Human Urinary mRNA as a Biomarker of Cardiovascular Disease

Supplemental Digital Content is available in the text.


Isolation of RNA from urine supernatant
Frozen urine samples from the low-sodium and sodium-loaded conditions (approximately 4-5 mL) were retrieved from -80°C storage and thawed at room temperature. Once thawed, samples were vortexed vigorously for 30 seconds in order to break up uromodulin protein aggregates and resuspend extracellular vesicles into solution. Samples were then prepared for RNA isolation using an Exosome RNA Isolation Kit (Norgen Biotek, Thorold, Ontario). Per manufacturer's recommendations, samples were centrifuged at 1000 rpm (approximately 200 xg) for 10 minutes to avoid shearing cells. The supernatant was transferred to a fresh tube, and then centrifuged at 2500 rpm (approximately 1200 xg) for 10 minutes in order to pellet cells. The manufacturer's recommended protocol was then followed, with one modification: all centrifugation steps were performed at 4°C. Once isolation was complete, each 50 μL RNA sample was immediately stored at -80°C until used in qPCR experiments.

qPCR Assay Design
To achieve high sensitivity and specificity, we designed locked nucleic acid probe-based qPCR assays. cDNA sequences for MR target genes and control genes were retrieved from the Ensembl genome browser (http://www.ensembl.org/index.html). Each target gene sequence was archived in Seqbuilder (Lasergene, Madison WI) to facilitate annotation of primer and probe binding sites.
We selected the locked nucleic acid probes from a universal probe library according to the target-gene cDNA sequences (http://qpcr.probefinder.com/organism.jsp; Roche, Basel, Switzerland). Once appropriate probes had been selected, the 200 bp region surrounding each probe was entered into Primer3 software, and appropriate primer pairs were designed to span an intron. We confirmed that the primers did not amplify DNA at 40 cycles of PCR (not shown).

S3
Due to the presence of RNases in the urine, fragmentation of RNA was a concern.
Bioanalyzer (Agilent Technologies) analysis revealed urinary RNA to comprise small fragments less than 200 bp. Therefore, we employed a strategy of probing multiple sites along the relevant gene sequences to increase the likelihood of detecting any truncated RNA molecules. We designed up to three separate primer sets --when possible in view of all specificity-driven assay design constraints --for each gene of interest, positioned to amplify distinct regions in these genes. We designed three sets of primer-probe assays for: NR3C2, SCNN1A, SCNN1B, SCNN1G, SGK1, UMOD. Genes for which we were able to design one or two primer-probe assays within the design constraints included: AQP1, AQP2, GAPDH, HSD11B2, TSC22D3.
To distinguish primer sets within the same target gene of interest, we used the following naming convention: gene name, followed by "-" and a number. For example, the three assays within the SCNN1A gene are named SCNN1A-1, SCNN1A-2, and SCNN1A-3.

Reverse Transcription and Pre-amplification of RNA
RNA samples were retrieved from -80°C storage and thawed on ice. While samples were thawing, reverse transcription master mix was prepared using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems); components were combined per manufacturer's instructions. 10 μL aliquots of master mix were then combined with 10 μL of RNA template for a final reaction volume of 20 μL. The reactions were run on a thermal cycler following the RT protocol outlined in Supplemental Figure 1.
The low nucleic acid content of urine prompted us to evaluate whether target-specific PCR amplification prior to qPCR was required in order to detect molecules by qPCR. In our preliminary experiments, pre-amplification of cDNA was found to improve reliability of target sequence detection. Therefore, for the study samples, pre-amplification reactions were set up S4 once RT was complete. Target-specific primer sequences were generated for each gene of interest (Integrated DNA Technologies, Coralville, IA), and 25 μL PCR reactions set up using the protocol outlined in the iTaq DNA Polymerase kit (BioRad); 6.25 μL of cDNA template was added to each reaction. Fifteen PCR cycles were performed following protocol outlined in Supplemental Figure 1.

Detection of Biomarker Candidates Using Fluorescence Reporter Probe qPCR
qPCR reactions were prepared using the same target-specific primer sequences used in the preamplification reactions. TaqMan 2x Universal PCR Master Mix was combined with primers and Universal Probe Library probes (Roche), and 15 μL aliquots were loaded onto 384 well qPCR plates. 5 μL of pre-amplified cDNA template was then added to each well. PCR reactions were carried out on a 7900HT thermal cycler (Applied Biosystems). Human kidney RNA (Clontech) and no-template controls were included on each plate as positive and negative controls, respectively. We assayed commercial human kidney RNA (Clontech) within each plate. The technical replicates were averaged within each plate and these results were used to adjust Ct values on each plate to account for any interplate differences.

Examination of using Urinary Creatinine for mRNA Normalization
The optimal normalization strategy for urine supernatant mRNA assays is currently an open question. Historically, normalization to urinary creatinine has been performed when measuring a wide variety of urinary analytes. Normalization to urinary creatinine attempts to measure analytes independent of any effect of varying urinary concentration. Thus, we considered the possibility that the Ct value of urinary supernatant mRNA molecules would vary according to the urinary creatinine. Urinary creatinine was similar during low-sodium diet and after sodium loading (Supplemental Figure 2). Moreover, no clear or consistent relationship was observed S5 between urinary creatinine and Ct value for the RNA molecules we assayed (Supplemental Figure 3). Moreover, relative quantitation has been found to be unnecessary even in a crosssectional study of urinary mRNA. 4 For these reasons, Ct values are presented without normalization to another molecule.

Statistics
For the analysis of the clinical study, data are presented as means ± standard deviation. Betweengroup comparisons were made using the paired t test. Technical replicates' correlation with each other was analyzed using the Pearson correlation coefficient. Ct values from different qPCR assays within the same gene were evaluated using Pearson correlation coefficient, as was the relationship between Ct values and aldosterone or urinary sodium-creatinine ratio. A subset of assays probed participants' urine for the same gene target under the same dietary condition, but on a different qPCR plate (as an additional control)These repeated assays served as an additional internal control. To avoid counting these observations twice, all other analyses include only one plate's results for assays that were repeated on two plates for the same participant under the same dietary condition (as an internal control). The results retained were from the plate that either produced the most detected Ct values, or in the case of a tie, results from the plate prepared first were selected. We did not impute or otherwise assign a value when no mRNA was detected. A two-sided P<0.05 was considered significant, except where a more stringent threshold (arising from Bonferroni correction) is specified. Data were analyzed in R (R Foundation for Statistical Computing, Vienna, Austria).

Definition of Successful qPCR Result
We evaluated whether our technical replicates were tightly grouped, as expected if these assays performed well. We observed that when 2 or 3 of the technical triplicates for a sample yielded a S6 Ct value, the replicates were similar. For example, the median span of Ct values measured within sets of technical replicates was 0.14 cycles, with 75% of replicate sets spanning 0.23 cycles or less. Pearson correlation coefficients between replicates 1 & 2 and between replicates 2 & 3 were calculated for each assay. The correlation coefficients were greater than or equal to 0.990 in 86.7% of the 75 comparisons in this matrix (Supplemental Figure 7).

Distribution of C t Values
Focusing on the 17 individuals from whom we could assay urinary aliquots by qPCR for both high-and low-sodium diet conditions, we examined the distribution of all Ct values passing quality control. In this analysis, we included target genes (MR target genes) & control genes (not expected to respond to MR activation; Supplemental Table 2) during low-sodium diet and after sodium loading. After calculating the mean of the technical replicates, the 658 Ct values passing our quality control filter tended to segregate into two bins: 1) between 20-30 cycles or 2) undetected, with few results falling below 20 or above 30 cycles. 95% percent of detected Ct values fell between 19.4 and 30.9, within a total range of 15.9 -37.7 cycles. Although some correlations between age and specific gene products' Ct values were marginally statistically significant, we did not find compelling evidence of a relationship between age or sex and Ct values for the gene products we assayed.

Comparison of C t Values Detected By Different Primers Within the Same Gene
When feasible according to our stringent primer-probe design principles, 3 assays were designed with a gene. We compared the results of different qPCR assays within each gene using Pearson correlation coefficient. As expected, we found high pairwise correlations between results of the different assays within a gene (Supplemental Figure 8).