BREEDING SORGHUM FOR GRAIN, FORAGE AND BIOENERGY IN BRAZIL

– Sorghum is a versatile crop used for energy production, human and animal feedings, and raw material for industry. The grain sorghum is the fifth most important cereal worldwide, and Brazil is one of the top 10 sorghum producers. In the last 40 years, the sorghum grain yield in Brazil has increased to 32.70 kg ha -1 year -1 . Although this clear evolution, much remains to be done, and sorghum breeders in Brazil still face several challenges. This review discusses the main characteristics of sorghum genetics and breeding, aiming for sorghum improvement for food, fodder, feed, and fuel uses. Herein will be highlighted essential topics related to the genetic control of the main interest traits; conventional sorghum breeding; the development of sorghum lines and hybrids; the use of male-sterility in sorghum breeding; the implication of genotypes-by-environments interaction in sorghum, the use of genome-wide association studies, and genomic prediction to maximize the efficiency of the sorghum breeding programs.

hectares, concentrated mainly in the Rio Grande do Sul (RS) and São Paulo (SP) (Tabosa et al., 2019). Grain sorghum hybrids from Argentina crossed the frontiers, being introduced in RS and, lately, in SP, from were spread to other states in Brazil. In addition, due to government support and incentives, 50 sweet sorghum genotypes from the United States Department of Agriculture (USDA), Africa, and India were introduced in Brazil. The Embrapa Maize and Sorghum used these germplasms to start developing the Brazilian sorghum cultivars (Carmo, 1977). The Brazilian grain-sorghum yield was about 2400 kg ha -1 in 2020/2021. However, reports presented in Menezes et al. (2018) indicate that the Brazilian sorghum crop may achieve up to 7000 kg ha -1 of grain yielding. Furthermore, it is a consensus that only joint efforts of plant breeding, crop science, and crop protection, among others, will change the actual Brazilian sorghum grain yielding scenario, improving the productivity of the sorghum crop.

Types of sorghum and genetic inheritance of main interest traits
Breeding programs must have clear objectives to be successful. In this scenario, sorghum breeders must set breeding objectives based on farmers' needs. As presented by Rakshit and Gomashe (2013), regardless of sorghum types, the primary breeding objectives for sorghum are i) high and stable yielding; ii) tolerance and resistance to biotic and abiotic stresses. On the other hand, each type of sorghum has its ideotype, which considers the plant size and architecture, length of the production cycle, and sorghum end uses. Ideally, ideotype refers to a set of desired characteristics in a plant genotype to achieve a high yield for a given end-use (Donald, 1968).
In addition, agronomic traits such as uniformity in seed germination, flowering, and grain maturity, and responsivity to better soil fertility and low water demand, are desired traits in any sorghum crop (Rao et al., 2015). However, there are specific must-have traits for each type of sorghum. The sorghum crop is divided into five major commercial types: i) Grain sorghum; ii) Forage sorghum; iii) Biomass sorghum; iv) Sweet sorghum; and v) Broom sorghum.
According to Ribas (2014), these types can be subdivided according to their end uses: i) grain production; ii) forage for silage purposes; iii) forage for animal grazing purposes; iv) sweet sorghum for ethanol and v) biomass sorghum for energy generation.
In addition to table 1, broom and dualpurpose sorghum types have also their ideotypes.
Broom sorghum has value in the panicle, which small producers widely use to produce artisanal brooms. Furthermore, dual-purpose sorghum is an intermediate type with the potential to be exploited in grain and silage production or even forage cultivars with potential use in ethanol and sugar production.
Among the main traits of interest, earliness has become one of the essential sorghum traits for Brazil's growth conditions. Early genotypes optimize agricultural production by allowing crop rotation within the same growing season.
In addition, as fast as the plant completes its cycle, the crop will be exposed to unfavorable conditions, such as drought, for a short period. Therefore, earliness is a must-have trait for grain sorghum since this type of sorghum is mainly cultivated during the second crop season as an alternative to maize (Landau & Netto, 2015).
However, normal-late genotypes often present great biomass yielding than early genotypes, which is desired in cases of high biomass production needs.
Sorghum plants can be classified as sensitive or not to photoperiod. Although there is variation in response to photoperiod, sorghum is considered a short-day plant, meaning that, originally, the plant would bloom at times of the year with long nights. In cultivars sensitive to photoperiod, the apical bud remains in the vegetative stage until the critical photoperiod of 13 hours; floral induction occurs when the  (Rooney & Aydin, 1999). Dominant alleles at each locus contribute to late flowering, with Ma1 being the gene with the greatest effect on the trait (Hao et al., 2021). A dominant Ma1 locus results in photoperiod-sensitive genotypes, regardless of the other loci (Quinby & Schertz, 1970).
In the specific case of biomass sorghum,  (Rodrigues, 2015). and acid detergent (ADF) fiber. The optimal dry matter content should range between 30 and 35%, while DMD above 55% indicates high-quality silage (Tolentino et al., 2016). Furthermore, the good-quality silage should have at least 7% protein content and not exceed 55 and 40% of NDF and ADF, respectively, to maximize animal production (Tolentino et al., 2016;Rodrigues et al., 2021).
In addition, forage sorghum can be used for cutting or animal grazing purposes.

Sawazaki (1998) highlights an interspecific
crossing between Sorghum bicolor x Sorghum sudanense to obtain superior genotypes for cutting or grazing. In this scenario, fast growth and easy regrowth are highly desired. However, Simili et al. (2013)

Breeding for resistance to biotic and abiotic stresses
Regardless of the type of sorghum, breeding for abiotic and biotic stresses is of utmost importance. Sorghum is considered a drought-tolerant crop due to its root morphology and ability to reduce transpiration and metabolism in general (Schittenhelm & Schroetter, 2014).
However, sorghum yielding reduction might be observed under severe drought stress conditions (Albuquerque et al., 2011). Batista et al. (2017) highlight that yield losses depend on the length and intensity of the drought condition, and the plant cycle stage, which is more severe in reproductive stages (Anami et al., 2015).
In Brazil, due to the unstable climatic Due to its tropical origin, the sorghum crop is highly susceptible to cold conditions, negatively affecting seedling emergency, plant growth and development, tillering, height, dry matter accumulation, and flowering. Ortiz et al.
(2017) also reported a reduction in chlorophyll synthesis in cold conditions, reducing photosynthesis and production. Genetic variance for cold tolerance has been reported in grain, sweet, and biomass sorghum (Patanè et al., 2021;Franks et al., 2006), as well as QTLs, which can be helpful for marker-assisted selection (MAS) (Parra-Londono et al., 2018b;Burow et al., 2011). Furthermore, heterosis for cold tolerance has also been identified, and more tolerant hybrids development can be exploited (Schaffasz et al., 2019;Windpassinger et al., 2017).
To aim for a more sustainable agriculture system, sorghum breeders may focus on developing N and P efficient usage genotypes (Bollam et al., 2021;Bernardino et al., 2019;Hufnagel et al., 2014;Rodrigues et al., 2014 Like diseases, pests have great potential to reduce sorghum production, whether the products are grains, juice or biomass. In addition to quantitative damage, there is also the possibility of qualitative damage to production.
In the context of breeding, sources of resistance to the sugarcane aphid (Melanaphis sacchari) have been identified by Armstrong et al. (2015) and Boyles et al. (2018). Genomic regions associated with resistance to sorghum midge (Stenodiplosis sorghicola) and green aphid (Schizaphis graminum) have also been reported (Punnuri et al., 2013;Tao et al., 2003  The sorghum hybrids seeds are multiplied, and F 2 seeds are harvested from the F 1 plants. Finally, the F 2 seeds are sown in the field/greenhouse, generating the F 2 population. From the F 2 population, sorghum breeders can apply several selection methods to advance the genotypes and obtain sorghum inbred lines. Pedigree, bulk, and single seed descent (SSD) methods can be applied in this scenario.
In the pedigree method, starting from F 2 , the seeds of each selected plant will constitute a row in the next generation until homozygosity.
Then, in each generation, progenies tests are performed, and the best ones are selected. After, the selection is accomplished within and between families until the F 5 generation. Finally, single and multi-environment trials must be done. However, since the selection and plant genealogies start in the F 2 generation, the pedigree method requires much effort.
The bulk method is undoubtedly one of the most cost-effective segregating populations.
According to Ramalho et al. (2012), the bulk method prioritizes natural selection during the initial generations (F 2 to F 4 or F 5 ), which means that no artificial selection will be performed until the advanced stages of homozygosity. Briefly, in the bulk method, all the plants of the population are harvested, and their seeds are mixed.
From this set, a seed sample is taken and will constitute the next population. This process is repeated until to obtain the F 4 population.  Developing a double-haploid system in sorghum implies a revolutionary change for sorghum breeding, reducing breeding time, and improving the program's efficiency.

Heterosis and hybrid seeds production in Sorghum
Although sorghum is a self-pollinated crop, some levels of heterosis have been reported in sweet sorghum (Pfeiffer et al., 2010;Bunphan et al., 2015;Lombardi et al., 2018), grain sorghum (Crozier et al., 2020;Gomes et al., 2020;Santos, 2020;Menezes et al., 2017) and biomass sorghum (Packer & Rooney, 2014). As previously mentioned, preventing self-pollination in sorghum crops is necessary before performing the crosses. However, manual emasculation is not efficient on a commercial scale as it is time-consuming, which means high financial costs. Therefore, to commercially exploit heterosis in sorghum, an alternative approach is required. Reddy and Reddy (2019) report that the cytoplasmic-male sterility (CMS) and fertility restorers (Rf genes) were discovered in different sorghum varieties of the milo group (grain sorghum) by Stephens and Holland (1954). In general, a sorghum genotype will be male-sterile if it possesses, in homozygoses, the recessive alleles ms and rf in its cytoplasm and nucleus, respectively (Rakshit & Bellundagi, 2019).
Three distinct lines are used to commercially exploit CMS in Sorghum in this system (House, 1985) (Figure 2 In general, "A" and "B" are isogenic lines, except that the "B" line is male fertile due to MsMs cytoplasm. Due to cytoplasmic maternal inheritance, the "A" line is maintained through the A x B crossing, while the "R" line is used to restore fertility. Therefore, the in-field sorghum hybrid seeds are obtained through ♀A x R crossing. For hybrid seed production purposes, a ratio of 4:2 rows of female "A" and male "R" lines, respectively, is adopted (Figure 3a). In the case of higher pollination capacity, other ratios such as 6:2 can also be used to reduce costs Determining the best Sorghum hybrid combinations is crucial for successfully exploiting heterosis in a sorghum breeding program. In addition, the evaluation of the best combinations aids the sorghum breeders in deciding which ones will be synthesized on a commercial scale. Among several methodologies, the use of diallel crosses to determine general (GCA) and specific (SCA) combining ability is the most applied (Parmar et al., 2019;Chikuta et al., 2017;Menezes et al., 2014).
Using a partial diallel, Rocha et al. (2018) evaluated the combining abilities of four malesterile "A" lines and five fertility restorer "R" lines of sweet sorghum for ethanol production.  Figure 2. A-B-R system for hybrid seed production. Source: House (1985). This powerful genomic tool will be discussed in detail later in this review.

Recurrent selection in sorghum
Population improvement is a two-step approach involving broad genetic-based gene pools followed by recurrent selection methods.
In general, recurrent selection is a cyclic process that involves obtaining progenies by crossing, then evaluating, selecting, and recombining the best ones to develop an improved population without exhausting its genetic variability.  (Leite et al., 2020). Due to their stability, the male-sterility ms3 and ms7 alleles have been extensively used (Aruna et al., 2021), being the male-sterility conditioned to homozygous recessive genotypes. In a recent book by Embrapa, the use of genetic male-sterility in sorghum is discussed in detail (Oliveira et al., 2021). Aruna et al. (2021)  selecting. Menezes et al. (2021) highlight that choosing good parentals is the first and most crucial step to creating a good breeding population. In general, 10 to 20 good agronomic performing parentals are required to obtain a population with high yielding and genetic variability .
Once defined, the parental lines need to be crossed with a source of male-sterility genes (msms). The F 1 generation is composed of plants in heterozygosity (Msms) for male-sterility, that is, male-fertile plants. Therefore, proceeding with a self-fertilization generation is necessary to obtain male-sterile genotypes (msms). The F 2 seeds generated from self-fertilization present a

Genotype-by-Environment interaction in sorghum
The genotypes-by-environments (GE) interaction is a significant challenge in plant breeding. GE interaction can be defined as the differential genotype responses to types of environments. In severe GE interaction, genotypes re-raking across the environments might be observed, making the selection process more challenging. According to Hunt et al. (2020), plant breeders may adopt one of two strategies to manage GE interaction: i) ignore GE interaction by selecting for broad adaptation, or ii) exploit GE interaction by performing genotype selection for both broad and specific adaptation.
In both scenarios, multi-environment trials (METs) are of utmost importance. Plant breeders may use METs information to understand GE interaction better and select or recommend the best genotypes for a set of target environments or mega-environments.

Many agronomic and nutritional traits in sorghum
are harshly affected by GE interaction. For example, Aruna et al. (2020)  Many methodologies to study the GE interaction were developed over the years. Most of them aim to estimate the genotype adaptation, adaptability, and stability. For example, Adugna (2007) and Souza et al. (2013) adopted linear regression models (Finlay & Wilkinson, 1963;Eberhart & Russell, 1966) and other methodologies (Wricke, 1962;Shukla, 1972;Francis & Kannenbert, 1978;Lin et al., 1986;Annicchiarico, 1992) to evaluate the adaptability and stability of sorghum genotypes to perform selections and recommendations. Rakshit et al. (2017) and Aruna et al. (2020) applied Additive Main Effects and Multiplicative Interaction (AMMI) (Gauch, 1992) model to study GE interaction in several traits of sorghum.
On the other hand, the GGE biplot methodology was used by Phunke et al. (2017) and Diatta et al. (2021). In addition, a fascinating study was performed by Enyew et al. (2021) using both AMMI and GGE biplot methodology to study 324 sorghum genotypes evaluated across three environments in Ethiopia. These works confirm the above mentioned methods as the most common and effective multivariate models in the study of stability, adaptability, genotype ranking, and selection of suitable mega-environments. A detailed description of the cited methodologies can be found in Ramalho et al. (2012) andBorém et al. (2017).
In tropical conditions of Brazil, the GE interaction is even more challenging. Grain sorghum is one of the best alternatives for soybean crop succession. Therefore, being

Genomic prediction in sorghum breeding
Although the tremendous genetic gain haploids, which are widely applied in maize breeding but also can be exploited in sorghum (Hussain & Franks, 2019 flowering, and wet and dry biomass production. In its turn, significant work was conducted by Bernardino et al. (2021), who applied association mapping and genomic selection for sorghum adaptation to tropical soils of Brazil.
The authors mapped two candidate P efficiency genes ALMT and PHO2, on chromosomes 6 and 9. In addition, they achieve genomic prediction accuracies for grain yield, around 0.22 (absence of dominance effects) and 0.30 (dominance effects plus GWAS-derived fixed cofactors). This work highlighted the importance of combining genome-wide association and genomic prediction. The genomic prediction approach has been undergoing constant adjustments over the years. Many efforts have been allocated to applying genomic prediction to several crops, such as maize (Dias et al., 2020;Beyene et al., 2019), wheat (Juliana et al., 2020), rice (Labroo et al., 2021b), and sorghum (Oliveira et al., 2018;Bernardino et al., 2021), and to define the best prediction models and approaches (Santos et al., 2015;Dias et al., 2018;Fristche-Neto et al., 2018).

Final considerations
Sorghum is a versatile crop cultivated for several end-uses, such as grain production, forage, alcohol, and biomass. According to the breeding program's goals, the sorghum breeder can define the best strategies to obtain and select superior genotypes.
Most sorghum breeding programs target traits of earliness, high grain yielding, quality, biomass production, resistance, and tolerance to biotic and abiotic factors. For quantitative traits, progeny testing methods (Bulk, pedigree, and SSD) are more efficient than mass selection.
However, mass selection benefits high heritable traits, such as male sterility.
Heterosis has been reported for agronomic traits in Sorghum; therefore, breeders may also choose to exploit it through sorghum hybrid seed production. Furthermore, using an efficient male-sterility technique is of utmost importance to exploit heterosis commercially.
Although its exceptional adaptability to a wide range of environments, the genotypes-byenvironments interaction is a complicating factor in sorghum breeding, and several methodologies are available. They can be applied to better study and understand this interaction, aiming to minimize or exploit GE interaction. Finally, genomic prediction tools might increase in the next couple of years to reduce costs and improve genetic gain per unit, increasing sorghum breeding program efficiencies.