Construction of SNP-PARMS Fingerprints and Analysis of Genetic Diversity in Taro (Colocasia esculenta)

Abstract

Taro (Colocasia esculenta) is the fifth most cultivated root crop in the world. During the asexual reproduction of taro, the frequent mutation of somatic cells leads to high genetic diversity. With the continuous increase in the amount of taro germplasm resources collected, efficiently and accurately genotyping taro has become a major problem. The identification of taro resources using penta-primer amplification refractory mutation system single-nucleotide polymorphisms (SNP-PARMS) is a relatively efficient method. After resequencing 29 taro resources in this study, approximately 86.95 million SNPs were obtained. Then, 252 specific SNP loci were screened. Based on these 252 specific SNP loci, 36 pairs of PARMS-SNP markers were formed. Among them, 9 pairs of PARMS-SNP markers with a sample loss rate > 15% were eliminated, and finally 27 pairs of PARMS-SNP markers were determined. The average values of minimal allele frequency (MAF), polymorphic information content (PIC), gene diversity (GD), and heterozygosity of these markers are 0.63, 0.34, 0.49, and 0.45, respectively. We analyzed the population structure and the evolutionary group, and the results showed that the 72 taro resources could be divided into 6 groups. The clustering result of the 72 taro resources based on phenotypic traits showed a potential congruence with the result of grouping in the evolutionary tree, with only a few differences detected between the two classifications. Using these markers, DNA fingerprint maps of 72 taro resources were constructed, and all taro resources were differentiated. Some resources show potential similarities in DNA fingerprint maps, as well is in their phenotypic traits, confirming the validity of the fingerprint. The study’s findings serve as a reference for the analysis of the genetic diversity of taro resources.

1. Introduction

Colocasia esculenta (L.) Schott, commonly known as taro, belongs to the Araceae family and is a perennial root crop. It is often cultivated as an annual plant and contains more starch than potatoes and sweet potatoes. According to the Food and Agriculture Organization (FAO), in 2020, taro was cultivated in an area of 39,000 hectares globally. Currently, it is considered an important root crop globally, with particular significance as a staple food in some Pacific Island nations, while being widely cultivated as a supplementary food crop in tropical and subtropical regions, including China [1,2]. In China, taro has been cultivated for over 2300 years. Historical records and modern research have documented a rich diversity of local varieties, often classified into cultivar groups based on corm shape like globose or cylindrical, cormel quality, and adaptation to specific regions like the well-known ‘Liyu’ taro from Hunan and ‘Fenyang’ taro from Shanxi [3].

In breeding programs, germplasm resources are a prerequisite, and accurate evaluation of these resources greatly improves their utilization. While the general advantages of DNA marker techniques over traditional morphological identification—such as environmental stability and high throughput—are well-established, the application of these technologies to characterize the extensive but understudied taro germplasm in China remains of particular value. Owing to its precision in identifying varietal purity and enabling superior parent selection, plant DNA fingerprinting technology is extensively applied in breeding programs to accelerate genetic gain. For taro, which is predominantly clonally propagated, accurately distinguishing genetically similar cultivars is crucial for conservation and breeding. The PARMS technology, a third-generation molecular marker, enables the genotyping of SNPs and InDels through fluorescence signal detection. Distinguished by its high sensitivity, accuracy, cost-effectiveness, and polymorphism, it is now established as a leading method in molecular genotyping. Thus, techniques based on DNA markers have been widely used to analyze the genetic diversity of germplasm resources and establish fingerprint maps [4,5]. Single-nucleotide polymorphisms (SNPs) have become the markers of choice due to their availability in large quantities, stable heritability, and simple and fast operation [6,7].

Notably, only a few researchers have analyzed the genetic diversity among taro resources, especially within the Chinese germplasm which hosts considerable variety. In 2020, Wang et al. used whole-genome SNP markers to analyze the genetic diversity of 234 taro germplasm resources from 16 provinces in China and divided them into eight populations [8]. Recently, Pan et al. divided 121 taro germplasm resources into three groups after genetic diversity analysis based on high-density insertion–deletion simple sequence repeat (InDel-SSR) markers covering the entire genome [9]. However, none of them screened the core markers for subsequent utilization. Guo et al. screened 22 pairs of SSRs from 100 pairs and analyzed the genetic diversity of 109 taro resources [10], providing useful markers for the identification and evaluation of taro resources. However, SSR markers rely on PCR amplification and electrophoresis, which have lower flux and genetic stability and require more time compared to SNPs [11].

Recently, molecular fingerprinting has been introduced to the identification of germplasm resources. This method directly reflects the differences in DNA levels among individual organisms and helps identify elite plant varieties. Molecular marker technology has been used to construct genetic maps of numerous species such as Ipomoea batatas [12], Lonicera japonica Thunb [13], Zanthoxylum [14], Vernonia anthelmintica [15], Taraxacum mongolicum [16], Camellia sinensis [17], Cigar tobacco [18], and Raphanus sativus [19].

Therefore, this study performed simplified genome resequencing of 29 taro germplasms collected from Hunan Province and some other provinces of China during the third national crop resource census to identify SNP markers. After further filtering, 252 SNP loci were screened to develop penta-primer amplification refractory mutation system (PARMS-SNP) markers, and 27 original PARMS-SNP markers were successfully obtained. Genotyping of 72 taro resources was conducted using these 27 SNP markers, and further population structure analysis, principal component analysis (PCA), and evolutionary analysis were used. Combined with field phenotype and clustering results, a significant amount of gene mixing was speculated between diploid and triploid taro. Finally, a fingerprint map of taro germplasm resources was constructed. The study’s findings will provide a reference for evaluating and identifying taro germplasm resources, promoting the breeding of novel varieties.

2. Materials and Methods

2.1. Plant Materials

A total of 101 taro germplasm resources provided by the Vegetable Research Institute (VRI) of the Hunan Academy of Agricultural Sciences were used in this study. Based on the preliminary investigation of phenotypic traits, 29 taro resources with significant phenotypic differences were used for SNP screening (

Table 1. Information for 29 taro resources used for simplified genome sequencing.

Sample ID	Source	Ploidy	Classification	Group	Sample ID	Source	Ploidy	Classification	Group
0	Hunan	2×	Corm type	A	114	Hunan	2×	Corm type	A
3	Hunan	2×	Corm type	A	133	Hunan	2×	Corm type	A
4	Hunan	2×	Corm type	A	34	Hunan	2×	Leaf type	B
11	Hunan	2×	Corm type	A	45	Hunan	2×	Leaf type	B
23	Hunan	2×	Corm type	A	60	Hunan	2×	Leaf type	B
38	Hunan	2×	Corm type	A	2	Hunan	3×	Corm type	C
65	Hunan	2×	Corm type	A	21	Yunnan	3×	Corm type	C
69	Hunan	2×	Corm type	A	75	Hunan	3×	Corm type	C
70	Hunan	2×	Corm type	A	78	Hunan	3×	Corm type	C
71	Fujian	2×	Corm type	A	80	Hunan	3×	Corm type	C
77	Hunan	2×	Corm type	A	85	Hunan	3×	Corm type	C
79	Hunan	2×	Corm type	A	94	Hunan	3×	Corm type	C
83	Hunan	2×	Corm type	A	96	Jiangxi	3×	Corm type	C
98	Hunan	2×	Corm type	A	115	Hunan	3×	Corm type	C
111	Hunan	2×	Corm type	A

). They were divided into three groups based on ploidy and edible use: A, B, and C. Due to the limited availability of taro resources in this resource garden and the lack of significant differences, only three materials were selected and named Group B. The diploid stem was named Group A using taro, and the triploid stem was named Group C using taro. Specific SNPs were selected by using common Y75 and Y98 for groups C and A, respectively. Group B has only three materials, so they are compared pairwise. A total of 72 taro resources were used for fingerprint verification (Supplementary Table S1). All materials were planted at the Gaoqiao Research Base of the Academy of Agricultural Sciences in Gaoqiao Town, Changsha County (Hunan Province, China). The phenotypic traits of these resources were analyzed following the national standards set for the taro germplasm resources.

2.2. Simplified Genome Resequencing

Fresh leaves were collected from the germplasm resources in the field and rapidly frozen in liquid nitrogen, and genomic DNA was extracted from these samples with a kit (Tiangen, Beijing, China). The quality of the extracted DNA was assessed by electrophoresis (0.8% agarose gel), and the quantity was assessed with an ultraviolet spectrophotometer (Agilent, Cary 7000, NC, USA). Then, a library with 400 inserted segments was constructed using Illumina’s TruSeq DNA PCR free prep kit (Takara, Dalian, China), as per the manufacturer’s instructions, and double-terminal sequencing was performed on an Illumina NovaSeq sequencing platform.

2.3. Data Processing and SNP Discovery

The raw data obtained after simplified resequencing were filtered using fastp software (v0.20.0), following the sliding window method; contaminated adapters, bases with an average Q-value of less than 20, and reads shorter than 50 bp were removed to generate high-quality sequences. The high-quality data were further compared to the taro reference genome [20] using the bwa (0.7.12-r1039) mem program [21]. Then, GATK software (version 4.1.2.0) [22] was used to make the SNP calls. Further, the obtained SNP loci were filtered according to the following criteria: (1) Fisher test of strand bias (FS) ≤ 60, (2) Haplotype Score (HS) ≤ 13.0, (3) Mapping Quality (MQ) ≥ 40, (4) Quality Depth (QD) ≥ 2, (5) ReadPosRankSum ≥ −8.0, and (6) MQRankSum > −12.5. Finally, the filtered SNPs were annotated using ANNOVAR software (version hg18) [23].

2.4. SNP Screening and Primer Designing

Further filtering of SNP loci was carried out using the following criteria: (1) minimum allele frequency (MAF) > 0.4; (2) missing rate < 0.25; (3) heterozygous rate < 0.6; (4) no other mutations in the 50 bp region on each side of the SNP site. Then, a 50 bp region was selected upstream and downstream of the SNPs. The sequences were aligned onto the reference genome using TBtools software (version 2.056), and a single aligned sequence was selected to design primers with Primer5 software (version 5.0). The designed primers were synthesized by Shanghai Biotechnology Co., Ltd. (Shanghai, China).

2.5. Development of PARMS-SNP Markers

The typing reagent for PARMS-SNP was obtained from Wuhan Jingpeptide Biotechnology Co., Ltd. (Wuhan, China) [19]. There are four final results for genetic variant classification: homozygous reference (HOM_REF), heterozygous (HET), and homozygous alternate (HOM_ALT), undetermined. The HOM_ALT classification indicates that both chromosomal copies at a particular position carry the variant allele rather than the reference sequence. The proportion of undetermined should not be high. A total of 29 samples were selected for preliminary primer screening, using water as the negative control. Then, primers with good genotyping results were selected for analyzing the genetic diversity among the 72 germplasm resources. PowerMarker software [24] was used to calculate the genetic diversity parameters, namely polymorphic information content (PIC), MAF, GD (gene diversity), and HE (heterozygosity).

2.6. Analysis of Genetic Diversity

The genetic diversity was analyzed on the typing results obtained from 72 materials (Supplementary Table S1). Then, PCA was performed using the ade4 R package (R4.1). In addition, an evolutionary tree was generated using PowerMarker (V3.25) and uploaded to https://Itol.embl.de/ (accessed on 11 October 2024) for beautification. Structure Harvester software (version 2.3.4) was used to calculate the number of groups in the population (K-value) and obtain the optimal K-value. Further, Admixture software (version 1.3.0) (http://dalexander.github.io/admixture, (accessed on 25 November 2024)) was used to analyze the genetic structure of the population based on SNP information, with the K-value set between 2 and 10 (assuming the existence of two to seven ancestral populations), a mixed model, and default settings for all other parameters. Cluster analysis was also performed based on the phenotypic traits of the taro resources using Euclidean distance and the sum of squared deviations with DPS software (version 20.00).

2.7. Construction of Taro Fingerprint Spectra

A fingerprint map was generated based on the SNP genotyping results with the Excel program.

Table 2. The information of the 27 core PARMS-SNP markers used to generate the fingerprint.

ID	Chr	Position	Ref	Alt	Sequence of Primer X	Sequence of Primer Y	Sequence of Primer Z
SNP1	LG01	142740049	C	T	GAAGGTGACCAAGTTCATGCTTGACTTCTGTGGATTATAGCTCGC	GAAGGTCGGAGTCAACGGATTTGACTTCTGTGGATTATAGCTCGT	TCCATGCATTAAACTTGCCATTCA
SNP2	LG01	208737408	C	T	GAAGGTGACCAAGTTCATGCTTTTCTTGTAACACCCCGAAAATTC	GAAGGTCGGAGTCAACGGATTTTTCTTGTAACACCCCGAAAATTT	CGATAATGGCACGTATGTTTTTGG
SNP3	LG02	64737033	T	C	GAAGGTGACCAAGTTCATGCTGCTAAAGAATCAAAGTTAACAATAGATATCT	GAAGGTCGGAGTCAACGGATTGCTAAAGAATCAAAGTTAACAATAGATATCC	TCCAAGCAATAGTATGATCGCTGA
SNP4	LG02	65417421	T	C	GAAGGTGACCAAGTTCATGCTCACTAATTTATTATTTAGAAGAATGGCAGTT	GAAGGTCGGAGTCAACGGATTCACTAATTTATTATTTAGAAGAATGGCAGTC	TTGAGCATTGTCAGGTTGAATTCC
SNP5	LG08	8979131	C	G	GAAGGTGACCAAGTTCATGCTGGAAGAGACAGTTCTGATAGAAACAC	GAAGGTCGGAGTCAACGGATTGGAAGAGACAGTTCTGATAGAAACAG	TGCCATAGACCACTACATTCGATT
SNP6	LG01	22020255	G	T	GAAGGTGACCAAGTTCATGCTGGACCATTTAATTATCAAACATATGTAAAATAG	GAAGGTCGGAGTCAACGGATTGGACCATTTAATTATCAAACATATGTAAAATAT	TCTCTCTATCTCTCTCCTCTCTCT
SNP7	LG01	48214697	A	T	GAAGGTGACCAAGTTCATGCTCCCGGGAAGCACAAATATGTATTAA	GAAGGTCGGAGTCAACGGATTCCCGGGAAGCACAAATATGTATTAT	TATTTCCTTCTTCATCCTTGGCCA
SNP8	LG12	99501267	G	A	GAAGGTGACCAAGTTCATGCTCTGAACAGTAGCCACAAGTGACTG	GAAGGTCGGAGTCAACGGATTCTGAACAGTAGCCACAAGTGACTA	GGAGAAACTCATCAAGGGACCTTA
SNP9	LG10	109034243	T	G	GAAGGTGACCAAGTTCATGCTCAATTCAGCTGCTTACACACAAGA	GAAGGTCGGAGTCAACGGATTCAATTCAGCTGCTTACACACAAGC	TCCTTGTTGACGATCCAAGAAGAT
SNP10	LG05	165161699	A	C	GAAGGTGACCAAGTTCATGCTATTTTACCTTTCTTTAAGTAAGCTTTTGA	GAAGGTCGGAGTCAACGGATTATTTTACCTTTCTTTAAGTAAGCTTTTGC	GTCCATTAAACCCCAAACAACGAT
SNP11	LG04	144128461	T	C	GAAGGTGACCAAGTTCATGCTCGATCGACCGATAATGAAAGCCT	GAAGGTCGGAGTCAACGGATTCGATCGACCGATAATGAAAGCCC	GATTGCCTTGATCCACAATCGATT
SNP12	LG14	867311	A	G	GAAGGTGACCAAGTTCATGCTGTAATTACACTGGATTAAAAATATAACCTATGA	GAAGGTCGGAGTCAACGGATTGTAATTACACTGGATTAAAAATATAACCTATGG	GGGAGAAGTTTTGAGAAACAGTGG
SNP13	LG04	148250105	T	G	GAAGGTGACCAAGTTCATGCTGGTTTCCTCGTGTCAAACAGAAT	GAAGGTCGGAGTCAACGGATTGGTTTCCTCGTGTCAAACAGAAG	GTTTTCGGACTGCTCATCTCAAG
SNP14	LG05	153062049	C	T	GAAGGTGACCAAGTTCATGCTGACAACAAGAGGCAGCCTGC	GAAGGTCGGAGTCAACGGATTGACAACAAGAGGCAGCCTGT	TATAGAGTCGTGCTGTCAAAGACC
SNP15	LG11	89325049	T	C	GAAGGTGACCAAGTTCATGCTTCTGTGTGGTTCTCCGGCA	GAAGGTCGGAGTCAACGGATTTCTGTGTGGTTCTCCGGCG	TATGCTCAGGAGATGAAGAAGAGG
SNP16	LG08	170119348	C	A	GAAGGTGACCAAGTTCATGCTCGAATTTGTCAGAGATTGCTTCCC	GAAGGTCGGAGTCAACGGATTCGAATTTGTCAGAGATTGCTTCCA	TCAAAACAAGCTGTAGAGACCTCA
SNP17	LG10	7248248	A	G	GAAGGTGACCAAGTTCATGCTCCTGGCTTTCCTGTTCACGA	GAAGGTCGGAGTCAACGGATTCCTGGCTTTCCTGTTCACGG	AGGTATTTTCCATGATTCCTGGGT
SNP18	LG06	130247652	G	T	GAAGGTGACCAAGTTCATGCTGCACTCATGGAGGGGGCG	GAAGGTCGGAGTCAACGGATTGCACTCATGGAGGGGGCT	AAGTGGGTATGAGAGAAGAACCAC
SNP19	LG08	68668940	G	A	GAAGGTGACCAAGTTCATGCTTTGGTTATTCCATATTTAGATCTATAGATAGAG	GAAGGTCGGAGTCAACGGATTTTGGTTATTCCATATTTAGATCTATAGATAGAA	TAATAGTAGCGGGGCTTTACATCC
SNP20	LG08	17307221	T	C	GAAGGTGACCAAGTTCATGCTGTACACGTCCCTCTGCTTCAT	GAAGGTCGGAGTCAACGGATTGTACACGTCCCTCTGCTTCAC	TCAAGCTCTCACTGTGATGGTTTA
SNP21	LG14	102781792	T	C	GAAGGTGACCAAGTTCATGCTAGATCTCTTCACTTGTCAAGGATAAATAT	GAAGGTCGGAGTCAACGGATTAGATCTCTTCACTTGTCAAGGATAAATAC	AGCAGACAGATCAAGCTCTGAAAT
SNP22	LG13	51666285	T	A	GAAGGTGACCAAGTTCATGCTTCACAACTGGATACACTTTCAAAGT	GAAGGTCGGAGTCAACGGATTTCACAACTGGATACACTTTCAAAGA	GGAAAGTCTTGTTAAGGCACACTC
SNP23	LG01	206486904	T	A	GAAGGTGACCAAGTTCATGCTATGCCTACAATGCACACCTTCTCT	GAAGGTCGGAGTCAACGGATTATGCCTACAATGCACACCTTCTCA	TCTCAATCATCTCCTCCAGCAAAA
SNP24	LG04	64252159	T	C	GAAGGTGACCAAGTTCATGCTTAGGTCTCCATGTTCAAGAGGAAT	GAAGGTCGGAGTCAACGGATTTAGGTCTCCATGTTCAAGAGGAAC	AATTTGGATTACCACGTGTTGCTT
SNP25	LG02	185181757	G	A	GAAGGTGACCAAGTTCATGCTTTGGGAAAGGGCTGACAAGG	GAAGGTCGGAGTCAACGGATTTTGGGAAAGGGCTGACAAGA	CATCCACATCGAGAAGATTTTCGG
SNP26	LG12	83510795	A	G	GAAGGTGACCAAGTTCATGCTAGCTAAATAGTGAGTCAAAGGATCA	GAAGGTCGGAGTCAACGGATTAGCTAAATAGTGAGTCAAAGGATCG	CCTTACATCACAGTGCTTCACAAG
SNP27	LG01	179908017	A	T	GAAGGTGACCAAGTTCATGCTGAAGGCCTTGCTGAGAAGAGA	GAAGGTCGGAGTCAACGGATTGAAGGCCTTGCTGAGAAGAGT	GTGGACATAAACCTTGTGAAGAGC

Note: X primer had FAM fluorescent matching adapter GAAGGTGACCAAGTTCATGCT; Y primer had HEX fluorescent matching adapter GAAGGTCGGAGTCAACGGATT; Z served as a universal reverse primer.

shows the relevant sequences of 27 core primers. The genotypes of the optimal combination of SNP-PARMS markers were heat mapped for fingerprinting using the software.

3. Results

3.1. Simplified Genome Resequencing and Reference Genome Alignment

Simplified genome resequencing generated a total of 10 Gb bases with a Q30 of over 95% from the samples (Supplementary Table S2). After filtering, high-quality reads with over 55,978,374 entries were obtained, with the lowest comparison rate of 91.79% (Supplementary Table S3). After mapping the reads to the reference genome, the high-quality reads were aligned. The approach showed that 60, 34, and 45 are Leucocasia gigantea with a low alignment rate of only about 60%. The alignment rate of other samples was not less than 99.3%, and the repeat alignment rate was not higher than 18%. The sequencing depth of the samples was above 3× (Supplementary Table S4). These observations indicated the suitability of the data for the development and screening of SNP molecular markers.

3.2. SNP Diversity Analysis

After filtering the raw data, SNPs were obtained. Further filtering resulted in—high-quality SNPs; the number of homozygous genotypes that were inconsistent with the reference genome ranged from 1,804,236 to 8,781,972 (Supplementary Table S5), which were distributed across the genome (Figure 1A). The number of transitions (Ts) ranged from 5,801,036 to 21,053,229, accounting for 70.5% of the variant sites. Meanwhile, the number of transversions (Tv) ranged from 2,906,423 to 9,060,484, accounting for 29.5% of the variant sites (Figure 1B). The main SNP mutation were T:A>C:G and C:G>T:A. Further, SNP annotation detected 793,450 synonymous mutations, accounting for 92%, and 908,673 on-synonymous mutations (Supplementary Table S6). Insertions ranged from 533,007 to 2,931,928, while deletions ranged from 511,472 to 152,073 (Supplementary Table S7). Among these, 38,226 frameshift insertions or deletions resulted in changes in protein-coding genes (Table 2) distributed on 14 chromosomes (Figure 1B).

3.3. Development and Validation of High-Quality PARMS-SNP Markers

After filtering, 252 perfect SNP sites were screened from 86,596,792 SNPs. Subsequently, 36 sets of PARMS-SNP markers were obtained from these 252 SNPs. These 36 sets of primers were Further used to genotype the 72 taro samples. After removing 9 sets of primers with a sample loss rate > 15%, 27 sets were finalized.

Further analysis showed that these 27 sets of primers had MAF ranging from 0.51 to 0.92 (an average of 0.63), PIC ranging from 0.14 to 0.37 (an average of 0.34), HE ranging from 0.07 to 0.90 (an average of 0.49), and GD ranging from 0.15 to 0.50 (an average of 0.45) (Figure 2). These results indicate that the 27 primer sets possess high discrimination power and are suitable for analyzing taro genetic diversity.

3.4. Analysis of Genetic Diversity Among Taro Resources

Furthermore, we analyzed the population structure. The approach determined the number of groups in the population based on the cross-validation (CV) error rate. At K = 6, the CV error value was closer to 0, which indicated that the 72 taro resources could be divided into six groups (Figure 3). Although these resources mainly come from Hunan, some of them may have been introduced from other provinces very early, and their sources cannot be studied, so it is impossible to distinguish whether they are divided into six groups and related to their sources. PCA, using the SNP data (SNPs with MAF < 0.05), divided the 72 resources into multiple groups, wherein diploids and triploids did not aggregate (Figure 4A). The evolutionary tree confirmed this grouping (Figure 4B). Additionally, the clustering result of the 72 taro resources based on phenotypic traits showed a potential congruence with the result of grouping in the evolutionary tree; only a few differences were detected between the two classifications. For example, groups 2 and 8 clustered together based on the quality traits but were far apart in the evolutionary tree (Figure 4C). On the one hand, phenotypic traits are based on individual judgment, and different people might assess the same trait in different ways, and the same person might assess differently at different times. In addition, the investigation of phenotypic traits is limited and cannot represent all the characteristics of the plant.

3.5. Construction of PARMS-SNP Fingerprint Maps

Finally, fingerprint maps of 72 taro resources were constructed using the 27 PARM-SNP primer pairs. The approach showed a clear division of the 72 resources based on these markers. Among these resources, samples 40 and 121, 97 and 17, and 59 and 37 showed only one locus difference, suggesting potential similarities in the resources. These resources also showed similarities in their phenotypic traits, confirming the validity of the fingerprint (Figure 5).

4. Discussion

Taro resources are typically classified into leaf and taro stem types based on the edible part [25]. Among these, the taro leaf corresponds to Leucocasia gigantea (Blume) Schott. Aa a result, a low level of similarity was observed when comparing the taro leaf data to the reference genome (60%). We conducted SNP screening on three leaf taro plants and only identified two markers, so we ultimately abandoned the validation of leaf taro markers and materials. On the one hand, the lack of genomic information for wild taro may lead to information loss, and at the same time, the phenotypic differences between the three wild taro types may not be significant, which may also be the reason why screening SNP markers failed.

The corm taro is divided into Colocasia esculenta var. esculenta, called Dasheen, and Colocasia esculenta var. antiquorum, called Eddoe. Dasheen is believed to have a large taro, while Eddoes have more taro seeds [26]. However, when we actually conduct phenotypic trait surveys on the collected resources, some resources cannot be well distinguished. Some resources have 8–10 taro seedlings per plant, which are the same size as the mother taro; some resources have a large mother taro and small and few offspring taro. Therefore, we think there is another intermediate taro type with diploid and triploid forms [27]. In 1999, Tahara confirmed the homologous polyploid nature of triploid taro by analyzing germplasm resources in Yunnan and Nepal and suggested that the triploids evolved from diploids [28]. The genome sequencing of diploid and triploid materials showed a comparison rate of over 99% with the reference genome. At the same time, the specific primers we selected from diploid materials can also be used in triploid materials. The present study constructed an evolutionary tree using SNP information and found gene exchange between diploids and triploids, which cannot be completely distinguished. We cannot distinguish between diploids and triploids based on phenotype and genotype, but by detecting chromosome ploidy. Therefore, it is speculated that diploids might have undergone changes in chromosome numbers during reproduction, but the sequence has not changed significantly. This speculation explains the inability to distinguish between the diploid and triploid phenotypes.

The study, after multiple rounds of screening, 252 SNP markers were obtained via the resequencing of the genome of 26 stem type taros. However, only 37 usable markers were obtained after validation. The low abundance of SNP markers on the genome is probably due to the polyploidization of taro resources. During screening, many primers demonstrated good genotyping effects but were eliminated due to their high missing rate. This fact is also related to primer design. We initially used online software for batch primer synthesis (www.snpway.com), and after verification, we obtained a small number of available primers. Later, we manually designed and modified using primer5, and upon verification, the number of available primers obtained were twice that of the initial screening. However, because these primers could distinguish the samples in this resource garden, we did not continue to redesign and screen the first primers. This is also one of the reasons why we obtained a small number of primers through screening. In the future screening and synthesis process of primers, we recommend using Primer5 to synthesize them one by one instead of in batch. Therefore, during primer design, it is essential to align the sequences obtained via different approaches, screen the modifications, and then finalize the primers.

In 2020, Wang divided 234 taro germplasms into eight populations [18]. Although most of the materials in this experiment come from Hunan Province, some of the resources were introduced from other provinces in the early years, and their sources cannot be studied, so it is impossible to distinguish whether they are divided into six groups which relate to their sources. Through cluster analysis using SNP markers, different markers from different provinces such as 96 from Jiangxi, 21 from Yunnan, and 19 from Fujian were found to be in different branches. Famous local varieties formed in different regions, such as Fujian Fuding taro, Jiangxi Hongya taro, Lipu taro, Jiangyong fragrant taro, etc., exhibit phenotype differences. Although excellent taro resources have been widely introduced to various regions without any restrictions, there are still regional differences in flowering. For example, the introduction of taro from Lipu to Jiangyong resulted in flowering. Therefore, we speculate that different climatic environments have formed different resources, and population classification is closely related to the environment.

The present study attempted to group the taro germplasm resources of China based on both the phenotype and genotype. A subsequent comparison helped conclude that both approaches are necessary to evaluate the genetic diversity of taro resources, consistent with Oladimeji’s conclusion. Oladimeji et al. used data based on genotype and phenotype to group 114 taro resources. Although they found similarities in grouping, the correlation between the two approaches was low (R = 0.01) [26]. However, the present study found that the clustering of resources based on phenotype was consistent with the grouping based on genotype, indicating a better correlation.

Fingerprinting based on molecular markers has helped identify varietal purity, F1 hybrid offspring, and varietal authenticity [29]. SNP markers are widely distributed and abundant in animal and plant genomes and do not require fragment size analysis for detection. Moreover, the SNPs are easy to operate and can be detected in batches. The present study obtained 86,596,792 SNPs from the taro resources and ultimately screened 27 SNP markers, which could distinguish the 72 resources. Thus, based on the present study’s findings, we propose the use of SNPs for evaluating taro resources, identifying the elite ones, and reducing their duplication in germplasm gardens.