High Sensitivity and Specificity Platform to Validate MicroRNA Biomarkers in Cancer and Human Diseases

Abstract

We developed a technology for detecting and quantifying trace nucleic acids using a bracketing protocol designed to yield a copy number with approximately ± 20% accuracy across all concentrations. The microRNAs (miRNAs) let-7b, miR-15b, miR-21, miR-375 and miR-141 were measured in serum and urine samples from healthy subjects and patients with breast, prostate or pancreatic cancer. Detection and quantification were amplification-free and enabled using osmium-tagged probes and MinION, a nanopore array detection device. Combined serum from healthy men (Sigma-Aldrich, St. Louis, MO, USA #H6914) was used as a reference. Total RNA isolated from biospecimens using commercial kits was used as the miRNA source. The unprecedented ± 20% accuracy led to the conclusion that miRNA copy numbers must be normalized to the same RNA content, which in turn illustrates (i) independence from age, sex and ethnicity, as well as (ii) equivalence between serum and urine. miR-21, miR-375 and miR-141 copies in cancers were 1.8-fold overexpressed, exhibited zero overlap with healthy samples and had a p-value of 1.6 × 10⁻²², tentatively validating each miRNA as a multi-cancer biomarker. miR-15b was confirmed to be cancer-independent, whereas let-7b appeared to be a cancer biomarker for prostate and breast cancer, but not for pancreatic cancer.

1. Introduction

The incidence of cancer has not declined despite preventive efforts worldwide [1,2]. The cancer incidence rate increases steadily with age and reaches 1 in every 100 people aged ≥ 60 years and older [3]. Most adults undergo an annual or biannual physical medical examination that does not include a multi-cancer test. Specialized tests are available for some prevalent cancer indications, but they are invasive, often painful, and can lead to unacceptable false-positive or false-negative results [4,5,6,7,8,9]. Rigorous observational studies in Europe failed to determine the effects of mammography screening. Mammography screening only partially distinguishes patients with breast cancer from healthy women and increases the number of mastectomies performed [4,5]. A blood test for prostate-specific antigen (PSA) measures the level of PSA, which is a substance produced by the prostate. The levels of PSA in the blood might be greater in men with prostate cancer, but the PSA test exhibits a 15% false-positive rate, which leads to unnecessary surgical biopsies [6,7]. No single diagnostic test is available for pancreatic cancer. Serum levels of the antigen CA 19-9 higher than 37 U/mL are exploited for the diagnosis of pancreatic cancer in symptomatic patients, but they are not useful as screening markers in asymptomatic individuals because of their low positive predictive value (PPV) [8,9]. A definitive diagnosis requires a series of imaging scans, blood tests, and biopsies, which are typically performed after the appearance of symptoms. Pancreatic cancer is called a “silent” disease because it may cause patients to experience no symptoms until it is too late. Early detection is promising for cancer treatment and for saving lives. Early detection relies heavily on minimally invasive tests using blood samples, urine, saliva, or other biological fluids, so-called liquid biopsies [10,11]. Multiple cancer blood tests using known protein cancer biomarkers [12], circulating tumor DNA shed from tumors [13,14] and DNA methylation profiles [15] are currently being tested in large trials to replace exploratory invasive tissue biopsies and support clinical decisions in symptomatic subjects. Even asymptomatic individuals, especially those aged >60 years, with or without a family history of cancer, are worried about having cancer. A liquid biopsy test to label an asymptomatic individual as cancer-free will ease worries and reduce unnecessary medical procedures. Such a test, like the one proposed herein, will be a valuable addition to regular physical and medical examinations.

A 2001 seminal publication by Victor Ambros summarized findings regarding the function of miRNAs [16], a class of small noncoding RNAs with a length of 18 to 25 nucleotides [17,18]. Ambros proposed them to be the “tiny regulators” that control post-transcriptional gene expression, including that related to cell growth, differentiation, development and apoptosis. miRNAs are abundant in most eukaryotes, and approximately 2500 known miRNAs are common to all humans [18,19]. In 2008, Mitchell et al. illustrated that miRNAs are stable in the blood, which renders them viable biomarkers [20], and empowered more than 150,000 peer-reviewed studies worldwide [21,22,23,24,25,26,27,28,29,30]. Several studies have investigated the expression of miR-375 and miR-141 in multiple cancer indications, and many have confirmed that these miRNAs are overexpressed in the serum of cancer patients compared with healthy controls [20,23,26]. A recent review listed 29 medical studies with a total of approximately 7000 subjects in which elevated miR-21 levels were reported across neoplastic and non-neoplastic diseases [25]. miR-21 may not be a biomarker for a specific disease; however, it is a multi-cancer and multi-disease biomarker. One should be able to determine from a regular medical exam whether his or her miR-21 level is markedly higher than that found in healthy subjects (HL). In addition to miRNAs, no other biomarker has been explored intensely and found to be involved in cancer onset, progression, metastasis or survival. Often, the miRNA data from cancer patients overlap with the data from healthy controls, and it is only the median from cancer samples that differs from the median of healthy samples [27,31,32,33,34,35]. For example, droplet digital PCR (ddPCR) data from patients with urological cancers reported copy numbers per μL of plasma for miR-126, miR-141, miR-155, miR-182 and miR-375 in the range of 0–3000, 0.5–4, 2–40, 1–20 and 1–40, respectively [27]. Notably, a similar significant variation was observed in cancer and healthy samples [27]. This type of data is not atypical for miRNA quantification and has prevented the validation of selected miRNAs as cancer biomarkers.

The overlap of miRNA data between diseased and healthy subjects, the quantitative disagreement among studies and the conjecture that a certain miRNA acts as an oncogene or as a tumor suppressor have been attributed to differences in biospecimen collection methods, study protocols, choice of reference, analytical methods, population variation, disease stage, etc. [31,32,33,34,35]. To improve these statistics, a collective response from an miRNA panel has been proposed. To the best of our knowledge, no miRNA studies have reported zero data overlap between healthy samples and samples with a certain disease. The concentration of miRNAs in blood is in the low femtomolar (fM) range, which is a billion-fold less than the micromolar (μM) range required by typical UV-Vis analytical tools. Current methods for profiling the relative abundance of miRNAs in biological fluids or tissues include small RNA sequencing, reverse transcription–quantitative PCR (RT–PCR), droplet digital PCR (ddPCR) and microarray hybridization [27,31,32,33,34,35]. Although identification works well with these tools, the quantification accuracy and choice of reference have been questioned and may be partially responsible for conflicting conclusions [31,32,33,34,35].

Our earlier work estimated the levels of miR-21, miR-375 and miR-141 in cancer to be 2- to 3-fold higher compared to those in healthy controls. This estimate agrees with the 1.8-fold overexpression observed in a prostate cancer study [23], and the 1.7-fold overexpression of miR-21 observed in a lung cancer study [22]. The data described below were in excellent agreement with those reported in previous studies. A 2-fold overexpression is a small effect that is difficult to measure accurately using amplification-based techniques. This may explain the observed disagreement between studies and the overlap of data between cancer and healthy samples. A tentative cancer screening assay for asymptomatic subjects must be robust, reliable and accurate and should be preceded by studies with zero data overlap between cancer and healthy samples.

To achieve zero data overlap, an analytical assay for biomarker overexpression must exhibit a lower limit in cancer samples that is equal to or higher than the upper limit observed in healthy controls. Similarly, an analytical assay testing biomarker underexpression must exhibit a higher limit in cancer samples that is equal to or lower than the lower limit in healthy controls.

Table 1. (A) Correlation of measurement’s accuracy and miRNA x-fold overexpression to achieve zero overlap between healthy and diseased samples (1). (B) Correlation of measurement’s accuracy and miRNA x-fold underexpression to achieve zero overlap between healthy and diseased samples (1).

(A)
Accuracy of Measurement (+/−)	Normalized Control, Range	Average Control	Normalized Disease Overexpressed, Range	Average Disease	miRNA in Disease/Control, x-Fold
15%	0.85 to 1.15	1.0	1.15 to 1.55	1.35	>1.35
20%	0.8 to 1.2	1.0	1.2 to 1.8	1.5	>1.5
30%	0.7 to 1.3	1.0	1.3 to 2.4	1.85	>1.85
40%	0.6 to 1.4	1.0	1.4 to 3.2	2.3	>2.3
50%	0.5 to 1.5	1.0	1.5 to 4.5	3.0	>3.0
75%	0.25 to 1.75	1.0	1.75 to 12.25	7.0	>7.0
(B)
Accuracy of Measurement (+/−)	Normalized Control, Range	Average Control	Normalized Disease Underexpressed, Range	Average Disease	miRNA in Control/Disease, x-Fold
15%	0.85 to 1.15	1.0	0.63 to 0.85	0.74	>1.35
20%	0.8 to 1.2	1.0	0.54 to 0.8	0.67	>1.5
30%	0.7 to 1.3	1.0	0.4 to 0.7	0.55	>1.8
40%	0.6 to 1.4	1.0	0.3 to 0.6	0.45	>2.2
50%	0.5 to 1.5	1.0	0.18 to 0.5	0.34	>2.9
75%	0.25 to 1.75	1.0	0.036 to 0.25	0.14	> 7.1

(A) (1) To understand the examples shown in the table, let us consider the calculation for ±20% accuracy. The average normalized control was 1.0, with a lower limit (0.8) and an upper limit (1.2). Note that the ratio of the upper limit to the lower limit is 1.2/0.8 = 1.5. A zero overlap of data requires that the lower limit of disease measurement be equal to or greater than 1.2. For a normalized overexpressed disease, the same 20% accuracy yields a factor of 1.5, with a lower limit of 1.2 and an upper limit of 1.8. The other entries in this table are calculated in a similar manner. (B) (1) Small numerical differences between the last columns in Table 1A,B are due to rounding up of the calculations. In Table 1B the lower limit of the normalized control range serves as the upper limit of the normalized disease underexpressed range.

A,B illustrate the interplay between assay accuracy and biomarker/miRNA overexpression or underexpression. Notably, over- or underexpression of the specific biomarker did not change the correlation between x-fold expression and assay accuracy (see identical results in the last columns of Table 1A,B). For example, if an miRNA quantification assay is associated with a ±30% accuracy, then a disease associated with less than 1.85-fold miRNA overexpression will yield overlapping data between healthy and diseased samples, and this miRNA will not be validated as a biomarker. Alternatively, if the same assay investigates a disease associated with more than 1.85-fold miRNA overexpression, the data will exhibit zero overlap between healthy and diseased individuals. These correlations presume that the only significant parameter for miRNA levels is the presence or absence of a disease. It was also assumed that assay accuracy remained constant within the range of the tested measurements. Other parameters, like age, gender, race, etc., if they turn out to be significant, will need to be addressed separately by limiting the tested population. Typically, analytical assays require extensive real-life sample testing to determine assay accuracy, which delays development and implementation. An analytical tool with protocol-defined accuracy, constant across all relevant concentrations, such as that described below, will streamline testing, and enable miRNA validation.

The diverging literature reports and correlations in Table 1A,B highlight the need for a novel analytical tool. We opted for an amplification-free assay, a nanopore array detector suitable for single-stranded (ss) nucleic acid trace measurements and a bracketing protocol designed to deliver miRNA copies with approximately ± 20% accuracy across all concentrations. The proof of principle of this concept was reported in 2020 [36]. A recent study highlighted the implementation of this novel technique for quantifying miRNAs in serum and estimated 2- to 3-fold overexpression of known miRNA cancer biomarkers [37]. Here, we confirmed earlier serum results and, using let-7b as an example, detailed an experimental strategy for validating miRNAs as cancer biomarkers. Strikingly, miRNA measurements using serum and urine samples demonstrated equivalence, which advocates the replacement of blood drawn for urine collection. The technology implemented here is well positioned to overcome inconsistencies in the miRNA field and revolutionize the current thinking in validating miRNA biomarkers.

Nanopores have shown promise for trace measurements, and experimental platforms have been successfully used to quantify miRNAs [38,39,40,41,42,43,44,45,46]. The MinION device from Oxford Nanopore Technologies (ONT) is the only commercially available nanopore array detection device used in conjunction with a consumable flow cell with 2048 embedded nanopores. These proprietary protein nanopores permit the translocation of ss and prevent the translocation of double-stranded (ds) nucleic acids, as shown by ONT and us. The platform includes 512 independent detection channels, with one detection channel per four nanopores, and is promoted for sequencing long ss DNA/RNA. Sequencing combined with DNA-barcoded probes has been used for the detection of biomarkers, including miRNAs [47,48]. As an alternative, rolling circle reverse transcription was used to sequence miRNAs using MinION [49]. However, the lowest detection level of these techniques is in the picomolar range, which is 100- to 1000-fold higher than the miRNA levels in biological fluids, rendering these techniques unsuitable for miRNA quantification in liquid biopsies.

The MinION software, MINKNOW, all versions, reports the raw data, that is, the ion current (i), as a function of time (t) and can be exploited for ion conductance (sensing) experiments, as described by us [36,37,46] and others [50]. Our earlier study showed that selective osmium tagging of an oligo yields a chemically stable probe that hybridizes efficiently with complementary DNA, RNA or miRNA targets. These probes traverse size-appropriate proteins [45] and solid-state nanopores [44], including nanopores on the MinION platform [36,37,46]. Owing to the bulkiness of the osmium tags, the translocation of the probe was markedly slower than that of the intact RNA/DNA. Whereas other nanopore platforms can detect and report all translocations, MinION selectively detects our probes over intact nucleic acids. This is because of the relatively slow data acquisition rate at three data points/ms, which quantitatively detects our optimized probes but misses most intact nucleic acids. In the absence of the target, the probe is free, traverses the nanopores and is detected owing to an increase in the reported events over the background noise (see Figure 1, scheme in the middle). Hybridization between the target and the probe yields a hybrid that is too large to traverse the nanopore, resulting in no probe detection or silencing. Target quantification was based on 1:1 hybridization and a known probe concentration. The number of miRNA copies was determined by bracketing, that is, from the average of two experiments, one that yielded probe detection and another that yielded no probe detection, or silencing [36,37]. To obtain the desired ± 20% accuracy, the two experiments must differ by approximately 67% in either the probe or RNA. Typically, more than two sets of experiments were conducted to confirm the miRNA copy number determination. Therefore, the assay is currently low-throughput but delivers data with a protocol defined by ±20% accuracy.

2. Results and Discussion

2.1. Single-Molecule Ion Conductance Experiments Using the MinION from ONT

ONT provides protocols for sequencing long RNAs/DNAs but not for single-molecule ion conductance (sensing). MINKNOW recorded the raw data from the ion conductance experiments, which were subsequently analyzed using a publicly available algorithm, OsBp-detect, developed specifically for this application [51]. While most studies report relative miRNA abundance, our technology measures miRNA copies in aliquots, used for the nanopore experiments. Mitchell et al. reported miR-15b at ~10,000 and miR-16 at ~110,000 copies per 1 μL of plasma from three individuals [20]. We recently reported that miR-15b = 8855 and miR-16 = 105,125 copies per 1 μL of H6914 combined serum ([37] and Table 2). This consensus was the first demonstration of a MinION-based sensing assay. This agreement was partially fortuitous because the total RNA from the serum exhibited rather small sample-to-sample variation [52], rendering the H6914 RNA content comparable to the plasma RNA content of the individuals tested by Mitchell et al. [20]. Over the last three years, we purchased four different lots of H6914, isolated total RNA (~16ng/μL) from each and found reproducible copy numbers for six miRNAs (Table 2). This consistency is remarkable considering that (i) total RNA was isolated using different lots of the Monarch RNA isolation kit from New England Biolabs; (ii) nanopore measurements were conducted by different analysts; (iii) chemically distinct probes were used; (iv) different experimental protocols were used; (v) different versions of MinION flow cells (R9 or R10) were used; and (vi) different versions of MINKNOW software were used.

Table 2. miRNA copies measured per 1 μL of total RNA isolated from a biospecimen, serum or urine.

Biospecimen	H6914 (1) 1st Lot (HL) (2)	H6914 (1) 2nd Lot	H6914 (1) 3rd Lot	H6914 (1) 4th Lot	Serum H2 (3)	Urine1 H2 (3)	Urine2 H2 (3,4)
Total RNA, ng/mL	16.0	16.5	15.9	14.3	20.7	16.8	27.4 (4)
miRNA						Copies (+/−%)	Copies (+/−%)
miR-16	210,250
miR-15b	17,710	16,716	17,687		21,852	15,517 (7)
let-7b	12,150			8668 (5)			19,853 (37)
miR-21-5p	10,494		10,514		>2.0 × HL	9855 (14)	21,352 (22)
miR-375-3p	9240	9636	8292		1.5 to 2.0 × HL
miR-141-3p	6096		5341	4919 (5)	1.5 to 2.0 × HL	5313 (12)

(1) H6914 was the combined serum of healthy men and was purchased from Sigma-Aldrich. miRNA copies measured from the first and second lots were reported earlier per 1 μL of serum [37] and are listed here per 1 μL of total RNA, which equals 2 μL of serum. The accuracy values not listed here are all less than +/−20%. (2) HL stands for healthy level (H6914 1st lot is the control/reference in this study). (3) H2 is a healthy woman, in the age group 71–75, tested multiple times over a period of 3 years (see Table 3 and Table 4); only the serum HW data were reported earlier [37]. H2 (female) miRNA copies (six measurements for five miRNAs) normalized to the RNA content of the H6914 1st lot are listed in Table 4 for both urine samples and illustrate the match with H6914. (4) The original Urine2 sample contained RNA 82.3 ng/μL, which was diluted 1/3 with water before mixing with the probe. The number of miR-21 copies normalized to the RNA content in the H6914 1st lot was 21,352 × 16/27.4 = 12,468. The number of copies of H2-related miR-21 normalized to the number of copies of miR-21 in the H6914 1st lot was 12,468/10,494 = 1.19 (Table 3, 5th row, in the healthy urinary section). (5) Normalization to the 1st lot of H6914 yielded let-7b = 9698 (HL = 0.80) from 8668 × 16/14.3 (9698/12,150), and miR-141 = 5503 (HL = 0.90) from 4919 × 16.0/14.3 (5503/6096), both within experimental error comparable to the other lots.

Table 2. miRNA copies measured per 1 μL of total RNA isolated from a biospecimen, serum or urine.

Biospecimen	H6914 (1) 1st Lot (HL) (2)	H6914 (1) 2nd Lot	H6914 (1) 3rd Lot	H6914 (1) 4th Lot	Serum H2 (3)	Urine1 H2 (3)	Urine2 H2 (3,4)
Total RNA, ng/mL	16.0	16.5	15.9	14.3	20.7	16.8	27.4 (4)
miRNA						Copies (+/−%)	Copies (+/−%)
miR-16	210,250
miR-15b	17,710	16,716	17,687		21,852	15,517 (7)
let-7b	12,150			8668 (5)			19,853 (37)
miR-21-5p	10,494		10,514		>2.0 × HL	9855 (14)	21,352 (22)
miR-375-3p	9240	9636	8292		1.5 to 2.0 × HL
miR-141-3p	6096		5341	4919 (5)	1.5 to 2.0 × HL	5313 (12)

(1) H6914 was the combined serum of healthy men and was purchased from Sigma-Aldrich. miRNA copies measured from the first and second lots were reported earlier per 1 μL of serum [37] and are listed here per 1 μL of total RNA, which equals 2 μL of serum. The accuracy values not listed here are all less than +/−20%. (2) HL stands for healthy level (H6914 1st lot is the control/reference in this study). (3) H2 is a healthy woman, in the age group 71–75, tested multiple times over a period of 3 years (see Table 3 and Table 4); only the serum HW data were reported earlier [37]. H2 (female) miRNA copies (six measurements for five miRNAs) normalized to the RNA content of the H6914 1st lot are listed in Table 4 for both urine samples and illustrate the match with H6914. (4) The original Urine2 sample contained RNA 82.3 ng/μL, which was diluted 1/3 with water before mixing with the probe. The number of miR-21 copies normalized to the RNA content in the H6914 1st lot was 21,352 × 16/27.4 = 12,468. The number of copies of H2-related miR-21 normalized to the number of copies of miR-21 in the H6914 1st lot was 12,468/10,494 = 1.19 (Table 3, 5th row, in the healthy urinary section). (5) Normalization to the 1st lot of H6914 yielded let-7b = 9698 (HL = 0.80) from 8668 × 16/14.3 (9698/12,150), and miR-141 = 5503 (HL = 0.90) from 4919 × 16.0/14.3 (5503/6096), both within experimental error comparable to the other lots.

Table 3. MinION experiments targeting let-7b in healthy and cancer samples to illustrate validation strategy of let-7b as biomarker for prostate and breast, but not as biomarker for pancreatic cancer.

Subject ID (1)	Biospecimen	Condition (2)	Isolated Total RNA (ng/μL)	RNA (μL) (3)	Probe Let-7b 30 fM (μL) (3)	Probe Copies per 1 μL of RNA	Normalized Probe Copies (4)	Normalized Probe Copies/12,150 (5)	Ratio of Late to Early (Ir/Io)max (6)	Result (7)	Let-7b/ Let-7b HL (8)	Let-7b/ Let-7b HL (+/−) (8)
H1	urine	healthy	18.8	9.5	6.0	11,368	9675	0.80	R = 3.5, 4.8, 9.0 (increase ×2)	SIL	>0.80 *
“	“	“	“	6.3	6.0	17,143	14,590	1.20	R = 2.4, 1.2, 1.2 (decrease ×2)	DET	<1.20 *	1.00 (0.20)
H2 1/3 dilution	“	“	27.4	6.0	6.0	18,000	10,499	0.86	R = 2.1, 3.9, 1.1 (increase ×1, decrease ×1)	Note1
“	“	“	“	4.0	6.0	27,000	15,749	1.30	R = 6.7, 4.3, 1.9 (decrease ×2)	DET	<1.30 *
“	“	“	“	4.0	6.0	27,000	15,749	1.30	R = 1.6, 1.5, 1.1 (decrease ×1)	DET	<1.30 *
“	“	“	“	8.5	6.0	12,706	7411	0.61	R = 4.1, 3.7, 5.4 (increase ×1)	SIL	>0.61 *	1.07 (0.37)
H6914 4th lot	serum comb. men	“	14.3	10.5	5.0	8571	9590	0.79	R = 2.2, 1.5, 2.3 (decrease ×1)	DET	<0.79 *
“	“	“	“	6.8	5.0	13,235	14,809	1.22	R = 3.5, 3.8, 2.4 (decrease ×1)	DET	<1.22
“	“	“	“	13.0	5.0	6923	7746	0.64	R = 1.0, 1.5, 1.0 (increase ×1)	SIL	>0.64 *	0.72 (0.08)
SR16-690	serum	PAN cancer	16.4	9.0	5.0	10,000	9756	0.80	R = 1.2, 1.7, 1.2 (increase ×1)	SIL	>0.80 *
“	“	“	“	6.0	5.0	15,000	14,634	1.20	R = 5.3, 2.0, 1.9 (decrease ×2)	DET	<1.20 *
“ B	“	“	14.3	6.5	5.0	13,846	15,492	1.28	R = 3.6, 1.9, 1.6 (decrease ×2)	DET	<1.28
“ B	“	“	“	13.0	5.0	6923	7746	0.64	R = 1.0, 1.6, 0.8 (increase ×1)	SIL	>0.64	1.00 (0.20)
SR17-248 B	“	“	21.0	4.5	5.0	20,000	15,238	1.25	R = 1.6, 0.9, 1.6 (decrease ×1)	DET	<1.25
“	“	“	“	7.5	5.0	12,000	9143	0.75	R = 1.1, 0.6, 0.7 (decrease ×2)	DET	<0.75 *
“	“	“	“	9.5	5.0	9474	7218	0.59	R = 1.2, 2.3, 2.2 (increase ×2)	SIL	>0.59 *	0.67 (0.08)
SR23-6022	urine	PRO cancer	14.5	4.0	5.0	22,500	24,828	2.04	R = 3.7, 3.0, 1.0 (decrease ×2)	DET	<2.04 *
“	“	“	“	5.0	4.0	14,400	15,890	1.31	severely reduced events	SIL	>1.31
“	“	“	“	8.0	4.0	9000	9931	0.82	“	SIL	>0.82
“	“	“	“	4.0	4.0	18,000	19,862	1.63	“	SIL	>1.63 *	1.82 (0.20)
“	“	“	“	5.3	5.0	16,981	18,738	1.54	R = 1.6, 1.8, 1.3 (comparable)	SIL	Note1
SR23-6028	“	“	12.4	10.0	4.0	7200	9290	0.76	R = 0.7, 0.9, 1.7 (increase ×1)	SIL	>0.76
“	“	“	“	6.0	4.0	12,000	15,484	1.27	R = 1.9, 1.5, 2.7 (increase ×1)	SIL	>1.27 *
“	“	“	“	4.7	4.0	15,319	19,767	1.63	R = 2.7, 1.7, 2.0 (decrease ×2)	DET	<1.63 *	1.45 (0.18)
SR23-6016 1/4 dilution	“	BRE cancer	43.7	4.0	6.0	27,000	9886	0.81	R = 1.5, 2.9, 2.3 (increase ×2)	SIL	>0.81
“	“	“	“	2.7	6.0	40,000	14,645	1.21	R = 2.0, 2.3, 2.4	Note1
1/8 dilution	“	“	21.8	4.0	7.5	33,750	24,771	2.04	R = 6.6, 2.2, 3.6 decrease ×2)	DET	<2.04 *
“	“	“	“	4.0	6.0	27,000	19,817	1.63	R = 1.7, -, 3.2 (increase ×1)	SIL	>1.63 *	1.77 (0.28)
“	“	“	“	5.5	7.5	24,545	18,015	1.48	R = 1.1, 1.9, 3.2 (increase ×2)	SIL	>1.48
101499	serum match	“	17.1	5.5	6.0	19,636	18,373	1.51	R = 3.8, 3.9, 4.3 (increase ×1)	SIL	>1.51 *
“	“	“	“	4.0	6.0	27,000	25,263	2.08	R = 1.5, 0.8, 0.9 (decrease ×2)	DET	<2.08 *	1.80 (0.28)

(1) The logistics of the subjects are listed in Table 5. B at the sample ID represents the 2nd total RNA isolation. Samples with RNA concentrations > 35 ng/μL were diluted with Ambion water as shown. (2) PAN, PRO and BRE represent pancreatic, prostate and breast cancer, respectively. (3) Aliquots of total RNA and probe used for the ion conductance nanopore experiment. Probe for let-7b at 30 fM tagged with an average of 5.5 OsBp moieties (Table 6 in Section 2.1). (4) Normalization to the same RNA content (16 ng/μL H6914, 1st lot). Probe copies P were calculated from P = 600 × probe (μL) × 30 (fM). (5) The content-normalized probe copies were divided by the let-7b copy number (12,150 copies, H6914, 1st lot) to obtain x-fold healthy level (HL); see also under (8). (6) The assignment of an experiment as detection (DET) or silencing (SIL) is based on the R-factor, which is the ratio of the event counts of late (Ir/Io)max to early (Ir/Io)max. One test using this technology comprised three nanopore experiments (45 min each at −180 mV). The first experiment was the baseline or control experiment and was typically an experiment with a buffer only. The mixture sample (RNA + probe) was then loaded onto the flow cell and run twice on the same flow cell under the above conditions. The results of these three experiments were compared, and R for each was determined in the order of control, 1st run and 2nd run. (7) SIL and DET stand for silencing and detection, respectively. The R values of the 1st and 2nd runs were compared to the control and if at least one of them decreased, then the experimental result was “detection”. If at least one of them increased, then the experimental result was “silencing”. If one of the runs was an increase and the other a decrease, the experiment is considered currently inconclusive (Note1), even though preliminary evidence suggests that this experiment directly yielded the miRNA copy number. The decreasing or increasing of R must be statistically significant. (8) The two experiments with an asterisk (*) were used to determine the let-7b copy number, and the other entries are confirmatory. For a specific experiment with an x μL probe and y μL of sample RNA, probe molecules P = x μL (probe concentration in fM) × 600. If the experiment involved detection, then P > target miRNA molecules were within the y μL aliquot. If the experiment involved silencing, then P < target miRNA molecules were in y μL. miRNA copies per 1 μL of isolated RNA sample < or > P/y, depending on the experimental outcome.

Table 3. MinION experiments targeting let-7b in healthy and cancer samples to illustrate validation strategy of let-7b as biomarker for prostate and breast, but not as biomarker for pancreatic cancer.

Subject ID (1)	Biospecimen	Condition (2)	Isolated Total RNA (ng/μL)	RNA (μL) (3)	Probe Let-7b 30 fM (μL) (3)	Probe Copies per 1 μL of RNA	Normalized Probe Copies (4)	Normalized Probe Copies/12,150 (5)	Ratio of Late to Early (Ir/Io)max (6)	Result (7)	Let-7b/ Let-7b HL (8)	Let-7b/ Let-7b HL (+/−) (8)
H1	urine	healthy	18.8	9.5	6.0	11,368	9675	0.80	R = 3.5, 4.8, 9.0 (increase ×2)	SIL	>0.80 *
“	“	“	“	6.3	6.0	17,143	14,590	1.20	R = 2.4, 1.2, 1.2 (decrease ×2)	DET	<1.20 *	1.00 (0.20)
H2 1/3 dilution	“	“	27.4	6.0	6.0	18,000	10,499	0.86	R = 2.1, 3.9, 1.1 (increase ×1, decrease ×1)	Note1
“	“	“	“	4.0	6.0	27,000	15,749	1.30	R = 6.7, 4.3, 1.9 (decrease ×2)	DET	<1.30 *
“	“	“	“	4.0	6.0	27,000	15,749	1.30	R = 1.6, 1.5, 1.1 (decrease ×1)	DET	<1.30 *
“	“	“	“	8.5	6.0	12,706	7411	0.61	R = 4.1, 3.7, 5.4 (increase ×1)	SIL	>0.61 *	1.07 (0.37)
H6914 4th lot	serum comb. men	“	14.3	10.5	5.0	8571	9590	0.79	R = 2.2, 1.5, 2.3 (decrease ×1)	DET	<0.79 *
“	“	“	“	6.8	5.0	13,235	14,809	1.22	R = 3.5, 3.8, 2.4 (decrease ×1)	DET	<1.22
“	“	“	“	13.0	5.0	6923	7746	0.64	R = 1.0, 1.5, 1.0 (increase ×1)	SIL	>0.64 *	0.72 (0.08)
SR16-690	serum	PAN cancer	16.4	9.0	5.0	10,000	9756	0.80	R = 1.2, 1.7, 1.2 (increase ×1)	SIL	>0.80 *
“	“	“	“	6.0	5.0	15,000	14,634	1.20	R = 5.3, 2.0, 1.9 (decrease ×2)	DET	<1.20 *
“ B	“	“	14.3	6.5	5.0	13,846	15,492	1.28	R = 3.6, 1.9, 1.6 (decrease ×2)	DET	<1.28
“ B	“	“	“	13.0	5.0	6923	7746	0.64	R = 1.0, 1.6, 0.8 (increase ×1)	SIL	>0.64	1.00 (0.20)
SR17-248 B	“	“	21.0	4.5	5.0	20,000	15,238	1.25	R = 1.6, 0.9, 1.6 (decrease ×1)	DET	<1.25
“	“	“	“	7.5	5.0	12,000	9143	0.75	R = 1.1, 0.6, 0.7 (decrease ×2)	DET	<0.75 *
“	“	“	“	9.5	5.0	9474	7218	0.59	R = 1.2, 2.3, 2.2 (increase ×2)	SIL	>0.59 *	0.67 (0.08)
SR23-6022	urine	PRO cancer	14.5	4.0	5.0	22,500	24,828	2.04	R = 3.7, 3.0, 1.0 (decrease ×2)	DET	<2.04 *
“	“	“	“	5.0	4.0	14,400	15,890	1.31	severely reduced events	SIL	>1.31
“	“	“	“	8.0	4.0	9000	9931	0.82	“	SIL	>0.82
“	“	“	“	4.0	4.0	18,000	19,862	1.63	“	SIL	>1.63 *	1.82 (0.20)
“	“	“	“	5.3	5.0	16,981	18,738	1.54	R = 1.6, 1.8, 1.3 (comparable)	SIL	Note1
SR23-6028	“	“	12.4	10.0	4.0	7200	9290	0.76	R = 0.7, 0.9, 1.7 (increase ×1)	SIL	>0.76
“	“	“	“	6.0	4.0	12,000	15,484	1.27	R = 1.9, 1.5, 2.7 (increase ×1)	SIL	>1.27 *
“	“	“	“	4.7	4.0	15,319	19,767	1.63	R = 2.7, 1.7, 2.0 (decrease ×2)	DET	<1.63 *	1.45 (0.18)
SR23-6016 1/4 dilution	“	BRE cancer	43.7	4.0	6.0	27,000	9886	0.81	R = 1.5, 2.9, 2.3 (increase ×2)	SIL	>0.81
“	“	“	“	2.7	6.0	40,000	14,645	1.21	R = 2.0, 2.3, 2.4	Note1
1/8 dilution	“	“	21.8	4.0	7.5	33,750	24,771	2.04	R = 6.6, 2.2, 3.6 decrease ×2)	DET	<2.04 *
“	“	“	“	4.0	6.0	27,000	19,817	1.63	R = 1.7, -, 3.2 (increase ×1)	SIL	>1.63 *	1.77 (0.28)
“	“	“	“	5.5	7.5	24,545	18,015	1.48	R = 1.1, 1.9, 3.2 (increase ×2)	SIL	>1.48
101499	serum match	“	17.1	5.5	6.0	19,636	18,373	1.51	R = 3.8, 3.9, 4.3 (increase ×1)	SIL	>1.51 *
“	“	“	“	4.0	6.0	27,000	25,263	2.08	R = 1.5, 0.8, 0.9 (decrease ×2)	DET	<2.08 *	1.80 (0.28)

(1) The logistics of the subjects are listed in Table 5. B at the sample ID represents the 2nd total RNA isolation. Samples with RNA concentrations > 35 ng/μL were diluted with Ambion water as shown. (2) PAN, PRO and BRE represent pancreatic, prostate and breast cancer, respectively. (3) Aliquots of total RNA and probe used for the ion conductance nanopore experiment. Probe for let-7b at 30 fM tagged with an average of 5.5 OsBp moieties (Table 6 in Section 2.1). (4) Normalization to the same RNA content (16 ng/μL H6914, 1st lot). Probe copies P were calculated from P = 600 × probe (μL) × 30 (fM). (5) The content-normalized probe copies were divided by the let-7b copy number (12,150 copies, H6914, 1st lot) to obtain x-fold healthy level (HL); see also under (8). (6) The assignment of an experiment as detection (DET) or silencing (SIL) is based on the R-factor, which is the ratio of the event counts of late (Ir/Io)max to early (Ir/Io)max. One test using this technology comprised three nanopore experiments (45 min each at −180 mV). The first experiment was the baseline or control experiment and was typically an experiment with a buffer only. The mixture sample (RNA + probe) was then loaded onto the flow cell and run twice on the same flow cell under the above conditions. The results of these three experiments were compared, and R for each was determined in the order of control, 1st run and 2nd run. (7) SIL and DET stand for silencing and detection, respectively. The R values of the 1st and 2nd runs were compared to the control and if at least one of them decreased, then the experimental result was “detection”. If at least one of them increased, then the experimental result was “silencing”. If one of the runs was an increase and the other a decrease, the experiment is considered currently inconclusive (Note1), even though preliminary evidence suggests that this experiment directly yielded the miRNA copy number. The decreasing or increasing of R must be statistically significant. (8) The two experiments with an asterisk (*) were used to determine the let-7b copy number, and the other entries are confirmatory. For a specific experiment with an x μL probe and y μL of sample RNA, probe molecules P = x μL (probe concentration in fM) × 600. If the experiment involved detection, then P > target miRNA molecules were within the y μL aliquot. If the experiment involved silencing, then P < target miRNA molecules were in y μL. miRNA copies per 1 μL of isolated RNA sample < or > P/y, depending on the experimental outcome.

Table 4. miRNA copies (HL units, Figure 2) from experiments normalized as described in Table 3.

ID (1)	Indication	Isolated Total RNA, ng/μL (2)	miRNA Targets in HL Units (3)
			miR-15b	miR-21	miR-375	miR-141	miR-375 + miR-141
Cancer serum
CAN7	breast	6.9	0.79	1.60
CAN9	“	9.1	0.89	1.79		1.80
CAN4	prostate	12.0	0.88	1.79		1.81
CAN6	“	8.0	0.90	1.87		1.84
SR16-690	pancreatic	16.4	1.01	1.63
				1.88
SR17-248	pancreatic	14.4	1.00			1.88
Cancer urine
SR23 6016	breast	174.6	1.34	1.75		1.76	1.69
SR23 6017	“	88.8	1.12	2.13		1.73	1.66
						2.20
SR23 6018	breast	16.1				1.72
						2.25
SR23 6022	prostate	15.3			1.81	1.80
SR23 6028	“	13.3			1.81	1.82
SR23 6023	“	18.5			1.82	1.63
						2.00
SR23 6033	pancreatic	13.4			1.82	1.82
Healthy urine
		7.5, 22.9	0.97		1.04
		21.7			0.97
		9.5	1.00
H2		16.8, 82.3	0.84	0.90		0.83	0.84
				1.19
		8.7		1.01
		16.5					0.89
		12.2	1.00
		57.4	0.98				0.80
		163.0	0.84			1.02
		11.7					0.96
		25.6		1.03			0.87
		14.5					0.92
		16.4		1.17			1.16
		13.8	0.79	1.17			1.06
		12.4,15.7	1.02	1.42			1.30
		14.7		1.30			1.31

(1) The logistics of the subjects are listed in Table 5. Logistics for the H2 sample are in the caption of Table 2. (2) Multiple entries of RNA isolated from healthy samples corresponded to timely separate collections. When the amount of isolated RNA exceeded 35 ng/μL, it was diluted with Ambion water before mixing with the probe. (3) Cancer serum samples (second-, third- and fourth-row entries) were also used in an earlier study [37]; however, only miR-15b was measured earlier. The same RNA isolate was used for the additional miRNAs measured here. As described in Table 3, measured probe copies were normalized to the same RNA content and then divided by the corresponding miRNA copy number using as reference H6914 1st lot (Table 2). This yielded HL levels, from which a copy number was obtained and is listed in columns 4 through 8 as x-fold HL. The last column reports the results of a single experiment in which two miRNAs were simultaneously targeted (see Section 2.8). Experiments yielding silencing or detection, not followed by the corresponding test to determine miRNA copy number, are not included here, but were confirmatory.

Table 4. miRNA copies (HL units, Figure 2) from experiments normalized as described in Table 3.

ID (1)	Indication	Isolated Total RNA, ng/μL (2)	miRNA Targets in HL Units (3)
			miR-15b	miR-21	miR-375	miR-141	miR-375 + miR-141
Cancer serum
CAN7	breast	6.9	0.79	1.60
CAN9	“	9.1	0.89	1.79		1.80
CAN4	prostate	12.0	0.88	1.79		1.81
CAN6	“	8.0	0.90	1.87		1.84
SR16-690	pancreatic	16.4	1.01	1.63
				1.88
SR17-248	pancreatic	14.4	1.00			1.88
Cancer urine
SR23 6016	breast	174.6	1.34	1.75		1.76	1.69
SR23 6017	“	88.8	1.12	2.13		1.73	1.66
						2.20
SR23 6018	breast	16.1				1.72
						2.25
SR23 6022	prostate	15.3			1.81	1.80
SR23 6028	“	13.3			1.81	1.82
SR23 6023	“	18.5			1.82	1.63
						2.00
SR23 6033	pancreatic	13.4			1.82	1.82
Healthy urine
		7.5, 22.9	0.97		1.04
		21.7			0.97
		9.5	1.00
H2		16.8, 82.3	0.84	0.90		0.83	0.84
				1.19
		8.7		1.01
		16.5					0.89
		12.2	1.00
		57.4	0.98				0.80
		163.0	0.84			1.02
		11.7					0.96
		25.6		1.03			0.87
		14.5					0.92
		16.4		1.17			1.16
		13.8	0.79	1.17			1.06
		12.4,15.7	1.02	1.42			1.30
		14.7		1.30			1.31

(1) The logistics of the subjects are listed in Table 5. Logistics for the H2 sample are in the caption of Table 2. (2) Multiple entries of RNA isolated from healthy samples corresponded to timely separate collections. When the amount of isolated RNA exceeded 35 ng/μL, it was diluted with Ambion water before mixing with the probe. (3) Cancer serum samples (second-, third- and fourth-row entries) were also used in an earlier study [37]; however, only miR-15b was measured earlier. The same RNA isolate was used for the additional miRNAs measured here. As described in Table 3, measured probe copies were normalized to the same RNA content and then divided by the corresponding miRNA copy number using as reference H6914 1st lot (Table 2). This yielded HL levels, from which a copy number was obtained and is listed in columns 4 through 8 as x-fold HL. The last column reports the results of a single experiment in which two miRNAs were simultaneously targeted (see Section 2.8). Experiments yielding silencing or detection, not followed by the corresponding test to determine miRNA copy number, are not included here, but were confirmatory.

Table 5. Demographics of cancer patients whose samples are listed in Table 3 and Table 4.

Biobank	ID	Age Group	Gender	Cancer	T Stage	N Stage	Specimen
Tissue for Research, UK	101499	56–60	F	breast	-	-	serum matched
	SR23 6016	51–55	F	“	pT1b	pN0	urine
	SR23 6017	66–70	F	“	pT1b	“	“
	SR23 6018	51–55	F	“	pT1a	“	“
	SR23 6022	71–75	M	prostate	pT2	“	“
	SR23 6023	66–70	M	“	“	“	“
	SR23 6028	51–55	M	“	“	“	“
	SR23 6033	66–70	F	pancreatic	“	“	“
	SR16 690	51–55	M	“	pT2	“	serum
	SR17 248	51–55	M	“	pT1	“	“
Discovery Life Sciences, US	CAN4	66–70	M	prostate	newly diagnosed, pretreatment		serum
	CAN6	56–60	M	“			“
	CAN7	51–55	F	breast			“
	CAN9	56–60	F	“			“

Table 5. Demographics of cancer patients whose samples are listed in Table 3 and Table 4.

Biobank	ID	Age Group	Gender	Cancer	T Stage	N Stage	Specimen
Tissue for Research, UK	101499	56–60	F	breast	-	-	serum matched
	SR23 6016	51–55	F	“	pT1b	pN0	urine
	SR23 6017	66–70	F	“	pT1b	“	“
	SR23 6018	51–55	F	“	pT1a	“	“
	SR23 6022	71–75	M	prostate	pT2	“	“
	SR23 6023	66–70	M	“	“	“	“
	SR23 6028	51–55	M	“	“	“	“
	SR23 6033	66–70	F	pancreatic	“	“	“
	SR16 690	51–55	M	“	pT2	“	serum
	SR17 248	51–55	M	“	pT1	“	“
Discovery Life Sciences, US	CAN4	66–70	M	prostate	newly diagnosed, pretreatment		serum
	CAN6	56–60	M	“			“
	CAN7	51–55	F	breast			“
	CAN9	56–60	F	“			“

Table 6. Sequence and characterization of the probes used in this work.

ID: DNA Oligo Sequence Used for Probe	In Sequence mU is 2′-OMeU and dU is 2′-deoxyU	Concen, fM (1)	# OsBp Average (2)
Probe 375T5	(A)5dUCACGCGAGCCGAACGAACAAAC(T)5C(A)5	42.0	5.1
Probe m21T5	(A)5mUCAACAmUCAGmUCmUGAmUAAGCmUA(T)5C(A)6	27.1	4.4
Probe m141T5	(A)4CCAmUC(mU)3ACCAGACAGmUG(mU)2A(T)5(A)5	33.5	4.7
Probe 15bT5	(A)6dUGdUAAACCAdUGAdUGdUGCdUGCdUAT5A6	35.0	5.9
Probe let7bT5	(A)6CCACACAACCmUACmUACCmUCA(T)5(A)5	30.0	5.5

(1) Concentration of probe solution (fM) used for the nanopore experiments. It was obtained by 5/1000 or 10/1000 dilutions from the stock solution (μM) of the probe prepared by osmylation (T-OsBp)₅ of the oligo and characterized in-house by HPLC. (2) # OsBp, the average number of osmium label moieties on the probe (extent of osmylation) was measured using the following equation: absorbance at 312 nm/absorbance at 272 nm or R(312/272) = 2x(no osmylated pyrimidines/total nt) [37]. R is the ratio of the corresponding HPLC peaks, regardless of their shape (sharp or broad). An extra osmium tag was conjugated to a C or U base within the sequence. A single internal tag did not prevent hybridization, as shown by nanopore experiments.

Table 6. Sequence and characterization of the probes used in this work.

ID: DNA Oligo Sequence Used for Probe	In Sequence mU is 2′-OMeU and dU is 2′-deoxyU	Concen, fM (1)	# OsBp Average (2)
Probe 375T5	(A)5dUCACGCGAGCCGAACGAACAAAC(T)5C(A)5	42.0	5.1
Probe m21T5	(A)5mUCAACAmUCAGmUCmUGAmUAAGCmUA(T)5C(A)6	27.1	4.4
Probe m141T5	(A)4CCAmUC(mU)3ACCAGACAGmUG(mU)2A(T)5(A)5	33.5	4.7
Probe 15bT5	(A)6dUGdUAAACCAdUGAdUGdUGCdUGCdUAT5A6	35.0	5.9
Probe let7bT5	(A)6CCACACAACCmUACmUACCmUCA(T)5(A)5	30.0	5.5

(1) Concentration of probe solution (fM) used for the nanopore experiments. It was obtained by 5/1000 or 10/1000 dilutions from the stock solution (μM) of the probe prepared by osmylation (T-OsBp)₅ of the oligo and characterized in-house by HPLC. (2) # OsBp, the average number of osmium label moieties on the probe (extent of osmylation) was measured using the following equation: absorbance at 312 nm/absorbance at 272 nm or R(312/272) = 2x(no osmylated pyrimidines/total nt) [37]. R is the ratio of the corresponding HPLC peaks, regardless of their shape (sharp or broad). An extra osmium tag was conjugated to a C or U base within the sequence. A single internal tag did not prevent hybridization, as shown by nanopore experiments.

2.2. Can Urine Replace Blood as a Biospecimen for miRNA Determination?

For privacy reasons and to circumvent a blood draw, urine was explored as the miRNA source, as miRNAs have been found to be relatively stable in urine [53]. To directly assess whether miRNAs exhibit comparable copy numbers in serum and urine, we initially purchased a set of 12 samples: matched serum and urine samples from the same donor, using two donors each for breast, prostate and pancreatic cancer. Preliminary data support the hypothesis that urine can replace blood. However, the 1 mL urine volume afforded 0.05 mL of total RNA at a concentration of approximately 7 ng/μL, which is at the lower limit of our technology, and was not sufficient to run the number of experiments required to reach solid conclusions. We did not pursue this type of matching serum and urine study further because of the striking equivalence between miRNA copy numbers determined from H6914 serum and urine samples from healthy subjects (see later). Urine may contain up to 50-fold less RNA than the serum. While 0.2 mL of serum provides 0.1 mL of isolated total RNA, sufficient for multiple miRNA determinations, a much larger urine volume is necessary. A recently developed slurry kit from Norgen Biotek enabled the isolation of a 0.05 mL sample of total RNA from 5 to 10 mL of urine. miR-16 was found to be 12-fold more abundant in serum than miR-15b [20,37]. Attempts to measure miR-16 in urine failed, suggesting that miR-16 may be underexpressed in urine compared to serum. The other five miRNAs were measured in both the serum and urine samples (see Table 2, Table 3 and Table 4). Notably miR-141 has sequence similarity and belongs to the miR-200 family with members proposed as cancer biomarkers [54]. Whether the probe used here for miR-141 (see Table 6) also detects one or more of the members of the miR-200 family requires further testing.

A groundbreaking discovery was made when miRNA copies in the serum and urine (Urine1) of a healthy woman (H2) were found to be equivalent to the corresponding miRNA copies in the H6914 serum (Table 1). A second urine sample from H2 (Urine2) confirmed these findings after normalization to the H6914 RNA content (see caption of Table 1). Notably, the three samples from H2 were collected months apart and had distinct RNA content. To the best of our knowledge, this is the first demonstration of miRNA copy number equivalence between healthy serum (men, combined) and healthy urine (woman) collected at different times.

2.3. Biospecimen Used in This Study

The samples tested in this study included H6914 (3rd and 4th lots), serum and urine samples from cancer patients and urine samples from healthy subjects. The miRNA/urine study was approved by the Advarra IRB (see Section 3.1). The instructions for urine collection were identical for both healthy and diseased women and men. No formal follow-up was planned for the healthy group. Breast, prostate and pancreatic cancer samples were purchased from two blood banks, Discovery Life Sciences and Tissue for Research, with subject requirements for early-stage diagnosis (I or II) and before treatment, because miRNA levels may be influenced by disease stage and therapy. The selection of an early disease stage ahead of treatment is consistent with our objective of providing a validation strategy for miRNA biomarkers and developing a cancer screening test for asymptomatic individuals. The cancer samples were selected to be as inclusive as possible with respect to age, sex and ethnicity (see Table 5 in Section 2.1). Healthy urine samples were obtained from subjects who varied in sex and ethnicity and ranged in age from 30 to 75 years.

2.4. Validation Strategy for an miRNA Cancer Biomarker

The validation strategy was illustrated by targeting let-7b in three cancer indications. For simplicity, only a small number of samples were tested here, even though validation should include at least 10 samples each from healthy subjects and subjects diagnosed with a certain disease ahead of treatment, as mentioned earlier. The first three samples in Table 3 were from healthy subjects, while the fourth and fifth samples were serum samples from patients with pancreatic cancer. Within the accuracy of this technology, these five samples provided statistically indistinguishable let-7b copy numbers. To simplify the comparison, copy numbers were normalized to the H6914 1st lot RNA content of 16.0 ng/μL (reference), and then divided by the let-7b copy number (12,150) from this reference sample to give the HL number posted in the last column of Table 3. These five numbers (1.00, 1.07, 0.72, 1.00 and 0.67) give an average of 0.89 HL with RSD = 0.21. The tentative conclusion is that let-7b is not a pancreatic cancer biomarker that would be useful in discriminating pancreatic cancer from breast and prostate cancer. This finding should be confirmed using additional samples. The sixth and seventh samples were urine samples from patients diagnosed with prostate cancer and measured let-7b copy numbers at 1.82 and 1.45 HL (HL from H6914 1st lot). Compared to this study’s control at 0.89 HL, the sixth sample measured 2.0-fold higher, and the seventh sample 1.6-fold higher; that is, both were overexpressed by more than 1.5-fold (see Table 1A). The eighth and ninth samples were obtained from breast cancer patients, one from a urine sample, and the other from a serum sample with let-7b copy numbers measuring 1.77 HL and 1.80 HL. Both measurements were 2.0-fold higher than those of the control, 0.89 HL. These data suggest that 1.5 may serve as a threshold, whereby an miRNA level above 1.5 HL suggests cancer detection, and an miRNA level below 1.5 HL indicates the absence of cancer. Additional experiments were conducted with additional miRNAs, where a 1.5-fold threshold was applicable.

2.5. The miRNA Copy Number Is Proportional to the Total RNA Isolated from the Biospecimen

The number of miR-15b copies in the serum of H6914 and in the sera of healthy individuals and cancer patients was found to be directly proportional to the RNA content in the range of 9.1 to 20.7 ng/μL [37]. This observation was reported earlier ([37] (in 8 samples, 4 healthy and 4 diseased), confirmed here and extended by including serum and urine data (13 new samples, 8 healthy and 5 diseased) in the range of 6.9 to 174.6 ng/μL RNA (Table 4). These data suggest that miR-15b is not a cancer biomarker, in agreement with previous findings [20]. The observed independence in terms of age, sex or ethnicity is in bold contrast to studies that report a large data variation and attribute it to age, sex, ethnicity and other parameters. Our data irrevocably established that miRNA copies must be normalized to the same RNA content, which is currently not common practice.

Normalization to the same RNA content (16.0 ng/μL in H6914, 1st lot) for all tested miRNAs yielded copy numbers independent of age, sex, ethnicity and cancer indication (breast, prostate or pancreas), as well as biospecimen, suggesting that a urine sample may replace a blood draw. Further normalization, that is, dividing the miRNA copies of a sample by the corresponding miRNA copies from H6914 (1st lot), yielded two groups with zero overlap, one averaging 1.01 HL with RSD = 0.16 (40 counts: all miRNAs from healthy samples + miR-15b from cancer samples) and another averaging 1.83 HL with RSD = 0.09 (28 counts: miR-21, miR-375 and miR-141 from cancer samples) (Table 4 and Figure 2). The data yielded 100% sensitivity and 100% specificity; however, the sample size was small, and whether healthy subjects would remain free of cancer and for how long was not assessed. Notably, a p-value of 1.6 × 10⁻²² was determined by Excel’s t-test for the combined three cancer biomarkers in the healthy vs. the cancer group (sample size 52, see caption of Figure 2). For comparison, p-values of approximately 0.001 were used for miR-141 measurements by ddPCR, which is currently considered the most accurate method [27,32]. Despite the small study size, the unprecedented discrimination observed between healthy samples and samples from patients with breast, prostate and pancreatic cancer validates each of these three miRNAs as biomarkers for cancer. Overexpression of miR-21 [25] and miR-141 [20] has been associated with numerous cancer indications, in addition to breast, prostate and pancreatic cancers. Further testing of this set of miRNAs in samples from additional cancer indications should illustrate their usefulness as multi-cancer biomarkers. As long as a biomarker, such as miRNA, is elevated by 80% or more between diseased and healthy samples, our protocol-defined ± 20% accuracy in miRNA copy number determination is compatible with a 1.5 HL threshold to assign an unknown sample as healthy or cancerous (Table 1A and Figure 2). The data in Table 4, in conjunction with miRNA studies conducted worldwide during the last 25 years, suggest that elevated levels of miR-21, miR-375 and/or miR-141 in the serum or urine warrant consultation with a doctor, like a high fever would.

2.6. Bracketing the miRNA Copy Number Leads to High Accuracy

The nanopore technology used here for trace nucleic acid detection and quantification uses a bracketing approach for measurement, making it unique; it has no similarities to the currently used assays. The assay was developed and optimized earlier [36,37,46]. To the best of our knowledge, no other analytical assay yields measurements with a protocol-defined accuracy of ± 20%. When each measurement is ± 20% or better, a series of comparable measurements will exhibit an RSD < 0.2. Each process involved in this assay is outlined here. Total RNA (0.1 mL) was isolated from the serum (0.2 mL) using a Monarch kit from NEB, and total RNA (0.05 mL) was isolated from the urine (5 mL) using a Norgen slurry kit. The total RNA (ng/μL) was measured using a DS-11 DeNovix spectrophotometer. For accuracy reasons, isolated total RNA should contain more than 7 ng/μL RNA and have an absorbance ratio at 260 nm vs. 280 nm, A260/A280, better than 1.6. Capillary electrophoresis (CE) analysis, as described earlier [37], confirmed the dramatic variation observed in the total RNA isolated from individuals.

Samples for nanopore experiments were prepared by mixing a few μL of the isolated total RNA with a few μL of the probe complementary to the target miRNA. The mixture was stored at −20 °C overnight to ensure complete hybridization, and 75 μL of filtered ONT buffer was added to this mixture immediately before loading it onto the MinION flow cell. The nanopore experiment was conducted for 45 min at −180 mV. The flow cell was allowed to rest for 15 min and a second run was conducted under the same conditions. MINKNOW software ran the experiments and produced a fast5 file (the ion current (i) with time (t)), which was subsequently analyzed using OsBp-detect [51]. The latter yielded a tsv file that was opened in Excel. The data were grouped in the form of a histogram with 0.05 bins (see Figure 3 and Figure 4). An experiment using a buffer only instead of a sample primed the flow cell and served as a control for the experiment with the RNA/probe sample. If the test with the mixture of RNA and probe was determined to be silencing, then the next test may be designed using the same aliquot of probe but only 67% of the RNA aliquot. If the second test was determined to be a detection experiment, then miRNA copies per μL of RNA sample were determined from the average of the probe copies per μL of RNA sample from the two test samples. Typically, more than two tests were necessary before finding the set, one detection and the other silencing, which fulfilled the accuracy requirement. Owing to the low throughput of the assay and the approximately 15 h life span of the flow cells, all five miRNAs were not measured in every sample. This platform may be further optimized by developing a buffer that is optimal for ion conductance experiments to replace the currently used buffer.

2.7. Protein-Based vs. Solid-State Nanopores

A solid-state nanopore array is a more robust and cost-efficient alternative to the proteinic nanopores. Our technology can be implemented in other nanopore arrays with no or minor optimization, as shown using α-hemolysin [45] and silicon nitride nanopores [44]. The only requirement for this assay is that the width of the nanopore permits the translocation of ss nucleic acids and prevents the translocation of ds nucleic acids. Interestingly, the “bulkier” probe traverses nanopores of the same width as the unlabeled (intact) nucleic acids. We attribute this observation to the osmium tags extending parallel to the strand axis and not perpendicular to it. This configuration reduces the number of water molecules carried by the nucleic acids. For thermodynamic reasons, a naked ss nucleic acid (with no water solvation) is unlikely to exist in water; applying a voltage cannot eliminate the first solvation/hydration shell. The decreased hydration of the probe is envisioned as decreased “lubrication” during the sliding of the probe through the nanopore wall and may rationalize the dramatically slower translocation.

2.8. Potential, Limitations, Biomarker Multiplexing and Implementation of a Multi-Cancer Screening Test

This technology is not limited to the five miRNAs measured here. The probe is a DNA oligo complementary to the target sequence. Osmium tagging is straightforward, and the resulting probe is stable and characterizable (see Section 3.3 and Section 3.4). The design of the probe is general and has been optimized for efficient hybridization with a DNA or RNA target, and for nanopore detection. Probes for a limited number of human miRNAs have been manufactured and tested (Table 6 and [37]), and miRNAs from other species can also be targeted. However, this technology is not limited to miRNA quantification. Circulating tumor DNA (ctDNA), circular nucleic acids and practically any ss nucleic acid of a known partial sequence of interest can be detected and quantified. Liquid biopsies are non-invasive; however, this assay is not limited to liquid biopsies. Because of the availability of commercial kits for total RNA isolation (including miRNAs) from practically any tissue or organ, the latter can be used for ss nucleic acid detection and quantification.

Table 3 lists the results of the experiments in which two miRNAs, miR-375 and miR-141, were simultaneously targeted using the two corresponding probes. Targeting two miRNAs in one experiment yielded one copy number for both miRNAs, which reduced the number of experiments two-fold and could still serve as a screening test using an appropriate threshold value, as outlined above. Figure 2 illustrates the successful use of combined miRNAs to discriminate between cancerous and healthy samples. This approach can only be used when two targeted miRNAs exhibit similar copy numbers, as is the case for miR-375 and miR-141. This limitation is due to the inability of MinION nanopores to discriminate one probe from another with the current probe design (Table 6). Earlier work with α-hemolysin [45] and solid-state nanopores [44] illustrated that more osmium tags yielded deeper translocations with an earlier (I_r/I_o)_max. Adding another 2–3 osmium tags to the current 5 tags may yield the desired discrimination and enable individual miRNA testing in a single test targeting two miRNAs.

Implementation of a multi-cancer screening assay will not require extensive testing, as described in Table 3, which is suitable for miRNA validation studies. Instead of determining an miRNA copy number, the concept of the threshold value may be implemented as follows: considering an miRNA biomarker, like the ones studied here, where the HL level is close to 1.0 and the cancer level is close to 1.8, both at ~0.2 RSD, an experiment designed to target an miRNA level at the ~1.5 HL threshold (Yenos test) should yield detection with healthy samples and silencing with cancer samples or samples from asymptomatic individuals with elevated miRNA (Table 1A, Table 3 and Table 4 and Figure 2). We propose that elevated miRNA levels in two out of three tested miRNA cancer biomarkers indicate miRNA dysregulation that may be associated with the onset and/or presence of cancer and represent a warning sign for the tested individual. The number of false-positive and false-negative results from such a multi-cancer test targeting the ~1.5 HL should be practically nonexistent, as shown in Table 1A, Table 3 and Table 4 and Figure 2. Each single miRNA test included a control/baseline experiment and two runs using the same sample (see Figure 3 and Figure 4). A second test for an additional miRNA included one control and a second sample run twice. Six separate experiments were performed for two miRNA-related cancer biomarkers at the ~1.5 HL threshold, and the results collectively led to one conclusion, namely, whether the biomarkers were detected or silenced. We exploited this approach by testing consented individuals and showed that it works. One consented individual was a breast cancer survivor, and her miRNA levels, tested a few years later, were comparable to those of the healthy controls, suggesting that the miRNA levels recovered. Another individual who tested positive for cancer in the Yenos urine test underwent the Galleri test, which also exhibited cancer detection, thus confirming the Yenos test. Notably, the Galleri test evaluates DNA methylation, whereas the Yenos test evaluates miRNA overexpression, making them scientifically and technically independent. All four miRNAs tested exhibited comparable overexpression in the cancer samples (1.5 HL). Other miRNAs may exhibit a comparable or different overexpression, and if overexpression is at least 1.8-fold, our technology with a protocol-defined RSD ~0.2 will discriminate healthy from diseased samples.

3. Materials and Methods

3.1. Human Samples

Human sera from the USA, isolated via sterile filtration from male AB-clotted whole blood (H6914, 1st lot SLCH8785, 2nd lot SLCJ3635; data reported earlier; 3rd lot SLCL6534 and 4th lot SLCN9213; data reported here in Table 2) were purchased from Sigma-Aldrich over a period of 3 years. Serum samples purchased from Discovery Life Sciences (DLS, Huntsville, AL, USA) and Tissue for Research (Accio Biobank online, Suffolk, UK) were collected from informed consented individuals under the IRB/EC protocol. The selection of these samples from a large depository included both male and female donors, if applicable, and one each from African American, Hispanic and White ethnicity. Samples were collected from newly diagnosed, naïve and pretreated patients. The demographic information of the patients with cancer who provided their specimens is listed in Table 5. The proposal to include urine samples collected from consenting healthy subjects was reviewed by the Advarra Investigational Review Board (IRB). The protocol and consent form were reviewed, modified and approved by the Advarra IRB on November 15, 2023. Protocol: Yenos Analytical LLC-02. Quantification of selected microRNAs in the urine of healthy individuals (Pro00074065). Donors of urine samples reviewed and signed an informed consent form. They were then sent a kit/insulated box, cold bricks, and instructions to collect their biospecimen at home, freeze it and ship it overnight to the Yenos facilities. The healthy urine donors were 30 to 75 years old, female or male and of different ethnicities. For the isolation of total RNA from serum, we used the Monarch T2010S Kit (1st lot 10075450, 2nd lot 10141109 reported earlier and 3rd lot 10144556 used here). For the isolation of total RNA from urine, two kits were used: No. 29000 was used for 1 mL urine samples, and the slurry kit No. 29600 was used for 5 to 10 mL urine samples. All kits were used according to the manufacturer’s instructions.

3.2. Oligos, Probes and Other Reagents

The only ONT kit used for the experiments reported here was the Flow Cell Priming Kit XL (EXP-FLP002-XL), ONT flush buffer or ONT buffer. The ONT buffer is proprietary, provides the necessary electrolytes and must represent more than 80% of the approximate 80 μL sample volume. Custom-made DNA oligos and 2′-OMe-oligos synthesized at the 0.2 μM scale and purified by HPLC or PAGE by the manufacturer were purchased from Integrated DNA Technologies (IDT) and Millipore/Sigma-Aldrich, respectively. Oligos (sequences in Table 6) were diluted with Ambion nuclease-free water, untreated with DEPC, typically to 100 or 200 μM stock solutions, and stored at −20 °C. The oligo purity was confirmed to be >85% by in-house HPLC analysis [55]. Following osmium tagging (osmylation, see below), in-house HPLC analysis was used to determine the probe content, extent of osmylation and efficiency of probe/target hybridization [37]. Osmylated oligos were manufactured in-house, based on published methods at a concentration of approximately 30 μM. LoBind Eppendorf test tubes (1.5 mL) were used for serial 5/1000 or 10/1000 dilutions to yield probes at concentrations of 15–30 fM. Mixtures of a probe with isolated RNA were prepared in 0.5 mL RNase- and DNase-free sterile test tubes and stored at −20 °C overnight.

3.3. Osmylation of Nucleic Acids

The osmylation of nucleic acids using a 1:1 mixture of OsO₄ and 2,2′-bipyridine, abbreviated OsBp, was discovered 60 years ago [56], has been used extensively [57,58,59] and was optimized by us [60,61]. The detailed protocols for the synthesis, purification and quality control assays have been previously described [36,37]. OsO₄ is a hazardous material, and care must be taken for its use, storage and disposal [62]. Osmylation reactions require a 20-fold excess of OsBp over the reactive pyrimidine in monomer equivalents to ensure pseudo-first-order kinetics and to yield preferential labeling of thymidines (Ts) over the other pyrimidines. The osmylation reaction was quenched upon purification. Purification from excess OsBp (twice) was performed using spin columns (TC-100 FC from TrimGen Corporation) for 4 min at 5000 rpm. The flow-through solution is a probe that is chemically stable and can be stored at −20 °C for two years.

3.4. The Development, Optimization and Validation of Probes

The development, optimization and validation of probes accompanied by enhanced MinION detection have been previously reported [37]. The optimized probe comprises a sequence complementary to its target but extended at one end with four to five adjacent T residues and flanked by up to five adenosines (As) at either end (Table 6). This facilitated the entry of the probe into the nanopores. The adjacent Ts were tagged with an osmium label for quantitative detection. Within the probe sequence complementary to the target, Ts were replaced by uridine (U), 2′-OMe-U or dU, to minimize OsBp labeling because the osmylation kinetics of U and cytosine (C) are substantially slower than that of T [60]. HPLC analysis yields the probe concentration (content) using intact oligo as a standard because the absorbance of the probe at 260 nm is practically the same as that of the precursor intact oligo [60]. HPLC analysis provided evidence of the quantitative depletion of the OsBp reagent. Alternatively, a suitable spectrophotometer can be used to determine the content and extent of osmylation.

3.5. Single-Molecule Ion-Channel Conductance Experiments on the MinION (MinION Mk1B Platform)

One must register with the ONT and download the software MINKNOW to a computer/laptop with specifications provided by ONT. This platform consists of a device and the flow cell that fits within the device. The flow cell includes 2024 nanopores but not all 2024 are monitored by a detector. There are only 512 detection channels and, per the manufacturer’s description, when one nanopore becomes deactivated, the specific detection channel moves to the next nanopore. Per the manufacturer, there is one detection channel for every four nanopores. There were 512 signals recorded and reported in a single experiment. The fast5 file that is produced by the platform at the end of each experiment contains the ion current as a function of the time signal for each one of the 512 channels. All the functions necessary to test the hardware and flow cells and run the experiments were performed using MINKNOW software. The sample was loaded onto a flow cell that fitted within the MinION device. The experiment was run under the “start sequencing” mode. A direct RNA sequencing kit (SQK-RNA002) was used to initiate experiments. The flow cell type FLO-MIN106 was selected, and the run length (45 min) and bias voltage (−180 mV) were selected; basecalling was disabled, and the raw output bulk file (1–512) was checked in order to be generated. The output location was Library/MinKNOW/data/, and the output format was fast-5. All the experiments reported here were run for 45 min at −180 mV. The fast-5 file was analyzed using the OsBp_detect algorithm [51]. The number of events per channel from the OsBp_detect analysis was compared with the actual i-t trace of the specific channel using MATLAB visualization, and this algorithm, 2nd revision, was repeatedly validated. Currently, OsBp_detect can only be used with a 2017 or earlier version of MacBook Pro loaded with macOS 10.14 Mojave. Future work will include the adaptation of OsBp_detect to newer operating systems. While alternative parameters were explored, all experiments reported here were analyzed using the following threshold parameters: (i) event duration (in tps): 4–1200 (1.3–400 ms), (ii) lowest I_r/I_o <0.55, and (iii) all I_r/I_o <0.6, channels 1–512.

3.6. Data Analysis

A state-of-the-art laptop/computer requires approximately 5 min for OsBp_detect analysis, produces a file in tsv format, opens via Microsoft Excel and saves it as such. In the Excel spreadsheet, the algorithm-selected events (I_r/I_o data) are grouped in the form of a histogram with 0.05 bins, from 0.05 to 0.55, and plotted (Figure 3 and Figure 4). These events were added together and identified as total events. Typical histograms exhibit two maxima (I_r/I_o)_max: an early one at I_r/I_o = 0.15 and a late one at I_r/I_o = 0.30. These maxima may vary by ±0.05 units depending on the flow cell age. The events under late(I_r/I_o)_max and early(I_r/I_o)_max were noted, and their ratio R = (late(I_r/I_o)_max) / (early(I_r/I_o)_max) was calculated. These values (total events and ratio (R)) represent the criteria by which an experiment is judged as detection or silencing compared to the buffer control. Fewer total events than those in the control suggested silencing, whereas more total events suggested detection. A decreasing ratio R indicates detection, and an increasing ratio R indicates silencing. This assignment is consistent with an increased number of events owing to the presence of the probe, which traverses with (I_r/I_o)_max ~ 0.15, whereas intact RNA and background noise traverse mostly with (I_r/I_o)_max ~ 0.30. Each sample was run twice, and each run was compared to the buffer/control. Because the flow cells lost active pores during every experiment, the total number of events decreased during every experiment. The presence of the hybrid had an additional effect on reducing the number of events. This is because the hybrids are driven toward the nanopores, cannot pass through, and are pushed back by the alternating voltage of the platform. However, they remain in proximity and prevent other molecules from traversing the nanopores, that is shielding the nanopores.

4. Conclusions

This report presents the implementation of a novel analytical platform for the detection, quantification and validation of miRNA cancer/disease biomarkers from liquid biopsies with an accuracy of ±20%. This technology combines single-molecule ion conductance experiments using MinION functionality with an expertly optimized probe design. A general validation strategy applicable to any potential ss nucleic acid biomarker is presented. The copy numbers of five miRNAs, let-7b, miR-15b, miR-21, miR-375 and miR-141, were measured in the healthy and cancerous samples. Normalization of the copy number to the same RNA content was found to be critically important. miRNA copies from the combined serum of healthy men (H6914) were found to be equivalent to the miRNA copies measured in urine samples of healthy subjects—men and women—clearly illustrating the equivalence between serum and urine for the tested miRNAs. In contrast to miR-15b, which appears to be unrelated to breast, prostate and pancreatic cancers, and let-7b, which is not overexpressed in pancreatic cancer, the other three miRNAs were elevated in all three cancers, in agreement with the findings of multiple studies conducted over the last 25 years. The 1.8-fold overexpression of these cancer biomarkers agrees well with the overexpression observed in a prostate cancer study [23]. The 1.8-fold overexpression of miR-21 observed here is in excellent agreement with the 1.7-fold overexpression observed in a lung cancer study [22]. Strikingly, the normalized copy numbers of each of the five miRNAs appear to be independent of age, sex and ethnicity, in bold contrast to the variability observed using current miRNA quantification platforms. Compared with the corresponding miRNA copy number from the combined serum of healthy men (H6914 from Sigma-Aldrich), the data were grouped into healthy and cancer samples with no data overlap, that is, zero false negatives and zero false positives, yielding sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) all equal to 1. This unprecedented discrimination tentatively validated each miRNA (miR-21, miR-375 and miR-141) separately, as three-cancer biomarkers. The technology merits further testing in a larger sample size, as well as for other indications in addition to breast, prostate and pancreatic cancers. The ability of this platform to accurately quantify ss nucleic acid traces and validate potential miRNA biomarkers and its prospective adaptability to future solid-state nanopore platforms is unprecedented.