Davalliaceae in Peninsular Malaysia , a preliminary study based on trnL-F region

A preliminary molecular analysis based on trnL-F region is presented for 17 taxa of Davalliaceae in Peninsular Malaysia. Maximum parsimony and Bayesian analysis were conducted on the dataset in order to establish a robust phylogenetic relationship between taxa. The results of analysis indicate incongruence with morphological classification. All genera of Davalliaceae in the study area are paraphyletic except Araiostegia which is represented by only a single species. In addition it partially agrees with recent phylogeny base on rbcL data.


INTRODUCTION
Davalliaceae is a moderate-sized family of ferns with about 50-60 species in four currently accepted genera (Nooteboom 1992, 1994, 1998, Schneider et al. 2002), which is restricted to the Old World tropics and subtropics.In Peninsular Malaysia this family is represented by 17 species in five genera (Parris & Latiff 1997).However, in a recent revision of the family Nooteboom (1992Nooteboom ( , 1994Nooteboom ( , 1998) ) reduced the number of taxa, and lumped all Humata spp.plus Scyphularia spp.into Davallia, which he divided into two sections: sect.Davallia and sect.Scyphularia.
The first molecular study by Tsutsumi & Kato (2005) was based on five continuous chloroplast regions (atpB, rbcL, accd. atpB-rbcL spacer, and rbcl-accd spacer), and indicated that none of the genera in the family was monophyletic: Araiostegia and Davallia were divided into two and three clades respectively, and Humata and Scyphularia were paraphyletic.This finding did not support either the traditional classification of the genus or the division into two sections suggested by Nooteboom (1992Nooteboom ( , 1994)).The present study uses a fast-evolving chloroplast region, trnL-F, to infer the phylogeny of Davalliaceae in Peninsular Malaysia and to test the generic and sectional classification within the group.The suitability of the trnL-F region has already been tested in other studies on fern phylogeny (Schneider et al. 2004a, b, Skog et al. 2004).The present study also has a more complete taxon sampling of Malaysian members of the family than has been used before.

Ingroup sampling
All species of Davalliaceae reported to occur in Peninsular Malaysia by the most recent studies (Parris & Latiff 1997, Nooteboom 1998) have been collected from the field for DNA extraction, except for Humata parvula and Leucostegia pallida which were not found in reported localities during field work.
The DNA material for these two species was supplied by C. Tsutsumi (University of Tokyo, Japan) and from herbarium specimens (see Table 1 for details of specimens).

Outgroup sampling
Previous studies of fern phylogeny (Tsutsumi & Kato 2005, 2006) have reported the genus Oleandra as a sister clade to davallioid ferns.However, Leucostegia was also classified as an outgroup because this genus was originally placed within Davalliaceae but has since been indicated by molecular data to fall outside it (Tsutsumi & Kato 2006).In the present study, the outgroup consisted of three species, i.e., one representative of Oleandra and two of Leucostegia.

DNA extraction, sequencing and alignment
Procedures for extraction, amplification and sequencing in this study followed RBGE molecular lab protocols (Clark & Hollingsworth 2006).The primers used are those published previously for trnL-F region (Taberlet et al. 1991, Trewick et al. 2002).The sequence was aligned manually using MacClade v4.0 (Maddison & Maddison 2003).Ambiguous regions at the ends of the sequences were excluded.Fifteen gaps were coded using simple indel coding and the multistate gap region method (Simmons & Ochoterena 2000, Simmons et al. 2001).

Phylogenetic analysis
For the phylogenetic reconstruction two different types of analyses were performed: Maximum Parsimony (MP) and Bayesian Analysis (BA).We used PAUP v4.0b10 (Swofford 2002) for the reconstruction using Maximum Parsimony.
All characters were treated as unweighted and unordered.Multistate characters were interpreted as uncertain and gaps were treated as missing.
Heuristic searches were performed for all analyses with 10 000 RANDOM addition sequence replicates using TBR with MUL-TREES on, STEEPEST DESCENT off, ACCTRAN-optimization and branches collapsed if minimum branch length is zero.Descriptive tree statistics were given by the consistency index (CI), retention index (RI) and rescaled consistency index (RC).
Branch support analyses were carried out using Bootstrap (Felsenstein 1985) and Decay Indices (Bremer 1988).Bootstrap values were calculated using 10 000 replicates with the same Davalliaceae in Peninsular Malaysia, a preliminary study based on trnL-F region settings as above, except with only 1 RANDOM addition per replicate.Decay Index was calculated with default settings in AutoDecay v4.0 (Eriksson 1999).

Bayesian analysis
Parameters and the evolutionary model for the region was selected with the assistance of Modeltest v3.07 (Posada & Crandall 1998, 2001, Posada & Buckley 2004).The parameter and model based on the Akaike Information Criterion (AIC) was used.For the analyses, four independent Monte Carlo Markov Chains (MCMC) were run simultaneously for 1 million generations, starting with a random tree and with one tree saved every 100 generations.The analyses were also done with the inclusion of the gap matrix.The first 500 trees were discarded, and the burn-in for each run was determined by plotting the log likelihood of the cold chain versus the number of generations in Microsoft Excel.

Sequence alignment and model selection
The aligned trnL-F sequence data matrix contains 924 characters, of which 636 are parsimony informative.The AIC selected model for trnL-F was K81 uf+1 (two transversion-parameter model I unequal frequencies; Kimura (1981)).

Maximum parsimony
In the maximum parsimony (MP) analysis, the analysis including gap characters produced a single most parsimonious tree of 447 steps with CI = 0.88, RI = 0.94 and RC = 0.82.These values are relatively high indicating that the number of homoplastic characters is low.The bootstrap majority tree was less highly resolved, with a major polytomy near the base of the cladogram.Fig. 1 presents the singel most parsimonious tree.The topology in Fig. 1 shows that the ingroup consists of two lineages.The first lineage is formed by Araiostegia hymenophylloides, the second one is weakly supported (bs = 55 %, d = 1) and consists of a trichotomy of two partially resolved groups, a very weakly supported clade (DCII): including Scyphularia -D.solida, a well-supported clade consisting of all representatives of D. trichomanoides and a well-supported clade consisting of a Davallia clade (DC1) and a Humata clade (HC).

Bayesian analysis
The topology of the Bayesian majority rule tree is identical to the MP analysis, except that Scyphularia triphylla is sister to the Davallia clade with a low posterior probability value (pp = 0.76).The posterior probability for the nodes in the phylogeny ranges between 0.59 for the node joining the basal trichotomy to 1.00.

The monophyly of Davalliaceae
The phylogenetic relationships indicated by our analysis of the trnL-F region are incongruent with any previous morphological classification and indicates that neither of the large genera within Davalliaceae, i.e., Humata and Davallia are monophyletic.In this respect it supports the findings of Tsutsumi and others (Tsutsumi & Kato 2005, 2006, Tsutsumi et al. 2008).Davalliaceae (excluding Leucostegia) was shown by our data to comprise five major lineage or groups, namely the Araiostegia hymenophylloides (AC), D. solida, D. trichomanoides, a Davallia clade 1 (DC1) and a Humata clade (HC).The position of Scyphularia triphylla could not be established with any confidence.
The partitioning of clades/groups is almost the same as in Tsutsumi & Kato (2005), although the sister relationship of Araiostegia to all other clades or groups, is only weakly supported with bootstrap value 55 %, decay value 1 and posterior probability 59 %.In both analyses (MP and BA), D. solida and D. trichomanoides form a polytomy with other clades.In the Humata clade, Humata heterophylla is sister to all other Humata species, including D. corniculata, with high support (bs = 99 %, d = 10).As no study has been done on the same gene in other members of Davalliaceae, a further comparison with other data could not be made.
The present study has confirmed that Scyphularia triphylla is nested within Davallia, and should therefore be reduced to Davallia, but as only one Scyphularia species was examined the monophyly of this genus could not be tested.The present study also confirms that neither Davallia nor Humata are monophyletic, and based on this evidence plus that of Tsutsumi & Kato (2005, 2006), these genera need to be merged or redefined.The genus Humata could be made monophyletic by including Davallia corniculata but Davallia would then still not be monophyletic.Before formal taxonomic changes are made, however, it would be valuable to obtain additional data from further DNA regions, in order to confirm beyond doubt that the relationships resolved here reflect biological reality.molecular lab staff in the Department of Botany (Natural History Museum, London), herbaria of KEP, KEW, L, NHM, RBGE, SING and UKMB.The first author also like to thank R. Jaman (UKMB) for support in field work, C. Tsutsumi (University of Tokyo) for DNA material, M. Moller (RBGE) and H. Schneider (NHM) for the molecular analysis and valuable comments.This research was funded by Universiti Kebangsaan Malaysia and Malaysian government.