﻿Numerical analyses of seed morphology and its taxonomic significance in the genus Oxytropis DC. (Fabaceae) from northwestern China

﻿Abstract The lack of diagnostic taxonomic characteristics in some species complexes leave the species delimitation of Oxytropis DC. unresolved. Seed morphological features have proved to be useful diagnostic and taxonomic characteristics in Fabaceae. However, there are few systematic studies on the seed characteristics of Oxytropis. Here, we used scanning electron and stereoscopic microscopy to investigate the seed characteristics of 35 samples obtained from 21 Oxytropis species from northwest China. Our examination showed two main types of hilum positions, terminal and central, and five different types of seed shapes: prolonged semielliptic, reniform, prolonged reniform, quadratic, and cardiform. Seven different sculpturing patterns were identified: scaled, regulated, lophate with stellated testa cells, simple reticulate, rough, compound reticulate, and lophate with rounded testa cells. The seeds ranged from 1.27 to 2.57 mm in length and from 1.18 to 2.02 mm in width, and the length-to-width ratio ranged from 0.89 to 1.55 mm. The seed shape was constant within species and was useful for species delimitation within the genus Oxytropis when combined with other macroscopic traits. In contrast, the sculpturing patterns were highly variable at the species level and could not be used for species identification. Results of the cluster analysis and principal component analysis (PCA) indicated that the seed traits of Oxytropis species are useful for taxa identification at the species level, but have low taxonomic value at the section level.


Introduction
The genus Oxytropis DC. belongs to the tribe Galegeae (Fabaceae: Papilionoideae). It has been reported to be one of the largest groups of angiosperms, comprising approximately 330 species. The genus is distributed mainly in the cold mountainous regions of Asia, Europe, and North America (Polhill 1981;Zhu et al. 2010). It is thought to have been derived from Astragalus L. approximately 12-16 Ma, with which it shares many morphological features (Wojciechowski 2005). The genus Oxytropis is distinguished from Astragalus by beaked keels, asymmetrical leaflets, and acaulescent habit (Barneby 1952). Likely because of its relatively recent diversification, many taxonomic relationships within Oxytropis remain problematic (reviewed in Welsh 2001).
The genus Oxytropis was first established in 1802 by De Candolle (De Candolle 1802). It included 33 species, and he divided them into three groups according to whether stipules are adherent to stems or not and whether leaflets are opposite, verticillate, or neither. Bunge's (1874) comprehensive treatment of Oxytropis species in Eurasia identified four subgenera, 19 sections, and 181 species. His research also marked the beginning of modern Oxytropis research. Vassilczenko (1948) revised the work on Oxytropis in the Flora of USSR and separated the genus into six subgenera, 21 sections, and 276 species. Pavlov (1961) divided Oxytropis into four subgenera, 15 sections, and 124 species in the Flora of Kazakhstan. Leins and Merxmüller (1968) compiled 24 species, three subspecies, one variety, and two suspected species from Europe and divided them into two groups. Zhang (1998) recognized six subgenera, 146 species, 12 varieties, and three forms in Flora Reipublicae Popularis Sinicae. In contrast, Zhu and Ohashi (2000) recognized 125 species and four varieties in China. Welsh (2001) revised the genus Oxytropis in North America to include 57 taxa in only 22 species. Later, Zhu et al. (2010) taxonomically revised the genus in China and reported that it comprises three subgenera and 20 sections containing 133 species. These previous treatments of Oxytropis clarified many taxonomic problems. However, the lack of diagnostic taxonomic characteristics in some Oxytropis complexes has led to difficulties and differences in species delimitation, leaving the internal classification of Oxytropis unresolved.
Seed morphological features, such as seed shape, hilum shape, sculpturing pattern, and size, have been proven to be useful diagnostic and taxonomic characteristics in some genera of Fabaceae and other families (Lersten and Gunn 1981;Solum and Lockerman 1991;López et al. 2000;Al-Gohary and Mohamed 2007;Salimpour et al. 2007;Vural et al. 2008;Venora et al. 2009;Zorić et al. 2010;Celep et al. 2012;Kaya and Dirmenci 2012;Lantieri et al. 2013;Kamala and Aydin 2018;Rashid et al. 2018;Shemetova et al. 2018;Rashid et al. 2020). In the genus Trifolium L., Salimpour et al. (2007) reported that seed characteristics such as sculpturing pattern, shape, size, and hilum position, can be used as taxonomic markers within the section Lotoidea. In contrast, Zorić et al. (2010) concluded that the seed characteristics do not support infrageneric classification of Trifolium. Similarly, Shemetova et al. (2018) reported that seed shapes, colours, sizes, surface sculptures, and hilum positions are very diverse in Astragalus, and they emphasized that the systematic importance of seed characteristics needs to be evaluated in a phylogenetic context. However, Vural et al. (2008) found that Astragalus seed sculpturing pattern and seed shape can be used as taxonomically significant characteristics at the species level, if supported by other macromorphological characteristics. López et al. (2000) found that seed colour, weight, shape, and size, presence of an aril, and hilum position can be used as diagnostic characteristics for segregating two subtribes and delimiting lower taxonomic levels in the tribe Genisteae. Similarly, Rashid et al. (2020) concluded that seed shape, sculpturing pattern, and size are valuable characteristics for the identification and delimitation of species in the tribes Astragaleae and Trifolieae. Kamala and Aydin (2018) and Rashid et al. (2018) also reported that seed characteristics (coat, shape, colour, seed size, etc.) can be used to identify taxa in the tribe Vicieae.
Seeds of Oxytropis species were first studied by Solum and Lockerman (1991), who documented the seed coat patterns of Oxytropis riparia Litv. and Oxytropis campestris (L.) DC. Bojňanský and Fargašová (2007) studied the seeds of four European Oxytropis species and recorded their size, colour, and other information. Farrington et al. (2008) studied the morphological properties of seeds of 15 Alaskan Oxytropis taxa and found that seed coat micromorphology and anatomy can distinguish it from the genus Astragalus. Meyers et al. (2013) analysed the seed characteristics of 22 Oxytropis species in Alaska and concluded that seed coat types are highly variable at the species level and cannot be used for species identification. Erkul et al. (2015) studied the morphological properties of seeds of 13 Turkish Oxytropis taxa and found that seed characteristics have low taxonomic value in distinguishing subgenera, sections, and species. The infraspecific variation in seed traits has not been well addressed in most of the abovementioned studies because of sampling limitations. Only Meyers et al. (2013) studied whether seed traits were stable within species, along with studying the correlation between seed traits and the environment; however, they did not conduct any systematic analyses, such as cluster analysis.
Numerical taxonomy, also known as phenetics, mathematical taxonomy and multivariate morphometrics (Singh 2019), is mainly based on the overall affinity (similarity) at any taxonomic level. Quantitative traits have long been overlooked in taxonomic studies until numerical methodologies, such as cluster analysis, started to be widely applied in species delimitation (Thien et al. 1975). Recently, dendrograms and cladograms have been used instead of subjective analyses in many studies on the seed morphology of Fabaceae (Erkul et al. 2015;Fayed et al. 2019;Abusaief and Boasoul 2021). However, quantitative seed traits of the genus Oxytropis, such as length, width, length/width ratio, and weight, have not received much attention in taxonomic studies, possibly because these traits are considered fluctuating, and this fluctuation is random or excessive.
Northwest China is one of the main distribution regions of the genus Oxytropis (Zhang 1998;Zhu et al. 2010), but there is little research on the seed characteristics of Oxytropis in this area. Here, we carried out the first numerical analysis and microscopic investigation of 35 samples belonging to 21 Oxytropis species from northwest China using scanning electron and stereoscopic microscopy to elucidate the taxonomic significance of their seed micromorphology.

Materials and methods
The present study was mainly based on seeds collected in the field, with only a few seeds obtained from herbarium vouchers housed at the herbarium of Northwest Normal University (HWTC; Table 1). Voucher specimens collected from wild seeds are also kept at HWTC. The investigated species and their sources are listed in Table 1, and the classification of genera by Zhu et al. (2010) was adopted. Seed morphology was examined using a stereoscopic microscope (Leica M205 FA). For measuring seed length and width among the samples from the field, 80 mature and representative seeds per population were measured, while among the samples from herbarium specimens, 30 seeds per specimen were measured. The minimum-maximum range, mean, standard deviations in seed length and width, and length/width ratio were calculated. For SEM, the selected representative material was directly mounted onto aluminium stubs with double adhesive tape and coated with gold prior to observation with a HITA-CHIS-450 scanning electron microscope (NWNU University) at 25 kV.

Numerical analysis
Cluster analysis and principal component analysis (PCA) were performed using the Origin 2022 software (OriginLab Corporation 2022). The raw data matrix included quantitative traits, such as length, width, L/W ratio and weight, and qualitative characteristics, such as shape, sculpturing, and hilum position. The qualitative characteristics were coded using a presence/absence (0/1) matrix. Ward's method was used for cluster analysis using Euclidean distance to interpret the morphological similarities among species. In the cluster analysis, Euclidean distance is one of the most commonly used distance measurements in hierarchical clustering, which can reflect the absolute differences of individual numerical characteristics, and were applied to analyze differences in the numerical size of dimensions (Raymond and Sylvia 1993;Farhana and Safwana 2018). The Ward error sum of squares method applies the concept of ANO-VA to classification, resulting in richer clustering information that is rarely affected by abnormal data (Ward 1963;Szekely and Rizzo 2005). In the present study, to test the validity of the seed macro-and micromorphological traits, PCA was used to select taxonomically relevant qualitative and quantitative characteristics. It is usually used to distinguish between species within a given genus.

Seed morphology
The studied seeds, all from the genus Oxytropis, had two main types of hilum positions, terminal and central, and five different types of seed shapes: prolonged semielliptic, reniform, prolonged reniform, quadratic, and cardiform (   (Table 3).

Numerical analysis
In the present study, principal components analysis (PCA) indicates three groups of traits, which explain 82.81% of the total variation (Table 4). The first principal component (PC1) exhibited 41.51% of the variability, which had a high loading component of the seed length, width, and weight. The second PC (PC2) accounted for 22.18% of the variation and was strongly associated with L/W ratio and sculpturing, whereas the

Discussion
Seed morphology of the investigated species was determined for the first time in the present study. Seed characteristics, such as coat pattern, shape, and size, have been shown to be important for the classification within genera of Fabaceae species (Lersten and Gunn 1981;Solum and Lockerman 1991;López et al. 2000;Al-Gohary and Mohamed 2007;Salimpour et al. 2007;Vural et al. 2008;Venora et al. 2009;Zorić et al. 2010;Celep et al. 2012;De-Paula and Oliveira 2012;Kaya and Dirmenci 2012;Lantieri et al. 2013). Previous studies have shown that seed shape and hilum position are taxonomically significant and can therefore be used for the classification of taxa at the genus or even species level (López et al. 2000;Salimpour et al. 2007;Vural et al. 2008). The five main types of seed shapes observed in the present study were consistent with previous findings on Oxytropis (Erkul et al 2015). The seed shapes of different populations of the same Oxytropis species were highly consistent, indicating that they were relatively constant within species. Particularly, O. ochrocephala and O. kansuensis are easily confused, as they are morphologically difficult to distinguish and are both abundant in the northwest China (Zhu et al. 2010). However, our observations demonstrate that these two species can be distinguished based on their seed shape; O. ochrocephala has a quadratic seed, whereas O. kansuensis has a cardiform seed. These results indicate that seed shape might be a useful taxonomic marker for some Oxytropis species. However, similar seed shapes exist in other species of the genus Oxytropis and other groups of Fabaceae (Erkul et al. 2015). Thus, they should be considered in combination with other macro-morphological characteristics when applied to species identification within the genus Oxytropis.
The sculpturing pattern of seeds is thought to provide useful information for the infrageneric classification of some genera of Fabaceae (Salimpour et al. 2007;Vural et al. 2008;Kamala and Aydin 2018;Rashid et al. 2020). Farrington et al. (2008) proposed that Oxytropis seed coat micromorphology and anatomy can be used to distinguish Oxytropis from its sister taxon, Astragalus. However, studies have shown that the taxonomic value of seed sculpturing patterns in Astragalus and Oxytropis species is limited. For example, a study that examined 48 species of Turkish Astragalus found only two distinct seed coat morphological types (rugulate and rugulate-reticulate) (Vural et al. 2008). Similarly, Shemetova et al. (2018) recognised two main types of seed surface in the genus Astragalus: reticulate and indistinct primary sculpture. However, these seed sculpturing patterns have also been observed in the genus Oxytropis. Farrington et al. (2008) found that Alaskan Oxytropis (15 taxa) has rugulate, rugulatereticulate, and lophate sculpturing patterns. Consistently, Erkul et al. (2015) reported three types of seed sculpturing patterns in Oxytropis, namely rugulate, rugulate-reticulate, and lophate, and proposed that seed characteristics are not useful for separating the genera Oxytropis and Astragalus. Furthermore, Meyers et al. (2013) proposed that seed coat types among the Alaskan members of Oxytropis are highly variable at the species level and cannot be used for species identification. Our results supported this hypothesis because seed sculpturing patterns are variable within some species, including O. ochrantha (XL, NMP), O. bicolor (U, TM), O. myriophylla (EDG, AG, and MX), and O. imbricata (LC, TR), suggesting that seed sculpturing pattern has a limited taxonomic value. Interestingly, in the present study, the seed sculpturing pattern appeared to be conserved differently within different sections. Seed coat patterns were stable within some species in the section Mesogaea, such as O. ochrocephala, O. kansuensis, and O. xinglongshanica, but highly variable in the species of the sections Baicalia and Eumorpha. Therefore, the taxonomic significance of seed sculpturing pattern should be comprehensively analysed using a broader sample.
Previous studies on Oxytropis have suggested that seed characteristics, such as size (length, width, and length/width ratio), shape, surface sculpturing, and weight have low taxonomic value at the infrageneric level (Solum and Lockerman 1991;Bojňanský and Fargašová 2007;Farrington et al. 2008;Meyers et al. 2013;Erkul et al. 2015). However, most of these studies only subjectively compared their quantitative traits without a systematic analysis such as a cluster analysis. Only Erkul et al. (2015) systematically analysed the seed traits in 13 Oxytropis species from Turkey, but they did not explore the variation in seed traits at the species level because of sampling limitations. In the present study, the results of the cluster analysis showed that, except for O. myriophylla, different populations of the same species were clustered into one clade, indicating that the seed traits of Oxytropis are useful for the identification of taxa at the species level. However, species belonging to different sections were present in the same clade, indicating that seed characteristics have low taxonomic value at the section level. The results of the PCA also supported the former view that populations within the same species cluster together, while the distribution of samples of different species does not show a certain regularity. Furthermore, the first PC of the PCA provided a highly dominant variability of 41.51%, the characteristics with major scores that contributed to the formation of the groups were quantitative characteristics, such as length, width, and weight of seed. The second and third PCs are mainly qualitative characteristics, accounting for 41.3% of the total variance. These results suggest that even though quantitative traits and some qualitative traits, such as seed sculpturing patterns, are highly variable within species, these traits still play an important role in systematic analysis. Therefore, it is necessary to comprehensively analyse qualitative and quantitative characteristics in future research into Oxytropis seed morphology.
To date, a comprehensive phylogenetic study of the genus Oxytropis has not been carried out. Moreover, even though several studies have utilized DNA barcodes such as ITS, trnL-F, and psbA-trnH to investigate the molecular phylogeny of Oxytropis in northwest China, the low genetic difference of these barcodes among species makes it difficult to distinguish species within this genus and solve the phylogenetic relationship among its species (Li et al. 2011;Gao et al. 2013;Lu et al. 2014). Therefore, the phylogenetic reliability of seed traits in Oxytropis cannot be confirmed. More detailed molecular phylogenetic studies and more extensive taxon sampling are needed to discover the correlation between seed features and genus taxonomy.

Conclusions
Our results suggest that the seed traits of Oxytropis are helpful for identifying taxa at the species level, but have low taxonomic value at the section level. Seed shape was constant within species and was useful for species delimitation in the genus Oxytropis when combined with other macroscopic traits. The seed sculpturing patterns were highly variable at the species level and could not be used for species identification. Although quantitative traits and some qualitative traits, such as seed sculpturing patterns, are highly variable within species, these traits still play an important role in PCA and cluster analysis. The results of the PCA and cluster analysis showed that different populations of the same species were clustered into one clade, indicating that in Oxytropis, seed traits are useful for the identification of taxa at the species level. However, species belonging to different sections also clustered into the same clade, indicating that seed characteristics have low taxonomic value at the section level.