Phylogeny, infrageneric classification and species delimitation in the Malagasy Impatiens (Balsaminaceae)

Abstract The species-rich genus Impatiens (Balsaminaceae) is represented in Madagascar by no less than 260 species. We conducted molecular phylogenetic analyses of the Malagasy Impatiens based on nuclear and plastid data and 52 accessions (representing 33 species) to: 1) reassess the monophyly of the Malagasy Impatiens; 2) assess the monophyly of the sections Preimpatiens (Humblotianae and Vulgare groups) with spurs and Trimorphopetalum without spurs as delimited by Perrier de la Bâthie, as well as that of the subgenera Impatiens and Trimorphopetalum as defined by Fischer and Rahelivololona; 3) infer the current status of some morphologically variable species; and 4) test whether the species of Impatiens from the Marojejy National Park form a monophyletic group. The Malagasy Impatiens are further confirmed to be paraphyletic with respect of the Comorian I.auricoma. The present sectional and subgeneric classifications of the Malagasy Impatiens are partly supported, with strong support for the monophyly of the sect. or subgen. Trimorphopetalum. Section Preimpatiens was not supported as monophyletic and neither the Humblotianae group nor the Vulgare group is monophyletic. Impatienselatostemmoides, I. “hammarbyoides”, I.inaperta, I.lyallii and I.manaharensis are either para- or polyphyletic and may represent morpho-species. The Impatiens species from the Marojejy National Park do not form a monophyletic group and therefore are suggested to be derived from numerous independent colonisation events from all over Madagascar followed by subsequent diversifications.


Introduction
The genus Impatiens L. (Balsaminaceae) is a monophyletic and diverse genus with more than 1000 species (e.g. Yuan et al. 2004, Janssens et al. 2009) and is represented by at least 260 endemic species in Madagascar (e.g. Fischer and Rahelivololona 2002, 2004a, b, 2015a, b, c, 2016, Fischer et al. 2003, 2017. This continental island is one of the centres of species diversity for the genus, which is the largest flowering plant genus on the island (Perrier de la Bâthie 1934, 1948, Humbert 1955, 1956, Fischer and Rahelivololona 2002, 2004a, b, 2007a, b, 2015a, b, c, 2016, Fischer et al. 2003, 2017 (Figs 1, 2). Rahelivololona (e.g. 2002, 2004a, b) initiated the taxonomic studies of the Malagasy and Comorian members of Impatiens in an attempt to produce an updated Flora of the family Balsaminaceae for Madagascar and the Comoros. Since then, 75 new species have been described and at least another 75 new species will be published in the near future (E. Rahelivololona and E. Fischer, unpubl. data). The majority of the Malagasy Impatiens occurs in the montane regions of northern and eastern Madagascar (e.g. Tsaratanana National Park with 36 species, Marojejy National Park with 48 species, Masoala National Park with 59 species) (Figs 1, 2). Warburg and Reiche (1895) provided the first global infrageneric classification for Impatiens based solely on morphological data. Since then, a number of infrageneric classifications of the genus have been proposed for some tropical regions (including Madagascar). Perrier de la Bâthie (1934) established the first sectional classifications for the Malagasy and Comorian Impatiens, placing the species with obvious spurs and anthers with apical dehiscence in his section Preimpatiens. The author subdivided sect. Preimpatiens into two groups: the Vulgare group with pink, purple, white or orange flowers with shorter and slender spurs and the Humblotianae group with red, yellow or orange flowers with larger and broader spurs. Furthermore, Perrier de la Bâthie (1934) classified the Malagasy Impatiens species with anthers dehiscing apically but without spurs on the low sepals into two sections: the monotypic sect. Impatientella with deltoid and sharp anthers and sect. Trimorphopetalum with obtuse or truncate anthers (Fig. 2), both endemic to Madagascar. Fischer and Rahelivololona (2002) recognised Perrier de la Bâthie's sect. Impatiens with spurs ( Fig. 1) and sect. Trimorphopetalum without spurs (Fig. 4) at subgeneric level. They formally subsumed sect. Impatientella into subgen. Trimorphopetalum.
The category of species is widely accepted as the basic or working unit of biological classification (Rosell et al. 2010, Hohenegger 2014. From a phylogenetic point of view, a species of traditional taxonomy is often viewed as a species hypothesis tested by recovering either monophyletic or non-monophyletic units. Most species concepts (e.g. Baum and Shaw 1995) consider monophyly to be congruent with species hypotheses, while almost all species concepts view polyphyly as a rejection of species hypotheses (e.g. Rosell et al. 2010). Previous and contemporary workers (e.g. Perrier de la Bâthie 1934, Fischer and Rahelivololona 2002, Fischer et al. 2003, 2017, dealing with species circumscription in the Malagasy Impatiens, have encountered difficulties in deciding whether a taxon represents a "real" species or a morpho-species. Several species (e.g. I. elatostemmoides H.Perrier, I. "hammarbyoides" Eb.Fisch. & Raheliv. (nomen ) are known to be morphologically variable and this raises doubts as to whether these species deserve specific status. Molecular trees (phylogenetic hypotheses) can identify mono-, para-or polyphyletic taxa and can therefore be an important tool for assessing species delimitation. In addition, phylogenetic trees can also be used for assessing the various infrageneric classifications of the Malagasy Impatiens.
The first molecular phylogenetic study of the family Balsaminaceae by Yuan et al. (2004), based on nuclear ribosomal ITS (nrITS) sequence data and including 18 Malagasy Impatiens species (six species from subgen. Trimorphopetalum and 12 species from subgen. Impatiens), indicated that the Malagasy Impatiens species were polyphyletic, as they were resolved into three groups: a Humblotianae-Vulgare-Trimorphopetalum clade, a lineage with I. baroni Baker of sect. Impatiens and a Humblotianae-Vulgare clade consisting of seven Malagasy species of sect. Impatiens (I. anovensis H.Perrier to I. vilersii Costantin & Poiss.). The authors also showed the monophyly of the Malagasy sect. Trimorphopetalum only if sect. Impatientella, containing the spurless and entirely cleistogamous species I. inaperta, is included. Their results suggested a Malagasy origin of the Comorian species. In addition, each of the Malagasy Impatiens clades was nested within an African Impatiens lineage, suggesting multiple African origins of the Malagasy Impatiens. Moreover, subgen. Impatiens was not monophyletic. In contrast to Yuan et al. (2004), Janssens et al. (2006Janssens et al. ( , 2007Janssens et al. ( , 2009, who included representatives of subgen. Impatiens and Trimorphopetalum, strongly supported the monophyly of the Malagasy Impatiens (including the Comorian I. auricoma). These last three studies also confirmed a single African origin of the Malagasy representatives. However, all of the above-mentioned molecular studies (Yuan et al. 2004 with 17 species;Janssens et al. 2006Janssens et al. , 2007Janssens et al. and 2009 with six species) used a very limited sampling of the Malagasy Impatiens and, therefore, the monophyly of subgen. Trimorphopetalum was in need of being tested with a much larger sampling effort. More recently, based on both morphological and molecular evidence, Yu et al. (2015) divided the genus Impatiens into two subgenera, subgen. Clavicarpa and subgen. Impatiens. Moreover, the authors delineated seven sections in subgenus Impatiens: sect. Semeiocardium, sect. Tuberosae, sect. Racemosae, sect. Impatiens, sect. Scorpioidae, sect. Fasciculatae and sect. Uniflorae. Of these, the latter is characterised by short fusiform capsules and includes all Malagasy species of Impatiens, as well as several African and Asian species.
The Marojejy National Park is located in north-eastern Madagascar within the SAVA Region. With its tallest peak rising to 2137 m, the area is home to a diverse flora of upland species. The wide range of elevations and rugged topography of Marojejy create diverse habitats, which transition quickly with changes in altitude. There are four types of forests within the park: lowland rainforest below 800 m (Fig. 3); moist montane rainforest between 800 and 1400 m (Fig. 3); sclerophyllous montane cloud forest between 1400 and 1800 m; and ericoid shrub above 1800 m (Fig. 4) (Humbert 1955). The higher summits are covered by subalpine grassland with small ericaceous shrubs (Fig. 4), and are home to numerous local endemic species of Impatiens and of other large genera, such as Streptocarpus (Gesneriaceae) and Helichrysum (Asteraceae). The park has been recognised as a marked centre of plant endemism. For example, 32 palm species found in the Marojejy area are endemic to Madagascar and seven of them are restricted to the park. Of the 18 species of tree ferns, inventoried in the   Whether the endemic species of Impatiens from the Marojejy form a monophyletic group or are the result of a mixture of colonisation events from other regions through time have yet to be assessed.
The main objective of this study was to reconstruct a new and larger phylogeny of the Malagasy Impatiens, with a particular emphasis on taxa collected from Marojejy, using two nuclear AP3/DEF homologues (ImpDEF1 and ImpDEF2) and the plastid atpB-rbcL spacer. The resulting phylogeny was subsequently used to (i) reassess the monophyly of the Malagasy Impatiens as stated by Janssens et al. (2006Janssens et al. ( , 2007Janssens et al. ( , 2009); (ii) assess the monophyly of the sections Preimpatiens (Humblotianae and Vulgare groups) and Trimorphopetalum as delimited by Perrier de la Bâthie (1934), as well as that of the subgenera Impatiens and Trimorphopetalum (including sect. Impatientella) as defined by Fischer and Rahelivololona (2002); (iii) assess the current species status of the morphologically variable species I. elatostemmoides, I. "hammarbyoides", I. inaperta and I. manaharensis, using monophyly as the primary criterion (Backlund and Bremer 1998); and iv) test whether the species of Impatiens from the Marojejy National Park form a monophyletic group. The sectional classification proposed by Yu et al. (2015) could not be assessed, as our sampling was solely addressing the Malagasy and Comorian Impatiens species.

Taxon sampling
The taxon sampling for this study was focused on the Malagasy representatives of the genus Impatiens. We expanded the previous dataset of Yuan et al. (2004) and Janssens et al. (2009) with 29 new accessions (Table 1). In total, 52 accessions were included in our analyses, representing 48 Malagasy specimens (representing about 31-33 species), two Comorian accessions (representing one species) and two African species (Table 1). This sampling represented the three major groups defined by Perrier de la Bâthie (1934) as occurring in Madagascar and the Comoros: 31 accessions from sect. Preimpatiens (Perrier de la Bâthie, 1934) or subgen. Impatiens (Fischer & Rahelivelolona, 2002) (10 accessions representing six species of the Humblotianae group; 21 accessions representing 11or 12 species of the Vulgare group); and 22 accessions representing about 14 species from sect. Trimorphopetalum (Perrier de la Bâthie, 1934) or subgen. Trimorphopetalum (Fischer & Rahelivololona, 2002). The species showing some morphological variation were represented by more than one individual and, thus, were the subject of a test for monophyly. Twenty-seven accessions, represented by at least 17 species, were from the Marojejy National Park. Impatiens cecilii and I. hydrogetonoides, both from Africa, were used as outgroup based on Janssens et al. (2006Janssens et al. ( , 2009.

Molecular protocols
Total genomic DNA was isolated from silica-dried leaf material using a modified CTAB protocol (Doyle and Doyle 1987), which was optimised for Impatiens by Janssens et al. (2006Janssens et al. ( , 2009. The two nuclear AP3/DEF homologues (ImpDEF1 and ImpDEF2) and the plastid atpB-rbcL intergenic spacer were amplified following Janssens et al. (2007) and Janssens et al. (2006). PCR reactions for all three gene markers investigated in this study consisted of 2 min initial denaturation at 94 °C and 30 cycles of 30 s denaturation at 94 °C, 30 s primer annealing at primer specific temperature and 1 min extension at 72 °C. Primer annealing for ImpDEF1, ImpDEF2 and atpB-rbcL were at 57 °C, 55.5 °C and 51 °C, respectively. Amplification reactions were carried out on a Gene Amp PCR system 9700 (Applied Biosystems). Purified amplification products were sent to Macrogen, Inc. (Seoul, South Korea) for sequencing. Sequences obtained in this study will be deposited at GenBank (Table 1).

Data analyses
Contiguous sequences were assembled using Geneious v7.0.6 (Biomatters, New Zealand). Automatic alignment of the datasets was carried out with MAFFT (Katoh et al. 2002) under an E-INS-i algorithm, a 100PAM/k=2 scoring matrix, a gap open penalty of 1.3 and an offset value of 0.123. Subsequent manual fine-tuning of the aligned dataset was done in Geneious v7.0.6. Congruency between the nuclear and chloroplast datasets was inferred by a partition homogeneity test as implemented in PAUP*4.0b10a (Swofford 2000). The best-fit nucleotide substitution model for each plastid and nuclear dataset was determined using jModel Test 2.1.4 (Posada 2008) under the Akaike information criterion (AIC). The GTR+I+G model was found as best fit for ImpDEF1, whereas the GTR+G model was calculated as best substitution model for ImpDEF2 and atpB-rbcL. A mixed-model approach was used in which the combined dataset was partitioned in order to apply a different model of evolution on each DNA region (Ronquist and Huelsenbeck 2003). Bayesian Inference (BI) analyses were conducted with MrBayes v3.1 (Huelsenbeck and Ronquist 2001) on three individual data partitions and a combined data matrix. Each analysis ran two times for 10 million generations. Trees were sampled every 2500 generations. Inspection of chain convergence and ESS parameters was done with TRACER v1.4 (Rambaut and Drummond 2007). Bayesian posterior probability (BPP) values between 0.50 and 0.95 were considered to be weakly supported, whereas BPP values above or equal to 0.95 were taken into consideration to indicate well-supported branches (Suzuki et al. 2002, Alfaro et al. 2003. Maximum Likelihood analyses were carried out on the CIPRES web portal using RAxML v7.2.8 (Stamatakis et al. 2008) under the GTR-GAMMA model. Non-parametric ML bootstrapping analysis was calculated with 1000 bootstrap replicates.

Results
The aligned atpB-rbcL and ImpDEF1/ImpDEF2 matrices contained 924 bp and 812 bp, respectively. The phylogenetic tree, based on the combined data, is shown in Figure  5. The monophyly of the Malagasy Impatiens was not supported by our analyses, as the Comorian species I. auricoma was deeply nested within the Malagasy Impatiens Clade I (BS: 99; BPP: 1). Therefore, the Malagasy Impatiens are paraphyletic, unless the Comorian Impatiens species are included. The sect. Preimpatiens sensu Perrier de la Bâthie (1934)/subgen. Impatiens sensu Fischer and Rahelivololona (2002) was not resolved as a monophyletic group. Neither the Humbotianae group nor the Vulgare group was supported as monophyletic (Fig. 5). However, sect. Trimorphopetalum sensu Perrier de la Bâthie (1934)/subgen. Trimorphopetalum sensu Fischer and Rahelivololona (2002) was strongly supported as a monophyletic group (BS: 92; BPP: 1). The earliest diversified lineages in the Malagasy Impatiens clade fell into a large polytomy containing five groups ( The phylogenetic relationships amongst these major lineages were unresolved. Within the Impatiens Clade II, the morphologically variable I. manaharensis was not supported as monophyletic. Within the Impatiens Clade III, the morphologically variable species I. elatostemmoides appeared polyphyletic, while I. "hammarbyoides", I. lyallii and I. inaperta, also variable, seemed paraphyletic. Finally, the sampled species of Impatiens from the Marojejy National Park did not form a monophyletic group, as they were scattered across the tree (Fig. 5).

Discussion
The present analyses confirm the paraphyly of the Malagasy Impatiens with respect to the Comorian I. auricoma. This is consistent with Janssens et al. (2006Janssens et al. ( , 2007Janssens et al. ( , and 2009 but inconsistent with the polyphyly of the Malagasy Impatiens as shown by Yuan et al. (2004). The monophyly of sect. Trimorphopetalum (I. inaperta included) suggested by Yuan et al. (2004) is further strongly supported (BS: 92; BPP: 100) by the present study. In other words, subgen. Trimorphopetalum, as delimited by Fischer and Rahelivololona (2002), is supported. Spurless, greenish, brown to blackish or yellowish (never white, pink or purple) flowers with boat-shaped lower sepals, obtuse or truncate and apically dehiscing anthers and the lack of extrafloral nectaries on leaf lamina and petioles are the synapomorphic characters for this lineage, which seems to have evolved from a common ancestor with spurs (Fig. 5). In Yuan et al. (2004), the Trimorphopetalum clade was resolved as the most derived within Impatiens. The authors argued that this spurless lineage could not be recognised at sectional or subgeneric level, as proposed by Perrier de la Bâthie (1934) and Fischer and Rahelivololona (2002), respectively, because this taxonomic decision seems to make sect. Preimpatiens sensu Perrier de la Bâthie (1934) or I. subgen. Impatiens sensu Fischer and Rahelivololona (2002) paraphyletic. Our results do not support or reject Yuan et al. (2004)'s claims, as the Malagasy Impatiens clade (including the Comorian I. auricoma) is largely unresolved (Fig. 5). Neither the sampled species from the Vulgare group nor those from the Humblotianae group form a monophyletic group, a result consistent with Yuan et al. (2004). Therefore, our results provide no support for the Vulgare group characterised by shorter and slender spurs or for the Humblotianae group defined by larger and broader spurs, as delimited by Perrier de la Bâthie (1934). Furthermore, our analyses do not support or reject the monophyly of sect. Preimpatiens sensu Perrier de la Bâthie (1934) or subgen. Impatiens sensu Fischer and Rahelivololona (2002). To summarise, this study partly supports the sectional and subgeneric classifications of the Malagasy Impatiens proposed by Perrier de la Bâthie (1934) and Fischer and Rahelivololona (2002). More molecular data are needed to further assess the monophyly of sect. Preimpatiens or subgen. Impatiens.
In addition, results of this molecular phylogenetic study further highlight the difficulties that the Impatiens taxonomists have faced when dealing with the species delimitation of the Malagasy Impatiens (e.g. Perrier de la Bâthie 1934 ;Humbert 1956;Fischer and Rahelivololona 2002, 2004a, 2015a. Impatiens manaharensis seems polyphyletic and this supports our suspicion in the field that at least two taxa with very different morphology could be distinguished within this variable species. Impatiens elatostemmoides seems polyphyletic, while I. inaperta, I. "hammarbyoides" and I. lyallii (Fig. 5) appear paraphyletic. Therefore, these morphologically variable species may well represent morpho-species, meaning that they represent a group of several different species or are parts of a species complex. As a consequence, this study indicates that the current species delimitation of these para-or polyphyletic species is in need of revision.
Finally, the Impatiens taxa from the Marojejy National Park do not form a monophyletic group, as they are spread across the tree (Fig. 5). This suggests that they are the result of numerous independent colonisation events from elsewhere in Madagascar, followed by subsequent diversifications. In other words, they seem to have had multiple origins.

Conclusions
The Malagasy Impatiens are paraphyletic with respect of the Comorian I. auricoma. The present subgeneric and sectional classifications of the Malagasy Impatiens are partly supported, with strong support for the monophyly of subgen. Trimorphopetalum. Neither the Humblotianae group nor the Vulgare group forms a monophyletic group. Impatiens elatostemmoides, I."hammarbyoides", I. inaperta and I. manaharensis are either para-or polyphyletic and may represent morpho-species. The Impatiens species from Marojejy do not form a natural group. A further study based on a much larger molecular data set and sampling from the entire geographic ranges of Impatiens in Madagascar is needed to produce a well-resolved phylogeny. This will hopefully allow for a retest of the monophyly of sect. Preimpatiens sensu Perrier de la Bâthie (1934) or subgen. Impatiens sensu Fischer and Rahelivololona (2002), as well as molecular dating and biogeographic analyses of the Malagasy Impatiens.
Laboratory of Plant Systematics, Institute of Botany and Microbiology, Leuven for technical and financial support of the molecular laboratory work, P. Küpfer, Y.-M. Yuan, S. Wohlhauser, M. Callmander and P. Druart (all formerly University of Neuchâtel) for introducing her to molecular phylogenetics, technical training on molecular biology and financial support at the beginning of her studies; the DGF (Direction Générale des Forêts) and the MNP (Madagascar National Parks) in Madagascar for issuing collecting and exportation permits to EMR and EF; the Parc Botanique et Zoologique de Tsimbazaza (PBZT), Antananarivo, Madagascar, for allowing her to conduct the research on Impatiens; and the two anonymous reviewers and the Associate Editor (Hanno Schaefer) for their comments on an early version of the paper. The second author (EF) is indebted to the Missouri Botanical Garden and the Parc Botanique et Zoologique de Tsimbazaza for arranging our research and collecting permists. The Akademie der Wissenschaften und Literatur Mainz kindly sponsored the field trip of EF. Special thanks go to our colleagues Damien Ertz (Meise), Bernard Goffinet (Connecticut), Roger Lala Andriamiarisoa (Antananarivo) and Emmanuël Sérusiaux (Liège) who accompanied us in the field. Without the skills of Roger Lala Andriamiarisoa, the fieldwork would have been impossible.