Morphological Variation in Leaf Dissection of Rheum palmatum Complex (Polygonaceae)

Aims Rheum palmatum complex comprises all taxa within section Palmata in the genus Rheum, including R. officinale, R. palmatum, R. tanguticum, R. tanguticum var. liupanshanense and R. laciniatum. The identification of the taxa in section Palmata is based primarily on the degree of leaf blade dissection and the shape of the lobes; however, difficulties in species identification may arise from their significant variation. The aim of this study is to analyze the patterns of variation in leaf blade characteristics within and among populations through population-based sampling covering the entire distribution range of R. palmatum complex. Methods Samples were taken from 2340 leaves from 780 individuals and 44 populations representing the four species, and the degree of leaf blade dissection and the shape of the lobe were measured to yield a set of quantitative data. Furthermore, those data were statistically analyzed. Important Findings The statistical analysis showed that the degree of leaf blade dissection is continuous from lobed to parted, and the shape of the lobe is also continuous from broadly triangular to lanceolate both within and between populations. We suggested that taxa in section Palmata should be considered as one species. Based on the research on the R. palmatum complex, we considered that the quantitative characteristics were greatly influenced by the environment. Therefore, it is not reliable to delimitate the species according to the continuously quantitative vegetative characteristics.


Introduction
Species are fundamental units of systematic, ecological and evolutionary studies, and species delimitation has long been the focus of biologists' attention. The discovery and description of species is a major endeavor in the field of systematics. However, most species continue to be circumscribed based on morphological comparisons of museum specimens [1]. Published descriptions of new plant and animal species almost always include a herbarium specimen designated as a holotype and a list of diagnostic morphological features [1]. It may be insufficient to delimit species merely based on herbarium specimens because such specimens usually represent discontinuous sampling and only a handful of individuals from a limited number of localities or populations. Discontinuous sampling may fail to reveal the total geographic-morphological variations that occur within species, and such ignored variations may represent important characteristics that demonstrate the continuity within/among populations. Discontinuous sampling may also lead to the publication of many microspecies within a genus, and these species are often later considered as a species complex. Therefore, the delimitation of species and infraspecific taxa should be based on the study of natural populations [2].
For some genera of Chinese flora, microspecies were published by several authors based on the limited specimens collected from various localities by different collectors. Chinese botanists have devoted attention to some species complexes in recent decades [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19], but only a small number of taxa have been investigated. Rheum palmatum complex comprises all taxa (i.e., R. officinale Baill., R. palmatum Linn., R. tanguticum Maxim. ex Balf., R. tanguticum var. liupanshanense C. Y. Cheng et T. C. Kao and R. laciniatum Prain) within Rheum Sect. Palmata. These taxa differ from other members of genus Rheum in the possession of palmately lobed leaf blades [20], and they are monophyletic [21]. The key characteristics within the species complex are the degree of leaf blade dissection and the lobe shape [20]. The blades of R. officinale and R. palmatum are lobed; the lobes of R. officinale are broadly triangular, and those of R. palmatum are narrowly triangular. The blades of R. tanguticum and R. laciniatum are parted, and the lobes are narrow and triangularlanceolate. The difference between R. tanguticum and R. laciniatum is that the lobelets of the former are narrowly lanceolate, and those of the latter are linear. However, the transitional morphologies between species are often observed in both the herbarium specimens and the field individuals. Previous DNA-based studies have drawn conflicting conclusions, and the interspecific relationships remain to be resolved. Using AFLP (amplified fragment length polymorphism) markers, Suo et al. [22] concluded that R. officinale and R. palmatum are mutually sisters and are together sisters with R. tanguticum based on samples from Sichuan and Gansu provinces. Using DNA sequences from the chloroplast gene matK, Yang et al. [23] reported that R. officinale and R. tanguticum are mutually sisters and are together sisters with R. palmatum based on samples from Qinghai, Sichuan, and Gansu provinces. Based on chloroplast DNA trnL-F sequences from samples from Chongqing and Qinghai provinces, Wang et al. [21] showed that the individuals of R. palmatum, R. officinale, and R. tanguticum are nested one with one another on the phylogenetic tree. Based on the nuclear ITS (internal transcribed spacer) sequences from eight samples from Qinghai and Gansu provinces, Li et al. [24] indicated that the only sample of R. officinale in their study is sister with their seven samples of R. palmatum and R. tanguticum. According to our previous study, R. palmatum is the most widely distributed among the distribution of the species complex; the distribution of R. tanguticum overlaps with that of R. palmatum in northwestern distribution of the species complex, the distribution of R. officinale overlaps with that of R. palmatum in southeastern distribution of the species complex, and the distribution of R. officinale also overlaps with that of R. tanguticum [25]. R. laciniatum only inhabits the northern Sichuan province and embeds in the distribution areas of the other three species [20]. Based on their geographic distribution patterns and other evidence, we propose that these four species of Sect. Palmata might not be truly distinct species [25] and that this characterization might be attributable to the discontinuous sampling. The key characteristics (e.g., the degree of leaf blade dissection and the shape of the lobes) among the species in the complex are quantitative. Therefore, the main objective of the present study is to analyze the pattern of variation of these characteristics within and among populations through populationbased sampling covering the entire distribution range of R. palmatum complex.

Ethics statement
According to regulations of the People's Republic of China on the protection of wild plants, permits are required only for the species included on the list of state-protected plant species at the time of collection. Sect. Palmata is not on the list of state-protected plant species [26] (Regulations of the People's Republic of China on the protection of wild plants, http://www.people.com.cn/ item/faguiku/zrzyf/U1020.html). Thus, no specific permits were required for the described field studies, and no harm was caused to the plants and their habitats.

Materials
Morphological analyses were conducted on 2340 leaf blades from 780 individuals from 44 populations in the field, representing all taxa within R. palmatum complex, i.e., R. officinale, R. palmatum, R. tanguticum, R. tanguticum var. liupanshanense, and R. laciniatum, from the entire distribution range of Sect. Palmata (Fig. 1, Table 1). Vouchers were deposited in the herbarium of Shaanxi Normal University (SNNU) (Xi'an, China).

Measurement and analysis of leaf blade characteristics
The degree of leaf blade dissection and the shape of the lobes were measured. Because the two basal-most leaves of all investigated individuals are entire or waved, the third, fourth and fifth leaves from the bottom of the stem were measured in each individual. The mean value was obtained from three leaves for each individual. The leaf blades of the species complex are fivepalmate lobed with five main veins, and the five lobes and veins were accordingly named the central, lateral and basal lobes and veins, respectively, and the clefts between the central and lateral and the lateral and basal lobes were called the lateral and basal clefts, respectively (Fig. 2).
To characterize the degree of leaf blade dissection and the shape of the lobes, the following parameters were measured on the leaf blade: AB, the length of the central vein, to represent the length of leaf blade; AC, from the connection point of the blade and petiole to the lowest point of the sinus between the central and lateral lobes, to represent the depth of the lateral cleft; CD, the distance between the two lowest points of the sinus of the two lateral clefts, to represent the width of the central lobe; and BE, the length of the central vein from the leaf blade apex (B) to CD, to represent the length of the central lobe (Fig. 2).
To intuitively indicate the changes in the leaf lobe width and the leaf cleft depth, the above lines (i.e., parameters) were linked to form two triangles, ABC and BCE. In triangle ABC, if the length of the central vein (AB) is regarded as a unit of length, then the values of AC/AB vary with the depth of the leaf cleft. A smaller AC/AB value indicates that the leaf cleft is deeper (i.e., parted), whereas a larger AC/AB value shows that the leaf cleft is shallower (i.e., lobed). Likewise, if BE, the height of triangle BCD, is regarded as a unit of length, then the values of CD/BE vary with the width of the leaf lobe. A smaller CD/BE value indicates that the leaf lobe is narrower (i.e., lanceolate), whereas a larger CD/BE value shows that the leaf lobe is wider (i.e., broadly triangular). Therefore, the analyses of inter-population and intra-population variations were conducted at the individual and population levels based on the values of AC/AB and CD/BE. Variation among the populations can be observed from the scatter diagram comparing the two indices of leaf lobes and from the histogram indicating the number of individuals in different ranges of leaf cleft indices. Variation within the populations can be obtained from the histogram indicating the range of variation of leaf cleft indices in different individuals.
Using the parameters measured in the field, the maximum and minimum values of AC/AB and CD/BE were obtained at the individual and population levels, and the difference between the maximum and minimum values for each parameter were then assessed. Because the scatter diagram showed that the AC/AB and CD/BE values among different populations were continuous within taxa of R. palmatum complex, the differences between the maximum and minimum values of AC/AB or CD/BE were divided into three equal intervals, and each interval corresponded theoretically to a morphological variation range of leaf blade. Thus, each morphological variation range of leaf blade was obtained. All measurements were obtained using a rule precise to 0.1 cm. The correlation analysis among the morphological parameters and the correlation analysis of these morphological parameters with environmental factors were performed with the statistical software SPSS 17.0 (Chicago, IL, USA).

Leaf variations among populations
The scatter diagram showed that the values of both AC/AB and CD/BE were continuous at the individual (Fig. 3A) and population ( Fig. 3B) levels and that neither distinct individuals nor populations could be clearly grouped. The two indices are significantly correlated (r = 0.949, P,0.01) ( Table 2). If a leaf has a larger AC/AB value, then it must have a larger CD/BE value, indicating that the blade is lobed with broadly triangular lobes; in contrast, if a leaf has a smaller AC/AB value, then it must have a smaller CD/BE value, indicating that the blade is parted with lanceolate lobes. The number of individuals of different AC/ AB values (Fig. 4A) showed a near-normal distribution, and the AC/AB values are not disconnected. In other words, some individuals in the R. palmatum complex possessed deeper leaf divisions, but most individuals were characterized by moderate leaf divisions. Despite also showing a continuous distribution, the pattern of CD/BE values (Fig. 4B) differed slightly from that of AC/AB values (Fig. 4A). Most individuals had smaller CD/BE values (,1.000).

Leaf variations within populations
The AC/AB values ranged from 0.0970 to 0.8485 at the individual level in the R. palmatum complex, and the difference between the maximum and minimum values is 0.7515 (Tables 3,  4). As shown in the scatter diagram (Fig. 3), both AC/AB and CD/BE values were continuous in all taxa of the R. palmatum complex, which displayed three gradient-like ranges in the leaf morphology, and the median value of each range may represent a leaf blade type, i.e., lobed and broad triangular for R. officinale, lobed and triangular for R. palmatum, and parted and lanceolate for R. tanguticum and laciniatum. Therefore, the range of AC/AB values was divided into three equal intervals. In other words, the theoretical maximum range of variation in AC/AB values was 0.2505 for each of the three morphological ranges, and the three variation ranges were 0.0970-0.3475, 0.3476-0.5980 and 0.5981-0.8485 (Tables 3, 4 The AC/AB values ranged from 0.1658 to 0.7271 at the population level in the R. palmatum complex (Table 4), and the difference between the maximum and minimum values was 0.5613. The difference of AC/AB values was divided into three equal intervals. In other words, the theoretical maximum variation range of AC/AB values was 0.1871 for each of the three morphological ranges, and the three variation ranges were 0.1658-0.3529, 0.3530-0.5400 and 0.5401-0.7271. These three ranges should theoretically correspond to the depth of the blade division of R. officinale, R. palmatum and R. tanguticum, respectively. Similarly, the CD/BE values ranged from 0.1798 to 1.6383 at the population level (Table 4), and the difference between the maximum and minimum values was 1.4585. The difference of the CD/BE value was also divided into three equal intervals. In other words, the theoretical maximum variation range of CD/BE values was 0.4862 for each of the three morphological ranges, and the three variation ranges were 0.1798-0.6660,  The two indices of the leaf blade cleft also differed greatly among the different individuals within populations, and the differences between populations were also not identical (Table 5; Fig. 3). The difference in the AC/AB values varied from 0.0959 to 0.3771, among which the differences for nine populations were larger than the trisection of the difference between the maximum  and minimum of AC/AB values (0.2505). The population with the smallest difference was Pop43, followed by Pop41, and that with the largest difference was Pop22, followed by Pop12. The difference in the CD/BE values varied from 0.1061 to 1.0269, among which the differences for 18 populations were larger than the trisection of the difference between the maximum and minimum of the CD/BE values (0.6483). The population with the smallest difference was Pop6, followed by Pop43, and that with the largest difference was Pop22, followed by Pop16. The difference had no correlation with the population size.

Relationships between leaf variations and elevation, longitude and latitude
Correlation analysis revealed that the AC/AB and DC/BE values were both significantly correlated with latitudes and altitudes ( Table 2, P,0.01and P,0.05, respectively). With higher latitudes and/or altitudes, the blade divisions became deeper, and the blade lobes became narrower, especially on a single mountain, such as the four populations on the north slope of Mt. Taibai, Shaanxi province. The indices of leaf division showed no significant correlation with the longitudes or with the different longitudes at approximately the same latitudes.
In general, individuals from populations in the southeastern area of the distribution of the species complex tended to have shallower lateral clefts of leaf blades and wider central lobes than those from the northwestern area. The differences among all of the populations are gradational rather than distinct.

Discussion
Traditional classification and species delimitation are mainly based on herbarium specimens, and this strategy introduces many limitations. For example, herbarium specimens are often partial but not entire plants; furthermore, very few specimens are collected from the same location, rendering it difficult to reflect the within-population variation. Coupled with the impact of the concept of type species, some scholars believe that a specimen can represent all the characteristics of a given species. Many new species are established based on the differences between the specimens in hand and holotype according to one or two morphological traits. Individuals that are distinct from the holotype cannot be established as new taxa [2].
In the present study, the morphological traits of leaf blades for discriminating the species in R. palmatum complex were analyzed at the population level. Unfortunately, species within the species complex cannot be discriminated by the degree of leaf blade dissection and the shape of the lobes. As shown in Figure 3, the depth of the lateral cleft (represented by AC/AB) and the width of the central lobe (represented by DC/BE) are continuous. The present study indicates that the two indices (AC/AB and DC/BE) are continuous among different species of the complex, and even the within-population variation could be larger than the interpecific variation. Our findings are consistent with those of previous morphological studies. Ge and Hong [5,7] studied morphological characteristics such as leaf shape, tooth number and size of leaf margins of Adenophora potaninii complex, and no discontinuities were found among them. Based on their findings, they recognized only one species and two subspecies within the species complex. Yang et al. [27] investigated the morphological variation pattern of Medicago sativa complex, and they showed that the morphology of stems and leaves was too similar to be the key characteristics for the complex.
Leaves are photosynthetic organs. To absorb sufficient light energy, leaves must be as wide as possible. Meanwhile, to facilitate gas exchange (CO 2 , O 2 and H 2 O), leaves tend to be flat and thin. Therefore, leaves become the organ with the largest contact area with the environment, and they are thus heavily influenced by the environment. Many quantitative characteristics in plants, including size, weight and number, are controlled by polygenes [28], and the variation of these characteristics is continuous within and/or among populations. Therefore, these characteristics are less meaningful for species delimitation. Although such quantitative characteristics are greatly influenced by the environment, their intrinsic genetic variety may not be high. Thus, they are not good taxonomic characteristics [2]. The traditional identification of R. palmatum complex was indeed based on unreliable morphological characteristics. Morphological characteristics based on the degree of leaf blade dissection and the lobe shape are continuous among and within the studied populations. Yang and Zhang published a new species, R. qinlingense, which was later treated as a synonym of R. palmatum [29]. According to our analysis, R. qinlingense actually represents the intermediate type of R. officinale and R. palmatum. Wu et al. [29] also suggested that R. officinale and R. palmatum as recorded in Flora Tsinlingensis [30] should actually be considered as a single species. Our field survey showed that different leaf blades described as R. officinale and R. palmatum can be found within a single population, such as Pop14 (Xiabansi, Mt. Taibai, the major peak of Mts. Qinling). This phenomenon can also be found in R. palmatum and R. tanguticum, such as Pop38 (Jimai Village, Dari County, Qinghai Province). As    Table 3. Cont.  tanguticum. Similar findings also exist for other taxa, e.g., Gallego et al. [31]. Our analysis did not support R. officinale, R. tanguticum or R. laciniatum as independent species. It appears more reasonable to recognize these species as synonyms of R. palmatum. AC/AB and DC/BE values were found to be significantly correlated with each other, indicating that the degree of leaf blade dissection and the shape of the lobes were not meaningful for identifying the species in R. palmatum complex. Moreover, AC/ AB and DC/BE were both significantly correlated with latitudes, altitudes and longitudes (Table 2). Continuous variation in morphological characteristics among species in R. palmatum complex may be caused by geographical and ecological factors (e.g., altitude, latitude, longitude). With the latitude rising, the growth environment of R. palmatum complex becomes increas-   in the introduction. It should be noted that these studies [22][23][24] have involved only a limited number of samples for each species. It is clear that the morphological analysis must be complemented by an analysis of molecular characteristics, such as ITS (internal transcribed spacer) sequences or single (or low) copy nuclear genes, for better resolution within Sect. Palmata and to examine the interspecific relationships proposed in this study. A molecular study is currently underway to gain further insight into this issue. As for R. laciniatum, its difference from R. tanguticum lies only in the shape of the lobelets, which are linear and lanceolate for R. laciniatum and R. tanguticum, respectively. Although the terminal lobes were not analyzed in the present study, leaf lobe types, which varied from no secondary lobes to trilobate lobes, differ greatly among different populations, among different individuals within the same population, and even among different blades of the same individual. Whether secondary lobes or trilobate lobes, their shape also varied from triangular to narrowly lanceolate.
In summary, we suggest that R. palmatum complex be considered as a single species. Rheum officinale, R. tanguticum (including var. liupanshanense) and R. laciniatum are synonyms of R. palmatum. However, our hypothesis has yet to be confirmed by further studies, possibly using various molecular markers.