Advertisement
Open AccessJournal of UrologyAdult Urology1 Sep 2020

Expression of Small Noncoding RNAs in Urinary Exosomes Classifies Prostate Cancer into Indolent and Aggressive Disease

View All Author Information

Abstract

Purpose:

This is the first report of the development and performance of a platform that interrogates small noncoding RNAs (sncRNA) isolated from urinary exosomes. The Sentinel™ PCa Test classifies patients with prostate cancer from subjects with no evidence of prostate cancer, the miR Sentinel CS Test stratifies patients with prostate cancer between those with low risk prostate cancer (Grade Group 1) from those with intermediate and high risk disease (Grade Group 2-5), and the miR Sentinel HG Test stratifies patients with prostate cancer between those with low and favorable intermediate risk prostate cancer (Grade Group 1 or 2) and those with high risk (Grade Group 3-5) disease.

Materials and Methods:

sncRNAs were extracted from urinary exosomes of 235 participants and interrogated on miR 4.0 microarrays. Using proprietary selection and classification algorithms, informative sncRNAs were selected to customize an interrogation OpenArray™ platform that forms the basis of the tests. The tests were validated using a case-control sample of 1,436 subjects.

Results:

The performance of the miR Sentinel PCa Test demonstrated a sensitivity of 94% and specificity of 92%. The Sentinel CS Test demonstrated a sensitivity of 93% and specificity of 90% for prediction of the presence of Grade Group 2 or greater cancer, and the Sentinel HG Test demonstrated a sensitivity of 94% and specificity of 96% for the prediction of the presence of Grade Group 3 or greater cancer.

Conclusions:

The Sentinel PCa, CS and HG Tests demonstrated high levels of sensitivity and specificity, highlighting the utility of interrogation of urinary exosomal sncRNAs for noninvasively diagnosing and classifying prostate cancer with high precision.

Abbreviations and Acronyms

AMC

Albany Medical Center

CS

clinically significant

DMC

Downstate Medical Center

GG

Grade Group

HG

high grade

miRNA

micro RNA

MRI

magnetic resonance imaging

NEPC

no evidence of prostate cancer

NPV

negative predictive value

PCa

prostate cancer

PPV

positive predictive value

PSA

prostate specific antigen

sncRNA

small noncoding RNA

snoRNA

small nucleolar RNA

PSA screening has been shown to reduce prostate cancer mortality by 20% to 32%.1,2 However, the benefit of PSA as a screening test is limited because of the high false-positive rate of 65% to 75%3 and the risk of over diagnosis and overtreatment of clinically insignificant cancer. Use of PSA as a screening tool results in a large number of negative biopsies, patient anxiety, and financial and personal cost.4 MRI has recently emerged as an initial alternative to biopsy in men at risk for prostate cancer.5,6 However, the incorporation of MRI, with or without MRI guided biopsy, as an early step in the diagnostic process has many issues related to cost, availability, reliability, access and generalizability from centers of excellence. Many men are unable or unwilling to have MRI, and MRI requires transrectal ultrasound fusion and fusion targeted biopsy, which introduces additional sources of error and the requirement for a second procedure.7 In a recent randomized multicenter study, the positive predictive value of a positive MRI was only 35%.8 However, 2-year followup data show that a baseline MRI before confirmatory biopsy reduced failures of surveillance, with fewer patients progressing to higher grade cancer.9

Approximately 85% of patients with elevated PSA levels (greater than 3 ng/ml) referred for prostate cancer diagnosis will have a transrectal ultrasound guided or magnetic resonance imaging guided core needle biopsy; of these, about 40% to 50% will have histologically identifiable cancer.9,10 Further, about 15% of patients with a normal PSA harbor significant cancer.11,12 There is an unmet need for a cost-effective, widely available, easily interpretable noninvasive test that allows for accurate risk stratification and disease characterization prior to biopsy.

Exosomes are small extracellular vesicles that originate in the endosomal compartment of eukaryotic cells. They are found in biological fluids including blood, urine, semen and cerebrospinal fluid.13 Exosomes contain small noncoding RNAs, including miRNAs and small nucleolar RNAs which are derived from the cytoplasm and nucleolar region of the cell.14 The presence of exosomes and extracellular vesicles in the tumor microenvironment has been correlated with malignancy in a number of tumor types including prostate cancer and other urological cancers.15,16

miRNAs, their primary transcripts (pri-miRNAs) and immediate precursors (pre-miRNAs), are the most studied of the sncRNAs, and their biogenesis and mechanisms of action have been described in considerable detail.17 The function of miRNAs in prostate cancer has been the subject of intense investigation,18,19 and several studies have examined using them as biomarkers for diagnosis,20 monitoring active surveillance21 and for classifying advanced disease.22 snoRNAs serve as guide RNAs that facilitate the post-transcriptional modification of ribosomal RNA, transfer RNAs and small nuclear RNAs.23 snoRNAs are frequently subdivided into the 2 structural classes of 1) SNORDs, that are known to target methylation of specific recognition sequences in target RNAs, and 2) SNORAs, which facilitate the site specific enzymatic conversion of uridine to pseudouridine.23 A number of SNORDs also exhibit modification independent functions, particularly in differential splicing, as exemplified by SNORD27 which regulates alternative splicing of E2F7 pre-mRNA.24

The Sentinel Platform described here consists of 3 related but independent tests that rely on the interrogation of exosomal sncRNAs isolated from urine. The Sentinel PCa Test classifies patients with prostate cancer and those subjects with no cancer. For patients diagnosed with cancer, the Sentinel HG Test delineates patients with high grade (GG3-5) cancer who are candidates for therapeutic intervention from those with lower grade disease (GG1 and GG2) who are candidates for active surveillance. These latter patients can be monitored longitudinally for evidence of molecular progression using either the Sentinel HG Test or, for more conservative management, the Sentinel CS Test, which distinguishes between low grade disease (GG1) and higher (GG2-5).

We describe the development of a novel biostatistical based approach for the diagnosis and classification of prostate cancer based on the interrogation of miRNAs and snoRNAs isolated from urinary exosomes. The performance characteristics of this interrogation methodology demonstrate that the diagnostic tests have the potential to significantly reduce unnecessary core needle biopsies and incorrect disease classification, reducing morbidities and cost associated with the diagnosis and prognosis of prostate cancer.

Materials and Methods

Study Populations

Two independent patient groups were used for the development and validation of the miR Sentinel Tests. First, the discovery phase for the development of the Sentinel PCa and Sentinel CS Tests was a group of 235 participants from Albany Medical Center (Albany, New York) and SUNY (State University of New York) Downstate Medical Center (Brooklyn, New York). The clinical and demographic characteristics of these 235 participants are shown in table 1. The urine samples from these participants were interrogated on Affymetrix miR 4.0 arrays as described.

Table 1. Demographics and clinical characteristics of cohort used to develop classification algorithm

Control Ca Grade
GG1 GG2 GG3 GG4 GG5
No. Total (%) 89 (37.9) 90 (38.3) 34 (14.5) 9 (3.8) 7 (3.0) 6 (2.5)
Age:
 Range 23–89 53–83 50–81 60–81 51–74 59–74
 Mean ± SD 65.6 ± 13.0 67.3 ± 6.7 66.3 ± 6.6 72.1 ± 6.6 63.0 ± 7.5 67.2 ± 5.5
BMI (kg/m2):
 Range 18.9–54.1 20.7–53.3 22.0–44.1 26.1–36.1 16.0–44.1 27.5–41.2
 Mean ± SD 29.3 ± 5.6 28.4 ± 4.9 28.2 ± 3.8 30.1 ± 3.3 29.2 ± 8.5 32.7 ± 5.2
No. Race (%):
 NonHispanic white 72 (80.9) 78 (86.7) 30 (88.2) 7 (77.8) 7 (100) 6 (100)
 NonHispanic black 3 (3.4) 7 (7.8) 4 (11.8) 1 (11.1) 0 (0) 0 (0)
 Others 14 (15.7) 5 (5.5) 0 (0) 1 (11.1) 0 (0) 0 (0)
PSA (ng/ml): Not available
 Range 0.55–28.2 2.13–49.9 5.6–28.2 6.9–40.8 5.5–17.8
 Mean ± SD 6.3 ± 4.0 7.7 ± 7.9 10.5 ± 7.4 18.8 ± 10.7 8.7 ± 4.6

Under exempt study status, PSA levels were not available for patients with NEPC.

Secondly, for the training and validation of the miR Sentinel Tests, a total of 1,436 subjects were analyzed in a case-control design. Two patient groups were analyzed. One group was a retrospective sample of 613 patients obtained from the University Health Network GUBioBank (University of Toronto, Ontario, Canada). The other was a cohort of 823 subjects from Albany Medical Center and SUNY Downstate Medical Center, where urine samples were collected prior to biopsy and analyzed from participants with suspected prostate cancer, during a 2-year period (2017 to 2019). The clinical and demographic characteristics of these combined data sets are shown in table 2.

Table 2. Demographics and clinical characteristics of case-control sample

NEPC Ca Grade
GG1 GG2 GG3 GG4 GG5
No. total (%) 568 (39.5) 437 (30.4) 162 (11.3) 131 (9.1) 66 (4.6) 72 (5.0)
No. suspicious for PCa 374 376 150 123 63 69
Age:
 Range 23–90 46–93 50–96 49–95 50–93 54–91
 Mean ± SD 65.8 ± 9.0 70.3 ± 8.8 71.1 ± 8.2 74.2 ± 8.7 72.4 ± 9.7 72.7 ± 9.5
No. race:
 NonHispanic white 458 233 96 47 26 36
 NonHispanic black 61 20 11 5 9 3
 Others 49 184 55 79 31 33
PSA (ng/ml): Not available
 Range 0.21–108 1.24–32.0 1.67–138 1.93–1,400 1.98–199
 Mean ± SD 6.4 ± 6.2 7.4 ± 4.7 14.0 ± 19.4 57.6 ± 194.0 25.9 ± 36.0

Under exempt study status, PSA levels were not available for individual patients with no evidence of cancer. Suspicion of prostate cancer is based on elevated PSA and/or suspicious digital rectal examination.

Urine Collection and Processing

Nondigital rectal examination urine samples collected from Albany Medical Center and SUNY Downstate Medical Center were collected on the day of visit for clinical workup. Samples for the retrospective sample were retrieved from the GUBioBank, University Health Network, Toronto, Ontario, Canada and shipped frozen at -20C in bulk to the miR Scientific laboratories. Participant information was collected and anonymized as approved by the institutional review board at each participating site. Prostate cancer diagnosis was obtained by histopathological grading of core needle biopsies at each site.

After centrifugation to remove free cells and debris (15 minutes at 1,500 g at room temperature), sncRNAs were extracted from urine using Exosome RNA Isolation Kits (Norgen Biotek, Ontario, Canada) according to the manufacturer's instructions. sncRNA yields were quantified by fluorimetry and sncRNA samples were stored at -80C until analysis.

Microarray Analysis of Total Exosomal sncRNAs

sncRNAs were interrogated using the Affymetrix GeneChip™ miRNA 4.0 Array following the manufacturer's instructions. MIAME (Minimum Information About a Microarray Experiment) compliant raw data files for the 235 patients analyzed on these arrays have been deposited in the NCBI Gene Expression Omnibus25 and are accessible through the GEO Series accession GSE138740 (https://www.ncbi.nih.gov/geo.cgi?acc=GSE138740).

The small noncoding RNA entities interrogated for each participant were analyzed using proprietary selection and classification algorithms. The most informative sequences for distinguishing between cancer and noncancer, between GG 1 and GG2-5, and between GG1-2 and GG3-5 were identified using the same algorithmic approach. These sequences were then used as the basis of the QuantStudio™ OpenArray™ platform.

QuantStudio OpenArray Based Interrogation of Exosomal sncRNAs

cDNA Synthesis, Pre-Amplification of Selected miRNAs

For analysis of exosomal miRNA, total sncRNA was reverse transcribed in separate reactions with 3 specific miRNA stem-loop primer pools with the TaqManTM MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific) as recommended by the manufacturer. The miRNA cDNA pools were enriched individually with Pre-Amp primer pools for 16 cycles (95C for 10 minutes, 55C for 2 minutes, 72C for 2 minutes, 95C for 15 seconds and 60C for 4 minutes repeated for 16 cycles, and 99.9C for 10 minutes) and interrogated on the QuantStudio OpenArray on 3, 56-entity subarrays following the manufacturer’s recommendations.

cDNA Synthesis, Pre-amplification and Interrogation of Selected snoRNAs

Total sncRNA was reverse transcribed with the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific) as recommended by the manufacturer. snoRNA cDNA products were enriched by preamplification with a single Pre-Amp primer pool (95C for 10 minutes, 95C for 15 seconds and 60C for 4 minutes repeated for 18 cycles, and 99C for 10 minutes) and interrogated on 2, 56-entity subarrays.

Statistical Analysis

The selection algorithm identifies the group of exosomal sncRNAs that are associated with the outcome group of interest, for example NEPC versus prostate cancer (Sentinel PCa Test). The algorithm assesses the simultaneous association of sncRNAs without regard to type, which is particularly important since it is well established that transcription of many sncRNAs (particularly miRNAs) is coordinately modulated and sncRNAs frequently target the same mRNA transcripts, influencing mRNA stability, translatability or splicing. The importance measure assigned to a given sncRNA designates the impact of the sncRNA on predicting group outcome beyond that provided by all other sncRNAs.

The 1,436 participants described in table 2 were used to train and validate the interrogation of these sncRNAs using the OpenArray for the Sentinel PCa, Sentinel CS and Sentinel HG Tests to ensure that the output of the classification algorithm was not altered by the change platform. For these studies the validation data set consisted of 600 participants stratified so that an equal number of participants were biopsy negative versus positive (300 each), and that of the biopsy positive cases 200 were GG1-2 (146 GG1 and 54 GG2) and 100 were GG3-5. Patients from AMC/DMC constituted 36% of the validation data set. The remaining 836 participants formed the training data set, 52% of whom were from AMC/DMC.

The Sentinel PCa, Sentinel CS and Sentinel HG Tests are based on a classification algorithm that takes as input the sncRNA expression signature for a participant with unknown disease status (emulated by the validation set in this paper) and produces a Sentinel Score. The participant is classified by comparing this score to the predetermined cutoff value, referred to as the classification boundary (obtained from cross-validation in the training data set) that controls sensitivity for classifying a future patient with unknown disease status (but known expression signature), at a user-defined level (typically 95% or greater). For each test the classification boundary is shown as a vertical dashed line. The Sentinel PCa, CS and HG Tests operate analogously, using unique sncRNA expression signatures, however the classification rules and boundaries are different among the 3 tests. The tests are stand-alone and do not incorporate any information about other clinical markers such as prostate specific antigen, % core involvement, CAPRA (Cancer of the Prostate Risk Assessment) Score or Prostate Cancer Prevention Trial criteria.

RESULTS

Development of miR Sentinel PCa and CS Test Platform

Using the selection algorithm we identified the most informative sncRNA sequences that discriminate between cancer and no cancer. The sequences form the basis of the Sentinel PCa Test. For individuals classified as having prostate cancer, a similar approach was used to determine the most informative sncRNA sequences that discriminate between GG1 (indolent, low risk cancer) and GG2-5 (aggressive, intermediate and high risk cancers). These sequences form the basis of the Sentinel CS Test. Additional analyses demonstrated that the same collection of sncRNAs can be used to classify the patients into GG1+GG2 (low and favorable-intermediate risk cancer) versus GG3-5 (high risk cancers), which are used for the Sentinel HG Test.

To establish robust data sets for the miR Discovery Sentinel Tests, the participants in the “no cancer” group were selected from age matched men who were seen at urology clinics for issues unrelated to urological oncology (59) and from men who had 1 or more 12-needle diagnostic core needle biopsy that showed no evidence of prostate cancer (30). Patients in the “cancer” group (146) were selected based on the histopathology of the core needle biopsies, and 90 patients were classified as having GG1 cancer and 56 with GG2-5 (table 1). Exosomal sncRNAs, were interrogated using the Affymetrix miR 4.0 microarrays to define an expression signature for the Discovery PCa Test (fig. 1, A) and Discovery CS Test (fig. 1, C). Based on these data, the selection algorithm was used to identify the most informative sncRNAs for each test (fig. 1, B and D). The resulting Sentinel PCa Test incorporates the expression levels of 145 unique sncRNAs (85 miRNAs and 60 snoRNAs). Similarly, the resulting Sentinel CS Test uses 196 unique sncRNAs (130 miRNA and 66 snoRNAs) and the Sentinel HG Test examines 147 unique sncRNAs (122 miRNA and 25 snoRNAs). Of the sncRNAs 38 (23 miRNAs and 15 snoRNAs) are informative in all 3 tests, and a further 89 miRNA and 9 snoRNAs are common to more than 1 of the tests.

Figure 1.Classification and selection of informative sncRNAs from urinary exosomes using expression generated using miR 4.0 microarray analysis. In Discovery PCa Score scatter plot of cancer/no cancer status in training data set with classification boundary shown as vertical dashed line (A). Positive Discovery PCa Score indicates prostate cancer and negative Discovery PCa Score indicates no cancer. In identification of informative sncRNAs for Sentinel PCa Test (B), top 35 most informative sncRNA entities from selection algorithm are shown. With Discovery CS Score (C), patients are classified as having clinically insignificant cancer (GG1) or clinically significant cancer (GG2-5). Grade grouping assignments of GG1 or GG2-5 based on histopathology of core needle biopsies. In identification of informative sncRNAs for Sentinel CS Test (D), top 35 most informative sncRNA entities from selection algorithm discriminating between GG1 and GG2-5 cancer are shown.

Figure 1. Classification and selection of informative sncRNAs from urinary exosomes using expression generated using miR 4.0 microarray analysis. In Discovery PCa Score scatter plot of cancer/no cancer status in training data set with classification boundary shown as vertical dashed line (A). Positive Discovery PCa Score indicates prostate cancer and negative Discovery PCa Score indicates no cancer. In identification of informative sncRNAs for Sentinel PCa Test (B), top 35 most informative sncRNA entities from selection algorithm are shown. With Discovery CS Score (C), patients are classified as having clinically insignificant cancer (GG1) or clinically significant cancer (GG2-5). Grade grouping assignments of GG1 or GG2-5 based on histopathology of core needle biopsies. In identification of informative sncRNAs for Sentinel CS Test (D), top 35 most informative sncRNA entities from selection algorithm discriminating between GG1 and GG2-5 cancer are shown.

Validation of the Classification Algorithm in a Case Control Sample

Using the data set described in table 2, a testing data set of 600 subjects was randomly selected from the 1,436 participants while the remaining 836 subjects were used for training of the classification algorithms using data generated on the OpenArray platform of the most informative 280 sncRNAs. The scatter and sorted plots, and ROC curves for the training and validation studies for each test are shown in figures 2, 3 and 4. In each of these figures, the performance of the fully cross validated training data sets are shown in the upper panel. In the validation studies, the Sentinel PCa Test correctly classified 281/300 = 94% of patients with cancer as having cancer and 275/300 = 92% of participants with no cancer as having no cancer (sensitivity 94%, specificity 92%; empirical PPV 92%, empirical NPV 94%) (fig. 2). The Sentinel CS Test correctly classified 143/154 (93%) GG2-5 cases as high grade and 132/146 (90%) GG1 cases as not high grade (sensitivity 93%, specificity 90%; empirical PPV 91%, empirical NPV 92%) (fig. 3). The Sentinel HG Test correctly classified 94/100 (94%) GG3-5 cases as high grade and 191/200 (96%) GG1-2 cases as not high grade (sensitivity 94%, specificity 96%; empirical PPV 91%, empirical NPV 97%) (fig. 4). The performance characteristics of the Sentinel tests for both the training and testing studies are summarized in table 3. For each of the tests, the fully cross validated training study exhibits a slightly better overall sensitivity and specificity than the validation data set, however the performance characteristics for these tests in the validation study are still excellent.

Figure 2.Case control study of high throughput OpenArray interrogation of urinary exosomal sncRNA using miR Sentinel PCa Test. Training analysis of case control population of 836 subjects (268 no cancer, 568 cancer) of paired scatter (A) and sorted (B) plots of predicted cancer status. Corresponding ROC calculated shown with user defined false-negative rate of 0.05 highlighted in red (C). Testing analysis of cross-validation case control of 600 subjects (300 no cancer, 300 cancer) of paired scatter (D) and sorted (E) plots of predicted cancer status, and ROC curve (F).

Figure 2. Case control study of high throughput OpenArray interrogation of urinary exosomal sncRNA using miR Sentinel PCa Test. Training analysis of case control population of 836 subjects (268 no cancer, 568 cancer) of paired scatter (A) and sorted (B) plots of predicted cancer status. Corresponding ROC calculated shown with user defined false-negative rate of 0.05 highlighted in red (C). Testing analysis of cross-validation case control of 600 subjects (300 no cancer, 300 cancer) of paired scatter (D) and sorted (E) plots of predicted cancer status, and ROC curve (F).

Figure 3.Case control study of high throughput OpenArray interrogation of urinary exosomal sncRNA using miR Sentinel CS Test. Training analysis of case control population of 568 patients (291 GG1, 277 GG2-5) of paired scatter (A) and sorted (B) plots of predicted tumor classification into clinically significant (GG2-5) vs clinically insignificant (GG1). ROC plot and user defined false-negative rate of 0.05 are shown (C). Testing analysis of cross-validation case control of 300 patients (146 GG1, 154 GG2-5) of paired scatter (D) and sorted (E) plots of predicted tumor classification, and ROC curve (F).

Figure 3. Case control study of high throughput OpenArray interrogation of urinary exosomal sncRNA using miR Sentinel CS Test. Training analysis of case control population of 568 patients (291 GG1, 277 GG2-5) of paired scatter (A) and sorted (B) plots of predicted tumor classification into clinically significant (GG2-5) vs clinically insignificant (GG1). ROC plot and user defined false-negative rate of 0.05 are shown (C). Testing analysis of cross-validation case control of 300 patients (146 GG1, 154 GG2-5) of paired scatter (D) and sorted (E) plots of predicted tumor classification, and ROC curve (F).

Figure 4.Case control study of high throughput OpenArray interrogation of urinary exosomal sncRNA using miR Sentinel HG Test. Training analysis of case control population of 568 patients (399 GG1-2, 169 GG3-5) of paired scatter (A) and sorted (B) plots of predicted tumor classification into low grade (GG1-2) vs high grade (GG3-5) disease. ROC plot and user defined false-negative rate of 0.05 are shown (C). Testing analysis of cross-validation case control of 300 patients (200 GG1-2, 100 GG3-5) of paired scatter (D) and sorted (E) plots of predicted low and high grade tumor classification, and ROC curve (F).

Figure 4. Case control study of high throughput OpenArray interrogation of urinary exosomal sncRNA using miR Sentinel HG Test. Training analysis of case control population of 568 patients (399 GG1-2, 169 GG3-5) of paired scatter (A) and sorted (B) plots of predicted tumor classification into low grade (GG1-2) vs high grade (GG3-5) disease. ROC plot and user defined false-negative rate of 0.05 are shown (C). Testing analysis of cross-validation case control of 300 patients (200 GG1-2, 100 GG3-5) of paired scatter (D) and sorted (E) plots of predicted low and high grade tumor classification, and ROC curve (F).

Table 3.
1-Error Rate Numerator Denominator Proportion 95% Lower CI 95% Upper CI
Training data set
miR Sentinel PCa:
 Sensitivity 533 568 0.938 0.917 0.956
 Specificity 257 268 0.959 0.93 0.978
 PPV 533 544 0.98 0.966 0.989
 NPV 257 292 0.88 0.839 0.914
miR Sentinel CS:  
 Sensitivity 257 277 0.928 0.893 0.954
 Specificity 265 291 0.911 0.874 0.939
 PPV 257 283 0.908 0.871 0.938
 NPV 265 285 0.93 0.896 0.955
miR Sentinel HG:  
 Sensitivity 161 169 0.953 0.913 0.978
 Specificity 383 399 0.96 0.937 0.976
 PPV 161 177 0.91 0.861 0.945
 NPV 383 391 0.98 0.962 0.99
Testing data set
miR Sentinel PCa:
 Sensitivity 281 300 0.937 0.905 0.96
 Specificity 275 300 0.917 0.882 0.944
 PPV 281 306 0.918 0.884 0.945
 NPV 275 294 0.935 0.903 0.959
miR Sentinel CS:  
 Sensitivity 143 154 0.929 0.88 0.962
 Specificity 132 146 0.904 0.848 0.944
 PPV 143 157 0.911 0.859 0.948
 NPV 132 143 0.923 0.871 0.959
miR Sentinel HG:
 Sensitivity 94 100 0.94 0.88 0.975
 Specificity 191 200 0.955 0.919 0.978
 PPV 94 103 0.913 0.846 0.956
 NPV 191 197 0.97 0.938 0.987

Discussion

As reported here, the performance characteristics of the Sentinel PCa (sensitivity 94% and specificity 92%) and Sentinel HG (sensitivity 94% and specificity 96%) make it possible to accurately distinguish between no cancer and cancer, and low grade disease and high grade disease by interrogating the sncRNA present in urinary exosomes. The combination of the Sentinel PCa and Sentinel HG Tests has the benefit of identifying subjects who have no evidence of prostate cancer, and those patients who harbor high grade disease, from a single urine sample. The performance characteristics of these tests compare very favorably to other recently published diagnostic and prognostic tests designed for the same purpose.26,27 Since some men with GG2 cancer are candidates for active surveillance, the availability of both the Sentinel HG Test (for GG3 or higher) and the Sentinel CS Test (for GG2 or greater) for longitudinal monitoring of patients with low (GG1) or intermediate risk (GG2) cancer accommodates a range of criteria for patient eligibility for active surveillance.

The performance characteristics of all 3 Sentinel tests are high. The Sentinel PCa Test identifies 94% of patients with prostate cancer. The Sentinel HG clearly delineates those high grade (GG3-5), high risk cancers that need immediate therapy from lower grade (GG1 and GG2), low and intermediate risk cancers, and the Sentinel CS Test effectively distinguishes between low grade (GG1) and intermediate and high grade cancers (GG2-5).

Discordance between the Sentinel test results and the core biopsy outcomes may reflect pathological miss of higher grade cancer or a true test misclassification. Given the known false-negative rate of core needle biopsies, we estimate the apparent false-positive rate of the Sentinel PC is 6% to 12% based on the 95% CI. This compares favorably to the 50% to 60% false-positive rate reported for systematic transrectal ultrasound guided core needle biopsies9 and 30% to 40% false-positive rates reported for various MRI targeted biopsies.10,28 The combined apparent false-positive and negative rates of the Sentinel HG Test with biopsy outcome is around 10%. This performance is within the confidence limits of the well-established rate of misattribution of grade resulting from systematic biopsies. Therefore, it is plausible that in this case the apparent false-positive cases resulting from the Sentinel HG Test may in fact be those who harbor higher grade cancer missed on the systematic biopsy. An alternative explanation is that some of these represent actual false-positive test results. To further investigate this issue we are currently performing a large retrospective study comparing the Sentinel Scores with radical prostatectomy pathology.

The statistical methodologies used for the selection algorithm for the sncRNAs, and classification algorithm for the Sentinel PCa and Sentinel HG Tests are data driven, and do not require any a priori knowledge of the biological function of the selected sequences. The number and complexity of interactions between miRNAs and snoRNAs, and mRNA translation and translatability make identification of informative sncRNAs virtually impossible using bioinformatic approaches, largely due to the current lack of functional data for the snoRNAs in the knowledge base. In addition, since it is unclear whether exosomal RNA contents result from selective or nonselective bulk packaging of sncRNA into exosomes, it is not possible to infer which intracellular pathways might be dysregulated in the cell using bioinformatic approaches.

Conclusion

In this comprehensive evaluation of the urinary exosome based miR Sentinel PCa, CS and HG Tests, high accuracy for identifying the presence of cancer and the presence of high grade cancer was demonstrated. These data demonstrate that the evaluation of a panel of urinary exosomal sncRNA offers the ability to accurately and noninvasively screen, diagnose, characterize and monitor prostate cancer.

Acknowledgments

Clinical coordinators (Laura Davey and Brenda Romeo at AMC, Zaheer Bukhari at DMC and Heidi Wagner at GUBioBank University Health Network BioBank) were responsible for the identification of patients, and collection and delivery of the specimens used in these studies. James Criss, Shannon Lavender, Samantha Mall and Siti Zainuddin from miR Scientific provided excellent technical assistance.

References

  • 1. : Prostate-specific antigen-based prostate cancer screening: reduction of prostate cancer mortality after correction for nonattendance and contamination in the Rotterdam section of the European Randomized Study of Screening for Prostate Cancer. Eur Urol 2014; 65: 329. Google Scholar
  • 2. : A 16-yr follow-up of the European Randomized Study of Screening for Prostate Cancer. Eur Urol 2019; 76: 43. Google Scholar
  • 3. : Screening and prostate cancer mortality: results of the European Randomised Study of Screening for Prostate Cancer (ERSPC) at 13 years of follow-up. Lancet 2014; 384: 2027. Google Scholar
  • 4. US Preventive Services Task Force: Screening for prostate cancer: US Preventive Services Task Force recommendation statement. JAMA 2018; 319: 1901. Google Scholar
  • 5. : MRI-targeted or standard biopsy for prostate-cancer diagnosis. N Engl J Med 2018; 378: 1767. Google Scholar
  • 6. : Prostate magnetic resonance imaging, with or without magnetic resonance imaging-targeted biopsy, and systematic biopsy for detecting prostate cancer: a Cochrane systematic review and meta-analysis. Eur Urol 2020; 77: 78. Google Scholar
  • 7. : What needs to be done before prostate magnetic resonance imaging can safely be applied in the general population. Eur Urol 2019; 75: 397. Google Scholar
  • 8. : Active Surveillance Magnetic Resonance Imaging Study (ASIST): results of a randomized multicenter prospective trial. Eur Urol 2019; 75: 300. Google Scholar
  • 9. : Randomized study of systematic biopsy versus magnetic resonance imaging and targeted and systematic biopsy in men on active surveillance (ASIST): 2-year postbiopsy follow-up. Eur Urol 2020; 77: 311. Google Scholar
  • 10. : The current state of MR imaging-targeted biopsy techniques for detection of prostate cancer. Radiology 2017; 285: 343. Google Scholar
  • 11. : Prevalence of prostate cancer among men with a prostate-specific antigen level ≤4.0 ng per milliliter. N Engl J Med 2004; 350: 2239. Google Scholar
  • 12. : Screening for prostate cancer: systematic review and meta-analysis of randomised controlled trials. BMJ 2010; 341: c4543. Google Scholar
  • 13. : Biogenesis, secretion, and intercellular interactions of exosomes and other extracellular vesicles. Ann Rev Cell Dev Biol 2014; 30: 255. Google Scholar
  • 14. : Vesiclepedia 2019: a compendium of RNA, proteins, lipids and metabolites in extracellular vesicles. Nucleic Acids Res 2019; 47: D516. Google Scholar
  • 15. : Prostate cancer-derived urine exosomes: a novel approach to biomarkers for prostate cancer. Br J Cancer 2009; 100: 1603. Google Scholar
  • 16. : Extracellular vesicles in urological malignancies: an update. Nat Rev Urol 2020; 17: 11. Google Scholar
  • 17. : Post-transcriptional control of miRNA biogenesis. RNA 2019; 25: 1. Google Scholar
  • 18. : MicroRNAs in prostate cancer: an overview. Oncotarget 2017; 8: 5024. Google Scholar
  • 19. : MicroRNAs in prostate cancer: from function to biomarker discovery. Exp Biol Med 2018; 243: 817. Google Scholar
  • 20. : A plasma biomarker panel of four microRNAs for the diagnosis of prostate cancer. Sci Rep 2018; 8: 6653. Google Scholar
  • 21. : Assessment of serum microRNA biomarkers to predict reclassification of prostate cancer in patients on active surveillance. J Urol 2018; 199: 1475. LinkGoogle Scholar
  • 22. : Circulating miRNAs as non-invasive biomarkers to predict aggressive prostate cancer after radical prostatectomy. J Transl Med 2019; 17: 173. Google Scholar
  • 23. : The ribosome biogenesis-cancer connection. Cells 2019; 8: E55. Google Scholar
  • 24. : Dual function of C/D box small nucleolar RNAs in rRNA modification and alternative pre-mRNA splicing. Proc Natl Acad Sci USA 2016; 113: E1625. Google Scholar
  • 25. : Gene expression omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res 2002; 30: 207. Google Scholar
  • 26. : A urine-based DNA methylation assay, ProCUrE, to identify clinically significant prostate cancer. Clin Epigenetics 2018; 10: 147. Google Scholar
  • 27. : Multicenter optimization and validation of a 2-gene mRNA urine test for detection of clinically significant prostate cancer prior to initial prostate biopsy. J Urol 2019; 202: 256. LinkGoogle Scholar
  • 28. : Comparison of targeted vs systematic prostate biopsy in men who are biopsy naive: the prospective assessment of image registration in the diagnosis of prostate cancer (PAIREDCAP) study. JAMA Surg 2019; 154: 811. Google Scholar

No direct or indirect commercial, personal, academic, political, religious or ethical incentive is associated with publishing this article.

Advertisement