In the present study, we have used human biopsies collected in the frame of COMET biobank to evaluate the impact of several homogenization methods on total RNA recovery. We have found that different disruption techniques and homogenizing buffers influence the purity and some quality markers of RNA and can also impact quantification of mRNAs by RT-qPCR.
Our study shows that homogenization methods clearly impact the recovery of RNAs from human tissues. GentleMACS Dissociator combined to QIAzol reagent appears to be the best method to obtain the highest RIN value and 28S/18S ratio for all the tissues tested, except for muscle. As compared to FastPrep-24 Instrument, GentleMACS Dissociator seems to be more efficient to dissociate human tissues in one cycle, avoiding repeated lysis cycles and thus preserving RNA integrity. Such a difference was also evidenced on skin biopsies (which are very difficult to dissociate), where GentleMACS Dissociator is the only system allowing a complete homogenization of human skin samples, as compared to FastPrep-24 Instrument13. Manual disruption by syringe/needle can be a good alternative method which provides RNA with comparable quality to GentleMACS Dissociator, but only in case of soft tissues14 or isolated cells15. As also shown for skin biopsies13, we observe that RNA yields tend to be higher with GentleMACS Dissociator, but no statistical correlation can be found between biopsy weight and RNA levels, i.e., larger biopsies does not necessarily yield to the highest amounts of RNA. This variation can be due to heterogeneity in cellular composition and transcriptional activity in human tissues/organs, with implications on RNA profiles. In human liver biopsies16, numerous cell types can be identified by single cell RNA sequencing, including but not limited to hepatocytes, duct cells, endothelial cells, hepatic stellate cells, Kupffer cells, and immune cells. Moreover, the authors also found a transcriptional heterogeneity between hepatocytes according to their spatial localization along the liver lobule16. Adipose tissues are also heterogeneous in their composition, as they contain adipocytes, preadipocytes, mesenchymal stem cells, vascular cells, and inflammatory cells17. Furthermore, at least three distinct subpopulations of white preadipocytes have been evidenced in mice, with unique gene expression profile18. In skeletal muscles, single nuclear sequencing of different nuclei present in synticial myofibers revealed transcriptomic heterogeneity is not related to fiber type differences19.
For all homogenization methods used and tissue types, the A260/280 ratio reaches values ≥ 1.8, which are in the range of what is found in human tissues and cell lines20, suggesting limited protein contamination. The A260/230 ratio is however ≤ 1.8, with the lowest value obtained with GentleMACS Dissociator. This difference is also observed for DNA extraction, where GentleMACS Dissociator leads to inferior an A260/230 ratio when compared to bead beating-based homogenization methods21. A low A260/230 ratio is indicative of the presence of organic compounds such as phenol or guanidine, which is inherent to methods using chaotropic lysis buffers12. Nonetheless, guanidine thiocyanate contamination was shown not to interfere with downstream molecular applications, up to 100 mM in RNA samples22. Indeed, applications such as cDNA synthesis, in vitro transcription20, sequencing23, or microarrays24 were not affected by low A260/230 ratios.
To assess RNA quality, we used the RIN value, now preferred to the 28S/18S ratio that shows a high variability for RNA extracted from human samples, which is not necessarily related to poor RNA quality20. Nonetheless, for RNA extracted from adipose tissues with FastPrep-24 Instrument, the 28S/18S ratio indicated the absence of the 28S rRNA band, preventing the calculation of a RIN value, which may correspond to a mechanical breaking of 28S rRNA. Indeed, this rRNA is more susceptible to be affected by enzymatic degradation, mechanical shearing, or freezing procedure than 18S rRNA, as a result of its important size and the presence of “hidden break” in its polynucleotide chain25, 26. In addition, we observed that RINs were lower in liver than in other tissues, probably due to the fact that hepatic tissue contains high levels of RNAses10.
RT-qPCR is another method used to evaluate RNA integrity through analysis of mRNA expression of target genes. To avoid underestimation of RNA fragmentation/degradation, we used oligo(dT) primers for reverse transcription, rather than random primers, as previously suggested27. In our experiments, we observe a certain degree of variability between patients’ samples, suggesting an impact of their pathophysiological conditions on RNA profile. However, we find some housekeeping genes suitable for normalization of qPCR experiments such as β-actin in adipose tissues and muscle or TBP in liver. Nonetheless, when using human samples, variability between patients can only be minimized, as shown by Kim et al.28. The authors observed that no housekeeping gene was found to vary by less than twofold in liver diseases such as cirrhosis or hepatocellular carcinoma.
When normalizing Cts according to RIN (the Ct/RIN ratio), QIAzol and GentleMACS Dissociator gave improved results as compared to other buffer and extraction methods. In addition, an inverse correlation between RIN and Cts is found for β-actin in liver and muscle but not for other genes and tissues. To explain such discrepancies, we suggest that the β-actin mRNA is more sensitive to degradation by RNases or has a shorter half-life in these tissues, as it was shown previously in human cell lines29. In addition, we cannot exclude that RNA isolation methods differentially influence the recovery of some mRNA species, as previously suggested in studies comparing phenol extraction or RNA isolation kits30, 31. For adipose tissue, Cts remains unchanged regardless RIN observed. This low RIN may not necessarily reflect low quality mRNA despite the observed degradation of 28S rRNA. This suggests that, in adipose tissues, 28S rRNAs undergo degradation by mechanical breaking rather than degradation by enzymatic activity of RNases, with consequential reduction in RIN.
In the literature, the relationship between RIN and subsequent analysis of mRNA expression levels by RT-qPCR is not clear. For instance, a correlation between RIN and apparent expression levels of housekeeping genes is found during the progressive degradation of RNA samples, of unknown tissue origin26. Concerning human samples, Imbeaud et al.20 found that RNA quality metrics (Degradometer and RIN analysis) are predictive of relative gene expression analysis in RNA samples of disparate quality. In another study performed on a lung biobank5, the authors observed that gene expression measurements are more influenced by interpatient differences than other variables. Even for two samples issued from the same individuals, they observed that variability in mRNA expression levels was mostly within the range of a factor of 2.
Nonetheless, Kap et al. propose to divide RNA samples into “fit for purpose” groups according to RIN, with samples having RIN above 5 suitable for RT-qPCR32. When RIN are lower than 5, they suggest that samples can be also used for RT-qPCR, but only for amplification of small size amplicons32. For genome-wide studies such as microarrays or RNA sequencing, inclusion of samples with highest quality metrics is recommended to avoid experimental bias, as suggested by the literature32,33,34,35.
We also evaluated an alternative method for sample preparation for biobanking, using a 24 h-incubation in RNA stabilizing reagents like RNAlater before freezing6. For COMET biobank, we have chosen direct freezing of tissues to allow several applications to be performed on the same type of sample, such as genomics, proteomics, metabolomics, and histological analysis. However, to avoid tissue damage, we have reduced the time of ex-vivo warm ischemia to a minimum, as the biopsy is immediately collected after removal from the patient. Moreover, we prepared samples at low temperature within a short period of time before snap-freezing (15 or 20 min depending on the tissues). However, using RNA stabilizing reagents have the advantage of avoiding the need of freezing facilities in the clinic surgical block. Nonetheless, the impact of protective reagents is not clear in the literature, as a previous study showed that gene expression analysis is not altered after cold ischemia of 180 min, independently of pre-treatment with RNAlater5. Therefore, RNAlater does not necessarily improve RNA quality, if samples are prepared in a limited period of time before freezing. This data is in accordance with our experiments showing good quality of RNA when using QIAzol/ GentleMACS Dissociator. The effect of RNAlater is more obvious in fat and liver tissues when tissue homogenization is performed with FastPrep-24 Instrument, with both tissue types becoming harder and thus more resistant to the stress induced by mechanical dissociation.
In conclusion, we have demonstrated that different disruption techniques and homogenizing buffers impact the purity and some quality markers of RNA isolated from metabolic tissues and can also affect mRNA quantification by RT-qPCR. To optimize quality of RNA preparation, we suggest to prepare human samples at low temperature in a short period of time and to implement a gentle homogenization method like GentleMACS Dissociator combined to QIAzol reagent.
