Molecular phylogeny of the cosmopolitan aquatic plant genus Limosella (Scrophulariaceae) with a particular focus on the origin of the Australasian L. curdieana
Limosella is a small aquatic genus of Scrophulariaceae of twelve species, of which one is distributed in northern circumpolar regions, two in southern circumpolar regions, two in the Americas, one endemic to Australia, and six in tropical or southern Africa or both. The Australasian L. curdieana has always been considered distinct but its close phylogenetic relationships have never been inferred. Here, we investigated the following alternative phylogenetic hypotheses based on comparative leaf morphology and habitat preferences or floral morphology: (1) L. curdieana is sister to the African L. grandiflora; or (2) it is closely related to a group of other African species and the northern circumpolar L. aquatica. We tested these hypotheses in a phylogenetic framework using DNA sequence data from four plastid DNA regions and the nuclear ITS region. These were analyzed using maximum parsimony and Bayesian inference. We obtained moderately resolved, partially conflicting phylogenies, supporting that accessions of L. grandiflora form the sister group to the rest of the genus and that L. curdieana groups with the African taxa, L. africana and L. major, and L. aquatica. Thus, the molecular evidence supports the second hypothesis. A biogeographic analysis suggests an out-of-southern Africa scenario and several dispersal events in the Southern Hemisphere. Past dispersal from southern Africa to Australasia is suggested, yet it cannot be excluded that a route via tropical Africa and temperate Asia has existed.
KeywordsAquatic plants Biogeography Dispersal Lamiales Phylogenetic inference
While many of the species are either distributed in Africa or geographically close to Africa (Limosella aquatica L.), a few species are distributed in distant regions, i.e., L. americana Glück, L. curdieana F. Muell., and L. subulata E. Ives. Among these, L. americana described by Glück (1934) is sometimes recognized as a distinct taxon in Central and South America but was recently merged into L. aquatica (Brako and Zarucchi 1993; Cook 2004). Limosella subulata from North and South America has been recognized in some regional Floras (Brako and Zarucchi 1993; Crow and Hellquist 2000), but occasionally treated as a synonym of L. australis R.Br. (Cook 2004). Limosella curdieana, an Australasian endemic species, in contrast, is remarkable as it has never been synonymized, nor has its phylogenetic origin been inferred (Barker 1986, 1999; Harden 1992; Moore 1961).
We aim to test these two competing hypotheses for Limosella curdieana in Australasia and examine whether the species is phylogenetically related to an African species, L. grandiflora, or to a Northern circumpolar species and other African species, i.e., L. africana and L. major. For that purpose, we employed simultaneous molecular phylogenetic analyses of plastid DNA (subsequently referred to as ptDNA) and nuclear ITS (subsequently referred to as nrITS) DNA sequences based on our worldwide taxon sampling of Limosella.
Materials and methods
Specimen and voucher information for the taxa included in this study
Goldblatt P. & Porter L. 12488 (NBG)
Hagstroem & Acock 1162 (S)
Manning J.C. 2854 (NBG)
South Africa: Eastern Cape; Elandsberg
Namibia: Hunsberge; Nuobrivier
W. Giess & M. Muller 14313 (PRE)
South Africa: Central Cape; De Kom/Aarfontein
N.H. Helme 1541 (NBG)
Japan: Saitama; Koshigaya
New Zealand: Canterbury; Lake Forsyth
Australia: Victoria; Gunbower Isl.
Australia: Queensland; South Glen
South Africa: Northeast Cape; Springbok
A. Le Roux 2355 (PRE)
South Africa: Western Cape; Riebeeckasteel
South Africa: Western Cape; Knolfontein
South Africa: Western Cape; Swartruggens
Ethiopia: Bale Prov.; Bale Mts. Natl. park
Ethiopia: Begemdir Prov.; Simien
Ecuador: Napo; Río Chalupas
Lagaard 101769 (AAU)
South Africa: Western Cape; Swartruggens
DNA extraction, amplification, and sequencing
Total genomic DNA was extracted from silica gel-dried leaf tissues or from herbarium specimens using the CTAB method described in Ito et al. (2010). Parts of the plastid DNA regions, ndhF, rbcL, rps16 and trnT-trnF, and nrITS were PCR amplified with the following primers: ndhF-F2 (Oxelman et al. 1999) and ndhF-1955R.re (5′-CGATTATAKGACCAATTATATA) modified from Olmstead and Sweere’s (1994) ndhF-1955R for ndhF; rbcL-F1F (Wolf et al. 1994) and rbcL-1379R (Little and Barrington 2003) for rbcL; rps16-1F and rps16-2R (Oxelman et al. 1997); “a” and “b” for trnT-trnL and “c” and “f” for trnL-trnF (Taberlet et al. 1991); ITS-4 and ITS-5 for nrITS (Baldwin 1992). PCR amplification was performed using TaKaRa Ex Taq polymerase (TaKaRa Bio, Shiga, Japan), and PCR cycling conditions were 94 °C for 60 s; then 30 cycles of 94 °C for 45 s, 52 °C for 30 s, 72 °C for 60 s; and finally 72 °C for 5 min. The PCR products were cleaned using ExoSAP-IT (GE Healthcare, Piscataway, NJ, USA) purification and amplified using ABI PRISM Big Dye Terminator ver. 3.1 (Applied Biosystems, Foster City, CA, USA) with the same primers as those used for the PCR amplifications. DNA sequencing was performed with an ABI PRISM 377 DNA sequencer (Applied Biosystems). Automatic base-calling was checked by eye using Genetyx-Win ver. 3 (Software Development Co., Tokyo, Japan). The sequences generated and their metadata were submitted to the DNA Data Bank of Japan (DDBJ), which is a GenBank data provider (Table 1).
We assembled two datasets of Limosella: (1) ptDNA (ndhF, rbcL, rps16, and trnT-trnF); and (2) nrITS. Sequences were aligned using Mafft ver. 7.058 (Katoh and Standley 2013) and then inspected manually. We used “leave gappy regions” option in Mafft to code gaps found in ndhF, rps16, trnT-trnF, and nrITS.
The incongruence length difference test (Farris et al. 1994) in PAUP* ver. 4.0b10 (Swofford 2002) was employed to test for phylogenetic congruence among the four ptDNA regions using a partition homogeneity test with 1,000 replicates. This test did not reveal significant heterogeneity among genes (P value >0.05), and all subsequent analyses were therefore performed with a combined data set of ptDNA.
Phylogenetic inference was performed using maximum parsimony (MP) in PAUP* (Swofford 2002) and Bayesian inference (BI; Yang and Rannala 1997). In the MP analysis, a heuristic search was performed with 100 random addition replicates involving tree-bisection-reconnection (TBR) branch swapping, with the MulTrees option in effect. The MaxTrees option was set at 100,000. Bootstrap analyses (Felsenstein 1985) were performed using 1,000 replicates with TBR branch swapping and simple addition of sequences. The MaxTrees option was set to 1,000. Gaps were treated as binary characters.
BI analyses were conducted with MrBayes ver. 3.2.2 (Ronquist and Huelsenbeck 2003; Ronquist et al. 2012) run on the CIPRES portal (Miller et al. 2010) after the best models had been determined in MrModeltest ver. 3.7 (Nylander 2002); these models were GTR + I + Γ and GTR + Γ for ptDNA and nrITS, respectively. For gap characters, the “datatype = standard” option of MrBayes was used and default prior settings were applied, i.e., “ratepr = variable”. Analyses were run for 670,000 and 415,000 generations for ptDNA and nrITS, respectively, until the average standard deviation of split frequencies dropped below 0.01, sampling every 1,000 generations and discarding the first 25% as burn-in. The convergence and effective sampling sizes (ESS) of all parameters were checked in Tracer ver. 1.6 (Rambaut et al. 2014). All trees were visualized using FigTree ver. 1.3.1 (Rambaut 2009). Nodes are recognized as strongly, moderately, or weakly supported with: ≥95% bootstrap support (BS), ≥0.99 Bayesian posterior probabilities (PP); ≥70% BS, ≥ 0.95 PP; or <70% BS; <0.95 PP, respectively. The data matrices and the MP and BI trees are available at Treebase (TB2:S19401).
Species trees considering all samples were reconstructed for biogeographic analysis. Limosella macrantha was excluded due to incongruent positions between ptDNA and nrITS trees (see “Results”). A multispecies coalescent method (Heled and Drummond 2010) implemented in BEAST ver. 1.7.2 (Drummond et al. 2006; Drummond and Rambaut 2007) was performed. We ran *BEAST using the two data sets (ptDNA and nrITS) from 21 samples from seven ingroup taxa and assigning them to six terminal species, namely L. aquatica, L. australis, L. curdieana, L. grandiflora, L. major, plus Limosella sp.
We performed two independent runs of ten million generations of the MCMC chains, sampling every 1,000 generations. Convergence of the stationary distribution was checked by visual inspection of plotted posterior estimates using Tracer ver. 1.5 (Rambaut and Drummond 2007). After discarding the first 1,000 trees as burn-in, the samples were summarized in the maximum clade credibility tree using TreeAnnotator ver. 1.6.1 (Drummond and Rambaut 2007) with a posterior probability limit of 0.5 and summarizing mean node heights. The results were visualized using FigTree ver. 1.3.1 (Rambaut 2009).
Reconstruction of historical biogeography of Limosella was performed using RASP ver. 3.2 (Reconstruct Ancestral State in Phylogenies) (Yu et al. 2015). The Bayesian Binary Method (BBM; Ronquist and Huelsenbeck 2003) was selected because it: (1) accepts polytomies; (2) tends to suggest single distribution areas for ancestral nodes more often than others (Müller et al. 2015); and (3) is capable of providing unambiguous and informative results (Ito et al. 2016). BBM was conducted using the post burn-in species trees that resulted from the *BEAST analysis. The following seven biogeographic areas were defined: (a) Europe; (b) temperate Asia; (c) North America; (d) tropical Africa; (e) southern Africa; (f) Australasia; (g) South America (including Falklands). For the distribution of the species we used the area including the locality of accessions from the molecular phylogenetic analysis of the present study. Multiple ancestral states were allowed. The number of generations was set to 10 million and the first 10% of the samples were discarded as burn-in. All other parameters were kept at default settings.
Percentage of missing characters and gaps by DNA regions
Percentage of missing characters (%)
Total alignment length
Percentage of gaps (%)
Limosella was divided into two strongly-supported groups: a clade of four accessions of L. grandiflora (group I; 100% BS; 1.0 PP) and the rest of the genus (100% BS; 1.0 PP). In the latter clade, Limosella sp. was placed as sister to the remaining accessions (89% BS; 1.0 PP). The other supported groups were (1) L. australis and L. subulata (group II; 88% BS; 1.0 PP) and (2) L. aquatica, L. curdieana, L. macrantha, and L. major (group III plus L. macrantha; <50% BS; 0.98 PP). Three accessions of L. africana were resolved in a polytomy with L. australis-L. subulata and L. aquatica-L. curdieana-L. macrantha-L. major.
The nrITS alignment consisting of 22 ingroup and two outgroup sequences had a total length of 691 bp. In total 121 characters including four binary-coded indels were polymorphic, of which 63 were parsimony-informative. Percentage of missing characters and gaps were: 5.63 and 1.75%, respectively (Table 2). Analysis of this data set yielded 22 MP trees (tree length = 166 steps; consistency index = 0.86; retention index = 0.89). The strict-consensus MP tree and MrBayes BI 50% consensus trees showed no incongruent phylogenetic relationships. Therefore, the better resolved MrBayes tree is presented (Fig. 4b).
A topology similar to that of ptDNA was recovered, with two strongly-supported clades: one comprising four accessions of Limosella grandiflora (group I; 87% BS; 1.0 PP) and another consisting of the remaining taxa (99% BS; 1.0 PP). In the clade of 17 accessions, L. africana UPS:BOT:V-120091 was placed as sister to the remaining accessions (68% BS; 0.99 PP). A clade of L. macrantha and L. africana β (87% BS; 0.98 PP) positioned sister to the remaining accessions (<50% BS; 1.0 PP). Group II, consisting of L. australis and L. subulata (88% BS; 0.99 PP) and group III (L. aquatica, L. curdieana, and L. major; 83% BS; 1.0 PP) formed a clade in the BI analysis (<50% BS; 0.97 PP), which was placed as sister to L. africana N.H. Helme 1541 (NBG).
The ancestral area of Limosella was inferred in southern Africa (1.0 PP); ambiguous results were obtained for the ancestral area of the most recent common ancestor (MRCA) of L. aquatica, L. curdieana, and L. major with highest probability for Australia (70.9 PP) (Fig. 5; Online resource 2).
Phylogeny of Limosella and the position of L. curdieana
The present study reconstructed the most detailed molecular phylogeny of the genus Limosella to date with a primary aim to test two competing hypotheses for ancestral relationships of the Australasian species, L. curdieana, i.e., whether the species is phylogenetically closely related to (1) the African species, L. grandiflora, or (2) a group composed of the two African species, L. africana and L. major, and the Northern circumpolar species, L. aquatica. The topologies recovered based on ptDNA and nrITS, respectively, are mostly congruent, except for a single accession of L. macrantha (see below). Both topologies resolve L. grandiflora as sister to the remaining species, including L. africana, L. aquatica, L. curdieana, and L. major (Fig. 4). The molecular evidence thus clearly rejected the hypothesis proposed by Glück (1934) and instead supported a relationship of L. curdieana with L. africana, L. major and L. aquatica.
Topological incongruence between ptDNA and nrITS
We detected a topological incongruence caused by our single accession of Limosella macrantha. In the ptDNA phylogeny this tropical African species (Ghazanfar et al. 2008; Glück 1934) is resolved in a clade including multiple accessions of the likewise tropical African species, L. major, whereas in the ITS phylogeny it is sister to southern African L. africana. Although morphological evidence suggests a close relationship between L. macrantha and L. australis (including specimens previously segregated as L. subulata) (Cook 2004; Glück 1934), the single accession of L. macrantha obtained from previous phylogenetic studies (Kornhall and Bremer 2004; Oxelman et al. 2005) was resolved as distantly related to L. australis (Fig. 4). Further taxon and data sampling will reveal whether this single specimen is part of an introgressive hybrid swarm or represents a hybrid. Alternatively, incomplete lineage sorting cannot be excluded as cause of the topological conflicts since this is common in closely related lineages such as Limosella.
Biogeography of Limosella and implications for dispersal
The present study supports a basal diversification of Limosella in southern Africa with moderate to strong support (Fig. 5), confirming a southern African origin of the genus as indicated by species richness of the ingroup and the distribution of sister genera. Thus, southern Africa seems to be a cradle for not only a number of drought-adapted taxa such as Crassulaceae (Mort et al. 2001), Scrophulariaceae (Oxelman et al. 2005), Thesium L. (Moore et al. 2010), Amaryllidoideae (Rønsted et al. 2012), and Asparagus L. (Norup et al. 2015) but also aquatics such as Limosella.
Bell et al. (2010) provided an estimated origin of Scrophulariaceae earlier than the mid-Paleogene, since tribes Myoporeae and Scrophularieae diversified ca. 51–53 mya. As Limosella is derived much later in the evolution of the family (Oxelman et al. 2005), vicariance due to plate tectonics is not an option to explain the discontinuous distribution worldwide and instead seed dispersal is more likely as is mentioned by Cook (2004): “the disseminules are small seeds, dispersed in mud and perhaps otherwise.” The current distribution of L. australis (including L. subulata) in e.g., Ecuador, New Zealand, and the Falklands could be a case of “wind highways” existing in the Southern Hemisphere (Muńoz et al. 2004).
The biogeographic origin of Limosella curdieana remains uncertain, but a close relationship with L. aquatica from Europe, temperate Asia, and North America, and L. major from tropical Africa (Ethiopia) is moderately supported (Fig. 4). The biogeographic analysis suggests Australasia as the ancestral area for the MRCA of the three species and thus implies a dispersal route from southern Africa to Australasia, and then to temperate Asia and tropical Africa (Fig. 5). Alternatively, considering the facts that (1) migratory bird routes are well documented between Asia and Australasia (Boere and Stroud 2006), (2) case studies exist that reveal disjunct distributions between Asia and Australasia (Lobelia L.: Kokubugata et al. 2012; Solenogyne Cass.: Nakamura et al. 2012), and (3) young plant groups in Australia have predominantly migrated from Asia (Crisp and Cook 2013), dispersal to Australasia via temperate Asia may also explain the geographic isolation of L. curdieana in Australia.
Implications for taxonomy of Limosella
Our molecular phylogenetic results provide additional insights into the taxonomy of Limosella. The three specimens of L. africana show close relationships, especially in ptDNA analysis, but do not form a clade in either of the analyses. No significant morphological differences are observed among the specimens except leaf shape, i.e., linear-lanceolate leaves for L. africana α and ovate leaves for L. africana β and L. africana γ. Still, all three specimens could belong to L. africana var. africana Glück, even in the strict sense, because none of these have much-clustered, short-pedicellate flowers or fruits (Cook 2004; Glück 1934). The sterile Limosella sp. shows affinities to L. africana var. africana in vegetative morphology, i.e., ovate leaves. Additional taxon sampling in flower and fruit from the locality and/or nearby habitats will provide further insight into the affinities of this accession, e.g., whether it belongs to either of the varieties, or represents an undescribed taxon. Since non-monophyletic taxa may be the result of various processes such as incomplete lineage sorting, a proper investigation of the species status of L. africana would require sampling from multiple loci in a species tree context.
The single accession of L. subulata from Ecuador forms a clade with three accessions of L. australis from New Zealand and the Falklands. This result agrees with the taxonomic treatment by Cook (2004) who synonymized L. subulata with the widespread L. australis. Further studies based on increased taxon and data sampling focusing on the taxonomic status of these taxa are needed.
Within a phylogenetic framework we provided new insights into the evolutionary origin of L. curdieana, the only Australasian endemic species in the genus. Our molecular phylogenetic analyses based on ptDNA and nrITS data sets revealed a basal diversification in southern Africa and close relationships between L. curdieana and northern circumpolar as well as a species from tropical Africa. The biogeographic analysis points to an out-of-southern Africa into the Northern Hemisphere and subsequent dispersal back into the Southern Hemisphere. Past dispersal from southern Africa to Australasia is suggested, however, we cannot exclude the existence of a route via tropical Africa and Asia.
The authors thank R. Kaul (NEB) for providing Limosella materials, T. Trinder-Smith (BOL), J. Palmer (CANB), C. Gallagher, P. Milne (MEL), C. Cupido (NBG), and E. van Wyk (PRE) for arranging loans from their institutions and/or for hospitality during our research visits; M. Hjertson (UPS) for sending scanned images of voucher specimens of Limosella at UPS; C. Ishii (Tsukuba) for help with DNA sequencing; J. Guerin (South Australian Seed Conservation Center, Australia) for providing L. curdieana photos. We would also like to thank P. B. Pelser (CANU), J. Murata, H. Ikeda, and T. Ohi-Toma (TI), and J. Li (Kunming) for their continuous encouragements and supports. YI and AMM received plant collecting permit in Western Cape, South Africa from Cape-Nature. This research was supported by FY 2012 Researcher Exchange Program between Japan society for the promotion of science (JSPS) and Royal Society of New Zealand (RSNZ) to YI, JSPS KAKENHI Grant Number 25440224 to NT, and the South African National Research Foundation (NRF) to AMM.
- Barker WR (1986) Limosella. In: Jessop JP, Toelken HR (eds) Flora of South Australia, vol 3., Polemoniaceae to compositeABRS/CISRO, Melbourne, pp 1282–1284Google Scholar
- Barker WR (1999) Limosella. In: Walsh NG, Entwisle TJ (eds) Flora of Victoria, vol 4., Dicotyledons cornaceae to asteraceaeInkata Press, Port Melbourne, pp 497–498Google Scholar
- Boere GC, Stroud DA (2006) The flyway concept: what it is and what it isn’t. In: Boere GC, Galbraith CA, Stroud DA (eds) Waterbirds around the world. The Stationery Office, Edinburgh, pp 40–47Google Scholar
- Brako L, Zarucchi JL (1993) Catalogue of the flowering plants and gymnosperms of Peru. Monographs in Systematic Botany volume 45, Missouri Botanical Garden, St. Louis, MOGoogle Scholar
- Cook CDK (2004) Aquatic and wetland plants of southern africa. Backhuys Publishers, LeidenGoogle Scholar
- Crow GE, Hellquist CB (2000) Limosella. In: Crow GE, Hellquist CB (eds) Aquatic and wetland plants of northeastern North America, vol 1., Pteridophytes, gymnosperms, and angiosperms: dicotyledonsThe University of Wisconsin Press, Madison, pp 327–329Google Scholar
- Darwin C (1872) The origin of species by means of natural selection. John Murray, LondonGoogle Scholar
- Ghazanfar SA, Hepper FN, Philcox D (2008) Scrophulariaceae. In: Beentje HJ, Ghazanfar SA (eds) Flora of tropical East Africa. Published on behalf of the East African governments by Royal Botanic Gardens, Kew, UK, pp 1–211Google Scholar
- Godfrey RK, Wooten JW (1981) Limosella. In: Godfrey RK, Wooten JW (eds) Aquatic and wetland plants of southeastern United States Dicotyledons. Univ Georgia Press, Athens, p 649Google Scholar
- Gorshkova SG (1997) Limosella L. In: Schischkin BK, Bobrow EG (eds) Flora of U.S.S.R. vol 22, pp 367–369Google Scholar
- Harden GJ (1992) Limosella. Flora of New South Wales, vol 3. New South Wales University Press, Australia, pp 563–564Google Scholar
- Hilliard OM, Burtt BL (1986) Notes on some plants of southern Africa chiefly from Natal: XII. Notes R Bot Gard Edinb 43:189–228Google Scholar
- Hong DY, Yang H, Jin CL, Holmgren NH (1998) Scrophulariaceae. In: Wu ZY, Raven PH (eds) Flora of China. Science Press, Beijing, pp 1–212Google Scholar
- Ivanina LI (2001) Limosella. In: Fedorov AA (ed) Flora of Russia, the European part and bordering regions, vol 5. A. A. Balkema, Rotterdam, pp 325–347Google Scholar
- Miller MA, Pfeiffer W, Schwartz T (2010) Creating the CIPRES science gateway for inference of large phylogenetic trees. In: Proceedings of the Gateway Computing Environments Workshop (GCE), 14 Nov. 2010, New Orleans, LA, USA, pp 1–8Google Scholar
- Moore LB (1961) Limosella. In: Allan HH (ed) Flora of New Zealand, vol l., Government printerWellington, New Zealand, pp 846–847Google Scholar
- Nylander JAA (2002) MrModeltest v.1.0. Program distributed by the author. Department of Systematic Zoology, Uppsala University, Uppsala. Available at: http://www.ebc.uu.se/systzoo/staff/nylander.html
- Philcox D (1990) Limosella. In: Launert E, Pope GV (eds) Flora Zambesiaca, volume 8, part 2. Flora Zambesiaca Managing Committee, London, pp 73–75Google Scholar
- Rambaut A (2009) FigTree v1.3.1: Tree Figure Drawing Tool. http://tree.bio.ed.ac.uk/software/figtree/
- Rambaut A, Suchard MA, Xie W, Drummond AJ (2014) Tracer. Ver 1.6. http://beast.bio.ed.ac.uk/Tracer
- Rønsted N, Symonds MRE, Birkholm T, Christensen SB, Meerow AW, Molander M, Mølgaard P, Petersen G, Rasmussen N, van Staden J, Stafford GI, Jäger AK (2012) Can phylogeny predict chemical diversity and potential medicinal activity of plants? A case study of Amaryllidaceae. BMC Evol 12:182CrossRefGoogle Scholar
- Swofford DL (2002) PAUP*: Phylogenetic analysis using parsimony (*and other methods), version 4.0b. Sinauer, Sunderland, Massachusetts, USAGoogle Scholar
- Webb DA (1972) Limosella. In: Tutin TG, Burges NA, Chater AO, Edmondson JR, Heywood VH, Moore DM, Valentine DH, Walters SM, Webb DA (eds) Flora Europaea, vol 3, pp 205–216Google Scholar
- Yamazaki T (1993) Limosella. In: Iwatsuki K, Yamazaki T, Boufford DE, Ohba H (eds) Flora of Japan, vol IIIa. Angiospermae Dicotyledoneae Sympetalae, Kodansha, p 334Google Scholar