Heat Tolerance of Durum Wheat (Tritcum durum Desf.) Elite Germplasm Tested along the Senegal River

The Senegal River basin (Guinea, Mali, Mauritania, and Senegal) is a key agricultural production area in sub-Saharan Africa. Here, rice fields are left fallow during the cooler winter season, when the night temperatures reach 16 °C but the maximum daily temperatures remain above 30 °C. This season was used for the first time to conduct multi-environmental trials of durum wheat. Twenty-four elite breeding lines and cultivars were tested for adaptation during seasons 2014-15 and 2015-16 at two stations: Kaedi, Mauritania and Fanaye, Senegal. Phenological traits, grain yield and its components were recorded. Top grain yield was recorded at 5,330 kg ha and the average yield at 2,484 kg ha. The season lasted just 90 days from sowing to harvest. Dissection of the yield in its components revealed that biomass and spike fertility (i.e. number of seeds produced per spike) were the most critical traits for adaptation to these warm conditions. This second trait was confirmed in a validation experiment conducted in 2016-17 at the same two sites. Genotype × environment interaction was dissected by AMMI model, and the derived IPC values used to derive an ‘AMMI wide adaptation index’ (AWAI) to asses yield stability. The use of a selection index that combined adjusted means of yield and AWAI identified three genotypes as the most stable and high yielding: ‘Bani Suef 5’, ‘DAWRyT118’, and ‘DAWRyT123’. The last two genotypes were also confirmed among the best in a validation trial conducted in season 2016-17. The data presented here are meant to introduce to the breeding community the use of these two research stations along the Senegal River for assessing heat tolerance of wheat or other winter cereals, as well as presenting two new ideal germplasm sources for heat tolerance, and the identification of spike fertility as the key trait controlling adaptation to heat stress.


Introduction
The area along the Senegal River represents a major agricultural basin in Sub-Saharan Africa with the potential of 375,000 ha of arable and irrigated land.Today, a portion corresponding to approximately 200,000 ha is intensively cultivated with double seasons of rice (FAO, 2016).However, the cool season between middle of November to early March is not suitable for rice cultivation, and fields are mostly left at fallow.Preliminary Here is presented an attempt to investigate further the suitability of durum wheat production as a replacement for the fallow season by means of full scale breeding trials conducted over locations and years.The adaptability of a variety over a diverse environment is usually tested by the degree of its interaction with it (Ashraf et al., 2001).The importance of genotype × environment (G×E) interactions in breeding programs has been demonstrated in many major crops, including wheat (Najafian et al., 2010;Zali et al., 2011).This interaction complicates the identification of superior genotypes for a range of environments and calls for the evaluation in multiple sites to determine their true genetic potential (Yaghotipour & Farshadfar, 2007).Various statistical methods have been proposed to study G×E interactions (Lin et al., 1986;Becker & Léon, 1988;Crossa, 1990;Lin & Binns, 1994;Mohammadi & Amri, 2008;Malosetti et al., 2013).The additive main effect and multiplicative interaction (AMMI) model was developed specifically for analysis of G×E interaction in multi-locations varietal trials (Zobel et al., 1988).It estimates the total G×E effect of each genotype and partitions it into interaction effects with environmental components (Malosetti et al., 2013).
Hence, the aim of this research was to identify stable and high yielding durum wheat genotypes well adapted to the Senegal River Basin through multi-year and multi-location trials, as well as pinpointing the main traits critical for adaptation to heat stress.To the best of our knowledge, this is the first time that such an effort is conducted for this region.

Argo-Environmental Conditions
The experiments were carried out in two irrigated Savanah-type experimental stations: Fanaye, Senegal (FAN: 16°53′ N; 15°53′ W) and Kaedi, Mauritania (KED: 16°14′ N; 13°46′ W) during winter seasons 2014-15, 2015-16 and 2016-17.FAN is located 150 Km inland from the Senegal River delta, while KED is 300 km further away from the coast and its mitigating effect, and therefore tends to be warmer (Figure 1).FAN has sandy-clay soil with higher organic matter and good water holding capacity, while KED has lighter sandy-loam-clay soils with intermediate water holding capacity.All soils are rich in phosphorus (P) and low in the other nutrient, as typical for the 'Sahara effect' (Boy et al., 2008).
Table 1.Durum wheat genotypes used for field evaluation, their best linear unbiased estimator (BLUE) for grain yield (GY) across two sites in two seasons along the Senegal River and its summary statistics A second set of genotypes identified as 'validation' experiment was conducted only in season 2016-17.It included twenty durum wheat elites selected from the 39 th International Durum Observation Nurseries (IDON39), the three best (DAWRyT123, DAWRyT118, Bani Suef 5) and the one earliest (Oussara3) genotypes from the two previous seasons used as checks (Table B1).
All experiments were performed in alpha lattice design with six sub-blocks of size four repeated two times.The genotypes were grown in experimental plots of 7.5 m 2 at a sowing density of 120 kg ha -1 .A total of 150 kg of nitrogen were provided in three equal split applications, while 50 kg of phosphorus and potassium were provided as base fertilization before planting.
Weeds were chemically controlled during season 2014-15 by using a tank mixture of Derby (DowAgroscience, florasulam and flumetsulam) and Cossack (Bayer, sulfonylurea and safener) applied at Zadoks stage 14 (Z14, Zadoks et al., 1974), followed by a tank mixture of Derby and Pallas (DowAgroscience, pyroxsulam) at tillering stage (Z23).Mechanical weeding was also conducted as needed to ensure clean paddocks.For seasons 2015-16 and 2016-17 only mechanical weeding was conducted due to the unavailability of the chemical herbicides.
During 2014-15 season nine gravity irrigations were performed at intervals of 7-10 days in KED and in FAN for a total estimated of 320 mm and 410 mm of water provided, respectively.During 2015-16 season, the same number of gravity irrigations were performed in FAN, but reducing the quantity of water to approximately 360 mm total, while the number of irrigation was increased to 13 in KED for a total of approx.380 mm of water.For season 2016-17 a total of approx.380 mm of water were provided at the two stations via at intervals of 7-10 days.

Data Recording
The days to heading (DtH) was recorded as the number of days elapsed from sowing to the moment that 50% of the plot showed spikes emerging from the flag leaf (Z59).Before maturity (Z83-87), the number of fertile spike per meter square (Spk/m 2 ) were counted.Days to maturity (DtM) was recorded when 50% of the spikes turned yellow (Z91-92).A proxy of grain filling period (GFP) was then computed as the difference between DtM and DtH.Plant height (PLH) was measured in cm from the ground to the top of a representative ear excluding its awns.For each plot, only the middle rows were harvested for a total surface of 4.5 m 2 , dried and the biomass (Biom) weighted before threshing.The weight of the threshed grains was converted into yield (GY) expressed as kg ha -1 .The ratio between GY and Biom was expressed as harvest index (HI).One thousand grains were weighted in grams as 1000-kernels weight (TKW).The number of grain per meter square (Gr/m 2 ) was imputed using the weight of the grains harvested from 4.5 m 2 area and the average weight of one kernel derived from the TKW value, as per: Harvested weight of plot 4.5 m 2 × TKW 1000 (1) The number of grains per spike (Gr/spk) was derived from dividing the imputed number of grains per unit area by the number of spikes recorded for the same area, as follows: DtM, GFP, and Spike/m 2 were not recorded for season 2014-15.

Data Analysis
Both genotypes and environments were considered as fixed effects.Best linear unbiased estimators (BLUEs) of all traits were obtained using META-R (Multi Environment Trial Analysis with R for Windows) version 5.0 (Alvarado et al., 2015).Analysis of variance was computed for each environment using R version 3.2.1 (R Core Team, 2015), while combined ANOVA was obtained with GEA-R (Genotype × Environment Analysis with R for Windows) version 2.0 (Pacheco et al., 2015).Heritability was calculated based on the modified method suggested by Burton and Devane (1953) as follows: Where, σ 2 g is genotypic variance, σ 2 p is phenotypic variance, MSg is the mean square for the genotype, MSe is error mean square, MSG×E is the mean square of the interaction, r is the number of replicates and e is the number of environments considered.
The ratio of variance accounted by each source of variations (G, E, and G×E) was calculated dividing the sum of square of each for the total sum of square of the experiment.
For grain yield, G×E was partitioned by additive main effects and multiplicative interaction 2 (AMMI) model using R software (version 3.2.4) on R Studio.The 'AMMI wide adaptation index' (AWAI) was calculated using the following formula: Where, i is the number of significant IPCs determined by classical Gollob F-test in R Studio corresponding to 4 IPC in this specific case, si is the percentage of total G×E variance explained by each IPC, and PC is the actual IPC value.AWAI values close to '0' are obtained for the most widely adapted and stable germplasm (Bassi & Sanchez-Garcia, 2017).A performance index was generated by simultaneously selecting the best one third of the genotypes based on stability (AWAI) and one third best for average yield (BLUE).Genotypes that met both criteria were selected as the most suitable for cultivation along the Senegal River.

Heat-Prone Field Stations along the Senegal River
Temperatures along the Senegal valley varied across sites and years with much warmer temperatures during the season 2015-16 (Figure 1) mainly at the flowering windows.Planting was completed on the 6 th of December in FAN15, then the 17 th of December in FAN16, and further delayed to the 24 th December in FAN17.Sowing occurred on the 3 rd December in KED15, 10 th December in KED16, and 18 th December in KED17.The delay of sowing at both sites were due to late harvesting of rather long rice seasons.
During all growing seasons in FAN, average minimum night temperatures oscillated between 14 °C and 18 °C, while in KED the minimum night temperatures rarely descended below 22 °C.Maximum day temperatures oscillated between 30 °C and 33 °C in FAN15, while reached between 34 °C and 37 °C in FAN16 and FAN17.In KED16 the maximum temperatures remained constant between 33 °C and 35 °C while reached 37 °C during the jas.ccsenet.

period of g a weather
Figure 1 Kaedi wit

Two New Wheat Experimental Stations for Discriminating Heat Tolerance
The stations of Fanaye, Senegal and Kaedi, Mauritania were selected to represent the agro-environmental diversity that occurs along the Senegal River, with a particular focus on the delta and middle valley, respectively.The E effect of the experiment captured 77.7% of the total variance for GY, suggesting that these two stations are adequately contrasting to conduct significant multi-locations breeding selection for heat tolerance (Figure C).The three seasons used for testing, 2014-15, 2015-16 and 2016-17 had clear differences in temperatures during the phase of flowering, mostly caused by the delay in sowing.In fact, GY at FAN16 and KED16 were 60% and 29% lower compared to the timely sown season 2014-15 at the same sites, respectively.In FAN16 the germplasm was exposed to the highest temperatures (37 °C) during the time of flowering time, which in turn caused a severe drop in productivity.The following season (FAN17) planting was further delayed, but a drop in temperature to 34 °C occurred at the time of flowering, and this pushed the average GY to nearly double of what achieved in FAN16.This result shows the level of damage that the increase of just 3 °C in temperature can cause to the productivity of durum wheat if it occurs at the time of heading.

Selecting the Most Heat Tolerant Genotypes
The two stations over the two first seasons generated significant (p < 0.01) G×E interaction for GY, indicating that the tested genotypes did not respond equally to the changes in temperatures and sowing time.However, several genotypes were found to be stable and high yielding regardless of these changes, such as 'DAWRyT118', 'Bani Suef 5' and 'DAWRyT123'.These lines were among the top yielders in all environments and their AWAI score showed good stability (AWAI < 0.11).In particular, 'DAWRyT118' and 'DAWRyT123' were also confirmed as best performers in the 'validation' trials, which indicates that these lines carry heat tolerant traits capable of maintaining GY performance under stressed conditions.The two entries are in fact sister lines derived from top crossing the two most successful cultivars of the ICARDA durum program ('Om Rabi 5' and 'Cham 1') to Triticum dicoccoides collected in the surroundings of Aleppo (Table 1).Zaim et al. (2017) already described the usefulness of using T. dicoccoides in breeding durum elites, and identified 'DAWRYT118' as a top performer across drought prone sites in North Africa, with strong disease resistance, and good industrial processing qualities.Hence, their use in crossing schemes by durum breeders targeting heat tolerance is highly advised.

Traits to be Targeted by Durum Wheat Breeders to Increase Tolerance to Heat Stress
Heat stress has many detrimental effects on wheat at its various growth stages.Phenology traits (DtH and DtM) interacted among one another, but did not affect grain yield.Also, there was no significant difference between Top and Worst yielding lines for phenology in the two first season.Only the 'validation' set showed sufficient phenological variation to identified significant differences between Top and Worst.This would suggest that rather phenology is not an important characteristic for heat tolerance, when temperatures are constantly hot throughout the growing cycle as instead previously suggested by Hossain et al. (2012).Or more likely, the difference in results could be due to a limited amount of variation expressed for phenology by the 'discovery' set, while it was sufficient in the 'validation' set.Hence, phenology would instead represent a critical target that must be fixed through breeding first in order to then identify additional useful traits for heat tolerance.
High temperatures also shorten the tillering phase, resulting in poor setting of fertile tillers (Baldy, 1984).In addition, when heat occurs at the time of flowering it can reduce the vitality of the pollen and fertilization during pollen formation (Barlow et al., 2015;Draeger & Moore, 2017).Instead, during the grain filling period, heat stress reduces grain size and its weight (Dias & Lidon, 2009).Therefore, all these yield components appear of interest for improving heat tolerance.The number of Gr/spk and Biom showed positive correlation to GY, and also scored as significantly different among Top and Worst genotypes in two 'discovery' and one 'validation' environments.This is in good agreement with previous research that has also shown that Biom plays a decisive role in favoring GY (Mekhlouf & Bouzerzour, 2000;Abbassene et al., 1997;Masoni et al., 2007;Bahlouli et al., 2008).FAN16 was the environment with the most severe temperatures extremes during the flowering phase, and Gr/spk was identified as the only trait significantly different between Top and Worst elites at this location.Therefore, the ability of the best genotypes to maintain good fertilization under the severe heat resulted in better seed setting (Gr/spk) and ultimately higher yields.This is in agreement with Barnabas et al. (2008), and Hatfield and Prueger (2015), who found that the moment of fertilization is one of the most heat sensitive phase.Gr/spk represents therefore the single most appealing target trait for breeding better heat tolerance.The genotypes 'DAWRyT118' and 'DAWRyT123' were selected for their performances and stability.Their strategy for adaptation in fact relied mostly on the capacity of maintaining high spike fertility (Gr/spk) regardless of the temperatures, and to produce more Biom early in the cycle.Conversely, T. dicoccoides has been already praised by other authors for its capacity to produce vast biomass as well as for the fertility of its spikes (Merchuk-Ovnat et al., 2016a, 2016b;Merchuk-Ovnat et al., 2017).It is therefore not surprising that the two genotypes derived from it maintained these positive traits and used them to maximize heat tolerance.

Conclusion
The results presented here suggest that Senegal Valley provides ideal conditions for testing heat tolerance in wheat.A total of three genotypes identified as stable and well performing under these conditions ('DAWRyT118', 'DAWRyT123' and 'Bani Suef 5') showed good heat tolerance through the production of large biomass and maintenance of spike fertility.Breeders targeting improvement for this or similar regions should then focus on these traits, and possibly combining it with better harvest index.The Senegal River basin is regarded as a key place to bring social stability and food security to sub-Saharan Africa.Our results indicate that durum wheat is a suitable replacement of the fallow cycle and monoculture of rice.The area of possible expansion of wheat cultivation corresponds to the 200,000 ha currently grown as rice.Multiplying this area by the average yield of 3 t ha -1 reached by the three best lines, suggests the potential of producing 600,000 t of new food in sub-Saharan Africa, a potentially life-changing impact.

Appendix
Appendix A. ANOVA tables of all traits in all environments Table A1.ANOVA tables of days to heading

(
Figure additive were plac

Table 2
a Partitioni org grain filling in station at the t .Maximum an th indications of Phenology VA (Table2, T for the enviro ents of the two %).The 'valida 6 %.The E ex for the larges le 2).PLH mea E accounted f nd

Table A7
nd minimum te on the date of Table A1 and T onments (E), o first seasons ation' experime xplained 33.7% st part (35.8%) an across envi for 30.1% and for DtM revea ED17) for the g ys with a CV o

Table 4
Worst in all environments.Top and Worst genotypes had significant difference for Gr/spk and Biom at two 'discovery' environment, and at one of the two 'validation' sites.PLH and HI showed significant differences only in one environment.All other traits had no significant differences between Top and Worst yielders during the two first seasons.Instead, in the 'validation' trials in 2016-17 DtH and DtM became significant at both sites.

Table A4 .
ANOVA tables of thousand kernel weight

Table A9 .
ANOVA tables of spikes per m 2TableA10.ANOVA tables of grains per spike