Systematic Placement of the Enigmatic Southeast Asian Genus Paralamium and an Updated Phylogeny of Tribe Pogostemoneae (Lamiaceae Subfamily Lamioideae)

Paralamium (Lamiaceae) is a monotypic genus within the subfamily Lamioideae and has a sporadic distribution in subtropical mountains of southeast Asia. Although recent studies have greatly improved our understanding of generic relationships within Lamioideae, the second most species-rich subfamily of Lamiaceae, the systematic position of Paralamium within the subfamily remains unclear. In this study, we investigate the phylogenetic placement of the genus using three datasets: (1) a 69,276 bp plastome alignment of Lamiaceae; (2) a five chloroplast DNA region dataset of tribe Pogostemoneae, and (3) a nuclear ribosomal internal transcribed spacer region dataset of Pogostemoneae. These analyses demonstrate that Paralamium is a member of Pogostemoneae and sister to the monotypic genus Craniotome. In addition, generic-level phylogenetic relationships within Pogostemoneae are also discussed, and a dichotomous key for genera within Pogostemoneae is provided.

The genus Paralamium was originally described by Dunn (1913) and reported to be endemic to southeast Asia with a sporadic distribution in humid regions of southwestern China (subtropical Yunnan), northern Vietnam, northern Burma, and eastern India (Assam) (Li and Hedge, 1994;Harley et al., 2004;Suddee and Paton, 2004). The genus is distinguished from other Lamioideae genera mostly based on calyx morphology. Paralamium has unequal calyx-lobes, with the posterior calyx tooth being the largest and having a truncate apex flanked by smaller triangular lateral lobes, and lanceolate-triangular anterior lobes (Figure 1). Harley et al. (2004) called this unique calyx morphology a 1/2/2 split, while Li and Hedge (1994) recognized this shape as a 1/4 split. In addition, this genus is characterized by possessing very small pollen grains with the polar length and/or equatorial width less than < 18 µm (Harley et al., 2004), which is an uncommon feature within Lamiaceae.
Paralamium is monotypic, with the sole species, P. gracile Dunn (1913) described on the basis of a specimen collected from Yunnan, China (Henry 10636). However, before the description of this species, Hooker (1885) described Plectranthus griffithii Hook.f. based on a collection from Eastern Assam, India (Herb Griffith 4056). After careful examination of the type materials, Suddee and Paton (2004) suggested that Plectranthus griffithii Hook.f. and Paralamium gracile Dunn. were conspecific. Thus, they formerly transferred the former species to Paralamium and a new combination, Paralamium griffithii (Hook.f.) S. Suddee & A.J. Paton, was created, making the latter species (Paralamium gracile) as a synonym.
The systematic position of Paralamium has been enigmatic ever since its original description. When establishing the genus, Dunn (1913) noted that the calyx is the "most striking" character of Paralamium and similar to Orthosiphon Benth. (Nepetoideae), Coleus Lour. (Nepetoideae) and Teucrium L. (Ajugoideae) by virtue of the following calyx characters: a broad upper calyx tooth with recurved decurrent margins and a conspicuously veined calyx tube. However, in the protologue for Paralamium (Dunn, 1913), the genus was also considered to be closely related to Lamium L. (Lamioideae) based on nutlet and corolla characters, hence the name "Paralamium" which can be translated to mean "resembling Lamium." Studies on the genus after its original description have been scarce. Li (1977) placed Paralamium within subtribe Lamiinae of tribe Lamieae in subfamily Lamioideae sensu Briquet (1895Briquet ( -1897 based on its morphology provided in the protologue (Dunn, 1913). Later, Cantino and Sanders (1986) considered Paralamium as an anomalous genus within Lamiaceae because of its morphology similar to various genera in different subfamilies, but discreetly suggested that it could probably be related to Lamium based on their similar tricolpate and twocelled pollens observed by Abu-Asab and Cantino (1994). Harley et al. (2004) also placed Paralamium within Lamioideae in their comprehensive classification of Lamiaceae. In the most recent classifications of Lamioideae based on molecular data (Scheen et al., 2010;Bendiksby et al., 2011), Paralamium was provisionally treated as incertae sedis within Lamioideae but additionally suggested to be a member of tribe Pogostemoneae based on nutlet morphology (e.g., small glossy nutlets) (Bendiksby et al., 2011). While in the updated online synoptical classification of Lamiales, Olmstead (2016) placed Paralamium within tribe Stachydeae of Lamioideae. However, Paralamium has never been included in a published molecular phylogenetic analysis, making the above empirical placement of Paralamium within Lamioideae untested.
The main reason that Paralamium has not be included in any molecular phylogenetic studies is a lack of suitable leaf tissue for DNA extraction. However, during collecting expeditions in the Yunnan province of China in 2018 and 2019, we discovered two populations of P. griffithii. These collections allowed us to investigate the phylogenetic position of this monotypic and enigmatic genus based on molecular data. Here, using both plastid and nuclear ribosomal DNA markers, we present molecular phylogenetic analyses using different sampling strategies to finally establish the tribal affinities of Paralamium within Lamioideae and provide an updated phylogeny of the tribe Pogostemoneae. Furthermore, we provide a dichotomous key for genera within Pogostemoneae.

Field Collections
Specimens from two populations of Paralamium griffithii were collected from Malipo County (Liu et al. 7859) and Jinping County (Z.Y. Cai and X.E. Ye czy-36) within the Yunnan Province of China. Fresh leaves were collected and dried with silica gel. Voucher specimens were deposited in the Herbarium of Kunming Institute of Botany (KUN), Chinese Academy of Sciences.

Taxon Sampling and Genetic Markers Selected
In order to better evaluate the systematic position of Paralamium and assess the phylogenetic relationships of this enigmatic genus and related genera, we experimented with three datasets. The first dataset included 79 plastid protein-coding genes within Lamiaceae (dataset CP79) aiming to confirm the subfamilial position of Paralamium. In total, 84 accessions from 84 species and 63 genera of Lamiaceae were included for this initial analysis, covering 11 of the 12 subfamilies recognized by Li et al. (2016) and Li and Olmstead (2017 (Refulio-Rodriguez and Olmstead, 2014;Liu et al., 2020).
GenBank accession numbers and the source publications for taxa in this dataset are provided in Supplementary Table 1. We used the phylogenetic results from this first set of analyses as a basis for a more focused second round of analyses.
Because the first set of analyses demonstrated that Paralamium has affinities with tribe Pogostemoneae of Lamioideae, we expanded the sampling of Pogostemoneae in a second round of analyses. These analyses focused on further exploring the placement of Paralamium within Pogostemoneae and explicating relationships among genera of the tribe. Chen et al. (2014) demonstrated that the monotypic genus Holocheila (Kudô) S. Chow is a member of Pogostemoneae, so we also included this genus for analysis. In total, for the first time, all 12 genera (including Paralamium) of Pogostemoneae were included as part of our Pogostemoneae-wide analyses. This comprehensive generic sampling offers the opportunity to clarify generic relationships of Pogostemoneae using five plastid regions (matK, rbcL, rps16, trnH-psbA, trnL-trnF; dataset CP5) and the nuclear ribosomal internal transcribed spacer (dataset nrITS). In total, 56 sequences were newly sequenced for 13 species in 8 genera, while others were taken from previous studies (Chen et al., 2014;Yao et al., 2016) or downloaded from GenBank ( Table 1). Outgroups for the dataset CP5 and the dataset nrITS were sampled from tribe Gomphostemmateae (Chelonopsis souliei (Bonati) Merr., Gomphostemma lucidum Wall. ex Benth., and Gomphostemma sp.) according to Yao et al. (2016).

DNA Extraction, Amplification, and Sequencing
Total genomic DNA was extracted from fresh or silica-gel-dried leaf fragments using the CTAB procedure of Doyle and Doyle (1987), then dissolved in double-distilled water and kept at −20 • C for future polymerase chain reaction (PCR) amplification.
Primers and PCR thermal cycler settings for matK and rbcL followed Chen et al. (2014), and those for nrITS, trnL-trnF, rps16, and trnH-psbA were as described by Xiang et al. (2013b). Amplified PCR products were visualized on 1% TBE agarose gel, stained with ethidium bromide and then sequenced by an ABI-PRISM3730 sequencer after purification with a QIAquick PCR purification Kit (BioTek, Beijing, China). Voucher information for newly sequenced species and GenBank accession numbers for all sequences used in the current study are listed in the Table 1.

Plastome Sequencing, Assembly, Annotation, and Gene Region Extraction
The DNA concentration of Paralamium griffithii was at least 35 ng/µL as measured by a NanoDrop spectrophotometer 2000 (Thermo Scientific, Carlsbad, CA, United States). DNA integrity was detected and purified by 1% Agarose Gel Electrophoresis for 40 min at 150 V. Subsequently, the DNA samples were sheared into 300 bp fragments for paired-end library construction according to manufacturer's instructions (Illumina, San Diego, CA, United States), details are provided in Zhao et al. (2020a).
Prior to genome assembly, adapter sequences and low-quality reads were removed using the ea-utils package 1 . Quality control of raw sequence reads was carried out using FastQC 0.11.8 (Andrews, 2018) with the parameter set as Q ≥ 25. We used the GetOrganelle pipeline (Jin et al., 2020) for the de novo assembling. The software Bandage v. 0.8.1 (Wick et al., 2015) was employed for contig visualization and editing. Lastly, in order to validate the assembly error, the raw reads were mapped to the assembled plastid genome sequences by the Bowtie2 (Langmead and Salzberg, 2012) plugin in Geneious v. 11.0.3 (Kearse et al., 2012). In addition to the newly sequenced plastome of Paralamium griffithii and downloaded plastomes of 54 species from GenBank (Supplementary Table 1), 32 data from the Sequences Read Archive (SRA) were included for reassembling. 1 https://code.google.com/p/ea-utils/ The Initial annotations were implemented in the Plastid Genome Annotator (PGA) (Qu et al., 2019), and the published plastome of Phlomoides betonicoides (Diels) Kamelin & Makhm (MN617020; Zhao et al., 2020b) was set as a reference, then Geneious v.11.0.3 (Kearse et al., 2012), and tRNAscan-SE service (Lowe and Chan, 2016) were used adjusting of the putative starts, stops, intron positions, and tRNA boundaries as described by Zhao et al. (2021) and Xiang et al. (2020). Finally, the circular physical map of the plastome of Paralamium (Supplementary Figure 1) was drawn by the Organellar Genome DRAW tool (Lohse et al., 2013). The coding regions (CR) were extracted from the annotated complete plastome sequences for phylogenetic analyses.
Since topological incongruence between the combined cpDNA and nrITS data was reported in Yao et al. (2016), the nrITS and cpDNA datasets were not combined for analyses here. However, because plastome regions typically have a shared genetic history, the five plastid DNA regions were combined for phylogenetic analyses. All datasets were analyzed using Maximum Likelihood (ML) and Bayesian Inference (BI) algorithms on the CIPRES Science Gateway ( 2 Miller et al., 2010). The ML analyses were implemented with RAxML v.8.2.9 (Stamatakis, 2014), bootstrap probabilities were generated by conducting 1000 bootstrap iterations, and details for parameter settings are described by Xiang et al. (2020). Bayesian inference analyses were performed using MrBayes v.3.2.2 (Ronquist et al., 2012). The best-fit nucleotide substitution models were selected under the Akaike Information Criterion (AIC) using jModelTest v.3.7 (Posada, 2008). The models used were the GTR+I+G for dataset CP79, TVM+I+G for dataset CP5, and for the nrITS dataset. In addition, a partitioned strategy for the dataset CP5 also used for Bayesian inference analyses (GTR + G for matK, GTR + I for rbcL, TVM + G for rps16, TPM1uf + G for trnH-psbA, GTR + I for trnL-trnF). Specific steps for analyses are described in detail in Chen et al. (2016) and references provided therein. Finally, we used FigTree v.1.4.2 (Rambaut, 2014) to visualize and edit all resulting trees. We defined branches with posterior probabilities (PP) ≥ 0.95 and bootstrap values (BS) ≥ 80% as strongly supported, PP = 0.90-0.95 and BS = 70-80% as moderately supported, while PP < 0.90 and BS < 70% were defined as weakly supported.

Nutlets Morphology
Mature nutlets were collected from both wild-collected or herbarium plant specimens from the Germplasm Bank of Wild Species in Southwest China, Kunming Institute of Botany, for light microscope (LM) and scanning electron microscope

Genome Assembly, Features and Gene Content of Paralamium griffithii
The newly sequenced and annotated plastome was submitted to the National Center for Biotechnology Information (NCBI) database with the accession number MW201575. Illumina paired-end sequencing generated 20,321,882 clean reads, with coverage of 179 × for P. griffithii. The plastome size was 152,664 bp and displayed the typical quadripartite structure consisting of a pair of IR regions (25,617 bp) separated by the large single copy (LSC; 83,788 bp) and small single copy (SSC; 17,642 bp) regions (Supplementary Figure 1). In total, 114 unique genes (80 protein-coding genes, 30 tRNAs, and 4 rRNAs; Supplementary Table 2) were identified (duplicated genes in IR regions were counted only once). We used 79 common proteincoding genes for phylogenetic analyses based on Zhao et al. (2021) with the exclusion of the ycf 15 gene because it could not be extracted from most plastome reassembed from SRA database.

Sequence Characterization
Properties for different datasets are summarized in In the second set of analyses, the combined cpDNA dataset was 3,439 bp (832 bp for matK, 574 bp for rbcL, 880 bp for trnL-trnF, 861 bp for rps16, and 292 bp for trnH-psbA) after excluding ambiguously aligned characters. The nrITS matrix contained 656 aligned positions ( Table 2).

Phylogenetic Analysis
For each combined dataset (CP79, CP5, and nrITS), ML and BI analyses yielded identical topologies, respectively (Figures 2-4;   Supplementary Figures 2-8). Therefore, only the trees resulting from maximum likelihood analysis of each dataset are presented, with posterior probability values from BI analyses indicated.
This recognition guided the second set of analyses, which aimed to further clarify the position of Paralamium, reassess generic relationships within Pogostemoneae, and update the phylogeny of Pogostemoneae by including as comprehensive taxon sampling as possible using both nrITS and cpDNA data. In all analyses, Pogostemoneae is robustly supported as monophyletic (Figures 3, 4), but the topologies differed between the nrITS and cpDNA phylogenetic trees. In the nrITS phylogeny, Pogostemoneae was found to have two major clades (labelled A and B in Figure 3). Clade A, or the Pogostemon group, includes Pogostemon Desf., Anisomeles R. Br., and Microtoena Prain, in which the former two genera formed a clade (100%, 1.00) sister to Microtoena (98%, 1.00). Clade B is poorly supported (59%, -) and includes nine genera. Clade B in turn is comprised of two subclades: one containing Colebrookea Sm., Paralamium + Craniotome Rchb., weakly supported (57%, -); and another subclade composed of Holocheila and the "Achyrospermum group" (i.e., Achyrospermum Blume, Eurysolen Prain, Leucosceptrum Sm., Comanthosphace S. Moore, and Rostrinucula Kudô), also poorly supported (63%, -).
In Rostrinucula, the nutlets are narrowly ellipsoid with curved hook-like apices, brown, pubescent outside with glands and eglandular trichomes (R. sinensis (Hemsl.) C.Y. Wu, Figures 6Q-T). Nutlets of Comanthosphace are obovate, light brown, and the surface is rough and has subsessile and eglandular trichomes (C. ningpoensis (Hemsl.) Hand.-Mazz., Figures 6U-X). In Leucosceptrum canum Sm. the nutlets are oblong, brown, with sharp edges or ribs apically, and a surface more or less smooth but with sparse subsessile glands (Figures 7A-D). Nutlets of Eurysolen are also obovate, dark brown, dull, and densely glandular along the ventral side (Figures 7E-H). Only one species of Achyrospermum, A. wallichianum (Benth.) Benth. ex Hook. f., was included for this study. Achyrospermum wallichianum has somewhat elliptic light brown nutlets that are hairy at apex and reticulate on the surface (Figures 7I-L). Nutlets of Craniotome (Figures 7M-P) and Paralamium (Figures 7Q-T) are subspheric, brown and black respectively, and slightly reticulate outside. Nutlets of Colebrookea (Figures 7U-X) are obovoid to oblong, light brown, with apices and fruit navels densely covered with glands, and a surface that is smooth and sometimes with subsessile glands.

Paralamium as a Member of Pogostemoneae in Subfamily Lamioideae
The resulting topologies of Lamiaceae from the dataset CP79 are consistent with that of previous studies  based on five cpDNA regions and relationships among these subfamilies are well resolved. Moreover, all tribes of Lamioideae are strongly supported as monophyletic (Figure 2), which is in concordance with previous studies (Scheen et al., 2010;Bendiksby et al., 2011;Roy and Lindqvist, 2015;Zhao et al., 2021). Cantino and Sanders (1986) considered Paralamium as an anomalous genus within Lamiaceae because of its morphological similarities to genera from various subfamilies (i.e. Orthosiphon and Coleus of Nepetoideae, Ajuga of Ajugoideae, and Lamium of Lamioideae). However, the presence of tricolpate and twocelled pollens in Paralamium suggested its placement within Lamioideae (Cantino and Sanders, 1986). The genus was suggested to be closely related to Pogostemoneae by Bendiksby et al. (2011) based on nutlet morphology, but they explicitly treated it as incertae sedis within Lamioideae due to the lack of molecular phylogenetic data. Here, both the plastid and nuclear DNA data (Figures 2-4) support that Paralamium is a member of tribe Pogostemoneae, and is sister to the monotypic genus Craniotome.

Circumscription and Relationships Within Pogostemoneae
The monophyly of Pogostemoneae was supported by most studies (Scheen et al., 2010;Bendiksby et al., 2011;Chen et al., 2014) based on cpDNA sequences, but not by Roy and Lindqvist (2015) using PPR data, who revealed that genera of Pogostemoneae were included in two separate clades. The first FIGURE 2 | Phylogeny of Lamiaceae inferenced by maximum likelihood (ML) based on 79 coding regions (dataset CP79), with ambiguously aligned sites excluded from analysis. Bootstrap values ≥ 50% in ML and posterior probability values ≥ 0.90 in BI analyses displayed on the branch follow the order ML BS /BI PP ("-" indicates a support value BS < 50% or PP values < 0.9). Subfamilial classification of Lamiaceae is based on Li et al. (2016) and Li and Olmstead (2017). clade was referred as the Achyrospermum group (i.e., subclade A in Figure 2 sensu Roy and Lindqvist, 2015), forming the firstdiverging clade within Lamioideae. The second clade consist of Pogostemon, Anisomeles, and Craniotome (i.e., subclade B in Figure 2 sensu Roy and Lindqvist, 2015), forming the second diverging clade sister to remainder of Lamioideae.
In addition to the confirmation of the systematic position of Paralamium and sister relationship between Paralamium and Craniotome, some other well supported groups within Pogostemoneae are also recovered in this study, which enables us to further discuss the relationships within the tribe. Based on nrITS phylogeny (Figure 3), two subclades (i.e., clade A and clade B) can be recognized. Clade A is strongly supported and composed of three genera (Pogostemon, Anisomeles, and Microtoena), while clade B is composed of the remaining genera of Pogostemoneae. Although clade B is weakly supported (0.59, -), this split is supported by nutlet morphology. In the present study, nutlets of 17 species representing 11 out of 12 genera (except Holocheila) of Pogostemoneae were included for analyses. Based on our LM and SEM observations, we found that nutlets of genera in clade A (Figure 3; Pogostemon, Anisomeles, and Microtoena) are glossy and relatively glabrous (Figures 5, 6A-P), and the sclerenchyma region is very distinctive (Bendiksby et al., 2011), while genera in clade B (Rostrinucula, Comanthosphace, Leucosceptrum, Eurysolen, Achyrospermum, Paralamium, Craniotome) have dull and glandular nutlets (Figures 7Q-X), and the sclerenchyma region is often absent or indistinct (Ryding, 1994a(Ryding, , 1995. Within clade A, Anisomeles is sister to Pogostemon, with Microtoena sister to the Anisomeles-Pogostemon clade (Figures 3, 4). The three genera form a clade referred as clade A, which was supported by previous molecular phylogenetic studies (Scheen et al., 2010;Bendiksby et al., 2011). Cantino (1990Cantino ( , 1992a) suggested a close relationship between Anisomeles and Pogostemon based on their bearded staminal filaments and lustrous pericarps, as well as the presence of minute glands with unicellular caps on the leaf epidermis. Later, Abu-Asab and Cantino (1994) found that the two genera have very similar pollen grains with regular polygonal lumina and large perforations (see also Bean, 2015).
The close relationship between Microtoena and the Pogostemon-Anisomeles clade has been reported in previous studies (Scheen et al., 2010;Bendiksby et al., 2011;Chen et al., 2014;Roy and Lindqvist, 2015). The three genera are similar in terms of calyx morphology, with the calyx splitting the upper two and bottom three lobes up to ca. 1/2 of its length. Furthermore, linear bracts are present in Anisomeles and most species of Microtoena, while lanceolate or ovate bracts can be found in some species of Microtoena and Pogostemon (Wang, 2018). Geographically, most of the species of clade A are distributed in tropical East Asia (Scheen et al., 2010), although some species occur on islands within the Pacific and West Indian Oceans (Anisomeles), Africa (Pogostemon), and the Himalayas (Craniotome and Pogostemon glaber).
Another subclade (i.e., Achyrospermum group) composed of Achyrospermum, Eurysolen, Leucosceptrum, Rostrinucula, and Comanthosphace is also strongly supported in both the nrITS (Figure 3) and cpDNA trees (Figure 4), among which Rostrinucula and Comanthosphace are consistently resolved as sister genera (Figures 3, 4). The Achyrospermum group was first reported by Bendiksby et al. (2011) using cpDNA markers and subsequently recovered by Roy and Lindqvist (2015) based on the PPR region, but neither of them sampled Leucosceptrum. Species of the Achyrospermum group are distributed mainly in tropical East Asia and share several morphological characters. For example, the sclerenchyma region in the fruit pericarp is present in most lamioid members (Ryding, 1995;Bendiksby et al., 2011), but is obsolete, indistinct, or absent in the Achyrospermum group (Ryding, 1994b(Ryding, , 1995. Moreover, genera in this subclade have dull and glandular nutlets (Figures 6Q-X, 7A-L), while other genera within Pogostemoneae have glossy and glabrous nutlets (Bendiksby et al., 2011). Stamens long-exserted from the corolla are rare in Lamioideae, and are restricted to Comanthosphace, Rostrinucula, and Leucosceptrum in the Achyrospermum group, as well as a few species of Pogostemon in clade A. As suggested by Scheen et al. (2010), this character may be a synapomorphy for the small clade consisting of Comanthosphace, Rostrinucula, and Leucosceptrum. Molecular phylogenetic and morphological studies based on a broader sampling and more DNA sequences may further help to elucidate relationships within Pogostemoneae and identify morphological synapomorphies for the tribe.

Incongruence Between Nuclear and Plastid Phylogenies
In this study we provide the first comprehensive molecular phylogenetic study of Pogostemoneae. Though the intergeneric relationships within this tribe are generally well resolved, the placement of four monotypic genera (Colebrookea, Holocheila, Paralamium and Craniotome) is still uncertain due to incongruent topologies between nrITS and cpDNA trees. In the nrITS phylogeny (Figure 3), the Paralamium-Craniotome clade is sister to Colebrookea but weakly supported (57%, -). The Paralamium-Craniotome-Colebrookea clade is then sister to a clade including Holocheila and the Achyrospermum group, which is also weakly supported (-, 0.90) again. In the cpDNA tree (Figure 4), however, Holocheila is the first diverging clade, followed by Colebrookea, the Achyrospermum group, and then Paralamium-Craniotome + clade A, which is largely consistent with the topology of Chen et al. (2014). Most genera in clade B (excepting Achyrospermum, 25 spp.), all other genera are monotypic (Colebrookea, Craniotome, Eurysolen, Holocheila, Leucosceptrum, Paralamium, Rostrinucula) or oligotypic (Comanthosphace, 4 spp.) and mainly distributed in East Asia.
Incongruence between genomes have been noted within several genera in Lamiaceae, and ancient hybridization and chloroplast capture has often been posited to have contributed to the discordance (e.g., Albaladejo et al., 2005;Drew and Sytsma, 2013;Drew et al., 2014;Deng et al., 2015;Walker et al., 2015;Hu et al., 2018). Roy and Lindqvist (2015) suggested ancient reticulation events are likely to be responsible for the discordance between the plastid and PPR topologies of Pogostemoneae. They also demonstrated that ancestors of Pogostemoneae may have undergone rapid diversification during the middle Miocene in East Asia, which may have been triggered by climatic changes resulting from the uplift of the Qinghai-Tibetan Plateau (QTP) (Roy and Lindqvist, 2015). Considering that incomplete lineage sorting (ILS) among taxa is often associated with rapid radiations (Enard and Paabo, 2004;Pollard et al., 2006), ILS may also be a cause of the incongruences between the nuclear and plastid trees of Pogostemoneae. In the present study, two clades (clade A and clade B) are recognized based on nrITS phylogeny, but clade B is weakly supported by nrITS data and not recovered using cpDNA data. Although nutlet morphology supported the division of these two clades, futures studies involving next-generation sequencing and increased taxon sampling are need to provide insights into the complex evolutionary history of this group.

Key to All Genera of Pogostemoneae
The following circumscription of Pogostemoneae is based on this as well as previous studies (Scheen et al., 2010;Bendiksby et al., 2011). We provide a key to the 12 genera of Pogostemoneae below. Thus, studies using broad sampling of low-copy and/or single-copy intrageneric phylogenies and detailed comparative morphological investigation are needed.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.