenostomata. The samegenera from Ctenostomata were used in both of these studies withthe exception of Triticella used by Fuchs et al. (2009), suggestingthis difference in results is not simply due to limited taxon sam-pling in one study. Further investigation into these relationshipsusing a wider variety of taxa and/or loci is required to resolve suchdiscrepancies.

New Zealand is considered to be a bryozoan biodiversity hot-spot, with 953 known marine species, 61% of which are endemic(Gordon et al., 2009). The morphological taxonomy of New Zealandspecies is well known, with about two-thirds now fully described.The 35 New Zealand species sequenced in this study add consider-ably to the taxonomic coverage of cheilostome molecular phyloge-nies covering 25 families (nine of which have not previously beenrepresented), 13 superfamilies and three infraorders within the or-der Cheilostomata. Here we present the first analysis of thehypothesised evolutionary relationships within the Cheilostomata.Hypotheses directly tested include multiple origins of the ascus,multiple origins of the four ascophorine frontal shield types andmultiple umbonuloid origins of lepralioid frontal shields.

2. Materials and methods

2.1. Sample collection and DNA extraction

Samples of 15 species were collected by D. Gordon in Welling-ton Harbour, New Zealand in 2006, with further sampling in June2008 at Mathesons Bay and Northland, in addition to Tor Bayand North Head in Auckland, New Zealand, yielding another 43samples. Combined, these samples represent 25 families, 13 super-families and three infraorders of the order Cheilostomata (Table S1in Supplementary material). Collection involved scraping or cut-ting portions of the cheilostome colonies (and one cyclostome col-ony collected from Mathesons Bay) from various substrata andstoring them in 95% ethanol. These were identified to species byD. Gordon, using conventional light microscopy. The samples werethen refrigerated at 4 �C in 95% ethanol until DNA extraction. Vou-cher specimens were retained for each of the samples and are lo-cated at the School of Biological Sciences at the University ofAuckland (available upon request).

DNA was extracted using a modified phenol/chloroform/iso-amyl alcohol (PCI) protocol adapted from Milligan (1998). Roughly2 mm3 of tissue (4–5 individual zooids) from encrusting specieswere crushed into a fine powder using a micropestle within a1.5 ml tube containing RSB buffer and SDS. Proteinase K to a finalconcentration of 1 mg/ml was added and the sample was incu-bated at 55 �C overnight for digestion. The homogenate was thenextracted once with phenol, once with PCI (phenol–chloroform–isoamyl alcohol at a ratio of 24:24:1) and finally with CI (chloro-form–isoamyl alcohol at a ratio of 24:1). DNA precipitation usingethanol was carried out with the samples sitting at �20 �C

overnight. The extracted DNA was washed with 70% ethanol, air-dried and finally resuspended in TLE and stored at �20 �C.

2.2. PCR conditions and sequencing

Sequencing was attempted for five loci from each sampleusing the primers outlined in Table S3, which include new prim-ers designed from sequence alignments. These loci included 16S,cytochrome oxidase subunit I, 18S, 28S and elongation factor1a, providing a variety from both mitochondrial and nuclear ori-gins with varying evolutionary rates. Owing to the high diversityamong species, the chemical and temperature conditions of thePCR were not universal across all samples for each locus. PCRreactions used 15 ng of extracted DNA in a standard 25 ll mixcontaining PCR buffer, 1.5–2.5 mM magnesium chloride, 0.16–0.2 mM dNTPs, 0.4–0.8 mM of both the forward and reverse prim-ers, 0.1–0.25 U of Taq ti polymerase (Fisher Scientific) and dis-tilled water to volume. Manipulations of the magnesium, bufferand primer concentrations were performed for some samples. Cy-cling conditions were 94 �C for 3 min; 35 cycles of 94 �C for 30 s,45–60 �C for 40 s, and 72 �C for 40 s; and 72 �C for 7 min. Theproducts were visualized on a 1.6% agarose gel, stained with ethi-dium bromide.

Samples were prepared for sequencing via the SAPEX protocol(using shrimp alkaline phosphatase and Exo I) modified from Wer-le et al. (1994). Sequencing reactions were run using a 1/8 dilutionof BigDye version 3.1 dye terminator mix (Applied Biosystems).Dye-terminator removal was performed using CleanSEQ (Agen-court) and a magnetic SPRIPlate 96R. Sequencing products wererun on a 3130XL capillary sequencer (Applied Biosystems) at theSequencing Facility at the University of Auckland. Sequences wereimported into Geneious (Drummond et al., 2009) for editing andassembly. All sequences were subjected to similarity searchesusing BLAST at the NCBI database and/or FastA at the EuropeanBioinformatics Institute (http://www.ebi.ac.uk/Tools/fasta/in-dex.html) to screen for contamination.

Additional sequences were selected from the GenBank databaseto broaden taxonomic and geographic coverage (see Tables S1 andS2). Searches were performed across all loci through the NCBIsearch function (found at http://www.ncbi.nlm.nih.gov/). Se-quences from cheilostome species that were available across therange of target loci were selected. Additional sequences from theorder Ctenostomata were also added based on results from Fuchset al. (2009) suggesting that Cheilostomata is not monophyletic.

2.3. Alignments

Sequences for each locus were aligned using ClustalW (Thomp-son et al., 1994) as implemented in Geneious (Drummond et al.,2009) and were edited manually. All alignments were trimmedto remove long terminal gaps.

The COI and EF1a regions are both protein coding genes and thetranslated amino acid sequences were used to assist alignment ofthe nucleotide sequences. The EF1a region was amplified as an in-tron flanked by two coding regions, but this intron was discardedas it was too variable to confidently align.

Consideration of rRNA secondary structure has been shown toimprove estimates of phylogenies (Telford et al., 2005). Tsyga-nov-Bodounov et al. (2009) aligned bryozoan 18S sequences basedon the structural properties of the region. This alignment was ob-tained from the authors, and the sequences used in this analysiswere aligned against this to allow for variation due to structureand not necessarily sequence composition. Unfortunately, verifiedstructural alignments of the 16S and 28S were not available forbryozoans; therefore, these were aligned computationally usingClustalW (Thompson et al., 1994) and edited by eye.

354 S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362

2.4. Substitution saturation

Substitution saturation tests were performed in DAMBE (Xiaand Xie, 2001) using the method developed by Xia et al. (2003)for the protein coding genes COI and EF1a. This method calculatesan index of substitution saturation (Iss) and tests it with standardstatistical tests against critical values (Iss.c) obtained from com-puter simulations (Xia et al., 2003). For the COI locus, Iss was signif-icantly lower than that of Iss.c for the 1st and 2nd positions,indicating little saturation at these positions. However, for the3rd codon position, Iss was significantly larger than Iss.c, suggestingthat this position held no useful phylogenetic signal. Therefore, thisposition was removed from further analysis, as similarities be-tween sequences could not be considered homologous (Lemeyet al., 2009). This reduced the length of the COI sequence for phy-logenetic analysis from 631 base pairs to 420 base pairs.

EF1a showed a different pattern. For the first and second codonpositions, Iss was smaller than Iss.c, although not significantly, sug-gesting that substantial substitution saturation had occurred atthese positions. For the 3rd codon position, Iss was significantly lar-ger than Iss.c, suggesting, as with COI, that this position held no use-ful phylogenetic signal. Since these results indicated that all threepositions did not contain a reliable phylogenetic signal, the aminoacid sequence of this region was used instead in subsequentanalyses.

2.5. Model selection

Likelihood scores for different models of nucleotide substitutionwere calculated for the final alignment of each locus through thejModelTest program (Guindon and Gascuel, 2003; Posada, 2008).The AIC selection strategy was then used to determine the mostappropriate model of nucleotide substitution and these were usedin the further analysis of each region. For protein sequences, anequivalent program for protein models called ProtTest (Abascalet al., 2005) was used.

The best-fit model determined by jModelTest for 16S, 18S andCOI was GTR (Tavare, 1986) with both a proportion of invariablesites (I) and variation among sites (G). For 28S, the best-fit modelwas HKY (Hasegawa et al., 1985) with a proportion of invariablesites (I) and variation among sites (G). The best-fit model for theEF1a amino acid sequences determined by ProtTest was the Day-hoff model, again with a proportion of invariable sites (I) and var-iation among sites (G).

2.6. Phylogenetic analysis

Both maximum likelihood and Bayesian approaches were usedto evaluate each of the individual loci separately. The outgrouptaxon Plagioecia patina (order Cyclostomata) was used for all lociwith the exception of EF1a as this sequence was not available,and Plumatella repens (Class Phylactolaemata) was used instead.

Maximum likelihood analyses were conducted with PhyML(Guindon and Gascuel, 2003) as implemented within Geneious(Drummond et al., 2009), using the best-fit models of nucleotidesubstitution detected by jModelTest. Support for nodes was esti-mated by analyzing 1000 bootstrap pseudoreplicates for eachlocus.

Bayesian analyses were conducted with MrBayes, again usingthe best-fit models of nucleotide substitution detected with jMod-elTest. Four chains were run for each locus, with one chain heatedat a setting of 0.2. Ten million generations were run for each anal-ysis, with the first 2.5 million generations (25%) discarded as burn-in. Trees were sampled every 5000 generations. Here we presentonly the Bayesian results from the 16S rRNA gene, as it is the mostcomplete and well-supported tree, but all other single-locus trees

(including the maximum likelihood analyses) are available inKnight (2010).

The sequence data were also examined using a partitioned mul-ti-locus analysis implemented in the software package BEAST(Drummond and Rambaut, 2007). This analysis uses a coalescentapproach and provides the ability to analyze multiple data parti-tions simultaneously while allowing for the parameters of the sub-stitution model, the rate model among sites, the rate model amongbranches, the tree, and the tree prior to be ‘linked’ or ‘unlinked’ be-tween the different gene partitions (Drummond and Rambaut,2007). This provides the single best tree to be estimated from inde-pendent and potentially conflicting loci, without the need to sim-plify any assumptions in the analysis. EF1a and 28S wereexcluded from these multi-locus analyses due to limited taxon cov-erage and short sequence lengths. The 16S, 18S and COI alignmentswere analyzed, with independent parameter sets for each of thegene partitions. For each run the substitution models and clockmodels were unlinked, allowing for independent estimates foreach locus, and the trees were linked, causing BEAST to produceone consensus tree based on the data across all three loci. The sub-stitution models determined using jModelTest (Posada, 2008) wereimplemented independently for each locus, and the clock modelwas set to uncorrelated lognormal for each locus. The tree priorwas set to ‘Speciation – Yule Process’, as the data deal with smallnumbers of different species rather than multiple representativesfrom the same species. The chain length was set to 400 million,with the program logging the data every 40,000 iterations, whichresulted in the retention of 10,000 trees. At the completion of therun the log file was checked in Tracer to ensure the MCMC chainhad run for a sufficient length of time. The tree data produced weresummarized using the program TreeAnnotator with a burn-in pro-portion of 20% or 2000 trees.

Two datasets were used in the multi-locus analyses. The first,full dataset contained all 56 species in which sequences from atleast two of the three loci (16S, 18S and COI) were available. Inthe second restricted dataset (44 species), all species were requiredto have both 16S and 18S but a missing COI sequence was allowed.This was decided upon due to the larger length of both the 16S and18S compared to COI, as well as the better resolution these treesgave when analysed independently (not presented here; seeKnight, 2010), giving these sequences more weight in the analysisand therefore adding a larger bias to the results if these regionswere missing. The removal of species missing either the 16S or18S regions did not have any major consequences on the taxo-nomic breadth of the analysis (Table S1). The COI sequences(where available) provide additional phylogenetic information par-ticularly at the species and genus levels.

2.7. Hypothesis testing

Three hypotheses formulated from prior morphological analy-ses (Gordon, 2000) were tested using a range of constrained treetopologies (Table 1). The constrained topologies were constructedin BEAST and tested with Bayes factors calculated from compari-sons with the unconstrained analyses (Kass and Raftery, 1995;Suchard et al., 2001). The more positive the value, the higher thesupport for the unconstrained topology over the constrained topol-ogy. Values from 0 to 0.5 show no evidence of a difference betweenthe topologies, 0.5–1 shows substantial evidence, 1–2 showsstrong evidence and >2 shows decisive evidence for a differencebetween the topologies (Kass and Raftery, 1995; Suchard et al.,2001). For a record of the species included in these hypothesistests, refer to Table S1.

The first hypothesis of multiple origins of the ascus was testedby confining Ascophorina to monophyly (topological constraint (a)in Table 1). Ascophorina is partly defined by the presence of an

Table 1The morphological hypotheses tested and the topological constraints used to do so.

Hypothesis Topological constraints

1. Multiple origins of the ascus a. Monophyly of the infraorderAscophorina

2. Multiple origins of the fourascophorine frontal shieldtypes

b. Monophyly of the ascophorine gradeHippothoomorpha

c. Monophyly of the ascophorine gradeUmbonulomorphad. Monophyly of the ascophorine gradeLepraliomorphae. Monophyly of the ascophorine gradeAcanthostegomorpha

3. Multiple umbonuloid originsof lepralioid frontal shieldtypes

f. Monophyly of the ascophorine gradeLepraliomorpha withinUmbonulomorpha (Fig. 2)g. Monophyly of the combined gradesLepraliomorpha and Umbonulomorphatogether (Fig. 2)

S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362 355

ascus and if this constraint of monophyly can be rejected, it can beassumed that this feature has arisen on more than one occasion orhas been lost by subsequent lineages. A similar approach was takento test the hypothesis of multiple origins of the four ascophorinefrontal-shield grades. If each grade is monophyletic, then eachfrontal shield type clearly had a single, distinct origin. If mono-phyly of a grade is rejected, it indicates that its frontal shield typehas either arisen more than once, or has evolved into another of thefour grades in one or more of the descendent lineages. In eithercase, rejection of monophyly would bring into question the currentassumption of single, distinct origins of the four grades. Possibleparaphyletic relationships require further investigation of thebest-supported phylogeny.

The hypothesis of multiple umbonuloid origins of lepralioid-shielded bryozoans was also tested using a stepwise procedure.Firstly, if this hypothesis is true, neither grade would be expectedto be monophyletic as multiple lepralioid clades would be embed-ded within Umbonulomorpha. Therefore the monophyly of eachgrade was tested (topological constraints (c) and (d) in Table 1).

Fig. 2. A diagrammatic representation of topological constraints (f) and (g) (TableUmbonulomorpha (blue). The two grades are then further constrained to monophyly togcombines Lepraliomorpha and Umbonulomorpha and constrains them to monophyly asother. Again the rest of the species are left to vary without constraints. (For interpretatioversion of this article.)

Subsequently, a single origin of lepralioid frontal shields from anumbonuloid ancestor was examined by constraining Lepraliomor-pha to monophyly within Umbonulomorpha (Fig. 2; topologicalconstraint (f) in Table 1). If this constraint can be rejected it wouldsuggest that a single origin of lepralioid shields from umbonuloidshields would be highly unlikely. Finally, to examine the close phy-logenetic relationship between Lepraliomorpha and Umbonulo-morpha these two grades were confined to monophyly as awhole, with no constraints on the relationships within these twogrades (Fig. 2; topological constraint (g) in Table 1).

The hypothesis tests (with topological constraints (a)–(f)) werefirst conducted on the dataset where only missing COI data waspermitted. This dataset was chosen for the hypothesis tests fortwo main reasons. Firstly, the COI locus is known to be a poor pre-dictor of higher taxonomic relationships (Ekrem et al., 2006) whilethe 16S and 18S loci are considered to have moderate and slowevolutionary rates respectively. This results in the 16S and 18S locibeing more reliable in their inference of higher taxonomic relation-ships which is the focus of the proposed hypotheses. Secondly, thesequences obtained from the 16S and 18S loci were of a longerlength than the COI locus (particularly once the 3rd codon positionwas removed due to evidence of substitution saturation). Thereforethis reduces the impact of a missing locus. While these initialhypothesis tests proved valuable (see Section 3.4), it was decidedthat more could be gained by examining a more restricted dataset.A very small number of species (four species; see Table S1) repre-sented in the above analysis were placed well away from othermembers of their taxonomic classifications. It was of interest toexamine if the remaining species supported the hypotheses if theseanomalous members were excluded from the analyses. This pro-vided a more conservative dataset that would ensure any conclu-sions drawn from the hypotheses tests were not the result of afew anomalous species. The samples removed included Chorizo-pora brongniartii (CbroTB01) due to its disjunct placement fromother members of Hippothoomorpha, Escharoides angela (EAMB01)and Arachnopusia unicornis (Bry14UnkB) due to their disjunctplacement away from other members of Umbonulomorpha, andCalyptotheca immersa (CIMB01 and CIMB02) again due to its disjuctplacement away from other members of Lepraliomorpha (Fig. 5).After the removal of these samples from the data set, another

1). Constraint (f) has Lepraliomorpha constrained to monophyly (green) withinether. The rest of the species are left to vary without any constraints. Constraint (g)a whole, with no restrictions on the placement of these groups in respect to eachn of the references to colour in this figure legend, the reader is referred to the web

Table 2The number and length of both novel sequences obtained from the New Zealandsamples and sequences chosen from the GenBank database for further analysis as wellas the number of species these sequences cover. The COI locus was reduced to 420 bpafter the results of substitution saturation discussed in Section 2.4.

Loci Length(bp)

Newsequencesobtained

Speciescoverageof newsequences

SequencesfromGenBank

Totalspeciescoverage

Totalnumberofsequences

16S 620 52 35 49 77 10118S 663 55 33 51 81 106COI 631

(420)41 27 38 60 79

EF1a 496 23 13 6 18 2928S 353 35 27 19 45 54

356 S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362

unconstrained tree was constructed in BEAST, along with treeswith topological constraints affected by the excluded species (con-straints (b), (c), (d), (f) and (g)). This time a reduced chain length of300 million was used but 10,000 trees were still sampled. The tracefiles of each analysis were checked to ensure the MCMC chain hadrun for a sufficient length.

3. Results

3.1. Samples and sequencing

DNA for this analysis was extracted from 58 specimens fromNew Zealand. There was variable success in the amplification andsequencing of each of the specimens across the different loci withthe 18S region being the most successful with 55 sequences of663 bp in length obtained (Table 2; see Table S1 in Supplementarymaterial for a comprehensive list of sequences obtained). Furthersequences were obtained from GenBank, as described earlier, to in-crease the taxonomic and geographic coverage of the speciesanalysed.

Owing to issues in the amplification of some samples across thedifferent loci, new primers were designed in an attempt to over-come this (Table S3 in Supplementary material). These were allable to be matched with previously designed primers to providea variety of combinations. Each of these newly designed primerswas successful with all of them yielding clean products from atleast a few troublesome species. This was particularly prominentfor the COI region where the most reliable combination of primerswas the newly designed primer F4bryCOI with HCO2198. However,since these primers were all designed from the alignments of pre-viously published sequences, their target regions were withinthose of previously designed primers, resulting in the productsbeing shorter. Also, limited success was had with the primerF1bry18S designed to amplify a larger segment of the 18S locus,with a large number of samples still failing.

3.2. Single-locus analyses

Among the phylogenetic trees constructed from each locus,some differences were observed in both the placement of speciesand the level of lineage support. This is expected since each locushas a different evolutionary rate and therefore provides differingdegrees of support for different hierarchical levels within the phy-logenetic analysis. Only the 16S Bayesian analysis is presented here(Fig. 3), as it conveys the most complete single-locus results. (Allsingle-locus analyses are available in Knight (2010).) Consistenciesacross the loci were observed, with Chorizopora being phylogenet-ically distinct from the rest of Hippothoomorpha and a great pro-portion of Umbonulomorpha and Lepraliomorpha being closelyrelated. The Bayesian analysis of the 16S locus shown in Fig. 3

demonstrates these patterns. Some sequences stand out as beingvery anomalous, and were felt likely to be erroneous in speciesidentification, DNA sequence or alignment. These sequences weresubsequently removed from further analyses to ensure inferenceswere conservative and to avoid inappropriate conclusions beingdrawn from likely misleading data. For example, Parasmittina sp.has an extraordinary branch length (Fig. 3), indicating there maybe doubt about the validity of the sequence or its alignment. Thesequences removed for subsequent analyses are highlighted inTable S1 in Supplementary material, and we feel that they requirefurther validation before incorporation in phylogenetic analyses.

3.3. Multi-locus analyses

These analyses provide much higher support for the inferredclades when compared to the single-locus analyses, highlightingthe strengths of these approaches. Many consistencies are seen be-tween the phylogenetic relationships produced by the two datasetsused for the multi-locus analyses (Figs. 4 and 5). The full datasetcontains all 56 species with sequences from any two of the 16S,18S and COI loci (Fig. 4) while the restricted dataset is confinedto 44 species with sequences at both the 16S and 18S loci and al-lows for missing sequences only from the COI locus (Fig. 5).Although not as taxonomically complete, the tree constructed fromthe second dataset is likely to be more reliable. Within Cheilosto-mata the infraorder Ascophorina is consistently polyphyletic withmembers from Hippothoomorpha and Acanthostegomorpha inthe same lineage as Flustrina. The ascophorine family Hippothoi-dae is monophyletic (posterior probability = 1); however theassociated grade Hippothoomorpha is not, with the family Chorizo-poridae (represented by C. brongniartii) being distinct from Hippo-thoidae. The majority of umbonuloid- and lepralioid-shieldedbryozoans are seen to fall in the same lineage, with the exceptionof the umbonuloid-shielded bryozoan A. unicornis, which insteadfalls into a clade otherwise containing species from the infraorderFlustrina (Fig. 4). Suborder Malacostegina is consistently mono-phyletic, with Membranipora and Electra falling together.

Notable differences were also found between the two datasets.Steginoporella magnifica, which had been grouped with speciesfrom the suborder Malacostegina in the full analysis (Fig. 4), shiftedin the restricted analysis into the clade representing the majorityof the ascophorine umbonuloid- and lepralioid-shielded bryozoanswith posterior support of 0.81 (Fig. 5). The branch length, however,is relatively long making this placement uncertain.

3.4. Hypothesis testing from topological constraints on multi-locusdata

The results from the two sets of hypothesis tests are given in Ta-bles 3 and 4. Each table shows the Log10 Bayes factor calculated be-tween the tree constructed with the monophyletic topologicalconstraint in question and the unconstrained analysis.

The first set of hypothesis testing was undertaken on the samedataset used to construct the unconstrained tree in Fig. 5. As seenin Table 3, the unconstrained analysis is a significantly more likelytopology (given the data provided) than all but one of the con-strained topologies, allowing the monophyly of these groupingsto be rejected. The exception is the Acanthostegomorpha, whosemonophyly is more likely than the topology shown in the uncon-strained analysis (Fig. 5).

The second set of hypothesis testing was run on a reduced data-set with a small number of anomalous species removed. Thesetests show if the rejection of monophyly in the previous testswas driven entirely by these few anomalous species or if the lackof monophyly is pervasive in these groups. As seen in Table 4, thisnow results in monophyly not being rejected for the ascophoran

Fig. 3. The topology constructed from the Bayesian analysis of the 16S locus. Green = Ascophorina, red = Flustrina, orange = Malacostegina, blue = Ctenostomata andblack = Cyclostomata (outgroup). The values near the nodes are Bayesian probabilities and the scale is the expected substitutions per base. (For interpretation of thereferences to colour in this figure legend, the reader is referred to the web version of this article.)

S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362 357

grade Hippothoomorpha (Log10 Bayes factor = 0.25) and the com-bined ascophoran grades Lepraliomorpha and Umbonulomorpha(Log10 Bayes factor = �0.03), suggesting that these topologies arepossible explanations of the data.

4. Discussion

The results from this study provide valuable insight into theevolutionary history within the bryozoan order Cheilostomata.

Fig. 4. The topology constructed from the BEAST analysis on the first set of data containing at least two sequences from 16S, 18S and COI for each species.Green = Ascophorina, red = Flustrina, orange = Malacostegina, blue = Ctenostomata and black = Cyclostomata (outgroup). The values near the nodes are Bayesian probabilitiesand the scale is the expected substitutions per base. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

358 S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362

Fig. 5. The topology constructed from the BEAST analysis for the second data set where each species was required to have both 16S and 18S sequences but allowed missingdata for the COI locus. Green = Ascophorina, red = Flustrina, orange = Malacostegina, blue = Ctenostomata and black = Cyclostomata (outgroup). The values near the nodes areBayesian probabilities and the scale is the expected substitutions per base. (For interpretation of the references to colour in this figure legend, the reader is referred to the webversion of this article.)

S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362 359

Table 3Results from the first set of hypothesis tests. The Log10 Bayes factors reported arefrom tests against the unconstrained topology. Negative values indicate greatersupport for the hypothesis (constrained topology), positive values indicate greatersupport for the unconstrained topology. Values between 0–0.5 indicate no evidence ofa difference, 0.5–1 indicate substantial evidence for a difference, 1–2 indicates strongevidence for a difference and >2 indicates decisive evidence for a difference in thelikelihood of the topologies (Kass and Raftery, 1995). Asterisks indicate a monophy-letic grouping that is not rejected.

Topological constraint (i.e. the monophyly of_) Log10 Bayes factor

a. Ascophorina 21.47b. Hippothoomorpha 7.57c. Umbonulomorpha 100.57d. Lepraliomorpha 95.51e. Acanthostegomorpha� �33.92�

f. Lepraliomorpha within Umbonulomorpha 89.06

Table 4Results from the second set of hypothesis tests excluding anomalous taxa. The speciesremoved from the data set for this set of tests are indicated in Table S1 and in Fig. 5.Asterisks indicate a monophyletic grouping that is not rejected (see Table 3 forinterpretation of Log10 Bayes factors).

Topological constraint (i.e. the monophyly of_) Log10 Bayes factor

b. Hippothoomorpha� 0.25�

c. Umbonulomorpha 61.28d. Lepraliomorpha 63.53f. Lepraliomorpha within Umbonulomorpha 64.04g. Lepraliomorpha and Umbonulomorpha� �0.03�

360 S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362

The genetic sequence data for Cheilostomata is greatly improved inboth size and diversity with 206 novel sequences from 58 speci-mens obtained in New Zealand, adding 33 newly sequenced spe-cies and nine newly sequenced families to the phylogenies of thisgroup. Branch support for the higher taxonomic groups is not par-ticularly high in the single-locus analyses. However, the combina-tion of loci pulls together the strengths of the data and largelyremoves any counterintuitive relationships from the single-locusanalyses. While this does produce more informative relationshipsoverall, it can result in reduced branch support for some clades ow-ing to conflicting signals for them from different loci. There is ahigh likelihood of some conflicting signals between loci for cheilos-tome species, owing to the rapid radiation of this group during themid-Cretaceous (see Section 1). Despite these constraints, strongconclusions can be drawn from the data, detailed below.

4.1. Hypothesis 1: Multiple origins of the ascus

The first hypothesis to be tested was that there are multiple ori-gins of the ascus. The ascus is a defining feature in the InfraorderAscophorina and this group was shown to be polyphyletic in thetopologies of the phylogenetic trees constructed (Figs. 3–5) andwas rejected as being monophyletic in the hypothesis tests (Ta-ble 3), supporting previous studies (Hao et al., 2005; Fuchs et al.,2009; Tsyganov-Bodounov et al., 2009). Polyphyly of Ascophorinais expected when hypotheses of frontal shield evolution are con-sidered. Gymnocystal frontal shields found in Hippothoomorphaare thought to have arisen from a cribrimorph ancestor via thereduction of the costal field and subsequent increase in the sizeof the gymnocyst (Gordon, 2000). This is quite different fromumbonuloid frontal shields, which are thought to have evolvedby the overgrowth of an adventitious kenozooid (a zooid withouta polypide) providing an extra layer over the costal field whichthen became calcified (Gordon and Voigt, 1996; Gordon, 2000).Different again are spinocystal frontal shields, seen in Acanthoste-gomorpha, which form by the fusion of spines. This, combined with

the results obtained by this study, adds to the evidence that therewere multiple origins of the ascophorine ascus and associated skel-etal and tissue features involved in lophophore extension. Recentlyit has been suggested that the origins of an ascus had a high degreeof historical contingency, although it was also considered likely tooccur given the sufficient length of time and strong selective pres-sures of predation (Dick et al., 2009).

4.2. Hypothesis 2: Multiple origins of the four ascophorine frontalshield types

The second hypothesis tested was that each of the four ascoph-orine frontal shield types arose multiple times. While all ascopho-rine grades, except the Acanthostegomorpha, were rejected asbeing monophyletic, the potential for multiple origins of the gymn-ocystal frontal shield was particularly evident. The monophyly ofHippothoomorpha only upon exclusion of Chorizopora brongnartiiin the hypothesis tests, combined with the disjunct placement ofthis species with respect to the rest of Hippothoomorpha through-out all the analyses, provides strong evidence for at least two inde-pendent origins of the gymnocystal frontal shield in these lineages.Chorizoporidae is distinguished by disjunct zooids separated byshort tubes and a semicircular orifice as well as the presence of agymnocystal frontal shield (Gordon, 2000). Members of this familyhave long been associated with Hippothoidae because of thegymnocystal shield, but alternative origins have been put forwardbased on morphological evidence (Powell, 1967; Gordon, 2000).Gordon (2000) proposed a possible calloporid origin of the family.This could not be specifically tested due to the paucity of calloporidsequences that resulted in a single species having data from boththe 16S and 18S loci (as required to be included in the analysis).To thoroughly investigate this relationship a wide diversity of cal-loporid species would be required as well as a larger number ofChorizopora specimens. While the results from this study provideevidence for multiple origins of the gymnocystal frontal shield, fur-ther investigation is required to infer the origins of Chorizopora.

Both Umbonulomorpha and Lepraliomorpha were also rejectedas being monophyletic and are discussed in more detail in Sec-tion 4.3. Monophyly of Acanthostegomorpha could not be rejectedalthough these results are weak. This monophyletic constraint in-volved only three species with just one of these being placed in adifferent lineage from the other two in the unconstrained analysis(Fig. 5). As Acanthostegomorpha (as defined) contains fourteenfamilies from four superfamilies, this test does not have a great de-gree of power and further investigation is required.

4.3. Hypothesis 3: Multiple umbonuloid origins of lepralioid frontalshield types

The third hypothesis tested was multiple umbonuloid origins oflepralioid frontal shields. The results from this study show lep-raliomorphs dispersed amongst members of Umbonulomorpha(Figs. 4 and 5). This polyphyly can be explained morphologicallyby the hypothesis of multiple umbonuloid origins of lepralioidfrontal shields (Gordon, 2000). Molecular evidence from the 18SrRNA gene published by Tsyganov-Bodounov et al. (2009) supportsthis hypothesis, with Celleporidae and Smittinidae being nestedwithin Escharella, Escharoides and Umbonula (all Umbonulomor-pha) and is further supported by the results of this study.

As suggested by their names, Umbonulomorpha is defined bythe presence of an umbonuloid frontal shield and Lepraliomorphais defined by the presence of a lepralioid frontal shield. Both ofthese grades share the synapomorphies of a frontal shield, areolaeand associated hypostegal coelom thought to have been derivedfrom frontally expanded kenozooids (Gordon and Voigt, 1996). Ithas been proposed by Gordon (2000), upon examination of

S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362 361

morphological evidence, that throughout the history of Cheilosto-mata there have been multiple umbonuloid origins of lepralioidfrontal shields. The transition from an umbonuloid to a lepralioidfrontal shield is postulated to have occurred via the concurrentreduction of the umbonuloid component of the frontal shield andincreased invagination of the ascus (Gordon, 2000). The ascuswould then open adjacent to the operculum, explaining the dis-tinct primary orifice and well-defined operculum typical of lep-raliomorphs (Harmer, 1957; Gordon, 1993, 2000). Umbonuloidand lepralioid frontal shield formation differs significantly; how-ever, the end results can appear very similar, leading to the mis-classification of some shield types (Banta and Wass, 1979;Gordon, 1993). In fact it is now apparent that some taxa, such asthose from the family Smittinidae, show components of both typesof frontal shield adding to the evidence of a close relationship be-tween these groups (Gordon, 2000).

Topologies from the independent analyses of the 16S and 18SrRNA genes (the 16S Bayesian topology can be seen in Fig. 3) aswell as those from multi-locus analyses (Figs. 4 and 5) show a clus-tering of umbonuloid- and lepralioid-shielded bryozoans. To fur-ther examine the hypothesis of multiple umbonuloid origins oflepralioid frontal shields, a series of topological constraints weretested (Section 2.7). The monophyly of each grade was testedand, as expected, both were rejected in the first set of tests andagain in the second, even with unusually placed taxa removed (Ta-bles 3 and 4). With this confirmed, a single origin of lepralioid fron-tal shields from an umbonuloid ancestor was examined (Fig. 2) andagain this was strongly rejected in both sets of hypothesis tests(Tables 3 and 4), making a single origin of lepralioid shields fromumbonuloid shields highly unlikely. Finally a topological con-straint for a monophyletic Lepraliomorpha plus Umbonulomorphagroup was examined (topological constraint (g) in Table 1). Thiswas not rejected (Table 4) even though the most likely multi-locustrees (Figs. 4 and 5) do not place them in a monophyletic clade.This test provides evidence for a close association between thesetwo grades and therefore adds support for the morphologicalhypothesis proposed by Gordon (2000). Unfortunately there wereinsufficient representatives from each of the groups in this studyto be able to demonstrate unequivocally that umbonuloid shieldsare basal to lepralioid shields or to conduct more definitive testssuch as which umbonuloid lineages gave rise to which lepralioidlineages. Therefore this does not convincingly show that it wasumbonuloid-shielded bryozoans that gave rise to lepralioid-shielded bryozoans, although fossil evidence suggests that umbo-nuloid shields arose in the Cretaceous prior to the appearance oflepralioid shields in the Paleogene (Gordon and Voigt, 1996). Thereis scope for future tests of this type to clarify the more specific rela-tionships between these grades when sequence data from a largerrange of taxa become available.

5. Conclusions

This study provides the most complete molecular analysis todate of the hypothesised evolutionary relationships within thebryozoan order Cheilostomata. The polyphyly of infraorderAscophorina is demonstrated, suggesting more than one origina-tion of the ascus and associated structures involved in lophophoreextension. This is further supported by the lack of monophylywithin three of the four ascophorine grades defined by frontal-shield morphology. The disjunct relationship of two families ofHippothoomorpha provides evidence for more than one origina-tion of the ascophoran gymnocystal shield. More comprehensiveevidence is also produced to support the morphological hypothesisof multiple umbonuloid origins of lepralioid frontal shields, using astep-wise set of topological hypothesis tests combined with exam-

ination of multi-locus phylogenies. While further sequences fromeach of these speciose groups will allow more refined tests of thephylogenetic relationships among the Cheilostomata, this studyadds greatly to our understanding of the emergence of this order.

Acknowledgments

We would like to thank Dr. Howard Ross from the University ofAuckland for his advice and guidance regarding the phylogeneticanalyses. Additionally we would like to thank Peter Tsai, also fromthe University of Auckland, for his great amount of time and effortin getting the programs required for analysis up and running effi-ciently. Thank you also to Sarah Wyse for her hard work doingthe initial pilot study that led to this research and the MolecularEcology lab at the University of Auckland for their support. Finally,thank you to the University of Auckland for funding this research.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.ympev.2011.07.005.

References

Abascal, F., Zardoya, R., Posada, D., 2005. ProtTest: selection of best-fit models ofprotein evolution. Bioinform. Appl. Note 21 (9), 2104–2105.

Banta, W.C., Wass, R.E., 1979. Catenicellid cheilostome Bryozoa. I. Frontal walls.Aust. J. Zool. 68, 1–70.

Dick, M.H., Williams, L.P., Coggeshall-Burr, M., 2000. Use of mitochondrial DNAsequences to investigate sister-group relationships among gymnolaemataebryozoans. In: Herrera, C.A. (Ed.), Proceedings of the 11th InternationalBryozoology Association Conference. Smithsonian Tropical Research Institute,Balboa, pp. 197–210.

Dick, M.H., Lidgard, S., Gordon, D.P., Mawatari, S.F., 2009. The origin of ascophoranbryozoans was historically contingent but likely. Proc. Biol. Sci. 276, 3141–4148.

Drummond, A.J., Rambaut, A., 2007. BEAST: Bayesian evolutionary analysis bysampling trees. BMC Evol. Biol. 7 (1).

Drummond, A.J., Ashton, B., Cheung, M., Heled, J., Kearse, M., Moir, R., Stones-Havas,S., Thierer, T. and Wilson, A., 2009. Geneious v4.7. <http://www.geneious.com>.

Ekrem, T., Willassen, E., Stur, E., 2006. A comprehensive DNA sequence library isessential for identification with DNA barcodes. Mol. Phylogenet. Evol. 43, 530–542.

Fuchs, J., Obst, M., Sundberg, P., 2009. The first comprehensive molecular phylogenyof Bryozoa (Ectoprocta) based on combined analyses of nuclear andmitochondrial genes. Mol. Phylogenet. Evol. 52 (1), 225–233.

Gordon, D.P., 1993. Bryozoan frontal shields: studies on umbonulomorphs andimpacts on classification. Zool. Scr. 22 (2), 203–221.

Gordon, D.P., 2000. Towards a phylogeny of cheilostomes – morphological modelsof frontal wall/shield evolution. In: Herrera Cubilla, A., Jackson, J.B.C. (Eds.),Proceedings of the 11th International Bryozoology Association Conference.Smithsonian Tropical Research Institution, Balboa, pp. 17–37.

Gordon, D.P., Taylor, P.D., 2008. Systematics of the bryozoan genus Macropora(Cheilostomata). Zool. J. Linn. Soc. 153, 115–146.

Gordon, D., Voigt, E., 1996. The kenozooidal origin of the ascophorine hypostegalcoelom and associated frontal shield. Bryozoans in space and time. In: Gordon,D., Smith, A., Grant-Mackie, J. (Eds.), Proceeding of the 10th InternationalBryozoology Conference. National Institute of Water and Atmospheric ResearchLtd., Wellington, New Zealand, pp. 89–107.

Gordon, D.P., Mawatari, S.F., Kajihara, H., 2002. New taxa of Japanese and NewZealand Eurystomellidae (Phylum Bryozoa) and their phylogeneticrelationships. Zool. J. Linn. Soc. 136, 199–216.

Gordon, D.P., Taylor, P.D., Bigey, F.P., 2009. Phylum Bryozoa: moss animals, seamats, lace corals. In: Gordon, D.P. (Ed.), New Zealand Inventory of Biodiversity:Kingdom Animalia. Radiata, Lophotrochozoa, and Deuterostomia, vol. 1.Canterbury University Press, Christchurch, NZ.

Guindon, S., Gascuel, O., 2003. A simple, fast and accurate algorithm to estimatelarge phylogenies by maximum likelihood. Syst. Biol. 52 (5), 696–704.

Hao, J., Li, C., Sun, X., Yang, Q., 2005. Phylogeny and divergence time estimation ofcheilostome bryozoans based on mitochondrial 16S rRNA sequences. Chin. Sci.Bull. 50 (12), 1205–1211.

Harmer, S.F., 1957. The Polyzoa of the Siboga Expedition. Part 4. CheilostomataAscophora II. Siboga Expeditie. 28d(i–xv), 641–1147, pls 1142–1174.

Hasegawa, M., Kishino, K., Yano, T., 1985. Dating the human-ape splitting by amolecular clock of mitochondrial DNA. J. Mol. Evol. 22, 160–174.

Hausdorf, B., Helmkampf, M., Nesnidal, M.P., Bruchhaus, I., 2010. Phylogeneticrelationships within the lophophorate lineages (Ectoprocta, Brachiopoda andPhoronida). Mol. Phylogenet. Evol. 55 (3), 1121–1127.

362 S. Knight et al. / Molecular Phylogenetics and Evolution 61 (2011) 351–362

Jablonski, D., Lidgard, S., Taylor, P., 1997. Comparative ecology of bryozoanradiations: origin of novelties in cyclostomes and cheilostomes. Palaios 12,505–523.

Jackson, J.B.C., Cheetham, A.H., 1990. Evolutionary significance of morphospecies: atest with cheilostome Bryozoa. Science 248 (4955), 579–583.

Jebram, D.H.A., 1992. The polyphyletic origin of the ‘‘Cheilostomata’’ (Bryozoa). J.Zool. Syst. Evol. Res. 30, 46–52.

Kass, R.E., Raftery, A.E., 1995. Bayes factors. J. Am. Stat. Assoc. 90 (430), 773–795.Knight, S., 2010. Molecular Phylogenies of the Bryozoa: Testing the Concordance of

Molecular Phylogenies with Morphological Hypotheses on the Evolution of theBryozoan Order Cheilostomata, Masters of Science. School of BiologicalSciences, University of Auckland, Auckland.

Landing, E., English, A., Keppie, J.D., 2010. Cambrian origin of all skeletalizedmetazoan phyla—discovery of Earth’s oldest bryozoans (Upper Cambrian,southern Mexico). Geology 38, 547–550.

Lemey, P., Salemi, M., Vandamme, A.-M. (Eds.), 2009. The Phylogenetic Handbook: APractical Approach to Phylogenetic Analysis and Hypothesis Testing. CambridgeUniversity Press.

McKinney, F.K., Jackson, J.B.C., 1989. Bryozoan Evolution. Unwin Hyman Ltd.,London, UK.

Posada, D., 2008. JModelTest: phylogenetic model averaging. Mol. Biol. Evol. 25 (7),1253–1256.

Powell, N.A., 1967. Polyzoa (Bryozoa) Ascophora from North New Zealand. Discov.Rep. 39, 199–394.

Ryland, J.S., 1970. Bryozoans. Hutchinson University Library, London.Santagata, S., 2008. The morphology and evolutionary significance of the ciliary

fields and musculature among marine bryozoan larvae. J. Morphol. 269, 349–364.

Schäfer, P., Fois, E., 1987. Systematics and evolution of Triassic Bryozoa. Geol.Palaeontol. 21, 173–225.

Schwaninger, H.R., 1999. Population structure of the widely dispersing marinebryozoan Membranipora membranacea (Cheilostomata): implications forpopulation history, biogeography, and taxonomy. Mar. Biol. 135, 411–423.

Silén, L., 1942. Origin and development of the cheilo-ctenostomatous stem ofBryozoa. Ark. Zool. 33A (12), 1–130, 139 pls.

Suchard, M.A., Weiss, R.E., Sinsheimer, J.S., 2001. Bayesian selection of continuous-time Markov chain evolutionary models. Mol. Biol. Evol. 18 (6), 1001–1013.

Tavare, S. (Ed.), 1986. Some Probalistic and Statistical Problems in the Analysis ofDNA Sequences. Some Mathematical Questions in Biology – DNA SequenceAnalysis. American Mathematical Society, Providence, RI.

Taylor, P.D., 1990. Bioimmured ctenostomes from the Jurassic and the origin of thecheilostome Bryozoa. Palaeontology 33, 1–17.

Taylor, P.D., 1994. An early cheilostome bryozoan from the Upper Jurassic of Yemen.Neues Jb. Geol. Paläont., Abhandl. 191, 331–344.

Telford, M.J., Wise, M.J., Gowri-Shankar, V., 2005. Consideration of RNA secondarystructure significantly improves likelihood-based estimates of phylogeny:examples from the Bilateria. Mol. Biol. Evol. 22 (4), 1129–1136.

Thompson, J.D., Higgins, D.G., Gibson, T.J., 1994. CLUSTAL W: improving thesensitivity of progressive multiple sequence alignment through sequenceweighting, position-specific gap penalties and weight matrix choice. Nucl.Acids Res. 22 (22), 2673–4680.

Todd, J.A., 2000. The central role of ctenostomes in bryozoan phylogeny. In: HerreraCubilla, A., Jackson, J.B.C. (Eds.), Proceedings of the 11th InternationalBryozoology Association Conference. Smithsonian Tropical ResearchInstitution, Balboa, pp. 104–135.

Tsyganov-Bodounov, A., Hayward, P.J., Porter, J.S., Skibinski, D.O.F., 2009. Bayesianphylogenetics of Bryozoa. Mol. Phylogenet. Evol. 52 (3), 904–910.

Werle, E., Schneider, C., Renner, M., Volker, M., Fiehn, W., 1994. Convenient single-step, one tube purifiation of PCR products for direct sequencing. Nucl. Acids Res.22, 4354–4355.

Whitfield, J.B., Lockhart, P.J., 2007. Deciphering ancient rapid radiations. TrendsEcol. Evol. 22 (5), 258–265.

Xia, X., Xie, Z., 2001. DAMBE: software package for data analysis in molecularbiology and evolution. J. Hered. 92 (4), 371–373.

Xia, X., Xie, Z., Salemi, M., Chen, L., Wang, Y., 2003. An index of substitutionsaturation and its application. Mol. Phylogenet. Evol. 26, 1–7.

Yagunova, E.B., Ostrovsky, A.N., 2008. Encrusting bryozoan colonies on stones andalgae: variability of zooidal size and its possible causes. J. Mar. Biol. Assoc. UK88 (5), 901–908.