Characterization of semi-arid Chadian sweet sorghum accessions as potential sources for sugar and ethanol production

Sweet sorghum (Sorghum bicolor (L.) Moench) is an important crop in Chad that plays an economic role in the countryside were stalks are produced mainly for human consumption without any processing. Unfortunately, very little information exists on its genetic diversity and brix content. Studies performed in 2014 and 2015 showed that there were significant variations (p < 0.001) for all assessed quantitative traits. Potential grain yield (0.12–1.67 t ha−1), days to 50% flowering (68.3–126.3 days), and plant height (128.9–298.3 cm) were among traits that exhibited broader variability. Brix content range from 5.5 to 16.7% across accessions, was positively correlated to stalk diameter and plant height, but negatively correlated to moisture content in fresh stalk and potential grain yield. Fresh stalk yield range from 16.8 to 115.7 Mg ha−1, with a mean value of 58.3 Mg ha−1 across accession. Moisture content in fresh stalk range from 33.7 to 74.4% but was negatively correlated to fresh stalk yield. Potential sugar yield range from 0.5 to 5.3 Mg ha−1 across accession with an average of 2.2 Mg ha−1. Theoretical ethanol yield range from 279.5 to 3,101.2 L ha−1 across accession with an average of 1,266.3 L ha−1 which is significantly higher than values reported under similar semiarid conditions. Overall, grain yields were comparatively low. However, two accessions had grain yield of more than 1.5 t ha−1; which is greater than the average 1.0 t ha−1 for local grain sorghum varieties in Chad. These could have multi-purpose uses; grains, sugar and bioenergy production.

Experimental design. Field experiments were conducted at the ITRAD Research Centre in Bébédjia (9° 55′ N Latitude North and 15° 8′ Longitude East), during the growing season from April to October for two consecutive years; 2014 and 2015, on a poorly desaturated sandy clay soil type. The experimental design was an α-lattice with three replications, each of which was subdivided into twelve plots of ten lines. Seeds from each Table 1. Rainfall, regional distribution, and origins of sweet sorghum accessions used in study. where CSY is conservative sugar yield (Mg ha −1 ), FSY is fresh stalk yield (Mg ha −1 ), DSY is dry stalk yield (Mg ha −1 ), JCY is juice yield (Mg ha −1 ), and SGY is sugar yield (Mg ha −1 ). Sugar concentration of juice (SCJ) was determined as 75% of brix expressed in g kg −1 sugar juice 35 :

Regions Departments Number of villages Number of accessions
Theoretical ethanol yield (TEY, L ha −1 ) from extracted juice was calculated as sugar yield (kg ha −1 ) multiplied by a conversion factor of 0.581 L kg −1 sugar 41 : Total soluble sugar (y, %) was estimated using equation by 23 : where x is the Brix of stalk juice. Data analysis. Analysis of variance (ANOVA) and Newman-Keuls test were performed to determine if the average of the quantitative characteristics varied significantly (at 5.0, 1.0, and 0.1% probability thresholds) between accessions. Some parametric distributions were calculated to see the dispersion of the values of the characteristics based on the average, the minimum, the maximum and the coefficient of variation. For all characteristics, genetic variabilities were estimated from the components of the analysis of variance, and broad heritability (H 2 ) was calculated using the formulas by 16 . Bivariate analysis was carried out, using the Pearson correlation coefficient to see the link between two characteristics. Multivariate analyses were performed through Principal Component Analysis (PCA) to highlight uncorrelated characters, which were used to build the dendrogram from the Hierarchical Ascending Classification (HAC). Then the Discriminant Factor Analysis (DFA) was carried out to characterize the group from the HAC. Variability in agro-morphological traits. Analysis of qualitative traits (Table 2) showed a great diversity among sweet sorghum accessions from Chad. Generally, 64% of seedlings were green and 36% violet, and these were accession specific. Seedling vigor ratings showed frequencies of 31.4, 32.9, and 21.2% for low, fair, and good vigor respectively. The color of the leaf midrib (a phenotypic characteristic of sweet sorghum) was green for 69.5% and white for 29.5% of the accessions. Most accessions had loose panicles (51.9%) with straw colored glumes (80.9%), red seed (84.5%), often very floury in vitreousity (69.7%) and mainly from the caudatum (64.55%) or bicolor (35.45%) race. The analysis of variance (Table 3) showed highly significant differences (p < 0.001) between accessions in all assessed quantitative agronomic characteristics. The coefficients of variation (CV) for most characters was above 30%, reflecting variability amongst accessions. Heritability (H 2 ) range from 58.6 to 98.8%. The strongest agronomic heritability (H 2 > 80%) were recorded for number of days to flowering (95.6), number of days to heading (95.5), plant height (94.9), number of internodes (94.5), internode length (90.2), panicle length (90.2), stalk diameter (83.9), dry stalk weight (82.1), and perultimate leaf width (81.0). The lowest heritability was scored for potential yield (64.8), panicle grain weight (64.8), panicle weight (61.5) and panicle width (58.6).
Other traits (see Supplementary Table 1) showed significant variability with some Chadian accessions performing better than the ICRISAT checks. Compared to the average of ICRISAT checks, 22 Chadian accessions were taller, 63 had shorter days to heading, 54 had shorter days to flowering, 46 had higher number of internodes, 20 had longer internode lengths, 16 had longer perultimate leaf, 7 had wider perultimate leaf, 19 had longer panicle, 48 with wider panicle, 41 with higher fresh stalk biomass, 74 with higher dry stalk biomass, and 33 with broader stem diameter. Five of the 105 accessions screened had potential grain yields equal to or higher than the average 1.0 t ha −1 of grain sorghum production in semi-arid Chad.
All assessed quantitative characteristics linked to sugar production showed significant variability (p < 0.001) within and when compared to ICRISAT checks ( Table 4). The coefficient of variation was high (CV > 30%) for most characteristics except for Brix (19.8%), total soluble sugar (20.6%) and the sugar concentration on juice (20.1%). The heritability was very high for all assessed characteristics except for the Juice yield (59.6%) and moisture content of fresh stalk weight (57.6%). Heritability was strongest for brix (98.8%), sugar concentration of juice (98.8%) and total soluble sugar (98.8%), followed by dry stalk yield (82.1%) and then fresh stalk yield (79.8%).
Late maturity correlated positively to larger stem diameter, number of internodes, fresh and dry stalk yields, juice and sugar yields but negatively to thousand grain weight, with no significant correlation to potential grain yield. Taller plants with longer panicles, high brix content, high fresh and field dried stalk yields, and high juice and sugar yields were negatively related to potential grain yield. Table 4. Analysis of variance of twelve characteristics determining sugar and ethanol production in 105 accessions of semi-arid Chadian sweet sorghum. ***Indicates significance at p < 0.001, CV; Coefficient of variation, MFS; moisture content of fresh stalk, FSY; fresh stalk yield, DSY; field dried stalk yield, CSY; conservative sugar yield, JCY; juice yield, SGY; sugar yield, SCJ; sugar concentration of juice, TEY; Theoretical ethanol yield, y; total soluble sugar. ± standard error.  www.nature.com/scientificreports/ Cluster analysis and diversity organization. Hierarchical clustering, using Ward's method and truncation on level 175, grouped the 110 sweet sorghum accessions into four major groups (Fig. 2). Group 1 was composed of 31 accessions and was further subdivided into 2 sub-groups; 1a (13 accessions) and 1b (18 accessions). Group 2 had the largest number of accessions (50) and was subdivided into 2 sub-groups; 2c (38 accessions) and 2d (12 accessions). Group 3 consisted of 27 accessions divided into 2 sub-groups; 3e (2 accessions) and 3f. (25 accessions). Group 4 was composed of 2 accessions. The Mahalanobis distance obtained through the Discriminant Factor Analysis (DFA) showed long distance between all the groups (p < 0.001 and p < 0.01), which signified that the four groups were quite distinct ( Table 6).
The mean values of the accessions in each cluster are given in Table 7, obtained through the Discriminant Factor Analysis (DFA). Group 1 accessions were characterized as short, early flowering, with lowest fresh stalk weight and the lowest average brix content (10.7%). Groups 2 and 3 showed mean values of all characteristics studied, Figure 2. Clustering of 110 sweet sorghum accessions into 4 groups using the standardized squared Euclidean distance of Ward's hierarchical clustering method. Table 6. Distances of Mahalanobis and statistical significant of Fisher. **,***indicates significance at p < 0.01 and p < 0.001 respectively.  www.nature.com/scientificreports/ with an average brix value of 11.6% and 13.7% respectively. The two accessions in group 4 were characterized as the tallest plants, late flowering, with the heaviest fresh stalk weight and the highest average brix value (15.2%). The structuring of the individuals in each group (Table 8)  Promising genotypes for sweet sorghum improvement program in Chad. This multi-year evaluation of the 105 Chadian sweet sorghum accessions showed higher performance in juice and sugar yields of 10 accession compared to the 5 improved sweet sorghum from ICRISAT considered as high performers ( Table 9). The "Balnda" accession had 16.7% brix, 42.6 Mg ha −1 juice yield, 5.3 Mg ha −1 potential sugar yield, and a theoretical ethanol yield of 3,101.2 L ha −1 . It was followed by the accession "Sian Guebeuh" with 2,712.2 L ha −1 of theoretical ethanol yield, and "Var137" with 2,669.5 L ha −1 of theoretical ethanol yield. The "Zimikaye Combole" accession, despite its 13.7% brix, achieved a high sugar yield (4.4 Mg ha −1 ) and ethanol yield (2,570 L ha −1 ). The best ICRISAT varieties was IS23541 with 14.1% brix, 27.4 Mg ha −1 juice yield, 2.9 Mg ha −1 potential sugar yield, and 1695.0 L ha −1 theoretical ethanol yield.

Discussion
The agro-morphological diversity study of Chadian sweet sorghum showed significant variability in qualitative and quantitative characteristics. In fact, two main leaf midrib colors were observed; white and green which are also the main colors of this type of sorghum 28 . According to research by 28 , accessions with white midrib were not Table 8. Regional distribution of the accessions of Chadian sweet sorghum following Ward's hierarchical clustering into four major groups.  www.nature.com/scientificreports/ very juicy. However, our study showed that some white midrib accessions were very juicy. The sweet sorghum accessions from the Sudanese zone were identified to be from the caudatum and bicolor race, unlike the dryseason sorghum which are from the durra race 14 . According to 28 , sweet sorghum belongs to bicolor, caudatum, durra and hybrid bicolor-guinea race. Ritter et al. 33 suggested that sweet sorghums are of polyphyletic origin, with relatives among kafir, caudatum, and other grain sorghum types. Murray et al. 26 identified three separate groups of sweet sorghum which often are classified together. He classified these major types as: syrup types; (historical and some modern) which were from the caudatum race, modern sugar and energy types; associated with the kaffir/bicolor races, and amber types; mainly durra and bicolor races.

Regions Group 1 Group 2 Group 3 Group 4 Total
Generally, the 0.5 t ha −1 average grain yield production in this study was quite low compared to the average 1.0 t ha −1 observed in Chadian grain sorghums or 0.87 t ha −1 for dry-season grain sorghum in the same region 14 . According to 7 , sweet sorghum is characterized by reduced grain yield as compared to grain sorghum. However, two of the accessions Begon (1.7 t ha −1 ) and GWS lache (1.5 t ha −1 ) showed higher than average grain yields and could be used to improve grain yield of sweet sorghum. Plant height, stem diameter, number of internodes, internode length and other morphological characteristics showed high variability in the accessions studied.
The brix from the current study differed significantly between accessions and it value ranged from 5.5 to 16.7%, with an average of 11.9%. Brix content was lower than that reported by 8,28 , who obtained brix value ranging from 8.9 to 21.8% and 11.8 to 22.5%, respectively. The "Balnda" and "Chian Woua" accessions had 16.7% brix, which was higher than what has been recorded in many sweet sorghum studies. According to 30 , optimal harvesting stage for sweet sorghum is when the juice contains 15.5-16.5% brix which is one of the most important characteristics necessary to obtain juice of high fermentable quality and thus maximize ethanol yield per hectare.
The moisture content of fresh stalk for the 105 landraces ranged from 33.7 to 74.4%, averaging 48.7%. The average value is lower than moisture contents obtained by 40 (76.0%; using a single cultivar) and 10 (81.0%; using three cultivars). But the value obtained on this current study were high than that obtained by 5 (16.5%; using 73 sweet sorghum accessions). The higher end of the observed range of juice yield in this study were similar or higher than values obtained by 43 , but similar to values obtained by 41 working with 31 sweet sorghum lines in Arizona.
Estimated sweet sorghum sugar yields in current study showed high level of diversity (p < 0.001) amongst the 105 cultivars, the average value (2.2 Mg ha −1 ) being lower than estimated mean values (4.0 Mg ha −1 and 4.0 to10.7 Mg ha −1 ) reported by 38,41 ; both evaluating 4-6 sweet sorghum lines at variable planting dates and locations respectively. However, 2.2 Mg ha −1 was higher than values reported by 34 (1.8 Mg ha −1 , 1 variety with variable NPK fertilization management), 44 (1.8 Mg ha −1 in 2009, using 1 hybrid at variable N-fertilization rates) and 35 (1.7 Mg ha −1 using 5 cultivars across 3 years). The average amount of theoretical ethanol yield (1,266 L ha −1 ) were lower than value (2,854 L ha −1 ) obtained by 41 , but higher than that obtained by 35 (1,025 L ha −1 ) and 2 (1,000-1,149 L ha −1 , using 12 cultivars). According to 19 , up to 13.2 Mg ha −1 of total sugars, equivalent to 7,682.0 L ha −1 of ethanol can be produced by sweet sorghum under favorable conditions.
The correlation matrix showed positive correlations of interest between Brix content with plant height, stem diameter, number of internodes and number of days to flowering. Brix is a measure of dissolved sugar to water mass ratio of a liquid; it was positively influenced by the maturity. According to 37 , all known sweet sorghums are tall, and prior research identified a positive correlation between height and sugar accumulation. The positive relation between brix and maturity suggest that early maturity may not be a desirable characteristics for sweet sorghum variety development since plants will need more days to accumulates more biomass and store energy in its stalk throughout the growing period 27,32,36 . Positive relations between plant height and days to flowering suggest that taller plants tended to flower later as observed by 8,42 . The study showed plant height, number of days to flowering and fresh stalk weight negatively affected the moisture content in fresh stalk. Similar results were obtained by 5 working with sweet sorghum cultivars from the U.S. National Plant Germplasm System collection.
Yields of sugar and theoretical ethanol were significantly (p < 0.05) and negatively influenced by potential grain yield and thousand grain weight. Indeed, sweet sorghum accessions are characterized by the accumulation of carbohydrates in their juicy stems 32 to the detriment of the grains which remain rather poorly filled. Clerget et al. 4 suggested a potential negative interaction between stem development and grain yield in sweet sorghum. This is contrary to grain sorghum accessions where accumulation of carbohydrates is done in favor of grains. Sweet sorghum landraces usually have small panicles and the stem sweetness is commonly attributed to low panicle strength 22 . However, the interactions and trade-offs between panicle size, grain filling and stem development are complex and can be complicated in photoperiod-sensitive sweet sorghums which tend to show great phenotypic plasticity. While competition for carbohydrates between grain filling and sugar storage in stems has been suggested by other studies 2,31 , the remobilization of stem reserves towards grain was frequently reported as being small 21 . Thus, varieties in current study which combined high stem reserves with comparatively good grain yield will be great candidates for dual-purpose (food-fuel) sweet sorghum breeding. These varieties are currently under consideration for a potential collaborative project between sorghum scientists in Chad and the United States. Such varieties were reported 25 as being cultivated by farmers in semi-arid Mali.

Conclusion
The current study showed that sweet sorghum accessions from Chad were from caudatum and bicolor race, and most of them had loose panicles with red seeds. Brix values ranged from 5.5 to 16.67% were found and greatly differed (p < 0.001) among accessions. Two accessions "Balnda" and "Chian Woua" with a high brix value of (16.67%) were identified and could be used as source of sugar to improve grain sorghum in Chad. The yields of potential sugar and theoretical ethanol showed the values ranging from 0.45 to 5.3 Mg ha −1 and 279.5 to 3,101.2 L ha −1 respectively. The study showed high variability (p < 0.001) for all assessed quantitative characters within sweet sorghum accessions from Chad, and identified four major groupings within accessions, each with multiple subgroupings except cluster 4 which had 2 accessions. This study provides valuable findings which could be used to Scientific Reports | (2020) 10:14947 | https://doi.org/10.1038/s41598-020-71506-9 www.nature.com/scientificreports/ improve sweet sorghum production in Chad for dual purpose use. Furthermore, the ten accessions with higher brix content than the five improved ICRISAT varieties could be used in biofuel (ethanol) breeding programs in similar geographical production regions. These accessions are currently been considered for improvement in a collaborative project with US based scientists.