﻿Pulvinatusia (Brassicaceae), a new cushion genus from China and its systematic position

﻿Abstract The new genus and species Pulvinatusiaxuegulaensis (Brassicaceae) are described and illustrated. The species is a cushion plant collected from Xuegu La, Xizang, China. Its vegetative parts are most similar to those of Arenariabryophylla (Caryophyllaceae) co-occurring in the same region, while its leaves and fruits closely resemble those of Xerodrabapatagonica (Brassicaceae) from Patagonian Argentina and Chile. Family-level phylogenetic analyses based on both nuclear ITS and plastome revealed that it is a member of the tribe Crucihimalayeae, but the infra-/intergeneric relationships within the tribe are yet to be resolved.

growing on high-altitude mountains and are thought to be associated with dry and cold environments, such as the high Andes and Patagonia, Himalayas, and New Zealand Alps (Aubert et al. 2014;Boucher et al. 2016). Hauri and Schröter (1914) compiled the first worldwide list of cushion plants which included 338 species of 34 families and 78 genera. A century later, Aubert et al. (2014) updated the cushion plants catalogue in which they recognized 1,309 species of 63 families and 273 genera. An online database was also created for easy access and timely update (http://www.cushionplants.eu/).
The mustard family (Brassicaceae) is distributed primarily in temperate areas, and many of its species grow on high mountains. Aubert et al. (2014)  Sun. Although many other Brassicaceae species were described as cushion plants and found to be occurring in China (Zhou et al. 2001;Al-Shehbaz 2015), they were not included in Aubert et al.'s catalogue (2014).
From 2000 to 2019, 58 new species of Brassicaceae from China were described (Du et al. 2020), the number of Chinese Brassicaceae species has grown to ca. 500 (Chen et al. 2019). During an expedition in August 2015 to Mt. Xuegu La, Damxung County, Xizang (Fig. 1), China, we collected a mustard plant with typical cushion characters and whitish pink flowers ( Fig. 2G-H). We went back to the above-mentioned locality in August 2019 and collected fruiting material of this plant ( Fig. 2A-F). Morphological studies family-wide revealed that it represents a new genus and species, hereafter recognized as Pulvinatusia xuegulaensis. We also carried out molecular studies to verify its systematic position within the family.

Taxon sampling and data collection
To assess the identity and systematic position of the new taxon, a family-level sampling strategy was adopted. Two datasets, the nuclear ITS and plastomes, were utilized to reconstruct the phylogeny of Brassicaceae. The ITS dataset included 125 species representing 98 genera, of which two accessions of the novelty were newly sequenced. The plastome dataset included 74 species representing 70 genera, of which 16 accessions representing 16 species were newly sequenced. The plastome of Bivonaea lutea (Biv.) DC. was extracted from raw sequencing data SRR8528386 deposited under NCBI BioProject PRJNA518905. Cleome lutea Hook. was chosen as outgroup for ITS and plastome datasets. Both ITS and plastome datasets comprised all 52 currently recognized tribes and nine genera which were not assigned to tribes within Brassicaceae.  Data downloaded from GenBank and newly generated for this study are listed in Appendices 1 and 2, respectively.

DNA extraction, amplification, and sequencing
Total genomic DNA was extracted from silica gel-dried fresh leaves using the Plant Genomic DNA Kit (Tiangen Biotech, Beijing, China) following the manufacturer's protocol. The ITS region of one sample of Pulvinatusia xuegulaensis (voucher specimens ZBFC-510) was amplified with the primers ITS-18F as modified by Mummenhoff et al. (1997) and ITS4 (White et al. 1990). A 25-ml polymerase chain reaction (PCR) included 1-2μL sample DNA (approx. 1-10 ng), 12.5μL Premix Taq TM (Takara Biomedical Technology, Beijing, China), 1μL of 10 μM stock of each primer, adjusted to 25 μL with ddH 2 O. The PCR program included a hot start with 4 min at 94 °C, and 30-32 cycles of amplification (1 min denaturing at 94 °C, 45-60 s annealing at 52-53 °C, 60-80 s extension at 72 °C), and a final elongation step for 10 min at 72 °C. The sequencing primers are the same as amplified primers. While the ITS region and plastome sequences of another sample of P. xuegulaensis (voucher specimens ZJW3454), together with the plastome data of 15 species listed in Appendix 2 were generated by genome skimming. Libraries for pair-end 150-bp sequencing was conducted using the Illumina HiSeq 2000 platform at Novogene Co. (Beijing, China).

Data assembly and annotation
For the genome skimming data, low-quality reads were filtered, and the clean data were assembled using the GetOrganelle pipeline (Jin et al. 2020). The nuclear ITS and plastomes were also annotated using Geneious 8.2.4 (Kearse et al. 2012) with the published ITS of C. himalaica (AY662283) and plastome of Rudolf-kamelinia korolkowii (Regel & Schmalh.) Al-Shehbaz & D.A. German (KX886350) as the reference, respectively. Positions of start and stop codons of plastome sequences were checked manually.
Maximum parsimony (MP) and Bayesian Inference (BI) analyses were performed for the ITS dataset, while for the 75 CDS dataset, Maximum Likelihood (ML) method was utilized. No substitutional saturation was detected in 75 CDS dataset, as the index of substitution saturation (Iss) values were both significantly smaller than the critical Iss (Iss.c) values as defined by Xia et al. (2003). MP analysis was performed with heuristic searches of 1000 replicates with random stepwise addition using tree bisection reconnection (TBR) branch swapping as implemented in PAUP* 4.0a168 (Swofford 2020). All characters were weighted equally, and gaps were treated as missing data. BI and ML analyses were carried out with MrBayes v.3.2.6 (Ronquist et al. 2012) and RAxML 8.2.12 (Stamatakis 2014) implemented in the CIPRES Science Gateway v.3.3 (Miller et al. 2010), respectively. The best-fit model for nucleotide sequences was evaluated using jModeltest 2.1.6 (Darriba et al. 2012). Corrected Akaike Information Criterion (AICc) method was used to select the best-fit models. The SYM+I+G model were selected for ITS dataset in the BI analyses. Two independent runs each with four Monte Carlo Markov chains (MCMCs) were run for five million generations, and one tree sampled every 1000 generations. The first 1250 trees (25% of total trees) were discarded as burn-in. The remaining trees were summarized in a 50% majority-rule consensus tree, and the posterior probabilities (PP) were calculated. The ML analyses were conducted using the GTR+G model for 75 CDS dataset, with the option of rapid bootstrap of 1000 replicates.

Morphological evaluation
With a single pivotal root, very short internode and compact branches, Pulvinatusia xuegulaensis forms a hemispherical (dome) shape (

Nuclear ITS and plastome assemblies
The ITS sequences for two accessions of the novelty were 628 bp long. Most of the 16 newly sequenced plastomes were assembled into complete circular genome, except one or two gaps remained in the noncoding regions of three accessions. Gaps information, voucher records, and GenBank accession numbers are provided in Appendix 2.

Phylogenetic analysis
The aligned ITS matrix was 496 bp long with 261 (52.6%) parsimony-informative sites. The aligned plastome CDS matrix was 61,713 bp long with 7,730 (12.5%) parsimony-informative sites. The resolution of MP analyses was relatively weaker than the outcome of BI analyses, thus only the topologies of Bayesian phylogenetic analysis were shown for ITS dataset. As our aim was to assess the systematic position of Pulvinatusia xuegulaensis, only clades containing this taxon were concerned. In the ITS phylogeny, two accessions of P. xuegulaensis clustered together and embedded in a clade consisting of Crucihimalaya species. This P. xuegulaensis/Crucihimalaya clade is sister to Ladakiella klimesii (Fig. 3). In the plastome phylogeny (Fig. 4), only three Crucihimalaya species and one accession for each of P. xuegulaensis and L. klimesii were sampled. The sequence of P. xuegulaensis formed a clade with L. klimesii, and then sistered to a clade composed of three Crucihimalaya species. Therefore, both nuclear and chloroplast phylogenies indicated that P. xuegulaensis should be assigned to the tribe Crucihimalayeae. Diagnosis. As indicated above, the monospecific Pulvinatusia xuegulaensis and Ladakiella klimesii are the only members of the tribe Crucihimalayeae with pulvinate and scapose habit and pink to whitish pink petals. The former differs by having simple and fewer forked trichomes, thin papery leaves, solitary flowers, caducous sepals, and glabrous, somewhat flattened fruits. By contrast, L. klimesii has subdendritic trichomes with finely branched rays, thick and fleshy leaves, 2-4-flowered racemes, persistent sepals, and pubescent and terete fruits.
Description. Herbs perennial, cespitose, scapose, pulvinate, with well-developed caudex covered with petioles of previous years. Trichomes simple, mixed with fewer forked stalked ones. Leaves densely imbricate, sessile, thin, papery, densely long ciliate, midvein obscure, adaxially concave to nearly flat, base attenuate, apex subacute. Flowers solitary on short pedicels originating from axils of basal leaves. Fruiting pedicels stout, erect or ascending, often hidden among basal leaves. Sepals oblong, abaxially with trichomes similar to those on leaves. Petals whitish pink to pink; blade obovate to suborbicular, apex obtuse, rounded or rarely acute, claw subequaling or slightly shorter than sepals. Stamens 6, slightly tetradynamous; filaments unappendaged, free; anthers ovate or oblong, obtuse at apex. Ovules 2 or 3 per ovary, placentation parietal. Fruits dehiscent, latiseptate, ovoid to ellipsoid, inflated; valves thick leathery, carinate; replum rounded, visible; septum complete; style obsolete or short and to 0.4 mm long, stout; stigma capitate, entire, unappendaged. Seeds aseriate, wingless, oblong, seed coat smooth, not mucilaginous when wetted; cotyledons accumbent. Name derivation. The generic name is derived from the pulvinate habit of the plant, and the species epithet from the Xuegu La (Xizang, China), where the type collection was made.

Discussion
Pulvinatusia xuegulaensis displays typical cushion-plants morphology, which belongs to the dome type of Aubert et al.'s category (2014). Many ball-shaped individuals grow together along alpine slopes and form a community with spectacular landscape (Fig. 1B-C). Without flowers and fruits, one can easily misidentify P. xuegulaensis as Arenaria bryophylla Fernald, a member of Caryophyllaceae family and one of the most typical cushion plants in the Sino-Himalayas. This might partially explain why this new taxon remained unrecognized until now; even the type locality is nearby a county road (Fig. 1A). Only with its conspicuous cruciform pink flowers and ovoid silicles, one can easily recognize it as Brassicaceae. To date, only one population of P. xuegulaensis has been found, within the family and the six cushion taxa (as mentioned in the Introduction) listed by Aubert et al. (2014) occurring in China, P. xuegulaensis is most similar to Ladakiella klimesii in gross morphology. Whereas it differs from the latter by more (vs. less) compact branches; imbricate (vs. rosulate) leaves; solitary flowers (vs. 2-4-flowered raceme) and stout (vs. slender) fruiting pedicel. By contrast, these distinct characters of P. xuegulaensis are also shown in Xerodraba patagonica (Speg.) Skottsb. (Eudemeae, Brassicaceae) (Table 1), a South American species endemic to southern Argentina and Chile at an altitude of 20 -1050 m (Salariato et al. 2015a), demonstrating morphological homoplasy between unrelated taxa of different continents.
In both nuclear and chloroplast phylogenies, Pulvinatusia xuegulaensis fell in a clade consisting of Ladakiella and Crucihimalaya species, indicating that the new taxon is phylogenetically close to these two genera, which had been assigned to the tribe Crucihimalayeae by German and Al-Shehbaz (2010). This study therefore supported Pulvinatusia to be the third genus within Crucihimalayeae. However, the intergeneric relationship within this tribe was not resolved. In the nuclear rDNA (ITS) phylogeny, two accessions of P. xuegulaensis were embedded in a clade consisting of nine Crucihimalaya species and then sister to L. klimesii (Fig. 3). This indicates that the genus Crucihimalaya as currently delimited (German 2005; Al-Shehbaz et al. 2011) is not monophyletic. In fact, generic delimitation and systematic position of Crucihimalaya have been in dispute for a long time. This genus was first established by Al-Shehbaz et al. (1999) to accommodate nine species excluded from Arabidopsis based on morphological and molecular evidences (Price et al. 1994;O'Kane et al. 1995). This delimitation was followed by Zhou et al. (2001) and Appel and Al-Shehbaz (2003), and the genus had been assigned to the tribe Camelineae by Al-Shehbaz et al. (2006) in their first scheme of tribal classification. However, subsequent molecular studies revealed that Crucihimalaya is phylogenetically distant to taxa from Camelineae but  Koch et al. 2007;Warwick et al. 2008;German et al. 2009). These species then had been transferred to Crucihimalaya and resulted in a heterogeneous genus including 13 species (German and Ebel 2005;German 2005;Al-Shehbaz et al. 2011), whereas a new genus Ladakiella was created to accommodate L. klimesii excluded from Alyssum (German and Al-Shehbaz 2010). Both Ladakiella and Crucihimalaya s.l. were assigned to the newly proposed tribe Crucihimalayeae (German and Al-Shehbaz 2010). The ITS phylogeny constructed in this study suggested either to combine P. xuegulaensis with Crucihimalaya s.l. or split the latter genus into several segregates. Pulvinatusia xuegulaensis is very similar to L. klimesii in gross morphology as they both share pulvinate habit and inflated ovoid silicles. These morphological similarities corresponded to their phylogenetic relationships revealed in the plastome phylogeny, within which these two species formed a clade sister to three Crucihimalaya species (Fig. 4). The discrepancy between nuclear and chloroplast phylogenies revealed in this study might be attributed to two main reasons: 1) sampling difference, i.e., there are nine species from Crucihimalaya s.l. sampled in the ITS phylogeny, but only three species sampled in the plastome phylogeny, especially lack of C. bursifolia and C. rupicola. 2) reticulate evolution caused by hybridization and/or introgression, of which evolutionary processes have been proposed for numerous taxa in the mustard family (Mummenhoff et al. 2004;Lihová et al. 2006;Dierschke et al. 2009;German and Friesen 2014;Mandáková et al. 2017;Hohmann and Koch 2017;Chen et al. 2020). To clarify inter-and infrageneric relationships within Crucihimalayeae, studies with comprehensive sampling and more molecular markers are needed. The discovery of Pulvinatusia xuegulaensis added one new genus and species to the cushion plant list compiled by Aubert et al. (2014). The cushion habit had long been considered a good example of evolutionary convergence among various plants in alpine and arctic regions (Aubert et al. 2014). It had been suggested to evolve independently four times in South American Brassicaceae (Salariato et al. 2015b) and happened at least 115 times in whole angiosperms (Boucher et al. 2016). Characterized by dense branches and compact structure, cushion plants usually form hemispheric or mat shapes, which enables them to adapt to cold and/or dry harsh environments and also facilitate other alpine plant species by nurse trait effects (Körner 2003;Yang et al. 2010;Chen et al. 2015;Chen et al. 2017;Yang et al. 2017). However, nothing is known about the underlying genetic basis of adaptation to alpine environments of cushion plants. All the three genera of Crucihimalayeae coexist in Qinghai-Tibet Plateau, and all species of Crucihimalaya are not pulvinate, while both L. klimesii and P. xuegulaensis are cushion species, thus provide an excellent system to decode the genetic basis of the formation of cushion structure and study the adaptive evolution of cushion plants, and the available genome of C. himalaica (Zhang et al. 2019) can facilitate this process.