**3. MiRNA profiling and data analysis**

using the miRCURYTM method produced RNA concentrations above the lower detection limit of 2 ng/μL of the NanoDrop instrument and less than 10 ng/μL. Moreover, the miRCURYTM RNA isolation method showed a high average A260/280 ratio (1.9 ± 0.1) but low average

36 Biomarker - Indicator of Abnormal Physiological Process

**Figure 4.** Bioanalyzer analysis of serum exosomal small RNA. Exosome total RNA derived from serum was analyzed using Small RNA Kit in an Agilent 2100 Bioanalyzer. The electropherograms show the size distribution in nucleotides (nt) and fluorescence intensity (FU) of ladder (A) and exosome total RNA (B). The peak at 4 nt is an internal standard. The Small RNA region is visible in the interval of 0–150 nucleotides, including the miRNAs in the sizes between 10 and

**Figure 3.** Flow analysis on exosome sub-populations (A). Exosomes isolated with CD63 showed that serum-derived exosome in our samples express high level of CD63 marker (82.6%) and a moderate level of CD9 markers (32.8%) (B).

40 nucleotides.

miRNA expression profiling has allowed the identification of miRNAs that are involved in many biological processes, including organism development and establishment and maintenance of tissue differentiation [23, 24]. Thus, miRNAs are being explored as elements for cell-fate reprogramming in stem-cell applications or as biomarkers for identifying the origin of cancers of unknown primary sites. The expression pattern of miRNAs is widely different being tissue specific and related to developmental stages. Measuring miRNA expression can also be useful for system-level studies of gene regulation, especially when miRNA profiles are integrated with mRNA profiling. Circulating extracellular miRNAs, including exosome miRNA, are quantifiable in a range of specimen types including serum, plasma, urine and formalin-fixed tissue block. Hence, they are important as non-invasive biomarkers for many molecular diagnostic applications, including cancer [25, 26], cardiovascular and autoimmune diseases [27] and forensics [28]. **Table 1** shows new and promising miRNAs as potential biomarkers for diagnosis and prognosis of different cancers.

#### **3.1. General consideration for high-throughput miRNA profiling**

The yield of RNA extracted from biofluids is usually very low, in the order of 1–10 ng. If we consider that miRNAs represent only 0.01% of total RNA, the strategies used for their detection and quantification are crucial. Some general consideration must be made in high-throughput miRNA profiling. First, the length of mature miRNA (19–25 nucleotides) is too low to allow annealing to the traditional primers during the reverse transcription step. Second, unlike mRNAs, miRNAs lack poly(A) tail, a region frequently used to anneal complementary and universal primers for RNA enrichment or reverse transcription. Third, miRNAs exist in different isoforms, so-called isomiRs, that are functional and evolutionarily important and, inside the same family (e.g., the let-7 family), can differ by a single base to the reference miRNA sequence. Depending on the goals of an miRNA profiling experiment, measurement of different forms may be required, even if the large majority of miRNAs typically show only modestlength heterogeneity. Another challenge for miRNA high-throughput profiling is variance in miRNA GC content that is reflected in different melting temperatures (Tm) of annealing reactions. To date, three major approaches are used for miRNA profiling: quantitative reverse transcription PCR (qRT-PCR), hybridization-based methods (microarrays) and next-generation sequencing (NGS) (RNA-seq). Microarrays were among the first hybridization-based methods to be used for parallel analysis of large numbers of miRNAs.


different hybridization efficiencies of each probe, due to their different content in GC, the different melting temperature, due to the reduced size of miRNAs, the bias due to the enzymatic labeling and the relative low dynamic range. Microarray-based methods generally require a larger amount of starting material than qRT-PCR, and it can be challenging to develop probes and hybridization conditions that work well to detect many different miRNAs at once [43, 44]. miRNA profiling by NGS platforms may be the most promising approach, as it largely avoids many miRNA measurement pitfalls [45]. However, NGS remains expensive and labor

Circulating MicroRNA Profiling in Cancer Biomarker Discovery

http://dx.doi.org/10.5772/intechopen.75981

39

To date, oligonucleotide-miRNA microarray analysis is the most common high-throughput technique for assessment of disease-specific expression of hundreds of miRNAs. The Affymetrix GeneChip® miRNA microarrays platform provides the most sensitive, accurate and complete measurement of small non-coding RNA transcripts involved in gene regulation. It represents miRNA sequences from all organisms present in miRBase (http://www. mirbase.org), as well as small nucleolar RNAs (snoRNA) and small Cajal body-specific RNAs (scaRNA) included in snoRNABase (http://www.snorna.biotoul.fr/) and Ensembl (http:// www.ensembl.org). This platform has been used in our laboratory with success to profile exosome non-coding RNA from serum samples (unpublished data, manuscript in preparation). The protocols for non-coding RNA profiling include the use of the Affymetrix FlashTag Biotin HSR RNA Labeling Kit (Affymetrix, Santa Clara, CA, USA) and the detection by fluorescent emission. Total RNA samples or total RNA samples enriched for low molecular weight (LMW) RNA were starting materials. The process begins with a poly(A) tailing reaction followed by the ligation of a biotinylated signal molecule to the target RNA samples. The labeled RNA samples are included in the hybridization mix and hybridized overnight to Affymetrix GeneChip miRNA arrays (Affymetrix, Santa Clara, CA, USA), followed by a procedure to wash, stain and scan the array to acquire probe cell intensity data (CEL file). To date, the GeneChip miRNA 4.0 array is the updated array designed to interrogate all mature miRNA sequences contained in miRBase release 20, including 30,424 probe sets, covering 203 organisms and requiring low sample input (130–1000 ng total RNA). CEL files analysis by the Expression Console Software coupled with Transcriptome Analysis Console (TAC) Software (Affymetrix, Santa Clara, CA, USA) leads to a simple, fast and free analysis for Affymetrix

Raw data processing begins with quality control analysis by assessing the performance of internal controls and analyzing replicates to detect biases. For example, microarrays have well-known geographic biases, because some areas of the array perform differently from others. Data normalization, which is the next step after quality assessment, is crucially for obtaining accurate results [46]. The goal of normalization is to adjust the data to remove variation across samples not related to the biological condition and therefore allowing the identification of relevant biological differences. The normalization step is particularly important because some of the discrepancies between miRNA profiling studies are in part due to the application

intensive, both in the sample preparation and data analysis.

**3.3. Affymetrix GeneChip microarray platform**

GeneChip expression arrays.

**3.4. Analysis of raw data**

**Table 1.** Circulating miRNAs as potential biomarkers in different cancers.

#### **3.2. Microarray miRNA profiling**

Microarrays provide a high-throughput approach to profile all annotated miRNAs, in different types of samples including biological fluids. This method is relatively less expensive than others, such as qRT-PCR and next-generation sequencing (NGS). Moreover, the integrated analysis of the expression profile of miRNAs and their target genes, together with the analysis of each miRNA gene target pathway, is able to provide information on the function of each miRNA for a given sample. The technique generally begins with the enzymatic or chemical marking of targets followed by their hybridization to oligonucleotides fixed on a solid support. The signal generated by each probe is detected by a scanner and analyzed by specific software able to process the signal intensity. The variables related to the method include the different hybridization efficiencies of each probe, due to their different content in GC, the different melting temperature, due to the reduced size of miRNAs, the bias due to the enzymatic labeling and the relative low dynamic range. Microarray-based methods generally require a larger amount of starting material than qRT-PCR, and it can be challenging to develop probes and hybridization conditions that work well to detect many different miRNAs at once [43, 44]. miRNA profiling by NGS platforms may be the most promising approach, as it largely avoids many miRNA measurement pitfalls [45]. However, NGS remains expensive and labor intensive, both in the sample preparation and data analysis.

#### **3.3. Affymetrix GeneChip microarray platform**

To date, oligonucleotide-miRNA microarray analysis is the most common high-throughput technique for assessment of disease-specific expression of hundreds of miRNAs. The Affymetrix GeneChip® miRNA microarrays platform provides the most sensitive, accurate and complete measurement of small non-coding RNA transcripts involved in gene regulation. It represents miRNA sequences from all organisms present in miRBase (http://www. mirbase.org), as well as small nucleolar RNAs (snoRNA) and small Cajal body-specific RNAs (scaRNA) included in snoRNABase (http://www.snorna.biotoul.fr/) and Ensembl (http:// www.ensembl.org). This platform has been used in our laboratory with success to profile exosome non-coding RNA from serum samples (unpublished data, manuscript in preparation). The protocols for non-coding RNA profiling include the use of the Affymetrix FlashTag Biotin HSR RNA Labeling Kit (Affymetrix, Santa Clara, CA, USA) and the detection by fluorescent emission. Total RNA samples or total RNA samples enriched for low molecular weight (LMW) RNA were starting materials. The process begins with a poly(A) tailing reaction followed by the ligation of a biotinylated signal molecule to the target RNA samples. The labeled RNA samples are included in the hybridization mix and hybridized overnight to Affymetrix GeneChip miRNA arrays (Affymetrix, Santa Clara, CA, USA), followed by a procedure to wash, stain and scan the array to acquire probe cell intensity data (CEL file). To date, the GeneChip miRNA 4.0 array is the updated array designed to interrogate all mature miRNA sequences contained in miRBase release 20, including 30,424 probe sets, covering 203 organisms and requiring low sample input (130–1000 ng total RNA). CEL files analysis by the Expression Console Software coupled with Transcriptome Analysis Console (TAC) Software (Affymetrix, Santa Clara, CA, USA) leads to a simple, fast and free analysis for Affymetrix GeneChip expression arrays.

#### **3.4. Analysis of raw data**

**3.2. Microarray miRNA profiling**

D = diagnostic; P = prognostic; R = response predictor.

Microarrays provide a high-throughput approach to profile all annotated miRNAs, in different types of samples including biological fluids. This method is relatively less expensive than others, such as qRT-PCR and next-generation sequencing (NGS). Moreover, the integrated analysis of the expression profile of miRNAs and their target genes, together with the analysis of each miRNA gene target pathway, is able to provide information on the function of each miRNA for a given sample. The technique generally begins with the enzymatic or chemical marking of targets followed by their hybridization to oligonucleotides fixed on a solid support. The signal generated by each probe is detected by a scanner and analyzed by specific software able to process the signal intensity. The variables related to the method include the

**Cancer Samples miRNAs Types of** 

210, miR-126

133a, miR-133b

Lung cancer Serum miR-182, miR183, miR-

38 Biomarker - Indicator of Abnormal Physiological Process

Breast cancer Serum miR-1, miR-92a, miR-

Gastric cancer Serum miR-1, miR-20a, miR-27a,

Colorectal cancer Plasma miR-409-3p, miR-7,

Hepatocellular cancer Exosome miR-101, miR-221, miR-

**Table 1.** Circulating miRNAs as potential biomarkers in different cancers.

Head and neck Plasma miR-21 qRT-PCR D [29]

Ovarian cancer Serum miR-34a P [12] Prostate cancer Plasma miR-21, miR-141, miR-221 qRT-PCR D [13]

Renal cancer Serum miR-378, miR-451 qRT-PCR D [35] Pancreatic cancer Serum miR-16 and miR-196a qRT-PCR D [36]

Melanoma Serum miR-221 qRT-PCR D, P [41] Lymphoma Serum miR-221 qRT-PCR P [42] Leukemia Exosome miR-29a qRT-PCR P [8]

miR-34, miR-423-5p

miR-93

221, miR-224

Serum miR-125b qRT-PCR R [31]

Serum miR-125b qRT-PCR R [33]

miR-21 qRT-PCR R [37]

miR-126 qRT-PCR R [39]

**biomarkers**

qRT-PCR D [30]

Microarray D [32]

qRT-PCR D [34]

qRT-PCR D [38]

qRT-PCR D [40]

**Reference**

Raw data processing begins with quality control analysis by assessing the performance of internal controls and analyzing replicates to detect biases. For example, microarrays have well-known geographic biases, because some areas of the array perform differently from others. Data normalization, which is the next step after quality assessment, is crucially for obtaining accurate results [46]. The goal of normalization is to adjust the data to remove variation across samples not related to the biological condition and therefore allowing the identification of relevant biological differences. The normalization step is particularly important because some of the discrepancies between miRNA profiling studies are in part due to the application of different normalization approaches. GeneChip array data are normalized by the use of a tool that uses Robust Multichip Analysis (RAM) plus detection above background (DABG) algorithms, as default analysis. RAM is a robust linear normalization model, to minimize the effect of probe-specific affinity differences, and consists of three steps: background adjustment, quantile normalization and summarization. DABG is a detection metric generated by comparing perfect match probes to a distribution of background probes.

an independent gene expression profiling method. Most researchers choose qRT-PCR as the preferred method for the validation of microarray data using both TaqMan and SYBR-

Circulating MicroRNA Profiling in Cancer Biomarker Discovery

http://dx.doi.org/10.5772/intechopen.75981

41

The TaqMan qRT-PCR method uses a stem-loop RT primer, specifically designed to detect the 3′ end of individual mature miRNAs generating a unique template for RT. In the qPCR step, cDNA is amplified with specific primers and product accumulation is monitored using a fluorogenic probe (TaqMan probe), complementary to the target gene. In SYBR-green-based qRT-PCR, miRNA is typically poly-adenylated at the 3′ end, and oligo-d(T) is used as an RT primer while a double-strand DNA binding dye (SYBR-green) allows for the detection of PCR

We routinely use TaqMan® microRNA assays (Applied Biosystems, Inc.) to validate microarray data on the same clinical samples as described [51] and perform qPCR in triplicate reactions and the 2-ΔΔCt method to estimate the relative quantity of each miRNA [52]. An important issue is the choice of normalizer. Typically, "housekeeping" genes selected as endogenous controls allow normalization of qPCR data as they are affected by the same experimental variability as the target genes. In a cellular context, stable small RNA controls, such as RNU44, RNU48 and RNU6, are usually used. However, for circulating miRNAs, there is growing evidence that the abovementioned small RNAs are highly variable or not stably detectable [53]. This lack of consensus has resulted in the generation of various normalization strategies. An approach widely used and employed in our group is the selection, from each microarray

Circulating miRNAs are attractive as clinical biomarkers for diagnostic purposes, as well as for monitoring disease progression and response to treatment. However, the nature of circulating miRNAs places several challenges. The success of the circulating miRNA profiling requires rigorous control of pre-analytic and analytic variables, specifically when investigating potential circulating miRNA markers. Here, we provided a consistent and reproducible method for circulating miRNA detection, profiling and analysis. We discussed the main issues associated with miRNA measurement that is crucial for miRNA profiling, especially for exosomal circulating miRNA. In addition, it needs to take into account that it is difficult to measure specific miRNA levels because they are short and conserved sequences, paralogs or distinguish between precursor and mature forms. In our work experience, the use of standard protocols for sample preparation, and of exogenous synthetic miRNA as the standard control, helps to solve a part of these problems. Microarray data processing such as normalization procedures among different samples is challenging especially for extracellular miRNA. In our work experience, it can be concluded that the use of robust algorithms and software may avoid errors and false positive discovery. However, the validation of array results by the use of an alternative methodology, especially when using different protocols and platforms for

study, of several genes as a normalizer based on their stable expression.

green assay.

products during qPCR.

**4. Conclusion**

profiling purposes, is mandatory.

miRNA profiling experiments typically involve comparisons between two or more groups, and therefore the next stage of analysis is usually the calculation of differential miRNA expression between groups. The degree of fold difference that constitutes a meaningful difference depends on the experimental context, although it is always useful to assess the statistical significance and false discovery rate that is associated with the differential miRNA expression. This comparison yields a p-value, which is, then, combined into a probe set level p-value using the Fischer equation. Statistical analysis is performed using TAC and a fold change of two is commonly adopted to describe the signal changes between groups.

#### **3.5. Circulating miRNA profiling challenges**

Circulating exosome miRNAs are surprisingly stable and show distinct expression profiles among different fluids. Given the instability of most RNA molecules in the extracellular environment, the presence and apparent stability of miRNAs in body fluids such as serum and other body fluids, that are known to contain ribonucleases, suggest that secreted miRNAs are packaged in some manner to protect them against RNase digestion. miRNAs could be shielded from degradation by packaging in lipid vesicles, like exosomes, in complexes with RNA-binding proteins or both [47]. This view supports the idea that extracellular miRNAs are prepared for export in one cell. They can be recognized, taken up and utilized by another cell, working as mediators of cell-cell communication [48–50]. The growing interest in developing circulating miRNAs as blood-based biomarkers requires very careful consideration of the effects of various pre-analytical procedures, such as handling and storage conditions of the sample before processing, which can affect the reliability and reproducibility of circulating miRNA quantification. The main technical difficulty to study miRNA expression profiles is the efficient extraction of miRNAs from biological samples, because of their small size and their attachment to lipids and proteins. The use of commercial extraction kits has become available to optimize the extraction of small RNAs and normalize sample-to-sample variations in isolation procedures. Therefore, it is important to establish standardized protocols for blood collection, sample storage conditions, inclusion of exogenous and endogenous miRNA controls for each clinical sample and standardized calculations for normalization of the results to ensure the reproducible and accurate quantification of circulating miRNA levels so that miRNA analysis can be implemented in the clinical laboratory setting.

#### **3.6. Validation of microarray results by qRT-PCR**

miRNA microarrays are less expensive but inclined to have a lower sensitivity and dynamic range and are therefore best used as discovery tools rather than as quantitative assay platforms. Current publication guidelines require that all microarray results are confirmed by an independent gene expression profiling method. Most researchers choose qRT-PCR as the preferred method for the validation of microarray data using both TaqMan and SYBRgreen assay.

The TaqMan qRT-PCR method uses a stem-loop RT primer, specifically designed to detect the 3′ end of individual mature miRNAs generating a unique template for RT. In the qPCR step, cDNA is amplified with specific primers and product accumulation is monitored using a fluorogenic probe (TaqMan probe), complementary to the target gene. In SYBR-green-based qRT-PCR, miRNA is typically poly-adenylated at the 3′ end, and oligo-d(T) is used as an RT primer while a double-strand DNA binding dye (SYBR-green) allows for the detection of PCR products during qPCR.

We routinely use TaqMan® microRNA assays (Applied Biosystems, Inc.) to validate microarray data on the same clinical samples as described [51] and perform qPCR in triplicate reactions and the 2-ΔΔCt method to estimate the relative quantity of each miRNA [52]. An important issue is the choice of normalizer. Typically, "housekeeping" genes selected as endogenous controls allow normalization of qPCR data as they are affected by the same experimental variability as the target genes. In a cellular context, stable small RNA controls, such as RNU44, RNU48 and RNU6, are usually used. However, for circulating miRNAs, there is growing evidence that the abovementioned small RNAs are highly variable or not stably detectable [53]. This lack of consensus has resulted in the generation of various normalization strategies. An approach widely used and employed in our group is the selection, from each microarray study, of several genes as a normalizer based on their stable expression.
