Blood collection tube and RNA purification method recommendations for extracellular RNA transcriptome profiling

Abstract

Blood-based extracellular RNA (cell-free RNA; exRNA) biomarkers require validated sample collection, processing, and quantification procedures. No study to date has systematically tested pre-analytical variables affecting transcriptome-wide exRNA analysis. By evaluating their impact on deep transcriptome profiling of microRNAs and mRNAs in blood plasma or serum, we compared ten blood collection tubes, three blood processing time intervals, and eight RNA purification methods. In addition, we assessed interactions among a selected pre-analytical variable set, resulting in 456 extracellular transcriptomes. Blood preservation tubes failed to stabilize exRNA and RNA purification methods differed significantly in performance, causing variations in concentration, detected gene numbers, replicability and observed transcriptome complexity. Critical interactions between tubes, purification methods and time intervals were identified. We provide 11 analytical performance metrics for exRNA quantification methods and put forward recommendations for both users and manufacturers of RNA purification methods and blood collection tubes, collectively, essential groundwork for exRNA-based precision medicine applications.

Fig. 1. Workflow in the extracellular RNA Quality Control (exRNAQC) study.

To evaluate the 8 exRNA purification methods (upper left panel), 2 blood draws from a single individual were performed to separately apply mRNA capture or miRNA sequencing. To compare RNA purification performance, 9 performance metrics were calculated. Blood was drawn from 9 individuals to evaluate 10 blood collection tube types, including 5 classic and 5 preservation tube types (upper right panel), at 3 time intervals between blood draw and processing. Preservation tubes were processed immediately (T0) and after 24 (T24) and 72 (T72) hours and classic tubes were processed immediately (T0) and after 4 (T04) and 16 (T16) hours. Both mRNA capture and miRNA sequencing were performed, and the data was analyzed using 5 performance metrics. Based on the number of miRNAs and mRNAs detected and replicate variability metrics, a dedicated selection of precise and sensitive exRNA purification methods and blood collection tubes was further evaluated in exRNAQC phase 2. For both mRNA capture and miRNA sequencing in phase 2, 5 individuals were sampled to test 3 blood collection tubes and 4 RNA purification methods. Interactions between RNA purification methods, blood collection tubes and processing time intervals were assessed by 6 performance metrics. MAP=MagNA Pure method, MAX=Maxwell method, MIR=miRNeasy method, MIRA=miRNeasy Advanced method, MIRV=mirVana method, MIRVE=mirVana method with purification protocol for small RNA enrichment, NOR=Norgen method, NUC=NucleoSpin method, QIA=QIAamp method. Designed with Freepik (free license) and Servier Medical Art (CC BY 4.0).

Fig. 2. RNA purification methods strongly influence mRNA and miRNA sequencing.

Performance metrics are shown for both mRNA capture (left panels) and miRNA (right panels) sequencing. For each unique RNA purification-plasma input volume combination, 3 technical replicates were analyzed (n = 39 for mRNA capture & n = 45 for miRNA sequencing). Absolute numbers of detected mRNAs (a) and miRNAs (b) that reached the count threshold (see “Methods”) are shown. High numbers indicate good performance. Endogenous mRNA (c) and miRNA (d) concentration. Values are log rescaled to the lowest mean of all methods and transformed back to linear scale. The mean and 95% confidence interval are shown. High concentrations indicate good performance. Replicate variability based on ALC at mRNA (e) and miRNA (f) level, respectively. Small ALC indicates good performance. Overview of all performance metrics at mRNA capture (g) and miRNA (h) sequencing level, respectively, after transforming the values to robust z-scores. High z-scores indicate good performance. Rows and columns of the heatmaps are clustered according to complete hierarchical clustering based on Euclidean distance. Average z refers to the mean of robust z-scores for a specific RNA purification method. The number that follows the name of the purification method is the plasma input volume (in ml). MAX=Maxwell method, MIR=miRNeasy method, MIRA=miRNeasy Advanced method, MIRV=mirVana method, MIRVE=mirVana method with purification protocol for small RNA enrichment, NOR=Norgen method, NUC=NucleoSpin method, QIA=QIAamp method.

Fig. 3. Preservation tubes do not show robust performance over time.

Per blood collection tube and per performance metric, a summary of mean fold-changes (FC) between immediate processing and the 2 selected processing time intervals is given for both mRNA (a) and miRNA (b) profiling. Ideally, the mean FC of the performance metrics approaches 1, indicating that there is little change over time. Per tube type (preservation and classic) tubes are ranked by mean value across all metrics from low (top) to high (bottom). Note that different donors were sampled and that tubes were processed at different time intervals for preservation and classic blood tubes.

Fig. 4. Blood collection tubes impact release of immune cell RNA over time.

Colored cells represent adjusted p-values < 0.05 from beta regression models with random effects for all cell types. Tukey's method was used for pairwise comparisons (two-sided testing) while correcting for multiple testing. P-values smaller than the minimum representable value in R (1e-16) are annotated as < 1e-16. Non-significant p-values point towards blood collection tube stability over time. T0=immediate blood processing, T04, T16, T24, T72=plasma prepared 4, 16, 24 and 72 hours after blood draw, respectively. Note that different donors were sampled and that tubes were processed at different time intervals for preservation and classic tubes.

Fig. 5. RNA purification method selection for exRNAQC phase 2 for mRNA and miRNA analysis.

Median robust z-score (see “Methods”) per method-input volume combination (13 for mRNA, 15 for miRNA) shown for the number of detected mRNAs (a) or miRNAs (b) and replicate variability metrics. The number in the labels is the plasma input volume (in ml). MAX=Maxwell method, MIR=miRNeasy method, MIRA=miRNeasy Advanced method, MIRV=mirVana method, MIRVE=mirVana method with purification protocol for small RNA enrichment, NOR=Norgen method, NUC=NucleoSpin method, QIA=QIAamp method.

Fig. 6. Interactions between pre-analytical variables should be considered when comparing RNA purification method or blood collection tube performance.

For both mRNA capture and miRNA sequencing, 5 biological replicates were used for each of the 18 unique tube (n=3), purification method (n = 2) and time interval (n=3) combinations (total n=90). Shown are the interactions between pre-analytical variables for mRNA capture (a) and miRNA (b) sequencing. P-values correspond to the Wald test for the terms in the linear mixed-effects model (two-sided testing).

Fig. 7. Recommendations for users and RNA purification and blood collection tube manufacturers.

The left panel summarizes the exRNAQC recommendations for users. The right panel describes the exRNAQC recommendations for manufacturers of RNA purification methods and blood collection tubes. gDNA=genomic DNA, MAX=Maxwell method, miRNA=microRNA, MIR=miRNeasy method, MIRA=miRNeasy Advanced method, mRNA=messenger RNA, QIA=QIAamp method. Designed with Freepik (free license).

Fig. 8. The exRNAQC study represents the most comprehensive analysis of pre-analytical variables in the context of exRNA profiling.

The heatmap shows studies (raw data and references in Supplementary Data 1) evaluating pre-analytical variables based on both mRNA and miRNA (top part), miRNA only (middle part) and mRNA only (bottom part). The numbers in brackets after the column names indicate the scale range. The darker the coloring, the more items were studied. Studies marked with a solid black circle evaluate interactions between pre-analytical variables. The exRNAQC study outperforms previous studies analyzing pre-analytical variables impacting exRNA analyzes in terms of the combination of evaluated metrics and shows uniqueness by studying the impact on both miRNA and mRNA.