Evaluation of techniques and suitable sample number for soluble carbohydrate extraction in stem bases of elephantgrass

Methods for extracting soluble carbohydrates were compared, aiming at optimizing sample number and size, for level determination using Teles' reagent. Oven-dried ground stem basal segments of elephantgrass (Pennisetum purpureum, Schum.) cv. Napier, carbohydrate standards (glucose, maltose, sucrose, and starch), and sample-standard combinations (CSs) were extracted through the following procedures: (a) REF: reflux for 180 min (in 0.6N HCl); (b) SHA: orbital shaking for 60 min (in 50% ethanol), followed by an acidic hydrolysis (1.2N HCl) in water bath (90-100 degrees C) for 60 min, and neutralization with 5% NaOH; (c) HET: 80% ethanol overnight and water bath for 180 min; (d) ETH: 0.6N HCl (in 80% ethanol) overnight and water bath for 180 min; and (e) DIG: 80% ethanol overnight, addition of 1.2N HCl, and water bath for 60 min. REF yielded the highest extraction level in the elephantgrass sample, followed by ETH and SHA. Levels with HET and DIG were the lowest. However, REF was excessively time consuming and effected starch extraction, being not the best choice for soluble carbohydrate evaluations. HET showed no sucrose extraction and DIG did not perform well with the sample, both being considered not applicable. ETH, although exhibiting a marked sample extraction, showed the poorest standard results, specially with glucose. SHA was found to be the best method, because it was the only one that effected total sugar (except for maltose) and no starch extraction. Results pooled for all treatments revealed non-significant paired-mean contrasts between separate and combined (CSs) sample-standard associations. This indicates that the strategy of carrying out comparisons with standards was appropriate. For the SHA method, the Student's t distribution of the means revealed that the minimum number of samples for maintaining the highest possible precision was 5 replications. On the other hand, increasing sample number up to 9 replications did not improve data precision.