﻿ Hydrangeamarunoi (Hydrangeaceae), a new species from Osumi Peninsula, southern Japan

﻿Abstract Hydrangeamarunoi Tagane & S. Fujii, from the Kimotsuki Mountains in the Ohsumi Peninsula, southern Japan, is described and illustrated. It is morphologically similar to H.alternifolia in having three-petaloid calyx lobes in marginal flowers, but is distinguished by the larger stamen number, and longer styles and seeds. Multiplex ISSR genotyping by sequencing (MIG-seq) demonstrated that the new species is monophyletic and closely related to H.amamiohsimensis and H.moellendorffii rather than H.alternifolia.


Introduction
Hydrangea L., including approximately 270 natural species (De Smet et al. 2015) and four widely cultivated species (Fulcher et al. 2016), is a well-known genus in Hydrangeaceae. Based on phylogenetic analysis, De Smet et al. (2015) proposed a broad circumscription of Hydrangea comprising approximately 200 species distributed across East and Southeast Asia and the Americas. Most Hydrangea species are shrubs or lianas. However, the species of Hydrangea L. sect. Cardiandra (Siebold & Zucc.) Y.De Smet & Samain are herbs that have been treated as members of the genus Cardiandra (Ohba 1985a(Ohba , b, 2001Wei and Bartholomew 2001;De Smet et al. 2015). In the current broad circumscription of Hydrangea, it is treated as a section of the genus Hydrangea, which additionally also includes eight groups previously known as genera i.e. Broussaisia Gaudich, Decumaria L., Deinanthe Maxim., Hydrangea s.str., and Pileostegia Hook. f. & Thomson, Platycrater Siebold & Zucc., and Schizophragma Siebold & Zucc. In the phylogeny by De Smet et al. (2015), Hydrangea sect. Cardiandra is monophyletic and a sister to the sect. Deinanthe, which comprises two known herbaceous species from China to Japan.
During our floristic survey in Kagoshima Prefecture, southern Japan in 2021, we collected an unknown flowering species of Hydrangea sect Cardiandra. It is similar to H. alternifolia in appearance, but it differs from it in its habitat and some floral characters. To clarify the relationship between the unknown species and the other species of Hydrangea sect. Cardiandra in Japan, we examined the phylogenetic relationships of 52 samples of H. alternifolia, H. amamiohsimensis, H. moellendorffii, and the unknown species using multiplex ISSR genotyping by sequencing (MIGseq, Suyama and Matsuki 2015) and compared this with our observations of morphological characteristics. Multiplexed inter-simple sequence repeats (ISSR) genotyping by sequencing (MIG-seq) is a technique used to obtain many single nucleotide polymorphisms (SNPs) throughout a genome, which is valuable for determining molecular phylogenetic trees. It has been successfully applied to resolve the taxonomy of closely related taxa, including in Hydrangea ). Based on the phylogenetic hypotheses resulting from MIG-seq analysis and subsequent morphological observations, we describe Hydrangea marunoi, sp. nov.

Morphological observation and assessment of conservation status
To assess the novelty of the unknown species, we consulted the taxonomic literature (Ohba 1985a(Ohba , b, 2001Wei and Bartholomew 2001;Ohashi 2017) and herbarium specimens at FU, KAG, KAP, and TI, as well as the digitized specimen images of FKSE, TRPM, and those available at the Shimane Nature Museum of Mt. Sanbe available on the web (Digital herbarium of Shimane University Faculty of Life and Environment Sciences http://tayousei.life.shimane-u.ac.jp/harbarium/).
The conservation status was calculated following the IUCN Red List categories and criteria v3.1 (IUCN 2012) and IUCN guideline (IUCN 2019). The Extent of Occurrence (EOO) and Area of Occupancy (AOO) were calculated using the GeoCAT software (Bachman et al. 2011).

Taxon sampling for phylogenetic analysis
To perform the phylogenetic analysis, 52 samples of the Hydrangea sect. Cardiandra were gathered from both our field surveys in Japan and herbarium specimens deposited at the Kagoshima University Museum (KAG): 10 H. marunoi samples, two H. amamiohsimensis samples, four H. moellendorffii samples, and 36 H. alternifolia samples (Suppl. material 1: Table S1; Fig. 1). Additionally, one sample of H. bifida (Maxim) Y.De Smet & C.Granados of Hydrangea sect. Deinanthe was used as the outgroup. During the field survey, a small piece of leaf was cut, placed in a tea bag, and dried with silica gel in a zip-lock bag.

MIG-seq analysis
Total DNA was extracted from dried leaves using the cetyl trimethylammonium bromide (CTAB) method (Doyle and Doyle 1990). For de novo SNP detection, MIG-seq (Suyama and Matsuki 2015) was performed according to the protocol described by Suyama et al. (2022). To prepare the MIG-seq library, a two-step PCR amplification process was performed: ISSR regions were amplified using the first PCR, and Illumina sequencing adaptors and indices were added to the first PCR products during the second PCR. Sequencing was performed on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) using a MiSeq Reagent Kit v3 (150 cycles; Illumina). We skipped the sequencing of the first 17 bases of reads 1 and 2 (SSR primer regions and anchors) using "DarkCycle." Low-quality and extremely short reads containing adapter sequences were removed using Trimmomatic 0.39 (Bolger et al. 2014). The Stacks 2.60 pipeline (Rochette et al. 2019) was used for de novo SNP genotyping with the following parameters: the minimum depth of coverage required to create a stack (m) = 3, the maximum distance between stacks (M) = 2, and the maximum mismatches between loci when building the catalog (n) = 2. Three criteria were used for the SNP filtering. First, any SNP site where one of the two alleles had less than three counts was filtered out owing to the difficulty in distinguishing polymorphisms from sequencing errors when the minor allele count of SNPs is extremely low (Roesti et al. 2012). Second, SNPs with high heterozygosity (Ho ≥ 0.6) were removed because excess heterozygosity may have resulted from artifactual loci built from several paralogous genomic regions. Third, SNPs with a genotyping rate of < 30% were eliminated. Using the third criterion, the SNPs retained by at least 16 samples were included in the SNP dataset.
Maximum likelihood phylogeny based on SNPs was inferred using the RAxML 8.2.10 software (Stamatakis 2014). We used a GTRCAT model with an ascertainment bias correction using the Lewis method and performed 1,000 replicates of parallelized tree search bootstrapping.

Results
Among the 17,753,114 raw reads (334,964 ± 34,812 reads per sample) obtained, 13,254,044 reads (250,076 ± 28,521 reads per sample) remained after quality control. After de novo SNP detection and filtering, 1875 loci and 4506 SNPs were identified. Hydrangea bifida (JPN4970) was removed from the SNP dataset because of its high proportion of missing data (0.982). The ten H. marunoi samples were monophyletic and formed two geographically defined groups; populations from Mt. Nokubi (K1658-1661) and that from the Oda River (JPN9950, JPN10103, K1633, K1637, K1638, KAG088891) (Fig. 2). Hydrangea marunoi was sister to a clade that included H. amamiohsimensis and H. moellendorffii. Hydrangea marunoi, H. amamiohsimensis, and H. moellendorffii were all supported as monophyletic by bootstrap values of 100%. Hydrangea alternifolia was sister to a clade that included these three species.
Etymology. The species epithet marunoi is named after Mr. Katsutoshi Maruno, a local botanist who made significant contributions, including elucidating the flora of Kagoshima Prefecture and collecting specimens of this species, as cited earlier.
Vernacular name. Kimotsuki kusa-ajisai (suggested here). 'Kimotsuki' named after the Kimotsuki Mountains in Osumi Peninsula where the species occur and 'kusaajisai' is the common Japanese name for the species of Hydrangea sect. Cardiandra.  (Fig. 1) and the total number of individuals is estimated to be fewer than 1000. Based on the specimen records, the extent of occurrence (EOO) is calculated to be 162 km 2 by GeoCAT (Bachman et al. 2011) and the area of occupancy (AOO) is 40 km 2 . Some of the habitats are located within the protected area of the Kirishima-Kinkowan National Park and the Inaodake Nature Conservation Area, and the habitat has not been disturbed. Given this situation, it is assessed here as Vulnerable according to the IUCN criterion D (IUCN 2012, 2019).

Conservation status. Vulnerable (VU). Hydrangea marunoi is known from several populations in Osumi Peninsula
Notes. The style length is one of the critical characteristics in delimiting the taxa of the Japanese Hydrangea sect. Cardiandra (Ohba 1985a(Ohba , b, 2001Ohashi 2017). Ohashi (2017) described the Hydrangea alternifolia style length as 1-1.5 mm (fruiting), whereas Ohba (1985bOhba ( , 2001 described it as 0.5-1(-1.2) mm. Our examination of the H. alternifolia specimens resulted in the style length varied from 0.6-1.2 mm, supporting Ohba's description. One possible explanation is that Ohashi (2017)

Discussion
In appearance, H. marunoi is more similar to H. alternifolia, typically having threepetaloid calyx lobes in marginal flowers, than to H. moellendorffii, which has two-petaloid calyx lobes or to H. amamiohsimensis without petaloid calyx lobes. However, the MIG-seq tree (Fig. 2) clearly exhibited that H. marunoi is more closely related to the clade consisting of H. amamiohsimensis endemic to Amami-Oshima, an island located 583 km south of Kyushu Island, and H. moellendorffii of Iriomote Island, located 282 km east of Taiwan, than to H. alternifolia widely distributed on Honshu, Shikoku, and Kyushu islands (Fig. 2) The MIG-seq tree also revealed that a clade H. alternifolia consisting of three samples JPN10093, 10106 and 10107, collected from the Sea of Japan (western) side of Honshu Island (yellow squares in Fig. 1, designated as clade A) is highly differentiated from the rest of H. alternifolia samples (designated as H. alternifolia s.str.) (Fig. 2). Further morphological studies based on additional materials are required to characterize this clade.