The systematics of Lobelioideae was explored by checking literature works through online libraries and journals. We explored previous works to understand the debates and contentions that had been there previously and the steps that had been taken to solve the contentions. We also wanted to have a general overview of the taxonomic progress with regards to this subfamily (Zuccarini 1832; McVaugh 1941, 1943a, b, 1949a, b, 1955; Wimmer 1943, 1953, 1968; Bowden 1959, 1961; Moeliono and Tuyn 1960; Carlquist 1962, 1969, 1974; Carlquist et al. 1965; Mabberley 1974, 1975; Thulin 1986a, 1991; Thulin et al. 1986b; Phillipson 1989; Ayers 1990; Lammers 1990, 1991, 1993, 2007a, b, 2010, 2011; Murray and Cameron 1990; Harvey 1992; Lammers and Hensold 1992; Lammers et al. 1993; Knox and Kowal 1993; Knox et al. 1993; Cosner et al. 1994; Givnish et al. 1994; Givnish 1995, 1998; Gustaffson and Bremer 1995; Murata 1995; Vieira and Shepherd 1998; Serra et al. 1999; Givnish 2000; Schultheis 2001a, b; Eddie et al. 2003; Givnish et al. 2004; Knox et al. 2004, 2008a, b; Murray et al. 2004; Knox 2005, 2014; Koopman and Ayers 2005; Knox et al. 2006; Antonelli 2008, 2009; Givnish et al. 2009; Haberle et al. 2009; Givnish 2010; Givnish et al. 2013; Crowl et al. 2014; Lagomarsino et al. 2014, 2016; Chen et al. 2016; Crowl et al. 2016; Knox and Li 2017; Uribe-Convers et al. 2017; Hunter 2018).

We aimed to include as many of the Lobelioideae species as possible. Nineteen loci were obtained, that is, eighteen plastid gene loci (atpB-rbcL spacer, atpB, atpF, atpF-atpH spacer, atpH, matK, ndhF, psbA-trnH spacer, psbA-trnK spacer, petD, rbcL, rpoC1, trnL-trnF spacer, trnT-trnL spacer, trnV-trnK spacer, trnK-matK spacer, rpl32-ndhF spacer, rpl16) and one nuclear gene, internal transcribed spacer (ITS). These sequences were generated using the NCBI ENTREZ UTILITY (Kans 2020) program (Accessed 1^ST April 2020) and double-checked manually at the GenBank database. Additionally, almost all available Campanulaceae complete plastid genomes were manually accessed from the GenBank. The respective plastid regions were extracted using NCBI BLASTN v. 2.9.0+ (Camacho 2018) with default settings. We included nine outgroup taxa to increase the chances of recovering the early branching of Campanulaceae (Knox 2014). The outgroups included: Abrophyllum ornans (F.Muell.) Benth. (Bentham and Mueller 1864), Carpodetus serratus J.R.Forst. & G.Forst. (Forster 1776), Corokia cotoneaster Raoul (Raoul 1846), Cuttsia viburnea F.Muell. (Mueller 1865), Pentaphragma ellipticum Poulsen (Poulsen 1903), Phelline lucida Vieill. ex Bail. (Baillon 1872), Roussea simplex Sm. (Smith 1789), Scaevola sp. L. (Linnaeus 1771) and Stylidium adnatum R.Br. (Brown 1810). Taxa voucher information and GenBank accession numbers are provided in the Suppl. material 1: Lobelioideae data matrix.

All the gene regions were aligned separately using MAFFT v. 7.429 (Katoh and Standley 2013) with an adjust-direction and 1000 maximum iterations options. The alignment of each region was manually checked and taxa with short sequences (≤ 200bp) were edited using GENEIOUS Pro v. 5.6.4 (Kearse et al. 2012). Edited sequences were analyzed using PHYUTILITY v. 2.2.6 (Smith and Dunn 2008) to delete gaps and ambiguous sequences. The indels within the sequences were treated as missing data and they were therefore excluded from the analysis. The trnF-trnL spacer region had the highest number of sequences while atpF recorded the least (Table 1). Each dataset was analyzed using JMODELTEST v. 2.1.10 (Darriba et al. 2012) to determine the best evolution substitution model (Table 1). Maximum Likelihood (ML) analysis for each of the aligned dataset was done using RAxML v. 8.2.12 (Stamatakis 2014). Datasets with unavailable models were analyzed using the GTRCAT model. Each dataset was analyzed using 100 bootstrap values to measure clade support. After pilot phylogenetic analyses, nine plastid regions, that is, atpB, atpF, atpF-atpH spacer, atpH, matK, psbA-trnK spacer, petD, rbcL, and rpoc1, were selected (Table 1) as they had a better phylogenetic resolution. The nine plastid regions were concatenated to form a combined plastid (cp) dataset and used for ML analyses with the best-selected model. The selection of the best substitution model of the combined dataset did not employ the use of partitioning in this analysis. ITS region was also subjected to ML analyses separately since it formed a tree with a poor resolution when combined with the plastid regions.

Biogeographic analyses were conducted in MESQUITE v. 3.61 (Maddison and Maddison 2019) using the parsimony ancestral states reconstruction method. The biogeographic regions were divided into Africa (Madagascar, tropical, and southern Africa), America (North, Central, and South America), Asia (tropical and temperate Asia), Australasia (Australia and New Zealand), Mediterranean (northern Africa, Cyprus, Sicily, Sardinia, and Crete) and the Pacific Islands (Hawaii, Kaua’i, French Polynesia, Rarotonga, and the Marquesas Islands) according to Chen et al. (2016). The ancestral regions for the outgroups species, Campanulaceae sp. and Lobelia sp. were unclear and therefore were not assigned any value (region), however, the reconstruction method employed was set to consider missing and inapplicable data.

The data underpinning the analysis reported in this paper are deposited in the Dryad Data Repository at https://doi.org/10.5061/dryad.3xsj3txfw.

Many scholars have expressed their insights with the existing systematics of the Lobelioideae genera (Lammers 2007a; Givnish 2010; Chen et al. 2016; Knox and Li 2017). The uncertainty in circumscription among different lineages in Lobelioideae has been a result of rapid diversification and divergence of this subfamily approximately 20 million years ago (Knox and Li 2017). After extensive literature search and reviews, we found three main areas that were previously in contention: South African origin of Lobelioideae (Mabberley 1975; Knox and Li 2017), herbaceous habit, and Asian origin of giant lobelioids (Carlquist 1962; Mabberley 1974; Chen et al. 2016), and the convergent evolution of giant rosette (perennial monocarpic herbs mostly occurring in alpine and subalpine bogs) lobelioids (Antonelli 2009; Givnish 2010). Currently, agreements regarding these contentions appear to have been reached and are in accord with our analyses.

The geographical origin of the Lobelioideae had been a point of contention, with different scholars having varying biogeographic theories. Mabberley (1975) suggested a South American origin of lobelioids. Mabberley (1975) postulated that the South American pachycaul lobelioids gave rise to plants, which spread to Chile and the Caribbean (Lobelia § Tylomium (C.Presl) Benth. (Bentham 1876)), Hawaii (Trematolobelia) and Brazil (Lobelia § Rhynchopetalum (Fresen.) Benth. (Bentham 1876)). He added that the rise of winged seeds in Hawaii permitted the inter-island spread of lobelioids and in Brazil, it allowed the Lobelia § Rhynchopetalum to travel to Africa. However, Knox et al. (1993, 2008a) stated that the South American species are mixed assemblage, possibly involving pantropical dispersal events. Knox et al. (2006), Chen et al. (2016), and Knox and Li (2017) concluded Lobelioideae originated from South Africa and underwent multiple cosmopolitan radiation events. Our results supported the ‘Out of Africa’ hypothesis and multiple cosmopolitan radiations of Lobelioideae, which corroborated Antonelli (2009), Chen et al. (2016), and Knox and Li (2017).

The ancestral habit type and origin of the giant Lobelioids have been in the limelight for years. Carlquist (1962, 1969), using wood anatomy, suggested an herbaceous origin of giant lobelioids. However, Mabberley (1974, 1975) challenged the above sentiment and suggested that the herbaceous species of lobelioids have been derived from large, thick-stemmed ancestors (Lobelia § Rhynchopetalum and Lobelia § Tylomium). He added that the herbaceous habit of lobelioids is an advanced character. Knox et al. (1993), using cpDNA restriction sites and inversions, supported Carlquist’s (1962) hypotheses of herbaceous ancestry. Givnish (2000), based on molecular phylogenetic analysis, showed that the ancestor of the Hawaiian lobelioids was most likely woody, corroborating Mabberley’s (1974, 1975) proposals. Also, Givnish et al. (2009) hypothesized that an Asian group – represented by the placeholder Lobelia nicotianifolia Roth (Roth 1821) – might have the ancestral stock from which both Pacific and African giant lobelioids had evolved. Most recently, Chen et al. (2016) confirmed (1) the herbaceous habit of lobelioids ancestors, and (2) the Asian origin of giant lobelioids. Knox and Li (2017) corroborated Chen et al. (2016) and added that extant Hawaiian/Pacific and Brazilian/African giant lobelioids are derived from herbaceous giant lobelioids (Knox and Li 2017). Our results corroborated that of Chen et al. (2016) and Knox and Li (2017), in which, herbaceous ancestry of giant rosette lobelioids was well illustrated.

The Hawaiian lobelioids form a remarkable clade, encompassing more species than any other plant clade restricted to a single oceanic island or archipelago, and their geographic source has been hotly debated (Givnish et al. 1994, 2009; Givnish 1995). They have long been viewed as one of the most spectacular cases of adaptive radiation in plants on oceanic islands (Carlquist et al. 1965; Carlquist 1974; Lammers 1990; Givnish et al. 1994, 2004; Givnish 1995; Givnish 1998; Givnish and Montgomery 2014). Wimmer (1953) and Mabberley (1974, 1975) postulated that fleshy-fruited genera are a product of a single colonization event while capsular-fruited taxa are products of more than one colonization event. However, Givnish (2000), using molecular phylogenetic analysis, illustrated that Hawaiian lobelioids are instead a product of a single immigration event. Antonelli (2009) suggested that the Hawaiian and African giant lobelioids appeared to have evolved from a single common ancestor. However, Givnish (2010) refuted those claims and argued that the giant rosette lobelioids are an exemplar of convergent evolution rather than single common ancestry. Our analysis corroborated one of Givnish’s (2010) illustrations on lobelioids’ convergent evolution theory, that is, compared to the rest of lobelioids species, only a minority number of species (Lobelia § Rhynchopetalum) have the giant rosette growth form adapted to alpine or mountain conditions, with non-rosette species forming the remainder of the clade. These ‘rosette-species’ are embedded within the non-rosette species (Fig. 1), a clear indication that indeed the giant rosette lobelioids are a result of convergent evolution. Knox and Li (2017) also used maximum-likelihood analyses of whole plastomes to conclude that the giant African lobelioids (including some descendants in South America) were sister to the Pacific giant lobeliads as a whole and with Lobelia boninensis Koidz. (Nakai 1920), from the Bonin Islands, then Apetahia/Sclerotheca from the Society Islands and the Cook Islands forming a sister to the Hawaiian lobelioids, and then all of them forming a sister to some Asian giant lobelioids, corroborating the proposal by Givnish et al. (2004). Our combined plastid data ties together Delissea-Brighamia, Trematolobelia, Lobelia § Galeatella, and Lobelia § Revolutella, all from Hawaii, closely related to giant African and South American lobelioids.

In addition, Knox and Li (2017) summarized the cosmopolitan radiation of lobelioids in four out-of-Africa dispersal scenarios. (1) The biogeographic pattern of South African species relative to lobelioids elsewhere in the world maps Lobelioideae ancestry to the modern-day Western Cape Province (Knox and Li 2017). Lobelia anceps L.f. (Linnaeus 1782), for instance, originated from South Africa and subsequently dispersed to many other southern hemisphere sites, including New Zealand (Knox et al. 2006). Madagascar acted as a stepping-stone to eastern Asia where the robust, herbaceous, hemicryptophyte growth form evolved from (Knox and Li 2017). (2) The Amphi-tropical dispersal from the Western Cape to the Mediterranean region did occur and with rapid subsequent dispersal to the North America region (Knox and Li 2017). (3) The dispersal from South Africa to South America stood the greatest likelihood of success if the initial colonization occurred at a similar latitude with a similar habitat (Knox and Li 2017). The circumscription of Lobelia xongorolana E.Wimm. (Wimmer 1935) (Endemic in Angola) as the sister lineage to the Brazilian species (Mabberley 1974), would implicate Angola as a stepping-stone in dispersal to Brazil, whereas a true sister-species relationship with Lobelia stricklandiae would suggest that dispersal to Brazil originated from East Africa and that dispersal to Angola was a separate event (Knox and Li 2017). (4) Successful colonization of Australia from South Africa also would have been favored by latitudinal and habitat similarity. The Western Australia endemic Isotoma hypocrateriformis Druce (Druce 1917), is sister to the remaining species in this predominantly Australasian clade that subsequently diversified in most Australian habitats, dispersed on three separate occasions to New Zealand (Knox et al. 2008b; Heenan et al. 2008), and dispersed twice to eastern Asia (Knox and Li 2017).

Lobelioideae consisted of up to 31 genera (Knox et al. 2008a). However, through our extensive literature review, we found a total of 33 currently documented genera (Table 3). We sampled 27 out of the 33 Lobelioideae genera in our combined plastid (cp) dataset (for easy understanding, subsequent discussion part is based on the combined (cp) plastid tree (Fig. 1; Fig. 2) and the ITS region tree. Our analyses found ten monophyletic Lobelioideae genera, that is, Monopsis, Wimmerella, Dialypetalum, Clermontia, Solenopsis, Legenere, Downingia, Burmeistera, Lysipomia, and Trematolobelia, three paraphyletic genera, that is, Apetahia, Sclerotheca, and Cyanea, and lastly, five polyphyletic genera, that is, Lobelia, Pratia, Centropogon, Siphocampylus, and Isotoma. Grammatotheca, Delissea, Brighamia, Palmerella, Porterella, Diastatea, Hypsela, Hippobroma, and Lithotoma had only one representative in each genus. Pratia borneensis Hemsl. (Hemsley 1886), Lobelia physaloides A.Cunn. (Cunningham 1838), Lobelia heterophylla subsp. heterophylla, and some members of the Lobelia § Delostemon (E.Wimm.) J.Murata (Murata 1995) formed the basal group of Lobelioideae with a BS value of 83 (Fig. 1).

Figure 2. 

Biogeography of Lobelioideae combined plastid (cp) region datasets using parsimony ancestral state reconstruction. Taxa color coding represents the geographical distribution of the species.

Lobelioideae originated from Africa and this corroborated Knox et al. (2006) and Knox and Li (2017). Nemacladoideae forms a sister to the Lobelioideae group. This group is endemic to North America. However, their position and endemicity in North America do not affect the African origin of Lobelioideae (Fig. 2). Knox et al. (2006) stated that the removal and/or inclusion of Cyphia clade (Endemic to Africa) as the sister to Lobelioideae would not interfere with the African origin of this subfamily, and this corroborated with our biogeographic results. Besides, it is also evident that multiple dispersal events occurred in this subfamily. The basal group consisted of species with their ancestral region in Australasia and Africa. The African group nested some species from Asia, America, and the Pacific Islands. This depicted possibilities of long-distance dispersal and diversification events in some species.

The following is a discussion of specific genera within the Lobelioideae subfamily. The order of the discussion is according to the positioning of the genera in the phylogram, starting from the basal position (Fig. 1). Pratia is polyphyletic. Pratia borneensis is a sister (BS = 83) to a clade formed by Lobelia physaloides, L heterophylla subsp. heterophylla and seven members from the Lobelia § Delostemon. Pratia angulata Hook.f. (Hooker 1844), forms a clade with Lobelia chinensis Lour. (de Loureiro 1790) with a BS value of 86. Pratia nummularia A.Braun & Asch. (Braun 1861), on the other hand forms a clade with Lobelia angulata with a BS value of 100, while Pratia purpurascens (R.Br.) E.Wimm. (Wimmer 1953), forms a clade with Lobelia arnhemiaca E.Wimm. (Wimmer 1948) with a BS value of 61 (Fig. 1). Biogeography: Murray et al. (2004) described New Zealand as the ancestral region of Pratia. This was also echoed by Knox et al. (2008b) in their work on the phylogenetic position of Lobelia glaberrima Heenan (Heenan et al. 2008), in New Zealand. Our analysis placed this genus in Australasia as the ancestral region (Fig. 2).

Grammatotheca has only one species, Grammatotheca bergiana C.Presl (Presl 1836). It is nested within some members of the Lobelia § Delostemon with a BS value of 62 (Fig. 1). Our results confirmed that of Antonelli (2009) and Chen et al. (2016). Biogeography: Knox et al. (2006) indicated South Africa as the origin of this genus. Chen et al. (2016) also placed Grammatotheca in Africa. This genus is embedded within Lobelia § Delostemon clade which has its ancestral region in Africa. Knox et al. (2006) stated that holopogonoid Lobelia gave rise to this particular genus which diversified in South Africa and was later introduced to Australia via hay shipped with cattle from South Africa. Our results corroborate the above-mentioned studies and placed this genus in Africa as its ancestral area (Fig. 2).

Monopsis forms a monophyletic group with a BS value of 91. It forms a clade with members of the Lobelia § Delostemon with a BS value of 100 (Fig. 1). This result corroborates that of Chen et al. (2016). Biogeography: Phillipson (1989) indicated Cape Province, South Africa as the ancestral region of Monopsis. Knox et al. (2006) also indicated South Africa as the ancestral region of this genus. Our result placed this genus in Africa as its ancestral region (Fig. 2).

Wimmerella forms a clade with a BS value of 100. They form a sister to Lobelia anceps L.f. (Linnaeus 1782) with a BS value of 100 (Fig. 1). This result corroborates that of Knox and Li (2017) and Chen et al. (2016). Biogeography: Knox et al. (2006) placed this genus in Western Cape, South Africa as its ancestral region. Chen et al. (2016) and Knox and Li (2017) corroborated Knox et al. (2006) results. Our analysis corroborates the above-mentioned studies and placed this particular genus in Africa (Fig. 2).

Delissea and Brighamia form a clade with a BS value of 86, a result similar to that of Murata (1995), Antonelli (2008), Givnish et al. (2009), Chen et al. (2016), and Knox and Li (2017). Biogeography: These genera were placed in Kaua’i or some older island as their ancestral region (Givnish et al. 1994, 2004, 2009; Givnish 1995). Our results place these genera in the Pacific Islands as their ancestral area which corroborates both Givnish et al. (2004) and Knox and Li (2017).

Trematolobelia forms a clade with a BS value of 99 (Chen et al. 2016) (Fig. 1). This clade forms a sister to Lobelia kauaensis (A.Gray) A.Heller (Heller 1897), with a BS value of 90. Biogeography: Givnish (1998) placed Trematolobelia on Kaua`i as its ancestral area. Givnish et al. (2009) indicated the Hawaiian archipelago as the ancestral region of this particular genus. Our analysis placed this genus in the Pacific Islands as its ancestral area which corroborates the above-mentioned studies (Fig. 2).

Apetahia and Sclerotheca form a clade with a BS value of 64. This result confirms that of Givnish et al. (2009) and Chen et al. (2016). We sampled only three Apetahia and eight Sclerotheca species. The low BS value might have been a result of incomplete sampling in these two genera. Sclerotheca margaretae F.Br. (Brown 1935), Sclerotheca viridiflora Cheeseman (Cheeseman 1903), Sclerotheca oreades E.Wimm. (Wimmer 1948), Sclerotheca arborea DC. (Candolle 1839), Sclerotheca forsteri Drake (Del Castillo 1892), and Sclerotheca magdalenae J.Florence (Florence 1996) form a clade with a BS value of 60. Biogeography: Chen et al. (2016) placed these genera in French Polynesia as their ancestral area. Knox and Li (2017) indicated the ancestral location of Apetahia longistigmata (F.Br.) E.Wimm. (Wimmer 1948) to be in Marquesas and S. viridiflora to be in Rarotonga, both in the South Pacific Islands. Our results placed the genera in the Pacific Islands as their ancestral area which concur with the above-mentioned studies.

Cyanea forms a paraphyletic group (Fig. 1). This result corroborates with Antonelli (2009), Chen et al. (2016), and Hunter (2018). Cyanea aspleniifolia Hillebr. (Hillebrand 1888), Cyanea koolauensis Lammers, Givnish and Sytsma (Lammers et al. 1993), and Cyanea fissa Hillebr. (Hillebrand 1888) form a clade with a BS value of 98 that is sister to Clermontia (BS=77) (Fig. 1). Clermontia forms a clade (Givnish et al. 2009, 2013; Chen et al. 2016) with a BS value of 90 (Fig. 1). More broadly, Hunter (2018) used phylogenomic data from hundreds of single-copy nuclear genes and whole plastomes to infer that most of Clermontia are sister to the purple-fruited clade of Cyanea (see Givnish et al. 1994; Givnish 1995), with the orange-fruited clade of Cyanea sister to both. Biogeography: Givnish et al. (1994, 2009, 2013) placed the origins of both Clermontia and Cyanea on Kaua`i or some older island. Chen et al. (2016) and Knox and Li (2017) placed these two genera in the Hawaiian Islands as their ancestral region. Our results placed the genera in the Pacific Islands as their ancestral area, corroborating the above-mentioned studies (Fig. 2).

Solenopsis is monophyletic with a BS=100. Lobelia urens L. (Linnaeus 1753) (Lobelia § Stenotium (C.Presl) Lammers (Lammers 2011)) formed a sister to Solenopsis with the BS value of 87 (Fig. 1). Our results corroborate that of Knox and Li (2017). Biogeography: Crespo et al. (1998) indicated the Mediterranean as the ancestral region of this genus. Knox and Li (2017) also indicated the Mediterranean as the ancestral area for Solenopsis. Our result corroborates that of Crespo et al. (1998) and Knox and Li (2017) and places this genus in the Mediterranean region as its ancestral region (Fig. 2).

Downingia formed a monophyletic clade with a BS value of 96. Porterella is sister to Downingia with a BS value of 98. This result corroborates Chen et al. (2016). Legenere limosa (Greene) McVaugh (McVaugh 1943a), and Legenere valdiviana (Phil.) E.Wimm. (Wimmer 1953) form a clade (BS=100), which is a sister to Downingia and Porterella. Palmerella forms a sister to Downingia, Porterella, and Legenere with a BS of 90 (Fig. 1). Biogeography: McVaugh (1941) indicated western North America as the ancestral area of Downingia. Schultheis (2001a, b) corroborated McVaugh (1941). Chen et al. (2016) placed Downingia, Porterella, and Legenere in North America. Our analysis places the genera in North America as their ancestral region (Fig. 2).

Diastatea is clustered with Lobelia nana Kunth (Kunth et al. 1976) with a BS value of 62 (Fig. 1). However, our phylogenetic results contradicted that of Chen et al. (2016) that appeared to form a clade with Solenopsis. Diastatea was differentiated from genus Lobelia by two main features: a superior ovary, and a persistent corolla lacking a dorsal fissure (McVaugh 1940). Albeit these characters have been used to separate the two genera, some species in the Lobelia § Stenotium (featured by partially inferior to the superior ovary) possess the same characteristics e.g. Lobelia xalapensis Kunth (Kunth et al. 1976), L. nana, and Lobelia diastateoides McVaugh (McVaugh 1940). Biogeography: Knox et al. (2008a) placed the genus in South America as the ancestral area. Chen et al. (2016) also indicated America as the ancestral area of Diastatea. Our biogeographic results corroborate both Knox et al. (2008a) and Chen et al. (2016) and place the genus in South America as its ancestral area.

Hippobroma is monotypic and is nested within members of the Lobelia § Tylomium (Fig. 1). This result corroborates Chen et al. (2016). Biogeography: Knox and Li (2017) indicated Mexico as the ancestral region of Hippobroma. The ancestral area of § Tylomium, which nestles this genus, is in North America (Lammers 2011). This genus might have resulted following morphological diversification. This corroborates Chen et al. (2016) that placed the genus in Central America. Our result places the genus in Central America as its ancestral area which corroborates the above-mentioned studies.

Isotoma is polyphyletic. Isotoma hypocrateriformis Druce (Druce 1917), is sister to P. angulata, P. nummularia, P. purpurascens, Hypsela, Lithotoma, and some Lobelia species belonging to the Lobelia § Hypsela (C.Presl) Lammers (Lammers 2011), with a BS value of 96. All these genera are from the Australasian region except for Lobelia loochooensis Koidz. (Koidzumi 1929), and L. chinensis that are from Southeast Asia and Lobelia oligophylla (Wedd.) Lammers (Lammers 1999), from South America. Isotoma tridens (E.Wimm.) Lammers (Lammers 1999), forms a clade with Hypsela tridens E.Wimm. (Wimmer 1943) with a BS value of 100. Isotoma fluviatilis F.Muell. ex Benth. (Bentham and Mueller 1869) is sister to L. chinensis and P. angulata with a BS value of 81. Isotoma axillaris Lindl. (Lindley 1826) forms a clade with Lobelia petraea with a BS value of 91 (Fig. 1). According to Givnish et al. (2009), Isotoma formed a sister to the tropical American taxa. Our results corroborate that of Chen et al. (2016). Biogeography: Bussell et al. (2002) and Chen et al. (2016) placed Isotoma in Australia as its ancestral region. Knox and Li (2017) placed Hypsela and Lithotoma in Australasia as their ancestral region. Our results place Isotoma, Lithotoma, and Hypsela in Australasia as their ancestral area which corresponds to the above-mentioned studies (Fig. 2).

Lysipomia pumila (Wedd.) E.Wimm. (Wimmer 1937), Lysipomia cuspidata McVaugh (McVaugh 1955), and Lysipomia sphagnophila Griseb. (Lechler 1857) forms a clade with a BS value of 100 (Fig. 1). This result is consistent with that of Antonelli (2008) and Chen et al. (2016). Biogeography: McVaugh (1955), in his revision of Lysipomia, indicated South America as the ancestral area of this particular genus. Knox and Li (2017) stated diversification in S. America generated Lysipomia. Our results placed Lysipomia in South America as its ancestral region which corroborates McVaugh (1955) (Fig. 2).

Siphocampylus and Centropogon are polyphyletic and intercalates with each other, albeit their statistical support values are low (<50) (Fig. 1). Centropogon dianae Lammers (Lammers 1998), Siphocampylus sparsipilus E.Wimm. (Wimmer 1924), Centropogon brittonianus Zahlbr. (Zahlbruckner 1897), Centropogon gloriosus Zahlbr. (Zahlbruckner 1897), Centropogon simulans Lammers (Lammers 1998), Centropogon peruvianus (E.Wimm.) McVaugh (McVaugh 1949a), Centropogon luteynii Wilbur (Wilbur 1977), Centropogon costaricae (Vatke) McVaugh (McVaugh 1943a), and Siphocampylus manettiiflorus Hook. (Hooker 1848) forms an early clade with the members of the Lobelia § Tylomium. Centropogon nigricans Zahlbr. (Zahlbruckner 1915) is sister to the Burmeistera clade (Fig. 1). Burmeistera forms a clade with a low support value (BS=43) (Fig. 1). The low BS values for Burmeistera and Siphocampylus may be due to inadequate taxon sampling (Uribe-Convers et al. 2017). Biogeography: The ancestral areas for Centropogon, Burmeistera, and Siphocampylus are in South America (Antonelli 2009; Knox and Li 2017; Uribe-Convers et al. 2017). Our analysis placed these genera in S. America which corroborates the above-mentioned (Fig. 2).

Lobelia is the ‘core genus’ among members of the Lobelioideae group (Knox et al. 2006). Lammers (2011) classified this genus into eighteen sections based on morphological characteristics. Our analysis included fifteen out of the eighteen sections: four monophyletic, five polyphyletic, one paraphyletic and five had only one representative each. The unsampled sections were Speirema (Hook.f. and Thomson) Lammers (Lammers 2010), Trimeris (C.Presl) A.DC. (Candolle 1839), and Plagiobotrys Lammers (Lammers 2010), (Table 2).

Lobelia § Holopogon Benth. (Bentham and Mueller 1869). This section had only one out of fourteen species (Lammers 2011) sampled in a combined plastid dataset, that is, Lobelia heterophylla subsp. heterophylla. It occurred at the basal position of the phylogram and formed a clade with L. physaloides with a BS value of 92 (Fig. 1). Our analysis corroborated that of Antonelli (2008) and Knox et al. (2006). Biogeography: Lammers (2011) indicated Australia as the ancestral area of L heterophylla subsp. heterophylla. Our biogeographic results corroborate the above-mentioned studies and place this species in Australasia as its ancestral area (Fig. 2).

Lobelia § Colensoa (Hook.f.) J.Murata (Murata 1995). Only one species was sampled in the combined plastid dataset. According to Lammers (2011), this section is monotypic. Lobelia physaloides forms a sister clade with L heterophylla subsp. heterophylla (§ Holopogon) with a bootstrap value of 92 (Fig. 1). In the ITS phylogram, it is embedded between the Lobelia § Stenotium and Lobelia § Delostemon. Biogeography: Lammers (2011) placed this section in New Zealand’s North Island as the ancestral location. Our results place this section in Australasia as its ancestral region which corroborates Lammers (2011).

Lobelia § Delostemon. This section is paraphyletic. We sampled fourteen out of forty-four species recorded by Lammers (2011) in our combined plastid dataset. Lobelia baumannii Engl. (Engler 1894), Lobelia hartlaubi Buchenau (Buchenau 1881), Lobelia malowensis E.Wimm. (Wimmer 1948), Lobelia patula L.f. (Linnaeus 1782), Lobelia neglecta Roem. and Schult. (Roemer and Schultes 1819), Lobelia coronopifolia L. (Linnaeus 1753), Lobelia holstii Engl. (Engler 1894), and Lobelia tomentosa L.f. (Linnaeus 1782) form a clade with a BS = 84. Lobelia thermalis Thunb. (Thunberg 1794) is sister to a clade of Lobelia aquatica Cham. (Chamisso 1833) and Lobelia zeylanica L. (Linnaeus 1753) with a BS value of 88. However, this clade is intercalated by Grammatotheca. Lobelia capillifolia A.DC. (Candolle 1839) and Lobelia linearis Thunb. (Thunberg 1794) form a clade with a BS value of 97 and is sister to Monopsis with a BS value of 100 (Fig. 1). Our result corroborates that of Chen et al. (2016), Antonelli (2008), and Knox et al. (2006). Antonelli (2008) clustered Grammatotheca and L. aquatica together and further indicated the similarities between them, that is, both are slender annual herbs and have smaller dorsal corolla lobes. These similarities are also observed in Monopsis debilis (L.f.) C.Presl (Presl 1836), (Phillipson 1986). Lammers (2011) described plants in the Lobelia § Delostemon as perennial with prostrate, decumbent, and ascending stems, sessile or petiolate leaves, bilabiate corolla, and capsular fruit. These features are also found in Monopsis alba Phillipson (Phillipson 1986), Monopsis simplex (L.) E.Wimm. (Wimmer 1948), and Monopsis stellarioides Urb. (Urban 1881), (Phillipson 1986). Antonelli (2008) suggested that if a cladistic approach of classification were to be observed strictly, then Grammatotheca and Monopsis would have been placed under this section. A suggestion that is highly supported by our analysis. Biogeography: Our analysis indicates Africa as the ancestral area of this section. However, it embeds two species; L. aquatica and L. zeylanica which were placed in South America and Southeast Asia respectively. Our results corroborated that of Antonelli (2009) and Lammers (2011).

Lobelia § Mezleriopsis Lammers (Lammers 2011). This section had only one out of seven species (Lammers 2011) sampled in the combined plastid dataset. Lobelia sonderiana (Kuntze) Lammers (Lammers 1999) forms a sister to the remaining members of the Lobelioideae group except for Grammatotheca, Monopsis, Lobelia § Delostemon, Lobelia § Colensoa, Lobelia § Holopogon, and P. borneensis (BS = 56) (Fig. 1). This result corroborates Antonelli (2008). Biogeography: Lammers (2011) indicated the ancestral area of this section to be in South Africa with L. sonderiana extending up to Kenya. This corroborates with our results and places Africa as the ancestral region of this section (Fig. 2).

Lobelia § Stenotium. We sampled ten species out of a hundred and forty-four proposed by Lammers (2011) in our combined plastid dataset. This section is polyphyletic. Lobelia laxa McOwan (MacOwan 1890), L. erinus, Lobelia graniticola E.Wimm. (Wimmer 1948), and Lobelia fervens Thunb. (Thunberg 1794) formed a clade with a BS value of 100. Lobelia anceps (BS = 100) is sister to a clade of Wimmerella pygmaea (Thunb.) Serra M.B. Crespo and Lammers (Serra et al 1999) and Wimmerella hederacea (Sond.) Serra and Lammers (Serra et al 1999). Lobelia muscoides Cham. (Chamisso 1833) forms a clade with Lobelia jasionoides (A.DC.) E.Wimm. (Wimmer 1943) with a BS value of 100. Lobelia urens forms a sister to Solenopsis with a BS value of 87. Lobelia nana and Diastatea form a clade with a BS value of 62 (Fig. 1). According to our phylogenetic analysis, this section appears to be polyphyletic, a suggestion that corroborates Antonelli (2008) and Knox et al. (2006). More so, Solenopsis, just like Wimmerella, has corolla completely fused (Knox et al. 2006). Lammers (2011) suggested the inclusion of Wimmerella in this section. A suggestion that is well supported by our phylogenetic analysis. Biogeography: Our analysis placed the ancestral area of this section in Africa. However, L. nana and L. xalapensis have their ancestral areas in South America while L. urens has its ancestral region in the Mediterranean. Our results corroborated that of Lammers (2011).

Lobelia § Jasionopsis Lammers (Lammers 2011). Only one sample was analyzed in our combined plastid dataset. This section is monotypic. The sampled species included L. jasionoides which form a clade with L. muscoides (§ Stenotium) with a BS value of 100 (Fig. 1). This corroborates Knox and Li (2017). Biogeography: Lammers (2011) described this species as endemic to the Cape provinces of South Africa. Chen et al. (2016) placed it in Africa as its ancestral area. Knox and Li (2017) corroborated both analyses. Our result places Africa as its ancestral region (Fig. 2) which corroborates the above-mentioned studies. The close relationship between L. jasionoides and L. muscoides, both statistically (BS = 100) and geographically (both in S. Africa), suggest a biphyletic nature of this section and/or the inclusion of L. jasionoides in Lobelia § Stenotium.

Lobelia § Rhynchopetalum (Giant Lobelioids/Rosettes). We sampled forty-eight out of the sixty-one species (Lammers 2011) in our combined plastid dataset. This section is polyphyletic. Dialypetalum, Brighamia, Delissea, Trematolobelia, and Lobelia § Revolutella E.Wimm. (Wimmer 1948) are all embedded within this section. Lobelia sessilifolia Lamb. (Lambert 1811) is sister to the members of this section however with a low support value of 26 (Fig. 1). This corroborates Chen et al. (2016), Crowl et al. (2016), and Knox and Li (2017) that the ancestor to giants lobelioids might have its ancestral region in S.E Asia. Biogeography: Lammers (2011) described this section as almost pantropical with species in three disjunct areas, that is, Southeast Asia, tropical Africa, and South America. According to our results, species with S.E. Asia as their ancestral region formed the basal group of this section. Lobelia sessilifolia forms a sister to the members of this section. Our analysis placed this taxon in S.E. Asia as its ancestral area, corroborating Chen et al. (2016) that the ancestor of the giant lobelioids could have been from S.E Asia. Our result corroborates that of Lammers (2011), Chen et al. (2016), and Knox and Li (2017) (Fig. 2).

Lobelia § Revolutella. We sampled three species out of nine (Lammers 2011): Lobelia niihauensis St.John (John 1931), Lobelia yuccoides Hillebr. (Hillebrand 1888), and Lobelia hypoleuca Hillebr. (Hillebrand 1888). They form a clade with a BS value of 100 (Fig. 1). This corroborates Givnish’s (1998), Antonelli’s (2008), Givnish et al.’s (2009), and Chen et al.’s (2016) results. Biogeography: Lammers (2011) and Chen et al. (2016) indicated the Hawaii archipelago as the ancestral area of sampled members of this section. A more detailed phylogenomic analysis by Hunter (2018) placed the origin of § Revolutella in Kaua`i. Our analysis corroborates the above-mentioned studies and places this section in the Pacific Islands as its ancestral region (Fig. 2).

Lobelia § Galeatella E.Wimm. (Wimmer 1948). In this section, we sampled one species, that is, L. kauaensis. Lammers (2011) included five species in this section, however, L. kauaensis was not amongst those included. Lammers (2007b) indicated that this species is a hybrid of natural taxa (nothotaxon). It forms a sister to Trematolobelia with a BS value of 90. Our results corroborate with that of Chen et al. (2016). Biogeography: Lammers (2007b, 2011) and Hunter (2018) stated that this section has its ancestral region in the Hawaiian archipelago. This corroborates with our results as it places this section in the Pacific Islands as its ancestral region.

Lobelia § Lobelia. Eleven out of twenty-two species (Lammers 2011) were sampled in our combined plastid dataset. This section is monophyletic. Lobelia inflata L. (Linnaeus 1753), and Lobelia kalmii L. (Linnaeus 1753) form a clade with a BS value of 68 and form sister to members of this section with a BS value of 96 (Fig. 1). This result corroborates Antonelli (2008). Biogeography: Lammers (2011) indicated North America as the ancestral location of this section. Our analysis concurs with Lammers (2011) and places North America as the ancestral area of this section (Fig. 2).

Lobelia § Cryptostemon (E.Wimm.) J.Murata (Murata 1995). We sampled two out of nine species included by Lammers (2011) in this section. The combined plastid dataset included Lobelia irasuensis Planch. & Oerst. (Planchon and Oersted 1857), whereas Lobelia fenestralis Cav. (Cavanilles 1791) was included in the ITS dataset. Lobelia irasuensis forms a clade with Lobelia divaricata Hook. and Arn. (Hooker et al. 1838) with a BS=96 (Fig. 1) while L. fenestralis forms a clade with Lobelia laxiflora Kunth (Kunth and Bonpland 1820) with a BS = 59. Biogeography: Chen et al. (2016) placed L. irasuensis in Central America as its ancestral region, which corroborated Antonelli (2009). Our analysis places this section in Central America as the ancestral region which corroborates with the above-mentioned studies (Fig. 2).

Lobelia § Homochilus DC. (Candolle 1839). We sampled two out of five species (Lammers 2011). Lobelia laxiflora Kunth (Kunth and Bonpland 1820) and Lobelia aguana E.Wimm. (Wimmer 1935) form a clade with a BS value of 48 (Fig. 1). Biogeography: Givnish et al. (2009) and Chen et al. (2016) indicated the ancestral area of these two species to be in Central America. Our results corroborate the above-mentioned and places this section in Central America.

Lobelia § Tylomium. Six out of thirty-eight species were sampled. This section is paraphyletic. Hippobroma longiflora (L.) G.Don (Don 1834) is nested within this section. Lobelia portoricensis Urb. (Urban et al 1899), Lobelia kraussii Graham (Graham 1830), Lobelia martagon Hitchc. (Hitchcock 1893), Lobelia stricta Sw. (Swartz 1788), and Lobelia vivaldii form a clade (Fig. 1). Biogeography: Lobelia assurgens L. (Linnaeus 1759), L. portoricensis, L. martagon, and L vivaldii were placed in the Greater Antilles as their ancestral region while L kraussi and L. stricta were placed in the Lesser Antilles as their ancestral region (Lammers 2011). Chen et al. (2016) indicated Central America as their ancestral area which corroborated our results.

Lobelia § Hypsela. Thirteen out of forty-three species within this section were sampled. This section is polyphyletic. It is intercalated with Isotoma, Hypsela, Pratia, and Lithotoma species. Lobelia fugax Heenan, Courtney & P.N.Johnson (Heenan et al 2008), Lobelia ionantha Heenan (Heenan et al 2008), Lobelia fatiscens Heenan (Heenan et al 2008), and Lobelia carens Heenan (Heenan et al 2008) form a clade with a BS value 98. Lobelia roughii Hook.f. (Hooker 1864), Lobelia linnaeoides Petrie (Petrie 1890), Lobelia macrodon (Hook.f.) Lammers (Lammers 1998), Lobelia glaberrima Heenan (Heenan et al 2008), and L. oligophylla also form a clade with a BS value of 70. Lobelia arnhemiaca forms a clade with P. purpurascens with a BS value of 61. Lobelia oligophylla is sister to L. angulata, L. roughii, L. macrodon, L. glaberrima, and L. linnaeoides (Fig. 1). This result corroborates that of Antonelli (2008). In the ITS region dataset, Pratia, Isotoma, Hypsela, and L. chinensis form a clade with a BS=100. Lammers (2011) indicated the chromosome number of this section as 2n=12, 14, 28, 42, 56, 70, 77, 84, 91 and 140. Pratia also shows these same chromosome number variations, consistent with it being an exemplar of interspecific hybridization. Pratia angulata is 2n=70 while P. perpusilla is 2n=42, the hybrids between these two species have 2n=77, 91, and 140 chromosome numbers reported (Murray et al. 2004). According to Knox et al. (2008b), Isotoma was distinguished by floral fusion with adnate filaments, Pratia on the other hand was classified using berry fruits, and Hypsela was differentiated by having both floral fusion and berry fruits. These features are similar to those used by Lammers (2011) used to describe the Lobelia § Hypsela. More so, Lammers (2011) suggested the inclusion of Isotoma in this section, a suggestion that is well supported by our phylogenetic analysis. Our analysis proposes the inclusion of Hypsela, Pratia, and Lithotoma in this section too. Biogeography: Lammers (2011) described this section as Amphi-pacific with a majority of the species in the southern hemisphere. Our analysis placed this section in Australasia as their ancestral region although L. loochooensis and L. chinensis were placed in Asia as their ancestral region. Lobelia oligophylla, on the other hand, has its ancestral region in South America. Our analysis corroborated that of Lammers (2011) (Fig. 2).

Lobelia § Tupa (G.Don) Benth. (Bentham 1876). We sampled all four members of this section in our combined plastid dataset. The sampled species included Lobelia polyphylla Hook. & Arn. (Hooker and Arnott 1830), Lobelia bridgesii Hook. & Arn. (Hooker and Arnott 1830), Lobelia tupa L. (Linnaeus 1753), and Lobelia excelsa Bonpl. (Bonpland et al. 1816), and form a monophyletic group with a BS value of 99 (Fig. 1). So far, this corroborates Lammers and Hensold (1992) and Antonelli (2008) that species belonging to this section might be monophyletic due to the uniform occurrence of an unusual chromosome number (2n=42). Biogeography: The ancestral area of this section is Chile (Lammers 2011). Chen et al. (2016) corroborated Lammers’s (2011) results. Our result is consistent with both of them and places South America as the ancestral region of this section.