Genotyping by sequencing for identification and mapping of QTLs for bioenergy-related traits in sweet sorghum

Sweet sorghum (Sorghum bicolor L. Moench) is a promising bioenergy crop. To increase the productivity of this crop, marker-assisted breeding will be important to advance genetic improvement of sweet sorghum. The objective of the present study was to identify quantitative trait loci (QTLs) associated with bioenergy-related traits in sweet sorghum. We used 188 F7 recombinant inbred lines (RILs) derived from a cross between sweet sorghum (Wray) and grain sorghum (Macia). The RILs and their parental lines were grown at two locations in 2012 and 2013. Genotyping-by-sequencing analysis of the RILs allowed the construction of a map with 979 single nucleotide polymorphisms. Using the inclusive composite interval mapping of additive QTLs, major QTLs for flowering time and head moisture content were detected on chromosome 6, and explained 29.45% and 20.65% of the phenotypic variances (PVE), respectively. Major QTLs for plant height (29.51% PVE) and total biomass yield (16.46% PVE) were detected on chromosome 7, and QTLs for stem diameter (9.43% PVE) and 100 seed weight (22.97% PVE) were detected on chromosome 1. A major QTL for brix (39.92% PVE) and grain yield (49.14%) PVE co-localized on chromosome 3, was detected consistently across four environments, and is closely associated with a SWEET sugar transporter gene. Additionally, several other QTLs for brix identified in this study or reported previously were found to be associated with sugar transporter genes. The identified QTLs in this study will help to further understand the underlying genes associated with bioenergy-related traits and could be used for development of molecular markers for marker-assisted selection.


Introduction
Bioenergy is an alternative and renewable energy derived from biomass resources, which includes waste (agricultural production waste, mill wood waste, urban organic waste, etc.), standing forests, and agricultural crops (Demirbaş 2001). It is expected that bioenergy will provide 30% of the world's energy by 2050 (Guo et al. 2015).
Sweet sorghum (Sorghum bicolor L. Moench) is a leading bioenergy crop (Calviño and Messing 2012; Carpita and McCann 2008;Zegada-Lizarazu and Monti 2012). Sweet sorghum is highly productive, with low input requirements, and is drought-tolerant (Rooney et al. 2007;Zegada-Lizarazu and Monti 2012). Sweet sorghum accumulates large amounts of carbohydrates in its stalk and produces total biomass as high as 30 Mg ha -1 (Bihmidine et al. 2015;Rooney et al. 2007). Stalk carbohydrates are easily converted to ethanol via fermentation of stalk juice.
The pressed stalk, after juice extraction, can be compressed into pellets, which are combustible. Thus, sweet sorghum could be used for both biofuel and thermo-electrical energy production (Zegada-Lizarazu and Monti 2012).
To be an efficient energy crop, sweet sorghum should be genetically improved. Biotechnology, genomics, and marker-assisted breeding will be important for genetic improvement of sweet sorghum (Madhusudhana 2014;Rooney et al. 2007). Genomic regions linked to complex traits can be identified by genetic mapping and QTL (quantitative trait locus) analysis (Shehzad and Okuno 2014). The fundamental idea underlying QTL analysis is to associate genotype and phenotype in a population exhibiting a genetic variation (Broman and Sen 2009). Analysis of QTLs is the first strategy in marker-assisted selection (MAS) for phenotypic traits (Shehzad and Okuno 2014).
Thus, the objectives of this study were 1) to identify QTLs associated with bioenergy-related traits in sweet sorghum using GBS, and 2) to confirm previously identified QTLs in an independent genetic background. QTLs for the bioenergy-related traits of sweet sorghum: flowering time, plant height, head moisture content, biomass yield, stem diameter, brix (sugar content), grain yield, and 100 seed weight were detected for 188 F 7 recombinant inbred lines (RILs) derived from a cross between Macia (grain sorghum) and Wray (sweet sorghum) (Bihmidine et al. 2015).
These QTLs could help identify genes that influence biomass productivity of sweet sorghum, and could be used for development of molecular markers to select individuals with valuable bioenergy-related traits within a breeding population using these markers.

Plant material
Two hundred F 6 and F 7 RILs derived from a cross between Macia and Wray were used for the QTL evaluation in this study. Macia (SDS 3220, PI 565121) is a grain sorghum developed by the Botswana National Agricultural Research System (Saadan et al. 2000). Macia is 1.4-1.6 m tall, with 7.3% stalk sugar content, grain yield about 3.3 Mg ha -1 , and biomass yield about 10.1 Mg ha -1 (Makanda et al. 2009;Setimela et al. 1997). Wray is a sweet sorghum developed by U.S. sugar-breeding programs (Broadhead et al. 1978;Hills et al. 1990 (Bihmidine et al. 2015;Pedersen et al. 2013 The experimental design was an alpha lattice incomplete block design with 15 incomplete blocks of 14 plots each per replication (15x14 alpha lattice), with two replications in each environment. The total was 420 plots for each environment. A plot was a single row measuring 4.5 m long, with 0.75 m between rows. A single row was sown at a rate of 50 seeds per row.
Phenotyping of bioenergy-related traits 1) Flowering time was measured as the duration of days from planting until 50% of the plants within a plot were shedding pollen. In 2012, harvesting dates were September 18 th -26 th at Havelock and October 2 nd -5 th at Mead. In 2013, harvesting dates were September 23 rd -27 th at Havelock and October 1 st -9 th at Mead. On average, harvesting was performed 125 days after planting. At harvesting times, five middle plants within a plot were arbitrarily selected to measure plant height, total biomass, stem diameter, brix (sugar content in stem juice), head and stem moisture content, grain yield, and 100 seed weight. 2) Plant height was measured as the distance in cm from the base of the plant to the tip of the panicle. 3) Total biomass in Mg ha -1 was collected when plants reached their physiological maturity by cutting the plants near the soil surface and separating them into panicles (heads), leaves, bottom stems (the length of 0-20 cm above ground), and remaining stems. The sub-samples were bagged and weighed immediately to obtain the wet weight and placed into an oven at 120-160 o C for ten days to completely dry the samples. Dried subsamples were reweighed and were summed up to obtain the total dry weight. Total biomass was calculated as follows: ( ℎ -1 ) = ℎ ( )/1,000,000 ( 2 )/10,000 4) Stem diameter (cm), and 5) brix readings were obtained from the bottom stem sections before drying. Brix degree ( o Brix) as a measure of soluble solids (mainly sucrose) in the stem juice was measured using a hand-held refractometer (MASTER-T Brix 0-32% with ATC, Atago Co., LTD, Tokyo, Japan). Because of the high volume workload in the field, stem diameter and brix readings were measured in the lab within 2-3 days. The bottom stems were kept in covered plastic boxes and put in a cold room (4 o C) to minimize sugar metabolism and to maintain brix level until the measurement. 6) Head moisture content and 7) stem moisture content were calculated as the percentage difference between wet biomass and dry weight. 8) Grain yield was calculated as follows: ( ℎ -1 ) = ℎ ( )/1,000,000 ( 2 )/10,000 Sub-samples of panicles were threshed and cleaned. Seeds were randomly sampled for 9) 100 seed weight (g).

Phenotypic Analysis
Analysis of variance (ANOVA) for bioenergy-related traits was performed for each environment and combined environments using the PROC MIXED procedure (Littell et al. 2006) of SAS version 9.4 (SAS Institute, 2008). Pearson's correlation coefficients between traits were calculated using the PROC CORR procedure of SAS.
Frequency distribution for bioenergy-related traits was performed for each environment using the PROC CAPABILITY. Narrow-sense heritability with the standard error was determined using SAS code which is available at http://www4.ncsu.edu/~jholland/ heritability/Inbreds.html (Holland et al. 2003).

Genotyping-by-sequencing (GBS)
A total of 190 genomic DNA samples, including 188 randomly selected RILs from 200 RILs and two parental lines (Macia and Wray), were sent to the Institute for Genomic Diversity, Cornell University for GBS. The

Linkage map construction
Linkage maps were constructed by IciMapping 3.2 (Wang et al. 2012). Before linkage map construction, binning of redundant markers, using the ICiMapping BIN function was performed to calculate the missing rate, and linkage maps were pre-constructed to evaluate a 1:1 segregation ratio of markers using the chi-square test (P≤0.01).
SNPs with ≥80% missing rate and significant SNPs from the chi-square test were removed from the data set.
Finally, genetic linkage maps were reconstructed with 979 selected SNPs.

QTL analysis
QTLs analysis, on biparental populations (BIP), was performed by IciMapping 3.2. Inclusive composite interval mapping of additive QTLs (ICIM-ADD) was used to conduct QTL mapping. Genetic linkage maps, means of phenotypic data for bioenergy-related traits for 188 RILs across four environments, and genotypic data of 979 SNPs were used for QTL analysis. The Kosambi mapping function was used to convert recombination frequency to mapping distance.
Step in scanning was assigned at one centiMorgan (cM). A thousand-permutation test was applied to each data set to determine LOD (log 10 of the likelihood odds ratio) threshold (P≤0.01), with LOD score >3 was used to determine significant QTLs. Phenotypic variation explained by the marker (%PVE) and the additive effect of QTLs were estimated by IciMapping.

Phenotypic data
Significant differences (P<0.05) were observed among genotypes and genotype x environment interactions for all traits measured in each environment and combined environments ( The highest positive correlation was observed between flowering time and head moisture content (r=0.73) ( Table 2). Sweet sorghum lines with later flowering time had higher head moisture content. Total biomass yield was positively correlated with all traits except head moisture content (Table 2). Total biomass yield correlated best with plant height (r=0.67). Brix was positively correlated with head moisture content (r=0.06) and 100 seed weight (r=0.15) and was negatively correlated with stem moisture content (r=-0.41), grain yield (r=-0.39), and stem diameter (r=-0.13) ( Table 2).

Linkage map construction
A total of 979 SNPs from GBS were mapped onto 10 linkage groups (chromosomes) based on the physical position of the SNPs. The linkage map spanned a total of 1,707.11 cM with an average inter-marker distance 1.74 cM ( Table 4). The maximum distance between markers was 47.11 cM (Chr 8) (

QTLs for brix
There were six QTLs associated with brix, which were located on chromosomes 2, 3, 4, and 7 and explained 5.28-39.92% PVE. Two QTLs on chromosome 3 were major QTLs. The major QTL with the highest PVE was on chromosome 3 (39.92% PVE) and was detected consistently across all four environments. There were negative additive effects of the QTLs detected on chromosomes 2, 3, and 7 and ranged from -0.62 to -1.71. QTLs detected on chromosome 4 had positive additive effects ranging from 0.79-0.82.

QTL for grain yield
Only one major QTL for grain yield was detected on chromosome 3 with 49.14% PVE and was detected consistently across four environments. This major QTL co-localized with the highest PVE QTL for brix, and its additive effect was 0.97 Mg ha -1 .
QTLs for 100 seed weight There were four QTLs associated with 100 seed weight, which were located on chromosomes 1, 4, 6, and 7, and explained 5.31-22.97% PVE. QTLs on chromosome 1 and 7 were major QTLs. The major QTL with the highest PVE was on chromosome 1 (22.97% PVE) but was detected only at Havelock in 2013. There were positive additive effects of the QTLs detected on chromosomes 1, 4, and 6 and ranged from 0.06-0.13 g, while the QTL detected on chromosome 7 had a negative additive effect (-0.09 g).
These results suggest that QTLs for flowering time on chromosomes 6 and 1 may be contributed by Ma1 and SbEHD1, respectively.
Similar results were also observed in previous QTL mapping for flowering time in sweet x grain sorghum populations. Ritter et al. (2008) reported a QTL for flowering time on chromosome 1 associated with marker ACC/CA3 or at 40.1 cM, whereas the peak position of the QTL located on chromosome 1 in this study was at 61 cM. In previous studies, QTLs for flowering time in sweet sorghum were mostly mapped on chromosome 6 at 0-48 cM, which contains the Ma1 and Ma6 genes (Felderhoff et al. 2012;Murray et al. 2008b;Ritter et al. 2008;Shiringani et al. 2010).
Co-localization of QTLs for flowering time, plant height, and biomass yield Plant height and flowering time are highly related traits in grasses, in which apical growth is terminated by flowering (Lin et al. 1995). QTLs associated with plant height are linked with loci controlling flowering, and have been previously reported in wheat (Börner et al. 1993), maize (Khairallah et al. 1998;Thornsberry et al. 2001), rice (Yan et al. 2011), barley (Chen et al. 2009), and sorghum (Börner et al. 1993;Higgins et al. 2014;Khairallah et al. 1998;Lin et al. 1995;Madhusudhana and Patil 2013;Morris et al. 2013;Shiringani et al. 2010;Thornsberry et al. 2001). In this study, a QTL for plant height (PH6, S6_41974107-42311647, 6.89-8.56 cM) co-localized with a major QTL for flowering time on chromosome 6. This result supports the finding that the dwarfing locus, dw2, is linked to the maturity gene, Ma1, which had been investigated in sorghum (Higgins et al. 2014;Lin et al. 1995;Madhusudhana and Patil 2013;Morris et al. 2013;Shiringani et al. 2010). We also found that total biomass yield correlated best with plant height (r=0.67), and major QTLs for these traits co-localized on chromosomes 6 and 7.
Similarly, Ritter et al. (2008), Shiringani et al. (2010), and Shiringani and Friedt (2011) also reported the colocalization on chromosome 6, while Murray et al. (2008b) and Felderhoff et al. (2012) reported the co-localization on chromosome 7. The results indicated that total biomass yield is highly determined by plant height.
QTLs for brix associated with SbSWEET1A, SbSUT5, and SbSUT6 Sugar content in this study was measured using brix. Brix is used as a practical analysis of total solute content (mostly sugars) (Kawahigashi et al. 2013). Ritter et al. (2008) suggested that brix is a simpler phenotypic trait compared with quantifying sucrose and sugar content, which are more time-consuming and expensive to measure. Moreover, they identify similar QTLs. In the current study, QTLs for brix were identified on chromosomes 2, 3, 4, and 7. Similar results were also observed in previous QTL mapping for brix on chromosome 2 (Guan et al. 2011;Shiringani et al. 2010), chromosome 3 (Felderhoff et al. 2012;Guan et al. 2011;Murray et al. 2008b), chromosome 4 (Bian et al. 2006Felderhoff et al. 2012;Lekgari 2010;Shiringani et al. 2010), and chromosome 7 (Bian et al. 2006;Guan et al. 2011;Lekgari 2010;Murray et al. 2008b;Shiringani et al. 2010). The QTLs for brix identified in the present study may be associated with known candidate genes controlling sugar content in sorghum.
Interestingly, Milne et al. (2013) reported that SbSUT5 and SbSUT6 may function in sucrose phloem loading in leaves and sucrose accumulation in stems. However, a previous qRT-PCR expression analysis of SbSUT5 and SbSUT6 in Wray and Macia stems did not support the hypothesis that these genes play a prominent role in sucrose phloem loading or stem sugar accumulation (Bihmidine et al. 2015). Further functional analyses of these genes will determine their contributions to whole-plant carbohydrate partitioning and sugar accumulation in stems.
It is also noteworthy, although we did not detect similar QTLs in our analyses, several of the other studies investigate the molecular functions of these genes and their potential involvement in regulating stem sugar content.
Finally, we also found another QTL for brix on chromosome 2 that was not correlated with known sucrose transport genes in sorghum. Shiringani et al. (2010) reported a QTL for brix on chromosome 2 at positions 30-40 cM, while the QTL for brix located on chromosome 2 in this study was at position 56.3-59.4 cM. It is possible that these QTLs for brix will be contributors for identifying other sugar transporter genes or novel genes affecting sugar metabolism.
Pleiotropic effects of QTLs associated with brix and grain yield A major QTL for brix and grain yield was co-localized on chromosome 3 (BR3_1 and GY3, S3_62297204-68877230, 158.76-176.55 cM. Felderhoff et al. (2012) also reported QTLs for brix and grain yield located nearby genomic regions on chromosome 3. Pleiotropic effects of genes that control more than one quantitative trait might explain this result. On the other hand, two or more nearby genes that control different traits might explain the results of co-localized regions. It is interesting to note that a related SWEET gene in maize and rice controls grain filling (Sosso et al. 2015). In our study, brix was negatively correlated with grain yield (r=-0.39) as expected for competing sinks (Bihmidine et al. 2013). Sweet sorghum varieties that traditionally have high concentration of sugar in stems have a small panicle because of competition for carbohydrates between grain filling and sugar storage in stems (Bihmidine et al. 2015;Gutjahr et al. 2013).

QTLs for stem moisture content
QTLs for all traits were detected for combined environments, except QTLs for stem moisture content. The reliability of the confidence intervals associated with QTL locations depends on the heritability of the individual QTL (Kearsey and Farquhar 1998). QTLs related to stem moisture content could not be detected in combined environments, but they were detected in one environment (Mead 2012) on chromosomes 3 and 7. Stem moisture content is easily affected by environment. The heritability of stem moisture content for combined environments was as low as 0.25, whereas that of Mead in 2012 was 0.48. The low heritability of the stem moisture content QTLs is likely the reason that we failed to detect them in the analysis. This result is similar to that found by Felderhoff et al. (2012). They reported no QTLs for percent moisture across all environments, but they found a QTL for moisture on chromosome 3 at only one of four locations.
QTLs detected in this study were similar in genomic regions with previous studies. Some QTLs were located on the same position, and genomic regions were narrowed in this study. QTLs for head moisture content were identified on chromosome 1, which is similar to the data of Lekgari (2010), who reported a QTL related to head moisture content on chromosome 1 at position 52.6 cM. In this study, QTLs for this trait were detected on chromosome 1 at positions 34 and 65 cM. QTLs for stem diameter were detected on chromosomes 1, 4, and 6. Two QTLs on chromosome 1 were located at the same position as QTLs reported by Shehzad and Okuno (2015) and by Shiringani et al. (2010), but both regions from this study were smaller than the previous studies.
Similarly, for 100 seed weight, we detected QTLs on the same chromosomes as found in previous studies, but ours were located at different positions. In this study, QTLs for 100 seed weight were detected on chromosomes 1, 4, 6, and 7 at positions 148.7-149.3, 105.1-107.4, 0-6.9, and 60.4-62.1 cM, respectively. Han et al. (2015) and Murray et al. (2008b) found QTLs for 100 seed weight at overlapping regions of 0-19 cM on chromosome 1. Shehzad and Okuno (2015) reported two QTLs for 100-grain weight on chromosome 4 at position 0-12 cM and 17.9-30.9 cM. Murray et al. (2008b) found a QTL for 1000 seed weight on chromosome 6 at position 9.6-30.8 cM. Han et al. (2015) and Shehzad and Okuno (2015) also reported QTLs for 100 seed weight on chromosome 7 at position 150.9-161.4 and 76.3-93.6 cM, respectively. These differences are likely attributable to differences in genotypes evaluated and environmental conditions.
In conclusion, GBS increased the precision of the QTL analysis and related genes could be identified. The bioenergy-related QTLs in sweet sorghum detected in this study were located in genomic regions associated with known genes linked to the traits (flowering time, plant height, and brix). Other QTLs we identified overlap with previous studies. Tantalizingly, the association between SbSWEET1A and the major QTL for brix on chromosome 3 suggests a potential mechanism underlying sugar accumulation in the sorghum (and possibly other bioenergy grasses) stem. Future validation of the bioenergy-related QTLs and identification of associated candidate genes should be conducted for functional genomic improvement of bioenergy-related traits in sweet sorghum.