Introduction

Considering the irregular rainfall in tropical and semiarid regions, producing roughage in these sites requires strategies to supply forage to the herd throughout the year. Ensiling maize (Zea mays L.) and sorghum crops (Sorghum bicolor L.) stands out as a great alternative to supply forage during the dry seasons when there is forage paucity1. Comparing these two crops, sorghum plants display more adaptability and forage yield stability than maize plants in semiarid conditions2. Moreover, ensiling sorghum genotypes is crucial for safe animal feeding since the anaerobic fermentation process, plus the pH sharp decrease, might limit the enzymatic activity of β-glycosidase as well as the activity of microorganisms producing β-glycosidases, avoiding toxic release of hydrogen cyanide (HCN)49.

There are numerous forage sorghum cultivars available for sale; however, in recent years, the dry matter yield (DMY) potential and DMY stability of these commercial forage sorghum genotypes have been revealed as inferior to that of genotypes in the breeding phase, primarily in semiarid regions. Santana et al.3 evaluated the DMY of three forage sorghum materials, finding higher production in the experimental genotypes 201813B and 201814B (12.0 and 12.3 t/ha, respectively) than in the commercial genotype Volumax (4.8 t/ha). Also, Moura et al.4 assessed the DMY of five forage sorghum genotypes and observed higher yields in the experimental ones 12F02006 and 12F03033 (20.7 and 16.6 t/ha, respectively) than in the commercial ones BRS655 and Volumax (10.9 and 9.4 t/ha, respectively).

Conversely, assessing fermentation and nutritional traits in silages made of high-yield forage sorghum genotypes must also be considered to select them for ensiling5,6. Information on nutritive value, fiber digestibility, and dry matter losses during the ensiling process is crucial to select genotypes with high DMY and suitable ensilability7,8. In this scenario, studies integrating genetic diversity and silage fermentation evaluations are scarce, comprising only agronomic responses and/or silage parameters assessed separately8. Combining genetic and silage quality parameters is necessary to achieve more accurate genotype selection and improve the release of future cultivars18. Furthermore, rumen kinetics of high-DMY sorghum genotype silages are understudied, especially in regions with irregular rainfall regimes5. Thus, searching for more productive forage sorghum genotypes that allow for producing good-quality silages is determinant for livestock sustainability in semiarid regions.

This study hypothesized that selecting high-DMY forage sorghum genotypes for ensilage can preserve fermentation quality and increase yields of forage and digestible nutrients per area. Based on these premises, this study aimed to evaluate agronomic traits and genetic diversity of 18 sorghum genotypes to select the most promising for forage accumulation. Concomitantly, the study aimed to assess fermentation, chemical composition, and rumen kinetics characteristics of seven of the 18 genotypes.

Methods

Experiment 1—agronomic traits and genetic diversity in forage sorghum genotypes

Experiment site and climatic details

The field trial was conducted during the harvest seasons of 2021/2022 and 2022/2023 in the Experimental Field of Gorutuba at EMBRAPA Maize and Sorghum located in Nova Porteirinha municipality, Minas Gerais state, Brazil (15°45’26” S, and 43°16’53’’ W). The site soil is classified as Red Yellow Latosol with a medium texture9. The site climate is mesothermic tropical, sub-humid, and semiarid with irregular rainfall (800 mm per year) and relative humidity of 65%. The predominant climate in the region is Aw according to the Köppen classification10. Climatic data during the experimental period are detailed in Fig. 1.

Fig. 1
Fig. 1
Full size image

Monthly rainfall and average temperature during the harvest seasons of 2021/2022 and 2022/2023 in the Experimental Field of Gorutuba, Nova Porteirinha, Minas Gerais, Brazil.

Experimental design and treatments

The experiment was conducted in a complete block design with three repetitions (experimental plots) and two harvest seasons (harvest season 1 in 2021/2022, and harvest season 2 in 2022/2023). Eighteen (18) genotypes from the Germplasm Active Bank of EMBRAPA Maize and Sorghum (Table 1) were assessed regarding agronomic traits in experimental plots.

Table 1 Forage sorghum genotypes (Sorghum bicolor L.) assessed in the study and their respective identification number, origin center, status, and time to harvesting (days).
Full size table

Cropping management of forage sorghum genotypes

Soil samples were collected in a zig-zag way to perform analyses of soil physical and chemical properties in a specialized laboratory, before beginning the trial. Physical characteristics, pH, and base saturation in the 0–20 cm soil layer in the first year were: pH = 5.8; V = 69%; 59, 8, and 33 dag/kg of sand, silt, and clay, respectively. In the second year, the pH was 5.9 and V was 70%. Planting fertilization was performed using 300 kg/ha of the NPK (08-28-16) commercial formula. Moreover, we performed a topdressing fertilization using 150 kg/ha of urea from the V6 to the V8 phenophases, that is, from six-true-leaf to eight-true-leaf stages. Such a period was chosen because it is critical to define the crop potential for grain and dry matter yields11.

Seeds used in the experiment were gently donated by EMBRAPA Maize and Sorghum. Sowing was performed manually on 26 November 2021 (harvest season 1) and on 11 November 2022 (harvest season 2), with no supplementary irrigation. The experimental plots comprised two 5-m rows spaced 0.7 m apart, totaling 3.5 m2. Sowing density was from 200 to 250 thousand seed/ha, and the excess of seedlings were removed using manual thinning from the germination up to the V5 phenophase aiming for reach from 10 to 12 plants per linear meter, that is, a planting density of 157 thousand plants/ha, on average. Such management was realized to achieve the planting density recommendation from 100 to 170 thousand plants/ha in the sorghum cropping11. The experiment adhered to procedures approved by the Unimontes Animal Use Ethics Committee under protocol no. 008/2023.

Agronomic traits

Agronomic traits were assessed within experimental plots measuring 10.5 m2 (four 5-m long rows spaced 0.7 m apart), specifically in the two central rows (5 × 0.7 m = 3.5 m2). The evaluated agronomic characteristics were flowering time, number of plants, plant height, fresh matter yield (FMY), and DMY. The flowering time was estimated in days, and when half of the plants within the plot emitted their panicles, counting from the sowing day. Plant height was estimated from the ground to the panicles’ tops using a proper rule. Then, the number of plants was determined by manual counting within the plot and converted into plants/ha, right before the forage harvest.

FMY was estimated from the forage harvest within the plot area. The collected material was weighed on a digital scale, and the result was converted into t/ha of fresh matter. Furthermore, the leaf-stem ratio (LSR) and the panicle percentage were calculated from the material fractionation into leaves, stems and panicles. For the DMY estimation, forage samples were chopped into 1 to 2 cm particle size using a stationary machine (JF 50 MAXXIUM®, São Paulo, Brazil), and dried in a forced-air oven at 55 °C for 72 h according to the INCT methods12 to obtain the dry matter content (DM, INCT G-001/1 method). Moreover, the parameter of Shukla stability variance13 on the DMY was calculated due to ambient variations throughout the harvest seasons (2021/2022 and 2022/2023). In this context, DMY data variance between the two harvest seasons was calculated for each genotype using an appropriate statistical method (more details were provided in the statistical analysis). Lower variance coefficients (σi2) mean fewer DMY differences caused by climatic variations over the years, thus indicating higher stability for DMY.

Statistical analysis

Agronomic traits data were subjected to analysis of variance (ANOVA) and covariance, besides a test of normality for residuals (residuals × quantiles) using the PROC MIXED of SAS On Demand for Academics14. No missing data or outliers were handled, since we found normal residual distributions for all variables. Genotype, harvest season, and their interaction were considered fixed effects, while block and its interactions were considered random effects. The harvest season was analyzed as a fixed effect to observe genotype-environment relations. Means of genotypes were compared using the Scott Knott test using the software Genes v. 1900.2023.9315. Means of harvest seasons were compared using the probability of difference (“pdiff”) adjusted by the Tukey test. All tests were significant at 5% of error probability (P < 0.05). The mathematical model was:

Yijk = µ + XGi + XAj + ZBk + X(GA)ij + Z(GB)ik + Z(AB)jk + Z(GAB)ijk + eijk.

Where Yijk is the observed value; µ is the overall mean; XGi is the fixe effect of forage sorghum genotype (1 to 7); XAj is the fixed effect of harvest season (1 to 2); ZBk is the random block effect (1 to 3); X(GA)ij is the interaction between genotype and harvest season, while Z(GB)ik, Z(AB)jk and Z(GAB)ijk are the interactions between block and fixed effects, and eijk is the residual error for each observation.

The parameter of Shukla stability variance (σi2) on the DMY was determined considering the two harvest seasons. The model was analyzed using the PROC MIXED of SAS OnDemand for Academics using the script created by Piepho16. How lower the variance, the higher the stability for DMY17. Genetic diversity was measured using the general distance of Mahalanobis (D2ij) and the formation of a dissimilarity matrix. Then, the Tocher grouping method was performed to gather genotypes based on lower genetic distances15. Both Mahalanobis distance and Tocher grouping analyses were chosen because they are accurate in identifying phenotypic divergences among genotypes, as they are based on variance data rather than means15. Agronomic traits data were subjected to principal component analysis (PCA), and genotypes were plotted on a 2D graph comprised of PC1 and PC2. Genetic diversity and PCA were performed using the software Genes v. 1990.2023.9315.

Experiment 2—Silage of the forage sorghum genotypes

Fermentation parameters and chemical composition

Silage of seven of the 18 forage sorghum genotypes were assessed concomitantly with the field agronomic experiment, in the 2021/2022 and 2022/2023 harvest seasons. Experimental genotypes number 3 (202110F007), 4 (202110F008), 10 (CMSXS6000), 11 (CMSXS6001), and 14 (CMSXS6002) were previously chosen within the 18 ones for ensiling because they were promisor in the EMBRAPA forage sorghum breeding program for high DMY18. Their silages were compared to the 16 and 18 ones (BRS658 and Volumax silages, respectively). As mentioned before, forage sorghum was harvested when panicle grains displayed a pasty-to-floury texture.

After forage weighing for calculations of FMY and DMY, the material was chopped into 1 to 2 cm particle size using a stationary machine (JF 50 MAXXIUM®, São Paulo, Brazil). The material was compacted into experimental silos composed of PVC tubes with a 100-mm diameter and 50-cm height, closed with PVC lids equipped with Bunsen valves. Forage was compacted to achieve a silage density of 550 kg/m3, then the filled PVC tubes were weighed and stored at room temperature in the Laboratory of Animal Nutrition of Unimontes for 120 days for further analyses. Silages fermented during 120 days to emulate the farmers’ forage conservation strategy in some semiarid regions5.

After opening the silos, the silage was removed and homogenized on a plastic tray, and about a 500-g sample was collected to analyze the chemical composition and rumen kinetics variables. This 500-g sample was dried in a forced-air oven at 55 °C for 72 h and mowed to pass 2-mm and 1-mm sieves using a Willey-type mowing. Another aliquot was separated and pressed on a hydraulic machine to extract the effluent to determine pH values, ammonia nitrogen19, and organic acids. Other silage fractions were put into identified plastic pots, each containing a datalogger. Each ‘plastic pot + datalogger + silage’ set was maintained under controlled ambient conditions in a proper room to determine the aerobic stability time20.

Lactic, acetic, propionic, and butyric acids, as well as ethanol concentrations, were analyzed using a liquid chromatograph (Shimadzu™ Prominence System model 20 A, Kyoto, Japan) and methods proposed by Price21. The lactic-to-acetic ratio was calculated by dividing the lactic acid by the acetic acid concentration.

Chemical composition variables of silages were analyzed using the INCT methods12. We measured contents of DM (INCT-CA G-003/1 method), ash (INCT-CA M-001/2 method), crude protein (CP, INCT-CA N-001/2 method), ether extract (EE, INCT-CA G-004/1 method), neutral detergent fiber (NDF, INCT-CA F-001/1 method), acid detergent fiber (ADF, INCT-CA F-001/2 method), and lignin (LIG, INCT-CA F-005/2 method). Additionally, non-fiber carbohydrate contents were calculated using the equation proposed by Sniffen et al. 22:

NFC (g/kg) = 1000 – Ash – CP – EE – NDF.

Where CP is the crude protein content, EE is the ether extract content, and NDF is the neutral detergent fiber content. Organic matter content was obtained by subtracting the ash from the DM content12. The total digestible nutrient content (TDN) was calculated for lactating Zebu cows using the equation proposed by Valadares Filho et al. 23:

TDN (g/kg) = (0.78 × CP) + (0.95 × NFC) + (0.65 × NDF) + 2.25 × (0.86 × EE) – 7.16.

Where CP is the crude protein content, NFC is the non-fiber carbohydrate content, NDF is the neutral detergent fiber content, and EE is the ether extract content.

Rumen kinetics

The use of animals adhered to procedures approved by the Unimontes Animal Use Ethics Committee under protocol no. 008/2023. Two Zebu-Holstein crossbreed bulls weighing 450 kg and rumen-fistulated were used in an in-situ assay of rumen kinetics24. They fed a diet containing 3.0 kg concentrate and 25.0 kg BRS Ponta Negra forage sorghum silage during the experiment. Samples were incubated at a 20 mg/cm2 ratio using non-woven fabric (NWF) bags measuring 7 × 7 cm with 150-µm porosity. The NWF bags were positioned on the ventral sac rumen region and remained there for 0, 3, 6, 12, 24, 48, 72, 96, 120, 144, and 288 h to simulate the rumen retention time of sorghum silage24.

The remaining residues after incubation were analyzed for DM and NDF contents using the INCT methods12. Degradation rate data was fitted by the non-linear method of Gauss-Newton using the SAS software according to the equation proposed by Ørskov and McDonald25. The potential degradability of DM (pdDM) over time (‘t’ given in hour) was obtained using the equation:

pdDM = a + b (1- e− ct).

Where ‘a’ is the soluble fraction, ‘b’ is the potentially degradable insoluble fraction, and ‘c’ is the rate of degradation of fraction ‘b’ per hour. The effective degradability of DM (edDM) was calculated using the calculation proposed by Alderman and Blake26:

edDM = a + (b × c ÷ c + k).

Where ‘a’ is the soluble fraction, ‘b’ is the potentially degradable insoluble fraction, ‘c’ is the rate of degradation of fraction ‘b’ per hour, and ‘k’ is the passage rate of particles throughout the rumen. Moreover, the potential degradability of NDF (pdNDF) over time (‘t’ given in hour) was interpreted using the model of Mertens and Loften27:

pdNDF = b × e− ct + I.

Where ‘I’ is the undegradable fraction, ‘b’ is the potentially degradable insoluble fraction, and ‘c’ is the rate of degradation of fraction ‘b’ per hour. For the effective degradability of NDF, the following model was used:

edNDF = bp × c ÷ (c + k).

Where edNDF is the effective degradability of NDF, ‘bP’ is the standardized potentially degradable fraction, ‘c’ is the rate of degradation of fraction ‘b’ per hour, and ‘k’ is the passage rate of particles throughout the rumen. DM and NDF fractions were standardized according to Waldo et al.28 to determine the indigestible ones (ipDM and ipNDF). Moreover, we could calculate the effective degradable dry matter yield (DDMY) of silage using the equation:

DDMY (t/ha) = (DMY – 15%) × (edDM ÷ 1000).

Where DMY is the dry matter yield (t/ha) of the Experiment I, 15% is the potential DM losses during the ensiling process, and edDM is the effective DM degradability (g/kg).

Statistical analysis

Chemical composition, fermentation parameters, aerobic stability, and ruminal kinetics data were subjected to analyses of variance (ANOVA) and covariance, besides a test of normality for residuals (residuals × quantiles) using the PROC MIXED of SAS On Demand for Academics14. No missing data or outliers were handled, since we found normal residual distributions for all variables. Genotype effect was the fixed effect while the effects of harvest season, block, and their interactions were random. Means of genotypes were compared using the Scott Knott test using the software Genes v. 1900.2023.9315. All tests were significant at 5% of error probability (P < 0.05). The mathematical model was:

Yijk = µ + XGi + ZAj + ZBk + Z(GA)ij + Z(GB)ik + Z(AB)jk + Z(GAB)ijk + eijk.

Where Yijk is the observed value; µ is the overall mean, XGi is the fixed effect of forage sorghum genotype (1 to 7); ZAj is the harvest season random effect (1 to 2); ZBk is the block random effect (1 to 3), Z(GA)ij, Z(GB)ik, Z(AB)jk and Z(GAB)ijk are the interaction effects, and eijk is the error for each observation.

Results

Experiment 1—agronomic traits and genetic diversity in forage sorghum genotypes

Agronomic traits and genetic diversity

There were isolated effects of forage sorghum genotype and harvest season on plant height (P < 0.001) (Table 2). Sorghum plants were taller in harvest season 1 for all genotypes. The tallest genotypes were the experimental ones (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, and 14), while the commercial genotype number 15 was the shortest.

Table 2 Plant height (PH), number of plants (NP), flowering time (FLOWR), and leaf-stem ratio (LSR) in forage sorghum genotypes (Sorghum bicolor L.) harvested in seasons 1 (2021/2022) and 2 (2022/2023). Means followed by the same lowercase letter within rows or uppercase letter within columns, for the same variable, do not differ by the Scott-Knott test at a 5% probability of error (P < 0.05). G = genotype effect. H = harvest season effect. G × H = genotype × harvest season interaction (represents the genotype-environment interaction). SEM = standard error of the mean. Letters in the row and columns related to the ‘Mean’ cell compare isolated harvest season and genotype effects, respectively. Legend of genotypes’ nomenclature: 1 = 202110F005. 2 = 202110F006. 3 = 202110F007. 4 = 202110F008. 5 = 202110F010. 6 = 202110F011. 7 = 202110F012. 8 = 202110F013. 9 = 202110F016. 10 = CMSXS6000. 11 = CMSXS6001. 12 = CMSXS6003. 13 = 202110F020. 14 = CMSXS6002. 15 = BRS655. 16 = BRS658. 17 = AGRI001E. 18 = Volumax.
Full size table

There was a significant isolated effect of genotype on the panicle percentage (P < 0.001). Commercial genotype number 15 exhibited the most robust panicle among all materials, while experimental genotypes 6 and 13 showed the lowest panicle proportionally (Table 3). Conversely, a genotype-season interaction effect was observed on FMY and DMY (P < 0.001). During harvest season 1, higher fresh and DM forage production were observed for almost all experimental genotypes (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, and 14), except for number 1, which displayed intermediate production. The lowest forage production was recorded in the commercial genotypes (15, 16, 17, and 18). During harvest season 2, the highest forage production was observed in genotype 6, and the lowest in genotype 15.

Higher parameters of Shukla stability variance were found in the genotypes 6 (σi2 = 48.30), 9 (σi2 = 55.90), 14 (σi2 = 51.47), and 18 (σi2 = 50.90). These genotypes showed less stability for DMY, showing considerable discrepancies in forage accumulation between harvest seasons 1 and 2 (Table 3).

Table 3 Panicle percentage (P), fresh matter yield (FMY), dry matter yield (DMY) and parameters of Shukla stability variance for DMY in forage sorghum genotypes (Sorghum bicolor L.) harvested in seasons 1 (2021/2022) and 2 (2022/2023). Means followed by the same lowercase letter within rows or uppercase letter within columns, for the same variable, do not differ by the Scott-Knott test at a 5% probability of error (P < 0.05). G = genotype effect. H = harvest season effect. G × H = genotype × harvest season interaction (represents the genotype-environment interaction). SEM = standard error of the mean. Letters in the row and columns related to the ‘Mean’ cell compare isolated harvest season and genotype effects, respectively. Legend of genotypes’ nomenclature: 1 = 202110F005. 2 = 202110F006. 3 = 202110F007. 4 = 202110F008. 5 = 202110F010. 6 = 202110F011. 7 = 202110F012. 8 = 202110F013. 9 = 202110F016. 10 = CMSXS6000. 11 = CMSXS6001. 12 = CMSXS6003. 13 = 202110F020. 14 = CMSXS6002. 15 = BRS655. 16 = BRS658. 17 = AGRI001E. 18 = Volumax.
Full size table

Regarding the PCA, PC1 and PC2 explained 82.06% (68.74 plus 13.52%) of the total data variation on agronomic traits (Fig. 2). DMY and NP were those variables most contributing to form PC1 and PC2, respectively (DMY displayed an eigenvector value of 0.8416 for PC1, and NP showed an eigenvector value of 0.9681 for PC2).

Six groups were formed from the sorghum agronomic traits and Mahalanobis genetic distances (Fig. 1). Group I comprised ten genotypes (1, 3, 5, 7, 8, 9, 10, 12, 13, and 14), Group II comprised three ones (11, 17, and 18) and Group III comprised two genotypes (2 and 16). Groups IV, V, and VI were formed by a single genotype each (15, 4, and 6, respectively). As represented in Fig. 1, the longest genetic distance was observed between genotypes 6 and 16, while the shortest distance was found between 5 and 7. Genotype number 8 stood out in the multivariate analysis of agronomic traits, and its position in the 2D projection (upper and further right) corroborates that. Higher values for PH, NP, FMY, and DMY were simultaneously observed in this genotype (Tables 2 and 3). Conversely, the commercial genotype 15 was a negative outlier for agronomic traits, with lower values for PH, NP, FMY, and DMY than the others. Genotype 15 position in the 2D projection (below and further left) corroborates that.

Fig. 2
Fig. 2
Full size image

2D projection of 18 forage sorghum genotypes (Sorghum bicolor L.) evaluated based on agronomic traits, Tocher grouping, and principal components 1 and 2 (PC1 and PC2). The longest Mahalanobis distance was between genotypes 6 and 15 (D2ij = 2339.54). The shortest Mahalanobis distance was between genotypes 5 and 7 (D2ij = 12.81). Legend of genotypes’ nomenclature: 1 = 202110F005. 2 = 202110F006. 3 = 202110F007. 4 = 202110F008. 5 = 202110F010. 6 = 202110F011. 7 = 202110F012. 8 = 202110F013. 9 = 202110F016. 10 = CMSXS6000. 11 = CMSXS6001. 12 = CMSXS6003. 13 = 202110F020. 14 = CMSXS6002. 15 = BRS655. 16 = BRS658. 17 = AGRI001E. 18 = Volumax.

Experiment 2—silage of the sorghum genotypes

Fermentation parameters and chemical composition

There was a genotypic effect on concentrations of ammonia nitrogen and organic acids in forage sorghum silages, but no significant effect on pH and aerobic stability results (Table 4). The average pH value was 3.66, and the aerobic stability lasted 127.4 h. Ammonia nitrogen contents observed in silages of commercial genotypes 16, and 18 were superior to those found in silages of the experimental genotypes 3, 4, 10, and 14.

Regarding organic acids, silage fermentation of genotypes 11 and 18 was more lactic than that of genotypes 3, 10, 14, and 16 (Table 4), with the number 10 genotype silage displaying the lowest concentration. In contrast, the highest acetic fermentation was found in genotype 10, followed by genotypes 11 and 18, and genotype 3 lastly. Numbers 16 and 18 genotype displayed the highest and second-highest propionic fermentations, respectively, while silages of genotypes 4 and 10 led to the lowest propionic fermentation.

Number 14 genotype silage displayed the highest butyric fermentation, followed by that of number 11, 4, 10, 16 and 18 (Table 4). Silage fermentation of genotype 16 presented no butyric acid (0.00 g/kg of DM). The highest ethanol concentrations were found in silages of genotypes 10 and 11, while the lowest alcoholic fermentation was observed in silages of genotypes 3, 14 and 16. Finally, the highest L/A ratio was observed in the number 3 genotype silage, followed by those of genotypes 4 and 14. The lowest L/A ratio was reported in the silage of genotype 10.

Table 4 Fermentation parameters, aerobic stability (AE) and chemical composition of silages made of forage sorghum genotypes (Sorghum bicolor L.) harvested in seasons 1 (2021/2022) and 2 (2022/2023). Means followed by the same letter within rows do not differ by the Scott-Knott test at a 5% probability of error (P < 0.05). pH = hydrogen potential. N-NH3 = ammonia nitrogen content. L/A ratio = lactic acid-acetic acid ratio. DM = dry matter content (on fresh matter basis). OM = organic matter content. CP = crude protein content. EE = ether extract content. NDF = neutral detergent fiber content. ADF = acid detergent fiber content. LIG = lignin content. NFC = non-fiber carbohydrate content. TDN = total digestible nutrient content for lactating cows (Valadares Filho et al. 2023). SEM = standard error of the mean. Legend of genotypes’ nomenclature: 3 = 202110F007. 4 = 202110F008. 10 = CMSXS6000. 11 = CMSXS6001. 14 = CMSXS6002. 16 = BRS658. 18 = Volumax.
Full size table

Number 16 genotype silage showed the highest DM content, followed by those of numbers 3 and 4 (Table 4). Silages made of number 10 and 16 genotypes had intermediate DM concentrations, while the lowest ones were recorded in numbers 11 and 14 genotype silages. Furthermore, number 14 genotype silage displayed the highest ash content, besides the lowest organic matter concentration as a consequence.

Silages from genotypes 16 and 18 had the highest CP concentrations (Table 4), in contrast to the lowest DMY these commercial materials showed (Table 2). The opposite occurred in experimental genotypes 4 and 10: lower silage CP contents but higher DMY. The highest EE concentrations were recorded in 3, 14, 16 and 18 genotype silages (Table 4). Number 10 genotype silage stood out, displaying the highest NDF concentration, and silages of commercial genotypes 16 and contrasted it showing the lowest NDF values. Silage ADF results exhibited a similar pattern to the NDF ones: the highest value in the experimental genotype 10, and the lowest ones in genotypes 16 and 18. More lignin was found in silages from the experimental genotypes 4, 10 and 11. Finally, the highest NFC and TDN concentrations were observed in silages of genotypes 4, 16, and 18, and the lowest ones were found in genotypes 10 and 14.

Rumen kinetics

The genotype effect changed fractions ‘a’ (soluble fraction) and ‘b’ (potentially degradable insoluble fraction) of forage sorghum silages, as well as affected pdDM, edDM at 2, 5, and 8% rates, besides ipDM and DDMY (Table 8). In contrast, no significant differences between silages were recorded for fraction ‘c’ (rate of degradation of the insoluble fraction per hour) or lag time for DM colonization by microbiota (Table 5).

The lowest fraction ‘a’, besides the lowest edDM at 2, 5, and 8% rate, was found in numbers 10 and 11 genotype silages. The highest ipDM values were found precisely in these silages. The highest fraction ‘b’ and the highest pdDM were recorded in genotype silages of numbers 16 and 18. Conversely, producing silages of the experimental genotypes 3, 4, 10, and 14 led to higher DDMY than producing silages of the commercial genotypes 16 and 18, and the experimental material number 11 (Table 5).

Table 5 Rumen kinetics and degradable dry matter yield (DDMY) in silages made of forage sorghum genotypes (Sorghum bicolor L.) harvested in seasons 1 (2021/2022) and 2 (2022/2023). Means followed by the same letter within rows do not differ by the Scott-Knott test at a 5% probability of error (P < 0.05). Fraction ‘a’ = soluble fraction. Fraction ‘b’ = potentially degradable insoluble fraction. Fraction ‘c’ = rate of degradation of fraction ‘b’ per hour (g/kg/h). pdDM = potential degradability of dry matter. EdDM 2% = effective degradability of dry matter at a 2% rate (k = 2%/h). EdDM 5% = effective degradability of dry matter at a 5% rate (k = 5%/h). EdDM 8% = effective degradability of dry matter at an 8% rate (k = 8%/h). ipDM = standardized indigestible fraction of dry matter. pdNDF = potential degradability of neutral detergent fiber. EdNDF 2% = effective degradability of neutral detergent fiber at a 2% rate (k = 2%/h). EdNDF 5% = effective degradability of neutral detergent fiber at a 5% rate (k = 5%/h). EdNDF 8% = effective degradability of neutral detergent fiber at an 8% rate (k = 8%/h). ipNDF = standardized indigestible fraction of neutral detergent fiber. Lag time = time for particle colonization by fermentation microorganisms. Legend of genotypes’ nomenclature: 3 = 202110F007. 4 = 202110F008. 10 = CMSXS6000. 11 = CMSXS6001. 14 = CMSXS6002. 16 = BRS658. 18 = Volumax.
Full size table

There were significant effects of genotype on pdNDF and ipNDF of forage sorghum silages, as well as on edNDF at 2, 5, and 8% rates (Table 5). In contrast, no statistical differences were recorded in fraction ‘c’ and lag time for NDF colonization. The highest pdNDF values were observed in numbers 16 and 18, while the lowest were recorded in numbers 10 and 11 genotype silages. The opposite was observed in ipNDF results: lowest values in numbers 16 and 18, besides the highest values in numbers 10 and 11 genotype silages. Regarding edNDF at 2, 5, and 8%, the highest rates were found in silages of sorghum genotypes 4, 14, and 18.

Discussion

Agronomic traits and silage parameters of the forage sorghum genotypes showed considerable differences among experimental (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, and 14) and commercial genotypes (15, 16, 17, and 18). Some key findings could be highlighted from that: (i) experimental were taller and more productive than commercial genotypes; (ii) in general, commercial genotypes displayed more DMY variability between harvest seasons; (iii) genotypes showed great genetic diversity that enables directed plant breeding; (iv) suitable fermentation parameters were found in all genotype silages; (v) commercial genotypes led to silages with superior nutritional value but the experimental ones allowed producing more digestible nutrients per area.

Plant height discrepancies among forage sorghum genotypes were influenced by both environmental and genetic factors. Rainfall in the harvest season 1 (2021/2022) was higher than in the harvest season 2 (20222/2023), so plants had more resources such as water and available soil macro and micronutrients to grow in the first trial period29.

Plant height in forage sorghum genotypes is controlled by recessive genes in dw1 to dw4 loci. Dwarf sorghum genotypes have recessive genes in two loci, resulting in shorter plants, while tall genotypes show recessive genes in only one locus30. Tall genotypes such as the experimental ones showed more capacity for vegetative growth and forage accumulation likely because of a favorable combination of maturity alleles (Ma1 to Ma6), enabling a longer phenological cycle and higher forage accumulation than the commercial genotypes30. In contrast, commercial genotypes were shorter and earlier because they are probably less sensitive to photoperiod, so they flowered before and stopped growing more quickly than the experimental ones31.

Dwarf sorghum genotypes, such as the commercial ones assessed in this trial, have morphological characteristics such as short internodes and stem length, which limited forage accumulation32. These genotypes often show high leaf-to-stem ratio, favoring better light use efficiency, and higher nutritional value33. Such traits were sought for a long time to select forage sorghum materials at the detriment of agronomic traits such as plant height and DMY.

Multivariate analyses and Tocher grouping revealed great genetic diversity among the genotypes. DMY and NP were the variables most contributing to such divergence. Ten experimental genotypes comprised Group I, which were taller and more productive than the commercial ones (15, 16, 17, and 18). The other groups were composed of few or only one genotype, indicating they are quite different from each other6. In general, experimental genotypes showed long genetic distances from the commercial ones, primarily from the number 15. Seeking the heterosis phenomenon, crossbreeding genotype number 15 with the experimental ones of Groups I, II, III, V, and VI can generate superior forage sorghum genotypes in the future, combining high DMY and ensilability34,35 .

Lack of significant differences in pH values (3.66, on average) of genotypes’ silages occurred because sorghum is a suitable crop for ensiling, with desirable traits for a good fermentation process such as DM content over 300 g/kg at harvest time, soluble carbohydrate concentration over 150 g/kg, and reduced buffering capacity36. Silage pH values between 3.6 and 4.2 indicate that lactic acid bacteria (LAB) and the lactic acid were mainly responsible for reducing pH20. It is worth noting that genetic variation and phenotypic characteristics, such as DMY and genotype size, did not affect the silage pH. Silages of all genotypes showed significant lactic and acetic fermentation rather than butyric acid and ethanol production, favoring rapid pH decrease, nutrient preservation, and low activity of enterobacteria, fungi, and yeasts. Furthermore, homolactic fermentation predominated in almost all silages since the L/A ratio remained over 3.007.

Higher ammonia nitrogen contents in silages of 11, 16, and 18 genotypes were derived from high CP concentrations found in these silages. Numbers 16 and 18 are dwarf genotypes with higher leaf-to-stem ratio, leading to increased CP content because most of the protein concentrates in leaves4,37. Conversely, ammonia nitrogen below 100 g/kg of total N were found in all silages, indicating no excessive proteolysis by Clostridium bacteria38.

Silage aerobic stability was kept for 118 h, on average, with no significant discrepancies among forage sorghum genotypes. Breaking aerobic stability occurs when silage temperature overcomes the room temperature by 2 °C, due to yeasts consuming residual organic acids after silo opening, and enterobacteria consuming residual soluble carbohydrates20. Aerobic stability over 100 h, such as found in this study, indicates suitable preservation after silo opening due to concentrations of lactic acid, which restricts enterobacteria proliferation, and concentrations of acetic acid, which diminishes fungi and yeast activities39,40.

Number 16 genotype silage excelled with the highest DM content because more panicle and grains were ensiled using this material. Panicles and grains are important for enhancing fermentation traits because they are highly digestible and rich in soluble carbohydrates37. In contrast, silages made of numbers 11 and 14 showed lower DM concentrations because of the low grain proportion at the ensiling time. However, all silages displayed at least 250 g/kg of DM, which is enough to ensure a good fermentation process41. The highest panicle and grain proportion also explained the higher EE concentration in silages of genotypes 16 and 18.

Number 14 genotype silage displayed the highest ash and the lowest OM concentrations. High-yield sorghum genotypes, such as the genotype 14, exhibit a high ability to remove mineral soil nutrients, and some of these accumulate in the vegetative tissue29. Nevertheless, high ash contents lead to low OM concentrations, which in turn is not favorable for silage fermentation since the fermentable soluble carbohydrates are within the OM38.

Silages of genotypes 16 and 18 showed more digestible nutrients and less fiber compounds, mainly because of the higher leaf-to-stem ratio at the harvest time. The highest NDF and ADF contents in the number 10 genotype silage suggest an important lignification to protect from lodging, pests, and diseases. However, this process results in lower digestibility36. In contrast, silages of the commercial genotypes 16 and 18 displayed higher TDN and NFC concentrations. Thus, they are alternatives for conditions where high forage digestibility is necessary (e.g., in high-performance cattle confinements).

Numbers 16 and 18 genotype silages displayed a high degradation rate and low concentrations of indigestible fractions during incubation, since these genotypes have lower lignin concentration than the others. In contrast, silages of genotypes 10 and 11 displayed high indigestible fractions and low potential degradability because of their high lignin contents. Nutrients bound to lignin considerably reduce feed digestibility42,43. However, the highest DDMY found in silages of experimental genotypes should be considered in sorghum breeding for ensiling, because it combines forage yield and the effectively degradable fraction inside the rumen, creating a digestible forage yield index. Thus, assessing silage DDMY can contribute to releasing sorghum cultivars seeking more efficient and sustainable livestock production systems.

Sorghum can be an alternative to corn genotypes for ensiling because they often show comparable nutritional value and agronomic traits, being more adaptable to drought conditions and low soil fertility44,45. In this study, commercial sorghum genotype silages (16 and 18) were comparable to some findings for corn materials in terms of digestibility. Santana et al.46 assessed the in-situ rumen kinetics of silages from four corn genotypes and found 636.7 g/kg of pdDM. We found similar pdDM results for 16- and 18-genotype silages (644.4 g/kg, on average). Silages of experimental genotypes (3, 4, 10, 11, and 14) performed worse compared to corn genotypes found in literature: Hristov et al.47 reported a pdNDF average of 793.0 g/kg when studying the in-situ rumen kinetics of five corn genotypes’ silages. In contrast, our experimental sorghum materials produced silages with only 415.8 g/kg. Nevertheless, the DMY of experimental genotypes 3, 4, 10, 11, and 14 averaged 20.8 t/ha, a superior production to some findings for corn in the literature: Peña et al.48 assessed the DMY of four corn genotypes before ensiling them, and recorded an average of only 12.5 t/ha.

In conclusion, genetic diversity observed in this study allows for selecting forage sorghum genotypes with high DMY and appropriate fermentation and nutritional parameters for ensiling. Crossbreeding the experimental genotypes 3, 4, 10, and 14 with the commercial number 15 can be useful in this sense. Experimental genotypes 3, 4, 10, and 14 enable producing silages with similar fermentation profiles to the commercial genotypes 16 and 18 but with more digestible nutrients per area. Evaluating the digestible dry matter yield as an index variable to select genotypes brings a new perspective on sorghum breeding for ensiling. Ensiling the genotypes 3, 10, and 11 should be incentivized in semiarid regions, considering their lower dry matter yield seasonality. Further studies can be conducted to assess these genotypes in the long term, and their silages can be tested in animal performance trials.