Deciphering performances of fifteen genotypes of Stevia rebaudiana in southwestern France through dry biomass and steviol glycoside evaluation

Industrial Crops & Products xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Industrial Crops & Products journal homepage: www.elsevier.com/locate/indcrop

Deciphering performances of fifteen genotypes of Stevia rebaudiana in southwestern France through dry biomass and steviol glycoside evaluation Cécile Hastoya,b, Patrick Cossona, Sébastien Cavaignacc, Philippe Boutiéb, Pierre Waffo-Teguod, ⁎ Dominique Rolina, Valérie Schurdi-Levrauda, a

Univ. Bordeaux INRA, UMR Biologie du Fruit et Pathologie, 1332, 71 avenue Edouard Bourlaux, 33883 Villenave d’Ornon cedex, France Oviatis SA, Le Bourg, 47150 Lacaussade, France c INVENIO, Domaine de Lalande, 47110 Sainte Livrade sur Lot, France d Univ. Bordeaux, UFR des Sciences Pharmaceutiques, Unité de Recherche Œnologie, EA 4577, USC 1366 INRA, ISVV, 210 Chemin de Leysotte, 33882 Villenave d’Ornon cedex, France b

ARTICLE INFO

ABSTRACT

Keywords: Stevia rebaudiana Phenotypic diversity Genotypes Aerial biomass Yield Descriptors Steviol glycosides

Stevia rebaudiana (Bertoni) is a perennial shrub native of Paraguay whose leaves naturally accumulate steviol glycoside (SG) sweeteners. Optimization of S. rebaudiana semi-perennial cultivation in Europe is based upon breeding adapted high-value genotypes. The basis of a pre-breeding program is to evaluate in field conditions the performances of a large set of genotypes that could then be used as genitors for new varieties. This study investigates the performances of fifteen clones gathered from different origins in the second year of production under the environmental conditions of southwestern France. The performances were evaluated in terms of SG yield on the basis of dry leaf biomass yield, SG content and composition. The fifteen genotypes showed a very wide range of performances, expressing ratios of 15, 3.5 and 4 between the worst and best genotypes for SG yield, dry leaf biomass yield and SG content respectively. The sweetest SG (RebA) could represent 72% of total SG content in the best genotype. We confirmed the major role of dry leaf weight in SG yield. Five canopy descriptors out of ten, number of principal stems, stem density (number of stems/m² of canopy area), canopy volume (m3), Harvest Index (HI) and Specific Stem Mass (g/m2), were relevant to describe aerial biomass variability and classify the genotypes. This study thus provides biomass descriptors as tools for future breeding purposes and highlights the urgent need to evaluate the great genetic diversity of S. rebaudiana as a prerequisite for breeding purposes.

1. Introduction

80% of world stevia leaf production came from China, with 50,00060,000 tons of dry leaves per year (Sun, 2016) harvested on around 80,000 ha. Other significant producing countries are located in Asia (Indonesia, India, Japan, Korea) and in America (Mexico, USA, Canada, Paraguay, Argentina) (Gantait et al., 2018). Recent regulatory approval and the opening of markets explain the need for production in Europe (Commission Regulation (EU), 2011; EC, 2017). Many experimental studies have been carried out recently on Stevia production in Portugal (Lankes and Grosser, 2015), Germany (Lankes and Zabala, 2011; Munz et al., 2018), Denmark (Grevsen and Sorensen, 2016), Italy (Andolfi et al., 2006; Karimi et al., 2015; Macchia et al., 2007; Tavarini and Angelini, 2013), Greece (Zachokostas, 2015), Sweden (Vouillamoz et al., 2015) and France (Barbet-Massin, 2015; Hastoy et al., 2016,

For centuries, Paraguayan Indians have been using Stevia rebaudiana leaves as a natural acaloric source of sweetening (Soejarto et al., 1982). The active sweet compounds in this species are the glycosylated diterpenes Steviol Glycosides (SG; Bridel and Lavieille, 1931). S. rebaudiana is the only species in the Stevia genus which exhibits an intense and persistent sweet taste (Soejarto et al., 1982), because of high SG content in the leaves. These molecules are 250-300 times sweeter than sucrose (Ceunen and Geuns, 2013a) and are considered as a natural alternative to controversial synthetic sweeteners. The global market is constantly increasing, leading to strong demand for S. rebaudiana leaves for direct use or SG extraction. In 2016,

Abbreviations: SG, steviol glycosides; ST, stevioside; RebA, rebaudioside A; RebC, rebaudioside C; DulA, dulcoside A; RebF, rebaudioside F; Rub, rubusoside; RebD, rebaudioside D; RebM, rebaudioside M; RebB, rebaudioside B; SB, steviolbioside; ACN, acetonitril; LMR, Leaf-Stem Mass Ratio; HI, Harvest Index; SLM, Specific Leaf Mass; SSM, Specific Stem Mass ⁎ Corresponding author. E-mail address: [email protected] (V. Schurdi-Levraud). https://doi.org/10.1016/j.indcrop.2018.09.053 Received 10 May 2018; Received in revised form 26 September 2018; Accepted 28 September 2018 0926-6690/ © 2018 Elsevier B.V. All rights reserved.

Please cite this article as: Hastoy, C., Industrial Crops & Products, https://doi.org/10.1016/j.indcrop.2018.09.053

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

2015). All these experiments have shown that Stevia production as a perennial crop is possible and economically viable in Europe (Ferrazzano et al., 2016). The identification of productive genetic resources that are adapted to European environmental conditions is clearly needed, however. Productivity of S. rebaudiana is evaluated by SG yield and composition. SG yield expressed in kg/ha or in t/ha depends on two main components: biomass production of dry leaves and their specific SG content according to their taste. S. rebaudiana dry leaf yield depends on genotype and genotype x environment. This leads to a wide range of yields from 710 to 7,420 kg dry leaf/ha in western countries (Parris et al., 2016; Taleie et al., 2012). As for many plants, plant biomass yield is determined by the efficiency of solar energy capture and the conversion of the captured solar energy into vegetative tissues (Demura and Ye, 2010). S. rebaudiana is highly sensitive to photoperiod. It is an obligate short-day species, with a critical day length of around 12-13 h leading to early flowering (Metivier et al., 1979; Ceunen and Geuns, 2013b). Long-day conditions (14h-16 h) are required to prolong vegetative growth and increase leaf yield (Ceunen and Geuns, 2013b; Metivier and Viana, 1979; Yoneda et al., 2017). Temperature regime ranging from 0–2 °C to 35 °C also plays a crucial role in SG content (Sumida, 1980). Lightly-textured and well-drained soil impacts biomass production because S. rebaudiana is sensitive to waterlogging (Angelini et al., 2018; Print et al., 2015) and to water deficit (Benhmimou et al., 2017). Nitrogen (Barbet-Massin et al., 2015a; Rodrigues et al., 2017) or nitrogen-phosphorus-potassium (Aladakatti et al., 2012; Pal et al., 2015) fertilization contributes positively to dry leaf yield. Planting date and plant density also affect leaf biomass production (Angelini et al., 2018; Angelini and Tavarini, 2014; Munz et al., 2018; Serfaty et al., 2013; Taleie et al., 2012). Increasing plant density to 40,000 plants/ha or 100,000 plants/ha gives a better dry leaf yield of 1.03 and 1.99 t/ha respectively (Taleie et al., 2012). A comparative study of three and two genotypes also showed that leaf yield is enhanced over the years until 2 years of production in France and 8 years in Italy, respectively (Andolfi et al., 2006; Barbet-Massin et al., 2016) This makes Stevia cultivation possible as a semi-perennial crop in Europe. Optimum production conditions and their effects on dry leaf yield have therefore been widely documented. But, canopy structure is poorly described by quantitative traits. Previous works on phenotypic diversity have revealed great variability (Abdullateef and Osman 2011). Tateo et al. (1998) classified the plant architecture in four categories. This has been simplified, in more recent studies, to two main architectures, either upright with several main stems or a bushy structure with many secondary axes on a limited number of main stems (BarbetMassin, 2015; Munz et al., 2018). However, very few biomass descriptors have been used to describe Stevia aerial architecture. Only Harvest Index, Dry weightleaf/Dry weightleaf + stem , or LMR, Leaf mass ratio, is commonly used. This descriptor has been demonstrated to vary between 37% and 58% and can affect biomass production (Tateo et al., 1998). Like plant biomass accumulation, total SG content depends highly on photoperiod, ontogeny, genotype and its interaction with the environment. A study carried out in southwestern France has shown that the phenotypic variability of 96 “Criola”-related genotypes during the first year of production led to a large range of SG content from 4.6 to 12.3 % w/wdryleaf (Barbet-Massin et al., 2016), while a study with 24 Brazilian and Paraguayan genotypes showed a wider range of SG content from 9.4 to 27.3 % w/wdryleaf (Montoro et al., 2013). During vegetative growth, SG content increases and reaches a maximum at budding stage (Barbet-Massin, 2015; Ceunen and Geuns, 2013b). Longday photoperiod improves SG accumulation by 30% compared to shortday conditions (Ceunen and Geuns, 2013b). For the second year of production, SG content could be enhanced by 12% to 152% according to the genotype (Barbet-Massin et al., 2016). Each SG has different taste characteristics according to its monosaccharide type, number and attachment position on its steviol aglycon

part (Ekman and Hall, 2016; Upreti et al., 2012). It has been demonstrated that RebA contributes to a sweet taste, whereas ST, RebC and DulA elicit a bitter aftertaste (Espinoza et al., 2014; Hellfritsch et al., 2012). With consumers demanding less bitterness and liquorice taste, the content of RebA (Fry, 2016) and also of minor SG RebM or RebD (Hellfritsch et al., 2012; Prakash et al., 2014) needs to be improved. A couple of studies have shown that SG profiles remain stable between different environments and years of production, indicating high genotypic determinism (Barbet-Massin et al., 2016). In S. rebaudiana, few improved varieties are available. Angelini et al. (2018) have recently inventoried 90 varieties of S. rebaudiana. Most of them are sold as heterogeneous seed populations, often produced through open pollination and improved through massal selection (Yadav et al., 2011). The mostly widely known and cultivated varieties are named “Criola”, “Eirete” and “Morita”. All these varieties and seed populations have been improved mainly for SG yield and SG taste through their rebaudioside A/stevioside ratio in and for environmental conditions that are very different from those in Europe. Only the “Gawi” genotype from the EUSTAS collection was selected in Germany for its adaptation to temperate climates (Zabala, 2011). In Europe, only a few genotypes from different origins have been evaluated in terms of productivity. Seven genotypes from the EUSTAS collection were evaluated in Alentejo, Portugal (Lankes and Grosser, 2015) and four of them were used in a Danish study (Grevsen and Sorensen, 2016). Ninety-six genotypes were evaluated in southwestern France (Barbet-Massin et al., 2016) but they were all randomly chosen from a population of “Criola” seeds. Therefore, there is an urgent need to evaluate the performances of a wide variety of genotypes in European conditions. The aim of the present study is to explore the performances of fifteen genotypes gathered from different origins under the soil and climate conditions of southwestern France. It was decided to evaluate their performances in the second year of production in order to avoid the crop establishment period. These performances were deciphered in terms of (1) aerial biomass components and the evaluation of ten descriptors, (2) SG quantification and composition through the identification and quantification of ten SG. 2. Materials and methods 2.1. Plant material Fifteen genotypes of different origins were collected for the trial (Table 1). The genotypes were gathered from different providers to try to collect the widest possible non-patented diversity. Three genotypes were provided by the EUSTAS gene bank (Hortilab, Telgte, Germany). These genotypes have already been evaluated in field conditions in Europe (Lankes and Grosser, 2015). Eleven genotypes came from the OVIATIS collection (Lacaussade, France) and were previously selected from providers in Argentina, Paraguay, Spain, and Israël. One genotype was an “Eirete” type. Genotypes from the EUSTAS gene bank and OVIATIS collection came from in vitro cuttings and were produced as clones under partially-regulated greenhouse conditions (22 °C-18 °C, shading when light intensity was above 500 W/m²) for 7 weeks. Acclimatisation started by 10 days with a saturated hygrometry level, followed by 2 weeks of gradual aeration. In vitro cuttings were transplanted to a Jiffy®7 pellet (42 mm diameter, Jiffy, France). All the genetic resources were planted on a private farm in Liposthey (44°17'56.9"N 0°53'14.7"W), South-West France. In this area, the soil type is black sandy soil, 85-90% coarse sand and 5-10% fine mud, with density of 1.3 t/m3. Soil characteristics were determined at 30 cm depth in 2017 (Table 2). The soil was airy with little organic matter. No fertilizer was added as 50 kg N/ha was judged to be adequate. Weather conditions from the beginning of the trial in June 2016 to the harvest in September 2017 are given in Table 3. All the 7-weekold plantlets were transplanted to the field at the end of June 2016. 2

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Table 1 Genetic resources of S. rebaudiana studied in field conditions in the South-West of France: provider, country of the provider, name of genetic resources, year of collection, type of plant material used and origin before selection when known. Provider

Country

Name

Year of collection

Plant material

Origin before selection

EUSTAS gene bank

Germany

OVIATIS collection

France

SteviaStore

Paraguay

C Gawi D E161718 E8 EspLac1 EspLac2 FP GF S6030-1 Lac4 Larrère Septo4 Septo5 Eirete

2016 2016 2016 2011 2011 2013 2013 2011 2014 2015 2013 2013 2014 2014 2015

In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings In vitro cuttings Cuttings

Colombia Unknown Paraguay Argentina Argentina Spain Spain Argentina Argentina Israel Argentina Argentina Argentina Argentina Paraguay

weeds were removed by hand at the foot of the plants. When Septoria leaf spot disease appeared, Score®250EC or Ortiva®25SC fungicides (Syngenta) were applied at 0.5 L/ha. Treatments were performed three times in 2016 and five times in 2017. Genotype performance was studied during the second year of production. The plant harvest was conducted at the flower budding stage on September 21st, 2017.

Table 2 Organic status and element composition at 30 cm depth in the sandy experimental field in Liposthey (France) in 2017 Organic status

(%)

Elements (g/kg)

Oligo-elements (mg/kg)

pH

Organic matter Total nitrogen

2.3 0.053

P2O5 K2 O

0.091 0.042

Water KCl

C/N ratio

25.6

MgO

0.076

EDTA Cu EDTA Mg EDTA Fe EDTA Zn

3.11 < 3.99 88.81 4.35

6.2 5.7

2.2. Yield components and canopy descriptors measurements Plant yield and architecture was assessed on 3 plants randomly selected per genotype and per block, among 21, at the flower budding stage on 21th September. The number of main stems, plant height and East-west and North-south diameters were measured. After measurement, whole plants were cut by hand around 10 cm above the ground, and aerial biomass harvested. Fresh leaves and stems were separated by hand before drying at 40 °C for 60 h. Dry material was weighed for each plant to obtain the dry leaf and stem weight variables used in the calculation of dry weight per plant. These six measured variables were used in the calculation of six variables, canopy area and volume, stem density (Clifton-Brown and Lewandowski, 2002), Harvest index (Barbet-Massin, 2015), Specific Leaf and Stem Mass (Valladares and Guzman, 2006) listed in Table 4. Harvest index, Dry weightleaf / Dry weightleaf + stem , has been called in other studies LMR for Leaf mass ratio (Barbet-Massin et al., 2015a; Rees et al., 2010).

Plants were transplanted manually on twelve rows. Rows are distant from 60 cm and 1 m successively. Each block consisted of three rows with 33 cm between plants in a row. The experiment consists of four randomised blocks. This resulted in a planting density of 3.75 plants/ m². Plastic mulching with drop-by-drop underground irrigation was installed. Irrigation was conducted for 1 h per day at 1 L/h per drip during summer, and was reduced to 30 min per day in autumn. Each genotype was repeated 21 times per block (7 plants x 3 rows). After 5.5 months of growth, all the plants were cut after the flowering stage in December 2016. A wintering veil (30 g/m²) was installed for crop protection over the winter until March 2017. Plant growth started in the middle of March 2017 and the regrowth rate was 93% on average (Table A1). In 2017, NovaTec® Solub 14-8-30 (Compo Expert) was supplied as fertigation at 40 kg nitrate/ha. During both growing years,

Table 3 Monthly weather records during the experimental period (June 2016-September 2017) in Liposthey (France): means of minimum and maximum temperature (°C), cumulative precipitation (mm), cumulative sunshine duration (h) and cumulative light intensity (kWh/m²) Year

Month

Minimum temperature mean (°C)

Maximum temperature mean (°C)

Precipitation mean (mm)

Sunshine duration mean (h)

Light intensity mean (kWh/ m²)

2016

June July August September October November December January February March April May June July August September

13.5 14.7 14.6 13.0 8.1 6.3 3.8 -0.1 4.7 6.9 5.5 12.1 15.8 16.2 15.2 11.6

23.5 26.6 29.1 27.1 20.0 13.9 13.4 8.7 14.4 16.9 19.6 23.9 26.8 26.1 27.8 22.3

92.8 10.6 9.6 69.6 17.8 73.4 15.2 29.8 95.4 75.8 34.6 105.6 122.2 43.6 46.8 104.2

NA 201 241.5 184.2 135.5 74.5 91 88.5 86.5 104.6 217.4 169 175.2 125.2 200.9 129.2

NA 175 164 118 78 41 36 39 53 86 149 160 167 147 156 103

2017

3

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Table 4 Quantitative variables used for the characterisation of S. rebaudiana phenotypic variability under Liposthey (France) field conditions. Variables used for multiple regression are in bold. Type

Variables

Unit

Biomass and architecture measured variable

Dryweightleaf Dryweightstem h

Number of stems Harvest Index Leaf mass ratio Canopy area

g g m m m nb % m2

Canopy volume

m3

Stem density Specific Leaf Mass Specific Stem Mass

stem/m2 g/m2 g/m2 g

Number of stem / Canopy area Dry weightleaf /Canopy area Dry weightstem /Canopy area Dry weightleaf × SGs contenti

SGs Contenti

% w /wDW leaf

SGs Contenttotal

% w /wDW leaf

i : ST , RebA , RebC , DulA , RebF Rub, RebD, RebM , RebB, SB 10 SGs SGs contenti/SGs contenttotal

NS

Ew

Biomass and architecture calculated variable

SGs calculated variable

%

SGs proportioni

2.3. Steviol glycoside extraction and quantification

Height North-south diameter East-west diameter Dry weightleaf / Dry weightleaf+stem NS 2

× 4 3

×

EW 2

×

NS 2

×

EW 2

×

h 2

explaining dry leaf weight variability among the ten desciptors in bold Table 4. The correlation matrix was constructed in order to select only independent variables (Spearman correlation coefficient r < 0.8) for multiple linear regression (Table A5). Then, the best linear model regression was obtained by exhaustive selection of independent predictors according to the Bayesian Information Criterion (Table A6), using the “leaps” package (Lumley and Miller, 2017). Heatmap on discriminant descriptors was constructed by clustering using the complete linkage method in the “stats” package (R Core Team, 2015) in order to reveal S. rebaudiana architecture. The mixed linear models described above were used to estimate genetic and environmental variance. The broad sense heritability was then estimated using the formula H2 =σ2g/[σ2g +σ2e/n], where σ2g is the genetic variance, σ2e the environmental variance and n the number of plants per accession. SG compositions within the genotype collection were analysed using Principal Component Analysis (PCA) with the “FactoMineR” (Le et al., 2008), and “missMDA” (Husson and Josse, 2016) packages to compute missing values. Hierarchical Clustering on Principal Component (HCPC) was applied to PCA results in order to classify the genotypes according to their SG profile. For data representation, the “ggplot2” (Wickham, 2009), “cowplot” (Wilke, 2017) and “extrafont” (Chang, 2014) packages were used.

Twenty mg of dried and ground leaf from whole leaf samples (Ceunen and Geuns, 2013b) was extracted with 2 mL of ultra-pure water at 80 °C during 2 h. Extracts were centrifuged (10 min, 14700 g) and the supernatant filtered (0.45 μm) before analysis. 5 μL of supernatant were injected into a C18 column (250 x 4.6 mm, 4 μm; Poroshell 120 EC-C18, Agilent, Germany) with a guard column on a Reversed Phase High-Performance Liquid Chromatography (RP-HPLC) system (Agilent LC1100, USA). SGs were eluted by a gradient phase of acetonitrile (HPLC grade, Sigma-Aldrich, France): (t0: 20% ACN; 1.5 – 11.5 m in. 20 – 50.7% ACN; 11.6 – 14.6 m in. 100% ACN; 14.7 – 26 min: 20% ACN, 26 min : stop), and with a flow rate of 1.5 mL/min. Ten SGs were detected at 202 nm (RebD, RebM, ST, RebA, RebC, RebF, DulA, Rub, RebB, SB) and previously identified by purified SG standard (Chromadex, USA). For each SG, a standard range between 5 and 1,000 ng/μL of purified standard was used to quantify each amount. Results were expressed as content per unit of leaf dry weight (% w/ wdryleaf) for each SG and total SGs, and as a proportion (%) of the content of each SG to total SG content (Table 4). 2.4. Statistical analysis The statistical analysis was performed using R software version 3.4.3 (R Core Team, 2015). The complete data frame was cleaned by deleting outliers using Bonferroni tests on Studentised residuals of linear models, using the “car” package (Fox and Weisberg, 2011). Mixed linear models from the “lme4” package (Bates et al., 2015) were used to study the continuous quantitative variables:

P ij = µ + genotype i + block j +

Description / Formula

3. Results 3.1. Evaluation of genetic resource performances by SG yield, dry leaf weight and SG content For each genotype, SG yield expressed in g/plant (Fig. 1A) was calculated from dry leaf weight and SG content (Table 4; Fig. 1B-C). A significant and strong genotype effect (p < 0.001; Table A2) was found for SG yield (Fig. 1A). Depending on the genotype, SG yields ranged from 0.9 g/plant for the “D” genotype to 13.9 g/plant for the “GF” genotype, corresponding to 15-fold variability. Taking into account the regrowth rate between the two years (Table A1), it corresponds to a 34.44 and 430.16 kg/ha for “D” and “GF” respectively. Multiple comparison analysis between all the genotypes revealed that SG yield for the worst group including “D”, “S6030-1” and “Eirete” ranged from 0.9 to 1.8 g/plant and was significantly different from the group including “EspLac1”,” EspLac2”, “FP” and “Gawi”, whose SG yields ranged from 5.7 to 8.2 g/plant. Finally, the best producer “GF” (13.9 g/plant) was significantly different from the rest of the genetic

ij

With Pij: continuous quantitative variables, μ the overall mean of the phenotypic data, “genotype” corresponds to the genetic differences among the genotypes, “block” accounts for the differences in microenvironmental conditions between the four blocks and “ε” is the residual. “Genotype” was considered as a fixed effect and “blocks” and interactions as random effects. The “car” package was used for Type II analysis of variance on mixed models (Fox and Weisberg, 2011). Marginal means and standard error were calculated on mixed models and significant differences between genotypes were determined on marginal means by Tukey’s Honestly Significant Difference (HSD) test with the “emmeans” package (Russell, 2018). Multiple regression analysis was used to identify descriptors 4

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Fig. 1. Yield component of S. rebaudiana genotypes during harvesting at flower budding stage obtained in Liposthey (France) field conditions: SG yield expressed in g/plant (A), dry leaf weigth expressed in g/plant (B) and SG content expressed in % w/w of leaf dry weight (C). Barplot represents Least-Squares Means of 4 randomized blocks, corresponding to a total of 12 plants per genetic resource, with standard deviation. Results of multiple comparison by Tuckey’s Honestly Significative Difference are indicated by letters, and the same letter shows no significant difference at p = 0.05 level.

resources. In order to decipher the SG yield, the genotypes were also compared according to their dry leaf weight expressed in g/plant (Fig. 1B). Dry leaf weight showed 4-fold variability between the genotypes studied. It ranged from 20.5 to 91.5 g of dry leaf/plant. Among the genetic ressources studied, only “GF” exhibited a significant difference from the others. In our field conditions, this genotype was the best leaf biomass producer with around 3,000 kg of dry leaf/ha, taking into account the regrowth rate (Table A1). The other genotypes exhibited a continuum from low leaf biomass production of 770 kg of dry leaf/ha for “D” to 2,139 kg of dry leaf/ha for “Esplac2”. Except for “FP” and “EspLac1”, the distribution of the genotypes was the same for dry leaf weight (Fig.1B) and SG yield (Fig. 1A). Among the genetic resources studied, total SG content varied between 4.5 and 18.4 % w/wdryleaf (Fig. 1C), corresponding to wide variability in S. rebaudiana. A significant and strong genotype effect (p < 0.001; Table A2) was found. For this trait, “EspLac1” presented significantly higher SG content (18.4 % w/wdryleaf) and was

significantly different from the rest of the genetic resources, except “FP” and “GF” with SG amounts of 15.1 and 14.33 % w/wdryleaf respectively. Six genotypes (“Lac4”, “Gawi”, “E161718”, “Septo4”, “Septo5” and “Larrere”) presented intermediate SG content ranging between 11.8 and 10.2 % w/wdryleaf. Three genotypes (“D”, “S6030-1” and “Eirete”) had the lowest SG content with less than 7 % w/wdryleaf (Fig. 1C). It is interesting to note that the genotype distribution for SG yield was partly maintained for SG content. “D”, “S6030-1” and “Ereite” appeared to be low performance genotypes. “GF”, which was the best genotype for SG yield, was also one of the best in terms of SG content. “FP” and “EspLac1” balanced out their SG yield by their high SG content. A significant positive Spearman’s correlation of 0.87 (p < 0.001) was found between SG yield and dry leaf weight (Fig. 2A), whereas a lower Spearman’s correlation coefficient of 0.66 (p < 0.001) was found between SG yield and SG content (Fig. 2B). Moreover, the lowest correlation was observed between dry leaf weight and total SG content (r = 0.33; Fig. 2C). Variance analysis on a linear regression of SG yield 5

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Fig. 2. Correlation between yield component during Stevia rebaudiana harvesting at flower budding stage : SGs and dry leaf weight (A), SG yield and SG content (B), dry leaf weight and SG content (B). Each point corresponds to 1 harvested plant, and the plots represent all the genetic resources (12 plants per genetic resource). Significant Spearman correlation coefficient at p = 0.05 level is indicated on the bottom (r). Red line represents resulting linear correlation.

Fig. 3. (in color) Main morphotypes of S. rebaudiana observed in a genetic resources collection in Southwestern France field conditions. (A) Genotype “Gawi”, from the EUSTAS collection, exhibited compact habit with large leaves. (B) Genotype “Eirete”, exhibited an intermediate architecture. (C) Genotype “C”, from the EUSTAS collection, exhibited a large canopy area. White rectangles represent 10 cm. Photo credit: C. Hastoy, Liposthey (France).

revealed that 68% of the variance was explained by leaf dry weight, 24% by SG content and 8 % by their interaction (Table A3).

regressors allowed the genetic resources to be classified into three main clusters (Fig. 4). Genetic resources such as “Gawi” belong to Group 1 and produce a condensed foliar biomass on an intermediate size ground area (Fig. 3A). The genotypes belonging to group 2 “Larrere” and “Eirete” (Group 2; Fig. 3B) correspond to airy biomass on a large ground area and low stem density. The genotypes belonging to the third group show intermediate stem densitie on small area leading to a weak canopy volume as “C”. (Group 3; Fig. 3C). Interestingly, “Gawi”, “EspLac2” and “Septo4” expressed the best dry leaf weight in our field conditions and were ranked in the three different groups, groups 1, 2 and 3 respectively.

3.2. Variables involved in S. rebaudiana aerial biomass variability In our field growth conditions, three main architectures were observed, ranging from compact plants with large leaves such as “Gawi” to airy plants with small leaves on many lateral branches such as “C” (Fig. 3). All the intermediate architectures could also be observed, such as “Eirete" (Fig.3). Thus, to understand what contributes to dry leaf weight, ten quantitative yield and canopy descriptors were determined to assess morphotype (Table 4, in bold). All ten canopy descriptors contributed significantly to dry leaf weight variability at p < 0.05 level (Table A4). After correlations evaluation (Table A5) and exhaustive selection (Table A6), the best linear model obtained had a high adjusted R2 = 0.8312 (Table A7) and was composed of 5 regressors involved in dry leaf weight variability:

y = 0.697

0.005456

+ 370.64 + 1.408 + 0.081

3.3. Evaluation of SG profiles SG composition needs to be characterised in order to evaluate the economic potential of the genotype according to the quality of SGs linked to sweet taste (RebA, RebD, RebM, RebB) versus bitter taste (ST, RebC, DulA, Rub, SB). Principal Component Analysis was used to characterise SG composition on raw data, corresponding to 12 plants per genotype (Fig. 5A). Among the 10 detected SGs, the proportions of the 5 SGs produced the most (ST, RebA, RebC, RebF and DulA) were enough to explain 98.98% of the phenotypic variability (Fig. 5A). SG composition identified a significant correlation between the major SGs (Fig. 5D). ST and DulA proportions were positively correlated (r = 0.62; p < 0.05) and highly negatively correlated to the proportion of RebA (r = - 0.87 and r = - 0.83 respectively; p < 0.05), while the RebF and RebC proportions are positively correlated (r = 0.98; p < 0.05) and negatively correlated to the proportion of ST (r = - 0.61 and r = - 0.57 respectively; p < 0.05). Interestingly, the proportions of these 5 SGs exhibited non-significant correlations with yield components such as dry leaf weight and total SG content, or a significant

89.534

where y is dry leaf weight expressed in g, is the number of principal stems, is stem density expressed in number of stems/m² of canopy area, is canopy volume expressed in m3, is Harvest Index expressed as the % of foliar biomass to whole aerial biomass, is Specific Stem Mass expressed in g/m². These five regressors therefore constitute quantitative variables allowing dry leaf weight variability to be described. Broad-sense heritability H2 was calculated for each descriptor in order to evaluate the genotypic contribution to phenotypic variance. All the H2 values were high, from 0.36 to 0.94 (Table 5). Their high levels indicate a large contribution of the genotype to these descriptors. The complete linkage method on these previously-identified 6

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Table 5 Estimates of broad-sense heritability (H2) of architectural traits for S. rebaudiana grown in Liposthey fields (France). Descriptor H2

Number of principal stems nb 0.60

Stem density number of stems/m² of canopy area 0.36

Canopy volume 3

m 0.55

Harvest Index (LMR)

Specific Stem Mass (SSM)

% of foliar biomass among whole aerial biomass 0.94

g/m² 0.92

Fig. 4. Classification of S. rebaudiana genetic resources according to morphotypes linked to foliar biomass production. Heatmap was constructed from a centered matrix calculated on marginal means of independent canopy descriptors. For each variable (column), the grey gradient represents the result for genetic resources (row) from lower (light grey) to higher (black) values. SSM: Specific Stem Mass; HI (LMR): Harvest Index (Leaf Mass Ratio).

correlation but with a low correlation coefficient (r < 0.5; Fig. B1). These results reveal the independence of yield components and SG profile phenotypes in S. rebaudiana. Hierarchical Clustering on Principal Component allowed us to identify 6 clusters of genotypes (Fig. 5B-C). The genotypes were mainly separated along the first axis (Fig. 5 B) which corresponds to the RebA/ ST split. The “C” genotype appeared to be the best in terms of sweet taste SGs with 72% of RebA and a RebA/ST ratio of 3.9. It is interesting to note that this genotype also showed a high level of minor sweet SGs RebM and RebD (2.5%; Fig. 5E), which may contribute to its taste quality. The other genotypes were discriminated by their RebA/ST ratio (Fig. 5B), ranging from 1.47 to 0.87 for cluster 2, 0.97 to 0.5 for cluster 3. Only “Eirete” (cluster 4) had 20% of RebC. “EspLac2” (cluster 5) showed a large proportion of ST (84%). The genotypes from cluster 6 (“E8” and “S6030-1”) did not have any RebA, RebF and RebC. In our field growth conditions, genotypes belonging to cluster 1 and 2 showed the greatest economic potential, thanks to their higher proportion of RebA. However, minor sweet SGs RebM and RebD of around 1.5% (Fig. 5E) could also contribute to taste quality for the “D”, “Larrere”, “Septo5”, “Esplac1”, “FP”, “GF” and “Esplac2” genotypes.

4. Discussion and conclusion For farmers, the main productivity trait of S. rebaudiana is SG yield. Improved varieties with enhanced SG yield and adapted to various environments must therefore be proposed to farmers. The SG yield trait encompasses dry leaf weight, SG content and SG composition, which have to be evaluated in large genotype sets. In our study, SG yield was evaluated in a set of fifteen different genotypes in southwestern France in the second year after planting. These genotypes were gathered from different providers in order to obtain diversified origins. The fifteen genotypes revealed the high phenotypic variability of S. rebaudiana, in terms of productivity through SG yield (34-430 kg/ha), dry leaf yield (770-3,000 kg/ha) and SG content (4.47-18.41 % w/ wdryleaf). In France, 3 genotypes selected from a “Criola” heterogeneous seed population were evaluated on 3 sites (Barbet-Massin et al., 2016). They revealed SG yields of between 55 and 292 kg/ha. In the western United States, SG yield varied between 260 kg/ha and 890 kg/ha for 6 improved genotypes from S&W company cultivated in 4 locations (Parris et al., 2016). Taken together, these results show the role of breeding improvement. The “C” and “Gawi” genotypes from the EUSTAS collection have been evaluated in several places in Europe. Thus they can be used as benchmarks for SG productivity according to 7

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

(caption on next page)

8

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Fig. 5. (in color) Characterisation of SG profiles on 15 genotypes of S. rebaudiana in Southwestern France field condtions. (A) Principal Component Analysis (PCA) variables factor map of 5 major SG proportions (DulA : dulcoside A; RebA : rebaudioside A; RebC : rebaudioside C; RebF : rebaudioside F; ST : stevioside). The 5 SG proportions are expressed as a percentage of total SG content. The two major principal components explain 98.98 % of the variance. (B) PCA individual factor map from 15 genotypes according to the variable factor map. Each color represents 1 genotype. Ellipses reveal 6 clusters obtained by Hierachical Clustering on Principal Component (HCPC). (C) SG proportion for 15 studied genotypes classified according to the HCPC results. Horizontal bars above histograms represent the clusters previously shown in Fig. 5B. (D) Pearson correlation matrix of 5 major SG proportions. Positive correlations are represented in green and negative correlations are in orange. Circle size is proportional to Pearson coefficient, indicated inside in bold. No significant correlation is represented by an empty cell inside the matrix (p > 0.05). (E) Proportion of minor SG profiles of 15 genotypes in field conditions according to the HCPC clustering results.

different sites. In our conditions, the “C” and “Gawi” genotypes from the EUSTAS collection reached 94 and 222 kg/ha respectively. These genotypes in experiments in Alentejo, Portugal, exhibited higher performances in SG yield, with 990 kg/ha and 1,150 kg/ha, respectively (Lankes and Grosser, 2015). In an experiment in Denmark, the “C” and “Gawi” genotypes also showed high performances in SG yield with 376 and 468 kg/ha, respectively (Grevsen and Sorensen, 2016). It is interesting to notice that under Danish conditions (Grevsen and Sorensen, 2016), the “C” genotype obtained higher dry leaf yield than “Gawi”, with 4.5 and 3.9 t/ha respectively. These results clearly illustrate the need to evaluate genotypes in different environments. It also highlights the difficulty of comparing S. rebaudiana dry leaf yields per hectare because of plant density. It can range from 37,500 plants/ha in the present study to 100,000 plants/ha (Munz et al., 2018; Vouillamoz et al., 2015). Kumar et al. (2014) comparing five densities per ha, between 37,000 and 110,000 plants /ha for non described genotypes showed that plant productivity is higher in weak density. Number of leaves and stems decrease with an increasing density, whereas LAI (ratio between leaf area to ground canopy area) and leaf dry yield increase. Therefore, expressing biomass productivity in g of dry leaf per plant seems to be more accurate. Our collection ranged from genotype “D” producing 20 g/plant to “GF” producing 91 g/plant. This plant productivity was then similar to what was observed for the “SW107” genotype with 75 g/plant in USA (Parris et al., 2016) or for “F” genotype with 83 g/plant in Denmark or for “Rebaudiana” genotypes in an experiment by (Munz et al., 2018) in Germany with 70 g/plant. Our results revealed great morphological variability. Quantitative canopy descriptors served to distinguish between the genetic resources according to three main aerial architectures. A first group of genotypes exhibits condensed foliar biomass on an intermediate ground area, with a condensed stem. The plant architecture of this first group seems to be the best architecture for greater dry leaf yield, compared to airy biomass on a large ground area, except for “Esplac2”. This result shows that different architecture can lead to similar yield. Similar plant architectures were compared in Germany (Munz et al., 2018). The authors showed that plants with broad leaves on an upright shape corresponding to a higher Leaf Area Index (LAI) led to larger leaf dry yield, compared to the plants with narrow leaves with horizontally oriented shoots. However, the authors concluded that the fraction of intercepted photosynthetically-active radiation was lower for condensed biomass architecture and higher LAI was linked to higher Specific Leaf Area (SLA; ratio of leaf area to leaf dry weight), corresponding to thinner leaves (Munz et al., 2018). These observations suggest that increased yield may be attributed to higher leaf photosynthesis response to irradiance at elevated LAI, as in improved maize hybrids (Dwyerl et al., 1991). This result could be explained by the positive relationship between SLA and photosynthetic rate demonstrated on 17 herbaceous species (Dubey et al., 2017). In our study, five significant canopy descriptors could be selected: number of principal stems/ha, stem density expressed in number of stems/m² of canopy area, canopy volume expressed in m3/ha, Harvest Index and Specific Stem Mass expressed in g/m². These canopy descriptors explaining aerial biomass variability exhibit high heritability. High heritabilites for Leaf yield, Leaf:stem ratio and Stevioside concentration had already been described in Brandle and Rosa (1992) in half-sib families grown in India. This makes these descriptors very

useful tools for breeding programmes. They have been already used in other species produced for their leaf biomass, such as in tobbaco (Maleki et al., 2011) or miscanthus (Clifton-Brown and Lewandowski, 2002). Besides leaf dry yield, SG content and composition are also important components of SG yield. Among the genotypes we studied, SG content varied between 4.5 and 18.4 % w/wdryleaf. This is similar to what was obtained for genotypes from “Criola” seeds studied in France over two years of production (Barbet-Massin et al., 2016). However, high-performance genotypes have also been described, such as the twenty-three genotypes from Brazil and Paraguay cultivated in Italy. They can accumulate 27% of SGs w/wdryleaf (Montoro et al., 2013). These differences can be partly attributed to genotype effect. However, it has also been described that SG content is highly influenced by the pedoclimatic conditions, the age of the plant stand and its phase of development (Barbet-Massin et al., 2016; Parris et al., 2017; Tavarini and Angelini, 2013). As demonstrated in previous studies (Barbet-Massin et al., 2016), five major SGs (ST, RebA, RebC, DulA and RebF) were enough to explain 98% of SG variability in our collection. A higher proportion of RebA and lower proportion of ST, RebC and DulA are desirable traits (Espinoza et al., 2014; Hellfritsch et al., 2012). Minor RebM and RebD are also highly targeted by breeders (Pure Circle, 2013). In our study, we observed one particularly interesting genotype, “C” from the EUSTAS collection, which accumulates 72% RebA among total SGs but also a significant amount of RebM and D. Four other genotypes (“Larrere”, “Septo5”, “EspLac1” and “FP”) appeared to be very interesting for SG profile improvement, as their RebA/ST ratio is above 1. We demonstrated that RebA proportion is negatively correlated to DulA and ST proportion, which is useful for breeding purpose. However, ST and RebC are negatively correlated. RebC elicitates bitter human receptors at lower concentrations than ST (Hellfritsch et al., 2012). It is involved in the maintenance of a bitter aftertaste. Therefore, selection clearly needs to minimise the proportion of RebC in SG profiles. SG composition is determined by genotype and is poorly influenced by the environment. Comparison of “C” genotype composition over sites and years demonstrates that SG profiles are stable. In the first year of production, the “C” genotype reached 71% of RebA among total SGs in a field experiment in France (Hastoy et al., 2016) and around 75% in Portugal (Lankes and Grosser, 2015). In the second year of production, “C” showed 72% of RebA in France and around 69% in Danish conditions (Grevsen and Sorensen, 2016). Results obtained in western France field conditions with a large set of fifteen different genotypes in the second year of cultivation confirmed that S. rebaudiana semi-perennial production is possible in France. This study confirmed that dry leaf weight is one of the main traits contributing to SG yield. Our diversity revealed high dry leaf weight per plant of as much as 91 g/plant, although our field design with low plant density of 37,500 plants/ha led to an intermediate yield per hectare. For the future, understanding the relationship between canopy structure, plant density, light interception and photosynthetic activity could improve knowledge of the leaf yield building pattern in S. rebaudiana for breeding purposes. According to the plant architecture, the canopy area and the plant density, an optimum has still to be found. In our conditions, the best dry leaf producer “GF” and “C”, 9

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

“EspLac1”, “Septo5”, “Larrere”, “D” and “FP” for the best SG composition could represent interesting genitors for breeding improvement of S. rebaudiana in southwestern France. This work highlights the interest to develop studies on enlarged collections.

Oviatis SA, France. Nouvelle-Aquitaine Region supported the work through Cifre support. The authors thank Justine Peirotte (Fraise Concept’, Douville, France) and the INVENIO team (Douville, SainteLivrade and Ychoux experimental sites, France) for plant production and cultural management, Pierre Jannot for support (Rouages, Agen, France) and all lab members for harvest help.

Acknowledgments C. Hastoy was supported by ANRT funding n° 2014/0915 and Appendix A .

Table A1 Dry leaf yield and SG yield expressed in kg/ha according to the regrowth rate between the two years of the experiment for the fifteen S. rebaudiana genotypes grown in Liposthey field (France) Genetic resources

Regrowth rate (%)

Dry leaf yield (kg/ ha)

SG yield (kg/ha)

D Eirete S6030-1 C E161718 E8 FP Larrère Lac4 EspLac1 Septo5 Septo4 Gawi EspLac2 GF

100 98 92 98 70 88 86 96 94 91 95 97 100 98 88

770.62 932.39 1020.11 1042.20 1142.63 1238.68 1275.37 1294.35 1362.01 1494.71 1506.53 1775.04 2022.99 2139.48 3001.15

34.44 64.24 57.06 94.37 120.60 122.63 192.96 132.40 160.95 275.24 156.74 187.49 222.88 212.60 430.16

Table A2 Probability values of ANOVA type II table on linear mixed model for SG yield and its component Variables

Genetic resources

SG yield Dry leaf weight SG content

p < 0.001 p < 0.001 p < 0.001

Table A3 Variance analysis for linear regression of SG yield by leaf dry yield and SG content Explicative variables

P-value

Variance proportion (%)

Dry leaf weight SG content Dry leaf weight : SG content

p < 0.001 p < 0.001 p < 0.001

68 24 8

Table A4 Result of ANOVA type II on linear model for dry leaf weight regression Explicative variables

P-value

h

p p p p p p p p p p

NS

EW

Number of stems Canopy area Canopy volume Stem density Harvest Index Specific Leaf Mass Specific Stem Mass

10

< < < < < < < < < <

0.001 0.001 0.001 0.01 0.01 0.001 0.01 0.001 0.001 0.001

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Table A5 Pearson correlation matrix of 10 canopy descriptors for selection of independant variables for multiple linear regression of dry leaf weight of S. rebaudiana

Black cross: non-significant correlation. Pearson correlation coefficient in black shows positive correlation and grey coefficient shows negative correlation. LMR, Leaf-Stem Mass Ratio; HI, Harvest Index; SLM, Specific Leaf Mass; SSM, Specific Stem Mass. Table A6 Exhaustive selection for dry leaf weight regression according to Bayesian Information Criterion (BIC)

LMR, Leaf-Stem Mass Ratio; HI, Harvest Index; SLM, Specific Leaf Mass; SSM, Specific Stem Mass. Table A7 Multiple linear regression of dry leaf weight: estimate coefficients, standard error, p-value and Variance Inflation Factor (VIF). Estimator Intercept Stem number Stem density Canopy volume HI (LRM) SSM

Estimate coefficient -89.534220 0.697269 - 0.0054564 370.644336 1.408996 0.081928

Standard error 7.065820 0.112679 0.011572 33.335757 0.097374 0.003638

Adjusted R² = 0.8312. BIC = 1238.865 LMR, Leaf-Stem Mass Ratio; HI, Harvest Index; SSM, Specific Stem Mass.

11

Pr

VIF -16

< 2x10 4.85 × 10-9 5.20 × 10-6 < 2x10-16 < 2x10-16 < 2x10-16

3.943557 3.941348 3.306356 1.982858 1.200936

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al.

Appendix B

Fig. B1. (in color) Pearson correlation matrix of SG yield, leaf dry yield, SG content and the 5 major SG proportion. Positive correlations are represented in green and negative correlations are in orange. Circle size is proportional to Pearson coefficient, indicated inside. No significant correlation is represented by an empty cell inside the matrix (p > 0.05).

Ekman, S., Hall, G., 2016. Flavour characterisation of Stevia products in water solution. In: Geuns, M.C. (Ed.), Proc. 9th Stevia Symp. 2016, EUSTAS, Sweden, 15-16th Sept. 2016, Jan, pp. 19–32. Espinoza, M.I., Vincken, J., Sanders, M., Castro, C., Stieger, M., Agosin, E., 2014. Identification, Quantification, and Sensory Characterization of Steviol Glycosides from Differently Processed Stevia rebaudiana Commercial Extracts. J. Agric. Food Chem. 62, 11797–11804. Ferrazzano, G.F., Cantile, T., Alcidi, B., Coda, M., Ingenito, A., Zarrelli, A., Di Fabio, G., Pollio, A., 2016. Is Stevia rebaudiana bertoni a non cariogenic sweetener? A review. Molecules 21, 1–12. Fox, J., Weisberg, S., 2011. An R Companion to Applied Regression, Second. ed. Sage publication, Thousand Oaks (CA). Fry, J.C., 2016. Stevia Innovation: Improved Leaf Extracts From Advanced Understanding of Taste. In: Geuns, M.C. (Ed.), Proc. 9th Stevia Symp. 2016, EUSTAS, Sweden, 1516th Sept. 2016, Jan, pp. 33–40. Gantait, S., Das, A., Banerjee, J., 2018. Geographical Distribution, Botanical Description and Self-Incompatibility Mechanism of Genus Stevia. Sugar Tech 20, 1–10. Grevsen, K., Sorensen, J.N., 2016. 2nd year cultivation results of the Danish “Green Stevia” project-a natural sweetener for organic food products. In: Geuns, M.C. (Ed.), From F. to Fork, Proc. 9th Stevia Symp. 2016, EUSTAS, Sweden, 15-16th Sept. 2016, Jan, pp. 115–126. Hastoy, C., Cosson, P., Sclaunich, E., Jannot, P., Boutié, P., Cavaignac, S., Rolin, D., Schurdi-Levraud, V., 2016. Organic stevia production and Development of Simple Sequence Repeat (SSR) and Single Nucleotide Polymorphism (SNP) Molecular Markers as a Tool for Stevia rebaudiana’s Germplasm Characterization and Future Breeding Purposes. In: Geuns, M.C. (Ed.), From F. to Fork, Proc. 9th Stevia Symp. 2016, EUSTAS, Sweden, 15-16th Sept. 2016, Jan, pp. 127–136. Hastoy, C., Cosson, P., Sclaunich, E., Jannot, P., Boutié, P., Rolin, D., Schurdi-levraud, V., 2015. Towards Breeding of new Stevia rebaudiana Cultivars adapted to Organic Production in the South West of France : First Results on Genetic and Phenotypic Evaluation. Proc. 8th Stevia Symp. 2015. EUSTAS, Univ., Bonn, Ger, pp. 225–232. Hellfritsch, C., Brockhoff, A., Staehler, F., Meyerhof, W., Hofmnann, T., 2012. Human Psychometric and Taste Receptor Responses to Steviol Glycosides. J. Agric. Food Chem. 60, 6782–6793. Husson, F., Josse, J., 2016. missMDA: A Package for Handling Missing Values with Multivariate Data Analysis. J. Stat. Softw. 70, 1–31. Karimi, M., Ahmadi, A., Hashemi, J., Abbasi, A., Tavarini, S., Guglielminetti, L., Angelini, L.G., 2015. The effect of soil moisture depletion on Stevia (Stevia rebaudiana Bertoni) grown in greenhouse conditions: Growth, steviol glycosides content, soluble sugars and total antioxidant capacity. Sci. Hortic. (Amsterdam). 183, 93–99. Kumar, R., Sharma, S., Sood, S., 2014. Yield components, light interception and marker compound accumulation of stevia (Stevia rebaudiana Bertoni) affected by planting material and plant density under western Himalayan conditions. Archives of Agronomy and Soil Science 60 (12), 1731–1745. Lankes, C., Grosser, P., 2015. Evaluation of Stevia rebaudiana genotypes at a location in the Alentejo region in Portugal. In: Geuns, J.M.C., Ceunen, S. (Eds.), Proc. 8th Stevia Symp. 2015, EUSTAS, pp. 167–176 Bonn, Ger. Lankes, C., Zabala, U., 2011. Evaluation of Stevia rebaudiana genotypes. In: Geuns, M.C. (Ed.), Proc. 5th Stevia Symp. 2011, EUSTAS, Belgium, Jan, pp. 75–87. Le, S., Josse, J., Husson, F., 2008. FactoMineR: A Package for Multivariate Exploratory Data Analysis and Data Mining. J. Stat. Softw. 25, 1–18. Lumley, T., Miller, A., 2017. leaps: Regression Subset Selection. [WWW Document]. R Packag. version 3.0.. . Macchia, M., Andolfi, L., Ceccarini, L., Angelini, L.G., 2007. Effects of Temperature, Light and Pre-Chilling on Seed Germination of Stevia rebaudiana (Bertoni) Bertoni Accessions. Ital. J. Agron. 1, 55–62. Maleki, H.H., Karimzadeh, G., Darvishzadeh, R., Sarrafi, A., 2011. Correlation and sequential path analysis of some agronomic traits in tobacco (’Nicotiana tabaccum’ L.) to improve dry leaf yield. Aust. J. Crop Sci. 5, 1644–1648. Metivier, J., Viana, A.M., 1979. The Effect of Long and Short Day Length upon the Growth

References Abdullateef, R.A., Osman, M., 2011. Influence of genetic variation on morphological diversity in accessions of Stevia rebaudiana Bertoni. International Journal of Biology 3 (3), 66–72. Aladakatti, Y.R., Palled, Y.B., Chetti, M.B., Halikatti, S.I., Alagundagi, S.C., Patil, P.L., 2012. Effect of nitrogen, phosphorus and potassium levels on growth and yield of stevia (Stevia rebaudiana Bertoni.). Karnataka J. Agric. Sci. 25, 25–29. Andolfi, L., Macchia, M., Ceccarini, L., 2006. Agronomic-productive Characteristics of Two Genotype of Stevia Rebaudiana in Central Italy. Ital. J. Agron 257–262. Angelini, L.G., Martini, A., Passera, B., Tavarini, S., 2018. Cultivation of Stevia rebaudiana Bertoni and Associated Challenges. Springer, Cham, pp. 35–85. Angelini, L.G., Tavarini, S., 2014. Crop Productivity, Steviol Glycoside Yield, Nutrient Concentration and Uptake of Stevia rebaudiana Bert. under Mediterranean Field Conditions. Commun. Soil Sci. Plant Anal. 45, 2577–2592. Barbet-Massin, C., 2015. Sélectionner et cultiver Stevia rebaudiana Bertoni en milieu tempéré : exploration de la variabilité de la teneur et de la composition en glycosides de stéviol. Thesis. pp. 211. Barbet-Massin, C., Giuliano, S., Alletto, L., Dayde, J., Berger, M., 2015a. Nitrogen Limitation Alters Biomass Production but Enhances Steviol Glycoside Concentration in Stevia rebaudiana Bertoni. PLoS One 10, 1–16. Barbet-Massin, C., Giuliano, S., Alletto, L., Dayde, J., Berger, M., 2016. Towards a semiperennial culture of Stevia rebaudiana (Bertoni) Bertoni under temperate climate: effects of genotype, environment and plant age on steviol glycoside content and composition. Genet Resour Crop Evol. 685–694. Bates, D., Maechler, M., Bolker, B., Walker, S., 2015. Fitting Linear Mixed-Effects Models using lme4. J. Stat. Softw. 1, 1–48. Benhmimou, A., Ibriz, M., Faïz, C. Al, Khiraoui, A., Amchra, F.Z., Lage, M., 2017. Productivity of new sweet plant in Morocco (Stevia rebaudiana Bertoni) under water stress. J. Med. Plant Stud. 5, 126–131. Brandle, J.E., Rosa, N., 1992. Heritability for yield, leaf:stem ratio and stevioside content estimates from a landrace cultivar of Stevia rebaudiana. Can. J. Plant Sci. 72, 1263–1266. Bridel, M., Lavieille, R., 1931. Le principe a saveur sucrée du Kaà-hê-é (Stevia rebaudiana, Bertoni). Académie des Sci. Paris Comptes Rendus 192, 1123–1125. Ceunen, S., Geuns, J.M.C., 2013a. Steviol Glycosides: Chemical Diversity, Metabolism, and Function. J. Nat. Prod. 76, 1201–1228. Ceunen, S., Geuns, J.M.C., 2013b. Influence of photoperiodism on the spatio-temporal accumulation of steviol glycosides in Stevia rebaudiana (Bertoni). Plant Sci. 198, 72–82. Chang, W., 2014. extrafont: Tools for using fonts [WWW Document]. R Packag. version 0.17. Clifton-Brown, J.C., Lewandowski, I., 2002. Screening miscanthus genotypes in field trials to optimise biomass yield and quality in southern Germany. Eur. J. Agron. 16, 97–110. COMMISSION REGULATION (EU), 2011. No 1131/2011 of 11 November 2011 amending Annex II to Regulation (EC) No 1333/2008 of the European Parliament and of the Council with regard to steviol glycosides. Off. J. Eur. Union amending A, L295/205 L295/211. Demura, T., Ye, Z.-H., 2010. Regulation of plant biomass production. Curr. Opin. Plant Biol. 13, 298–303. Dubey, P., Raghubanshi, A.S., Dwivedi, A.K., 2017. Relationship among specific leaf area, leaf nitrogen, leaf phosphorus and photosynthetic rate in herbaceous species of tropical dry deciduous in Vindhyan highlands. Ann. Plant Sci. 1531–1536. Dwyerl, L.M., Tollenaar, M., Stewartl, D.W., 1991. Changes in plant density dependence of leaf photosynthesis of maize (Zea mays L.) hybrids, 1959 to 1988. Can. J. Plant Sci. 71, 1–11. EC, 2017. Novel Food Catalogue, Stevia rebaudiana. [WWW Document]. Eur. Com. URL. http://ec.europa.eu/food/safety/novel_food/catalogue/search/public/index.cfm#.

12

Industrial Crops & Products xxx (xxxx) xxx–xxx

C. Hastoy et al. of Whole Plants and the Level of Soluble Proteins, Sugars, and Stevioside in Leaves of Stevia rebaudiana Bert. J. Exp. Bot. 30, 1211–1222. Montoro, P., Molfetta, I., Maldini, M., Ceccarini, L., Piacente, S., Pizza, C., Macchia, M., 2013. Determination of six steviol glycosides of Stevia rebaudiana (Bertoni) from different geographical origin by LC–ESI–MS/MS. Food Chem. 141, 745–753. Munz, S., Praeger, A., Merkt, N., Claupein, W., Graeff-Hoenninger, S., 2018. Leaf area index, light interception, growth and steviol glycoside formation of Stevia rebaudiana Bertoni under field conditions in southwestern Germany. Ind. Crops Prod. 111, 520–528. Pal, P.K., Kumar, R., Guleria, V., Mahajan, M., Prasad, R., Pathania, V., Gill, B.S., Singh, D., Chand, G., Singh, B., Singh, R.D., Ahuja, P.S., 2015. Crop-ecology and nutritional variability influence growth and secondary metabolites of Stevia rebaudiana Bertoni. BMC Plant Biol. 15, 1–16. Parris, C.A., Shock, C.C., Qian, M., 2017. Soil Water Tension Irrigation Criteria Affects Stevia rebaudiana Leaf Yield and Leaf Steviol Glycoside Composition. HortScience 52, 154–161. Parris, C.A., Shock, C.C., Qian, M., 2016. Dry Leaf and Steviol Glycoside Productivity of Stevia rebaudiana in the Western United States. HortScience 51, 1220–1227. Prakash, I., Bunders, C., Devkota, K.P., Charan, R.D., Ramirez, C., Priedemann, C., Markosyan, A., 2014. Isolation and characterization of a novel rebaudioside M isomer from a bioconversion reaction of rebaudioside A and NMR comparison studies of rebaudioside M isolated from Stevia rebaudiana Bertoni and Stevia rebaudiana Morita. Biomolecules 4, 374–389. Print, I., Pordel, R., Esfahani, M., Kafi, M., Nezami, A., 2015. Response of Stevia rebaudiana Bertoni root system to waterlogging and terminal drought stress Department of Agronomy Faculty of Agriculture, Ferdowsi university of Mashhad, Mashhad. Iran. J. Biodivers. Environ. Sci. 6, 238–247. Circle, Pure, 2013. PureCircle to start producing stevia based high-purity Reb D sweetener. Int. Sugar J. 115, 298. R Core Team, 2015. R: A language and environment for statistical computing. R Found. Stat. Comput., Vienna, Austria [WWW Document]. Rees, M., Osborne, C.P., Woodward, I., Hulme, S.P., Turnbull, L.A., Taylor, S.H., 2010. Partitioning the Components of Relative Growth Rate: How Important Is Plant Size Variation? The American Naturalist 176, 6. Rodrigues, M.A., Afonso, S., Ferreira, I.Q., Arrobas, M., 2017. Response of stevia to nitrogen fertilization and harvesting regime in northeastern Portugal. Arch. Agron. Soil Sci. 63, 626–637. Russell, L., 2018. emmeans: Estimated Marginal Means, aka Least-Squares Means. [WWW Document]. R Packag. version 1.1.. . Serfaty, M., Ibdah, M., Fischer, R., Chaimovitsh, D., Saranga, Y., Dudai, N., 2013. Dynamics of yield components and stevioside production in Stevia rebaudiana grown under different planting times, plant stands and harvest regime. Ind. Crops Prod. 50,

731–736. Soejarto, D.D., Kinghorn, A.D., Farnsworth, N.R., 1982. Organoleptic evaluation of Stevia leaf herbarium samples for sweetness. J. Nat. Prod. 45, 590–599. Sumida, T., 1980. Studies on Stevia rebaudiana Bertoni as a possible new crop for sweetening resource in Japan. J. Cent. Agric. Exp. Stn. 1–71. Sun, J., 2016. Development of Stevia and Steviol Glycosides Industry in China. In: Geuns, M.C. (Ed.), Proc. 9th Stevia Symp. 2016, EUSTAS, Sweden, 15-16th Sept. 2016, Jan, pp. 145–150. Taleie, N., Hamidoghli, Y., Rabiei, B., Hamidoghli, S., 2012. Effects of plant density and transplanting date on herbage, stevioside, phenol and flavonoid yield of Stevia rebaudiana Bertoni. Intl J Agri Crop Sci 4, 298–302. Tateo, F., Mariotti, M., Bononi, M., Lubian, E., Martello, S., Cornara, L., 1998. Stevioside content and morphological variability in a population of Stevia rebaudiana (Bertoni) Bertoni from Paraguay. Ital. J. Food Sci. 10, 261–267. Tavarini, S., Angelini, L., 2013. Stevia rebaudiana Bertoni as a source of bioactive compounds: The effect of harvest time, experimental site and crop age on steviol glycoside content and antioxidant properties. J. Sci. Food Agric. 93, 2121–2129. Upreti, M., Dubois, G., Prakash, I., 2012. Synthetic study on the relationship between structure and sweet taste properties of steviol glycosides. Molecules 17, 4186–4196. Valladares, F., Guzman, B., 2006. Canopy structure and spatial heterogeneity of understory light in an abandoned Holm oak woodland. Ann. For. Sci. 63, 749–761. Vouillamoz, J., Wolfram-Schilling, E., Carron, C., Baroffio, C., 2015. Agronomic and phytochemical evaluation of Stevia rebaudiana genotypes. In: Geuns, J.M.C., Ceunen, S. (Eds.), Proc. 8th Stevia Symp. 2015, EUSTAS, pp. 53–56 Bonn, Ger. Wickham, H., 2009. ggplot2 : Elegant Graphics for Data Anamysis. Springer-Verlag, New York. Wilke, C.O., 2017. cowplot: Streamlined Plot Theme and Plot Annotations for “ggplot2.” R Packag. cowplot version 0.9.2.. . Yadav, A.K., Singh, S., Dhyani, D., Ahuja, P.S., 2011. A review on the improvement of stevia [Stevia rebaudiana (Bertoni)]. Can. J. Plant Sci. 91, 1–27. Yoneda, Y., Shimizu, H., Nakashima, H., Miyasaka, J., Ohdoi, K., 2017. Effects of light intensity and photoperiod on improving steviol glycosides content in Stevia rebaudiana (Bertoni) Bertoni while conserving light energy consumption. J. Appl. Res. Med. Aromat. Plants 7, 64–73. Zabala, U.M., 2011. Optimierung von Wachstum und Ertrag (Süßstoffbildung) bei Stevia rebaudiana Bertoni unter mitteleuropäischen Standortbedingungen. Rheinischen Friedrich-Wilhelms-Universität, Bonn Thesis. Zachokostas, P., 2015. Effects of Chemical Weed Management on the Quantitative and Qualitative Characteristics of Stevia rebaudiana Bertoni, under Mediterranean Conditions (Karditsa, Thessaly, Greece). In: Geuns, M.C. (Ed.), Proc. 8th Stevia Symp. 2015, EUSTAS, Ger. 27-28th January 2015, Jan, pp. 5–12.

13