Introduction

Potato, the world’s fourth most important food crop, with an annual production of 330 million tons (FAO 2010), regularly undergoes drought stress in most of its growing regions. Drought occurs due to erratic rainfall, especially in tropical and subtropical regions where potato is mainly cultivated under rainfed conditions, or to inadequate irrigation techniques. Drought affects potato growth and production by limiting the amount of productive foliage (Jefferies 1993), decreasing the rate of photosynthesis per unit of leaf area (Ta et al. 2003) and reducing harvest index (Deblonde et al. 1999). It also influences tuber quality (Ta et al. 2003). By exacerbating drought events throughout the world, climate change is expected to strongly impact potato production. Yield losses in the world due to drought are expected to range between 18 % and 32 % in the next 30 years (Hijmans 2003). The response of potato to water stress is well documented (van Loon 1986). All stages of crop development are sensitive to drought stress, but the magnitude of the response highly depends on the timing of water limitation (Spitters and Schapendonk 1990). For example, insufficient water supply in the period between emergence and the beginning of tuber bulking leads to reduced growth rate of foliage, incomplete ground cover and reduced intercepted radiation (Tourneux et al. 2003).

The modern varieties of S. tuberosum L. are considered as susceptible to drought stress (Weisz et al. 1994). This is likely to be due to the relatively narrow genetic base of this group, where more than 80 % of the genes of modern varieties come from varieties developed and grown at the beginning of the 20th century (Contreras 2008). To improve potato yield in less optimum climatic conditions and adapt potato to drastic changes in climatic conditions, it is consequently important to explore the capacity of the available germplasm (improved varieties, genetic stocks and landraces) to provide useful traits and genes for facing these challenges. Of particular interest are the Andean potato landraces, extremely diverse and cultivated in a large range of agro-ecological conditions (Ritter et al. 2008). Andean potatoes include the species Solanum ajanhuiri Juz. and Bukasov, Solanum curtilobum Juz. and Bukasov, Solanum juzepczukii Bukasov, and the cultivar groups Andigenum, Chaucha, Goniocalyx, Phureja and Stenotomum within S. tuberosum ssp. andigenum (Spooner et al. 2007; Ovchinnikova et al. 2011). Due to their evolution and cultivation in the cold and dry Andean Altiplano, the Andean potatoes are supposed to be more drought tolerant than the non Andean S. tuberosum potatoes (Ritter et al. 2008). Information on drought tolerance of Andean native potatoes is however scarce. Vasquez-Robinet et al. (2008) reported that within S. tuberosum L., photosynthesis is less affected by drought in the Andigenum group than in modern varieties. Little is known, however, about the level of tolerance of the other Andean potatoes.

The objectives of the present study were to analyze the variation for drought tolerance in a large set of improved varieties, genetic stocks and landraces and identify drought tolerant cultivar groups and accessions.

Materials and Methods

Plant Material

A set of 918 potato accessions provided by CIP (Centro Internacional de la Papa) genebank has been used in this study. This set, previously used for diversity studies, was genotyped with micro-satellite markers (Ghislain et al. 2004). It is composed of 226 improved varieties, 74 genetic stocks and 618 landraces.

The subset of improved varieties included foreign varieties and advanced breeding clones extracted from the Late Blight Resistant (LBR) and Lowland Tropic Virus Resistant (LTVR) populations developed at CIP (Bonierbale et al. 2003). Two segregating diploid population (BCT and PD) genetic stocks used in the study were extracted from populations developed at CIP to improve late blight resistance and nutritional characteristics, respectively. The BCT backcross population was developed by Bonierbale et al. (1994). Briefly, an interspecific progeny was developed by crossing a dihaploid S. tuberosum clone (USW-2230) as female with an individual of the accession PI473331 of the wild species Solanum berthaultii Hawkes. One individual from the F1 progeny (M200-30) was then backcrossed with the S. tuberosum parent clone HH1-9 to generate the backcross progeny (Bonierbale et al. 1994). The diploid population PD progenies were produced from a cross between a native cultivated potato belonging to the Phureja cultivar group (with quantitative resistance to potato late blight) and the S. tuberosum dihaploid PS-3 highly susceptible to potato late blight (Ghislain et al. 2001).

The 618 landraces accessions belonged, following the taxonomy of cultivated potatoes proposed by Spooner et al. (2007) and Ovchinnikova et al. (2011), to the four species Solanum ajanhuiri Juz. & Bukasov (15), S. curtilobum Juz. & Bukasov (15), S. juzepczukii Bukasov (22) and S. tuberosum L. (566). In the present study, we distinguished, within S. tuberosum L., different cultivars groups. According to Spooner et al. (2007) cultivar groups are taxonomic categories used by the International Code of Nomenclature of Cultivated Plants to associate cultivated plants with traits that are of use to agriculturists. Based on origin, ploidy and previous classifications, in particular Huamán and Spooner (2002), we considered the cultivar groups Andigenum (261), Chaucha (113), Goniocalyx (45), Phureja (81), Stenotomum (43) and non Andean S. tuberosum (23), this last one referred by Ovchinnikova et al. (2011) as Chilotanum group. L. The three species S. ajanhuiri, S. curtilobum, S. juzepczukii and the five S. tuberosum cultivar groups were considered as eight different groups in the statistical analysis. Three materials (‘Revolucion’, ‘Achirana’ and ‘Compis’) were included as controls for comparison across trials and years. ‘Revolucion’ is an improved variety, well adapted to the arid conditions of the Peruvian Central Coast of Peru, but susceptible to drought. ‘Achirana’ is a drought tolerant improved variety, grown in several countries worldwide. ‘Compis’ is an Andigenum landrace, well adapted to the dry highlands of Peru.

The set was evaluated throughout 5 years, from 2004 to 2008, in CIP’s La Molina experiment station, located in the coast of Peru, at latitude −12.0778°, longitude −76.9110° and 240 m above sea level. La Molina climate is characterized by the absence of rainfall, high relative humidity (between 71.3 and 92.0 %) and low vapor deficit pressure (from 38 10−6 MPa in August to 58 10−6 MPa in November). Annual average minimum and maximum temperatures were 14.9 °C and 18.0 °C, respectively. Monthly maximum temperature ranged from 17.1 °C in August to 19.1 °C in November. Accumulated solar radiation increased toward the end of the cropping season and ranged from 1,671 to 4,808 W m−2. Weather inter-annual variation was very low.

The soil is an Entisol, containing 46 % sand, 32 % silt and 22 % clay. Bulk density ranged from 1.3 to 1.2 and field capacity from 24 % to 21 % moisture by weight down the soil profile to 100 cm. Soil organic matter content was 3.1 % and pH was 7.1. Electrical conductivity (EC) and cation exchange capacity were 2.7 DS/m and 10.7 meq 100 g−1, respectively.

Different accessions were evaluated each year with the exception of the controls, cultivated in all years. Planting was done the first week of August, and harvest the last week of November. Germplasm of the three subsets was evaluated in three different trials, due to their different plant type and earliness. Thirty plants per clone, in three replications, in one row and 10 plants per replication were arranged in the field, following an alpha design in a RCBD. Fertilization and weed and pest control were the same for each group and year.

Irrigation frequency and volume were estimated to allow significant differences between treatments and accessions. In the irrigated treatment, water was applied to the crop every 10 days. In the drought treatment, furrow irrigation was stopped 35 days after planting in the case of improved varieties and 42 days after planting in the case of the genetic stocks and landraces, to take into consideration the quicker early growth of the improved germplasm. The total quantity of water received by the crop over the 5 years varied between 375 and 414 mm for the irrigated treatment and 128 and 175 mm in the drought treatment. Soil water potential was measured every day in each treatment using granular matrix sensors (Watermark model 900 M-Monitor, the Irrometer Company, Riverside-California, USA), placed at 30 and 45 cm depth in three sites. Data were recorded every day at 2:00 p.m. with a data logger. Four weeks after stopping irrigation, soil water potential in the fully irrigated treatment ranged from 0.022 to 0.063 MP and exceeded by 0.2 MPa the drought treatment.

Measurements

Improved varieties that had a much shorter growing cycle were harvested 90 days after planting, while genetic stocks and landraces were harvested 120 days after planting. For each plot were measured tuber yield (g plant−1), tuber number (per plant) and average tuber weight (g). Tubers were collected and dried for the determination of tuber dry matter (%). For each accession, a Drought Susceptibility Index (DSI) was calculated as \( {\text{DSI = }}\left( {{\text{1 - Yd/Yw}}} \right){\text{/D}} \), where Yd=mean yield of the accession under drought, Yw=mean yield of the same accession under well-watered conditions, and D=environmental stress intensity (Fischer and Maurer 1978).

Experimental data were subject to ANOVA using SAS-Proc Mixed version 9.1 for WINDOWS in order to adjust the year effect on the accessions.

Results

Variance analysis (Table 1) revealed significant accession and treatment effects for tuber yield, tuber number, average tuber weight, and tuber dry matter. Year and subset effects were not significant. The interaction between treatments and accessions was highly significant. The interaction between subsets and treatments was also significant, except for tuber number. The interaction between years and treatments was not significant.

Table 1 Variance analysis for tuber yield, tuber number, tuber weight and tuber dry matter in the whole collection
Full size table

Average tuber yield under irrigated conditions was 813.7, 392.1 and 475.9 g in improved varieties, genetic stocks and landraces, respectively (Table 2). Under drought conditions, tuber yield decreased to 420.2, 223.0, and 299.7 g in improved varieties, genetic stocks and landraces, respectively. Highly significant correlations were noted between tuber yield under irrigated and drought conditions (r = 0.782, P < 0.001; r = 0.426, P < 0.001; and r = 0.742, P < 0.001) in improved varieties, genetic stocks and landraces, respectively. Within each subset, year effects were low or not significant. Accession and treatment effects were highly significant. The interaction between treatment and accession was significant, except for tuber number in the subset of genetic stocks.

Table 2 Variance analysis for tuber yield, tuber number, tuber size and tuber dry matter in the different subsets comprising the whole collection (improved varieties, genetic stocks and landraces)
Full size table

In landraces, a highly significant group effect was found on tuber yield, number, size and dry matter (Table 3). The Andean S. tuberosum cultivar groups Andigenum and Chaucha, the species S. curtilobum and the non Andean S. tuberosum accessions had higher tuber yield in average than S. ajanhuiri and S. juzepczukii, both under irrigated and drought conditions (Table 4). These last species had higher tuber number than S. curtilobum and the S. tuberosum cultivar groups Andigenum, Phureja and Stenotomum, both under irrigated and drought conditions. The highest tuber weights were noted in S. curtilobum and the non Andean S. tuberosum and the lowest in S. ajanhuiri. The highest tuber dry matter was observed in S. juzepczukii and the lowest in the S. tuberosum cultivar group Andigenum.

Table 3 Variance analysis for tuber yield, tuber number, tuber weight and tuber dry matter in landraces
Full size table
Table 4 Ranking of cultivar groups for tuber yield, tuber number, tuber weight and tuber dry matter
Full size table

A larger variation was noted for drought susceptibility index (DSI) in genetic stocks and landraces, compared to improved varieties (Fig. 1). Within the subset of improved varieties, higher DSI was found for the LBR and LTVR accessions, compared to the foreign varieties (Fig. 2a). Within the subset of genetic stocks, a higher DSI and a broader variation for this trait were noted among PD accessions (Fig. 2b). Among landraces, S. juzepczukii had the lowest DSI value (Fig. 2c). A broader variation for DSI was found in this group and to a lesser extent, in S. curtilobum and S. ajanhuiri.

Fig. 1
figure 1

Boxplot grouping for DSI in the whole collection

Full size image
Fig. 2
figure 2

Boxplot grouping for DSI in the subsets of improved varieties (a), genetic stocks (b) and landraces (c). LBR late blight resistant population; LTVR lowland tropic virus resistant population; ADG andigenum(1); CHA chaucha(1); AJH ajanhuiri(2); CUR curtilobum(3), JUZ juzepczukii(4), GON goniocalyx(1); PHU phureja(1); STN stenotomum(1); TBR tuberosum(5); for (BCT population and PD see explanations in the text) (1) Andean Solanum tuberosum cultivar groups; (2) Solanum ajanhuiri Juz. & Bukasov; (3) Solanum curtilobum Juz. & Bukasov; (4) Solanum juzepczukii Bukasov; (5) Non Andean Solanum tuberosum L

Full size image

Accessions combining high yield under irrigated conditions and low DSI were found mainly among improved varieties and landraces (Fig. 3). Table 5 provides a list of improved varieties, genetic stocks and landraces with tuber yield under irrigated conditions superior to their median, and DSI lower to their median. Among the checks, Revolucion had lower yield under drought and higher DSI than Compis, as expected. Achirana had the lowest DSI. The improved varieties possessing these traits (38 accessions) included 6 foreign varieties and 17 and 15 accessions extracted from the LBR and LTVR populations, respectively. Most of improved varieties had higher yield, both under irrigated and drought conditions, than the checks. The subset of genetic stocks (13) only involved BCT clones, characterized by low yield and low DSI. The landraces with high tuber yield under irrigated conditions and low DSI (114 accessions) included accessions from the cultivar groups Andigenum (58), Ajanhuiri (1), Chaucha (25), Curtilobum (5), Goniocalyx (7), Phureja (6), Stenotomum (7) as well as non Andean S. tuberosum accessions (5). They did not include S. juzepczukii accessions because of their low yield under irrigated conditions.

Fig. 3
figure 3

Plotting of accessions according to tuber yield per plant under irrigated conditions and drought susceptibility index within improved varieties (a), genetic stocks (b) and landraces (c). Genotypes with tuber yield under irrigated conditions superior to their median and DSI lower to their median (quadrant on the right and below part of the graphics) represent 16.8 %, 33.7 % and 18.5 % of the improved varieties, genetic stock and landraces, respectively. These genotypes are listed in Table 5. LBR late blight resistant population; LTVR lowland tropics virus resistant population; ADG andigenum(1); CHA chaucha(1); AJH ajanhuiri(2); CUR curtilobum(3), JUZ juzepczukii(4), GON goniocalyx(1); PHU phureja(1); STN stenotomum(1); TBR tuberosum(5); for BCT population and PD see explanations in the text. (1) Andean Solanum tuberosum cultivar groups; (2) Solanum ajanhuiri Juz. & Bukasov; (3) Solanum curtilobum Juz. & Bukasov (4) Solanum juzepczukii Bukasov; (5) Non Andean Solanum tuberosum L

Full size image
Table 5 Accessions combining high tuber yield under irrigated and low drought susceptibility index (ranked according to their irrigate yield) in each group
Full size table

Discussion

Highly significant differences for tuber yield, number of tubers, average tuber weight and tuber dry matter were found among the tested accessions, both under irrigated and drought conditions. The reduction of around 62 % of water supply resulted in reduction of 58, 80 and 71 % of tuber yield, tuber number, and average tuber weight, respectively. However, the effect of water shortage highly differed among the subsets of germplasm used in this study. Within the subset of improved varieties, the drought treatment resulted in a reduction of 48 % in tuber yield, 22 % in tuber number and 35 % in average tuber weight in comparison to full irrigation. Within the subset of genetic stocks, the reduction was 52 % in tuber yield, 30 % in tuber number and 30 % in average tuber weight. Finally, within the subset of landraces, water shortage resulted in 38 %, 20 % and 26 % reduction in tuber yield, number and weight, respectively, indicating lower yield loss under drought particularly in comparison with improved varieties. This increase in drought susceptibility observed from landraces to improved varieties could reflect a gradual narrowing of the genetic base.

As pointed out by Ober et al. (2004), it is important to differentiate between genotypes that have high yield under drought stress simply because of high inherent yield potential and those that also have greater drought tolerance per se, here estimated from the drought susceptibility index (DSI). In the present study, accessions with high yield under irrigated conditions also tended to yield well under drought conditions. This suggested that potential yield explained an important part of the productivity under drought stress, but also that yield under both irrigated and drought conditions depended on the adaptation of the different varieties (most of them adapted to high altitude environments) to the conditions of the testing location. Despite the general linear relationship between drought stress and irrigate tuber yields, there was sufficient dispersion to indicate that not all accessions responded similarly to drought. This was confirmed by the significant treatment by accession interaction for tuber yield in improved varieties, genetic stocks and landraces and the large variation for DSI across subsets and accessions.

Potato landraces cultivated in a large range of agro-ecological conditions and subjected to very diverse selection pressures (Ritter et al. 2008) had higher variation for DSI than improved varieties, confirming their value as a reservoir of potentially useful drought tolerance genes. Among improved varieties, higher DSI in LBR and LTVR accessions, compared to foreign varieties, is likely to be due to the high moisture selection pressure applied to screen for late blight and virus resistance in these populations (Bonierbale et al. 2003). Among genetic stocks, the PD accessions derived from a cross between Andean (Phureja) and non Andean S. tuberosum accessions (Ghislain et al. 2001) had higher DSI, but also higher variation for this trait than the BCT accessions obtained from a backcross population involving the wild potato species S. berthaultii Hawkes.

Comparing DSI and tuber yield under irrigated conditions permitted us to identify accessions combining high productivity and drought tolerance. The proportion of accessions with tuber yield under irrigated conditions superior to the median and DSI lower to their median, regarding to the total number of accessions tested, was slightly higher in landraces (18.5 %) compared to genetic stocks and improved varieties (17.6 and 16.8 %, respectively). Among native Andean potatoes, the highest proportion of accessions combining high yield under irrigated conditions and low DSI were noted in the S. curtilobum (33.3 %) and in the S. tuberosum cultivar groups Stenotomum (25.0 %), Andigenum (22.2 %) and Chaucha (22.1 %). The high proportion in these groups indicated the potential interest of landraces Andean potato to improve drought tolerance in S. tuberosum modern germplasm without affecting potential yield.

S. curtilobum derived from natural hybridization between S. juzepczukii Bukasov and Andigenum cultivars of S. tuberosum (Hawkes 1990) is considered by Ritter et al. (2008) as one of the most frost resistant group. Potatoes of the Andigenum group of S. tuberosum have a wide adaptation, being cultivated throughout the Andes from Venezuela to Northern Argentina (Huamán and Ross 1985) at altitudes between 1,950 and 4,050 m (Ochoa 2003). Sources of heat tolerance have also been identified in this group by Morozova and Volkova (1991). Their tetraploid genome allows for hybrid generation with improved S. tuberosum varieties (Kumar and Kang 2006). Therefore, accessions of the Andigenum cultivar group are good candidates for the introgression of drought tolerance traits into S. tuberosum modern varieties. Potatoes of the Andean cultivar group Stenotomum are cultivated in Central Peru and Bolivia between 3,200 and 4,000 m. Potatoes of the group Chaucha which resulted from hybridization between Andigenum and Stenotomum potatoes are cultivated in Central Peru and Bolivia from 3,300 to 4,000 m altitudes and have the same area of distribution as S. curtilobum.

Conversely, the proportion of accessions with tuber yield under irrigated conditions above the median and DSI lower than the median, with respect to the total number of accessions tested, was low in the the diploid groups Goniocalyx (15.6 %), Phureja (7.5 %), Ajanhuiri (6.7 %) and Juzepczukii (0 %). Goniocalyx is cultivated in Central and Southern Peru, between 3,100 and 4,000 m (Hawkes 1990). Phureja is characterized by the lack of tuber dormancy, a trait reflecting its adaptation to regions free from long periods of frost and drought, like the Andean valleys (Hawkes 1990). S. ajanhuiri is cultivated in high altitudes of Peru and Bolivia, and is resistant to frost. S. juzepczukii, a natural triploid hybrid between wild potatoes and Stenotomum potatoes, had the lowest potential yield and the highest drought susceptibility values. It is cultivated at very high altitudes (3,800 to 4,000 m) in Central Peru and Bolivia.

The average drought tolerance level of the different groups did not appear to be directly related to their eco-geographic distribution. Both Andigenum and Phureja cultivar groups have a wide area of distribution and are adapted to low altitudes and milder climates. Phureja had, however, higher drought susceptibility index and lower yield potential, compared to Andigenum. Similarly, S. curtilobum, S. ajanhuiri and S. juzepczukii which are cultivated at similar altitudes in Central Peru to Bolivia, had very different drought tolerance levels.

Drought tolerance appeared more related to level of ploidy and evolution of potato. The polyploid species S. curtilobum (5x) and polyploid S. tuberosum groups Andigenum (4x) and Chaucha (3x) had higher drought tolerance and potential yield than the diploid species and S. tuberosum groups. Stenotomum, the most primitive form (Hawkes 1990) was the most tolerant among the diploid groups evaluated. Chaucha and Curtilobum groups could have inherited drought tolerance from their progenitors. The Chaucha cultivar group derived from natural hybridization between Andigenum and Stenotomum potatoes, while S. curtilobum results from a cross between S. juzepczukii Bukasov and tetraploid cultivars of S. tuberosum (Hawkes 1990).

In the present study, the evaluation conducted in the absence of rainfall allowed a perfect control of the quantity of water brought to the crop. However, differential or poor adaptation of the germoplasm samples to lowland conditions could have slightly biaised the evaluation, particularly in the case of landraces, all of which originated from high altitude areas. Further multi-year and multi-locational evaluation of the tested germplasm will permit to confirm the present results, describe the adaptation patterns of the tolerant germplasm and confirm the potential interest of some accessions as potential sources of drought tolerance in breeding programs. We also started to evaluate the collection used in this study for several physiological drought tolerance related traits. The associations that will be observed between traits and yield components will allow identifying traits that could be used for indirect selection. As the accessions tested here have already been genotyped (Ghislain et al. 2004), the results obtained could also be used for genetic association work.