Introduction

The genus Drimia Jacq. (Asparagaceae, subfamily Scilloideae, tribe Urgineeae sensu APG III1) (alternatively Hyacinthaceae subfamily Urgineoideae sensu APG II2) comprises approximately 110 bulbous geophytic species3,4 distributed in Africa, Madagascar, the Mediterranean basin and Asia5. The majority of the species (~93) are native to Africa. Currently, a total of eight species of the genus Drimia have been recognized in India viz. D. coromandeliana (Roxb.) Lekhak & P. B. Yadav, D. govindappae (Boraiah & Fatima) Lekhak & P. B. Yadav, D. indica (Roxb.) Jessop, D. nagarjunae (Hemadri and Swahari) Anand Kumar, D. polyantha (Blatt. & McCann) Stearn, D. raogibikei (Hemadri) Hemadri, D. razii Ansari and D. wightii Lakshmin6. Among them, seven species are endemic to the subcontinent6,7. Squill (European squill, D. maritima) is one of the most ancient medicinal plants. Since Stoll et al.8 isolated and crystallized scillaren A, a large number of bufadienolides have been reported from the bulbs of squill9. Bufadienolides (a class of cardiac glycosides) are the C-24 steroids with an α-pyrone group at position 17β9,10,11. The principle bufadienolides, i.e. scillaren A and proscillaridin A, isolated from Indian squill, D. indica10,12,13,14,15 are the same as those of the European squill, D. maritima8,10,16,17. Different species of Drimia show remarkable morphological similarities resulting in taxonomic misinterpretations5,18,19,20,21.

Several taxonomic revisions of Indian members of Drimia have been published18,22,23,24,25,26,27,28,29, relying solely on morphological characters for species delimitation6. Lekhak et al.7 and Yadav et al.6 inferred that, morphological characterization alone may not be sufficient to delimit interspecies relationship in this genus. To address this problem, a few studies have been conducted so far based on cytotaxonomy, karyotype, palynology, interspecific hybridization, nuclear DNA content, RAPD and SRAP markers, ITS and matK sequence data6,7,30,31,32,33. However, the phylogenetic relationships among the Indian members of Drimia still remain unclear.

Molecular phylogenetic studies have conventionally relied on comparison of homologous nucleotide sequences to establish a degree of similarity between closely related species. The use of nuclear and/or organellar non-coding sequences has greatly assisted our understanding of relationships and circumscriptions at all levels of the taxonomic hierarchy in plant phylogenetic studies34,35,36,37,38. Analysis of plastid DNA sequences has proven to be very useful in the phylogenetic study of Hyacinthaceae5. The potential use of leaf anatomical characteristics in the species level phylogeny has also been well documented in different monocot plant groups, particularly in Hyacinthaceae39,40,41. As far as we are aware, no comparative study has been carried out on leaf morpho-anatomical features of Indian species of Drimia. The stomatal traits of the monocot leaves have been considered as important taxonomic markers in different levels of systematic hierarchy37,42. Similarly, pollen grain characteristic, especially the exine micromorphology has also been reported to be very useful in defining the evolutionary trends in many plant families43,44,45,46.

The aim of the present study is to infer the phylogenetic relationships within the Indian species of Drimia based on cpDNA trnL intron, rps16-trnK intergenic spacer, atpB-rbcL intergenic spacer and rDNA ITS1-5.8S-ITS2 sequences, along with leaf morpho-anatomical, stomatal and pollen exine micromorphological characteristics.

Results

Out of the eight Indian species6, the present study deals with 12 accessions representing seven species of Drimia (Table 1). To investigate the phylogenetic relationships cpDNA non-coding (trnL intron, rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer) and nuclear rDNA ITS1-5.8S-ITS2 sequences of the collected species of this genus were analysed along with the characterization of leaf morpho-anatomical and pollen exine morphological features.

Table 1 List of collected Indian species of Drimia with their site of collection, somatic chromosome number7,32,33,51 and NCBI GenBank accession numbers.
Full size table

Phylogeny of Drimia (subfamily Scilloideae) inferred from cpDNA trnL intron sequences

The maximum likelihood (ML) phylogenetic tree comprising a total of 54 taxa was rooted with two closely related outgroup taxa47 i.e. Tradescantia pallida and Weldenia candida (Fig. 1). All ingroup members representing six different subfamilies of the family Asparagaceae fell into three major clades (I, II and III). The clade-I was subdivided into two subclades representing Scilloideae and Brodiaeoideae with bootstrap values (BS) 98% and 95% respectively. The clade-II consisted of only Agavoideae (BS 94%), while clade-III was subdivided into three subclades representing Nolinoideae (BS 95%), Lomandroideae (BS 98%) and Asparagoideae (BS 99%). In subfamily Scilloideae of clade-I, all Indian members of Drimia, viz. four populations of D. indica, two populations each of D. polyantha and D. wightii, and the single populations of D. coromandeliana, D. govindappae, D. nagarjunae and D. razii grouped together and originated from a single node (BS 84%), supporting the monophyly of the Indian members. Within the clade, D. wightii, D. govindappae and D. nagarjunae grouped with weak support (BS 64%) (Fig. 1). The three non-Indian species (D. maritima and D. undata from Europe and D. sanguinea from southern Africa) were weakly supported (BS 54%) as sister to the Indian species (Fig. 1).

Figure 1
figure 1

Maximum likelihood phylogeny of the genus Drimia (subfamily Scilloideae) based on cpDNA trnL intron sequences. Numbers beneath nodes are Bootstrap support (BS) indices. Black arrows indicate formation of six subclades. Dotted arrow indicates the origin of all studied Indian members of Drimia from a common node.

Full size image

Phylogenetic relationships among the Indian species of Drimia based on concatenated sequences of cpDNA rps16-trnK intergenic spacer, atpB-rbcL intergenic spacer and trnL intron

In order to clarify the interspecies relationships among the Indian members of the genus Drimia, a phylogenetic tree based on combined sequences of cpDNA rps16-trnK intergenic spacer, atpB-rbcL intergenic spacer and trnL intron was reconstructed. Both MP and ML methods yielded identical topologies (only MP tree is shown in Fig. 2). The Indian species of Drimia were split into two distinct clades (Fig. 2). The clade-I was subdivided into two subclades (I- II), where subclade-I included D. coromandeliana, D. indica and D. polyantha, while the subclade-II included D. govindappae, D. nagarjunae and D. wightii. This topology was similar to the cpDNA trnL intron-based tree (Fig. 1). Among the members of subclade-I, all four populations of D. indica clustered together, while both the populations of D. polyantha originated from a single node (BS 100%) and D. coromandeliana emerged as a sister taxon. On the other hand, the clade-II of the cpDNA concatenated sequence-based tree consisted of D. razii (Fig. 2).

Figure 2
figure 2

Maximum parsimony phylogeny among the Indian species of Drimia based on concatenated sequences of cpDNA trnL intron, rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer. Numbers beneath nodes are Bootstrap support (BS) indices. Two selected morphological characters of taxonomic importance (pollen exine pattern and mesophyll cell characteristics) have been mapped on the tree.

Full size image

Phylogenetic relationships among the Indian species of Drimia based on rDNA ITS1-5.8S-ITS2 sequence

Phylogenetic relationships among the Indian species of Drimia were also assessed based on rDNA ITS1-5.8S-ITS2 sequences. The ML tree (Supplementary Fig. S1) resolved two distinct clades (I- II). The clade-I (BS 100%) comprising D. coromandeliana, D. indica and D. polyantha (Supplementary Fig. S1) was similar to that of the subclade-I of the cpDNA concatenated sequence-based tree (Fig. 2). The remaining species of Drimia i.e. D. nagarjunae, D. govindappae, D. wightii and D. razii formed the clade-II (BS 78%) on the basis of their rDNA ITS1-5.9S-ITS2 sequence complexity (Supplementary Fig. S1).

Morpho-anatomical and stomatal characteristics of leaf

Various qualitative and quantitative morphological parameters of leaf were evaluated (Supplementary Table S1). All Indian species of Drimia are characterized by green or glaucous, fleshy leaves. The species vary in the number of leaves per bulb (LN) (Table 2). The greatest number of leaves per bulb (13.8 ± 0.29 to 14.2 ± 0.24) was found in the four populations of D. indica, while D. razii has the fewest (6.2 ± 0.35). Three major types based on leaf shape (LS) i.e. lanceolate-straight (0), linear (1) and lanceolate-curled (2) were observed among the studied taxa. The leaves in D. indica, D. coromandeliana, D. polyantha and D. nagarjunae are lanceolate-straight [character state = 0], linear [character state = 1] in D. razii, and lanceolate-curled in D. wightii and D. govindappae [character state = 2] (Supplementary Fig. S2a; Table 2). The longest leaves (43.9 ± 0.64 cm) were recorded in D. indica population-IV and the widest in D. nagarjunae (4.0 ± 0.02 cm) while D. razii had the shortest (21.1 ± 0.38 cm) and narrowest (0.2 ± 0.0 cm) (Table 2).

Table 2 Morphological and anatomical characterization of leaves of 12 Indian accessions of Drimia*.
Full size table

Several qualitative and quantitative anatomical characters were also studied (Supplementary Table S1). The species were found to vary in the cross-sectional shape of the leaves (LST.S) (Table 2). The leaves of all species except D. razii were subulate in shape [character state = 0] whereas leaves of D. razii were polygonal in section [character state = 1] (Table 2). Anatomically, the basic leaf features of the genus Drimia include the presence of a thick cuticle, an epidermis of barrel or rectangular shaped cells, and a series of chlorenchymatous mesophyll tissues. The mesophyll tissues could be further categorized into two types i.e. compact, single layered palisade cells and loosely arranged irregularly shaped, multiple layered spongy cells with intercellular spaces aligned horizontally next to the inner layer of palisade cells. In addition, two types of palisade cells (PT) were observed among the studied samples, i.e., columnar [character state = 0] in D. wightii, D. razii, D. govindappae, D. nagarjunae (Supplementary Fig. S2b; Table 2) and spherical types [character state = 1] in D. indica, D. coromandeliana and D. polyantha (Supplementary Fig. S2c; Table 2). The length of palisade cells (LPC) varied significantly among the species (Table 2). The maximum LPC (22.8 ± 0.18 μm) was observed in D. razii followed by D. govindappae (22.0 ± 0.21 μm) and D. nagarjunae (21.5 ± 0.10 μm). Drimia polyantha and D. indica showed minimum LPC (ranging from 9.3 ± 0.03 μm to 9.5 ± 0.08 μm). The collateral vascular bundles were found at regular intervals, with adaxial phloem and relatively well-developed xylem abaxially. One to two layers of compactly arranged spherical bundle sheath cells were found around the vascular bundles of each species. Additionally, significant differences in leaf thickness in cross-section (LTT.S) ranging from 104.0 ± 0.10 μm to 334.0 ± 0.11 μm were observed (Table 2).

Stomatal traits are considered as one of the important taxonomic markers in delimiting species37,48. The present study revealed that all taxa were characterized by anomocytic type (without subsidiary cells) of stomata (Fig. 3a–g). However, the length and breadth of stomata were found to be species-specific (Table 3, Fig. 3a–g). The stomatal index (SI) varied approximately 1.8-fold among the species (Table 3). Maximum SI was recorded in D. polyantha while D. indica and D. wightii showed minimum SI (Table 3). The species also differed in the length (EL) and width (EW) of the surrounding epidermal cells. Maximum EL was recorded in D. nagarjunae (416.0 ± 0.04 μm; Fig. 3e) and minimum EL (197.1 ± 0.08 μm) and EW (16.9 ± 0.08 μm) in population-I of D. wightii (Fig. 3f; Table 3).

Figure 3
figure 3

Stomatal characteristics of seven Indian species of Drimia. (a) D. indica population-I; (b) D. coromandeliana; (c) D. razii; (d) D. polyantha population-I; (e) D. nagarjunae; (f) D. wightii population-I; (g) D. govindappae.

Full size image
Table 3 Stomatal characterization of leaves of 12 Indian accessions of Drimia*.
Full size table

Pollen exine morphology based on scanning electron microscopic (SEM) study

Pollen morphological traits have been used as taxonomic markers in Hyacinthaceae19,49. In this study we examined the exine surface architectures of pollen grains of all the collected species of Drimia except D. razii and D. nagarjunae (in which flowering was not observed) under SEM. All species had monosulcate, ellipsoidal grains (Fig. 4) but two distinct types of exine ornamentation were observed. Reticulate exine [character state = 1] was observed in D. indica, D. coromandeliana and in D. polyantha (Fig. 4a–c) while pollen grains of D. wightii and D. govindappae were characterized by perforate exine [character state = 0] (Fig. 4d and 6e). An earlier report on D. razii pollen grains showed them to be of the perforate type (or fine reticulate)19 and this was species was therefore coded as ‘character state = 0’ and used for further analysis on ancestral state reconstruction.

Figure 4
figure 4

Scanning electron microscopic images showing pollen exine ornamentation patterns in five Indian species of Drimia. (a–c) Reticulate exine in D. indica population-I, D. coromandeliana and D. polyantha population-I; (d–e) Perforate exine in D. govindappae and D. wightii population-I.

Full size image

Interspecies relationship based on UPGMA phenogram analysis using combined leaf morpho-anatomical and stomatal data

The cophenetic correlation for the obtained UPGMA phenogram (Fig. 5) was 0.964, indicating a good fit between the cophenetic value matrix and the average Euclidean distance matrix. The observed phenogram (Fig. 5) revealed the formation of three distinct clusters (I-III). The cluster-I consisted of D. razii, while the cluster-II included D. govindappae, D. nagarjunae and two populations of D. wightii. The cluster-III was composed of two populations of D. polyantha, D. coromandeliana and four populations of D. indica (Fig. 5) which was similar to the subclade-I of the cpDNA concatenated sequence-based tree (Fig. 2).

Figure 5
figure 5

UPGMA phenogram derived from average Euclidean distance between taxa of the genus Drimia using combined morphological, anatomical and stomatal data.

Full size image

Ancestral state assessment for type of palisade cell (PT) and pollen exine pattern (PEP)

We reconstructed the ancestral state for the type of palisade cells (PT) of leaf and pollen exine patterns (PEP) in Indian Drimia. A phylogenetic tree based on combined sequences of cpDNA trnL intron, rps16-trnK intergenic spacers and atpB-rbcL intergenic spacers was constructed and used as a backbone for tracing ancestral character (Supplementary Figs S3 and S4). The obtained maximum parsimony (MP) tree, rooted with D. razii, revealed the distribution of character states in the terminal taxa and the evolutionary history of both leaf and pollen characters was studied. The ancestral state reconstruction clearly showed that columnar palisade cells (Supplementary Fig. S3) and perforate exine of pollen (Supplementary Fig. S4) were the ancestral characters in Indian Drimia species. The exine ornamentation in D. nagarjunae was not observed in the present study. The results revealed that D. razii, D. wightii, D. govindappae and D. nagarjunae retained the ancestral states for leaf palisade cell characters, while D. polyantha, D. coromandelian and D. indica shared derived spherical palisade cells (Supplementary Fig. S3) and reticulate pollen exine architecture (Supplementary Fig. S4).

Discussion

This study demonstrated, for the first time, an explicit phylogenetic relationship among seven different Indian species of Drimia on the basis of cpDNA non-coding sequences (trnL intron, rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer), rDNA ITS1-5.8S-ITS2 sequence data, leaf morpho-anatomical, stomatal data and pollen exine morphological data.

Our analysis shows that the Indian species of Drimia that were sampled comprise a monophyletic group derived from a common ancestor (Fig. 1), thereby supporting the theory that India may be a secondary centre of evolution for this genus19. A previous analysis of systematic relationship among five Indian taxa of Drimia based on ITS and matK DNA sequence variations30 did not include the two important Indian taxa i.e. D. govindappae and D. nagarjunae, and may require re-evaluation. The present study included 12 accessions representing seven Indian species of Drimia and was made on the basis of concatenated sequences of cpDNA trnL intron rps16-trnK intergenic spacers and atpB-rbcL intergenic spacers (Fig. 2), and was further confirmed by analysing the rDNA ITS1-5.8S-ITS2 sequence-based tree (Supplementary Fig. S1). The phylogenetic trees rooted with two Asparagus species (outgroups) retrieved, D. razii as sister to the other six species, which grouped into two evolutionary lines, providing deeper insights into the interspecies relationships in the genus in India. Nath et al.33 also suggested the grouping of all the Indian species of Drimia into two major complexes based on karyological data.

Boraiha and Khaleel27 first recognized D. govindappae as distinct from D. indica and D. coromandeliana, although Dixit and Yadav18 reported successful hybridization between D. indica and D. govindappae. However, intermediate forms of these species have not been observed in nature so far. Based on floral morphology, Deb and Dasgupta25 treated D. coromandeliana, D. nagarjunae and D. govindappae as synonyms of D. indica. On the contrary, the present UPGMA analysis (Fig. 5) clearly demarcates these species on the basis of combined leaf morphological, anatomical and stomatal data. The UPGMA phenogram revealed the formation of three distinct clusters (I- III) among the studied Indian taxa of Drimia (Fig. 5), which is positively correlated with the results obtained from the molecular phylogenetic tree (Fig. 2).

The Indian species of Drimia include both day-blooming and night-blooming species with two types of life cycle patterns, viz. synanthous (leaves and flowers appeared simultaneously) and hysteranthous (leaves and flowers appear in different seasons)6,7,50. Yadav and Dixit19 observed peculiarity in the time of flower opening and closing among the closely related Drimia species. Differences in the timing of anthesis among the different taxa of Drimia also induced troubles in the hybridization experiments51. Peruzzi et al.40 discussed the utility of several leaf anatomical traits including different types of palisade cells in the grouping of different species of the genus Ornithogalum (Hyacinthaceae). The present study confirms the value of stomatal characteristics in interspecies delimitations among the genus Drimia in India. In our earlier work on Asparagus, the evolutionary significance of different stomatal as well as the surrounding epidermal cell traits in species-level phylogeny of the subgenus Protasparagus was clearly demonstrated37. Pollen exine morphology was highly congruent with molecular data (Fig. 2). Yadav and Dixit19 studied the pollen exine ornamentations of four Indian species and categorized them accordingly. Pehlivan and Özler49 characterized different taxa of Muscari (Hyacinthaceae) on the basis of pollen surface ornamentation.

The present phylogenetic analysis also sheds light on the evolution of two important leaf and pollen characters i.e. type of palisade cells (PT) and pollen exine patterns (PEP) of the Indian species of Drimia (Supplementary Figs S3 and S4). The parsimony ancestral states reconstruction using the combined sequences of cpDNA trnL intron rps16-trnk intergenic spacer and atpB-rbcL intergenic spacer suggested that the columnar palisade cell is ancestral in Drimia in India and spherical palisade cells are derived. Similarly, a possible trend of evolution from perforate to reticulate pollen exine ornamentation is also suggested in the present analysis. However, further investigations on different vegetative and floral characters of the genus Drimia including D. nagarjunae pollen grains and allied members of Scilloideae are needed to infer evolutionary significance of the present observation.

In conclusion, the present research work demonstrates an explicit phylogenetic relationship among seven Indian species of Drimia on the basis of both molecular and leaf morpho-anatomical characters for the first time. This study also highlights the possible evolution of exine ornamentation. Altogether, the present research work brings out new insights on species diversification of Drimia in India and provides important background information for further studies on their biogeography.

Methods

Taxon sampling

Out of eight Indian species recognized by Yadav et al.6, a total of 12 accessions representing seven species of Drimia were used in the present study (Table 1). Voucher specimens were deposited in the Herbarium of Shivaji University, Kolhapur (SUK). All the samples (bulbs) have been grown and maintained for more than eight years in the net house at the experimental garden of the Department of Botany, University of Calcutta (elev. 9 m, 22.5275° N, 88.3628° E). A representative number of individual plants for each taxon adapted in a similar environment have been used for the present phylogenetic analysis. A total of 54 accessions representing all the subfamilies of Asparagaceae (except the monogeneric subfamily Aphyllanthoideae) were analysed for cpDNA trnL intron sequence-based phylogeny of the genus Drimia. Among them, sequences of 42 accessions representing five subfamilies of Asparagaceae (Brodiaeoideae, Agavoideae, Asparagoideae, Lomandroideae and Nolinoideae) including two outgroup taxa (Tradescantia pallida: Accession no.: AM113705.1 and Weldenia candida: Accession no.: AJ387746.1) were retrieved from the NCBI public database (http://www.ncbi.nim.nih.gov) (Supplementary Table S2) following the taxonomic classification of APG III47.

Genomic DNA isolation and PCR amplification of cpDNA trnL intron, rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer regions

Genomic DNA was isolated from young leaves of each of the taxa using CTAB method52. The quality of DNA in each sample was checked by 1.0% (w/v) agarose gel electrophoresis. DNA concentration was measured using Eppendorf BioSpectrophotometer. The amplification of the cpDNA trnL intron region was performed in a programmable thermal cycler (Mastercycler Nexus, Eppendorf AG 22331 Hamburg) using the universal forward primer: 5′-CGA AAT CGG TAG ACG CTA CG -3′ and reverse primer: 5′-GGG GAT AGA GGG ACT TGA AC-3′, as described in Saha et al.37. PCR cycling conditions were followed according to Taberlet et al.53. For amplification of each of the cpDNA rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer region, a 25 μl reaction was setup with 2.5 μl of 10X PCR buffer along with15 mM MgCl2 (Genei, Bangalore), 0.5 μl of 10 mM dNTP mix (Genei, Bangalore), ~100 ng template DNA, 0.5 μl of Taq DNA polymerase (5U/μl) (Genei, Bangalore) and 1 μl of each primer (4.0 pM/μl). Both the forward (F) and reverse (R) primers specific to cpDNA rps16-trnK intergenic spacer (F: 5′-AAA GTG GGT TTT TAT GAT CC-3′ and R: 5′-TTA AAA GCC GAG TAC TCT ACC-3′35) and atpB-rbcL intergenic spacer (F: 5′-ACA TCK ART ACK GGA CCA ATA A-3′ and R: 5′-AAC ACC AGC TTT RAA TCC AA-3′54) were commercially synthesized by GCC Biotech (India) Pvt. Ltd. Kolkata, India. The PCR cycling conditions were as follows: for rps16-trnK intergenic spacer: initial denaturation at 95 °C for 3 min followed by 30 cycles at 95 °C for 30 sec, annealing at 48 °C for 30 sec, extension at 72 °C for 1 min; final extension was at 72 °C for 8 min and for atpB-rbcL intergenic spacer: initial denaturation at 94 °C for 2 min followed by 30 cycles at 94 °C for 1 min, annealing at 50 °C for 1 min, extension at 72 °C for 1 min; final extension was at 72 °C for 8 min.

PCR amplification of rDNA internal transcribed spacer region (ITS1-5.8S-ITS2)

For the amplification of rDNA ITS1-5.8S-ITS2 region, specific primers (forward primer: 5′-GAA TGG TCC GGT GAA GTG TTC GG-3′ and the reverse primer: 5′-CGC CTG ACC TGG GGT CGT G-3′) were designed using NCBI primer blast software (http://www.ncbi.nim.nih.gov) and were commercially synthesized by Integrated DNA Technologies (RFCL Limited, New Delhi, India). 25 μl PCR reaction mix contained 2.5 μl of 10X PCR buffer along with 15 mM MgCl2 (Genei, Bangalore), 1.0 μl of 10 mM dNTP mix (Genei, Bangalore), ~100 ng template DNA, 1.0 μl of Taq DNA polymerase (5 U/μl) (Genei, Bangalore) and 1 μl of each primer (4.0 pM/μl). The PCR cycling conditions were as follows: initial denaturation at 95 °C for 3 min followed by 30 cycles at 95 °C for 30 sec, annealing at 66 °C for 45 sec, extension at 72 °C for 1 min; final extension was at 72 °C for 8 min.

DNA sequencing

All the PCR amplicons of cpDNA trnL intron, rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer and rDNA ITS1-5.8S-ITS2 of the studied species of Drimia were sequenced using the Big Dye Terminator cycle sequencing method (Xcelris Labs Ltd, Gujarat, India; http://www.xcelrislabs.com). Chromatograms of all the DNA sequences were analyzed by using Bio-Edit.v.7.1.3 software (Ibis Biosciences, Carlsbad, CA 92008). Multiple sequence alignments were performed using ClustalW (http://www.genome.jp/tools/clustalw) with Gap Open Penalty: 15 and Gap Extension Penalty: 6.66. All the newly generated sequences have been deposited in the NCBI GenBank database (http://www.ncbi.nim.nih.gov) under accession numbers listed in Table 1.

Phylogenetic analysis using cpDNA non-coding sequences

Phylogenetic analysis using cpDNA trnL intron sequences of 54 accessions representing six subfamilies of Asparagaceae and two outgroup taxa (Supplementary Table S2) was performed by maximum likelihood method using the partitioned model option with MEGA 6.0655. Based on Bayesian information criterion (BIC) and Akaike information criterion, corrected (AICc) using MEGA 6.0655, the best-fit nucleotide-substitution model was found to be T92 + G (Tamura 3-parameter model), with the lowest BIC score (5267.637), and lowest AICc score (4414.472). Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value (-2098.72). The tree was drawn to scale, with branch lengths measured in the number of substitutions per site. A discrete Gamma distribution was used to model evolutionary rate differences among sites [5 categories (+G, parameter = 1.1548)]. Positions containing gaps and missing data were eliminated from the datasets (complete deletion option). Bootstrapping of the datasets was performed with 1000 replications56.

In order to clarify the interspecies relationship among the studied members of Drimia, a phylogenetic analysis was further conducted using the concatenated sequences of cpDNA rps16-trnK intergenic spacer, atpB-rbcL intergenic spacer and cpDNA trnL intron following Farris et al.57. A total of 14 accessions including 2 outgroup taxa viz. Asparagus officinalis [Accession nos.: AB613992.1 (rps16-trnK intergenic spacer), AY147755.1 (atpB-rbcL intergenic spacer), KJ774036.1 (trnL intron)] and Asparagus setaceus cultivar Pyramidalis [Accession nos.: AB613995.1 (rps16-trnK intergenic spacer), JF784417.1 (atpB-rbcL intergenic spacer), KJ774038.1 (trnL intron)] were aligned with the ClustalW programme (with Gap Open Penalty: 15 and Gap Extension Penalty: 6.66) in the MEGA 6.06 package55. The phylogenetic analysis of the aligned matrix was performed by both maximum likelihood (ML) and maximum parsimony (MP) methods. The T92 model (Tamura 3-parameter model) of nucleotide-substitutions for the ML analysis of the concatenated sequences of three cpDNA non-coding sequence data was determined by the lowest BIC (8118.609) and AICc scores (7897.829). The Subtree-Pruning-Regrafting (SPR) algorithm58 with search level 1 was used to obtain the MP tree. The initial trees were obtained by the random addition of sequences (10 replicates). Bootstrap analyses were performed on 1000 replicates56.

Phylogenetic analysis using rDNA ITS1-5.8S-ITS2 sequence

Phylogenetic analysis using the rDNA ITS1-5.8S-ITS2 sequence was also performed for further confirmation of the degree of relatedness among the studied members of Drimia. A total of 14 accessions including 2 outgroup taxa (Asparagus officinalis Accession no. KJ868767.1 and Asparagus setaceus cultivar Pyramidalis Accession no. KJ885623.1) were aligned with the clustalW programme with gap open penalty 15 and gap extension penalty 6.66 (http://www.genome.jp/tools/clustalw). The phylogenetic analysis was done by ML method with MEGA 6.0655 as mentioned in the above. The best-fit nucleotide-substitution model was found to be T92 + I (Tamura 3-parameter model), with the lowest BIC score (3793.582), and lowest AICc score (3601.744). The bootstrap method was employed with 1000 replications56.

Morphological and anatomical characterization of leaf

A minimum of 10 mature leaves from three separate individual plants of each taxon of the genus Drimia was used for evaluating different qualitative and quantitative morphological and anatomical parameters (Supplementary Table S1). The morphological parameters included shape of leaf (LS), number of leaves per bulb (LN), leaf length (LL) and width (LW). LL was measured from base to tip of the fully expanded leaf blade while LW was measured from margin to margin at the middle portion of the leaf blade. For leaf anatomical studies, free-hand cross sections were prepared from the middle part of the leaf with a sharp razor blade. At least three sections were used for analysis and scoring of different anatomical characters viz. shape (LST.S) and thickness (LTT.S) of leaf in t.s., type (PT) and length (LPC) of palisade cells according to Saha et al.37 and Chatterjee et al.59. Sections were observed under Leitz BioMed compound microscope and photographed with the attached ProgRes CT5 digital image documentation system. The observed qualitative characters (LS, LST.S and PT) were converted into character states (binary and multistate) and were finally scored for each taxon (Supplementary Table S1). The experiment was repeated thrice.

Stomatal characterization

For stomatal characterization, a minimum of 10 mature leaves from three separate individuals of each taxon of Drimia was studied. For uniformity, only the middle portion of the mature leaf was considered and epidermises were peeled off to analyse different stomatal characters (Supplementary Table S1) following the method described in Saha et al.37. Both the upper and lower epidermal peels were observed and photographed under Leitz BioMed compound microscope equipped with digital camera ProgRes CT5. Length (SL in µm) and width (SW in µm) of stomata, stomatal index (SI), length (EL in µm) and width (EW in µm) of the surrounding epidermal cells were measured from at least 10 stomatal complexes selected randomly. The stomatal index (SI) was determined by calculating the average number of stomata and the number of epidermal cells per microscopic field (area: 205892.61 µm2) following the protocol of Reginato et al.60. All kinds of measurements were done using the software package ProgRes Capture Pro 2.8.8 (Jenoptik Optical System). The experiment was repeated thrice.

Scanning electron microscopic (SEM) analysis of pollen grains

The exine surface architectures of the pollen grains of three separate individuals of five collected species and 10 accessions of Drimia (D. indica population-I, II, III and IV, D. coromandeliana, D. polyantha population-I and II, D. govindappae and D. wightii population-I and II) were studied by SEM analysis following the protocol of Talbot and White61. Pollen grains were mounted on the aluminium stubs using a double adhesive carbon tape and sputter coated with a 20–30 nm thick film of Au/Pd under S150 Sputter Coater. The sample containing stubs were examined at 15 Kv accelerating voltage and photographed under a SEM-EDX unit (SEM-Carl Zeiss Evo-40 EDX- Oxford Instrumentation) [GSI, Geological Survey of India, Kolkata].

Statistical analysis

Descriptive statistics including means and standard errors and one-way analysis of variance (ANOVA) was carried out to test the significance of variation in the leaf traits of the studied taxon62. Tukey’s B multiple range tests was used for post hoc analyses. The statistical analysis was conducted at 0.05 probability level using SPSS v16.0 statistical package. To determine the interspecies relationship among the members of Drimia based on morphological, anatomical, and stomatal characters, cluster analysis was conducted on the Euclidean distance matrix with the unweighted pair group method using arithmetic averages (UPGMA) with default data transformation and normalization option with the InfoStat version 2013d (Free version) software package. To calculate the average Euclidean distance, a combination of the observed variables (morphological, anatomical and stomatal) per accession was analysed, which included: eight quantitative (LN, LPC, LTT.S, SL, SW, SI, EL and EW), two binary (LST.S and PT) and one multistate (LS) characters (Supplementary Table S1).

Ancestral state reconstruction

To study the evolution of the leaf and pollen characters among the Indian species of Drimia, two important traits, viz. type of palisade cell (PT) and pollen exine pattern (PEP) were selected based on the earlier reports19,30,39,41. The character states of PT and PEP (Supplementary Table S1) of each taxon were used to reconstruct their ancestral states using Mesquite 3.31 software63. This software analyses the character state at the terminal taxa and graphically represents the history of character evolution. A phylogenetic tree was reconstructed based on combined sequence data from the three chloroplast non-coding DNA segments (cpDNA trnL intron, rps16-trnK intergenic spacer and atpB-rbcL intergenic spacer) using Mesquite heuristic search method. Alignment of the input sequences was done using Muscle 3.8.31 programme. The obtained phylogenetic tree was then served as a backbone to study the transition parameters for ancient and recent state reconstruction of morphological traits (PT and PEP) using maximum parsimony method.