Boesenbergia longiflora (Zingiberaceae) and descriptions of five related new taxa

Fig. 1. 

Gastrochilus longiflora (B. longiflora). Watercolour on paper by Vishnupersaud (1829) 

from the Wallich Collection, Kew. Reproduced with permission of The Board of Trustees of 

the Royal Botanic Gardens, Kew.

49

Boesenbergia longiflora and related taxa

---

To determine which taxon a 

B. aff. longiflora specimen belonged, the location, habitat, 

plant height, flower colour and underground morphology were important. Here, the 

“Additional specimens examined” under each described species are determined based 

on a combination of these parameters. The name 

B.  longiflora  (see “Taxonomy” 

below) is lectotypified. 

During 2008

-2012, c. 25 live rhizome divisions of 

B.  aff.  longiflora were 

obtained from a variety of sources:  botanical gardens, research institutions and various 

personal collections. Each accession was grown by the first author in Hawaii, USA 

under controlled, shade-house conditions. Details on the morphology and phenology 

were recorded over several growing seasons. DNA samples with a corresponding 

herbarium specimen were taken from each accession. 

In 2009

-

2012 field observations were made in Thailand with botanists from 

BK,  the  Thailand  Institute  of  Scientific  and  Technological  Research  and  Kasesart 

University. Similarly in India, observations were conducted with personnel from the 

Botanical Survey India (Shillong), the Assam Forestry Department and the Nagaland 

University. Digital images of all Boesenbergia 

observed in the field and in cultivation 

were recorded for later reference.

Early in the research, all Boesenbergia species known to have radical 

inflorescences were evaluated for possible inclusion in the B. longiflora complex. Based 

on molecular data (Mood et al., unpublished) B. basispicata K.Larsen ex Sirirugsa, 

B. tenuispicata K.Larsen  and B. trangensis K.Larsen belong to other Boesenbergia 

clades. DNA samples were not available for B. angustifolia (Hallier f.) Schltr., B. 

phyllostachya (Gagnep.) K.Larsen and B. siphonantha (King ex Baker) M.Sabu, 

Prasanthk. & Škorničk. Consequently, their phylogenetic position is undetermined at 

this time.

The final taxonomy presented here represents an incorporation of data from the 

phylogenetic analyses, field observations and detailed study of the cultivated plants. 

The descriptions are ordered with the yellow-flowered species first, followed by the 

white-flowered species. The distribution map (Fig. 20) was constructed by mapping 

types and selected specimens of each species (these are marked by an asterisk following 

the herbarium abbreviations in the list of additional specimens examined). 

Molecular phylogeny. Phylogenetic analyses were conducted to complement and 

assist in taxonomic decisions required to elucidate relationships among species of 

the 

B. longiflora clade as defined here. Taxon selection was based on an unpublished 

broader analysis of plastid trnK data for Boesenbergia relative to other genera of 

Zingiberaceae (authors in prep.). The bulk of the data for non-Boesenbergia samples 

are from the earlier family-wide analyses of Kress et al. (2002) with a few additional 

sequences from recently described genera such as from Kress et al. (2010) and Leong-

Škorničková et al. (2011). The data for the specimens cited in these references can 

be found at http://www.ncbi.nlm.nih.gov/genbank/ (GenBank 2011).

 A representative 

shortest maximum parsimony (MP) phylogram is shown in Fig. 3. Only names of 

relevant Boesenbergia samples are indicated on the phylogram. 

50

Gard. Bull. Singapore 65(1) 2013

---

Fig. 2. Hand-coloured lithograph published in Curtis’s Botanical Magazine 69, t. 4010, 1843 

as Gastrochilus 

longiflora. Reproduced with permission of The Board of Trustees of the Royal 

Botanic  Gardens,  Kew.  (Identified  by  Baker  (1890)  as G. jenkinsii. Described here as B. 

hamiltonii.)

51

Boesenbergia longiflora and related taxa

---

DNA was extracted from silica-dried leaf material or leaf tissue from herbarium 

specimens using standard methods as described in Kress et al. (2002). Thirty-seven 

samples of Boesenbergia were used in the DNA sequencing analyses. Two samples 

of B. pulcherrima (Wall.) Kuntze, the generic type, were designated as outgroup taxa. 

Two other samples, B. plicata var. lurida (Ridl.) Holttum and B. plicata var. plicata 

(Ridl.) Holttum were also included. The remaining samples were all members of the 

B. longiflora clade. All samples were designated by a collection or accession number 

and vouchered (Appendix B).

Both nuclear and plastid data were collected. The nuclear ribosomal ITS (nrITS) 

region was amplified using the 18S-F and 26S-R primers (Prince 2010) and Phusion 

high  fidelity  polymerase  (New  England  BioLabs,  Ipswich,  Massachusetts,  USA) 

with 5X GC Buffer per the manufacturer’s instructions at an annealing temperature 

of 62ºC.  The same primers were used for subsequent sequencing. The plastid trnK 

region was amplified and sequenced using conditions described in Kress et al. (2002) 

or Prince & Kress (2006). Two newly designed primers were synthesized to assist with 

sequencing of some difficult samples (5Fb: CTCTATGGATTTTCAAGGAT and 5Rb: 

AGACCAAAATGAAAATAATA). All samples were direct sequenced on a 3130xl 

Genetic Analyser at Rancho Santa Ana Botanic Garden (Claremont, California, USA).

Noisy sequences (electrophoregrams showing more than a few polymorphic 

sites) in the nrITS data were cloned using TOPO TA cloning kits (Invitrogen, now Life 

Technologies, Grand Island, New York, USA) with four to eight clones sequenced per 

sample. A small number of the trnK sequences were also noisy (due to polymerase 

“stutter” caused by a mononucleotide T repeat around 450bp). Those samples were 

re-amplified using Phusion polymerase, which has been shown to outperform other 

polymerases for this particular problem (Fazekas et al. 2010) and re-sequenced.

Individual sequences of each specimen were edited and a consensus sequence 

was generated in Sequencher v4.9 (Gene Codes Corporation, Ann Arbor, Michigan, 

USA). The consensus sequence was trimmed (18S and 26S data pruned from the 

ITS1+5.8S+ITS2 region, removal of amplification primer sequences from the trnK 

region) exported and aligned manually in Se-Al v2.0a11 (1996; available from A. 

Rambaut, Oxford, England, UK at http://tree.bio.ed.ac.uk/software/seal). Alignment 

was relatively straightforward due to the high degree of sequence similarity and length 

for the taxa involved. Ambiguous regions were generally restricted to the monomeric 

“T” repeat in the 5

ʹ trnK–matK intergenic spacer (IGS). All newly generated data have 

been deposited in GenBank and are available under accession numbers JX992748–

JX992840 (Appendix B).  

Data were analyzed under parsimony and likelihood criteria by genomic data 

partition, first independently, and later in combination. The decision to combine data 

partitions is often based upon the results of an incongruence-length difference, or ILD 

test (Farris et al. 1995). We did not perform this test as it has been suggested to be 

too conservative (Cunningham 1997, Struck 2007). We expected the test to fail due 

to the presence of two very different ITS copies detected for at least one sample in 

the ITS data partition. The decision to combine data partitions was instead based on a 

combination of tree topology congruence and branch support values. In most instances 

52

Gard. Bull. Singapore 65(1) 2013

---