Evaluation of miRNA recovery during isolation for different biofluids
The robustness of an assay is an essential parameter for miRNA detection. To test the miRNA recovery, the levels of a synthetic spiked-in miRNA (cel-miR-39) in six matched human serum, urine, bile and perfusate samples were measured using the four aforementioned methods. The direct output of the PCR results showed that the median Cq values were the lowest for the RN method in all 4 analyzed biofluids (23.15, 23.26, 22.3, and 22.85 for serum, urine, perfusate, and bile, respectively) and the most consistent, as determined by the smallest range, in urine, perfusate, and bile (23.01-24.45, 21.76-22.59, and 22.21-23.62, respectively). To determine the relative improvement in detection using the RN method in comparison with the QP, NG and CU methods, relative detection levels were converted to percentages with the mean of the RN method for each biofluid set to 100% (Fig. 1). The mean recovery of cel-miR-39 from serum using the RN method was significantly better than the recovery using the QP and CU method with levels (mean% ± SEM) reaching only 33.8% ± 14.4 and 45.8% ± 6.1 of the levels of the RN method, respectively. Recovery from serum using the NG method was also less than half of the RN method (36.1% ± 6.9), but not significantly different from RN (Fig. 1A). The recovery improvement of cel-miR-39 was even more pronounced in RNA isolated from urine samples. The RN method performed significantly better than methods QP, NG, and CU, with recovery levels of 32.5% ± 9.3, 41.5% ± 8.9, and 10.3% ± 3.3, respectively when compared with the RN method (Fig. 1B). The significantly better performance of the RN method was also observed for RNA isolated from perfusate with recovery levels reaching only 29.8% ± 4.4, 10.8% ± 2.3, and 22.2% ± 5.8 for QP, NG, and CU, respectively, compared with the RN method (Fig. 1C). Although recovery of RNA from bile was better with method RN, the levels were not significantly improved when compared with the QP and CU method (66.6% ± 15.5 and 56.1% ± 14.0). Recovery from bile using the NG method, however, was significantly lower (1.6% ± 0.3) (Fig. 1D).
The effect of the different RNA isolation methods on the yield of endogenous miRNA present in the four biofluids is shown in Fig. 2. For serum, miR-122, miR-222 and miR-21 were measured (Fig. 2A). For urine, miR-30a and miR-92e were determined (Fig. 2B), for the biofluids bile and PF both miR-122 and miR-222 were measured (Fig. 2C and D). The overall trend in detection levels of the different RNA isolation methods within one biofluid of the endogenous miRNAs is similar to that of the spiked-in miRNA cel-miR-39. Some clear differences, however, are also observed. Detection in RNA from serum after isolation with method CU was comparable to detection in RNA isolated with method RN (Fig. 2A).
Isolation methods differ in the co-purification of heparin
As shown in multiple studies, anti-coagulants can strongly affect the analysis of miRNAs for biomarker discovery. Heparin contamination is known to negatively influence the quantification of miRNAs in blood samples and urine, as well as in perfusate. To circumvent this specific inhibition, the use of heparinase I in cDNA synthesis was propagated [34,35,36]. To assess any possible differences in co-isolation of heparin between the different methods, four perfusate samples with proven heparin contamination were selected from a previous study [36]. Total RNA was isolated using methods QP, RN, and NG, and RT-qPCR was performed in the absence (−) or presence (+) of heparinase I. The RT-qPCR results for cel-miR-54 in the presence and absence of heparinase I showed a clear increase in relative detection levels for methods RN and QP, while the improvement with RNA isolated with the NG method is only marginal (Fig. 3A). Results for recovery of cel-miR-39 were presented as relative detection levels in Fig. 3B. Co-purification of heparin was most prominent in RNA isolated using methods QP and RN, where treatment with heparinase I resulted in a mean increase of the relative detection levels of 804-fold and 236-fold, respectively. Heparinase I treatment during cDNA synthesis of RNA isolated using the NG method, on the other hand, only resulted in a 2-fold improvement (Fig. 3B). As already shown previously, relative detection levels after heparinase I treatment were the highest in RNA isolated using method RN. We also tested the effect of heparin contamination on endogenous miRNAs miR-122 and miR-222, two miRNAs that have shown to be well detectable in perfusate samples [17]. Again, detection of miRNAs clearly improved when heparinase I treatment was applied to RNA isolated through methods QP and RN. RNA isolated with method NG, again, showed little, if any, evidence of heparin co-purification as no improvement was observed after treatment with heparinase I for miR-122 and miR-222 (Fig. 3C and D). The RT-qPCR results for cel-miR-54 in the presence of heparinase I also showed that no additional inhibitory compounds, significantly contributing to a decrease in detection levels, were isolated that are specific to one of these 3 methods as mean relative detection levels after heparinase I treatment for methods RN, QP, and NG were 856, 820, and 825, respectively (Fig. 3A).
The RNA isolation procedure NG is not completely insensitive to heparin contamination
Initially considered as a method that could isolate RNA without heparin contamination, results suggested that a certain level of heparin contamination could still be detected in the RNA samples isolated with the NG method in the absence of heparinase I (Fig. 3). To follow up on this observation and to determine to what extend and to what level heparin contamination is present in RNA samples isolated with the NG method, clean (un-used) UW solution was spiked with the standard amount of cel-miR-39 (200 amol) in combination with increasing amounts of heparin. RNA was isolated using either the NG or the RN method, and RT-qPCR detection of cel-miR-39 and cel-miR-54 - the latter added during cDNA synthesis - was performed in the absence of heparinase I. For both cel-miR-54 (Fig. 4A) and cel-miR-39 (Fig. 4D), the reduction of detection levels with increasing amounts of heparin was very similar between methods RN and NG, suggesting that the latter procedure does not prevent heparin contamination. To determine whether the observed discrepancy was caused by components only present in body fluids, urine and serum samples from four patients were pooled and also spiked with increasing amounts of heparin. Spike-in miRNAs cel-miR-39 and cel-miR-54 were determined as described for the clean UW samples. Nonlinear regression comparison (i.e. curve fitting) of dose-response data on log-transformed, normalized duplicates was used for statistical analysis. RNA isolated with the NG method from both serum and urine appeared significantly less contaminated with heparin when compared to RNA isolated with the RN method (p < 0.0001). This was determined for both for the spike-in added during the work-up procedure (Fig. 4B and C) as well as for the spike-in added during the cDNA synthesis procedure (Fig. 4E and F). This indicated the presence of components in body fluids that prevent co-isolation of heparin with the NG method.
Evaluation of co-purification of PCR-inhibiting compounds other than heparin
RT-qPCR analyses are prone to interference by compounds that co-purify with RNA [37]. Besides heparin, other interfering compounds can also be co-purified which cannot be counteracted by heparinase I treatment. As already shown, a substantial variation was observed in the detection levels of cel-miR-39, hinting at the presence of non-heparinous compounds. Therefore the RNA samples used previously (obtained from serum, urine, perfusate and bile of six patients), were spiked with cel-miR-54, and, in the presence of heparinase I, cDNA was synthesized, followed by RT-qPCR analysis. Detection levels of cel-miR-54 in serum were not significantly different between methods RN, QP, NG, and CU, with values ranging between 1157-1444, 1207-1465, 1122-1473 and 802-1441,respectively (Figure 5A). Results for urine were comparable, with respective values ranging from 1140-1457, 1310-1468, 1309-1637 and 1012-1591 for methods RN, QP, NG, and CU (Figure 5B). Results for perfusate were also not different between the four different methods with values ranging between 1408-2141, 1406-2181, 1471-2082 and 1049-2167 for RN, QP, NG and CU, respectively (Figure 5C). As the reference values of cel-miR-54 - indicated by the red line in Figure 5 - were 1358, 1571 and 2101 for serum, urine and perfusate, respectively, the co-purification of other PCR-inhibiting compounds from these biofluids seemed almost absent.
Contrary to serum, urine and perfusate, levels of spike-in cel-miR-54 in RNA samples from bile showed a larger extent of variation. Detection levels range from 1373 to 1917, 1288–1958, and 1515–2165 for the RN, QP, and NG method, but 32–1884 for the CU method (Fig. 5D). This increased range for cel-miR-54 levels in RNA from bile could be attributed to one sample that showed a distinct green discoloration. To follow up on this observation, the number of bile samples analyzed was increased to 10, and cel-miR-39 and cel-miR-54 levels were again determined. Four out of 10 RNA samples showed this discoloration after isolation (Fig. 6A), and this phenomenon was associated with lower detection levels of both cel-miR-39 and cel-miR-54, which was most likely caused by co-isolation of biliverdin when the CU method was used (Fig. 6B).