Genotype × environment interactions for potato yield and quality traits: Identification of ideotypes adapted in different ecological regions of Northwest China

Abstract

Background

As a multi-use cash crop, the yield and quality traits of potatoes are often affected by genotype × environment interactions (GEI). Understanding the influence of GEI on potato yield and quality across varying environmental conditions is crucial for selecting excellent varieties suitable for specific ecological regions.

Methods

Ten advanced potato lines and twenty-two cultivars were assessed at three pilot sites in Gansu Province, China over two successive years (2020–2021). Tuber yield, dry matter, starch, reducing sugar, carotenoid, vitamin C, pyridoxamine, thiamine, nicotinic acid, pyridoxal and pyridoxine were analyzed.

Results

An analysis of variance (ANOVA) showed significant effects of GEI on all traits, with the greatest effects observed for vitamin C, reducing sugars, and carotenoids. AMMI, the additive main effects and multiplicative interaction (AMMI), and the genotype main effect plus genotype × environment interaction (GGE) biplot analysis revealed that G25 exhibited high yield, high dry matter content, and high carotenoid content in Weiyuan County. In Anding District, G18 showed high yield, high thiamine content, and low reducing sugar content. Meanwhile, G7 demonstrated superior performance with high yield, high carotenoid and thiamine contents, along with high pyridoxine content in Yongchang County. Genotype recommendations based on the mean values and stability of a single trait are partial and prejudiced, while selections based on multiple traits are desirable. Five ideal genotypes (G23, G26, G25, G24, and G16) were selected by multi-trait stability index (MTSI) in consideration of both the mean values and stability of yield and key quality traits.

Conclusions

Multi-environment trials revealed that GEI significantly affected vitamin C, reducing sugars, and carotenoids in potato. Five high-yielding and high-quality potato genotypes adapted to different ecological regions of Gansu Province were identified using MTSI analysis. This study provides references for the regional cultivation and quality breeding of potatoes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-025-06741-1.

Background

Potato (Solanum tuberosum L.) is the most important food crop worldwide and ranks fourth in global production quantity. Potatoes have been extensively cultivated globally due to their high yield and abundance of essential nutrients required by the human body. Although high yield is the most critical trait of potatoes, quality also plays an important role in regarding their edibility and processing characteristics. In potato production practice, quality is a multifaceted trait heavily depending on the intended use of the final product [1, 2]. During steam cooking, the starch content is the most critical factor for potatoes in that steaming quality is positively correlated with dry matter content [3]. Dry matter is a critical quality criterion for frying (e.g., French fries and potato chips), because a high dry matter content in tubers ensures a low oil absorption rate [1]. Similarly, in frying processes, reducing sugar content is an important quality indicator, as it reacts with aspartic acid during the Maillard reaction, affecting the taste, color, and aroma of fried products, as well as the formation of neurotoxic carcinogenic acrylamide [4]. This poses a significant potential risk to human health. In addition, potatoes are recognized as an important source of antioxidants, including ascorbic acid and carotenoids [5]. Ascorbic acid acts as a key inhibitor of both enzymatic and non-enzymatic reactions during cooking, while carotenoids are recognized as the primary lipophilic components contributing to the total antioxidant activity of potatoes [6]. Potatoes are also considered rich in essential vitamins, with statistics indicating that a single potato tuber weighing 150 g can provide up to 8% of the recommended daily intake for niacin, 10% for vitamin B6, and 6% for folic acid [7].

Yield and quality traits of potato tubers are primarily quantitative traits influenced by environmental factors closely linked to genotype × environment interactions (GEI) [2, 8–10]. It is recognized that different traits are influenced by GEI to varying degrees. In many studies, multi-environment trials (METs) are often conducted to better understand the performance of genotypes across diverse environments by assessing their stability [11]. Several statistical methods have been employed for stability analysis in METs, including Analysis of Variance (ANOVA), linear regression (LR), and principal component analysis (PCA) [12]. Currently, the Finlay-Wilkinson (FW) regression [13, 14], the additive main effects and multiplicative interaction (AMMI) model [15, 16], and the genotype main effect plus genotype × environment interaction (GGE) biplot are frequently utilized to analyze GEI in METs [17, 18]. These models are based on single traits to evaluate genotypes; however, the merit of a genotype depends on its comprehensive performance across multiple traits. Furthermore, the high-quality genotypes identified using the same model based on different traits can vary. Conducting a comprehensive analysis using multiple models could mitigate potential errors associated with individual models and enhance understanding of genotype characteristics. The multi-trait stability index (MTSI) analysis facilitates the uniform quantification of multiple traits to assess genotypic excellence or inferiority [19], thereby preventing errors in identifying single traits and discrepancies in genotypes recommended for various traits. MTSI could be beneficial for breeders, as it integrates the mean values of multiple traits with stability, offering a straightforward selection process that considers the correlation structure among traits [20].

To overcome the constraints of potato production and quality, it is essential to evaluate the performance of different genotypes under specific cultivation conditions and climates to assess their adaptability and stability in terms of yield and quality traits. However, in this process, emphasis is placed on the agronomic excellence of the new cultivars compared to existing ones, particularly concerning potato yield and quality [21]. Little emphasis is paid on the interaction between the cultivars and the target environments, which is often unpredictable. Therefore, this study aimed to (1) explore the influence of GEI on the main traits of potato using ANOVA, the phenotypic coefficient of variation (PCV), the genotypic coefficient of variation (GCV), and broad-sense heritability (H²); (2) quantify the GEI effect using FW regression, the AMMI model and the GGE biplot; (3) identify ideal potato genotypes with high-yielding and high-quality suitable for cultivation in the three ecological regions of Northwest China.

Materials and methods

Plant materials and experiment sites

Ten advanced potato lines and twenty-two main cultivars (Table 1) from Gansu Agricultural University, Gansu province, China were used as experimental materials and were grown in replicated field trials at three pilot sites over two successive years (2020–2021) (Fig. 1). The selected sites represented contrasting environmental conditions across Weiyuan County (a chilling and humid region), Anding District (a semi-arid and arid region), and Yongchang County (an arid region) in Gansu Province, Northwest China (Fig. 1 and Table 2). A matrix of five site-year environments was created in this multi–environment trial: Weiyuan County 2020 (E1), Weiyuan County 2021 (E2), Anding District 2020 (E3), Anding District 2021 (E4), Yongchang County 2020 (E5), Yongchang County 2021 (E6) (Table 2). The soil characteristics (0–30 cm depth) at three the experimental sites are presented in Table S1. Weather data were collected from a weather station located near the experimental site. The monthly mean temperature, monthly rainfall and monthly sunshine du-ration are shown in Fig. S1.

Table 1.

Agronomic traits and quality characteristics of different potato genotypes

Genotypes	Genotypes entry	Commodity rate (%)	Tuber shape	Skin color	Flesh color	Eye depth
1402–1	G1	80.13 ± 14.01 d	Long ellipse	Light yellow	Light yellow	Shallow
1412–1	G2	97.51 ± 1.75a	Ellipse	Yellow	Yellow	Shallow
1416–5	G3	92.21 ± 4.41b	Ellipse	Light yellow	Milky white	Shallow
1422–1–12	G4	94.55 ± 5.44ab	Ellipse	Light yellow	Light yellow	Shallow
1423–1–8	G5	84.02 ± 14.00cd	Ellipse	Light yellow	Milky white	Shallow
1423–1–20	G6	97.43 ± 4.15a	Long ellipse	Light yellow	Milky white	Shallow
1425–1–13	G7	94.14 ± 9.55ab	Long ellipse	Light yellow	White	Shallow
1428–1–26	G8	89.73 ± 10.13bc	Ellipse	Light yellow	Light yellow	Shallow
1428–1–27	G9	96.00 ± 4.65a	Long ellipse	Light yellow	Light yellow	Shallow
1428–1–31	G10	92.00 ± 5.18b	Long ellipse	Light yellow	White	Shallow
Burbank	G11	82.45 ± 5.57cd	Long ellipse	Light yellow	White	Shallow
Innovator	G12	87.9 ± 8.23bc	Long ellipse	Brown	Light yellow	Shallow
Shepody	G13	84.65 ± 4.22cd	Long ellipse	White	White	Shallow
Beifang 002	G14	91.42 ± 7.12b	Long ellipse	Light yellow	Yellow	Shallow
Beifang 106	G15	87.40 ± 7.05bc	Long ellipse	Light yellow	White	Shallow
Dinshu 4	G16	88.80 ± 6.45bc	Ellipse	Light yellow	Yellow	Medium
Dongnong 310	G17	80.25 ± 8.63 d	Ellipse	Light yellow	White	Shallow
Gannongshu 7	G18	94.59 ± 4.56ab	Ellipse	Light yellow	Light yellow	Shallow
Gannongshu 9	G19	95.41 ± 3.45ab	Circle	Light yellow	Light yellow	Shallow
Lishu 13	G20	93.26 ± 4.56ab	Long ellipse	White	White	Shallow
Longshu 12	G21	83.23 ± 10.45cd	Long ellipse	Light yellow	Light yellow	Shallow
Longshu 4	G22	87.12 ± 9.87c	Long ellipse	Yellow	Yellow	Shallow
Longshu 10	G23	85.65 ± 8.56c	Ellipse	Yellow	Yellow	Shallow
Longshu 16	G24	84.75 ± 7.86cd	Ellipse	Light yellow	Light yellow	Shallow
Longshu 7	G25	80.70 ± 11.45 d	Ellipse	Yellow	Yellow	Shallow
Minshu 4	G26	93.37 ± 4.25ab	Circle	Red	Yellow	Shallow
Qingshu 10	G27	88.63 ± 5.22bc	Circle	White	White	Shallow
Tianshu 12	G28	87.40 ± 3.67bc	Circle	Light yellow	White	Shallow
Yunshu 901	G29	90.33 ± 4.75b	Ellipse	Light yellow	Light yellow	Shallow
Zhongshu 18	G30	80.45 ± 8.56 d	Long ellipse	Yellow	Yellow	Shallow
Zhongshu 22	G31	82.45 ± 7.68cd	Long ellipse	White	White	Shallow
Atlantic (Control)	G32	89.86 ± 5.78bc	Circle	Light yellow	Light yellow	Shallow

Fig. 1.

The location of the pilot sites

Table 2.

Agro-ecological characteristics of the three pilot sites

Location	Years (2020–2021)	Altitude (m)	Annual mean temperature (℃)	Annual rainfall (mm)	Annual sunshine duration (h)	Free frost period (d)
Weiyuan County	E1, E2	2240	5.3	550	2462.0	166
Anding District	E3, E4	1920	6.7	377	2116.0	141
Yongchang County	E5, E6	1954	4.8	185	2941.3	134

Experimental design and field management

The tubers were planted in Weiyuan County from April 25 to May 1 in both 2020 and 2021, with harvesting occurring from September 21 to 27. In Anding District, planting began from April 28 to May 1, followed by harvesting from September 20 to 25 in both years. In Yongchang County, the planting period extended from April 26 to 29, with harvesting from October 1 to 4 in both 2020 and 2021. Field experiments under three site–year environments were conducted using a randomized complete block design. Each potato genotype was planted in three replicate plots which consisting of a single ridge with two rows. The plot area was 1.0 m × 0.9 m with the row spacing 50 cm and the plant spacing 25 cm, and protected rows were planted around the test field to create a buffer. Before sowing, all experimental plots received a basal application of 900 kg ha⁻¹ of 12–25-10 (NPK) fertilizer. During the potato tuber bulking stage, an additional 900 kg ha⁻¹ of 18–0–25 (NPK) was topdressed. The fertilizers applied were potassium sulfate, di-ammonium phosphate, and urea. For water management, soil moisture was maintained at 60–70% of field capacity during the seedling stage and 75–80% during tuber initiation and bulking. Irrigation was applied immediately when soil moisture dropped below the target thresholds. Weed control combined mechanical inter-row cultivation with herbi-cide sprays (quizalofop-P-ethyl, clethodim, and haloxyfop-P-methyl). Potato early blight (Alternaria solani) and late blight (Phytophthora infestans) were managed with targeted fungicides upon symptom appearance.

After harvest, the tuber yield (TY), commodity rate (a single tuber weighs more than 70 g), tuber shape, skin color, flesh color, and eye depth were assessed. Mature tubers of similar size were selected after 20 days of cold storage at 4℃ in the dark. Some samples were directly measured for the contents of dry matter, reducing sugar, and vitamin C. The remaining samples were freeze-dried in an ultra-low temperature desiccator at −80℃ to determine the contents of starch, carotenoid, and B vitamins.

Determination of tuber quality traits

Dry matter content

The content of dry matter (DM) was determined according to Li et al. [22]. Three tubers were taken from each sample, washed and dried, then sliced and weighed (W1). The slices were placed into an electric blast drying oven (Taisite, 101-2 A, China), baked at 80℃ to a constant temperature, and weighed again (W2).

$$\mathrm{DM}=\frac{W1}{W2}\times 100\%$$ 1

Starch content

The starch content (ST) in the tubers was determined using the enzymolysis method [23]. A sample of 8–10 mg of freeze-dried powder was weighed, and then 1 mL of 85% ethanol was added. The sample was incubated in a water bath at 80 °C for 1 h, followed by centrifugation at 3000 rpm for 20 min. After discarding the supernatant, 1 mL of water was added, and the sample was gelatinized in a boil-ing water bath for 1 h. After cooling, 1 mL of α-amylase solution was added, and the sample was enzymolyzed in a water bath at 60 °C for 1 h. A volume of 100 μL of the test or blank control solution was mixed with 500 μL of anthrone and incubated in a boiling water bath for 10 min. Finally, the absorbance was measured at 620 nm.

Reducing sugar content

Reducing sugar content (RS) was extracted as described by Yuan et al. [24]. The fresh tubers 2.0 g were homogenized and incubated with 20 mL 80% (v/v) ethanol at 70℃ for 3 h. After centrifugated at 10,000 g for 20 min, the supernatant was collected, vacuum-dried, dissolved in deionized water and passed through a membrane filter (0.2-μm, Millipore). The concentrations of RS were determined by High performance liquid chromatography (HPLC) (Model 1100 series, Agilent Technologies) with an Amide-80 column (HW-40 F, TSKgel).

Carotenoid content

The content of carotenoid (CA) was analyzed according to the method described by Benmeziane et al. [25] with slight modifications. A sample of 300 mg of freeze-dried powder was weighed and placed in a 15 mL centrifuge tube and 10 mL carrot-like extract (ethanol: acetone = 1:1, containing 0.1% BHT) added. Ultrasonic extraction was performed at 30℃ for 30 min, followed by high-speed centrifugation at 4℃ and 8000 g for 15 min. Measure absorbance at 445 nm (extracted solution used as reference solution).

$$\mathrm{CA}=\frac{A\times V}{2500\times M}\times 10$$ 2

A is the absorbance at 445 nm, V is the total volume of the extract (mL), M is the mass (g) of the sample (g), and 2500 is the experimental light absorption coefficient of CA.

B vitamins contents

The contents of B vitamins were determined by HPLC according to Datta et al. [26]. A sample of 400 mg of freeze-dried powder was weighed and placed into a 2 mL brown centrifuge tube, a volume of 1.8 mL 0.01 mol·L⁻¹ metaphosphate (pH 3.1) was added and swirled until the freeze-dried powder was thoroughly mixed. The tubes were placed in a 40 prm shaker for 30 min, and an ultrasonic extraction was carried out for 30 min at 22℃, then the tubes were centrifuged at 15,000 rpm at 4 ℃ for 15 min. The supernatant was passed through a 0.22 μm microporous filter membrane and filtrate was stored at 4 ℃ for analysis.

Vitamin C content

Extraction of vitamin C content (VC): The extraction of VC was performed according to Yaman and Mızrak [27]. 80 ml of meta-phosphoric acid solution (3%) added to 5 g of fresh tubers was homogenized and mixed in a shaker for 10 min. Using meta- phosphoric acid solution (3%), set the volume to be 100 mL and the final solution was filtered using a CA filter (0.45 µm).

The VC content was determined by HPLC according to Klimczak and Gliszczyńska-Świgło [28]. The mobile phase was dissolving 1.24 g KH₂PO₄ in 1000 mL of distilled water, pH = 2.4. The C18 column (4.6 × 250.0 mm, 5 μm) (ACE, Scotland) was employed for separation of related compounds. The flow rate was 0.5 mL/min and the wavelength of the detector was 254 nm.

Determination of H², PCV, and GCV

Genotypic variance (${\mathrm{\sigma }}_{\text{G}}^2$), interaction variance between G and E (${\mathrm{\sigma }}_{\text{GE}}^2$) and error variance (${\mathrm{\sigma }}_{\text{E}}^2$) were calculated by SPSS. The phenotypic variance (${\mathrm{\sigma }}_{\text{PH}}^2$) was determined based on ${\mathrm{\sigma }}_{\text{G}}^2$, ${\mathrm{\sigma }}_{\text{GE}}^2$ and ${\mathrm{\sigma }}_{\text{E}}^2$. The formula is as follows [29]:

$${\sigma }_{\mathrm{G}}^2=\frac{{\mathrm{MS}}_{\mathrm{G}}-{\mathrm{MS}}_{\mathrm{GE}}}{\mathrm{re}}$$ 3

$${\sigma }_{\mathrm{GE}}^2=\frac{{\mathrm{MS}}_{\mathrm{GE}}-{\mathrm{MS}}_{\mathrm{E}}}{\mathrm{r}}$$ 4

$${\sigma }_{\mathrm{GE}}^2={\mathrm{MS}}_{\mathrm{E}}$$ 5

$${\mathrm{\sigma }}_{\text{PH}}^2={\mathrm{\sigma }}_{\text{G}}^2+{\mathrm{\sigma }}_{\text{GE}}^2+{\mathrm{\sigma }}_{\text{E}}^2$$ 6

where MS_G, MS_GE, and MS_E are the mean squares of G, GEI, and error; e and r represent the number of environments and replicates. The H² was determined as [29]:

$$H^2=\frac{{\sigma }_{\mathrm{G}}^2}{{\sigma }_{\mathrm{PH}}^2}$$ 7

$\overline{\text{x}}$ is the mean value of traits. The PCV and GCV were calculated as follows [30]:

$$\mathrm{PCV}\left(\%\right)=\left(\frac{\sqrt{{\sigma }_{\mathrm{PH}}^2}}{\overline{\mathrm{X}}}\right)\times 100$$ 8

$$\mathrm{GCV}\left(\%\right)=\left(\frac{\sqrt{{\sigma }_{\mathrm{G}}^2}}{\overline{\mathrm{X}}}\right)\times 100$$ 9

Statistical analysis

Best linear unbiased prediction (BLUP) analysis

The BLUP model is based on a linear mixture of fixed and random effects. The model can be expressed as follows [31]:

$$Y_{\mathit{ij}}=\mu +S_j+S{G}_{\mathit{ij}}+e_{\mathit{ij}}$$ 10

where Y_ij is the observed value of the genotype i in the environment j; μ is the population mean; S_j is the fixed effect of the environment j; SG_ij is the interaction between the environment j and genotype i; e_ij is the residual of genotype i in environment j which can be further decomposed into spatial autocorrelation error and measurement random error.

Genotype stability analysis

Genotype stability in different environments was assessed using FW regression, achieved primarily by regressing the specific performance of each genotype on the mean value of the environments. The regression coefficient for each genotype is directly related to stability. When the regression coefficient is close to 1, it indicates that the assessed genotype is stable across all environments. Deviations from 1 indicate higher or lower stability in specific environments [30]. The formula is described as follows:

$${\mu }_{\mathit{ij}}=\mu +G_i+E_j+b_i{E}_j+{\epsilon }_{\mathit{ij}}$$ 11

where μ_ij is the mean value of genotype i in environment j; μ is the population mean; G_i is the main effect of genotype i; E_j is the main effect of environment j; b_i E_j is the score value of the slope of genotype i in E_j; ε_ij is the residual of genotype i in environment j.

AMMI model

Both AMMI and GGE combine ANOVA for additive parameters and singular value decomposition (SVD) for multiplicative parameters, the principal components (PCs). According to Gauch et al. [15], the equation for the AMMI model is expressed as follows:

$$Y_{\mathit{ij}}=\mu +{\alpha }_i+{\beta }_j+\sum _{i=1}^N{\lambda }_n{\gamma }_{\mathit{in}}{\delta }_{\mathit{in}}+{\theta }_{\mathit{ij}}$$ 12

where Y_ij is the observed value of the genotype i in the environment j; μ is the population mean. The main effect of α_i genotype i. The main effect of β_j environmental j; λ_n is the singular value of the nth interaction principal component (IPC); γ_in is the genotypic IPC score of the nth IPC; δ_in is the environmental IPC score for the nth IPC; n is the total number of IPC factor; θ_ij is the residual of genotype i in environment j.

The AMMI stability value (ASV) can be calculated as follows [32]:

$$ASV=\sqrt{{((S{S}_{IPC1}/S{S}_{IPC2})\times IPC1)}^2+IPC{2}^2}$$ 13

where SS is the sum of squares. The principal component with the greatest variation is called IPC1 and the second is called IPC2. The lower the ASV value, the higher the stability. According to the ASV value, genotypes can only be ranked according to stability, and the mean value of genotypes is not considered. The genotype stability index (GSI), which combines mean value, takes into account mean value and stability. The GSI can be calculated as follows [32]:

$$GS{I}_i=RAS{V}_i+R{M}_i$$ 14

where RM_i is a ranking of genotypes i based on mean value; RASV_i is a ranking of genotypes i based on AMMI stability. The lower the GSI value, the better the mean value and stability of the genotype.

GGE biplot analysis

The GGE biplot is a graphical representation of a bidirectional data table. Constructing a GGE biplot involves two steps. The first step decomposes the data table into different PCs through SVD. The second step involves creating the GGE biplot based on the SVD results (i.e., PC1 and PC2). The GGE biplot generated by SVD based on environmental centralization data is the only biplot suitable for genotype and pilot evaluation [33]. The general formula can be expressed as follows:

$$Y_{\mathit{ij}}-\mu -{\beta }_j=\sum _{l=1}^k{\lambda }_l{\xi }_{\mathit{ij}}{\eta }_{\mathit{lj}}+{\epsilon }_{\mathit{ij}}$$ 15

where Y_ij is in the environment j genotype i yield or other characteristic values; μ is the population mean. The main effect of β_j environmental j; λ_l is the singular value of the lth PC; ξ_il is the characteristic vector of PCl; η_lj is the environmental feature vector of PCl; ε_ij is the residual of genotype i in environment j.

To generate GGE biplot for visual analysis, the singular value is decomposed into genotype feature vector and environment feature vector, the GGE model can be redefined as:

$$Y_{\mathit{ij}}-\mu -{\beta }_j=\sum _{l=1}^k{g}_{\mathit{ij}}{e}_{\mathit{lj}}+{\epsilon }_{\mathit{ij}}$$ 16

where g_il and e_lj are the values of genotype i and environment j on PCl; $g_{\mathit{il}}={\lambda }_l^{\mathit{fl}}{\xi }_{\mathit{il}}$, $e_{\mathit{lj}}={\lambda }_l^{1-f}{\eta }_{\mathit{lj}}$, f_l is the allocation factor of PCl.

MTSI analysis

MTSI is calculated based on the Euclidean distance between the genotype and the ideal genotype. It serves as a method to unify multiple traits into a quantifiable value within the framework of a linear mixed model [34]. The MTSI were calculated using the metan and Matrix packages in R. The MTSI acquisition process is as follows:

$$X=\mu +Lf+\epsilon$$ 17

where X represents the observed value of the p × 1 vector; μ is the population mean of the p × 1 vector; L stands for the factorial loadings of the p × f matrix; f is the common factor of the p × 1 vector; ε is the residual of the p × 1 vector; p and f denote the number of traits and common factors retained. The score was calculated as:

$$F=Z{(A^T{R}^{-1})}^T$$ 18

where F is the g × f matrix; g is the number of genotypes and f is the number of factors; Z is the g × p matrix with the standardized means; R is the p × p correlation matrix between variables; A is the p × f matrix of canonical loadings. Finally, MTSI was calculated as:

$${\mathit{MTSI}}_i={\left[\sum _{j=1}^f,{\left({F_i}_j,-,F_j\right)}^2\right]}^{0.5}$$ 19

F_ij is the jth value for genotype i; F_j is the jth value of the ideal genotype. The lower the MTSI value of the genotype, the closer it is to the ideal genotype for all target variables.

Data analysis

The data were obtained from three replicates and expressed as means ± SE. Data statistics and analysis were conducted using Microsoft Excel 2016, SPSS 22.0, and Origin 2021. Statistical analyses were performed using ANOVA and Duncan’s multiple range test, where p < 0.05 indicated significance. GenStat 21 st was employed for GEI analysis, while R version 4.1.0 was used for the calculation of BLUP values and MTSI.

Results

Effects of genotype, environment and GEI on yield and quality traits of potato tuber

PCA

The contribution rates of different traits to potato genotypes varied, prompting this study to conduct PCA on 11 traits across 32 potato genotypes. Uncorrelated PCs were utilized to better capture information regarding potato yield and quality and to quantify the influence of each trait. A simultaneous GEI analysis was conducted using the most significant traits influencing the specific performance of the genotypes, which greatly enhanced the accuracy of the analysis. The extracted eigenvalues for the first five PCs were all greater than 1, and the cumulative contribution rate reached 69.57%, indicating that these PCs encompassed the majority of the information relevant to the genotypes (Fig. 2). For the first principal component, the absolute values of the coefficients for DM, RS, VC, and thiamine content (VB1) were all less than 0.1, suggesting that the effects of these traits on the first principal component were negligible (Table 3). Similarly, variables with absolute coefficient values for the second, third, fourth, and fifth PCs were removed. The contribution rates of variance for the first five PCs were 22.46%, 15.31%, 12.02%, 10.57%, and 9.17%, respectively (Fig. 2b). Each of the five PCs selected 2, 2, 1, 1, and 1 primary indicators, with pyridoxine content (PN) and CA selected from the first principal component, VB1 and TY from the second, VC from the third, DM from the fourth, and RS from the fifth (Table 3).

Fig. 2.

Eigenvalues and percentage of explained variances of PCA

Table 3.

Eigenvectors corresponding to the principal components

Traits	Principal component
	1	2	3	4	5
TY	0.201	0.339	−0.110	−0.231	0.034
DM	0.071	−0.210	0.252	−0.434	0.127
ST	−0.140	0.226	0.375	−0.220	0.396
RS	−0.074	0.024	0.407	0.367	−0.600
CA	−0.265	0.204	−0.102	−0.095	0.047
VC	0.001	0.092	0.490	0.339	0.434
PM	0.206	0.215	0.242	−0.292	−0.363
VB1	0.064	0.412	−0.190	0.186	0.135
VB3	0.235	0.100	−0.099	0.416	0.136
PL	0.210	−0.325	−0.014	0.115	0.305
PN	0.344	0.023	0.139	−0.074	−0.066

TY tuber yield, DM dry matter content, ST starch content, RS reducing sugar content, CA carotenoid content, VC vitamin C content, PM pyridoxamine content, VB1 thiamine content, VB3 nicotinic acid content, PL pyridoxal content, PN pyridoxine content

ANOVA for potato tuber yield and quality traits

ANOVA was used to evaluate the primary traits affecting the expression of potato genotypes to assess the extent of GEI on these traits. The results of the ANOVA indicated that genotype significantly influenced all traits. Genotype was identified as the factor with the greatest contribution to variation, particularly for PN (70.01%), RS (66.10%), TY (53.10%), and CA (53.49%) (Table 4). The mean TY across all environments ranged from 23.90 t/hm² to 65.72 t/hm² (Table S2). Significant differences in TY values among different genotypes within the same environment were observed, particularly between L9 (75.50 t/hm²) and L21 (25.73 t/hm²) in Weiyuan County. Similar trends were also obtained for PN, RS, and CA (Table S2).

Table 4.

Combined ANOVA analysis of tuber yield and quality traits of thirty-two potato genotypes grown in three pilot sites

Source of variation	df	TY		DM		RS		CA
		SS	Per (%)	SS	Per (%)	SS	Per (%)	SS	Per (%)
G	11	1213.6***	53.10	205.0***	23.27	2.73***	66.10	13,887.2***	53.49
E	2	246.8***	10.80	315.0***	35.76	0.19***	4.60	1661.1**	6.40
GEI	22	638.7***	27.94	297.2***	33.74	1.04***	25.18	6253.4***	24.09
Block	6	29.4 ns	0.13	7.9 ns	0.90	0.02ns	0.48	810.5*	3.12
Error	66	157.1	6.87	55.8	6.33	0.15	3.63	3350.2	12.90
Total	107	2285.6		880.9		4.13		25,962.4
Source of variation	df	VC		VB1		PN
		SS	Per (%)	SS	Per (%)	SS	Per (%)
G	11	249.8***	37.62	78.0***	25.10	839.4***	70.01
E	2	52.0***	7.83	149.6***	48.13	5.4*	0.45
GEI	22	300.6***	45.26	76.5***	24.61	236.3***	19.71
Block	6	6.1 ns	0.92	0.8 ns	0.26	3.9 ns	0.33
Error	66	55.6	8.37	5.9	1.90	113.9	9.50
Total	107	664.1		310.8		1198.9

TY Tuber yield, DM Dry matter content, RS Reducing sugar content, CA Carotenoid content, VC Vitamin C content, VB1 Thiamine content, PN Pyridoxine content, df degree of freedom, SS The sum of square, Per Contribution percentage to the total, G Genotype, E Environment

*** Significant at P < 0.001

** Significant at P < 0.01

* Significant at P < 0.05; ns (not significant)

There was significant environmental variability observed for all traits, except PN. In DM (35.76%) and VB1 (48.13%), the percentage of total sum of squares attributed to environment was even higher than that of genotype. Consequently, the performance of the same genotypes in different environments. GEI significantly influenced all traits. Among the tuber quality traits, VC (45.26%) was the most affected by GEI. The effect of GEI on PN was the smallest at 19.71%, yet it was considerably higher than the environmental effects on the total sum of squares for TY (10.80%), VC (7.83%), CA (6.40%), RS (4.60%), and PN (0.45%).

Analysis of PCV, GCV, and H² for potato tuber yield and quality traits

The influence of GEI on yield and quality traits is not only reflected in the ANOVA, but also indicated by important indicators such as PCV, GCV, and H². The highest PCV values were observed for VB1 (71.35) and RS (50.05), indicating that the environment had a significant impact on these two traits (Table 5). Conversely, DM (10.28) exhibited the lowest PCV value, suggesting that genotype was the primary factor influencing DM (Table 5). The difference between PCV and GCV represented the range of influence of GEI on yield and quality traits. For instance, the difference between PCV and GCV values was greater for VC (30.07), and VB1 (25.95), indicating that these traits were considerably influenced by GEI. The H² was from 5.50% to 60.62%, with PN was more than 60%, suggesting that this trait was less affected by GEI. In contrast, DM (10.09%) was the most affected by GEI (H² = 5.50). Although the H² value of VC (16.43%) was higher than that of DM (10.09%), both remained relatively low. Therefore, considering ANOVA, H², PCV, and GCV, it can be concluded that VC, VB1, and DM were the most significantly affected by GEI.

Table 5.

Analysis of coefficient of variation and H²

Traits	PCV	GCV	H2 (%)
TY	31.18	20.80	44.50
DM	10.28	3.26	10.09
RS	50.05	30.62	27.79
CA	20.74	14.04	45.81
VC	36.93	6.86	16.43
VB1	71.53	45.58	24.74
PN	32.43	25.25	60.62

TY Tuber yield, DM Dry matter content, RS Reducing sugar content, CA Carotenoid content, VC Vitamin C content, VB1 Thiamine content, PN Pyridoxine content, PCV Phenotypic coefficient of variation, GCV Genotypic coefficient of variation, H² Broad sense heritability

Genotype selection in consideration of mean value and stability of single trait

BLUP values

Previous studies have shown that investigating genotype stability using BLUP values for each trait instead of mean values significantly improves the accuracy of the analysis [35]. The BLUP values of yield and quality traits among different genotypes in this study were obtained using the lem4 package in R, and specific values are presented in Table S3, which provides detailed data for each trait. In the following sections, GEI was quantified, and single-trait excellent genotypes were recommended based on the BLUP values.

FW regression analysis

FW Regression analysis could only obtain linear regression coefficients (i.e., the sensitivity shown in Fig. 3). Genotype performance was visualized by the mean values of yield and quality traits, along with sensitivity. The results indicated that G19 exhibited the highest stability in TY, followed by G8, G9, and G20. However, the mean TY value of G8 was lower and did not provide an advantage in TY. The genotypes showing both high and stable TY were limited to G9, G19 and G20 (Fig. 2a). Among the genotypes, G7 was identified as one of the most susceptible varieties to environmental changes, including TY, and RS, with sensitivities of 1.02, and 6.99, lower than that of the control, respectively (Fig. 3a and c).

Fig. 3.

The FW regression for potato yield and quality traits

Regarding quality traits, DM, and RS are considered fundamental quality traits of potato tubers, whether nutritional or for processing. RS plays a crucial role in potato storage, frying, and processing. Unlike other traits, lower RS content indicates better quality performance. In the RS analysis, G21 exhibited the highest stability, followed by G17, G16, G27, and G24. However, G18, G17, G32, and G14 had lower RS content. The genotypes displaying low and stable RS content were limited to G21, G17 and G18 (Fig. 2c). Similar methods were used to analyze the other quality traits, and it was found that the genotypes exhibiting high mean value and stability across different traits were not the same. According to the FW regression, G25 was identified as one of the genotypes with the best performance in this study, particularly in TY, CA, VC, and DM, being 67.84%, 56.65%, 56.61%, and 23.98% higher than that of control, respectively (Fig. 3 and Table S2).

AMMI model analysis

When decomposing the GEI using the AMMI model, it was observed that the sum of IPC1 and IPC2 for all traits were more than 60%, ranging from 62.95% (DM) to 81.73% (RS), exhibiting high reliability in explaining GEI (Table 6). The ASV were utilized to assess the stability of genotypes with different traits. For example, G7 (0.01) and G21 (0.01) displayed high-RS stability in various environments. However, this does not imply high stability in yield or other quality traits for G7 and G21 (Table 7). Additionally, high stability does not necessarily ensure high mean value. The GSI was derived from a ranking based on trait mean values and stability. A lower GSI value indicates better mean trait performance and greater stability of the genotype, making it an excellent genotype. For instance, G18 (7) and G21(7) had low GSI values for reducing sugars, indicating low reducing sugar content and high stability (Table 7). However, it is not possible for all traits of a genotype to have lower GSI values (Table 7). G25 was one of the excellent genotypes selected by the AMMI model in this study. The GSI values for G25 were significantly lower for DM, CA, and VC compared to the control (Table 7).

Table 6.

Combined ANOVA of AMMI model and GGE biplot in tuber yield, vitamin C, pyridoxamine and nicotinic acid of potatoes

	TY		DM		RS		CA		VC		VB1		PN
	SS	SS% of GEI	SS	SS% of GEI	SS	SS% of GEI	SS	SS% of GEI	SS	SS% of GEI	SS	SS% of GEI	SS	SS% of GEI
AMMI model
IPC1	257.3***	40.29	108.9***	36.64	0.54***	51.92	3506.3***	56.07	161.5***	53.73	56.5***	73.86	162.5***	68.77
IPC2	181.3***	28.39	78.2***	26.31	0.31***	29.81	747.3***	11.95	59.1***	19.66	0.25 ns	0.33	3.8 ns	1.61
IPC3	81.2***	12.72	25.5***	8.58	0.08 ns	7.69	350.2***	5.60	25.6***	8.53	0.08 ns	0.10	2.4 ns	1.02
GGE biplot
PC1	478.1***	74.86	159.9***	53.81	0.97***	94.05	4981.5***	79.66	224.9***	74.84	48.7***	63.73	111.1***	47.03
PC2	112.0***	17.54	91.0***	30.61	0.03 ns	3.50	865.5***	13.84	51.4***	17.10	23.6***	30.89	78.03***	33.01
PC3	20.8***	3.26	22.5***	7.58	0.01 ns	1.23	147.6***	2.36	6.7***	2.25	1.7***	3.63	34.3***	14.53

TY Tuber yield, DM Dry matter content, RS Reducing sugar content, CA Carotenoid content, VC Vitamin C content, VB1 Thiamine content, PN Pyridoxine content, IPC Interaction principal component, SS The sum of square

*** Significant at P ≤ 0.001

ns (not significant)

Table 7.

AMMI analysis of potato tuber yield and quality traits

Genotypes	TY		DM		RS		CA		VC		VB1		PN
	ASV	GSI	ASV	GSI	ASV	GSI	ASV	GSI	ASV	GSI	ASV	GSI	ASV	GSI
G1	97.26	54	1.48	33	0.33	37	4.28	39	3.86	36	0.36	34	0.69	45
G2	127.93	47	1.00	34	0.17	50	6.15	32	4.28	47	0.55	47	0.17	28
G3	86.38	28	1.21	34	0.40	58	2.83	56	4.28	50	0.55	30	0.74	32
G4	89.97	25	0.38	19	0.21	50	2.58	56	6.23	58	0.38	46	0.51	51
G5	155.77	34	1.26	26	0.14	46	1.47	37	3.56	46	1.64	34	0.45	21
G6	105.78	29	1.02	26	0.44	57	1.12	41	3.99	42	0.40	44	0.47	43
G7	179.83	33	1.27	43	0.01	25	0.82	36	9.56	48	0.41	48	0.41	33
G8	23.83	3	0.59	28	0.53	63	9.22	34	3.56	41	0.43	42	0.22	29
G9	58.54	10	2.20	50	0.08	48	1.01	20	0.96	35	0.38	42	1.24	37
G10	100.52	55	1.00	11	0.18	59	2.12	44	3.13	29	1.27	34	0.32	26
G11	52.58	37	1.23	37	0.09	51	0.09	24	2.90	13	0.11	9	0.49	43
G12	64.52	42	2.31	44	0.02	16	0.99	18	4.33	42	0.19	22	0.21	17
G13	181.52	59	1.69	32	0.04	27	1.56	32	0.88	3	0.18	25	0.59	39
G14	148.23	55	2.16	59	0.03	10	0.84	25	2.81	31	0.49	43	0.17	13
G15	37.65	30	1.76	53	0.10	43	0.83	18	4.57	37	0.47	26	0.80	30
G16	64.42	37	2.94	56	0.06	28	0.61	18	1.65	14	0.18	36	0.27	27
G17	134.39	51	0.73	7	0.03	13	0.54	25	4.06	40	0.11	28	0.30	43
G18	78.37	40	1.19	19	0.03	7	1.94	41	2.95	33	1.03	28	0.26	41
G19	39.22	28	1.19	25	0.04	20	1.06	39	3.37	20	0.14	21	0.49	29
G20	98.85	43	1.07	25	0.03	20	0.50	34	4.96	34	0.33	24	0.41	31
G21	94.36	40	2.34	60	0.01	7	1.43	32	2.94	24	0.25	35	0.18	31
G22	214.56	50	2.00	59	0.07	50	1.14	38	0.57	14	1.95	33	0.51	33
G23	52.96	25	0.85	29	0.05	43	1.60	28	7.66	32	0.71	36	1.06	37
G24	25.23	18	1.70	28	0.03	27	0.72	27	9.30	42	0.58	41	0.99	45
G25	68.88	30	1.37	26	0.05	29	1.72	27	1.03	9	0.37	22	0.35	34
G26	58.62	24	1.16	20	0.12	41	1.23	21	2.63	16	0.47	39	0.43	46
G27	74.02	29	0.59	24	0.02	14	0.69	31	2.86	38	0.11	23	0.11	31
G28	38.23	17	1.05	39	0.05	29	0.51	36	8.18	32	1.09	39	0.30	23
G29	67.92	25	1.68	33	0.03	9	0.18	23	3.38	35	0.46	34	0.07	25
G30	86.38	29	0.63	17	0.03	28	1.49	32	3.49	42	0.17	35	0.47	43
G31	91.88	11	0.38	17	0.05	35	3.10	59	6.90	43	2.38	35	0.22	41
G32	33.66	18	1.12	43	0.04	16	1.42	33	4.40	30	0.22	21	0.20	9

TY Tuber yield, DM Dry matter content, RS Reducing sugar content, CA Carotenoid content, VC Vitamin C content, VB1 Thiamine content, PN Pyridoxine content, ASV AMMI stability value, GSI Genotype stability index

GGE biplots analysis

The ANOVA of the GGE biplot showed results was similar to the AMMI model, with a significance at P ≤ 0.001 of the first three PCs for the same qualitative characteristics (Table 6). Finally, we employed GGE biplots to analyze GEI. The sum of PC1 + PC2 for all traits in the GGE biplot ranged from 80.04% to 97.55% (Table 6). The"mean and stability"biplot revealed that G25 demonstrated higher yield and stability across various environments (Fig. 4a). G21 exhibited the best yield stability; however, its yield was relatively low, being 36.21% lower than the control (Fig. 4a). Ideal genotypes possess high mean value and stability across all traits. However, it is challenging to achieve a thorough combination of high mean value and stability of key traits in actual breeding and production processes. Regarding quality traits, genotypes were classified into different categories based on the mean value and stability of key traits. For example, G25 exhibits high values for TY and CA, allowing for its preliminary identification as a high-yielding and high-CA variety (Fig. 4a and d). G18 performed well in TY, DM, RS, and VB1, particularly in RS, which had a lower mean and better stability, at 23.26% lower than the control. It was preliminarily identified as a frying processing variety (Fig. 4a, b, c and f).

Fig. 4.

Analysis of the mean and stability of yield and quality traits of different potato genotypes. The small circles in the plot represented the average environment. The one-way arrow points from the origin to the average environment as the average environment axis (AEA), where the direction represented the higher mean value. The lines perpendicular to the AEA axis are the average environmental coordinates (AEC). The dotted lines show the stability of the genotypes being the shorter line, the higher the stability of each genotype. The “△” represents genotype scores, the “ + ” represents environment scores, the “○” represents AEC

The"ranking genotype"biplots produced similar results compared to the"mean and stability"biplots shown in the figure above. However,"ranking genotype"biplots could distinguish and identify genotypes at the same stability level using concentric circles (Fig. 5). For example, in the TY biplot, G25 and G27 were positioned on the same level of concentric circles, indicating that their stability was also equivalent (Fig. 5a). Interestingly, some genotypes performed particularly well for certain traits while exhibiting mediocre performance for others, thus failing to achieve a balance between mean value and stability. For instance, G15 demonstrated high mean value and stability for PN, but had low mean values and stability for other traits (Fig. 5).

Fig. 5.

Analysis of the ranking genotype of yield and quality traits of different genotypes

The "which-won-where/what" biplots were used to identify the winning genotype in each environment and the mega-environments (MGEs) (Fig. 6) [10]. In the TY biplot, the genotypes were divided into three independent MGEs based on different levels of rainfall (Table 2 and Fig. 6a). The first MGE includes E1 and E2, characterized by high rainfall (annual rainfall > 500 mm), with G25 identified as the top-performing genotype within this MGE; The second MGE includes E3 and E4, characterized by moderate rainfall (annual rainfall > 300 mm) and G18 was the top-performing genotype; The third MGE includes E5 and E6, characterized by low rainfall (annual rainfall < 200 mm) and G7 was the top-performing genotype (Fig. 6a). Furthermore, it was observed that RS and VC could not completely differentiate between the MGEs represented by E1, E2, and E3, E4 (Fig. 6c and e).

Fig. 6.

Analysis of the adaptability of yield and quality traits of different genotypes. This plot consists of a polygon with perpendicular lines, called equality lines, drawn onto its sides. These lines divide the polygon into various sectors. Genotypes located on the polygon’s vertices are the best in each mega-environment from a particular sector

Combining yield and quality traits to recommend genotypes suitable for cultivation in each ecological region allowed for a more accurate identification of variety characteristics. In E1 and E2, G25 was identified as an excellent genotype, exhibiting high yield, high DM content, and high CA content (Fig. 6a, b, d and Table S2). In E3 and E4, G18 was recognized as an excellent genotype, characterized by high yield, high VB1 content, and low RS content (Fig. 6a, c, f and Table S2). In E5 and E6, G7 exhibited high yield; although its performance in CA, VB1, and PN was not optimal, it showed increases of 161.08%, 494.12%, and 513.95%, respectively, compared to the control (Fig. 6a, d, f, g and Table S2). Therefore, G7 was considered an excellent genotype with high yield and high contents of CA, VB1, and PN, making it suitable for cultivation in E5 and E6.

Genotype selection in consideration of mean value and stability of multiple trait

MTSI analysis

The genotypes of potato were ranked from the highest to the lowest according to MTSI value, with the highest MTSI value positioned in the center of the concentric circle and the lowest on the outermost periphery. The red circle in the figure denotes the selection intensity threshold (MTSI = 5.12) (Fig. 7). In this study, the preferred genotypes identified were G23, G26, G25, G24, and G16 at a 15% selection intensity (MTSI were 4.02, 4.45, 4.73, 5.08, and 5.12, respectively) (Fig. 7). Even though G18 was not chosen as the ideal genotype, it closely approached the red circle, indicating its specificity for several traits, such as TY and RS (Fig. 4). The investigation of genotypes closely situated around the selection intensity threshold is thought significant for future studies.

Fig. 7.

Genotype ranking and selected genotypes based on the MTSI considering a selection intensity of 15%

Based on the MTSI values, the selected genotypes G23, G26, G25, G24, and G16 exhibited increased mean values for all traits except VB1 and PN, compared to the other 27 genotypes (Table S2). However, the mean values for VB1 and PN in G23, G26, G25, G24, and G16 were slightly lower than the overall mean values, with no significant differences among them (Table S2). Overall, the selected genotypes showed favorable selection differentials for all traits. Similar findings were observed in the single-trait models, indicating that G23, G26, G25, G24, and G16 excelled in most traits (Figs. 3, 4 and Table 7). Specifically, G25 demonstrated significant increases in TY, VC, and DM by 67.81%, 56.60%, and 23.87%, respectively, compared to the control (Table S2). G24, G16, and G26 also performed well in TY and DM (Table S2). Therefore, MTSI is a valuable tool for plant breeders to select excellent genotypes across diverse environments and traits. Additionally, MTSI has proven to be an efficient and precise method for identifying excellent genotypes.

Discussion

Analysis of factors affecting yield and quality traits of potato tubers

Yield serves as a primary criterion for preliminary selection, identification, and classification of potato varieties for processing during field cultivation. When selecting excellent potato varieties, consideration should also be given to quality and other agronomic traits. ANOVA indicated that genotype played the most significant role in affecting potato tuber yield and the most of quality traits, followed by environmental factors and GEI (Table 4). This highlighted the extensive genotypic diversity in potatoes, which necessitated evaluation to optimize yield and quality traits. Similar results were obtained in soybeans and peppers [36, 37]. In addition to genotype, pedo-climatic conditions constitute the principal environmental influences on the qualitative attributes of potato tubers [38]. Therefore, we investigated the yield and quality traits of 32 genotypes across two different growing seasons and in three pilot sites.

The combined effects of G + GEI accounted for a significant proportion (49.71%−91.28%) of total variation in yield and quality traits. The results of this study were consistent with Tatarowska et al. [39], who pointed out that the main effect of G + GEI explained most of the variation in METs. According to the findings of Khan et al. (2021) [40], Shahriari et al. (2018) [41], and Nowosad et al. (2016) [42], the proportion of GEI in the total variance of all traits varies, accounting for the differential expression of genotypes in the studied environment. In this study, the GEI had the greatest effect on VC content (Table 4). The effect of GEI was also noticeable across most traits under normal conditions, indicating that the degree of adaptation to the environment hinges on the chosen genotype. Due to crossover interactions, none of the genotypes demonstrated excellence across multiple environments [43]. For example, G5 displayed significantly varied TY values across the three pilot sites, with the highest in Weiyuan County and the lowest in Yongchang County while G8 exhibited the highest VC value in Weiyuan County and the lowest in Yongchang County (Table S2). Similar trends were observed in DM and RS. This disparity was attributed to WY and YC being part of distinct ecological regions, with a substantial variance in average annual rainfall of up to 365 mm (Table 2). Previous research has indicated that rainfall stands out as the key environmental factor influencing yield and quality traits [30]. The impact of divergent rainfall patterns on potato yield and quality traits in this study aligns with earlier findings. These results suggested that improvements in crop performance linked to water adjustments, such as improved water use efficiency, will resonate in overall productivity and potato quality [30, 44, 45].

The results of this study revealed that VB1 and RS exhibited higher PCV values, suggesting a wider selection range for these traits, while DM and VC displayed a more restricted selection range (Table 5). At the same time, the GCV values of VB1 and RS were also high, indicating that genotypes had a greater influence on these traits, which may assist in trait selection (Table 5). The PCV values of DM, CA, and PN were close to the GCV values, implying that these traits were less responsive to environmental fluctuations and more inclined towards genetic control. This observation aligned with previous studies investigating the association between PCV and GCV [30, 46]. Conversely, the substantial disparity between PCV and GCV values for VC and VB1 suggested a pronounced environmental impact on these traits, indicating larger GEI effects and heightened trait instability, consistent with the results by ANOVA (Tables 4 and 5).

Broad-sense heritability serves as a pivotal parameter for gauging phenotypic variation attributable to genetic factors [47]. Our study revealed elevated heritability estimates for TY, CA, and PN, signifying strong genetic determination of these traits with limited environmental influence (Table 5). It is widely recognized that the trait of TY in potatoes, along with other crops, has high heritability [48]. CA and PN represented essential antioxidants in potatoes, showcasing diverse health-enhancing properties [49]. Certain vitamins, such as PN, play crucial roles in determining the quality of potato products during frying. High-temperature frying of potatoes is often associated with health risks due to the potential formation of carcinogenic acrylamide [50]. Research has shown that vitamins like PN can mitigate acrylamide formation, with VB1 and PN identified as the most potent inhibitors, achieving inhibition rates of over 70% [51]. Additionally, water-soluble vitamins, including biotin and VC, have demonstrated a more than 50% reduction in acrylamide formation [51, 52]. The high heritability estimates for TY, CA, and PN emphasized the potential for enhancing potato characteristics through targeted selection for yield and quality traits [30].

Genotype identification of single trait using FW regression, AMMI model, and GGE biplot

A larger GEI increases the instability of genotype performance and poses challenges for breeding programs, which have become major obstacles in identifying stable genotypes in METs [39]. FW regression provides comprehensive information in a single graph for each trait, focusing on sensitivity and the selection of optimal genotypes across environments [30]. ASV values in the AMMI model were used to evaluate genotypic stability in METs [32]. ASV values were heavily dependent on the first two IPCs, with IPC1 and IPC2 making substantial contributions to all traits in this study (Table 6). The"mean and stability"biplot in the GGE biplot was used for stability analysis. It visualized stability through the first two PCs, providing an intuitive understanding of genotype performance [53]. In this study, both PC1 and PC2 for all traits were significant at P ≤ 0.001 and accounted for a substantial proportion of the sum of squares (Table 6). The high variance explained by the first two PCs (IPCs) due to GEI highlights the necessity of examining stability through METs. This examination is crucial for identifying any repeatable GEI patterns [20, 54] stated that a reliable interpretation of GEI variation can be achieved if the first two IPCs in the AMMI model account for > 66% and the first two PCs in the GGE biplot account for > 60% of the GEI variation. The first two PCs (IPCs) for all traits in this study met the threshold requirements. Therefore, analyzing the effect of GEI on the traits selected by PCA using the AMMI model and GGE biplot is of significant research value and represents a comprehensive approach.

The results from FW regression, AMMI model, and GGE biplot were very similar, indicating that the genotype exhibited significant fluctuations across different environments, likely due to variations in environmental conditions. This result was consistent with previous studies in maize and wheat [55, 56]. The key to genotype identification lies in balancing stability and high yield. In this study, G25 outperformed other genotypes in TY, CA, and VC values, showing the highest average and stability in TY (Figs. 3, 4, and Table 7). Stability in CA and VC was moderate but remained significantly above average (Figs. 3, 4, and Table 7). This is likely due to the influence of abiotic factors (rainfall, temperature, and soil properties), which also contribute to the significant GEI effects observed [2]. Thus, G25 can serve both as a high-yield and specialized variety with high CA and VC content, enhancing its quality traits.

The GGE biplot also serves an important function through the"which-won-where/what"biplot, revealing the best-performing genotypes within each MGE and facilitating the identification of genotypes suitable for specific growth environments [57]. In this study, the"which-won-where/what"biplots for TY, DM, CA, VB1, and PN divided the pilots into three distinct MEGs. The primary distinction among these MEGs was the varying levels of rainfall, indicating that moisture is a crucial factor influencing these traits in potatoes [58]. However, for VC and RS, E1, E2, E3, and E4 were grouped into the same MEG, failing to effectively differentiate the pilot sites. This may be due to GEI being the main factor affecting VC, while genotype was the primary factor influencing RS, with environmental effects being relatively minor (Table 4). Additionally, the geographical proximity of E1 and E2 (Weiyuan County) to E3 and E4 (Anding District) may have contributed to the inability to distinguish the pilot sites in terms of VC and RS (Fig. 1).

The three models employed to elucidate GEI are complementary due to their diverse statistical foundations, and integrating their outcomes enhances the credibility and reliability of the conclusions, aligning with prior studies [59]. Using a combination of different models to verify a specific conclusion is advisable, particularly when employing multiple methods [30]. Although a single model can substantially simplify results for identifying the best genotype based on a single trait, its scope may be limited in comprehensively considering multiple traits when the experimental objective is to identify excellent varieties based on combined traits.

Genotype identification based on multiple traits using MTSI

The introduction of MTSI offers breeders an integrated approach for recommending genotypes across multiple traits [34]. Lower MTSI indicates a higher overall mean value and stability of the genotypes [60]. In this study, G23, G26, G25, G24, and G16 emerged with the highest overall mean values and multi-trait stability (Fig. 7 and Table S2). In the FW regression, AMMI model, and GGE biplot, similar results were also obtained. Additionally, the"which-won-where/what"biplot clearly showed that the selected genotype G25 performed best for TY in E1 and E2, G24 had the highest DM performance in E5 and E6, and G23 excelled in PN in E5 and E6 (Fig. 6). However, a limitation of recommending genotypes based on MTSI is that while it can rank genotypes by amalgamating multiple traits in diverse environments, it is unable to segment and recommend specific adaptive regions for genotypes [61]. Due to MTSI simplifies the selection process and delivers concise results, its practical utility has been initially demonstrated in multi-environment crop breeding and genotype recommendation [62]. Notably, MTSI distinguishes itself by its ability to flexibly adjust the importance of target traits and the inclusion of traits based on specific breeding goals [34]. The study identified that the selection response of the designated genotypes exhibited enhanced selection efficacy at a 15% selection intensity, aligning with findings by Reddy et al. [61]. This research presents a case study of potato genotype recommendation in northwest China using MTSI, presenting a rational and effective approach that can be extrapolated for genotype recommendation across other crops and regions.

Conclusion

This study demonstrated that GEI significantly affected potato yield and quality traits, with vitamin C content, reducing sugar content, and carotenoid content being most sensitive to environmental variability. Through multi-environment trials and stability analyses, we identified five widely adapted, high-yielding genotypes (G23, G26, G25, G24, and G16) for Northwest China. Notably, G25 was ideal for humid regions (Weiyuan County) with high yield, dry matter, and carotenoid content, while G18 (despite MTSI exclusion) performed well in semi-arid areas (Anding District) with low reducing sugars and high thiamine content. G7 outperformed others in arid regions (Yongchang County) for yield and pyridoxine content. The MTSI effectively integrates potato yield and quality traits, guiding environment-specific cultivation practices and optimizing breeding programs to identify widely adaptable and stable varieties. Overall, this study highlights how GEI-aware selection improves potato breeding and provides a scalable approach for selecting crops with multiple desirable traits in changing environments.

Authors’ contributions

YW and FZ designed the study. LC, JY and JL performed the experiments. BZ, JY, JL and FZ analyzed the data. BZ wrote the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Key Research and Development Program of Gansu Province (25YFNA035) and National Key Research and Development Program of China (SQ2022YFD1600328).

Data availability

The data sets supporting the results of this article are included within the article and its additional files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All co-authors have seen and agreed on the contents of the manuscript, and there is no financial interest in reporting.

Competing interests

The authors declare no competing interests.