Fruit morphology helps identifying evolutionary groups in Alpinieae (Zingiberaceae): inferences from phylogenetic analysis of gingers in Sri Lanka

: To assess the systematics and the phylogenetic placement of the members of the two important genera Alpinia Roxb. and Amomum Roxb. in Sri Lanka, molecular data of twelve in-group species together with three outgroup taxa of the family Zingiberaceae were extensively analysed for phylogenetic significance. The current analysis of the evolutionary relationships of the Sri Lankan members of the genera of interest, utilising DNA sequence data of the chloroplast genome regions trn L-F and trn S-fM, has resolved four groups of Alpinia and two major clades of Amomum with substantial parsimony analysis and Bayesian inferences consistency values. With new accessions from the entire native range of the family, this result points to the need for an inevitable re-circumscription of the genus Alpinia , and shows congruence to the recent reshuffling of the genus Amomum and the family Zingiberaceae, except for the placement of each genus as monophyletic groups in the context of our study. Here, we suggest swapping the group defining species of an Alpinia clade observed in previous studies. Finally, our study suggests the use of fruit morphology to distinguish among recognised groups for Sri Lankan species as they exhibit positive correspondence.


INTRODUCTION
The two largest genera Alpinia and Amomum of the family Zingiberaceae, not only show a wide geographical distribution (Kiew, 1982;Smith, 1990;Larsen et al., 1998;Wu & Larsen, 2000;Sirirugsa, 2001) but also possess an array of homoplasious and plesiomorphic characters that make the classification of the genera problematic (Kress et al., 2002;Lamxay & Newman, 2012). Recent phylogenetic analyses of Alpinia and Amomum revealed the polyphyletic nature of the genera (Rangsiruji et al., 2000a;Xia et al., 2004;Boer et al., 2018) rejecting the previous classification which was based solely on morphological characters (Schumann, 1904;Smith, 1990). Boer et al. (2018) presented a more thorough phylogenetic analysis of Alpiniae, concluding the presence of 10 Amomum clades and six Alpinia clades. Further, they proceeded with re-circumscription of the Amomum clade by resurrecting three genera (Conamomum, Meistera, and Wurfbania) and describing three new genera (Epiamomum, Lanxangia, and Sundamomum). However, the current phylogenetic and evolutionary analyses of the two genera are far from being completed as they lack many representative species (e.g. species from South and Southeast Asia). Furthermore, only a few of the Sri Lankan species have been used in the aforementioned studies [Amomum pterocarpum, Meistera echonocarpa (formerly Amomum echinocarpum), Alpinia abundiflora and Alpinia fax] despite the fact that there are several taxonomically paramount species occurring in Sri Lanka (Burtt & Smith, 1983). In this paper, phylogenetic placement of  Journal of the National Science Foundation of Sri Lanka 49 (3) the Sri Lankan members of the two genera of interest is comparatively and comprehensively discussed, using the inferences drawn from the DNA sequence analysis.
With the recognition of the polyphyletic origin of the previously recognised tribes of the family Zingiberaceae (Sakai & Nagamasu, 1998;Rangsiruji et al., 2000a;Kress et al., 2002;Harris et al., 2003;Kress & Specht, 2005), several generic-level phylogenetic studies addressed the evolutionary relationships of the genera Alpinia (Rangsiruji et al., 2000b; and Amomum (Kaewsri, 2006;Kaewsri et al., 2007;Droop, 2012) as well as the tribe Alpinieae (Kaewsri et al., 2007;Kress et al., 2007, Boer et al., 2018 nesting two genera in comparative assessments. A common conclusion of these studies was that the members of the tribe do not follow the previous classification system and instead they tend to show different inter-and infra-generic evolutionary relationships indicating the polyphyletic nature of Alpinia and Amomum. Nuclear and plastid DNA sequence analysis of Alpinia indicated that taxa described under this genus are closer to other genera of Alpinieae than they are to each other making classification of the tribe difficult (Rangsiruji et al., 2000a; due to its polyphyletic origin. Rangsiruji et al. (2000a) identified four separate clades of Alpinia whereas Boer et al. (2018) described the separation of the genus into six groups. However, none of these studies agreed with the comprehensive infra-generic classification of Alpinia by Smith (1990), where she identified two subgenera within the genus encompassing 221 species. Similarly, during an analysis to assess the species boundaries of Aframomum, Harris et al. (2000) suggested that Amomum might be paraphyletic. Later studies by Xia et al. (2004) demonstrated the segregation of the genus Amomum into three major clades and they used fruit morphology to explain the clades observed. In another study by Kaewsri et al. (2007) using amplified fragment length polymorphism (AFLP) makers to analyse the phylogenetic relationships of Thai Amomum showed the positive correspondence of fruit morphology to the clades observed. Droop (2012) explained the presence of at least seven groups of Amomum, inferred by the results of the phylogenetic analysis using Internal transcribed spacer (ITS) and matK data, whereas Boer et al. (2018) found 10 clades of Amomum using the same gene regions and they used both anther crest and fruit morphology in combination to discriminate the clades.
Although Alpinia and Amomum add the highest number of endemic species to the ginger family in Sri Lanka (Burtt & Smith, 1983), taxonomy and phylogenetic relationships of the members of these two genera has been poorly studied. The last taxonomic revision of the family Zingiberaceae in Sri Lanka was in 1983 (Burtt & Smith, 1983). Since then, several major revisions have taken place in the world utilising new phylogenetic tools, which resulted in major changes in the classification of the family (Rangsiruji et al., 2000a;Xia et al., 2004;de Boer et al., 2018). More importantly, none of the Sri Lankan endemic species had been used in any of the previous phylogenetic analyses. This observation justifies the need for a thorough phylogenetic account of Sri Lankan members of the family Zingiberaceae. As a pioneering step, the present study employed molecular data to determine evolutionary relationships among the members of the two genera Alpinia and Amomum in Sri Lanka and to compare the results with the current major revisions.
The goals of this study are: to construct phylogenetic hypothesis of interspecific and intergeneric relationships of the two genera primarily focusing on Sri Lankan species and compare the results with currently existing classifications; and to assess the congruence of several morphological characters with the molecular analysis results.

Ingroup taxa
Representative samples from 12 species of both genera were collected for the study (Figure 1 and 2). These included two native Alpinia species, four species of nonnative Alpinia, four native Amomum species, one endemic Amomum species, and a new record of Amomum from Sri Lanka [Wurfbainia villosa var. zeylanicus (formerly Amomum villosum zeylanicus)] - Karunarathne et al. (2014). Unfortunately, four species of endemic Amomum (A. nemorale, A. graminifolium, A. benthamianum, and A. trichostachyum), one species of endemic Alpinia (A. rufescens) and one native Amomum species (A. hypoleucum) were not found during comprehensive field sampling. Most of these species have not been collected for more than 100 years and almost all types of localities are under intense human activities. Therefore, we present Journal of the National Science Foundation of Sri Lanka 49 (3) September 2021 an adequate sampling of extant taxa of the two genera in Sri Lanka (Table 1).

Outgroup taxa
Two representative species of Zingiber (Z. cylindricum and Zingiber sp.) and one species of Hedychium (H. coronarium) were collected during the field sampling for out-group taxa. For a better comparison of the overall phylogeny and the geographical distribution of the phylogenetic clusters, DNA sequences of trnL-trnF intergenic spacer region of four Aframomum, three Renealmia, and one Elettariopsis accessions were also downloaded, following a nucleotide blast from GeneBank for outgroup taxa (supplementary Table 1). These three genera are sister groups to the studied two genera in the present study (Kress & Specht, 2005;Kress et al., 2007).

DNA extraction and sequencing
Fresh leaf samples dried with silica beads were used for DNA extractions. Genome DNA was extracted using a modified Doyle and Doyle (1987) CTAB protocol with few additional purification steps: suspended DNA in TE was treated with ProteinaseK and RNase followed by phenol chloroform extraction and sodium acetate precipitation of DNA. PCR amplification was done for two chloroplast gene regions trn L-F and trn S-ƒM, using trn L-F (Taberlet et al., 1991) and trn S-ƒM (Shaw et al., 2005) primer pairs, respectively. The chloroplast genome regions matK, and trn L-F together with nuclear ribosomal ITS regions have been explicitly used in evolutionary studies of the family Zingiberaceae (Johnson & Soltis, 1994;Baldwin et al., 1995;Kress et al., 2002). However, only the chloroplast intergenic spacer trn L-F and the non-coding region trn S-fM were used because in the initial steps, the primers that were used for ITS and matK regions failed to amplify the desired region for several species (A. echinocarpum, A. pterocarpum, and A. fax) and two species produced multiple bands for ITS region (A. acuminatum and A. fulviceps). The region trn S-fM has been documented to be informative in phylogenetic studies (Shaw et al., 2005;Minami et al., 2009 (Larkin et al., 2007). Alignment files were further edited in BioEdit to process unaligned bases. This was repeated for all sequences of the studied two DNAs separately. A combined analysis of the two regions was also performed to obtain better resolved trees and thereby to estimate phylogeny with the strongly supported results (Nixon & Carpenter, 1996).
Furthermore, to conduct a more inclusive sampling of the two genera, sequences of both Alpinia and Amomum were downloaded from GeneBank to analyse with the accessions from the present study. DNA sequences of trnL-trnF intergenic spacer region of six Alpinia and seven Amomum accessions were downloaded following a nucleotide blast from GeneBank (Supplementary Table 1).

Maximum Parsimony Analysis (MPA)
MPA of the aligned sequence data was accomplished using PAUP* 4.0 (Swofford, 2003) on a power Macintosh G4. Heuristic search strategies were used in 1000 Tree Bisection Reconnection (TBR) branch swapping replicates of random taxon addition with MULPARS in effect. The majority-rule strict consensus tree assembly was used to build the most parsimonious trees. All positions were weighted equally; gaps were treated as missing values. The Consistency Index (CI) and Retention Index (RI) values for tree topologies were also calculated with PAUP. Bootstrap analyses (Felsenstein, 1985) were conducted using PAUP with a simple addition of 1000 fast-swap replicates to 100 tree topologies with a 50 % confidence.

Bayesian Inferences
Bayesian analysis was used to draw model based phylogenetic inferences (Yang & Rannala, 1997;Larget & Simon, 1999) from the DNA sequence data. MrBayes (Huelsenbeck & Ronquist, 2001) was used for the analysis to obtain the posterior probabilities (PP). Models for the sequence evolution was determined using jModeltest (Posada, 2008) and the following parameters were used in the analysis; GTR, nst = 6, rates = inverse gamma, statefreqpr = fixed (equal). The analysis comprised generations of 10,000,000 Markov Chain Monte Carlo (MCMC) chains with a sampling frequency of every 10,000 generations. The initial 25 % samples from each run were discarded as burn-in, after which the run was summarised as a majority rule consensus tree and the posterior probability values were saved in MrBayes. Separate analysis of the two regions and the combined analysis produced a similar tree topology for the strict consensus. Therefore, only the combined tree is presented here.
Journal of the National Science Foundation of Sri Lanka 49 (3) September 2021

Analysis of trn L-F intergenic spacer data
A total of 449 parsimony-informative characters were present in the trn L-F region, which produced 5473 equally parsimonious trees in more than 270,000 rearrangement steps in parsimony analysis. Strict consensus of these trees (Supplementary Figure 1)      Bayesian analysis of the trn L-F and trn S-fM sequence data resulted in cladograms similar to that of the MPA. All the clades observed in the maximum parsimony analysis were supported with higher posterior probability (PP) values in the Bayesian analysis. The two genera are monophyletic with strong PP value of 1.00. The Alpinia group formed two clades in the analysis separating A. abundiflora and A. fax from the rest with a PP value of 0.98. Overall, the Bayesian inferences of the phylogenetic trees is congruence with the results of MPA.
The complexity of the classification of the two polyphyletic genera Alpinia and Amomum has been acknowledged by many authors (Smith 1990;Larsen et al., 1998;Kress et al., 2002). As a result, many researchers who were interested in the family Zingiberaceae have dealt with taxonomy, phylogeny and the classification of these two genera. Remarkably, the phylogenetic relationships of the members of the family in Sri Lanka have never been assessed and none of the endemic species have been assessed in any of the recent phylogenetic studies of the family (e.g. Kress et al., 2002, Xia et al., 2005, Boer et al,. 2018, Droop 2019. Hence, the current study attempted phylogenetic analysis of the Sri Lankan members of the two genera Alpinia and Amomum. Results of the present analysis provides better insight to the evolutionary relationships of the studied species in Sri Lanka. According to the latest phylogenetic analysis of Amomum group (de Boer et al., 2018), Sri Lankan taxa of the Amomum group consist of three genera (i.e. Amomum, Wurfbainia and Meistera). However, the present study has not resulted in similar groups, and was rather congruent with the previous studies of the family Zingiberaceae (Rangsiruji et al., 2000a;Kress et al., 2002) by resolving only two clades. This may be due to the limited sampling in our study representing all the species of Amomum group. Although the present analysis does not include accessions representing all the genera of the family, our study highlighted two Amomum clades with strong statistical support (Figure 3). Therefore, in congruence with infra-generic relationships that were illuminated in recent studies (Kress et al., 2007;Droop, 2012), it can be concluded that Sri Lankan members of the two genera of interest make two evolutionary groups. Reduced staminodes, capsule: globose, echinate or irregular shaped, and relatively short stamen can be listed as a combination of features that characterise the separation of the clade.

Alpinia
Although all the Alpinia species studied in this analysis formed a monophyletic group, the separation of the clades within the genus (i.e. Alp I, Alp II, Alp III and Alp IV) agrees with the broader studies of the family (Rangsiruji et al., 2000a;Kress et al., 2002). These groups are in congruence with the recent phylogenetic studies on the family (Rangsiruji et al., 2000a;Kress et al., 2002Kress et al., , 2007 and also confirm the incongruence with earlier morphological treatments (Schumann, 1904;Smith, 1990).

September 2021
Journal of the National Science Foundation of Sri Lanka 49 (3) A. abundiflora and A. fax were placed as the basal group (Alp I), separately from all the other members of the genus with strong support values in both MPA and Bayesian analyses of the molecular data. Recognition of this clade was also described in phylogenetic analyses of Alpinia by  and Rangsiruji et al. (2000b). As the latter study explained, these species (namely A. fax clade) are characterised by the presence of capitate, usually radical inflorescence surrounded by a world of sterile bracts; fertile bracts subtending a cincinnus of several flowers. Furthermore, this observation reinforces the suggestion by Kress et al. (2007) that these species require a new generic name.  also suggested that the inclusion of A. rufescens in this clade is appropriate since this species also shares this distinctive morphological character combination. A. rufescens has been recorded only in Sri Lanka. It is only known from the type locality and has never been recollected since its type gathering (1862). During our extensive field sampling efforts (comprised of more than 25 locations) it was observed that the type locality of A. rufescens is no longer a potential area to support these species since the area is under intense human activities. This species was not found from any other location either. After close and careful observation of the characters of this clade, it could be suggested that A. abundiflora be used as the clade defining species instead of A. fax because A. abundiflora possesses typical characters such as capitate inflorescence, occurrence of the inflorescence either (rarely) on the vegetative stem or on a separate basal peduncle and typically a cincinnus of seven (7) flowers. These are also shared with A. rufescens (except for the number of flowers in a cincinnus). On the other hand, the inflorescence of A. fax rather elongates with age and it bears a lesser number of flowers in a cincinnus (seldom up to six flowers). Further, Karunarathne et al. (2015) reported the confusion of morphological characters of A. fax that was also presumably recorded in South India (Kumar et al., 2002).
Although the inclusion of Alpinia galanga with the rest of Sri Lankan Alpinia (except Alp I) showed weak support values (bootstrap 56 %, PP = 0.58: combined analysis), the separation of groups Alp I, through Alp IV was strongly supported. The clade Alp I separated A. abundiflora and A. fax from the rest with strong support (bootstrap 75 %, PP = 0.98). Next to the A. abundiflora clade, Alp II isolated A. galanga basal to the A1 group. A. galanga clade is characterised by higher number of flowers in the branched inflorescence, smaller size of flowers, clawed labellum and the glabrous ovary. Inclusion of additional species such as A. nigra (not found in our field observations) might better explain these characters. Although A. calcarata was included as a member of the A. zerumbet clade in previous studies, the present results placed it as a separate clade (Alp III). Linear-lanceolate shape of the lamina and the reduced lamina width along with the smaller sized fruits can be listed as deviating characters. A. zerumbet and A. malaccensis formed the terminal clade with strong bootstrap support. Besides the robustness of the habit of the plant, relatively large, globose fruits are characteristic to this clade.

Amomum
All the studied Sri Lankan Amomum species resolved into two major clades (Am I and Am II). This exhibits congruence with recent evolutionary analyses of the genus (Xia et al., 2004;Kaewsri et al., 2007;Droop, 2012, Boer et al., 2018. Results of the current analysis also do not correspond to earlier infra-generic classification of the genus (e.g. Schumann 1904;Smith 1990).
A. pterocarpum was placed as the basal most lineage (clade Am I) of the Amomum group. The clade is characterised by the presence of ribbed capsule with wings, sub-globose inflorescence with quickly deciduous bract and bracteoles. Am II clade resembles the A. maximum clade which was observed in the study by Xia et al. (2004) and other studies (Kress et al., 2007;Droop, 2012). The main characteristic feature of this group is the presence of smooth or winged fruit coat (Xia et al., 2004). Am II placed all the other members of the genus (A. echinocarpum, A. masticatorium A. fulviceps and A. acuminatum, A. villosum zeylanicus) in a single clade with strong branch support values. The most conspicuous morphological characters that shared among these species are slender basal peduncle covered with scaly sheaths, ovoid to elongated inflorescence, brownred bracts and bracteoles, tri-lobed anther appendage, bifid labellum and echinate fruits. This is in congruence with the recent evolutionary analysis of the genus. Xia et al. (2004) identified this group as A. villosum clade with echinate fruits being a distinctive character and later studies confirmed this placement (Xia et al., 2004;Kress et al., 2007;Droop, 2012 this separation of polyphyletic clades was not observed in the present study. Moreover, after comprehensive observation of the morphological characters of each clade resulted in the current phylogenetic analysis, besides the abovementioned group specific characters, fruit morphology of the studied species can be effectively utilised to separate different clades of the Sri Lankan members of Alpinia and Amomum. Figure 4 shows a detailed attribution of the fruit morphology of the groups for this context. However, we admit that we are not the first to use the fruit morphology to separate phylogenetic clades in the family Zingiberacea (Kress et al., 2002;Xia et al., 2005;Droop et al., 2012;Boer et al., 2018). Nevertheless, the use of the fruit morphology for Sri Lankan species presents an efficient strategy.

Sri Lankan gingers in a global phylogeny
The three Alpinia groups observed in our analysis are in close congruence with the most recent thorough analysis of the genus . Most importantly the analysis shows that Alpinias are not monophyletic, contrary to our analysis only with Sri Lankan species (Figure 4). This is expected as several other evolutionary groups in the family included in the analysis, and as observed in the phylogenetic analysis of the family (Kress et al., 2002;2005;de Boer 2018). In contrast, the Amomum species in Sri Lanka formed a monophyletic group with strong bootstrap support (100 %). Furthermore, this resolution of one Amomum clade in Sri Lanka corresponds to the same Amomum villosum clade as in Xia et al. (2004) of the Amomum phylogeny. This suggests that all the Sri Lankan species of the genus

September 2021
Journal of the National Science Foundation of Sri Lanka 49 (3) evolved from one common ancestor although the genus itself in the region is not monophyletic. However, none of the other species of Amomum in this group were included in the previous evolutionary analyses of the genus. Therefore, a comparison of the results of this study is not appropriate. Most importantly, even though the evolutionary groups within the two genera slightly changed with the global phylogeny, our use of the fruit morphology still applies to the evolutionary groups found in Sri Lanka (Figure 4). Moreover, this observation can further be extended to identify corresponding clades of the two genera in other regions as well, as they show high congruence to previous studies by Xia et al. (2004),  and Boer et al. (2018).
Overall, the current analysis of the evolutionary relationships of Sri Lankan Alpinia and Amomum utilising DNA sequence data of the chloroplast genome regions trn L-F and trn S-fM and the combined dataset resolved four groups of Alpinia and two major clades of Amomum. This result is in congruence and consistent with recent phylogenetic analyses of the two genera using molecular data to a certain extent, except for the placement of each genus as monophyletic groups in the context in the present analysis. This discrepancy in comparison to other recent studies seems to arise from the fact that none of the endemic Sri Lankan species were used in previous studies. Finally, the present study suggests using the fruit morphology to distinguish the observed groups for Sri Lankan species because they exhibit positive correspondence with the fruit morphology, which is also consistent.