Development of a mulberry core collection originated in China to enhance germplasm conservation

China, origin of mulberry, has rich genetic resources usually. High expense, limited space and wavering environment of usually conservation in vivo poses dangerous situation for mulberry. The concept of core collection could takes priority for conservation of mulberry. In this study, 560 accessions were used with 40 morphological descriptors and stratified sampling strategies for a core collection. The core collection consisted of 28 accessions, accounting for 5% of the whole collection. The core collection included seven accessions belonging to Morus alba, one accession belonging to M. alba var. macrophylla, four accessions belonging to M. atropurprea, one accession belonging to M. nigra, three accessions belonging to M. australis, seven accessions belonging to M. multicaulis, two accessions belonging to M. wittorum and three accessions belonging to M. bombycis. The quality of core collection exceeded the evaluation criteria and could be a prioritized collection for high efficient and long-term conservation for mulberry.


INTRODUCTION
Natural silk fiber, known as 'Queen of textile', is treated as a type of luxurious asset produced by silkworm.Crude silk fiber exported from China accounted for over 80% of worldwide production.Mulberry, the sole food source for silkworm, plays an important role in sericulture industry.Increasing demand for natural silk fiber all over the world needs more silkworms, leading to the lack of mulberry leaves in China.Propagation of mulberry vegetatively, natural and artificial crossing have produced abundant genetic resources.Over 3000 genotypes in China were documented (Pan 2000).Mulberry is usually conserved in vivo, exposed to environmental degradation and disadvantageous climatic conditions, resulting in the loss of genetic resources easily.Moreover, conservation in vivo of all germplasms is unpractical and highly expensive in human and financial resources.Therefore, it has been an urgent and challenging task for mulberry germplasm conservation, causing a serious threat to sustainable sericulture industry.Frankel (1984) and Brown (1989) proposed the concept of core collection, a limited set of accessions of whole collection with minimum repetitiveness and maximum genetic diversity of a species and its relatives.Owing to representative Z Yanfang et al.
number, core collection is a promising and efficient method for conservation of crops to reduce the expense and space.Core collections of many crops such as persimmon (Zhang et al. 2009), flax (Diederichsen et al. 2013), mungbean (Schafleitner et al. 2015), peach palm (Cristo-Araújo et al. 2015), apple (Liang et al. 2015), etc. have been developed.Core collection has been a preferential collection for high efficient and low cost conservation such as cassava (Escobar et al. 2000), European elms (Harvengt et al. 2004), and garlic (Keller et al. 2012).
An appropriate construction strategy is prerequisite to develop a core collection.van Hintum (2000) described a general procedure for developing a core collection in the following sections: i) identify the total sampling ratio; ii) divide all accessions in whole collection into distinct groups; iii) decide the sampling proportion within group and iv) select entries from each group.Many researchers proposed other methods for development of a core collection such as PowerCore (Kim et al. 2007), Mstrat (Gouesnard et al. 2001), stepwise clustering (Hu et al. 2000), least distance stepwise clustering (Wang et al. 2007).These different methods depend on some factors such as genetic diversity of species, the size of the whole collection, grouping of the whole collection and data type (i.e.phenotypic or molecular data).For example, in Brazil the cultivated area of soybean increased drastically, the level of genetic diversity in the soybean collection was low (Gwinner et al. 2017).Nevertheless, these methods could lay foundation on the study by van Hintum (2000).Chen et al. (2008) selected 11 accessions as core collection of Morus multicaulis Perr from 46 accessions originated in Shandong and Hebei province, China.Zhang et al. (2011) defined a core collection of 16 entries from 73 Gelu ecotype mulberry accessions in China.Guruprasad et al. (2014) analyzed 850 mulberry accessions assembled from 23 countries with molecular and phenotypic markers, resulting in a core collection including 122 entries (about 14.4% of total sampling ration).These limited studies showed the small size of whole collection from China.In present study, we firstly used 560 accessions from Mulberry Genetic Resources Catalog (Sericultural Research Institute 1986) and Mulberry Varieties Records in China (Sericultural Research Institute 1993) as whole collection in order to obtain an appropriate mulberry core collection based on morphological descriptors in order to enhance germplasm conservation.1) for development of core collection.

Materials and data identification
All accessions have been grown in field at Shandong Institute of Sericulture, Yantai City, Shandong Province, China since the foundation of the institute in 1950's.Plants were distributed in a randomized complete design.For each accession, three plants were grown.Plants were spaced 70~80 cm between the rows and 70~80 cm within the row.All these data were recorded in detail for this study and were obtained from corresponding author.

Development procedure of core collection
The development procedures included grouping principle, the total sampling ratio, sampling proportion within group and sampling method within group (Figure 1).Grouping principle was carried out in term of traditional classification based on morphological characteristics.All   2003).Two sampling method within group were stepwise clustering (Hu et al. 2000) and least distance stepwise clustering (Wang et al. 2007).Thus, 48 candidate core collections were formed to define the most suitable core collection.At least one accession should be chosen from each group.

Evaluation of candidate core collections
In order to evaluate the efficiency of candidate core collections, eight parameters were selected including coefficient of variation (CV), ratio of phenotype retained (RPR), variance of phenotypic value (VPV), variance of phenotypic frequency (VPF), index of diversity (I), deviation of phenotypic mean (D mean ), deviation of phenotypic maximum (D max ) and deviation of phenotypic minimum (D min ) (Li et al. 2002).Significant differences of the above parameters were calculated by SPSS16.0 (P=0.05).Duncan's multiple range test was used to compare the differences of these parameters among total sampling ratios, among sampling proportions within group and among sampling methods within group.Based on the result of multiple comparisons, each of parameters was allotted a rank.The results were expressed by average ranks of the different parameters (Reviewer: 3 Show how the "average rank" is calculated.).The same value of average rank showed no difference while lower value of average rank showed better efficiency by multiple comparison of total sampling ratio, grouping of all accessions, sampling proportion within group and sampling method within group (Li et al. 2002).
A homogeneity test (F-test) for variances and a t-test for means (P=0.05) were performed to determine the ultimate core collection.The mean difference percentage (MD%), variance difference percentage (VD%), variable rate (VR%), coincidence rate (CR%) and coverage (%) were used for validation of core collection.The criteria were listed as follows: (1) no more than 20% of the traits have different means significantly (at P=0.05) between the core collection and the whole collection; (2) the CR% retained by the core collection is no less than 80%; (3) MD% should lower while VD%, CR% and coverage (%) showed higher (Hu et al. 2000, Kim et al. 2007).

Screening of efficient evaluation parameters for mulberry core collection originated in China
For total sampling ratio, CV, RPR, VPF, I, D max and D min had lower P value than 0.05, showing significant difference of core collections by different total sampling ratio; for sampling proportion ratio, CV and VPF had lower P value than 0.05, showing significant difference of core collections by different sampling proportion ratios.For sampling method within group, CV, RPR, I, D max and D min had lower P value than 0.05, showing significant difference of core collections by different sampling methods within group (Table 2).Therefore, we selected six other parameters for testing efficiency of sampling strategies including CV, RPR, VPF, I, D max and D min .

Development and evaluation of mulberry core collection originated in China
A core collection of Morus multicaulis Perr.had developed by Chen et al. (2008), containing 11 entries selected from 46 accessions originated in Shandong and Hebei province, China.Zhang et al. (2011) defined a core collection of 16 entries from 73 Gelu ecotype mulberry accessions in China.Guruprasad et al. (2014) developed a core collection of 122 entries by analyzed 850 mulberry accessions assembled from 23 countries.These core collections could not represent genetic diversity of mulberry in China.In our study, we used 560 characterized accessions originated in China as whole collection for higher representatives of core collection.
Given the total sampling ratio, 10% -30% of whole collection was suggested (Brown 1989, van Hintum 2000).Various total sampling ratios were compared for suitable one because of genetic diversity of one crop, accession number of base collection, available management of genetic resources and data type (i.e.phenotypic or molecular data), etc (Balas et al. 2014, Leroy et al. 2014, Taniguchi et al. 2014).In previous studies, there was no information of effects of total sampling ratio on development of mulberry core collection (Chen et al. 2008, Zhang et al. 2011, Guruprasad et al. 2014).In our study, according to comparison of average ranks of six total sampling ratios, the lowest value of average rank was 13.73 when total sampling ratio was 5%, showing significant difference at P=0.05 (Table 3).Therefore, 5% was taken as suitable total sampling ratio.
Comparison of average ranks of four sampling proportions within group showed that the average rank was lowest and different significantly, when logarithm proportion was used as sampling proportion within group (Table 3).Logarithm  proportion could be suitable for sampling within group.
Comparison of two sampling methods within group showed that least distance stepwise clustering had lower value of average rank (26.03), but there was no significant different average rank between stepwise clustering and least distance stepwise clustering (Table 3).It showed that there was no difference of effects on efficiency of development of mulberry core collection between two methods of sampling within group.It was not in accordance with the results of Wang et al. (2007).Our results showed similarity between lowest hierarchical level and least distance because of distinct grouping and details of characterization of the whole collection.Previous studies also indicated that rational grouping could enhance the efficient sampling for core collections (van Hintum 1995, Zhang et al. 2000, Wang et al. 2011).
Stepwise clustering and least distance stepwise clustering were compared for selection of the ultimate core collection based on VD%, MD%, CR%, VR% and Coverage (%) (Table 4).The two core collections by clustering meet all validation criteria with MD % lower than 20% and CR% higher than 80%.Moreover, showed that least distance stepwise clustering showed higher MD% than stepwise clustering methods.Therefore, we confirmed that the ultimate core collection was established by least distance stepwise clustering.
The most suitable core collection could be one by least distance stepwise clustering when total sampling ratio was 5% and sampling proportion within group was logarithm proportion.All accessions of the ultimate core collection were listed in Table 5.There were 28 accessions from all eight ecotypes including M. alba L., M. alba var.macrophylla Loud., M. atropurprea Roxb., M. nigra L., M. australis Poir., M. multicaulis Perr., M. wittorum Handelb-Mazett.and M. bombycis Koidz.
In conclusion, a systematic and suitable mulberry core collection originated in China is firstly developed.Compared with previous mulberry core collections, the larger size of whole collection, more scientific and systematic development strategies were defined.In addition, Mulberry core collection originated in China in our study took advantages of more ecotypes.The core collection can be considered as a preferential collection for conservation and characterization of mulberry.

Table 1 .
Forty descriptors used in establishment of mulberry core collection

Table 2 .
Variance analysis of eight parameters for 48 candidate core collections

Table 3 .
Comparison of average ranks of evaluation parameters

Table 4 .
Comparison of core collections and whole collection

Table 5 .
The list of mulberry core collection