The Qinling Mountain in China is the main sympatric areas of natural kiwiberries populations, including A. arguta var. arguta, A. arguta var. giraldii, and A. melanandra (Figure 1A). From 2010 to 2015, we collected 16 natural populations across a 700-km region of the Qinling Mountain (33°21′18.42″–34°40′2.9844.138 N and 105°44′6.32″–111°41′46.86″ E; Figure 1B, Table 1). In total, 119 plants with 64 A. arguta var. arguta, 28 A. arguta var. giraldii, and 27 A. melanandra were systematically investigated over the 5-year period. To better understand the ecological adaptability of kiwiberries, the climate and environmental data were collected and analyzed from each sample site. The altitude, latitude and longitude were recorded by GPS; and the climate data (30 years, 1980–2010) were obtained from China Meteorological Data Service Centre (http://data.cma.cn/) or Local Meteorological Bureau. Here, we calculate the mean value of 30 years climate data as follows: annual cumulative sunshine hours, extreme maximum and minimum temperature, and monthly maximum and minimum temperature, humidity and precipitation.

Kiwiberries plants were initially classified as A. arguta var. arguta, A. arguta var. giraldii, or A. melanandra following the taxonomic treatment of Actinidia (Li J. et al., 2007), and their phenotypic traits were preliminary recorded. In addition, the fruits were harvested and stored fresh in a refrigerator for trait assessment in Northwest A&F University, Yangling, China. Voucher specimens of the leaves were made and deposited in the Herbarium of Xian Botanical Garden (XBG) for microscopic examination using an Olympus BH-2 microscope coupled with a Nikon D800 camera (Tokyo, Japan). The dormant canes were pruned to graft onto rootstocks at the kiwifruit orchard at XBG for phenological investigations, and to sprout new leaves for ploidy analysis at Wuhan Botanical Garden.

The ploidy level of each individual was determined by the flow cytometric measurement (FCM) using a flow cytometry system (Partec Cyflow Space, Germany). The new leaves of each sample were chopped and lysed in nuclear extraction buffer (solution A of High Resolution Kit, Partec, Germany) to extract the cells, followed by chromosome staining of 6-diamidino-2-phenylindole. FCM was based on a linear relationship between the fluorescence signals of the unknown sample and known internal standards. In the present study, individual DNA ploidy levels were calculated after comparing the position of the fluorescence peak of the unknown sample and the internal standard Actinidia. chinensis var. chinensis “Hongyang” (2n = 2x = 58), whose chromosome number had previously been determined by counting. The detail experimental procedure followed the basic protocol of Li et al. (2010).

The organic size is an important aspect of polyploids that has been associated with crop yield. A total of 8 quantitative characters including leaf length and width, petiole length, flower diameter, fruit weight, length, greater, and less diameter were qualified in 119 samples at least 3 times. The measurement was carried out between 2011 and 2015 based on the following procedures. The second and third leaves on strong stems were chosen and over 30 leaves were measured in each sample using Vernier calipers (Table 2; Table S2). The diameters of 30 flowers per sample were also determined by Vernier calipers. Thirty-five fruits were randomly selected for size and average weight determination (Figure 2).

Qualitative characteristic analysis is an important feature for yield the performance of kiwiberries. Therefore, the content of soluble solids, ascorbic acid, total sugar, total acid, and total amino acids in all samples were measured according to the National Standard Methods of China. Total acid content, expressed as percentage of citric acid, was determined after titrating to pH 8.2 ± 0.1 using 0.1 M NaOH (SAC, 2008). Soluble solids content was estimated as the mean of digital refractometer (Atago; Japan) readings taken of juice expressed from the 10 mm end caps removed from opposite ends of the fruits. Total sugar was determined by a direct titrimetric method using Fehling's reagent (SB/T 10203–1994). Ascorbic acid (vitamin C) content was estimated by titration using the colored oxidation/reduction indicator 2, 6-dichlorophenolindophenol (SAC, 1994). Total and amino acids were determined by the ninhydrin colorimetric method (Chinese standard GB/T 5009, 2003) using an automatic amino acid analyser (model 8800; Hitachi Ltd., Japan). Upon determining that, A. arguta var. giraldii had obviously higher content of total amino acids than A. arguta var. arguta and A. melanandra, the content of 17 amino acids were further measured. In addition, the fruit of 35 samples (14 samples of A. arguta var. arguta, 9 samples of A. arguta var. giraldii and 12 samples of A. melanandra) showed brilliant red-fleshed color, and their total anthocyanins were extracted by methanol/formic acid and analyzed by reversed-phase high-performance liquid chromatography (HPLC) followed the method described by Comeskey et al. (2009).

Statistical calculations were performed using IBM® SPSS® Statistics 20 software (IBM SPSS Inc., Chicago, IL, USA). All data were assessed for normality and homogeneity of variance (Kolmogorov–Smirnov test) prior to further analysis to fulfill the requirements of statistical analysis of variance. Differences between the fruit, flower and leaf characters of A. arguta var. arguta, A. arguta var. giraldii, and A. melanandra were evaluated using one-way ANOVA at p ≤ 0.05 and 0.01. When ANOVA was significant, the means were discriminated using Duncan's test. Correlations between ecological and climate factors vs. cytotypes' distribution and morphological/fruit quality characters vs. ploidy levels were estimated using Pearson's correlation analysis.

Three DNA ploidy levels (4x, 8x, and 10x) were detected: A. arguta var. giraldii at 8x and 10x, and A. arguta var. arguta and A. melanandra at 4x. The most frequent cytotype was 4x, accounting for 76.47% of the samples studied (Table 1). Sympatric occurrence of A. arguta var. arguta (4x), A. arguta var. giraldii (8x and 10x) and A. melanandra (4x) was common in 16 sampled sites. In addition, all possible combination of cytotypes were found: 4x and 10x plants co-existed in the SZB, SMT and GHM populations; 4x and 8x plants in the HNS, HNT, and SHX populations; and 8x and 10x plants in the SHH and SBX populations.

The eco-geographical distribution of A. arguta var. giraldii, A. arguta var. arguta, and A. melanandra on multivariate vertical gradient, solar radiation, temperature, and precipitation are listed in Table S1 and depicted in Figure 1. Base on one-way ANOVA analysis, their eco-geographical distribution among three taxa had obvious difference of climate change, except for the maximum humidity (Figure 3). Of that, A. arguta var. giraldii (8x and 10x, 1,409 m asl) located at highest altitude, but it was not significantly different from A. melanandra (1,395 m). Furthermore, the individuals with higher ploidy levels (8x and 10x) did not show better adaptability for extreme temperature, higher solar radiation, lowest precipitation, and humidity. Similarity, Pearson correlation analysis demonstrated there was no significant relationship (P < 0.01) between ploidy levels and climate conditions (Table S3).

Dimensional variation of leaves, flowers, and fruits among Actinidia arguta var. arguta, A. arguta var. giraldii, and A. melanandra are listed in Table 2. Octaploid and decaploid A. arguta var. giraldii had larger leaves than tetraploid A. arguta var. arguta and A. melanandra. However, fruit sizes of tetraploid A. melanandra were larger than A. arguta var. arguta which were, in turn, larger than those of octaploid and decaploid A. arguta var. giraldii. There was no significant difference between the flower sizes (P = 0.11 > 0.05) and fruit length (P = 0.23 > 0.05) among the three taxa. Pearson's correlation analysis confirmed that, ploidy level of three taxa had a negative relationship with fruit size (−0.67 < r < −0.53; P = 0 < 0.01), but positively correlated with leaf size (0.36 < r < 0.48; P = 0 < 0.01) (Table S4).

Analysis of the traits related to flavor and nutrition are shown in Table 3. Obvious variations were detected among the three taxa. The total sugar content varied from 6.38 to 8.67 g/100 g F.W., while the total acid content varied from 0.89 to 1.17 g/100 g F.W. The soluble solids content ranged from 12.85 to 13.88, which did not show significant difference (P = 0.092 > 0.05). The highest contents of soluble solids, total sugar, and total acid content were observed in fruit of A. arguta var. arguta, while the lowest were observed in A. arguta var. giraldii.

It is obvious that fruit of A. arguta var. giraldii contained more ascorbic acid and total amino acids than fruit from A. arguta var. arguta and A. melanandra. Ascorbic acid of three taxa varied significantly from 0.05 to 0.09 g/100 g F.W. and the total amino acids varied from 0.93 to 1.63 g/100 g F.W. In particular, the contents of 17 amino acids in octaploid and decaploid A. arguta var. giraldii were generally higher than that in other two tetraploid taxa in present study (Table 3; Figure 4). Moreover, high standard deviations (e.g., total sugar in A. arguta var. giraldii: 6. 3786 ± 1.998; Table 3) within taxa indicate high intra-variability in natural kiwiberries.

Fruit morphology of three taxa examined was highly variable both in the shape and color. A. arguta var. arguta predominantly had green fruit, whereas the fruits of A. melanandra and A. arguta var. giraldii were red or purple in color that differed in intensity (Figure 2). The average anthocyanins contents in A. arguta var. giraldii (6.11 mg/100 g F. W.) and A. melanandra (6.82 mg/100 g F. W.) are higher than A. arguta var. arguta (1.61 mg/100 g F. W.) in studied samples. Fruit shape of A. arguta var. arguta and A. melanandra (including ovoid, round, globose, oblong, and ellipsoidal) was more variable than A. arguta var. giraldii, whose fruit was similar in length but much leaner (Table 2).

Leaf shape and texture were similar in all three taxa. There was some micro-variation in the characteristics of the lower surface of the leaves (Figure S1). First, the lower surface of A. melanandra leaves were glaucous (Figure S1B), while this covering was almost completely absent from the leaves of both A. arguta var. arguta (Figure S1A) and A. arguta var. giraldii (Figures S1C,D). Secondly, the mid-vein on the lower surface of the leaves of A. arguta var. giraldii had a curly tomentum (Figures S1C,D), which was absent or very sparse on leaves of A. arguta var. arguta and A. melanandra.

There was no significant difference between the flower morphology of the three taxa (Figure S2). The flowers typically had white to light green petals, white filaments, and dark brown or black anthers. However, there were significant differences in phenology (Figure S3). A. arguta var. arguta (4x) and A. melanandra (4x) flowered in late April, partially overlapping from 2014 to 2016, whereas A. arguta var. giraldii (8x and 10x) flowered much later in mid-May. The ploidy races (8x and 10x) of A. arguta var. giraldii did not differ significantly in flowering time.

High-throughput ploidy analyses based on flow cytometry have revolutionized the study of ploidy variations and cyto-geography of Actinidia species (Ollitrault-Sammarcelli et al., 1994; Yan et al., 1997; Li et al., 2010), the chromosomes of which are particularly small and numerous (e.g., decaploid, 2n = 10x = 290). Octoploids and decaploids were initially discovered in natural A. arguta var. giraldii var. giraldii, which scattered in Qinling Mountains. Tetraploid A. arguta var. giraldii var. arguta and A. melanandra are the predominant ploidy races (76.47%), consistent with previous studies in Japan (Kataoka et al., 2010) and China (Li et al., 2013). Diploid kiwiberry (A. hypoleuca) was documented in Japan (Watanabe et al., 1990; Kataoka et al., 2010) that normally localized in relatively warm Pacific regions. Surprisingly, diploid species, as ancestors of polyploid races, were absented from the distributional areas of A. arguta var. giraldii and A. melanandra. The competitive exclusion could lead to their elimination from sympatric areas by polyploid progenies, who frequently grow larger and faster, with higher yields and better resistance to disease (Fowler and Levin, 1984; Te Beest et al., 2012; Renny-Byfield and Wendel, 2014). To sum up, we cautiously concluded that natural kiwiberries (A. arguta var. giraldii and related species) has the ploidy levels as following: A. hypoleuca-2x, A. melanandra- 4x, A. arguta var. giraldii var. arguta-4x, 6x, 7x, and 8x, A. arguta var. giraldii var. giraldii −8x and 10x.

The distributional pattern in sympatric areas of A. arguta var. giraldii var. arguta, A. arguta var. giraldii var. giraldii, and A. melanandra is characterized by a high frequency of mixed-ploidy populations (56.25%). The cytotype mixture was considered to be an evolutionarily unstable pattern, likely reflecting in situ formation or frequent cytotype immigration, consistent with the minority cytotype exclusion model (Levin, 1975), resulting in the elimination of the minority cytotypes (Baack, 2005; Zozomová-Lihová et al., 2015). Recent studies, however, have shown that mixed-cytotype populations are frequent and that balancing selection is commonly observed in these populations (Burton and Husband, 1999; Keeler, 2004; Kao, 2007; Castro et al., 2012; Duchoslav et al., 2016). Theoretical studies suggest that the long-term sympatric growth of cytotypes can only be maintained when different ploidy races have strong post- or prezygotic isolation mechanisms (Levin, 1975; Rodríguez, 1996). In the present study, no obvious niche differentiation was observed within the population to balance the spatial segregation of polyploidy species. However, over the three years studied, tetraploids consistently flowered much earlier than octaploids and decaploids (Figure S3). Divergence in flowering time is a by-product of natural selection, potentially resulting in reproductive isolation to maintain the co-existence of cytotypes (Van Dijk and Bijlsma, 1994; Petit et al., 1997); in the present study, the prezygotic isolation mechanism also played an important role in the co-existence of cytotypes in the 4x and 8x (HNA and HNT) and 4x and 10x populations (SZB and GHM).

Polyploids are deem to be more resistant to extreme condition (Stebbins, 1985; Brochmann et al., 2004), such as higher altitude, cold, heat, or drought stress, which lead them more easier to invasion of new habitats (Te Beest et al., 2012). For example, a systematic investigation of the allopolyploid, autopolyploid, and diploid hybrid species along an elevation gradient from sea level to 4,500 m within British Columbia, Canada, provided evidence that polyploids were disproportionately present at high elevations (Vamosi and McEwen, 2012). Previous studies within Actinidia genus have shown that hexaploid A. chinensis var. deliciosa plants in China grow at higher altitudes than both tetraploid and diploid A. chinensis var. chinensis plants (Li et al., 2010) and hexaploid A. arguta var. giraldii var. arguta plants are geographically localized in the colder regions of Japan, whereas diploid plants of the closely related A. hypoleuca are located in warmer regions (Kataoka et al., 2010). In this study, A. arguta var. giraldii var. giraldii, with higher ploidy levels (8x and 10x), were scattered in higher altitude than A. arguta var. giraldii var. argute, but the difference is no significant with tetraploid A. melanandra. In particular, the tetraploid A. melanandra could survived in more challenging climate, extreme temperature and few precipitation for instance (Figure 2). With diversified species, the tetraploid accounts for the highest proportion of ploidy races in natural kiwiberries, and successfully colonize different environments (Kataoka et al., 2010; Li et al., 2013). The adaptability of kiwiberries to harsh environments and climates, therefore, could not consistently enhance in response to the elevated ploidy levels.

The association between ploidy levels and morphological or quality characteristics is certainly complex in Actinidia genus. Studies on ploidy manipulation further confirmed that the fruit of colchicine-induced autotetraploids of A. chinensis were 50 to 60% larger than those of their diploid progenitors (Wu et al., 2012, 2013). A previous study on A. arguta var. giraldii and related species confirmed larger leaf and fruit sizes in the tetraploid and hexaploid fruit of A. arguta var. giraldii in Japan (Kataoka et al., 2010). In the present study, the leaf size was obviously larger in these individuals and positively correlated with higher ploidy levels (8x and 10x). Particularly, the nutritional ingredients of A. arguta var. giraldii var. giraldii (8x and 10x), including the amount of ascorbic acid and amino acids, were much higher than tetraploid A. arguta var. giraldii var. arguta and A. melanandra, implying that A. arguta var. giraldii var. giraldii could serve as a useful germplasm to attain rapid genetic improvement with respect to improved nutritional ingredients. However, the disadvantage of high ploidy races should be highlighted to scientists and breeders, as the fruit shape significantly varied after polyploidization of diploid A. chinensis (Wu et al., 2012). In addition, reduced flesh firmness and dry matter and less intense golden flesh color were observed in autotetraploid plants compared with parental diploid plants (Wu et al., 2013). In the present study, the fruit size of A. arguta var. giraldii var. giraldii decreased with the increasing ploidy level (8x and 10x), and some commercially important characteristics associated with fruit quality (e.g., total sugar and total acid content) was also poor on 8x and 10x plants. Therefore, the morphological and quality characteristics were not necessarily positively and linearly correlated with continuously increasing ploidy levels (Table S4). Thus, higher ploidy is not inevitably better in the Actinidia genus.

Studies of the natural resources of A. arguta var. giraldii and related species have enhanced our current understanding of ploidy variation, distributional pattern, and co-existence mechanisms of cytotypes. The establishment of a genetic diversity center of A. arguta var. giraldii and related species near Qinling Mountain, where species with abundant ploidy variations and diversified phenotypes presenting colorful fruit fresh, variable fruit shape, different fruit size, and nutritional compositions are detected (Figure 2), would be advantageous. The relationship between ploidy levels and agronomic traits, such as the polyploid advantage on a higher content of ascorbic acid and amino acids, and the disadvantages regarding ecological adaptation, fruit size, and fruit flavor, will improve our knowledge of multi-polyploidization in plants.

Substantially, the present study of new genotypes, including higher nutrient content, edible skins, colorful fruit flesh (red, purple) and new flavors, is the first step in germplasm exploration, and we expect to extend the utility of such genetic material to ultimately improve traditional kiwifruit quality, which is characterized by brown hairy skin, green flesh and acid flavor. To achieve task, symmetrical large-scale surveys, and detailed evaluations (solid soluble content, dry matter, aroma, yield, etc.) of natural germplasms planted in kiwifruit orchards are needed. Similarly, whole genome re-sequencing of A. arguta var. giraldii var. arguta, A. arguta var. giraldii var. giraldii, and A. melanandra should be conducted, although there may be particular challenges for analyzing improvement traits, reflecting the high genomic heterozygosity and inadequacies of the analytic techniques used to examine polyploidy. We anticipate that these methodological challenges will be overcome by advances in genome sequencing technologies (Faino and Thomma, 2014). Finally, other powerful approaches (e.g., CRISPR-Cas system) (Kanchiswamy et al., 2015) for the examination of kiwifruit will improve association studies examining the genotype vs. phenotype, an essential prerequisite to targeted breeding efforts.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.