The sample of Sasa oblongula was collected from Hengjing Villiage, Baishi Town, Yunfu City, Guangdong, China during a field trip in July 2021. Observations and measurements were taken using a magnifier (SZ-6) and a ruler with a scale of 0.5 mm. Some minor characters such as indumentum on ligules of both culm leaves and foliage leaves were observed with a stereomicroscope (Mshot MZ101). The description was made based on both living and dried material as well as relevant literature (e.g. Hu 1987, 1996; Wang and Stapleton 2006; Xia and Lin 2009). Comparisons between S. oblongula and Pseudosasa cantorii were conducted based on protologue and type specimens, and relevant specimens involved in the protologue of Arundinaria cantorii (≡Pseudosasa cantorii). The descriptive terms follow Beentje (2016) and herbaria acronyms follow Thiers (2021).

For obtaining reliable results, a reasonable proof strategy with two steps was designed to identify the systematic position of S. oblongula. The first step is to test whether S. oblongula belongs to Sasa based on our plastid tree. The second step is to identify which genus S. oblongula belongs to based on SNP tree, mainly due to low discrimination rates for those ‘three-branched’ genera in plastid results (Guo et al. 2021). For the plastid tree, a total of 24 species from 11 genera were sampled. Bambusa multiplex and Dendrocalamus strictus were set as the outgroups. All accession numbers and voucher information are listed in Table 1. For the SNP tree, a total of 14 species from seven genera belonging to subtribe Arundinarieae were included. Chimonobambusa sangzhiensis was set as outgroup. Particular emphasis in our taxon sampling was placed on the inclusion that several key generic types were all involved in this study, including Acidosasa, Indosasa, Oligostachyum, Pseudosasa, and Sasa.

Young leaves at the vegetative growth stage were collected in the field. Total genomic DNA was isolated from silica-dried leaves following the manufacturer’s specifications TIANGEN Genomic DNA Extraction Kit (TIANGEN, Beijing, China). DNA samples of concentration up to standard (≥1 μg) were sheared into fragments using Covaris M220 (Covaris, Woburn, MA). Insert size of 350 bp fragments were enriched by PCR, and the paired-end (2 × 150 bp) libraries were constructed on NovaSeq 6000 platform. About 20G deep genome skimming (DGS) data were generated. Finally, adapters and low-quality reads were filtered from raw data using Fastp v 0.23.1 (Chen et al. 2018) software.

The filtered clean reads were utilized to de novo assemble complete chloroplast (cp) genomes using GetOrganelle v 1.6.2 pipeline (Jin et al. 2018). Six k-mer values, including 21, 45, 65, 85, 105,125, were set for plastid contigs connection. Subsequently, the filtered plastid reads were transferred to Bandage (Wick et al. 2015) software for visualization processing. Two opposite plastid sequences exported from Bandage were aligned with the reference sequence Phyllostachys edulis (GenBank accession No. HQ337796), and one that matched the genomic direction of the reference was retained. The final cp genomes were manually corrected in Geneious 9.1.4 (Kearse et al. 2012).

After referring to previous plastid phylogeny studies of Arundinarieae (Zeng et al. 2010; Zhang et al. 2012), eight plastid DNA regions (atpI-atpH, psaA-ORF170, rpl32-trnL, rpoB-trnC, rps16-trnQ, trnD-trnT, trnS-trnG and trnT-trnL) were selected to reconstruct plastid phylogenetic tree. Our cp genomes were annotated from eight DNA regions of Acidosasa purpurea with ≥ 70% sequence similarity in Geneious. Then, all the annotated plastid DNA regions were extracted from whole cp genomes. Sequence directions were visualized and adjusted using Mauve v 2.4.0 (Darling et al. 2004).

The latest high-quality genome sequence of moso bamboo (Phyllostachys edulis) (Zhao et al. 2018) was selected as the chromosome-level reference genome to build an index using the software SAMtools v 1.9 (Danecek et al. 2021) and Picard v 2.27.3 (Broad Institute 2019). After filtration of low-quality data, our clean reads were processed in removal of duplicates using Fastuniq v 1.1 (Xu et al. 2012). New filtered paired reads were aligned to the reference genome by Bowtie2 v 2.4.4 (Langmead and Salzberg 2012) with the parameter of minimum acceptable alignment score for L, 0.3, 0.3. After that, SAMtools was further employed to sort alignment (BAM files). Picard was utilized to remove duplicates again with the parameter “MarkDuplicates”. GATK v 4.2.2.0 (Van der Auwera and O’Connor 2020) was performed to anchor variant calling including SNP and InDel using the joint calling method “HaplotypeCaller” in the genomic variant call format (GVCF). Each sample based on reads with mapping quality was set as at least 10 and the kmer size was set as 10 to 25. After completion of variants calling, the tool “CombineGVCFs” in GATK was carried out to combine all the GVCF files. The tool “GenotypeGVCFs” was then utilized to identify joint-called variants. Subsequently, the tool “SelectVariants” was implemented to select single nucleotide polymorphic sites (SNPs). Filtration of SNPs of low quality was then conducted in the tool “VariantFiltration” with the parameter “QD < 2.0, MQ < 40.0, FS > 60.0, SOR > 3.0, MQRankSum < -12.5 and ReadPosRankSum < -8.0”. Finally, the tool “SelectVariants” was run to extract filtered SNPs.

For a reliable phylogenetic tree based on SNP dataset, we considered that filtered raw SNPs with high missing genotype rates and low minor allele frequency will affect the accuracy of the phylogenetic trees and thus should be removed. Therefore, plink v 1.90b4.6 (Purcell et al. 2007) was operated to filter those low-quality SNPs with parameter “geno” set as 0.1 and “maf” set as 0.01. Filtered variants were then pruned with the parameter “indep-pairwise” set as 50, 10, 0.2, representing its window size, a variant count to shift the window and pairwise r2 threshold for SNPs, respectively. Finally, new clean SNP dataset was generated, and the GVCF format was transferred to PHYLIP format for phylogenetic analysis using the python script “vcf2phylip.py” (Ortiz 2019).

Chloroplast DNA regions and SNP dataset were utilized to reconstruct the phylogenetic tree, respectively. Eight plastid matrices were aligned using MAFFT v 7.450 (Katoh and Standley 2013) and concatenated as a super matrix. Maximum likelihood (ML) tree was inferred for plastid and SNP datasets using IQTREE v 1.6 in SH-aLRT test and ultrafast bootstrap (UFBoot) value (Nguyen et al. 2015). Node supports rates with SH-aLRT ≥ 80% and UFboot ≥ 95% were reliable and shown on each node. The final results were visualized with Figtree 1.4.4 (Rambaut 2018).

Ssasa oblongula has leptomorph rhizome, glabrous culm internodes, white powdery infranodal region, flat or slightly prominent nodes and culm supranodal ridge, mostly solitary branch at lower culm nodes and two to three (Fig. 3E, if three branches, central slightly dominant than lateral) branches at mid and upper culm nodes, glabrous culm leaf sheath (Fig. 3G) with erect and lanceolate blades, falcate auricles and ligules with ciliolate margin, glabrous foliage leaves blades and conspicuous transverse veins. These vegetative characters mentioned above make it fit well with the circumscription of Pseudosasa Makino ex Nakai (1925), rather Sasa. After examining the specimens of similar species and referring to the related literature (Munro 1868; Chia et al. 1983), we found that S. oblongula is most similar to P. cantorii (Munro) P. C. Keng ex S. L. Chen et al. (Zhu et al. 2006) by sharing one to three branches per nodes, glabrous internodes, the white powdery infranodal region, slightly prominent supranodal ridge, culm leaf sheath with falcate auricles, erect and lanceolate blades with serrulate margin, foliage leaf sheath with ciliate margin and truncate ligules, glabrous foliage leaf blades with conspicuous transverse veins, but differs by having nearly solid (vs. hollow) culm internode with appressed (vs. patent) branches, intravaginal (vs. transferred) and glabrous (vs. setose) abaxially culm leaf sheath with ciliate upper (vs. wholly) margin and arched (vs. truncate) ligules with ciliolate (vs. glabrous) margin, 3–6 foliage leaves with irregular (vs. coplanar) arrangement clustered at the top ultimate branch, glabrous (vs. hirsute) abaxially foliage leaf sheath with 1–4 mm (vs. 5–13 mm) long length per adjacent sheath apex, small-medium-sized (7–10 × 1.5–2.6 cm vs. 12.5–25 × 2.5–3.2 cm) foliage leaf blades with 6–7-paired (vs. 7–9-paired) secondary veins. A more detailed comparison between the two species is provided in Table 2.

To make clear the position of S. oblongula and its relationship with P. cantorii, phylogenetic Maximum likelihood analysis was conducted based on plastid, and SNP dataset was shown with SH-aLRT and UFboot values noted at each node. Our plastid phylogenetic tree indicated that S. oblongula was distantly related to those Japanese Sasa species (including generic type) (Fig. 4, SH-aLRT=99.5% & UFboot=100%) and those Sinosasa species (SH-aLRT=100% & UFboot=100%) previously assigned to Sasa from China (Qin et al. 2021). The SNP phylogenetic tree suggested that S. oblongula was sister to P. cantorii with strong support (Fig. 5, SH-aLRT=100% & UFboot=100%), supporting our morphological study well.