RNA quality and quantity
Samples were divided into 6 groups, according to their i. extraction method (manual vs automated); ii. TURBO DNA Free treatment (treated vs untreated), and iii. starting volume of whole blood (16 𝜇l, 33 𝜇l, 50 𝜇l and 66 𝜇l). The elution volume was 50 𝜇l for all the extractions performed. Thus, we have used the Qubit (fluorescence-based) concentration values (ng/𝜇l) for comparative analysis. As expected, the RNA concentration increased parallelly to the increased volume of blood used for the extraction, with the concentration obtained from 66 𝜇l of blood being significantly higher as compared to the concentration obtained from 16 𝜇l of blood (Wilcoxon matched-pairs signed-ranks test, p = 0.0039, Fig. 4A, Table 1). Interestingly, the variability (measured by the standard deviation) increased parallelly to the amount of whole blood used for the extraction (Fig. 4A, Table 1), suggesting that the sampling might give more consistent results for lower volumes of blood.
Table 1
Summary of the RNA QC metrics according to the sample groups
Group
|
Qubit Concentration
(ng/𝜇l; Average ± SD)
|
RIN
(Average ± SD)
|
A260/A230
(Average ± SD)
|
A260/A280
(Average ± SD)
|
Manual
Untreated
50 𝜇l blood
|
11.62 ± 2.53
|
8.16 ± 0.26
|
1.45 ± 0.55
|
2.14 ± 0.14
|
Manual
TURBO Treated
50 𝜇l blood
|
7.95 ± 1.39
|
8.62 ± 0.12
|
1.99 ± 1.45
|
1.31 ± 0.09
|
Automated
TURBO Treated
16 𝜇l blood
|
3.50 ± 0.92
|
7.42 ± 0.51
|
0.63 ± 0.29
|
2.19 ± 0.41
|
Automated
TURBO Treated
33 𝜇l blood
|
7.82 ± 0.72
|
7.29 ± 0.42
|
1.04 ± 0.38
|
2.24 ± 0.21
|
Automated
TURBO Treated
50 𝜇l blood
|
8.45 ± 2.27
|
7.39 ± 0.47
|
1.2 ± 0.17
|
2.06 ± 0.23
|
Automated
TURBO Treated
66 𝜇l blood
|
11.61 ± 2.44
|
7.26 ± 0.47
|
1.16 ± 0.29
|
2.12 ± 0.09
|
When using 50 𝜇l of blood, we found a significant decrease of the RNA concentration in the manual TURBO-treated protocol and the automated TURBO-treated protocol as compared to the manual untreated protocol (Wilcoxon matched-pairs signed-ranks test, p = 0.0010 and p = 0.0129, respectively, Fig. 4B). While the RNA concentration values correlated significantly for the manual treated and untreated protocols (Spearman r test, p = 0.001) we found no significant correlation between the manual and automated protocols that included the TURBO DNA Free treatment, suggesting that workflow-specific steps might affect the RNA concentration more than the TURBO DNA Free treatment and biological variables (i.e. individual cell counts, Fig. 4C and 4D).
Overall, all the RNA isolated was of good quality. No significant difference in RIN value was observed across samples processed from different volumes of starting material (Fig. 4E). Nevertheless, the RIN values obtained from the different samples varied across the experimental groups with the manual extraction method producing overall higher RIN values as compared to the automated methods (Fig. 4F, Table 1).
The A260/A230 values varied across the experimental groups with the automated TURBO-treated samples of 16 𝜇l blood producing the lowest A260/A230 ratio (0.63 ± 0.29). The A260/A280 values were > 2 for all the experimental groups except for the manual TURBO-treated 50 𝜇l blood samples that displayed an average A260/A280 ratio of 1.31 ± 0.09.
We next sought to assess the effect of TURBO DNA Free treatment on the RNA yield and RIN values. For the manual protocol the TURBO DNA Free treatment resulted in an overall yield reduction > 25% (Table 2), while the RIN values increased slightly (Table 2). When we compared the yield and RIN values in the automated TURBO DNA Free protocol to the manual untreated protocol, we found a yield reduction similar to the one induced by the TURBO DNA Free treatment in the manual protocol (Table 3). However, the TURBO DNA Free treatment induced a RIN reduction between 3.36% − 11.51% in the automated protocol as compared to the manual untreated protocol (Table 3).
Table 2
RNA yield reduction and RIN increase induced by TURBO DNA Free treatment in the manual protocol
Subject ID
|
Average Yield (ng)
Manual Untreated
|
Average Yield (ng)
Manual Treated
|
Yield reduction (%)
|
Average RIN
Manual Untreated
|
Average RIN
Manual Treated
|
RIN increase (%)
|
S1
|
495.00
|
367.50
|
25.76%
|
8.33
|
8.70
|
4.50%
|
S2
|
720.00
|
485.00
|
32.64%
|
7.98
|
8.50
|
6.58%
|
S3
|
528.75
|
340.00
|
35.70%
|
8.18
|
8.68
|
6.12%
|
Table 3
RNA yield and RIN reduction induced by TURBO DNA Free treatment in the automated protocol (50 𝜇l blood)
Subject ID
|
Average Yield (ng)
Manual Untreated
|
Average Yield (ng)
Automated Treated
|
Yield reduction (%)
|
Average RIN
Manual Untreated
|
Average RIN
Automated Treated
|
RIN reduction (%)
|
S1
|
495.00
|
373.83
|
24.48%
|
8.33
|
7.37
|
11.51%
|
S2
|
720.00
|
546.67
|
24.07%
|
7.98
|
6.90
|
13.48%
|
S3
|
528.75
|
347.17
|
34.34%
|
8.18
|
7.90
|
3.36%
|
Gene expression analyses
We assessed the reproducibility of gene expression profiles obtained from the different RNA extraction methods. Out of the 60 samples, 2 samples extracted manually from the same donor, generated data of low quality and were removed from downstream analyses as the library output of these samples was much lower than the output recommended for Quantseq mRNA. This could be due to the low purity of the samples as their Nanodrop readings demonstrated a high A260/230 ratio. Two additional samples produce libraries of suboptimal yield and were labeled as “low conc. library” for further analyses.
When we looked at the distribution of the VST counts across the samples, we noticed an overall homogeneous distribution except for sample S1_B1_Man_TTA that displayed a higher VST median as compared to the sample set (Fig. 5A). Interestingly, this was one of the two samples that gave low library yield, suggesting that the suboptimal library preparation affected the VST count.
To explore the effect of the different variables assessed in the study on the complete transcriptomic data we have used principal component analysis (PCA).
The assignment of the samples to the three individuals accurately predicted their distribution in a three-dimensional space suggesting that their transcriptional signatures can be retraced to the individual biology (Fig. 5B). Contrarily, the different extraction methods and the DNase treatment seemed to have a negligible effect on the sample distribution (Fig. 5C), although samples processed manually displayed a higher variability. This might be explained by the fact that biological variables might have a larger effect on the transcriptomic data as compared to analytical variables (i.e., isolation method, DNase treatment). It should be noted that in the PCA plots displayed in Fig. 5B & Fig. 5C, the variance of PC1 was 45%, indicating that the transcriptomic data of the samples was overall quite similar.
Nevertheless, the correlation matrix identified an overall high degree of similarity across the samples isolated with the automation method as compared to the ones isolated manually, irrespective to starting blood volume and DNase treatment (Fig. 6A). When performing correlation analysis only on the samples isolated on the automated system, we found an almost perfect correlation of samples belonging to the same individual, irrespective to the starting blood volume (Fig. 6B), supporting the sampling of volumes of blood as low as 16 𝜇l as an efficient method for whole blood transcriptomic profiling.
Processing time/sample throughput comparison between the manual and automated workflow
We have also evaluated the processing time of the standard manual extraction protocol and the automated protocol developed in-house on the Hamilton NGS Star platform.
The manual workflow overall takes about 75 min hands-on-time and 65 min incubation time, while the automated workflow overall takes 20 min hands-on-time and 50 min incubation time. The above calculations refer to the processing of a batch of 24 samples. However, the sample throughput can be significantly increased in the automated workflow as the Hamilton NGS STAR system is equipped with 3x32 sample tube carriers and it can process 96 samples per batch. Additionally, faster bead clean-up steps can be adopted to this method if the liquid handler is equipped with 96-Multi Probe Head.