Rediscovery of Mazus lanceifolius reveals a new genus and a new species in Mazaceae

Abstract Mazus lanceifolius (Mazaceae) is a perennial herb with opposite leaves and endemic to central China that has not been collected for 130 years. Rediscovery of this enigmatic species in the wild allows for determination of its phylogenetic position within Mazaceae. Phylogenetic reconstruction of Mazaceae based on DNA sequences from four plastid markers (matK, rbcL, rps16 and trnL-trnF) and nuclear ribosome ITS consistently showed that Mazus was not monophyletic. Mazus lanceifolius is in the most basal clade within Mazaceae, as sister to the remaining species of three recognized genera Dodartia, Lancea and Mazus. These results support the separation of M. lanceifolius from Mazus as a new genus, which was established here as Puchiumazus Bo Li, D.G. Zhang & C.L. Xiang. Meanwhile, a collection from Shennongjia Forestry District of Hubei Province, China, misidentified as “M. lanceifolius” in previous molecular study, is here revealed to represent an undescribed species of Mazus, i.e., M. fruticosus Bo Li, D.G. Zhang & C.L. Xiang, sp. nov. Morphologically, Puchiumazus is clearly distinct from the other three genera by having quadrangular to somewhat ribbed stems, and obviously opposite leaves. In addition, we provide a taxonomic key to the four genera of Mazaceae.


Introduction
Mazaceae (Reveal 2011) is a small herbaceous family in Lamiales currently containing three genera: Dodartia L., Lancea Hook.f. & Thomson and Mazus Lour. (APG IV 2016;Olmstead 2016;Christenhusz et al. 2017). The monotypic genus Dodartia based on D. orientalis L., occurs mainly in southern Russia and western to central Asia (Fischer 2004) and is characterized by having scale-like leaves and much-branched stems. The genus Lancea is found only in the Qinghai-Tibetan Plateau (QTP) where it includes two species, L. tibetica Hook.f. & Thomson and L. hirsuta Bonati (Chi et al. 2018, of which the former species is widely used in traditional Tibetan medicine. Morphologically, Lancea is characterized by leaves in a rosette and a lower corolla lip with a distinct palate. Mazus is the largest genus in Mazaceae, including approximately 30 species of annual or perennial herbs (Hong et al. 1998; distributed in Asia, Australia and New Zealand (Li 1954;Barker 1991;Fischer 2004). China is considered to be the center of distribution and differentiation of the genus (Yang 1979;Hsieh 2000), with ca. 26 species and three varieties currently recorded (Hong et al. 1998;. Species delimitation in Mazus has been problematic because of relatively high levels of morphological variation (Li 1954;Hong et al. 1998). In general, Mazus can be distinguished from the other two genera by a combination of morphological characters: a strongly two-lipped corolla (3/2-bilabiatae), a palate with two longitudinal plaits and a capsule enveloped in a persistent calyx (Fischer 2004;. Dodartia, Lancea and Mazus were once placed in the traditionally circumscribed Scrophulariaceae (e.g. Von Wettstein 1891) but variably affiliated with tribe Gratioleae (Von Wettstein 1891; Thieret 1954Thieret , 1967 or Mimuleae (Dumortier 1829;Burtt 1965;Argue 1984;Fischer 2004). However, Scrophulariaceae were found to be polyphyletic and some genera were subsequently transferred to existing families like Orobanchaceae, Plantaginaceae, Phrymaceae and Stilbaceae, and some genera were separated as small monophyletic families, including Calceolariaceae, Linderniaceae, Mazaceae, Paulowniaceae, Schlegeliaceae, and Wightiaceae (Olmstead and Reeves 1995;Oxelman et al. 1999Olmstead et al. 2001;Albach et al. 2005;Rahmanzadeh et al. 2005;Tank et al. 2006;Liu et al. 2020), then leaving a much reduced Scrophulariaceae s.s. To date, a number of genera have not yet been sequenced and are still unplaced.
When redefining Phrymaceae based on molecular phylogenetics,  had first shown that Mazus and Lancea formed a well-supported group that was weakly supported as sister to the rest of Phrymaceae. Consequently, they tentatively included the two genera in the redefined Phrymaceae and assigned them to a provisional subfamily "Mazoideae" . However, subsequent studies did not recover the sister relationship between "Mazoideae" and the rest of Phrymaceae, and Mazus and Lancea were found to be sister to the Or obanchaceae+Paulowniaceae+Phrymaceae clade Albach et al. 2009;). Thus, a new family Mazaceae Reveal (2011) was estab-lished to accommodate this. When Dodartia was first included in a molecular analysis, Xia et al. (2012) found that this genus was closely related to Lancea and they together formed the sister clade of Mazus. Currently, Mazaceae Reveal (2011) with the inclusion of all these three genera has been widely accepted (Refulio-Rodriguez and Olmstead 2014;APG IV 2016;Olmstead 2016;Christenhusz et al. 2017). It was found to be a member of the clade comprising Lamiaceae, Mazaceae, Wightiaceae, Phrymaceae, Paulowniaceae and Orobanchaceae .
Within the genus Mazus, M. lanceifolius Hemsl. is a distinctive species through its quadrangular stems and narrowly lanceolate, mostly cauline, opposite leaves (Fig. 1). By contrast, the other species of Mazus have terete stems and leaves often in basal rosettes (Yang 1979;Hong et al. 1998). Therefore, M. lanceifolius was assigned to a monotypic section: sect. Lanceifoliae Bonati (1908), which was followed by Yang (1979). Since its description by Forbes and Hemsley (1890), M. lanceifolius has never been recorded by any specimens until two populations of the rare species were rediscovered in Sichuan Province of China in 2020. The rediscovery of M. lanceifolius after more than one century offers us a precious opportunity to test its phylogenetic position based on morphological and molecular data.
Since the establishment of the family Mazaceae (Reveal 2011), only one molecular phylogenetic study exclusively focused on its phylogeny , including one species from each Lancea and Dodartia, and 23 out of 30 species of Mazus. In that study,  notably included two samples named as "Mazus lanceifolius", and stated that "M. lanceifolius" can be easily distinguished from other Mazus species by having lanceolate leaves and a robust stem. After consulting the vouchers of "Mazus lanceifolius" (D.G. Zhang zdg6673, Fig. 2) sampled by  as well as the type specimens (Henry 7250, K001079356!; Henry 5837, K001079356!) and the original description of M. lanceifolius, we found that the plants of "Mazus lanceifolius" used by  have opposite to subopposite leaves, which may have led the authors to identify the plant as M. lanceifolius because this species is the only known Mazus species with opposite leaves. However, except for these opposite leaves, their "Mazus lanceifolius" is remarkably different from the type specimen of M. lanceifolius in many aspects. For example, the plants sampled by  are robust shrubs having numerous and much branched stems, leathery leaves that are acutely serrate on the apical half and multiflowered inflorescences ( Fig. 2; see also fig. 2C in , while the type material of M. lanceifolius is a slender herb having several unbranched stems, submembranaceous and almost entire leaves and remarkably sparse inflorescences with no more than six flowers (Fig. 1). We therefore have to conclude that the specimen sampled as "M. lanceifolius" by  was misidentified, with the identity of that sample needing to be confirmed.
In the present study, we carried out an updated phylogeny of Mazaceae, in order to (1) investigate the phylogenetic placement of the distinct and enigmatic species M. lanceifolius based on its rediscovered populations; (2) confirm the identity of the misidentified M. lanceifolius by ; and (3) further contribute to a comprehensive phylogenetic framework for Mazaceae.

Material and methods
Field work, taxon sampling and data collection Two populations of Mazus lanceifolius were rediscovered in June 2020 in Sichuan Province, China. One is located in the Qingchengshan Mountain near Dujiangyan City, and another was found in Qianfoshan Mountain near Mianyang City. Morphological observations were conducted based on wild individuals as well as the type specimens. Fresh leaves were collected in the field and dried with silica-gel for DNA extraction (Chase and Hills 1991). Voucher specimens are deposited in the herbarium of Shanghai Chenshan Botanical Garden (CSH).
In the present study, most DNA sequences are based on previous phylogenetic analyses ), but some problematic sequences were excluded for analyses. For example, the trnL-trnF sequences of Mazus japonicus (Thunb.) Kuntze 3 (KX807207) in the study of (Deng et  [Lamiaceae]) were selected as outgroups for the cpDNA dataset. While, because of the high divergence of nrITS sequences, only eight species from the above-mentioned families were selected as outgroups.

DNA extraction, amplification and sequencing
Total genomic DNA was obtained from silica-dried leaves using the CTAB procedure of Doyle and Doyle (1987). After extraction, the DNA was re-suspended in doubledistilled water and kept at -40 °C for polymerase chain reaction (PCR) amplifications.
The DNA amplifications were performed in a thermocycler (Eppendorf Scientific, Inc., Westbury, NY, USA). Based on , four cpDNA regions (matK, rbcL, rps16 and trnL-trnF) and nrITS were selected for phylogenetic reconstruction. Primers, protocols for PCR, sequencing followed those in  and references therein.

Phylogenetic analysis
Sequences were initially assembled and edited with Geneious v.7.1.7 (Kearse et al. 2012) and aligned using MUSCLE (Edgar 2004) as implemented in Geneious v.7.1.7 (Kearse et al. 2012). The final alignments were manually adjusted in PhyDe v.0.9971 . The four chloroplast DNA regions were combined directly because the plastid genome is mostly uniparentally inherited (Soltis and Soltis 1998) and supposedly safe to be combined in phylogenetic analyses (Olmstead and Sweere 1994). Nuclear (ITS) and the combined plastid data set were analyzed separately using maximum likelihood (ML) and Bayesian inference (BI) methods.
BI analyses using Markov chain Monte Carlo (MCMC) methods (Yang and Rannala 1997) were performed with MrBayes v3.2.2 (Ronquist et al. 2012) and implemented on the CIPRES Science Gateway (http://www.phylo.org/) (Miller et al. 2010). The optimal substitution models were selected using Model Finder (Kalyaanamoorthy et al. 2017) plugin in PhyloSuite . Model parameters were estimated directly during the runs. For each Bayesian analysis, four MCMC chains were run simultaneously for 20 million generations. Each run began with one random tree and sampled one tree every 1000 generations. At the end of the run, chain convergence and estimated sample size (ESS) parameters were assessed with Tracer v.1.6.0 (Rambaut et al. 2014). A 50% majority-rule consensus tree was calculated for each dataset after discarding the first 25% of the trees as burn-in. In the resulting summary tree, posterior probability values (PP) ≥ 0.95 were considered to be strongly supported (Suzuki et al. 2002).

Sequence and alignment characterization
Ten sequences were newly generated for this study (Appendix 1). The resulting combined and aligned cpDNA dataset contained 4514 positions (including gaps), of which 1287 positions belong to matK, 1266 to rbcL, 963 to the rps16 partition and 998 to the trnL-trnF spacer. Of these 1259 (27.89%) nucleotides were variable in the dataset (Table 1). The aligned nrITS dataset includes 641 nucleotides, of which 300 (46.80%) were variable (Table 1).

Phylogenetic analysis of Mazaceae
In all analyses, the monophyly of Mazaceae was strongly supported (Figs 3, 4; ML BS: 100%, BI PP: 1.00; all values reported in this order below). Because the taxon sampling is different in the datasets of cpDNA and nrITS, we did not combine them for analyses.
Three subclades can be identified in the cpDNA (Fig. 3) as well as nrITS trees (Fig. 4). The two individuals of M. lanceifolius consistently form a clade sister to the rest of Mazaceae. Within the rest of the family, Dodartia-Lancea clade is sister to Mazus (Figs 3, 4). In both ML and BI analyses, a sister relationship between Lancea and Dodartia is well supported (87%, 1.00 in cpDNA tree; 92%, 1.00 in nrITS tree). Monophyly of Mazus is also strongly supported (97%, 1.00) based on cpDNA dataset while moderately supported in nrITS analyses (62%, 0.93). Relationships within the genus Mazus are not fully resolved (Figs 3, 4). The "M. lanceifolius" misidentified in  was found to be grouped with M. sunhangii based on cpDNA analyses with low support values (Fig. 3), while emerging as an isolated lineage in nrITS analyses when ITS sequence of M. sunhangii was not available (Fig. 4).  Diagnosis. The new genus is characterized by having quadrangular to somewhat ribbed stems and opposite, narrowly lanceolate leaves (Figs 1, 5A1-A3). Puchiumazus is sister to a clade composed of Dodartia, Lancea and Mazus. Morphologically, it is most similar to Mazus, but it differs in having quadrangular stems, lanceolate leaves (vs. terete stems and usually obovate-oblong leaves).
Description. Perennial herbs. Rhizomes fleshy, white, horizontal. Root thin, fibrous. Stems erect, unbranched, glabrous, up to 30 cm tall, old stems quadrangular, glabrous, young stems inconspicuously quadrangular to obtusely ribbed, minutely puberulent. Leaves opposite, petiole inconspicuous to nearly absent; leaf blade narrowly  , who was a prominent Chinese taxonomist specializing in the taxonomy of Scrophulariaceae in the traditional sense) and "mazus", indicating that the new genus was separated from Mazus and is morphologically similar to it.
Common name (assigned here). Bu Qiu Cao Shu (补求草属; Chinese name). Distribution. According to our data, this genus is endemic to Central China. It is known only from Hubei (Jianshi), Sichuan (Dayi and Dujiangyan) and Chongqing (Wushan) and can be found under evergreen broad-leaf forest at elevations of 600-1250 m.  century. After checking all floras and literature dealing with Mazus in China, we are certain that M. lanceifolius has not been lectotypified before. Thus, we here propose the specimen A. Henry 5837 (Kew barcode: K001079356) as lectotype of M. lanceifolius (Fig. 1A) in accordance with article 9.3 of the International Code of Nomenclature for Algae, Fungi, and Plants (Shenzhen Code) (Turland et al. 2018). Diagnosis. Mazus fruticosus differs from all other conspecific taxa by being a shrub with numerous and much branched stems and having opposite to subopposite leathery leaves that are acutely serrate on apical half.
Etymology. The epithet of the new species refers to its shrubby habit. Common name (assigned here). Guan Zhuang Tong Quan Cao (灌状通泉草; Chinese name).
Distribution and habitat. Mazus fruticosus is currently known only from Shenlongjia Forest District in Hubei Province, central China. It frequently occurs on rocky cliffs or near evergreen mixed forests at an elevation of 1100-1250 m.

Discussion
We here reconstruct the phylogeny of Mazaceae based on a combined cpDNA dataset of four markers (matK, rbcL, rps16 and trnL-trnF), and nrDNA ITS dataset, which have been used previously to infer relationships within Mazaceae Yamamoto 2020) and among Lamiales (Refulio-Rodriguez and Olmstead 2014;Liu et al. 2020). The monophyly of Mazaceae is recovered as reported in previous work ) relying on the same molecular markers. The major difference is that the third clade identified in the present study was not sampled by . Based on our analyses (Figs 3, 4), Mazaceae is composed of four genera (Fig. 5), including the new genus Puchiumazus described here. Three major clades can be identified for a re-circumscribed Mazaceae, and the cladogram is accompanied by some general morphological characters and geographical distribution patterns. The first clade is composed of two individuals of the new monotypic genus Puchiumazus (Figs 1, 5A1-A3), which is currently only known from three provinces in central China. Morphologically, the new genus can be distinguished clearly from other genera by having quadrangular to somewhat ribbed stems and opposite, narrowly lanceolate leaves.
The second clade consists of Dodartia ( Fig. 5B1-B3) and Lancea ( Fig. 5C1-C3). Both genera have broader distribution area than Puchiumazus, with Lancea always found at high elevations in QTP and Dodartia distributed in southern Russia and western to central Asia; it is cultivated as medical herb which has increased its distribution. Morphologically, both genera have small scale-like leaves (with a basal rosette of larger leaves in Lancea). Another important character is that ca. half of the capsule is enclosed by fruiting calyx and that calyx-teeth are much shorter than the fruit (Fig. 5B3, C3). In Puchiumazus, the style is persistent and ca. 2/3 of the fruit is enclosed in the fruiting calyx with calyx-teeth being much longer than the fruit. Calyx of Mazus is usually at least 1-2 times longer than capsule (e.g., Fig. 5D3).
Species of Mazus comprise the third clade, which is well supported in the cpDNA tree (94%, 1.00; Fig. 3), but moderately supported in the nrITS phylogeny (62%, 0.93; Fig. 4). Mazus is the largest genus of Mazaceae and it is widely distributed in East Asia and Australia. It can be distinguished from the other three genera by the more or less secund inflorescences and a corolla with a palate on the lower lip. Using the same DNA markers,  produced a fully resolved phylogeny of Mazus in which five clades of the genus were highly supported (see Fig. 4 of their study). The interesting finding is that we cannot recover a similar topology, although the data of most species come from their dataset. Part of the reason for this may be that some sequences generated for their study were wrongly submitted to GenBank (see samples in Material and methods). Another possible reason is that they did not consider the topology incongruence between cpDNA and nrITS sequences, but concatenated the data for their analyses.
Phylogenetic analyses in our study did not support the sectional classification (i.e. Lanceifoliae, Mazus and Trichogymus) of Mazus proposed by Hong et al. (1998). At that time, Mazus lanceifolius was placed within Mazus, which we here recognize as a new genus. In addition, monophyly of the remaining two sections was also not supported, which was also the case in the study of . Accordingly, they proposed a new infrageneric classification of Mazus, with two subgenera, Mazus and Notomazus T. Deng, N. Lin & H. Sun. Subgenus Mazus comprises most of the spe-cies and is native to Asia, while subgenus Notomazus comprises all species native to Australia and New Zealand. However, the monophyly of the two subgenera were not supported in our study. In both cpDNA and nrITS trees, Mazus radicans (Hook.f.) Cheeseman from subgenus Notomazus is deeply nested in subgenus Mazus, indicating it is necessary to redefine subgenus Notomazus. Given the discordance between the trees presented here and the one presented in , on the basis of the same sequence data, we think some additional checking of the data, perhaps even resampling of M. radicans, is needed before any revision is made to the subgeneric classification of Mazus. In addition, a future study including more individuals of each species and more DNA markers (especially single and/or low copy nuclear genes) is necessary to clarify internal relationships within Mazus.
Previously, all species of Mazus are described as herbs (Yang 1979;Hong et al. 1998;Fischer 2004) Actually, M. caducifer, M. spicatus, M. celsioides have rigid stems that look woody, but are not actually forming wood, thus these should be recognized as having a herbaceous habit. The new species described in the present study is probably the only species with a shrubby habit in the genus Mazus. This interesting find will help us to better understand the character evolution of Mazus. If Mazus sp. in Deng et al.'s (2019) also has a shrubby habit, we can speculate this character originated independently at least twice within the genus.
The abovementioned findings mean that more intensive field collections are necessary even in the post-Flora time. Yang (1979) have noticed the morphological difference between Puchiumazus lanceifolius (≡ Mazus lanceifolius) and other Mazus species. He pointed out that the quadrangular stem is only found in this species, and the nearly entire lanceolate leaves are also rare in Mazus, thus he suggested that this species probably is generically distinct. At the same time, he also emphasized that, because no fully developed flowers could be investigated based on specimens, he placed this species within Mazus. In this study, the rediscovery of this species offers an opportunity to investigate morphological characters of P. lanceifolius and provide a chance to extract DNA for molecular phylogenetic analyses, which led to the establishment of the new genus in the present study.
In recent years, many plants of Lamiales were rediscovered from biodiversity hotspots of China, including Aeschynanthus monetaria Dunn (Gesneriaceae; Hu et al. 2020 Li et al. 2012) and Pedicularis humilis Bonati (Orobanchaceae;Li et al. 2016). Most of these species had only been collected once before. The new genus described in the present study was also only known from the type collections (A. Henry 5837, 7250) before it was rediscovered. The type specimens of this were, until recently, the only known collections, and as a result, studies on the species since the original 1890 publication have been wanting. The re-investigation of this species is not only providing a chance to amend its description, but also a chance for a recognition of a new genus and redefinition of the family. The study highlights the important roles of field collections for systematic and biodiversity studies, which are often neglected in this age of biodiversity informatics (Wen et al. 2015).