Concerning the yield of the extraction process for each of the methods tested in human Jurkat T cells, ANOVA and Tukey analysis on data from the quantification by spectrophotometry using NanoDrop™ system revealed significant differences (P <0.001) for all methods tested. Pair-by-pair comparison showed statistically significant differences for all methods, except for QIAgen™ vs. QIAcube™, and Canvax™ vs Maxwell® methods which do not substantially differ. The highest yields were obtained using TRIzolTM, QIAgen™, or QIAcube™ (Table 1). The semiautomated Canvax™ and automated Maxwell® extraction methods provided worse RNA yields with respect to the rest of the methods. Similarly, quantification by fluorimetry using Qubit™ showed, comparatively, a significant (P < 0.001) higher performance in the RNA samples obtained with TRIzolTM, QIAgen™, or QIAcube™ methods, and a substantially lower performance with Bio-Rad, Canvax™, and Maxwell®, respectively (Table 1). In all cases, the yield was within the range proposed for each method by its respective manufacturer. Six extraction methods were tested (QIAgen™, QIAcube™, Bio-Rad, Monarch®, Maxwell®, and TRIzolTM) in PBMC. Only QIAcube™ and TRIzolTM methods showed statistically significant pair-by-pair differences and the highest yields (Table 1). Fluorimetry confirmed the results, showing the lower performance when Maxwell® and Bio-Rad were used. All results were within the range proposed for each method by its respective manufacturer. Besides, a remarkable difference was observed depending on the matrix used for extraction. Best yields were obtained for the human Jurkat T cells, reaching up to 80% higher in QIAcube™ and QIAgen™ methods, remaining unchanged for the Maxwell® method.
From these findings, the best yield was obtained for QIAcube™ (automated), QIAgen™ (semi-automated), and TRIzolTM (manual) methods. In line with our results from Jurkat T cells, a previous work from Tavares et al. (2010) using SK-N-MC neuroblastoma cells reported that semi-automated methods had better yield than manual ones. However, for PBMC, a yield up to 1.18 times higher was found in manual extraction methods when were compared with commercial kits (Ruettger et al. 2010, Oliveira-Alves et al. 2016), in agreement with our study. The decrease in the yield of some semi-automated methods, such as Bio-Rad and CanvaxTM, may be due to the use of β-mercaptoethanol (Mommaerts et al. 2015), used to deactivate RNases but with denaturing effects on guanidinium isothiocyanate in the lysis buffer (Kwiatkowski and Kivins 2004). Additionally, the poorer results in Canvax™ and Maxwell® could be attributable to the probable contamination in the eluted RNA such as the remains of the magnetic microspheres, interfering in the spectrophotometric quantification, and even in subsequent applications of the RNA samples (Stulnig and Amberger 1994; Liu et al. 2009; Martín-Nuñez et al. 2012).
To evaluate the purity of RNA, absorbance ratios at and 260 nm/280 nm (DNA/protein) and 260 nm/230 nm (DNA/contaminants) were determined. Accordingly, results showed acceptable 260/280 ratio values ranged from 1.77 to 2.1 for all the methods tested, with exception of Maxwell®, which showed a 260/280 ratio of 4.2 and 6.27 in human Jurkat T cells and PBMC, respectively (Table 1). Only Maxwell® showed significant differences (P <0.001) when was compared with the rest of the methods. The lower 260/230 ratios were obtained with Maxwell® and TRIzolTM methods for human Jurkat T cells, and only using TRIzolTM in PBMC. Best results, among 1.96-2.15, were obtained for QIAcube™, Monarch®, and CanvaxTM methods (Table 1). In this respect, lower 260/230 ratios may indicate the presence of compounds absorbing at 230 nm such as proteins (Stulnig and Amberger 1994), guanidine HCL, EDTA, carbohydrates, lipids, salts, or phenol (Psifidi et al. 2015). In comparison with PBMC, the yield obtained in Jurkat T cells was about 5 times higher as corresponding to a more homogeneous matrix with fewer interfering substances than biological fluids.
The analysis of the integrity of RNA revealed the presence of both ribosomal 28S and 18S bands, showing the typical 2:1 proportion of intensity, in all RNA samples either from human Jurkat T cells (Fig. 1a) or from PBMC (Fig. 1b), irrespective of the method of extraction, as usually reported (Ruettger et al. 2010; Gan et al. 2016), excepting for Maxwell®. In this case, a single very intense band was observed in both biological matrices (Fig. 1), located close to the 28S band but not having correspondence with this. This could be due to the aforementioned elution of the RNA together with the magnetic microspheres used in this method, which might cause changes in the migratory patterns of these RNAs in agarose gels (Martín-Núñez et al. 2012; Mommaerts et al. 2015). For 5S band only was observed for Monarch, QIAgen™, QIAcube™, and Bio-Rad methods in human Jurkat T cells, but not for any method in PBMC.
To determine the functionality of RNA samples PCR and real-time PCR were performed in genes whose expression level was similar in both human Jurkat T cells and PBMC. All samples were positive for the genes studied, and an amplification band of the desired size was obtained regardless of the type of extraction method and the matrix evaluated (Fig. 1c, d). Multiple comparisons showed significant differences for CT values in the Bio-Rad, Maxwell®, and TRIzolTM methods (Fig. 1g, h). All RNA obtained in our study was shown as functional by conventional PCR and real-time PCR, as observed in related studies (Bayatti et al. 2014; Chauhan et al. 2018). However, statistically significant differences were found in the results of real-time PCR for Bio-Rad, Maxwell®, and TRIzolTM methods when were compared with the rest of the methods tested in both matrices. These results suggest these methods could compromise the functionality of the RNAs obtained, so results could not be completely reliable.
Finally, an important fact to consider when comparing nucleic acid extraction methods is to know the time, labour, and cost analysis for each method (Chacon-Cortes et al. 2012; Psifidi et al. 2015). The feasibility of each method in terms of time and costs per sample is reported in table 2. The fastest extraction methods were the Canvax™ and Maxwell®, while the most time-consuming were the QIAgen™, QIAcube™, and manual, which required overnight incubation. On the other hand, the manual was the cheapest one, followed by the Monarch® and Canvax™, two semiautomated methods requiring lower manipulation by specialized personnel than the manual method. Additionally, manual or semiautomated methods can process 24 samples, approximately twice those of automated methods QIAcube™ and Maxwell®. The analysis performed in this study showed that automated methods were quite expensive and had a low capacity to work with several samples simultaneously, while the manual method (TRIzolTM) extended the protocol more than one day, increasing the risk of degradation. Therefore, semi-automated methods could be more advisable because they have a more affordable price, a greater capacity than other methods, and less execution time.
In summary, our data revealed differences attributable to the method chosen. Given our results, QIAcube™ and semi-automated extraction methods were perceived as the best options because of a lower variability, good functionality, and lower cost. Noteworthy Monarch® appeared as the second-best option because it showed quality indicators closer to expected, which guarantees reliable results. Despite larger studies with a greater number of methods and matrices would be advisable, the variety of methods compared in this study emphasize the relevance of the choice of an optimal RNA extraction method in biomedical and nucleic acid research.