High stability of EV content during storage and handling at room temperature

The major source of variability in the biomarker discovery and validation are pre-analytical decisions regarding sample harvesting, storage and processing. The current consensus protocols for urine EVs favor storage at − 80 °C although freezing and thawing can cause some EV loss or artefacts, such as EV aggregation or selective loss of larger EVs (MVs)². In this study we have tested the use of a commercially available preservative that claims the conservation of urine RNA/microRNA/DNA and proteins for more than 2 years at room temperature (RT). We have focused on small nanosized vesicles (nsEVs) that are reported to be remarkably resistant to freeze–thaw cycles^19,20. These are likely enriched with exosomes, while possibly containing a fraction of vesicles that do not originate from multivesicular bodies. The stability assessment was done on samples from 3 healthy donors, with 2 sample aliquots independently processed from each donor, overall featuring 6 biological replicates.

To compare the stability of exosome-like particles concentration and overall protein content we have measured particle number via NTA analysis in pre-cleared urine, and in reconstituted MV and exosome pellets from UC and CP samples. We have also processed the same samples with immunobeads based IP protocol (as explained in “Study design” in “Material and methods” section) but beads-bound EVs don’t allow for analytical methods such as NTA or protein content measurement. That is why in experiments represented on Figs. 1 and 2 only UC and CP were compared while all three methods were used for EV miRNA extraction and quantification (Fig. 3). As expected from our prior experience, UC gave major EV yields. We could also appreciate significant detection of exosome-sized particles (mean size around 100 nm) in reconstituted MV fractions confirming the prior observed co-pelleting of also nanosized EVs with larger vesicles at lower centrifugation speeds. In samples analyzed after 6 months, the overall EV concentration and size do not appear significantly affected by the storage although the number of retrieved vesicles is a bit lower upon exposure to room temperature (RT) (Fig. 1A). Conversely, in one-month-old samples the EV concentration is better maintained at RT with respect to frozen samples (Fig. 1B).

Quantitative analysis of EVs protein content. EVs were purified from urine obtained from three healthy volunteers, and 2 independently processed aliquots from each donor. The urines were either frozen at – 80 °C (F) or adjuvated with the preservant and stored at room temperature (P), After 1 and 6-months long sample storage, EV isolation was done by differential ultracentrifugation (UC) and chemical precipitation (CP). The Figure shows the total protein content measured by BCA assay, and creatinine content determined in whole urine and in urine microvesicles (MV) pellet and EV isolates from a single representative donor (A). For the same donor samples the specific content of canonical EV proteins was addressed by Western Blot (C, namely CD63, CD9, Tsg101 and Alix) (C) CD63 and CD9 expression was also measured for all analyzed samples by an ELISA assay (BEVs prepared from LNCaP cell conditioned medium were used as standards for relative comparisons in ELISA assay. A non-parametric test for multi-group comparison (three replicates each) 1-WAY ANOVA followed by Tukey post-hoc test was used for analysis of ELISA data; All p-values < 0.05 are marked with *, < 0.01 with **, < 0.001 with ***.

Quantification of EV miRNA content. Total RNA is extracted from EVs purified from healthy donor-obtained urine samples (N = 3), stored either preserved at room temperature (P) or frozen (F), using ultracentrifugation based protocol (UC), chemical precipitation (CP) or immuno-precipitation with anti-CD9 coated beads (IP). RT-qPCR amplification is performed of miRNAs de-scribed to be contained in EVs, namely miR-16, miR-21, miR-210 and miR.451. We show the average relative miRNA expression from three biological replicates. Relative expression was calculated with respect to a reference sample with known and constant RNA input (EV RNA from LnCAP cell culture). Samples from three healthy donors, and 2 independently processed aliquots from each donor, are included in the assessment. 1-way ANOVA and Bonferroni post-hoc test were used to determine statistical significance of observed differences * indicates p-value < 0.05, ** p-value < 0.01 and *** p-value < 0.001.

NTA measurements were quite in line with the determination of overall protein content (Fig. 2A). After 6 months UC pellet had higher values in frozen samples (2.8-fold increase) with the trend that was the same but more pronounced with respect to the same samples measured in NTA (1.3-fold). The overall urine protein content did not seem affected by different storage conditions while higher creatinine levels were measured in frozen with respect to RT samples (1.6-fold). After short term storage instead, protein content of exosome UC pellet was significantly higher in samples stored at RT (1.5-fold), while urine creatinine levels had an opposite drift and dropped 1.8-fold respect to frozen samples. The creatinine variations did not result significant, but the opposite trend observed for creatinine with respect to EV protein content raised questions about the suitability of creatinine as a calibrator of EV protein content in urine. This issue was therefore further addressed in the second part of this study. Observed data on EV proteins preservation overall supports the use of RT storage of urine samples especially for 1 month’s storage.

We have further assessed the amount of acknowledged EV and exosome markers by common quantitative assays such as ELISA and Western Blot. Sandwich ELISA assay measuring the level of small vesicles (co)expressing CD9 and CD63 showed correlation with overall protein content and vesicle number measurement (Fig. 2B and Supplementary Fig. 1). CD9 and CD63 are exosome enriched tetraspanins known to be well expressed on urinary vesicles^4,6,10,21. The signals were higher (over three-fold) in samples conserved as frozen for 6 months, while, conversely, more prominent (~ two-fold) in those samples stored for 1 month at RT. Western blot analysis of UC pellet from the same samples showed concordant data with respect to the content of bona fide exosome proteins Alix and Tsg101 (Fig. 2C). The opposite trend for tetraspanins CD9 and CD63 revealed by WB in long term (6 months) conserved samples raises the possibility of contamination of pellets with membrane fragments and would be in line with compromised integrity of vesicles and surface displayed proteins upon long term storage at RT.

The EV RNA content is one of the fundamental indicators of vesicle integrity and functionality, being RNAs both putative biomarkers and effectors in pleiotropic functions exerted by vesicles in health and disease^3,7,9. We have assessed the stability of overall RNA content of vesicles pelleted from differentially stored urine samples, with a particular focus on a most abundant RNA category found in EVs, small ncRNAs. We have attempted to use Qubit 2.0 Fluorimeter (Qubit Small RNA kit) in addition to Bio-analyzer concentration estimates in order to quantify the levels of total RNA in urine EV samples. Bioanalyzer RNA profiles were in line with > 80% abundance of small RNAs (miRNA) in all samples, but RNA content resulted too low to enable accurate Qubit 2.0 measurements, with some samples containing < 250 pg/μl. The latter was expected as our samples were obtained from urine volumes as low as 1 ml¹¹. Therefore, we could not use a fixed RNA amount, but have decided to use an equal original urine volume for input normalization n amplification reactions. Such decision is in line with the common practice in diagnostic sampling²². MiRNAs are among the exosome cargo molecules that have elicited substantial interest and have been early approached for potential clinical interest. Different miRNAs have been described as associated to urinary vesicles in healthy or diseased subjects^{6,12,23,24,25}. For the purpose of stability testing in this study we have first picked up several miRNAs commonly expressed in EVs, also in urine^23,25. Within 1 month of storage, notoriously abundant miRNAs such as miR-21, 16 and 210 are successfully amplified (with Ct values ranging between 23 and 33) from exosome-sized vesicles purified from 1 ml of urine and showed high stability in samples preserved at RT (Fig. 3). The content of analyzed miRNAs was quantified as relative expression to that of a reference sample with known and constant RNA input (EV RNA from LnCAP cell culture)^13,14. As expected, miR-451 resulted low abundant but was still revealed in all the samples tested. Increased stability of EV RNAs upon storage at RT was confirmed also after 6 months, in particular in UC samples. All three employed methods gave the material that can be reliably amplified and quantified, with the lowest miRNA expression obtained from IP samples. This result has been anticipated by an acknowledged fact that immunoaffinity selects for EV subpopulations, trading off the yield for specificity and purity. Interesting exception is observed for miRNA 210 that resulted highly enriched in IP samples (Fig. 3).

Certain incoherency in yield and stability between different RNA species was observed across different isolation protocols used. miR-21- and 16 levels were more in line with independently measured vesicle parameters (vesicle counting by NTA or protein quantification). Possible selective loss of some species and some vesicle subsets may be dependent on isolation protocol, as prior re-ported^3,15. Noteworthy the advantage of RT preservation over freezing of urine was in particular evident for recovery of mRNAs after 6 months of storage (Fig. 4) as well as recovery of MV RNA content. In this study, we have assessed two RNAs widely reported in biofluid EVs (beta-actin) and in urinary EVs (PSA). The samples used for these experiments came from male donors, and PSA mRNA features as indicator of prostate derived vesicles. Overall, we can observe that preservation of urine at RT for sub-sequent EV RNA analysis is a good option after both long- and short-term storage. While long term storage at RT is not viable strategy for protein analysis, pursuing it in a short time setting (within a month) could be excellent solution favoring the stability, easy handling and shipment of urine samples for comprehensive EV analysis. Side indication from the first set of data that warrants more in-depth study is that both storage and isolation method could introduce a bias for loss or recovery of distinct EV populations in urine.

Quantification of EV RNA and MV RNA/miRNA content. Total RNA is extracted from MV pellets and from EVs purified from healthy donor-obtained urine samples (N = 3), stored either pre-served at room temperature (P) or frozen (F), using ultracentrifugation-based protocol (UC), chemical precipitation (CP) or immunoprecipitation with anti-CD9 coated beads (IP). RT-qPCR amplification of mRNAs prior described to be contained in urine EVs, namely β-actin and RNY RNA was done as de-scribed in Material and Methods. Both these mRNAs and miRNAs (miR-16, miR-21, miR-210 and miR.451) were analyzed also in MVs. Samples from three healthy donors, and 2 independently processed aliquots from each donor, are included in the assessment. One-way ANOVA and Bonferroni post-hoc test were used to determine statistical significance of observed differences * indicates p-value < 0.05, ** p-value < 0.01 and *** p-value < 0.001.

Normalization strategies for quantitative analysis of urinary EVs

In the second part of the study we have addressed several normalization options that would enable the quantification and interpretation of differences in exosome/EV proteins and RNAs abundances in the context of intrinsic variability of concentration of urine, and urine EV. In order to make a best choice that fits our scope of proposing the method feasible both for research and clinical practice, we have considered the following premises. Though for quantitative comparisons a 24-h urine collection might be desirable, collecting first morning urine is definitely a more practical and conventional procedure that guarantees higher patient compliance. Moreover, studies on proteinuria have shown that this method of collecting random urine samples is valid and that the rate of protein excretion correlates well with that found in 24-h urine¹⁶. Other frequent pre-analytical choice in common diagnostics is that the samples are volume-normalized. In case of blood testing, due to a strict homeostatic control, testing and comparing the same volumes of samples is a standard option. In urine, this approach leads to inconsistencies in quantification of EV markers. We have first assessed exosome concentration and content of acknowledged vesicular markers in longitudinal samples with the aim to identify the reliable and measurable EV parameters that can be used for normalization. To this purpose we have used a small number of proband samples from 2 healthy donors, one male (Caucasian, 43 years) and one female (Caucasian, 45 years), from which the urine samples have been collected in 3 consecutive days. Each sample was divided into two aliquots that were independently processed, increasing thus the number of biological replicates and decreasing the impact of a random operational error to a small sample size.

We have first measured exosome-sized particles concentration as well as protein content in both neat urine samples and in EV isolates prepared by ultracentrifugation. This isolation method was selected because it is still the major method used in urinary EV studies and has provided the best yield among the purification methods featured in this study. In addition, UC should not conceptually introduce a bias versus specific EV subpopulations, except for favoring the recovery of small EVs, while obtained EV isolates can be analyzed by different methods, including NTA, ELISA and Western Blot. NTA measurements revealed significant fluctuations of overall EV number within and between two healthy donors (Fig. 5A). Surprisingly, there was no correlation between EV counts in whole urine and in purified UC pellets while fine agreement was observed between protein content of whole urine and that of extracted exosome fractions (UC) (Fig. 5A, B and Supplementary Fig. 1). Analysis of canonical EV markers by Western Blot confirms day-to-day variations in EV protein content measured in the constant urine volume (Fig. 5C). Expression of bona fide exosomal markers Tsg101 and Alix apparently poorly correlated with a total protein content or vesicle counts and had different pattern from tetraspanins CD9 and CD63. To better assess individual protein patterns, we have performed densitometric analysis of gel bands (ImageJ, Fig. 5C) that highlighted CD9 signal as best reflecting other independent quantitative parameters of EV pellets (number and overall protein content of vesicles). Noteworthy CD9, according to densitometry, has excellent correlation with Tsg101 pattern but not with Alix pattern. Although CD9 is not exclusively confined to exosomes but is abundantly expressed also in MV (here not shown) it is known to be highly enriched in exosome-sized vesicles with respect to both parent cells and other vesicles subfamilies. The appeal of CD9 resides in the fact that it is displayed on the vesicle surface and thus can be exploited for isolation strategies and, of note for the focus of this study, for rapid vesicle quantification. As shown in Supplementary Fig. 1, the correlation between the EV protein content and the particle concentration measured in isolated small EVs is fair but not perfect (R = 0.781), while the EV protein content correlates better with the particle counts in the original neat urine (R = 0.817). In a similar way, also CD9 content was better aligned with the particle number measured in original neat urine (R = 0.860) than in EV pellets (R = 0.607), while it nicely correlated with protein content of isolated EVs (r = 0.931). Overall, among all independent indicators of EV content in urine, we found the best matching pattern between protein content of isolated EVs pellet, including CD9 measurement, and overall exosome-like (small EVs) particles counts in the original neat urine, Therefore, we retain these two to be most appealing EV-related parameters for mirroring their overall content.

Comparative analysis of EVs content in longitudinal urine samples. EVs (UC) and MVs were purified by differential (ultra)centrifugation-based protocol from urine obtained from two healthy volunteers, one female (marked as 1) and one male (marked as 2), 45 and 43 years old, in 3 subsequent days. For each sample two aliquots of urine were processed independently and results presented as averages + SD. Total protein content was measured in a whole urine, MVs and EVs by BCA assay (A) particle concentration determined by NTA measurement (B) and EV proteins assessed by Western Blot (C). The signal intensity on the blot was further analyzed and quantified by densitometric analysis using ImageJ software and plotted on the right.

Observed day-to-day variations in EV content are likely caused by diuretic effects. To equalize them, the samples can be normalized either i) using a volume correction factor based on a measurement of a parameter that reflects accurately EV-, or at least, urine concentration, without itself being additionally biased by any physiological of pathological variable, or by ii) using an optimal internal normalizer for each category of markers that are of our interest. Measurement of above identified EV related parameters can be used in both strategies. They can be used as preanalytical volume correction method prior to subsequent analysis of either proteins or RNAs in the EV preparations. They can also be used in second normalization option as the internal “housekeeping” EV markers. For objective protein biomarker measurement normalization to CD9 or Tsg101 level is a viable option, the choice of which would be based on methodological requirements (for instance an ELISA versus Mass Spectrometry, respectively). We will further discuss their use for normalizing the RNA expression data in the following chapter (Fig. 9).

To explore the first option, besides EV-intrinsic factors, we have also considered some common biochemical parameters measured directly in urine as part of a standard laboratory analysis (Supplementary Fig. 2 and Fig. 6). We have included creatinine as it is frequently used for adjustment of urine analytes, including EVs, although its accuracy for the purpose has been long debated. To identify potentially useful biochemical indicator(s), we have first assessed the degree of association between each biochemical parameter measured across the longitudinal samples and the relative measurements of a particle number and a protein content of extracted urine EVs (UC) (Fig. 6).

Analysis of urine biochemical parameters in longitudinal urine samples. Urine samples are obtained from urine obtained from two healthy volunteers, one female (marked as 1) and one male (marked as 2), 45 and 43 years old, in 3 subsequent days. These longitudinal urine samples were submitted to a routine biochemical urine analysis in a San Raffaele Hospital clinical laboratory within two hours from collection. Panels on the right show the correlation analysis performed by plotting the values of selected biochemical parameters against the EV parameters independently measured in the same urine samples. Explanation of acronyms used: CRJ2U, creatinine; ALBU2, albumin; GLUC3, glucose; TPU, total protein in urine; LH2, light immunoglobulin chains.

Of all biochemical parameters measured in original urine samples the one that best correlated with the small EV content was the total protein content in urine (TPU), and the urine albumin level. TPU and Albumin correlated well with total particle counts in urine (R = 0.875/0.961), with urine exosome protein content (R = 0.980/0.907) and with CD9 content (R = 0.937/0.922) compared to creatinine correlation that gave coefficients of 0.670, 0.843 and 0.869 respectively. Other biochemical parameters which correlation to EV content is worth considering is urine glucose concentration and the heavy immunoglobulin chain levels (Supplementary Fig. 1).

Observed correlations opened to an interested possibility to use such routinely measured biochemical urine analytes to compensate for diuretic dilution effects and correct/normalize overall urine EV content (Fig. 7). We have considered the most promising those correction parameters that diminished most the variation between compared longitudinal urine samples. Coefficient of variation (CV) was calculated by dividing relative standard deviation and average of all samples. For instance, the correction with urinary TPU levels indeed reduced the variation coefficient between samples in terms of both urine particle numbers and urine exosomes protein content. Correction with urinary creatinine instead showed poor compensation effect.

Normalization of EV content to urine biochemical parameters. Parameters retained most indicative pf urinary EV content in longitudinal urine samples, namely particle concentration in the whole urine, protein content and the CD9 signal measured in EV isolates, was normalized with urine biochemical parameters such as total urine protein (TPU) and creatinine (CRU). The scope of this normalization is to decrease the fluctuation of EV concentration, indicated here by the decrease of coefficient of variation (CV). Urine samples are obtained from two healthy volunteers, one female (marked as 1) and one male (marked as 2), 45 and 43 years old, in 3 subsequent days. The sample points are indicated in x axis.

Internal normalizer-based strategies for urine EV RNAs

According to conventional approaches to elaboration and normalization of gene expression data, analysis of exosomal RNA content may benefit from some acknowledged internal normalizers comprising the RNA species that are found stably expressed across different human tissues and sample types and are reported to be sorted also into vesicles, though are not vesicle specific. The examples of latter include β-actin or GAPDH mRNAs, or some miRNA species including miR-16, miR-191 or miR-24^3,15,25,26. In the past, small nucleolar RNAs (i.e. SNOU6) were often proposed for miRNA expression analysis, while the emerging policy is to normalize qPCR data using the same class of biomolecules.^27,28 In order to cope with the complexity of the biofluids, it would be extremely beneficial to identify and use the RNA species that are enriched in vesicles of interest, in this case exosome-like vesicles. Moreover, depending on the purpose of the study, using RNAs that can normalize for organ specific vesicles could be considered. We have assessed the expression of 3 different mRNAs in the set of longitudinal samples, chosen as to represent different RNA categories relevant for possible normalization strategies (Fig. 7). Of note, mRNAs were present both in EV (UC pellets) and MV fractions of all samples (Fig. 7), differently from a protein content, that was too low and thus not detectable in MV pellets (not shown). Beta-actin was selected as a commonly used gene expression normalizer. PSA transcript was suggested as a potential prostate-specific mRNA that can have both the value of a normalizer or as a disease biomarker^29,30. PSA mRNA has often been reported and assessed in urinary EVs. Finally, we have employed RNY4, a member of HY RNA family that have been detected as abundant RNA species in biofluids and in EVs^29,31,32.

All these mRNAs were detected upon RNA extraction from EVs, in a material corresponding to 1 ml of starting urine sample, with roughly similar trends observed across the experimental points (Fig. 8). Beta-actin, and, especially PSA mRNA, resulted not-abundantly expressed in analyzed samples. Poor expression of PSA mRNA would be consistent with its specific prostate origin and with the fact that we did not use any isolation method enabling specific enrichment of prostate derived vesicles. Surprisingly however, EV expression of PSA mRNA did not correlate with the gender of the donor. Higher signal for PSA mRNA was obtained in a female donor sample (1.1–1.3) raising thus doubts on the utility of this mRNA as a truly prostate specific marker. Instead, RNY4 was well amplified from all UC pellets. We have previously confirmed that this small RNA is multifold enriched in exosome-sized vesicles with respect to parent cells (~ ΔCt = 10). These two observations constitute premises to consider it a promising internal reference for urinary EV RNA analysis.

Detection of urine EV-mRNAs by RT-qPCR in longitudinal samples. Following ultracentrifugation, total RNA has been extracted from EV (UC) and MV pellets and RT-qPCR amplification of mRNAs of interest performed as described. Purification is performed on urine obtained from two healthy volunteers, one female (marked as 1) and one male (marked as 2), 45 and 43 years old, in 3 subsequent days. For each sample two aliquots of urine were processed independently and results presented as averages + SD. The results are shown both as Ct values (on the left) and as relative RNA expression (on the right) calculated as delta Ct versus reference sample (RNA extracted from Prostate Cancer LnCAP cell line exosomes).

At glance, absolute mRNA expression pattern (row Ct values measured) did not mirror the EV number and protein concentration measured in the same EV pellets (comparing Fig. 5 and Fig. 8). On the other hand, transforming Ct values into quantities (using comparative Ct method) (Fig. 8C) or converting them into a linear scale (assuming amplification efficiency of 100% for each mRNA, (2^(40−Ct))) reveals the pattern that correlates perfectly with overall small EV content in the urine (as observed in Fig. 5). RNY4 was most abundantly expressed in all samples and correlated with overall particle counts in urine, and with protein and CD9 content in urinary EVs with respective coefficients of correlation of R = 0.987, R = 0.0777 and R = 0.837 (Fig. 8). Noteworthy, relative RNY4 expression correlated well with urine total protein, in particular albumin content (R = 0.900), rather than with creatinine levels (R = 0.585) (Fig. 9). When we attempted to correct the RNY4 expression inter-sample variations by normalizing it with creatinine or urine total proteins³³, we did not observe the decrease in inter-sample CV, while the best correction is obtained by considering the respective CD9 levels in the sample. The latter decreased overall inter-sample variation to less than 10% (Fig. 9). This supports the use of CD9 as a “housekeeping” EV marker to correct for urine dilution-related variations in EV-RNA content.

Normalization of EV-RNA content variations to urine biochemical parameters and EV number and protein content. RV RNY4 RNA expression levels across the longitudinal urine samples were first plotted against selected urine biochemical parameters, namely urinary creatinine, albumin and total protein content, as well as against independent EV content indicators, such as particle number measured in the whole urine and the CD9 signal measured in the respective EV isolates (on the left). The same parameters were subsequently used in the attempt to normalize/compensate for variations in EV RNA signals across the samples, revealed as decrease in coefficient of variation (CV, on the right). EVs (UC) were purified by differential (ultra)centrifugation-based protocol from urine obtained from two healthy volunteers, one female (marked as 1) and one male (marked as 2), 45 and 43 years old, in 3 subsequent days. The experimental points are indicated at the x axis of the right panel. The indicators shown on the relative y axis are matched between left and right panel.

In the last part of our study, we tried to challenge our approach for identification of best EV normalizers by applying it to a commercial custom miRNA array. We have chosen the array featuring 24 miRNAs, including both putative normalizers for diverse tissues and biofluids, as well as putative markers for diverse diseases, including cancer, that we have already deployed for analysis of EV-miRNA expression in cancer plasma samples (not published). We have assessed this panel miRNA expression pattern in our small longitudinal samples set and we have identified two clusters (Fig. 10). The first one (cluster 1) comprises miRNAs that are notoriously and abundantly expressed across different human samples and tissues (i.e. miR-16, -21, -191, -24). By evaluating Ct values for the same urine volume input, these seem to be uniformly expressed in all samples regardless the prior observed variations in EV concentrations. In a control experiment with LnCAP culture-derived EVs, we noticed that these miRNAs were not enriched in EVs with respect to a respective parent cell lysate (Fig. 10B). Conversely, the levels of a second group of miRNAs encompassing miR-223, -150, -145. -636 and RNU6 (cluster 2) changed across the longitudinal samples and appear enriched in control standard EVs with respect to respective cells. Relative miRNA expression levels correlated better with the inter-sample fluctuations observed in EV number and protein content, in particular for miRNAs belonging to the second cluster (Fig. 10A and C). Best correlation of EV-miRNA expression was noticed with respect to overall particle counts in original urine (R = 0.967–0.986 for mRNAs encompassing miR-150, -636, -145 and -223, compared to R = 0.368–0.795 for miRNAs such as miR-191, -21. -21, -16) and with respect to CD9 levels in extracted urinary EVs (R = 0.821–0.904 for first cluster and R = 0.644–0.920 for the second one) (Fig. 10C). Accordingly, the relative expression of the miR-150, -636, -145 and -223 miRNA cluster (cluster 2) was well associated to urine albumin level (R = 0.886–0.943) but not to urine creatinine (R = 0.533–0.775), unlike the cluster comprising miR-191, -21. -21, -16 expression (cluster 1) that showed only week correlation to urine biochemical parameters (R = 0.381–0.775 for albumin and R = 0.568–0.746 for creatinine) (Fig. 10C).

Detection urine of EV-miRNAs by RT-qPCR in longitudinal urine samples. Following ultracentrifugation, total RNA has been extracted from EV and MV pellets and RT-qPCR amplification and quantification of a 24 miRNA Taqman Custom Array performed as described. The results are expressed both as Ct values (B) and as relative RNA expression (Table A, and panel B on the right) calculated as delta Ct versus reference sample (RNA extracted from Prostate Cancer LnCAP cell line exosomes). The correlation index was determined between the trend observed for miRNA expression levels across the samples and the respective values for selected EV parameters (particle concentration and protein content) as well as biochemical parameters measured in the same urine samples (creatinine, albumin and total protein content) (C). Finally, the normalization of miRNA expression levels at longitudinal points is attempted using best correlating EV and urine parameters (D). EVs (UC) were purified by differential (ultra)centrifugation-based protocol from urine obtained from two healthy volunteers, one female (marked as 1) and one male (marked as 2), 45 and 43 years old, in 3 subsequent days. Experimental points are shown on relative x axes. For each sample two aliquots of urine were processed independently and averaged.

We ultimately attempted to compensate the inter-sample variations of miRNA expression in EV preparations by normalizing the results by EV parameters, with the best results obtained when the particle count measured in the neat urine and the CD9 content in urine were used (Fig. 10D). Such a result was coherent with our prior described results.

Overall, such an approach would point out the cluster 2 as an appealing miRNA panel for further validation as urine EV miRNA content calibrator in the extended sample cohort.

We wanted to compare our approach and results to a widely used and acknowledged method for identification of reliable internal controls for normalization of miRNA/gene expression data. This method consists in the use of statistical algorithms that rank the reference genes by calculating their overall expression stability, in associations with other possible control genes. We indeed applied three of the most used and often combined software solutions, namely GeNorm, NormFinder and BestKeeper (supplementary Fig. 3A–C) to the input data comprising either relative expression (quantities) or absolute expression (Ct values) data collected from our pilot sample set, according to each algorithm requirements³⁴. The key assumption of all these algorithms is that the investigated candidate reference genes do not vary in their expression systematically across the samples being considered. Even if this is true in terms of impact of different physiological or pathological conditions, in the scenario of identification and implementation of urinary vesicles biomarkers, this assumption is violated in practice by day-to-day fluctuations in diuretic dilution of the urine. If we pursue the scenario applied in this study, featuring the use of the same input volumes of the samples, without normalizing the input of cDNA for RT-PCR, then the use of computational algorithms is innately flawed. The measurement of RNA/cDNA extracted from EVs isolated from low original urine volumes (1 ml) is difficult and imprecise even with the use of Qubit assays, and is time consuming and costly for routine or high throughput analysis. Indeed, all three algorithms employed to analyze miRNA expression data in a pilot set of longitudinal samples (6 experimental points, 12 independently processed samples), would indicate as most stable genes miR-24 and miR-16 that are apparently stable across the all samples without being impacted by the real differences in EV concentration. This approach has clear limitation for selection of reliable internal references in this case, where the lack of statistical effect does not account for the sample specific variation and is thus missing the specific problem addressed by this study.