Association mapping reveals novel genomic regions controlling some root and stolon traits in tetraploid potato (Solanum tuberosum L.)

Abstract

Tuber crops have measurable biological variation in root and stolon phenotyping and thus may be utilized to identify genomic regions associated with these variations. This is the first comprehensive association mapping study related to potato root and stolon traits. A diverse panel of 192 tetraploid potato (Solanum tuberosum L.) genotypes were grown in aeroponics to reveal a biologically significant variation and detection of genomic regions associated with the root and stolon traits. Phenotyping of root traits was performed by image analysis software “WinRHIZO” (a root scanning method), and stolon traits was measured manually, while SolCAP 25K potato array was used for genotyping. Significant variation was observed between the potato genotypes for root and stolon traits along with high heritabilities (0.80 in TNS to 0.95 in SL). For marker-trait associations, Q + K linear mixed model was implemented and 50 novel genomic regions were detected. Significantly associated SNPs with stolon traits were located on chr 4, chr 6, chr 7, chr 9, chr 11 and chr 12, while those linked to root traits on chr 1, chr 2, chr 3, chr 9, chr 11, and chr 12. Structure and PCA analysis grouped genotypes into four sub-populations disclosing population genetic diversity. LD decay was observed at 2.316 Mbps (r² = 0.29) in the population. The identified SNPs were associated with genes performing vital functions such as root signaling and signal transduction in stress environments (GT-2 factors, protein kinases SAPK2-like and protein phosphatases “StPP1”), transcriptional and post-transcriptional gene regulation (RNA-binding proteins), sucrose synthesis and transporter families (UGPase, Sus3, SuSy, and StSUT1) and PVY resistance (Ry sto). The findings of our study can be employed in future breeding programs for improvement in potato production.

Supplementary information

The online version contains supplementary material available at 10.1007/s13205-021-02727-6.

Introduction

In terms of human consumption, potato (Solanum tuberosum L.) ranked as the third most significant food crop all over the globe, after rice and wheat (Grossi et al. 2020). It is cultivated in over 100 countries with an annual production of 388 million tonnes (t) (FAOSTAT 2019) with considerable economic value contributing about 3% of the total GNP (gross national agricultural product) in Turkey and around 3.1% contribution in 27 countries of EU (Çalışkan et al. 2010). The genetic variation in potato genotypes for root architecture and stolon characteristics can be utilized to ameliorate the negative impacts of climate change (Wishart et al. 2013). Current breeding efforts in potato are focused merely on tuber characteristics or above-ground parts with little to no attention on roots and stolons. Stolons are lateral shoots in potato that arises from the basal node on the main shoot at the underground level. It plays a significant role in the storage of carbohydrates and later undergoes tuberization to form mature tubers. Roots and stolons are crucial to improve the quality and quantity of potato crop by acquiring nutrients and water in fluctuating environments (Schiavon et al. 2016).

Potato is a shallow-rooted crop and the distribution of roots and stolon’s in the field experienced problems related to soil texture, drought, low fertility, and pest attack (Iwama 2008). Unlike other crop species, the potato has a distinct growth pattern with characteristic roots and stolon’s concealed underground, making it harder to track the development of these traits. Special attention is, therefore, inevitable to dissect the genetic regulation of roots and stolons (underground traits) in potato. The phenotypic variation in underground traits might be influenced by multiple quantitative trait loci (QTLs) and controlled by minor-effect loci. Efficient phenotyping protocols are essential for quantitative genetic studies (Trachsel et al. 2013). It is difficult to evaluate these traits in the field because roots and stolons are sensitive and prone to damage during removal from the soil. Moreover, non-invasive imaging of underground traits in field conditions is not a cost-effective technique (de Dorlodot et al. 2007; Trachsel et al. 2011; Smith and De Smet 2012). This scenario necessitates the use of a fast, capable, and user-friendly system for the evaluation of underground traits. For precise and accurate analysis of underground traits, many indoor growing platforms have been developed, ‘aeroponic’ is among one of them. Phenotyping approaches also required imaging techniques to acquire the phenotyping data. For root imaging, more than 30 different software tools are available, but WinRHIZO is considered as more flexible and reliable software to perform many specialist tasks in a single run (Pierret et al. 2013).

Potato is a vegetatively propagated, highly heterozygous autotetraploid (2n = 4x = 48) crop with a genome size of 844 Mbp (Spooner et al. 2005; Consortium 2011). Keeping in view the complex genetic makeup of potato, improvements are usually done at the diploid level by utilizing bi-parental populations (Sharma et al. 2018). For instance, QTL mapping studies in potatoes have been performed by employing bi-parental populations (Bradshaw et al. 2008; Khu et al. 2008). However, bi-parental mapping represents various limitations like low mapping resolution due to a lesser number of recombination and segregations. The effectiveness of identified QTL for MAS (marker-assisted selection) studies were decreased due to these drawbacks (Stich and Melchinger 2010). To overcome the drawbacks of the breeding techniques such as classical breeding and bi-parental mapping, Next-generation sequencing (NGS) is one of the best alternatives. NGS offers rapid development of genome-wide markers (SNPs) that have opened opportunities to explore the relationship between genetic and phenotypic diversity with the resolution never reached before. Dominant (AFLP) markers were used by Björn et al. (2008) for marker-trait associations in potato, but recent studies showed limitations about AFLP such as they are anonymous and cannot readily link to the potato genome sequence (Ritland 2005). Single-nucleotide polymorphisms (SNPs) are co-dominant, and several studies endorse their efficiency in genome-wide association studies of potato (Stich et al. 2013; Mosquera et al. 2016; Vos et al. 2017; Sharma et al. 2018; Khlestkin et al. 2019; Charlotte et al. 2020).

Association Mapping (AM) using high-throughput dense DNA markers such as SNPs is a robust tool to find out the marker-trait associations in crops. AM helps to identify the specific functional variants (i.e., alleles) which can be used to develop selection markers for a specific trait (Oraguzie et al. 2007). AM works on the principle of Linkage Disequilibrium (LD). Briefly, LD is non-random association between two alleles in a population. A 5 cM genetic distance for genome-wide LD in potato was stated by Björn et al. (2010). Stich et al. (2013) indicated that LD decay occurs with an intensity of 275 bp in tetraploid potato. Contrarily, Vos et al. (2017) found LD values ranged between 0.7 and 0.9 Mb in tetraploid potato using LD_1/2,90 estimator. The 2–4 Mb distance is the general trend for LD decay values in potato crop. LD is linked to marker density and population composition, and rapid LD decay shows a low tendency of markers inherited together (Tamisier et al. 2020). Zia et al. (2020) are of the view that LD decays with almost 1.22 Mb to an r² value of 0.20, utilizing SNP markers generated from SolCAP 12k array.

In potato, SolCAP array (8300 marker SNP array) has been established in the near past by utilizing 69,011 high-confidence SNPs (Hamilton et al. 2011; Felcher et al. 2012). SNP genotyping of potato germplasm can be done efficiently at a considerable cost through this array. Furthermore, a 20K Infinium SNP array was developed by Vos et al. (2015) which facilitated in understanding of the breeding history of the potato. Population structure can be examined using SNP genotyping platforms (Illumina) with the availability of a set of random SNP markers and diverse potato germplasm. The historical recombination and knowledge of population structure can be utilized effectively in association genetic studies when modeling marker-trait relationships (Ortiz 2020). Several kinds of literature have been published related to AM in crops such as Arabidopsis thaliana (Hagenblad and Nordborg 2002), Wheat (Breseghello and Sorrells 2006), sugar beet (Kraft et al. 2000), sugarcane (Jannoo et al. 1999), barley (Kraakman et al. 2006), and soybean (Zhu et al. 2003).

Until now no research has been reported on the association mapping of root and stolon traits in potato, due to the challenges in growing, excavating, and phenotyping underground traits. The current study hypothesized that tuber crops have measurable biological variation in root and stolon phenotyping and thus may be utilized to identify genomic regions associated with these variations. In this study, we aimed to identify the SNPs and genes associated with 7 important root and stolon traits through an association mapping approach using a 25K SolCAP array in a diverse panel of 192 tetraploid potato genotypes from Germany and Turkey. A state-of-the-art aeroponic system was used for root and stolon phenotyping.

Materials and methods

Plant materials

In this experiment, an association mapping panel of 192 tetraploid potato genotypes was used. All the genotypes used in this study belong to Solanum tuberosum L. 91 (47.3%) genotypes were selected from the German breeding company (Norika GmbH) including 37 breeding lines and 54 commercial varieties, 86 (44.7%) genotypes were obtained from the Turkish breeding company (Doğa Seeds) consisting of 83 breeding lines and 3 commercial varieties, while 15 (7.81%) common genotypes were also included, grown in both Turkey and Germany. Most of the varieties and breeding lines were developed for processing quality traits. This panel was chosen to evaluate the underground traits of potato. Detailed information on these genotypes is provided in the supplementary dataset (Table S1).

Phenotyping experiment

The experiments were conducted in an aeroponic system to derive the phenotyping data for the underground traits (roots and stolon’s) during summer 2019 and winter 2020, in an automated environmentally semi-controlled greenhouse system at the Faculty of Agricultural Sciences and Technologies (FAST) Niğde Omer Halisdemir University Niğde, Turkey (37°56′32″ N, 34°37′25″ E, 1229 m elevation). The aeroponic system was electronically controlled and constructed with a 10-mm-thick polycarbonate sheet, mounted on a galvanized metal frame and covered with black polyethylene inside, and a Styrofoam (5 cm thick) used on the top as a production bed. The dimension was 1000 × 100 × 80 cm (length × width × depth) of each aeroponics production unit.

Potato sprouts were used as explants. The sprouts of 192 genotypes were transplanted on the aeroponic production beds with 20 and 10 cm distance between the rows and plants, respectively, in a completely randomized design (CRD). Seven replicates of each genotype were planted. The range of relative humidity in the greenhouse was 55–60 and 25–45% in the years 2019 and 2020, respectively. Minimum, maximum, and mean temperatures (°C) during the experiments are given in Fig. 1. In the summer season, fans and cooling pads were used for cooling and also a shade cloth that provides a cooling effect on the greenhouse roof that can be automatically opened and closed. In the winter season, a moderate heating system was used to avoid chilling effects, controlled by a central operating system. Plant roots were misted for 20 s at every 6 min with an improved and modified solution mentioned in Table 1. The pH and EC of the solution were maintained as 6.2–6.8 and 2.0 mS cm⁻¹, respectively. The nutrient solution was changed weekly. Standard potato production practices were followed during the growing period. Lahlou and Ledent (2005) reported that potato roots have attained maximum depth after 80 days of planting. Therefore, a standard 85 days of the growing period was opted to collect roots and stolons samples in both years.

Fig. 1.

Minimum, maximum and mean temperatures (°C) during the year 2019–2020

Table 1.

Nutrient solution used in aeroponic system

Nutrient	Per liter	%
NPK	0.540 g	18 + 18 + 18
KNO3	0.122 g	13.5%K + 45.5%K
K2SO4	0.205 g	51%K + 18%S
Ca (NO3)2	0.487 g	26.3%Ca + 15.5%N
MgSO4	0.281 g	9.1%Mg + 14%S

Phenotypic data acquisition and statistical analysis

Root Traits such as total root length (TRL), root volume (RV), root surface area (RSA), and root fresh weight (RFW), and stolon Traits such as stolon length (SL), the total number of stolon (TNS) and branching of stolon (BS) were investigated in the study. Five samples per genotype were taken for the examination of root and stolon traits. The phenotyping descriptors of the aforementioned traits were explained in Table 2. The entire root system was scanned using scanner XL-11000 and analyzed using WinRHIZO software (Arsenault et al. 1995). One-way analysis of variance (ANOVA) based on a completely randomized design was performed on the data of both years, 2019 and 2020 using Statistix 8.1 software. The mean values of all traits were used to calculate marker-trait association using the “GWASpoly” R package. Pearson correlations (r) were performed on the traits under study, using SPSS software, while correlation scatter matrix was analyzed through R package “psych”. The broad-sense heritability (H²) for each trait was estimated by an equation: H² = V_G/V_P, using genotypic (V_G) and phenotypic (V_P) variance components (Bahmankar et al. 2014; Ogunniyan and Olakojo 2014).

Table 2.

Root and stolon trait names along with their phenotyping description

Trait name	Description of phenotyping
Total root length (TRL)	Cumulative length of all roots measured with the help of WinRHIZO software in centimeters
Root surface area (RSA)	Cumulative surface area of entire root system measured with the help of WinRHIZO software in square centimeter
Root volume (RV)	Cumulative root volume of all roots measured with the help of WinRHIZO software in cubic centimeter
Root fresh weight (RFW)	Fresh weight of the entire root system was taken immediately at harvest by calibrated precision scale
Total number of stolons (TNS)	Counted manually at the base of plant excluding the BS
Stolon length (SL)	Cumulative of entire stolon length (excluding BS) in centimeters measured with aid of measuring tape from the base of plant to the end point of stolon
Branching of stolon (BS)	Collection of all nodes on stolon representing the branching at axillary buds

Genotyping

The extraction of genomic DNA was performed by utilizing Gene JET Plant Genomic DNA Purification mini kit (Thermo Scientific) as per instructions given by suppliers. The quality and concentration of DNA were determined by electrophoresis on 1.2% agarose gels and further confirmed using the BioSpec-nano (Shimadzu) UV/Vis spectrophotometer. SNP genotyping was done for 192 potato genotypes, using the SolCAP 25K potato genotyping array (Hamilton et al. 2011; Uitdewilligen et al. 2013). Ilumina HiScan SQ system was used to read the array. The Genome Studio software was used for an assignment of genotypes to each locus. Tetraploid allelic combinations from nulliplex to quadriplex, with a total of diverse 21,226 markers were obtained after SolCAP 25K array analysis.

Population structure and principal component analyses

Population structure is a main pillar for association mapping and it was analyzed through STRUCTURE software which worked on the methodology of Bayes Bayesian clustering (Pritchard et al. 2000). The numeric file was made for population structure analysis according to the following format, for example, SNP T/C: TTTT = AAAA = 0; TTTC = AAAB = 1; TTCC = AABB = 2; TCCC = ABBB = 3; CCCC = BBBB = 4. After making a numeric file, marker filtering was done based on the calculation of P and Q frequency, minor allele-frequency (MAF) > 0.05, missing SNPs ≤ 10%. After filtering, 13,606 SNP markers were obtained. The Admixture-based clustering model was applied with STRUCTURE v.2.3.4 (Pritchard et al. 2000). Set the simulation like 100,000 burn-in and 100,000 iterations. For each K value, ten independent runs were performed, ranging from 1 to 10 with 5 replications each. The sub-population (K) was chosen based on Evanno et al. (2005) using STRUCTURE HARVESTER (Earl and vonHoldt 2012). To visualize the population structure, a bar plot was used based on Q sorting with an optimum ΔK value. Prcomp package in R v. 3.6.3 was used for principal component analysis (PCA) to visualize the genotype clustering based on sub-population. The function ggplot2 in R was used to plot the first two principal components (PCs).

Assessment of linkage disequilibrium

TASSEL was used to determine the LD between marker pairs based on D′ and r² (Hill and Robertson 1968; Achenbach et al. 2008). A permutation test was applied to check the LD significance level among the various loci.

Association analyses

Association analyses were done using the GWASpoly R package, using the “Q + K” linear mixed model approach for autopolyploids (Rosyara et al. 2016). GWASpoly has its unique features; i.e., it can analyze the diverse panel of tetraploids and work based on a different type of polyploid gene actions, including general, additive, simplex, and duplex dominant (Ref/Alt). Additive gene action was considered in the current study. GWASpoly needs rrBLUP for its proper and accurate functioning. Quantile–Quantile (q–q) plots were also drawn through this package. Q–Q plots tell us whether phenotypic data are logical or in other words, shows a normal distribution of the data around the mean. A threshold of − log10(P) = 5.4 was calculated following Bonferroni correction method (ɑ/m; where “ɑ” refers to P value = 0.05 and “m” corresponds to number of markers equals to 13,606) (Cockram et al. 2015). However, this threshold value is very strict and may lead to false negative marker-trait associations (possibility of avoiding/missing some important associations present). Instead, an arbitrary corrected significant threshold of − log10(P) was set at 3.23, similar to the approach used in previous association mapping studies (Riaz et al. 2018; Phan et al. 2018). Manhattan plots were envisioned utilizing the GWASpoly R package.

Results

Phenotypic data analysis

The descriptive statistics and ANOVA showed substantial variation and highly significant (P ≤ 0.001) differences for all traits (Tables 3 and 4). The traits of TRL, RSA, RV, RFW, TNS, SL, and BS were ranged from 2057.55 to 20,149.5 cm, 392.76–6432.55 cm², 5.71–175.84 cm³, 3.81–99.12 g, 1–9.5, 3.83–460 cm, and 1.56–255, respectively. Box plots showed data distribution of each trait analyzed through R 3.6.3 software (Fig. 2). Pearson correlation (r) analysis was done for all the traits to identify the relationship between them. Mean values were used for the correlation analysis of root and stolon traits. A significant correlation (r) was observed among all the traits. A significantly strong correlation was found between TRL and RSA (r = 0.915**), RSA with RV (r = 0.916**) and RFW (r = 0.870**). SL was also highly correlated with BS (r = 0.848**) (Table S4). The correlation scatter matrix unveiled the pictorial summary of Pearson correlation results along with a histogram depicting the normal data distribution pattern (Fig. 3). Relatively high heritability (H²) was observed, ranging from 0.80 for TNS to 0.95 for SL (Table 3). Higher H² values in our experiment showed the genetic contribution to the observed variability in the measured root and stolon traits. It also elucidates the presence of biological variation among the phenotypic traits.

Table 3.

Descriptive statistics and broad sense heritability (H²) estimate of phenotypic data

Trait	Mean ± SE	Min	Max	Genotypic variance	Phenotypic variance	Heritability (H2)
TRL	8871.87 ± 254.57	2057.56	20,149.50	47,298,068.00	49,565,796.75	0.93
RSA	2408.72 ± 73.93	392.76	6432.55	4,117,332.50	4,385,513.75	0.92
RV	55.99 ± 2.11	5.71	175.84	3642.94	3905.20	0.90
RFW	28.71 ± 1.04	3.81	99.12	873.28	932.27	0.93
TNS	3.36 ± 0.11	1.00	9.50	9.84	12.14	0.80
SL	86.87 ± 5.64	3.83	460.00	23,455.85	24,533.45	0.95
BS	25.15 ± 1.92	1.56	255.00	1714.07	1882.28	0.93

Table 4.

Summary of ANOVA following completely randomized design for the root and stolon traits of potato in the year 2019 and 2020

Traits	SOV	F value
		2019	2020
Total root length	Genotypes	34.2***	8.26***
Root surface area	Genotypes	19.9***	7.86***
Root volume	Genotypes	17.3***	6.26***
Root fresh weight	Genotypes	20.5***	9.67***
Total no. of stolon	Genotypes	3.49***	5.33***
Stolon length	Genotypes	68.9***	10.1***
Branching of stolon	Genotypes	28.0***	8.11***

SOV source of variation

(***) shows highly significant results at P < 0.001

Fig. 2.

Box plots depicting mean data distribution of potato root and stolon traits

Fig. 3.

Pairwise scatter plot matrix (lower boxes), histograms depicting data distribution (diagonal boxes) and Pearson correlation coefficients (r) (upper boxes) of all root and stolon variables of potato

Population structure and clustering

The delta K = 4 value obtained after structure analysis revealed that the genotypes were grouped into four sub-populations or clusters (Figs. 4a, 5). The SNP Hapmap numeric matrix with tetra-allelic SNP dosage was used for PCA analysis to assess the genetic variation among potato genotypes. PCA results grouped the genotypes into 4 clusters, which authenticates the results of structure analysis (Fig. 4b). PC1 and PC2 explained 13.10 and 11.59% variation in the genotypic data, respectively. The bar plot showed an estimated membership coefficient (Q) based on the Bayesian clustering approach, aligned in four colors: red, green, blue, and yellow (clusters 1, 2, 3, and 4, respectively, as shown in Fig. 4b. Individuals having Q values ≥ 0.40 grouped in the same cluster, while those depicting Q values < 0.40 grouped as admixed (Fig. 5, Table S1). The obtained four clusters showed diverse segregation and some admixed distribution of the genotypes. These results indicated that the population has a genetic diversity with different structural dimensions. It may be due to the reason that the crossing parents of the genotypes included in the study belong to an entirely different origin and might be originated from some wild-type species as well. Admixed genotypes were intermingled with each other and shared a relatively common area exhibiting considerably similar origin. Detailed information on genotypes distribution in 4 clusters and admixed sub-population is shown in supplementary dataset Table S2.

Fig. 4.

a Delta K values over 10 runs (ΔK = 4); b PCA scatter plot of the first two principal components (PC1 and PC2) depicting the clustering of genotypes into 4 clusters as indicated by delta K results from STRUCTURE

Fig. 5.

Bar plot displaying population sub-structure of 192 genotypes at a population size delta K = 4. Sub-population clusters 1, 2, 3 and 4 were represented by colors red, green, blue and yellow, respectively, according to the Q values (estimated membership covariates). X axis represents serial no. of genotypes corresponding to each individual bar plot as mentioned in Table S1, whereas y axis symbolizes Q values obtained from structure software

Linkage disequilibrium (LD)

LD was performed on robust 13,606 polymorphic SNP markers obtained by filtering monomorphic markers from a total set of 21,226 SNPs. LD was found for 13,386 SNP markers which were mapped on the potato genome. The average r² value was observed to be 0.29 with a percent significant LD of 49.98% in the population (Table 5). It indicates that 49.98% of markers in our population had a non-random association among various loci. A non-linear trend line gives an assumptive estimate of LD decay if plotted between variables r² and genetic distance (Mbps). A proposed marker dataset showed LD decay at around 2.316 Mbps to an r² threshold value of 0.29 (Fig. 6) (Remington et al. 2001; Whitt and Buckler 2003; Hamblin et al. 2004). A part of LD is preserved and found with linkage within 2 Mbps depicting no recombination (Gupta et al. 2005; Stich et al. 2005).

Table 5.

Linkage disequilibrium in the tetraploid potato (Solanum tuberosum L.)

Chromosome	No. of markers	Chr. size (Mbp)	SNP markers/Mbp	r2	Significant LD %	LD decay (Mbps)
1	1050	88.58	11.85	0.26	48.09	2.2
2	453	48.49	9.34	0.21	52.54	1.6
3	2178	62.62	34.98	0.3	84	0.9
4	1489	72.14	20.64	0.29	27.92	2.5
5	631	51.99	12.14	0.31	46.36	1
6	983	59.36	16.55	0.3	39.46	2.6
7	1170	56.63	20.66	0.26	43.59	2.8
8	3650	56.8	64.26	0.3	60.52	1.8
9	485	61.46	7.89	0.21	55.33	1.8
10	510	59.67	8.54	0.46	54.05	6
11	495	45.42	10.89	0.27	38	2.2
12	292	61.14	4.77	0.3	50	2.4
Linked markers	13,386	723.94	18.49	0.290.29	49.98	2.31

Fig. 6.

LD (Linkage disequilibrium) decay plotted between r² and genetic distance (Mbps) with non-linear trend line. Mean r² threshold value is 0.29 with LD decay at 2.316 Mbps

Association mapping study

The normal distribution of the data was analyzed through Q–Q plots and revealed normal distribution among all the traits. Q–Q plots further illustrate the false positives, family structure, and population stratification by a scale of − log10 (P) between observed and expected P values.

We performed an association mapping analysis for the root and stolon traits of potato using mean data of both years on a diverse 13,606 tetra-allelic SNP panel of 192 genotypes. Manhattan plots were obtained using additive gene action in the GWASpoly R package, showing the significance threshold for each locus along with the location of SNPs corresponding to each phenotypic trait (Fig. 7). Additive gene action was used with an assumption that the SNP effect is proportional to the minor allele dosage. Association analysis revealing significant SNPs to each trait was taken with a P value ≤ 0.00059. The dotted line in blue color showed − log10 (P) threshold of ≥ 3.23, which allows for significant SNP identification. A total of 50 significant SNPs was detected (Table S3), among these SNPs, nine were associated with root traits such as TRL, RSA, RV, and RFW. 41 SNPs were associated with stolon traits: viz., TNS, SL, and BS. Two SNPs were related to multiple traits in our study. The chromosome number and position of the identified SNPs were determined from the GWASpoly R package and re-confirmed from “potato genome assembly” and found to be the same. In our GWAS analysis, we found SNPs about PotVar (54.00%), SolCAP (42%) and Ry_sto_YesuSong_g_LG11 (4%). The PotVar SNPs were referred to as the SNPs originating from Uitdewilligen et al. (2013), while SolCAP SNPs were discovered by Hamilton et al. (2011).

Fig. 7.

a Manhattan and; b Q–Q plot for total root length (TRL) trait showing significant SNPs at − log10 (P) threshold of 3.23 (blue dotted line). Normal distribution was depicted by Q–Q plot

Total root length is an important trait for every crop especially crops having shallow-rooted like a potato. It is more closely related to water and nutrient uptake than other root measures. Two SNPs, solcap_snp_c2_53918 (Uncharacterized) and solcap_snp_c2_50821 (plausible functioned as Trihelix transcription factor GT-2) significantly associated with TRL, were identified on chromosome 12 (Fig. 7, Table S3). A strong correlation exists between total root length and root surface area. The higher value of these variables results in a less marked decrease in yield. One SNP, i.e., PotVar0037982 (protein kinase) was identified, significantly linked with RSA on chromosome 2 (Fig. S1, Table S3). The root volume is another critical trait, as it plays a significant role in the reproductive and vegetative growth of the plant. While analyzing the association of this trait, two SNPs, i.e., solcap_snp_c2_2400 (Probable Xaa-pro aminopeptidase P) and PotVar0008219 (Kinase) were found to be significantly associated with RV, located on chromosomes 1 and 11 (Fig. S2, Table S3). Root fresh weight defines the ability of the plant to absorb water. It also defines the total biomass of the root in the soil. We found four SNPs, i.e., solcap_snp_c2_43483 (Eukaryotic translation initiation factor 4E), solcap_snp_c2_29350 (PRA1 family protein D), solcap_snp_c2_20758 (Serine/threonine-protein phosphatase PP1) and solcap_snp_c2_51273 (Phototropin-1) significantly associated with RFW positioned on chromosomes 3, 9 and 11 (Fig. S3, Table S3).

The total number of stolons is one of the imperative traits influencing final tuber yield since swollen stolon referred to as tubers. TNS were examined for association analysis and five significant SNPs, i.e., solcap_snp_c2_43722 (serine/threonine-protein kinase SAPK2), PotVar0000812 (Conserved gene for unknown function), PotVar0017431 (Glutamate receptor 3 plant), PotVar0017902 (Uncharacterized aarF domain-containing protein kinase, chloroplastic) and PotVar0069668 (Sucrose synthase 2) were found to be associated with this trait on chromosomes 4 and 7 (Fig. S4, Table S3). Literature suggested that stolon length showed a positive correlation with number of tubers (Haverkort et al. 1990). In our experiment, we obtained 16 SNPs such as PotVar0000800 + PotVar0000460 (Conserved gene for unknown function), solcap_snp_c1_6750 (Mechanosensitive ion channel protein 3, chloroplastic), PotVar0123654 (DM_SUT4), solcap_snp_c2_25283 (Sucrose transporter 4 (SUT4), PotVar0087095 (Erg28), PotVar0087316 (Homeotic protein knotted-1), PotVar0111414 + PotVar0111512 (Adenine phosphoribosyl transferase), solcap_snp_c2_52195 (Transcription factor MYB44/Sucrose-responsive element-binding factor), PotVar0075291 (WRKY transcription factor 2), PotVar0069782 (Sucrose synthase 2), PotVar0114434 (OTU-like cysteine protease family protein), PotVar0051603 (Homocysteine s-methyltransferase), Ry_sto_YesuSong_g_LG11 (PVY-resistance genomic region), and PotVar0064663 (UTP—glucose-1-phosphate uridylyl transferase) significantly associated with SL on chromosomes 4, 7, 9 and 11 (Fig. S5, Table S3). Branching of stolon plays a significant role in the tuber yield, providing sites for tuber formation. We identified 20 SNPs significantly associated with BS on chromosomes 1, 6, 7, 9, 11 and 12. The name of SNPs along with their putative functions are annotated as PotVar0032873 (Geranylgeranyl hydrogenase), solcap_snp_c2_9724 (Phosphatidylinositol 4-phosphate 5-kinase 8), PotVar0040983 + PotVar0040971 (CPN60A), PotVar0056956 (Histone H2B), solcap_snp_c2_55552 (Probable protein phosphatase 2C 24), solcap_snp_c2_50795 (peroxidase 47), solcap_snp_c2_50783 (Serine/arginine-rich splicing factor SR45a), solcap_snp_c2_50263 (transcription factor DIVARICATA), PotVar0012596 (Conserved gene for unknown function), PotVar0102857 (Stromal membrane-associated protein), solcap_snp_c1_5446 (uncharacterized), solcap_snp_c1_1000 (Protein DEFECTIVE IN EXINE FORMATION 1), PotVar0114434 (OTU-like cysteine protease family protein), solcap_snp_c1_988 (carboxyl transferase), PotVar0051241 (MRNA binding protein), Ry_sto_YesuSong_g_LG11 (PVY resistance genomic region), PotVar0097908 (Conserved gene for unknown function), solcap_snp_c2_24534 (malate dehydrogenase, mitochondrial), and solcap_snp_c2_24446 (thioredoxin-1) (Fig. S6, Table S3). The gene ID, SNP variant (Ref/Alt), chromosome number and position of each identified significant SNP associated with the various root and stolon traits along with their respective scores (− log10(P) values ≥ 3.23; P ≤ 0.00059) and effect is given in supplementary dataset (Table S3). The identified SNPs have a diverse gene function related to root signalling, transcriptional and post transcription gene regulation, transporter gene families and PVY resistance.

The association of two SNPs with multiple traits might be the result of overlapping traits. PotVar0114434 on chromosome (chr) 9 was found to be common in SL and BS. This identified SNP has a putative functional role as an OTU-like cysteine protease family protein. SL and BS shared another common SNP named Ry_sto_YesuSong_g_LG11 on chromosome 11 with a putative function as PVY resistance. A physical map based on chromosome and physical base pair distance between the SNP markers associated with the studied root and stolon traits is depicted in Fig. 8. The narrow region of 0.03 Mbp (0.506–0.544 Mbp) on chr 9 (having a pleiotropic SNP) was found to be associated with BS and SL. Interestingly, the same chromosome comprises of 13.88 Mbp region (47.58–61.46 Mbp) associated with TNS, SL and RFW. A 3.79 Mbp region on chr 6 from 55.57 to 59.36 Mbp was associated with BS. A region spanning 8.13 Mbp (64.01–72.14 Mbp) located on chr 4 was linked with TNS and SL, which indicates the presence of possible QTL in this particular genomic region controlling stolon characteristic of potato. A physical distance of 3.97 Mbp (41.45–45.42 Mbp) was present on chr 11 linked to RV and RFW (Fig. 8). This physical map gave useful information about the localization of SNP markers along with physical distances between them to understand the recombination points and possible identification of QTL region associated with root and stolon traits for further utilization in MAS studies after validation.

Fig. 8.

Physical map of chromosomes and physical base pair distance (bp) displaying the significantly associated markers to the traits under study. SNP markers associated with the specific trait are portraying in rectangles with specific color. Pleiotropic SNPs are indicated with *. TRL total root length (yellow), RSA root surface area (black), RV root volume (brown), RFW root fresh weight (green), TNS total number of stolons (grey), SL stolon length (red), BS branching of stolons (purple). The map was drawn with the aid of R package “LinkageMapView”

Discussion

Understanding of genetics underlying stolon and root traits may help in the improvement of potato tuber yield under optimum and adverse environmental conditions such as drought and heat. Our study is the first comprehensive report on the association analysis of root and stolon traits in potato using a diverse 25K SolCAP array. We used 192 potato genotypes from different countries to identify the significant SNPs associated with TRL, RSA, RV, RFW, TNS, SL, and BS. A considerable variation and highly significant correlation were found between the traits that show large phenotypic diversity among the genotypes. Previous studies showed the same high correlation among root traits in the potato genotypes (Zarzyńska et al. 2017). The phenotypic measurements were taken for 2 years and a large variation was observed between all traits in each year. Precise phenotyping is very important for association studies to avoid false positives. Generally, under field conditions, root and stolon can be greatly influenced by uncontrolled environmental factors and any non-uniformity in the medium will bring about the changes in these traits. Many studies showed the relationship between root and stolon traits and crop productivity under fluctuating environments (Fita et al. 2006; Schiavon et al. 2016). An efficient phenotyping platform is inevitable for quantitative genetic studies; however, the non-availability of robust root phenotyping platforms is a major lag in the utilization of root and stolon-related genetic data for breeding in huge mapping population (de Dorlodot et al. 2007; Chen et al. 2011; Trachsel et al. 2013). Currently, many new approaches have been developed for better evaluation of root traits in a field, for example, “Shovelomics”. But still, some hurdles are present which includes, high cost of non-invasive imaging, delay in data acquisition, and the limitations of detailed data on underground trait from “Shovelomics” (de Dorlodot et al. 2007; Trachsel et al. 2011; Smith and De Smet 2012). Root and stolon are very sensitive and can be damaged during removal from the soil. It is necessary to use fast and ideal methods for evaluation. Alternatively, rapid, non-destructive indoor growing systems such as aeroponics have been developed in the recent past for accurate and precise measurement of root and stolon characteristics of potato (Kumar et al. 2015). Aeroponic provides a comparatively sophisticated platform to study underground traits as an alternative to the field. Thus, it enhances the accessibility and high-throughput phenotyping of these critical underground traits in potato. Keeping this in mind, the aeroponic system was used to grow plants in the current study. We can observe the root phenology, perform non-destructive sampling, and rapidly evaluate root and stolon characteristics under aeroponics. There is a research gap related to the underground trait-marker association in the potato under the aeroponic platform. The current study is thus novel in this aspect. Recently, the genetics of wheat root architecture was investigated in an aeroponic system (Thaon 2018). Root phenotyping approaches also require efficient imaging techniques to obtain the best phenotyping results. There is about 30 imaging software available for root phenotyping analysis (French et al. 2009). Among these software’s, WinRHIZO is considered as more flexible and reliable software to perform many specialist tasks (Pierret et al. 2013). In our study, roots were scanned by scanner XL-11000 and then analyzed using WinRHIZO software (Arsenault et al. 1995). Wishart et al. (2013) also used WinRHIZO software for the analysis of root traits in potatoes.

We present the first application of SNP markers in diverse potato genotypes for finding the novel associations between molecular markers and some traits of root and stolon in potato. Till now, association studies have been done in potatoes for different traits such as disease resistance (Malosetti et al. 2007; Mosquera et al. 2016; Charlotte et al. 2020) and tuber quality traits (Björn et al. 2008; Byrne et al. 2020; Khlestkin et al. 2020), but a research gap is present for root and stolon traits.

Potato is a highly heterozygous crop and SNP calling is more difficult in highly heterozygous species as compared to other inbred lines (Gardner et al. 2014; Hyma et al. 2015). SolCAP protocol was implemented to get thousands of SNPs. A total of 21,226 SNPs was obtained, after filtering, the number of SNPs in our study was higher in comparison to other SolCAP studies in potato (Zia et al. 2020). SolCAP was first implemented to study the LD and population structure in the European potato germplasm comprising of 8 diploid and 36 various tetraploid clones and cultivars (Stich et al. 2013). In the near past, Berdugo-Cely et al. (2017) evaluated 809 various Colombian species for association analysis of several plant traits in potato.

To eliminate spurious marker-trait association, assessment of genetic relatedness and population structure is of great importance (Yu and Buckler 2006; Zhu et al. 2008). After structure analysis, our population was divided into four sub-populations. PCA results authenticate the results obtained from the structure software (Fig. 4). The study conducted by Zia et al. (2020) also reported high genetic variation among potato genotypes indicative by the clustering approach, similar to our findings. LD among markers has got a vital role in association mapping studies because it provides information related to the resolution and strength of mapping (Flint‐Garcia et al. 2005). D′ and r², were assessed for each pairwise combination of SNPs. 49.98% of significant LD was observed. 221 tetraploid potato genotypes were genotyped with AFLP markers and an LD decay of 5 cM was observed with an r² threshold of 0.1 (Björn et al. 2008), while 10 cM LD decay was reported by Simko et al. (2006). This distance is equal to the 2 and 4 Mbp genetic distance (Vos et al. 2017). LD in our study (2.316 Mbp) was within the stipulated range as mentioned in previous studies. Some studies, however, report slow LD decay at 275 bp (Stich et al. 2013), but it depends upon the non-linear regression trendline in combination with the choice of r² threshold. The LD estimation is proportional to the frequency of large haplotype blocks in the population (Björn et al. 2010). In association mapping, LD patterns and haplotype blocks affect the identification of actual SNP variants. Zia et al. (2020) obtained significant LD of 47.4% in a diverse panel of 237 potato genotypes, showing an LD decay of 1.22 Mbp against an r² value of 0.2. The most important factor that affects LD is the mating system of the species (Nordborg et al. 2002). These findings imply the presence of abundant genetic variation suitable to find marker-trait associations.

In our study, we found 50 novel genomic regions associated with root traits (TRL, RSA, RV, and RFW) and stolon traits (TNS, SL, and BS). Some of the identified SNPs were linked with more than one trait. SNPs associated with stolon traits were located on chr 4, chr 6, chr 7, chr 9, chr 11 and chr 12, while those linked to root traits were positioned on chr 1, chr 2, chr 3, chr 9, chr 11, and chr 12. This localization of SNPs on several chromosomes might be due to the polygenic nature of our traits, influenced by minor genes.

Gene annotation of the identified SNPs depicted that few genes are conserved with unknown functions while others perform many vital putative functions (Table S3). SNPs in the genes of trihelix transcription factor GT-2, protein kinases, and protein phosphatases were found to be associated with roots in the present study (Table S3). GT-2 factors are members of trihelix transcription factors, which perform a major role under abiotic stresses. Overexpression of GmGT-2B and FaGT-2 like genes in roots of soybean and strawberry improved tolerance against salt, cold, and drought stress (Xie et al. 2009; Feng et al. 2019). The association of GT-2 factors with roots might be critical in stress signal transduction pathways, as it activates the transcription of stress-related plant hormones such as abscisic acid and salicylic acid (Song et al. 2016). Calcium-dependent protein kinases were found to be abundant in the early elongating stolons and roots of Solanum tuberosum (Grandellis et al. 2012). Similar findings have been reported in the current study, for instance, gene annotation showed an association of SAPK2 (stress-activated protein kinase-2) with the total number of stolons. In Arabidopsis thaliana, protein phosphatases of serine/threonine class enhanced sensitivity to osmotic stress (País et al. 2009) and regulate the transport of auxins playing critical roles in root elongation (Blakeslee et al. 2008). Roots and stolons are the primary sites of stress perception and signaling (País et al. 2009). In potato, StPP1 and StPP2 genes (belonging to serine/threonine protein phosphatase family) are involved in defense response against salt stress. Genes found in both protein kinase and phosphatase families are responsible for adaptive plant response in stress conditions. Plant glutamate receptor-like genes (GLRs) are involved in bringing about explicit changes in root system architecture such as root branching, growth and meristem proliferation (Forde 2014). Sucrose transporters (StSUT1) and sucrose synthase genes (Sus3 and SuSy) are involved in the transport and storage of sucrose in sink organs such as potato tuber and sugar beet (Fu and Park 1995; Kühn et al. 2003; Kühn and Grof 2010; Stein and Granot 2019). This might be the reason that SNPs linked to stolon traits possess the genes with plausible functions as sucrose transporter 4, sucrose synthase 2, and sucrose responsive element-binding factor (Table S3). Stolon traits like stolon length and branching of stolons were associated with SNPs having putative gene function in pathways involving the conversion of sucrose to starch in potato tuber. These include mitochondrial malate dehydrogenase and UTP-glucose-1-phosphate uridylyltransferase (UGPase) (Szecowka et al. 2012). The mRNA-binding proteins are involved in transcriptional and post-transcriptional regulations. In potato, StUBA2a/b and StUBA2c are found to be expressed in roots (Na et al. 2015).

The SNP PotVar0114434 was referred to as pleiotropic since it was associated with SL and BS with gene annotation of OTU-like cysteine protease family protein (Table S3). These proteins perform a major role in nutrients signal transduction in roots. An increase in protease activity enables tubers to respond to the nutritional needs of developing plants (Kohli et al. 2012). Another SNP named as Ry_sto_YesuSong_g_LG11 in our study has a pleiotropic effect due to its association with SL and BS. Ry sto is a gene that regulates resistance to PVY in potato (Song and Schwarzfischer 2008) and localized on chromosome 11, with a putative functional role in PVY resistance in potato. Results of association mapping were affected by an increased number of genotypes, years, location/environment. In future, the identified root and stolon traits associated SNPs in the present study will be validated in an independent population using KASP (Kompetitive allele-specific PCR) markers. It will be a ground-breaking approach for the development of selection molecular markers related to underground traits for marker-assisted breeding of root and stolon traits in potato.

Conclusion

The main purpose of the current study was to find SNPs associated with seven root and stolon traits in diversified tetraploid potato panel using an association mapping technique. We observed 50 novel genomic regions that could be potentially utilized in future potato-breeding programs. Our results showed that two SNPs were associated with TRL, one with RSA, two SNPs were related to RV, while four were linked to RFW. Five SNPs were associated with TNS, sixteen SNPs were found related to SL, and twenty were linked with BS. Furthermore, out of these obtained SNPs, two of them were found to be associated with multiple traits (SL and BS). The identified SNPs had a putative gene function in various plant signalling pathways, signal transduction, plant defence response, cell differentiation/proliferation, and PVY resistance. All the genomic regions identified in our study were novel, they can provide a new way to proceed for potato breeders. After validation with a genotyping technology like Kompetitive allele-specific PCR (KASP) markers these can be used in future potato-breeding programs to accelerate the breeding process through marker-assisted selection (MAS) as well as to cope with the needs of the increasing population.

Supplementary information

Below is the link to the electronic supplementary material.

Author contributions

MEÇ: supervised and designed the study, MEÇ and UD: funding acquisition, MFY: set an experiment and phenotypic data acquisition, MFY and MN: phenotyping and genotyping data curation and analysis, UD: formal computational analysis, MFY: drafted the manuscript, MEÇ, UD, MFY and MN review, editing and approved the manuscript.

Declarations

Conflict of interest

The authors declare that they have no conflict of interests.