Introduction

The genus Curcuma is a rhizomatous perennial herb comprises of more than 110 species in the family Zingiberaceae, and is commonly distributed in tropical and subtropical regions (Dosoky and Setzer 2018). Curcuma kwangsiensis S. K. Lee & C. F. Liang, also known as Guangxiezhu, is a native species in South China since ancient times. It has been broadly cultivated as cut flower and pot plant for the purposes of unique foliage and colorful bract. As a traditional Chinese medicinal material, it can be used for treating injuries, abdominal coldness and hypochondria (Zeng et al. 2008). The bioactivities of curcuminoids and volatile oil are widely applied in medicine, food and cosmetics (Xu et al. 2018). Moreover, essential oils are considered as one of the most important active constituents of C. kwangsiensis, and have been reported to elicit diverse biological effects such as antioxidant, anti-inflammatory, antiviral and antitumor activities (Sacchetti et al. 2005). In addition to C. kwangsiensis, most Curcuma species are edible, ornamental and medicinal plants, including C. alismatifolia Gagnep. (Theanphong and Mingvanish 2017), C. amada Roxb. (Banerjee et al. 2012) and C. longa L. (Zhang et al. 2017). At present, the greatest challenge is the selection of elite genotypes and development of superior parental lines from these germplasm resources. Therefore, assessment of the genetic relationship between C. kwangsiensis and related species is of prime importance.

Although there is an increasing demand for the cultivation of C. kwangsiensis due to its immense economic and pharmaceutical values, little genomic information is available regarding its evolution process and genetic variation (Sun et al. 2018). For sustainable and effective utilization of this plant resource, the analysis of genetic diversity is of foremost importance. In previous genetic studies, various DNA markers have been implemented for the detection of genetic diversity in Curcuma spp. To date, molecular marker development and genetic diversity investigations have been mainly focused on C. alismatifolia and C. longa (Ismail et al. 2016). However, informative markers for elucidating the population genetics of C. kwangsiensis are still lacking.

Compared to other genetic markers, simple sequence repeats (SSRs) are always preferable due to their codominant inheritance, good transferability, multiple alleles and high reproducibility (Zhou et al. 2019). Traditional techniques (e.g., biotin-streptavidin capture method) are usually based on the sequencing of SSR colonies and double-enriched microsatellite library (Barboza et al. 2018). With no availability of genomic information, these technologies are costly, laborious and time consuming. However, the rapid identification of expressed sequence tag (EST)-SSR markers from expressed sequences can be performed at a much lower cost (Pan et al. 2018). Through the use of next-generation sequencing, massive amounts of EST-SSRs can be obtained from transcriptome data at a high-throughput scale. In recent years, many EST-SSR markers were discovered by high-throughput sequencing in different plant species, which have been proven to be reliable and cost effective (Zhang et al. 2019).

Managing the genetic resources of precious germplasm collections is important for conserving their genetic diversity (Duan et al. 2017). However, the heterogeneous structure, redundant genotypes, and lacking of trait information can influence the usefulness of genetic variability. Considering that it is difficult to fully characterize all the individual plants due to lack of time, space and labor, the construction of a core collection with smaller accessions for obtaining the genetic information of the initial collections is urgently needed (Di Guardo et al. 2019). Previous genetic diversity analysis has revealed that a representative genotype-based core subset is able to capture the maximum amount of genetic diversity in a plant population (Frankel 1984; Frankel and Brown 1984). At present, several core collections have been successfully developed in economically important and fruit trees, including Vitis vinifera L. (Le Cunf et al. 2008), Olea europaea L. (Belaj et al. 2012), and Ziziphus jujuba Mill. (Xu et al. 2016). However, the core collection of Curcuma has yet to be constructed.

In the present study, we aimed to (1) characterize EST-SSR markers based on transcriptome data at a high-throughput scale; (2) assess the genetic diversity of 75 Curcuma accessions through the use of these markers; and (3) construct a core subset for the entire population.

Material and methods

Total RNA and genomic DNA extraction

Fresh leaves of C. kwangsiensis were collected and immediately frozen in liquid nitrogen at − 80˚C for RNA isolation. Total RNA was extracted from each sample using the RNAprep pure plant kit (Tiangen Biotech, Beijing, China) according to the manufacturer’s instructions. The quality and quantity of RNA were examined using an Agilent 2100 Bioanalyzer and a Nanodrop 2000 spectrophotometer, respectively. For SSR marker analysis, 75 Curcuma accessions collected from different areas were employed as a test population. The information of the collection process are summarized in Supplementary Table S1. Meanwhile, genomic DNA was extracted from fresh leaf tissue by following the method of Ye et al. (2019). The integrity and quantity of DNA were assessed by 1% TAE agarose gel electrophoresis and a Nanodrop 2000 spectrophotometer, respectively.

RNA sequencing and functional analysis

cDNA library construction was performed with the RNA samples of C. kwangsiensis, and the sequence length was set to 200 bp. For paired-end RNA sequencing, Illumina HiSeq 2500 platform (Illumina, San Diego, CA) was used to sequence both ends of the cDNA fragments. The quality of raw reads was evaluated with SeqQC-V2.2 program, and high-quality reads (Q < 20) were assembled using the Trinity software with default parameters. Only the high-quality, filtered reads were utilized for subsequent analyses. The assembled sequences were then compared against the non-redundant protein database (nr) by running BLASTX with a threshold E-value of 10–6. The unigene annotations and mapping routines were conducted with gene ontology (GO) and kyoto encyclopedia of genes genomes (KEGG) analyses by searching and comparing public databases. Finally, the assembled unigenes were deposited to the figshare database archived at: https://doi.org/10.6084/m9.figshare.11955270.v4.

EST-SSR validation and polymorphism examination

Micro Satellite (MISA, http://www.pgrc.ipkgatersleben.de/misa) was employed to identify SSR markers based on the unigene sequences. The search principle and primer design of each EST-SSR were described previously (Wu et al. 2014). PCR amplification for 6 randomly selected samples was assessed with 1.5% agarose gel electrophoresis. The successfully amplified polymorphic markers was further examined with 8% polyacrylamide gel electrophoresis (PAGE). To determine genetic relationships, the forward primer of SSR markers was elongated from the M13 primer appended to the 5′-end (Schuelke 2000). Approximately 0.5 μL of PCR products with different fluorescent labels and sizes were pooled and detected using a DNA Analyzer.

Genetical and statistical analyses

Allele sizes were corrected with FlexiBin ver. 2.0 and GeneMarker ver. 2.20 (SoftGenetics LLC., State College, PA, USA). The allele number and polymorphic information content (PIC) of each polymorphic SSR loci were determined using PowerMarker ver. 3.25 program (Liu and Muse 2005). The basic genetic statistics, principal coordinate analysis (PCoA) and analysis of molecular variance (AMOVA) were conducted using GenAlEx ver. 6.0 software (Peakall and Smouse 2006) and POPGENE ver. 1.32 software (Yeh et al. 1999).

The population structure of 75 Curcuma accessions was estimated using STRUCTURE ver. 2.3 program (Pritchard et al. 2000). Delta K Evanno's method (Evanno et al. 2005) was used to calculate the optimal K value in STRUCTURE HARVEST (Earl and vonHoldt 2012). The bar plots for the probability of membership obtained from STRUCTURE data were visualized using DISTRUCT (Rosenberg 2004) and CLUMPAK (Kopelman et al. 2015). Genetic distance matrix was constructed by PowerMarker using the shared allele distance. Cluster analysis was performed to explore the relationships among 73 Curcuma accessions by using an unweighted pair-group with arithmetic mean (UPGMA) based on Nei’s unbiased genetic distances (Backeljau et al. 1996).

Construction of core collection

According to the number of accessions, a progressive sampling strategy was used to develop the core subset, by establishing 12 core collections to verify the optimum size. To ensure the reliability of the core subset establishment, 5 repetitive runs were performed with two different approaches (i.e., random search and simulated annealing) by using PowerMarker ver. 3.25 program. PowerCore software (Kim et al. 2007) was employed for subsequent data screening. The analysis was repeated 1000 times until meeting the requirement of representativeness or achieving the appropriate number of accessions. Lastly, genetic parameters-based T-test was used to determine the association between the initial collection and the core subset. The statistical analysis was carried out using SPSS ver. 18.0 package.

Results

Transcriptome assembly and annotation

In total, 8.17 Gb of raw reads were obtained and assembled into 77,976 unigenes with a mean length of 1433 bp (N50 = 1969 bp; Fig. 1a). The length of the unigenes ranged from 201 to 13,396 bp with a total of 87,071,461 nucleotides. The unigenes with a length of 1000–2000 bp (25,414, 32.6%) were the most abundant types. Besides, species classification results found that 78% of the unigenes were successfully annotated in the Musa acuminata Colla database (Fig. 1b). Other species displayed sequence similarities of less than 5%, including Elaeis guineensis Jacq. (4.6%), Phoenix dactylifera L. (3.6%), Ananas comosus (L.) Merr. (2.4%), etc.

Fig. 1
figure 1

Characterization and functional classification of C. kwangsiensis unigenes. a Unigene length distribution; b Species classification of the annotated unigenes; c GO annotation; and d KEGG pathway mapping

Full size image

For the functional analysis of unigenes in C. kwangsiensis, 24,321 unigenes were classified into biological processes (11,651, 47.9%), cellular components (7023, 28.9%) and molecular functions (5647, 23.2%) by using Blast2GO software (Fig. 1c). Within the biological processes category, cell process (22.7%) and metabolic process (21%) were the dominant terms. Of the 18 cellular component terms, cell (19.9%) and cell part (19.9%) were the most abundant components. Among the 10 molecular function terms, the binding (47.9%) component was most highly represented. KEGG analysis showed that 27,010 unigenes were classified into 158 pathways and categorized into 5 clades (Fig. 1d). Of the 19 subgroups, translation (2577, 9.5%) was the most represented pathway, followed by carbohydrate metabolism (2286, 8.5%), folding, sorting and degradation (1842, 6.8%), and overview pathways (1645, 6.1%).

Identification of EST-SSR markers

In total, 11,678 EST-SSR markers were obtained from the 77,976 identified EST sequences (Table 1). Trinucleotide repeats were found to be the most common (6955, 59.6%), while di- (4048, 34.7%), tetra- (371, 3.2%), hexa- (203, 1.7%) and penta-nucleotides (101, 0.8%) were the next most frequent in consecutive order (Fig. 2a). The number of SSR repeat motifs ranged from 5 to 36, and trinucleotides with 5–8 repeat motifs (55.9%) were the most dominant, followed by dinucleotides with 5–8 (22.9%) and 13–16 (7.7%) repeat motifs. In addition, the frequency distributions of major di- and tri-nucleotide repeat units were also evaluated. Among the dinucleotide motifs, AG/CT with 6 repeats (18.3%) was the most frequent repeat type (Fig. 2b). Moreover, the richest repeat type of trinucleotide motifs was GAA/TTC with5 repeats (12.4%). All the primer sequences were deposited to the figshare database archived at: https://doi.org/10.6084/m9.figshare.11948541.v1.

Table 1 Summary of EST-SSRs identified in C. kwangsiensis transcriptome
Full size table
Fig. 2
figure 2

Basic characteristics of the EST-SSR markers. a The frequencies of different repeat classes. b The distributions of di- and trinucleotide motifs

Full size image

Polymorphism detection and transferability of EST-SSRs

In total, 800 EST-SSR primers were ultimately synthesized according to the program criteria, including 268 (33.5%) dinucleotide, 420 (59.5%) trinucleotide and 112 (12.9%) other motif types. Of these primer pairs, 486 (60.8%) of them exhibited successful amplification with the correct sizes. However, 30 PCR products demonstrated larger sizes than the expected ones, suggesting the existence of an intron in the amplified regions. Considering that the remaining SSRs may not produce any PCR bands, they were excluded from subsequent analyses. Of the successfully amplified primers, 115 (23.7%) of them were found to be polymorphic in 8 Curcuma accessions.

Next, 24 EST-SSR markers were employed to determine the genetic relationship among 75 Curcuma accessions (Table 2). Altogether, 277 polymorphic alleles were screened with an average of 11.542 alleles for each primer pair. The parameters, such as Ne, I, Ho and He are shown in Table 3. It is worth noting that PIC can serve as a valuable index for assessing the performance of EST-SSR markers. In this study, the values of PIC ranged from 0.496 (JHH28) to 0.905 (JHH54) with an average of 0.754, implicating that the polymorphic EST-SSR markers can be used to analyze the genetic diversity of Curcuma species. Consequently, the established EST-SSR markers were employed to measure cross-species transferability and evolutionary conservation, and a high transfer rate of 73.1% was detected among 13 Curcuma species (Supplementary Table S2).

Table 2 Details of 24 functionally relevent EST-SSR markers
Full size table
Table 3 Polymorphic information of 24 EST-SSRs in 75 accessions
Full size table

Genetic diversity and population structure analyses

A clustering tree using neighbor-joining (NJ) method was constructed to analyze the genetic relationship, in which the 75 accessions were clearly divided into four clades (Fig. 3). All the C. alismatifolia cultivars were classified into clade 1, clade 2 and clade 3, comprising of 26, 10 and 16 accessions, respectively. However, there were some exceptions in clade 3, such as Cros, Ccor and Cpha01. The remaining accessions were assigned into clade 4, including ten Curcuma species. Clade 4 revealed a unique evolutionary pattern of this species, which were consistent with its origins and previous findings. The genetic parameters were highest in clade 4, whereas the lowest value were found in clade 2 (Fig. 4). At the population level, 114 private alleles were found in the four populations and distributed across 24 loci, with frequencies ranging from 0.019 to 0.400. Notably, 11 private alleles were observed in C. phaeocaulis Valeton, followed by C. kwangsiensis (n = 7) and C. cordata Wall. (n = 6). In overall, 44.7% of the private alleles were detected in clade 4, suggesting that the accessions in this group possess informative genetic diversity and may share the same ancestry.

Fig. 3
figure 3

Neighbor-joining tree of 75 Curcuma accessions based on 24 EST-SSR markers

Full size image
Fig. 4
figure 4

Graphical representation of the allelic patterns across different clades

Full size image

A two-dimensional scatter plot illustrating the dispersion among the 75 accessions was constructed using PCoA (Supplementary Fig. S1). The four clusters had good consistency with the corresponding clades produced by the clustering tree. Coordinate 1 explained 36.99% of the variance in the genotype data and distinguished clusters 3 and 4 from clusters 1 and 2. Coordinate 2 explained 21.25% of the variance in the genotype data and separated clusters 1 and 4 from clusters 2 and 3.

In the population structure analysis, the delta K approach generated an obvious peak at K = 4, where all accessions were classified into four main groups (Supplementary Fig. S2). Overall, the structure results corroborated the dendrogram with some exceptions due to the mixed ancestry (membership value < 80%). The highest number of samples was found in group 1 (29), followed by group 4 (24), group 2 (11) and group 3 (11). The accessions in group 1–3 were referred to C. alismatifolia cultivars with some admixture among groups. In particular, group 4 contained some accessions with high importance values, and most of them were typical ornamental and medical plants.

Construction of the core subset

Twelve core collections were established, accounting for 11–40% of the total accessions. The results demonstrated that the I, Na, Ne and PIC values of a core subset with 22 genotypes were increased, with the corresponding retention rates of 102.5%, 87.7%, 107.8% and 102%, respectively (Fig. 5). Meanwhile, the 22 core individuals were divided into four clusters using the newly developed markers. Seven Curcuma species were grouped into cluster 1, while the other 15 C. alismatifolia cultivars were grouped into the remaining three clusters. The core individuals in each cluster were highly consistent with the origins of each clade based on the NJ tree above (Supplementary Fig. S3). The allele frequencies of the core subset were highly correlated (R2 = 0.923) with those of the whole collection, indicating the excellent representation of the core collection (Supplementary Fig. S4).

Fig. 5
figure 5

Comparison of the genetic parameters among different core collections, including a Number of alleles. b Number of effective alleles. c Shannon's information index. d Polymorphic information content

Full size image

Discussion

Curcuma kwangsiensis, an important perennial herb in South China, has been widely applied in cosmetics, food and medicine (Zhang et al. 2011). Previous research on this species has mainly focused on plant tissue culture and medicinal component analysis (Komatsu et al. 2008). However, the exploration and decipherment of novel genomic information and genetic variation are yet to be performed. To improve its yield and industrial value, it is necessary for breeders to understanding the hereditary information of this crop at the molecular level. However, the limited number of molecular markers has severely hindered the genetic assessment of C. kwangsiensis and its breeding programs.

Next-generation sequencing has revolutionized the fields of transcriptomics and genomics, making it possible to identify large numbers of functional genes and novel molecular markers (Ellegren 2014; Taheri et al. 2018). This technology was used to generate large volumes of data for non-model species, due to its high throughput, low cost and high accuracy (Shukla et al. 2015). By taking advantage of this technology, a large fraction of the transcriptome was obtained for C. kwangsiensis. In total, 8.17 Gb of raw data were generated after sequencing. The N50 and mean length of the assembled unigenes were 1969 bp and 1433 bp, respectively, which was relatively comparable to those obtained from the de novo transcriptome in other species, including C. alismatifolia (1501 bp, 994 bp) (Taheri et al. 2019), C. longa (424 bp, 367 bp) and C. aromatica Salisb. (410 bp, 359 bp) (Sheeja et al. 2015). Notably, the higher N50 value and longer average length of C. kwangsiensis indicate a more accurate and effective assembly for the transcriptome. For gene annotation, 78% of the unigenes were matched with wild banana (M. acuminata), which might be attributed to the fact that they all belonged to Zingiberales and shared similar genome sequences in the process of evolution. Using the GO database, we identified 24,321 unigenes involved in molecular functions, cellular components and biological processes. In addition, 27,010 unigenes were annotated into 158 pathways and 5 subcategories according to KEGG analysis. To our knowledge, GO annotation helps us elucidate the function of unigenes at the macro and physiological levels, and KEGG analysis provides a basis for systematic gene function annotation with regard to gene product networks (Kanehisa et al. 2012). These findings may facilitate further research on the biological function-specific unigenes in C. kwangsiensis or other related species.

SSR marker is preferred for genetic analysis due to its high level of codominance, reproducibility and polymorphism (Barboza et al. 2018). Prior to this study, there are limited SSR markers available for this unique ornamental and medicinal crop. In the present work, we identified 11,678 EST-SSRs from the 77,976 EST sequences. The trinucleotide motifs were the most abundant, which was consistent with prior researches conducted on C. alismatifolia (Taheri et al. 2019), Zingiber officinale Roscoe (Awasthi et al. 2017), colored calla lily (Wei et al. 2016) and pummelo (Liang et al. 2015). However, Ye et al. (2019) found that dinucleotide motifs were the most abundant types in the crape myrtle genome, instead of trinucleotide motifs. Other studies also demonstrated that dinucleotide motifs was the most frequent type (Rowland et al. 2012; Li et al. 2018). All these findings imply that the dominant repetition motifs are highly variable among different plant species.

In addition, the 800 SSR primer pairs exhibited a higher amplification efficiency (60.8%) compared to other reported species, such as Taxodium (51.1%) (Cheng et al. 2015) and tree peony (47.3%) (Wu et al. 2014). However, 30 PCR products were larger than expected, which might be attributed to the occurrence of large insertion fragments, long intervening introns, large repeat numbers, or sequence assembly errors (Wei et al. 2011). Generally, the PIC value more than 0.5 reflects a high degree of molecular marker polymorphism (Bostein et al. 1980). Herein, a mean PIC value of 0.754 implicated that these primer pairs could be used to assess the genetic diversity and relationships among Curcuma species. The high abundance of polymorphisms may be attributed to the complex genetic architecture of the tested germplasm or the contingent selection of polymorphic EST-SSR markers.

Furthermore, the EST-SSR markers were chosen to evaluate the cross-species transferability among 13 Curcuma species. Notably, a high transfer rate of 73.1% was identified in most Curcuma species. Such excellent transferability may becaused by the mild conservation of the SSR-flanking sequences among the four species (Liu et al. 2019). However, only 6 out of 24 markers amplified the expected bands in C. rubescens Roxb., suggesting that it may differ from other species evolutionarily. Two markers (JHH2 and JHH12) exhibited strong cross-species transferability in all the Curcuma species. Hence, these reliable EST-SSR markers can be applied for genome mapping, ancestry tracing and evolutionary genetic analysisin the near future.

The population structure and genetic diversity of the whole collection were evaluated using 24 polymorphic EST-SSR makers. Our findings indicated that the grouping in STRUCTURE was in good agreement with the cluster analysis. Although few exceptions were observed in the four populations, the data of cluster analysis was convincing. The results of dendrogram analysis revealed that all the individual plants were primarily clustered according to their geographical origins and growth habits, which were partially consistent with those of Syamkumar and Sasikumar (2007) as well as Jan et al. (2011). Nearly all the C. alismatifolia cultivars were divided into 3 clades, while the other Curcuma species were clustered together. This indicates that genetic clustering can be used to assess whether the four populations share the same pedigrees or breeding strategies. However, few exceptions still exist, for example, Cros, Ccor and Cpha01 in clade 3. Given that the four Curcuma species were procured, transported and bred concurrently, the mislabeling may have happened because of the nearly similar crop types and indiscernible flowers.

Additionally, we found that clade 4 exhibited the highest value of genetic diversity, suggesting that the individuals in this clade can capture abundant genetic information. Previous studies also showed that the accessions of this clade were rich in the main active components, curcuminoids and volatile oil (Dosoky and Setzer 2018; Jain and Parihar 2019). In the present work, 11 private alleles were identified in C. phaeocaulis, followed by C. kwangsiensis (n = 7) and C. cordata (n = 6). The majority of private alleles (44.7%) were observed in clade 4, and it is strongly believed that these Curcuma species share a unique ancestor. Therefore, novel strategies are needed to sharply focus on the conservation of rare private alleles and utilization of precious germplasm resources.

Indeed, it is difficult and expensive to characterize the genetic diversity of the entire germplasm collection. A core subset with minimum repetitiveness and maximum genetic diversity should be constructed to represent the entire collection of a plant species (Anoumaa et al. 2017). Previous research has suggested that an appropriate sampling ratio should be employed according to the genetic features of various germplasm collections (Xu et al. 2016). Typically, the sample size of 10–30% is required to cover the high levels of genetic variability in a plant population. Several important indices of I, Na, Ne and PIC are commonly used to estimate the genetic diversity of a core collection. Herein, we established a core subset with 29.3% sampling ratio, with the highest values of I, Na, Ne, and PIC (retention rates = 102.5%, 87.7%, 107.8% and 102%, respectively). The results are comparable to the findings of recently published core collection studies, such as O. europaea (Belaj et al. 2012), Sinojackia huangmeiensis J.W. Ge & X.H. Yao (Zhao et al. 2016), and Z. jujuba (Xu et al. 2016). No significant difference was observed between the core subset and the entire collection (P < 0.05), implying that the core subset can be used to represent the entire collection. Therefore, a core subset with the highest retention rate and lowest sample size is highly recommended.

The core subset established in this study is particularly useful to breed C. kwangsiensis, which can also serve as applied for effective sampling and gene mining of a germplasm collection. However, it should be noted that the detection and identification of genetic variants in a core subset is not sufficient to the entire genetic information of a plant population. The sample size of the entire collection and limited genotype data can affect the performance of the core subset. Thus, further research should focus on the improvement of this core collection via morphological characterization of the accessions, incorporation of new accessions and enrichment of the available genetic data.

Conclusions

In this study, we developed 11,678 EST-SSR markers from the transcriptome data of C. kwangsiensis. Of the 115 polymorphic markers, a set of 24 selected markers showed high cross-species transferabilities among 13 Curcuma species. Based on the results of genetic diversity and population structure analysis, the 75 accessions could be effectively divided into four major groups according to their geographical origins and growth habits. Finally, a core collection (22 accessions) was constructed, which was useful for effective sampling and gene mining of the germplasm resources. In order to improve the accuracy and reliability of genetic study in Curcuma, more sample size and genotype data should be incorporated for the further research.