Introduction

As an increased frequency and intensity of drought episodes worldwide are more tangible than ever, so are the concerns about their consequences on crop yields and food production (Dai 2011; Dai 2012; Cheeseman 2016). The impact of environmental stressors on food security (Ejaz Qureshi et al. 2013), together with the increasing nutritional demands from the population, calls for a renewed exploration of untapped genetic resources in staple crops such as potato (Solanum tuberosum). One of the current limitations of the potato crop is its susceptibility to drought (van Loon 1981). Paradoxically, its water use efficiency (5626 kcal/m3) has been shown to be significantly higher than other popular crops like maize (3856 kcal/m3) and wheat (2279 kcal/m3) (Renault and Wallender 2000).

In addition to being a good source of carbohydrates, potato tubers have important nutrients, micronutrients and health-promoting bioactives such as resistant starch (RS), phenolic compounds and anthocyanins and therefore constitute a viable option to deliver additional nutritional benefits to consumers on a global scale (Zaheer and Akhtar 2016). Resistant starch in potato has prebiotic properties, as it promotes the production of metabolites such as short-chain fatty acids by intestinal microbiota, which are recognised for their health benefits such as controlling glucose, insulin and cholesterol levels. Furthermore, phenolic compounds and anthocyanins have been associated with health benefits in pathologies related to oxidative stress, such as cardiovascular problems and cancer, amongst others (Andre et al. 2007; Burlingame et al. 2009; Ezekiel et al. 2013; Giusti et al. 2014; Charepalli et al. 2015; Yang et al. 2016).

The existing link between plant secondary metabolites, environmental stresses and agronomic management could serve to enhance bioactive content in food products through proper agronomic management practices (Akula and Ravishankar 2011; Keutgen et al. 2019). Nevertheless, establishing a connection between environmental stresses and bioactive production has been elusive in potato tubers (Wegener and Jansen 2013; Wegener et al. 2015).

The aim of this study was to assess the effect of drought stress on agronomic parameters and the nutritional content of potato tubers to provide new insights into the link between environmental stress and bioactive production.

Materials and Methods

Plant Materials

Tubers of native potato cultivars are known to be very diverse in size, shape and colour. Both phenotypic diversity and availability were used as main criteria for the selection of accessions used in the current study. The eight native Chilean potato landraces used (Cabra Roja, Chona Negra, Michuñe Azul, Michuñe Blanca, Michuñe Roja, Montañera, Murta, Quila), plus one commercial variety (Désirée), were kindly provided by the Universidad Austral de Chile (UACH, Valdivia, Chile). A thorough description of the available information on their agronomical traits can be found in the catalogue written by Contreras and Castro (2008).

Chemicals

All reagents were of analytical grade. Trolox, gallic acid (GA) and cynarin-3-glucoside (C3G) were from Sigma-Aldrich, St. Louis, USA. A K-RSTAR kit was used for resistant starch measurements (Megazyme International Ireland Limited, Wicklow, Ireland).

Field Experimental Design

The present study was carried out at the experimental field station of the UACH in Santa Rosa, in the south of Chile in the region of Los Lagos. This region is regarded as the place of origin of one of the important potato gene pools currently existing (Contreras and Castro 2008). The field site was covered with a black plastic foil (CHILEMAT, Chile), in order to prevent the interruption of the drought treatments by rain. Water was supplied by a drip irrigation system installed below the plastic foil. The experimental design consisted of a split plot with four replicates. Drought treatments were placed as main plots, whereas the genotypes were placed as sub-plots and the cultivar Désirée was used for border rows. Nine potato plants were spaced 30 × 70 cm in three rows, resulting in a plant density of about 48,000 plants per ha. Pests were controlled every 2 weeks according to good agricultural practice with the fungicide Ridomil Gold (Syngenta Production France S.A.S., Gaillon, France) and the insecticide Karate Zeon (Syngenta Chemicals B.V., Seneffe, Belgium).

Control plots were well watered throughout the experiment by means of drip irrigation, on Mondays and Thursdays during the whole working day (~ 8 h). Drought treatments were carried out by discontinuing irrigation for 6 weeks during the early-bulking stage of tubers at 88 days after planting (88 DAP, treatment 1 or T1) or during the late-bulking stage at 110 DAP (treatment 2 or T2). Four plants representative of each plot were harvested individually at the end of the experiment (154 DAP). The impact of the drought treatments was assessed by counting the number of tubers and their combined weight (yield) per individual plant. In addition, tuber bioactive and starch content were analysed as described below.

Tuber Selection and Processing

Bioactive analyses were carried out on 162 samples consisting of nine genotypes, three replicates, three drought treatments (control, T1, T2) and two cooking treatments (raw, boiled). Based on the homogeneity of the blocks, three out of four blocks were selected for these postharvest assessments.

Four plants per plot were randomly selected after harvest for sampling and eight tubers from each selected plant were collected randomly to carry out the analyses. Four of these eight tubers were boiled with peel for 25 min, and the other four were kept raw. In total, 32 tubers were processed per plot and for each variety of potato (4 plants × 8 tubers/plant). Potatoes were processed with skin to reduce the potential solubilization of compounds of interest in water and reflect cooking methodologies that reduce waste generation (industrial and household). Tubers were dried by lyophilisation, ground into a powder and pooled for further chemical analyses. All analyses were performed on samples dried by lyophilisation, so the dry weight used for calculations was the weight after lyophilisation.

Determination of Resistant and Total Starch

Resistant starch content was determined using the Megazyme Resistant starch kit (K-RSTAR, Megazyme International Ireland Ltd., Wicklow, Ireland) according to AOAC 2002.02 method (McCleary and Monaghan 2002). Briefly, 100 mg of the lyophilised samples was incubated with pancreatic amylase (α-amylase 10 mg/mL) and amyloglucosidase (AMG 3 U/mL) for 16 h at 37°C to produce glucose from digestible starch (non-resistant starch). The resistant starch (pellet) was recovered by centrifugation and was washed three times with ethanol 50%. Then, resistant starch was solubilised using an alkali solution (KOH 2M) and hydrolysed to glucose using a concentrated amyloglucosidase solution (AMG 3300 U/mL) at 50°C for 30 min. Glucose was determined using an enzymatic kit determination (GOPOD reagent).

Determination of Total Phenolic Compounds (TPCs)

TPCs were determined using a modified Folin-Ciocalteu method (Singleton and Rossi 1965). Briefly, 50 mg of pulverised tuber was extracted with 2 mL acidified methanol (0.01% v/v HCl in methanol); then, a mixture was prepared of 3.75 mL of deionised water, 0.5 mL of extract, 0.25 mL of Folin-Ciocalteu phenol reagent (Merck KGaA, Darmstadt, Germany) diluted two-fold in deionised water and 0.5 mL of 10% (w/v) sodium carbonate (Merck KGaA, Darmstadt, Germany). Absorbance at 765 nm was determined after 1 h, and gallic acid was used as the standard.

Determination of Anthocyanin Content

The extraction and determination of anthocyanins were carried out according to Giusti and Wrolstad (2001). A total of 50 mg of pulverised tuber was extracted with 2 mL acidified methanol (0.01% v/v HCl in methanol) and anthocyanin content was determined using the pH-differential method carried out with an UV-Visible spectrometer (V-630, Jasco, Easton, USA). In brief, two sample solutions were prepared, one with potassium chloride buffer (pH 1.0, 0.025 M) and one with sodium acetate buffer (pH 4.5, 0.4 M). The absorbance of each dilution was measured at 700 nm and at the maximum absorbance wavelength against distilled water, according to the compound to be analysed or C3G, which was used as standard (530 nm). The monomeric anthocyanin concentration was determined considering the molecular weight of the standard (MW), the dilution factor (DF) used, the absorbance of the diluted sample (A) and the molar absorptivity of the standard compound (ε) using the following formula: Monomeric anthocyanin pigment (mg/L) = (A ×MW ×DF × 1000) / (ε × 1). The absorbance of the diluted sample corresponds to the difference between absorbance obtained at maximum wavelength and at 700 nm of both dilutions (pH 1.0 and pH 4.5) (A = (Aλvis-max − A700)pH 1.0 − (Aλvis-max − A700)pH 4.5).

Antioxidant Capacity Determination by the ORAC Method

The ORAC (oxygen radical absorbance capacity) method was used as a first approximation to the determination of the antioxidant capacity of the samples. This is a method that allows analysing the free radical scavenging of compounds with or without a lag phase in their antioxidant capacity, so it is of use in foods and/or complex samples that contain various compounds that cannot be individualised, which exert antioxidant activity. ORAC was determined according to Garrett et al. (2010) using fluorescein and 2,2′-azobis (2-amidinopropane) dihydrochloride (AAPH) (both chemicals from Sigma-Aldrich, St. Louis, USA). A total of 200 μL of fluorescein (108 nM in PBS buffer pH 7.4) and 20 μL of the extract were incubated at 37°C for 10 min and then, 75 μL of AAPH (79.7 mM in PBS buffer pH 7.4) was added to initiate the reactive oxygen species (ROS) generation. Fluorescence was followed for 60 min, using 485/538 nm excitation/emission wavelengths (Fluoroskan Ascent, Thermo Scientific, Vantaa, Finland). The result expressed as the area under the curve (AUC) of the fluorescence signal decrease, compared to the AUC of a curve made with a standard antioxidant, accounts for the antioxidant activity of the sample. Trolox was used as the standard and results were expressed as μmol trolox equivalents (TE)/100 g of lyophilised potato.

Bioactives per Hectare

Data per hectare were calculated considering 48,000 plants/ha and the average percentage of dehydrated matter (lyophilised matter) per tuber and the bioactive concentration.

Statistical Analyses

Analyses were performed using R version 3.3.1 (R Core Team 2016), RStudio (2015), and visualised using GraphPad Prism version 8 (GraphPad Software, San Diego, CA, USA). Data were normalised through a square root transformation and were analysed via linear mixed models, using the packages lme4 (Bates et al. 2015) and lmerTest (Kuznetsova et al. 2017).

The linear mixed model below was used to take into account the experimental design:

$$ {Y}_{\mathrm{i}\mathrm{j}\mathrm{k}\mathrm{l}}=\mu +{\alpha}_{\mathrm{i}}+{\beta}_{\mathrm{j}}+{\alpha \beta}_{\mathrm{i}\mathrm{j}}+{r}_{\mathrm{k}}+\left({\beta r b}_{\mathrm{j}\mathrm{k}\mathrm{l}}\right)+\left({\alpha r}_{\mathrm{i}\mathrm{k}}\right)+{\varepsilon}_{\mathrm{i}\mathrm{j}\mathrm{k}\mathrm{l}} $$

with Y the response variable, α the main genotype effect (i = 1, 2,…, 9), β the main drought treatment effect (j = 1, 2, 3), αβ the interaction between genotype and treatment, r the replicate effect (k = 1, 2, 3), βrb the random effect of the split plot (with l = 1, 2,…, 12; the number of sub plots), αr the interaction between genotype and replicate and ε the residual error. Direct correlations between variables and adjusted R2 were calculated through linear regressions.

Results and Discussion

Yield and Number of Tubers Produced per Plant

Drought has been shown to have a negative impact not only in diverse physiological processes including photosynthesis and respiration, but also on parameters of agronomic importance such as yield (Levy 2014; Yordanov et al. 2000). Average yields per plant differed significantly between treatments (P<0.05) (Fig. 1) with an average yield of 1059 g per plant for the well-watered control, whereas it was 611 g in the early bulking treatment (T1), and 813 g in the late bulking treatment (T2) (Fig. 1; Table 1). The average number of tubers produced per plant was also reduced (P<0.01) from 35 tubers in control plots to 21 in T1 and to 23 in T2, respectively (Fig. 1; Table 1).

Fig. 1
figure 1

Distribution of yield in g (a) and number of tubers produced per plant (b) under well-watered conditions (control), drought stress 88 days after planting (T1) or 110 days after planting (T2). The shape of the areas represents the distribution of the data for each condition. The continuous line represents the average, the dashed line depicts the median and the dotted lines depict the 10 and 90 percentiles respectively. The effects of treatment and genotype were significant at α = 5% (P<0.01), but not the interaction between these factors

Full size image
Table 1 Anthocyanins (mg C3GE/100 g DW), TPC (mg GAE/g DW), ORAC (μmol TE/100 g DW), dry matter (% fresh weight), average yield per plant (yield/plt in g) and average number of tubers per plant (tubers/plt). Values correspond to the average ± SD. Interactions between drought treatment and genotypes are reported at the bottom of the table (α = 5%; NS non-significant)
Full size table

Significant differences were also observed between varieties in the number of tubers produced (P<0.01) and yield (P<0.01) (Fig. 2). The average yield per plant in control plots ranged from 402 g for Michuñe Roja to more than 1.8 kg for Désirée, whereas the number of tubers ranged from 17 for Michuñe Blanca to 59 for Cabra Roja (Fig. 2; Table 1). Cabra Roja was the genotype most affected by early bulking stress (T1) achieving only 29% of the yield of the control treatment (Table 1). In contrast, the commercial variety Désirée performed well under T1 keeping 83% of the yield. The responses of the different varieties were also contrasting when the drought occurred during late bulking (T2). Cabra Roja was again the most penalised genotype achieving 37% of the yield of the control conditions (Table 1). The variety Désirée showed a yield comparable to the control conditions, despite producing less tubers (78%), confirming its known tolerance to drought stress.

Fig. 2
figure 2

Tuber phenotype and average yield and number of tubers per plant in control conditions. Labels correspond to the following landraces: 1, Cabra Roja; 2, Chona Negra; 3, Désirée; 4, Michuñe Azul; 5, Michuñe Blanca; 6, Michuñe Roja; 7, Montañera; 8, Murta; 9, Quila

Full size image

The landrace Michuñe Roja performed well under drought 88 DAP maintaining 80% of the yield compared to well-watered conditions, even though its absolute yield was significantly lower compared to the other landraces. Concerning the treatment 110 DAP, the native genotypes Michuñe Azul and Quila performed well by maintaining respectively 93% and 89% of the yield of the control (in weight).

Impact of Drought on Bioactives

Drought treatments had a detrimental effect on yield and tuber production. However, they did not impact the concentration of bioactives and antioxidant activity in the tubers (Tables 1 and 2). As a result, the production of compounds estimated per area cultivated (ha) was significantly reduced (P<0.01) as shown in Table 3.

Table 2 Non-resistant starch (NRS), resistant starch (RS), total starch (total) content and percentage of RS in total starch (% RS) for raw and cooked samples. Results expressed in % DW and presented per genotype and per drought treatment: control; treatment 1 (T1, 88 DAP); treatment 2 (T2, 110 DAP). Values correspond to the average ± SD. Interactions between drought treatment and genotypes are reported at the bottom of the table (α = 5%; NS non-significant)
Full size table
Table 3 Per hectare: anthocyanins (g C3GE/ha), TPC (kg GAE/ha), ORAC (mol TE/ha), non-resistant starch (NR, in kg/ha), resistant starch (kg/ha) and total starch (kg/ha) for raw and cooked samples. Average ± SD. Interactions between drought treatment and genotypes are reported at the bottom of the table (α = 5%; NS non-significant)
Full size table

Dry Matter

The percentage of dry matter in the tubers was significantly different between genotypes, both raw and boiled (P<0.01), with values that ranged in the boiled samples between 16% (Michuñe Azul) and 27% (Chona Negra) (Table 1).

Resistant and Non-resistant Starch

Genotypes differed significantly in their total starch content in both raw and boiled samples (P<0.01) (Table 2). In raw tubers, total starch content ranged from 34% of the dry weight (Michuñe Roja) to 54% (Chona Negra), while in boiled tubers, it ranged from 36% (Michuñe Azul) to 60% (Quila). On average, boiled samples displayed a higher total starch content than raw samples (+15%). The relationship between total starch of raw and boiled samples was significant (P<0.01) but with a relatively low R2 = 0.38.

Resistant starch was the main form of starch found in raw samples (81% of total starch), whereas non-resistant starch was predominant in boiled samples (89% of total starch) (Table 2). Similar values, between 71 and 87% of RS for raw potatoes, were reported by Bach et al. (2013). Starch content can vary generally between 10 and 19% of the fresh weight for commercial varieties (Bethke 2014; Schwärzel et al. 2016), whereas values in dry weight range from 61.5 to 75.8% DW, as reported for the varieties Imilla Negra and Kufri Bahar respectively (Negi and Nath 2002; Burlingame et al. 2009; Jiménez et al. 2009). Our measurements on raw potatoes showed starch values between 36 and 51% of the DW, which are significantly lower. Different cultivation practices, postharvest conditions and methodological analyses may be partially responsible for these contrasting values with literature. This is especially relevant when the natural diversity of an agronomic trait is being assessed. By subjecting all varieties to the same methodology, this study provides a good example of the diversity in starch content within Chilean potato landraces.

Resistant starch content was significantly different amongst genotypes (P<0.01), with values in raw samples between 25% DW (Michuñe Azul) and 44% DW (Montañera). This is in contrast with the uniformity observed of RS in commercial varieties as reported by Raatz et al. (2016) and highlights the importance of assessment of RS on native potatoes. Once boiled, resistant starch ranged from 4.5% DW for Montañera to 7% DW for Chona Negra, a landrace that was consistently on the higher range of resistant starch content compared to the other genotypes. The effect of boiling of tubers on resistant starch agreed with literature and no correlation could be found between the resistant and non-resistant starch contents of raw and boiled potatoes (R2 < 0.1) (Fig. 3e, f).

Fig. 3
figure 3

Correlation between raw and cooked samples for different variables amongst Chilean potato landraces (without the commercial variety Désirée)

Full size image

Total Phenolic Compounds

TPC content of raw tubers ranged from 0.8 mg GAE/g DW for the landrace Michuñe Blanca to 13.3 mg GAE/g DW for the commercial variety Désirée (Table 1). After boiling, the content of TPC decreased significantly for all genotypes (83% on average). With a TPC range from 67 to 1330 mg/100 g DW, our results are comparable to previous reports for raw samples’ TPC as reported by Kita et al. (2015) and Andre et al. (2007). Regarding boiled samples, values from 23 to 80 mg GAE/100 g in our samples were in a lower range than those reported by Xu et al. (2009) (80–224 mg GAE/100 g DW). Interestingly, the variety Désirée exhibited a higher TPC compared to the native genotypes in raw samples; however, most of these phenols (97%) were lost during boiling (Table 1). This commercial variety also behaved as an outlier in Inostroza-Blancheteau et al. (2018), with a chemical composition different from the Chilean landraces. The contrasting effect in TPC content after boiling between Chilean potatoes and the variety Désirée, added to the findings of Inostroza-Blancheteau et al. (2018), suggests that significant modifications in the chemical composition of tubers in modern varieties, compared to their original native varieties, may have taken place as a result of selection in breeding programs.

Considering only Chilean landraces, there was a strong correlation between TPC of raw and boiled samples (R2 = 0.65, P<0.01) (Fig. 3b). Additionally, TPC and anthocyanins of raw and boiled samples were highly correlated (R2 = 0.9 raw, R2 = 0.75 boiled, P<0.01) (Fig. 4a, b). The lower correlation in boiled samples suggests changes in the contribution of anthocyanins to the TPC in our samples. Interestingly, raw Désirée tubers had a high TPC content but very few anthocyanins. If polyphenols are the main contributors of AA in raw samples but are lost by leaching during boiling, it is possible that other compounds increase their contribution to the AA of boiled samples. For example, the thermic process has been shown to induce the hydrolysis of some glycosylated antioxidants, which will then be more active in their free form (Xu et al. 2007; Navarre et al. 2010; Andre et al. 2014).

Fig. 4
figure 4

Correlation between different bioactives (anthocyanins, polyphenols and ORAC) in raw and cooked samples of Chilean potato landraces

Full size image

Anthocyanin Content

Several authors describe potatoes as good sources of anthocyanins, especially native genotypes from South America (Brown et al. 2007; Lachman et al. 2009; Burgos et al. 2013; Tierno et al. 2015; Calliope et al. 2018). Anthocyanins in potato tubers have been reported in a range from 0 to 153 mg/100 g DW (Brown et al. 2007; Giusti et al. 2014), well within the range of the results from our study (0 to 97 mg/100 g DW).

Anthocyanin content was significantly different amongst landraces (P<0.01) ranging from 0 to 97.1 mg C3GE/100 g DW in raw samples and from 0 to 11 mg C3GE/100 g DW in boiled samples (Table 1). Flesh colour seemed to be a good indicator of tuber anthocyanin content (P<0.01) as described by Calliope et al. (2018) (Fig. 5b). Indeed, the dark blue–coloured genotype Chona Negra showed the highest content in anthocyanins, whereas the very light-coloured landrace Quila showed the lowest content (Table 1).

Fig. 5
figure 5

Anthocyanins (mg C3GE/100 g DW; panels a and b) and ORAC (μmol TE/100 g DW; panels c and d) in cooked Chilean potato landraces according to their skin and flesh colours

Full size image

Anthocyanins decreased 84% on average in all genotypes after boiling. The ranking of the genotypes according to their content of anthocyanins did not change after boiling and the landraces Chona Negra and Cabra Roja showed the highest values (Table 1).

A high correlation was found between anthocyanin content of boiled and raw tubers (R2 = 0.9, P<0.01). Therefore, anthocyanin content in boiled tubers (y) could be well estimated from the raw content (x) as determined by the formula y = 0.448 + 0.128x (Fig. 3a).

ORAC

ORAC was significantly different amongst genotypes (P<0.01) and between raw and boiled samples (P<0.01). However, it did not change significantly as a result of the drought treatments in the field (Table 1). Antioxidant activity in raw material ranged from 1380 μmol TE/100 g DW for Quila to 9913 μmol TE/100 g DW for Chona Negra. Boiling decreased ORAC on average 62% for all genotypes. The variety Désirée showed the lowest ORAC when boiled (554 μmol TE/100 g DW) while Chona Negra remained as the genotype with the highest ORAC (4177 μmol TE/100 g DW). Andre et al. (2007) reported ORAC values between 28 and 251 μmol/g DW for raw samples of 74 Andean potato cultivars. A similar range was observed by Navarre et al. (2011) and Brown et al. (2005).

Considering only the Chilean landraces, the correlation between ORAC raw and boiled was significant (P<0.01), although with a low R2 = 0.31 (Fig. 4c). This suggests that AA of raw potatoes is not a good predictor of antioxidant capacity of boiled material. This is especially relevant when health properties are inferred based on the antioxidant activity of raw tubers. There was a significant correlation of antioxidant activity with TPC and anthocyanins in tubers as reported by other studies such as Lee et al. (2016). In particular, a high correlation was found in raw samples between ORAC and TPC (R2 = 0.71), as well as between ORAC and anthocyanin content (R2= 0.67). The same trend was shown after boiling, with a R2 of 0.39 between ORAC and TPC and of 0.38 between ORAC and anthocyanins (Fig. 4c, d, e, f).

Conclusion

Despite the narrow genetic variability of current commercial potato varieties, there are abundant untapped genetic resources from South America available for further improvement of this crop. The assessment of physiological and agronomical aspects of potato landraces is fundamental if we are to give proper use to these resources. Field studies are difficult to carry out, but they are of the utmost importance as plant growth and development can be significantly different compared to studies carried under controlled climate chamber conditions. Our results show that yield and tuber number are affected differently to bioactive production under drought stress. To our knowledge, this is the first time that information on resistant starch in native potatoes from Chile is provided. This added to the assessment of several bioactives and allows a novel insight into the untapped opportunities offered by Chilean potatoes for the future development of this staple crop.