Variability in, variability out: best practice recommendations to standardize pre-analytical variables in the detection of circulating and tissue microRNAs

Common biological/environmental variables

Pre-analytical variables can be subdivided into biological, environmental, and technical factors [13]. Biological and environmental factors are predominantly common to both tissue and circulating miRNAs. Multiple studies have highlighted that both external and patient-intrinsic factors such as age, sex, race, body mass index (BMI), diet, fasting state, underlying illnesses/organ dysfunction, blood cell count, exercise, vitamin supplementation, medications, smoking, diurnal variation, chemical exposure and even altitude can influence miRNA concentrations and are often difficult to standardize [2], [13], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27]. Appropriate case and control subjects should be selected and specimen collection protocols designed and consistently implemented to minimize effects of confounding variables. Repeated sampling from the same individual can be considered as an additional method to attenuate the effects of pre-analytical variables [13]. Several recent reviews provide planning guides or outlines for circulating miRNA studies [1], [7].

Common technical variables

Technical factors include collection, processing, transport and storage and can also be controlled for in study design. Technical variables common to both tissue and circulating miRNAs are discussed below with detailed recommendations outlined in specimen-specific sections.

Hemolysis/blood cell contamination

Contamination from blood cells is a significant source of miRNAs and its impact has historically been underappreciated. This has led to the inappropriate use of blood cell-miRNAs as endogenous controls (miR-16) and their incorrect assignment as potential cancer biomarkers in non-blood cell diseases [3], [19]. A recent study demonstrated that blood cell counts and hemolysis can alter plasma miRNA concentrations by up to 50-fold and that 58% of reported tumor-associated circulating miRNAs are highly expressed in blood cells [19]. Among the miRNAs examined, the observed variation stemming solely from blood cells was greater than many literature-reported differences for potential cancer biomarkers, raising a concern that reported findings are related to the secondary effect of blood cell miRNA contamination. Blood cell counts and/or hemolysis also likely contribute to variable results observed between different methodologies (extraction methods, specimen type, etc.) and certainly hinder the establishment of a standard practice for miRNA profiling.

Although this phenomenon has been rigorously studied for circulating miRNAs, blood cell contamination is also relevant for tissue specimens and deserves special consideration. Tumor tissue can be highly vascularized and complete removal of blood cells from tissue is neither common practice nor practical. A more detailed discussion of the challenges associated with selecting cells or regions of interest in tissue is presented below (see “Tissue variability”).

To address this issue, there has been a collective effort to compile blood cell-miRNAs and their purported disease and biologic pathway associations [3], [6], [7], [28], [29], [30], [31]. Existing miRNA-disease relationship databases were recently reviewed and a selection guide created for specific research interests [32]. Investigators should carefully review available resources prior to designing or interpreting studies and eliminate candidate miRNAs that are highly expressed in blood cells. The development of a single comprehensive database of cellular expression patterns for miRNAs would also facilitate better selection of candidate miRNAs [7].

Specimen collection and processing

Collection and processing procedures significantly influence miRNA concentrations. As they are generally specific to either circulating or tissue-derived miRNAs, they are discussed in the corresponding sections below. A recent review provides helpful flow charts for miRNA profiling study design in both body fluids and tissue specimens, including factors to consider based on specimen type, purification requirements and study goals [1].

miRNA extraction and detection

After specimen collection and processing, miRNAs must be extracted. Extraction methods are broadly grouped into two categories: guanidine/phenol/chloroform (GPC)-based extraction and column- or bead-based commercial isolation kits. Most techniques are comparable within a single method category, but demonstrate significant variation between different methods [2]. A recent evaluation of extraction kits demonstrated good comparability between three of five tested kits [20]. In contrast to this study, miRNeasy and mirVana kits have shown increased RNA recovery in studies comparing extraction kits [33], [34], [35], [36], [37]. Despite some inconsistencies, there is evidence that column-based methods perform better than GPC extraction due to organic and phenolic contaminants in GPC-extracted nucleic acids [34], [37]. Size fractionation and gel purification via polyacrylamide gels can subsequently enrich for small RNAs (18–24 nucleotide range) [1]. This process specifically selects for mature miRNAs and reduces miRNA precursors. One proposal is direct miRNA detection (without extraction) as a superior method because it eliminates miRNA loss during extraction, removes the need to normalize to total RNA, and is easier to perform routinely in a clinical laboratory [9]. Overall, the lack of consensus regarding miRNA extraction highlights the importance of using the same extraction method throughout a study to minimize the introduction of confounding variables.

Normalization

Various detection methods [quantitative reverse transcription PCR (qRT-PCR), microarray, RNA sequencing (RNAseq), next generation sequencing (NGS)] are available for miRNA quantification from circulation and tissue alike [1], [9], [38]. All of these methods require normalization to account and correct for variation across samples. Varying normalization methods have likely also contributed to inconsistent miRNA results. Although a normalization consensus has not been established, exogenous miRNA “spike-in” [such as Caenorhabditis elegans (cel-miR-39, cel-miR-54, cel-miR-238) or Arabidopsis thaliana miRNAs] during extraction can assess extraction efficiency [3], [20], [35]. In one study, normalization to a spike in control miRNA reduced extraction bias between five different kits and improved correlation of qRT-PCR results [20]. A ubiquitously expressed endogenous miRNA can also be used to normalize; however, as discussed in the hemolysis section, great care must be taken to select an appropriate miRNA(s) whose concentration is largely unaffected by pre-analytic variables. A study on lung cancer biomarkers outlined an empirical approach to identify endogenous circulating miRNAs for normalization [39]. Another study categorized sets of reference miRNAs for tissue-based analysis of colorectal and pancreatic adenocarcinoma [40]. When a relatively large number of miRNAs are measured, endogenous control methods could be supplemented with a broad normalization approach such as global median normalization (the average expression of each miRNA relative to itself) [3], [5], [41]. Normalization to plasma or serum volume can also correct for variations in extraction efficiency since RNA concentration does not appear to correlate with the number of miRNAs detected [3], [9], [22], [35].

RNA preservation reagents

Multiple technologies have been developed to preserve circulating and tissue-based RNA species, including RNAlater, Tempus, PAXgene, and others. In general, miRNA species are much more stable than mRNA and most studies have not used these methods. Several studies have demonstrated successful isolation and quantitation of specific miRNAs or whole miRNomes from whole blood [42], [43], [44], [45] or tissue [46]. However, few studies have systematically compared RNA stabilization media with a control method. One study directly compared Tempus and PAXgene whole blood stabilization systems and identified small differences in miR30b quantitation [43]. Another detected variation in miRNA quantitation isolated from whole blood preserved in either RNAlater or PAXgene, but noted both systems were appropriate for miRNA analysis [45]. A different study of the PAXgene Tissue stabilization system identified high-concordance in miRNA quantitation between frozen and stabilized specimens (R²=0.95) and reduced concordance between stabilized and formalin-fixed, paraffin embedded (FFPE) tissues (R²=0.81). In summary, RNA stabilization systems are compatible with miRNA analysis, choice of reagent influences total RNA yield and miRNA quantitation, and represents a poorly studied pre-analytical variable. Insufficient data is available to offer a definitive recommendation on the use of RNA preservation reagents. Available data suggests these are compatible with miRNA analysis but we emphasize the need for consistency in study design.

Specimen type

Although miRNAs can be detected in many fluid types [51], we specifically focus on blood as the most widely studied source. Blood cell-derived miRNAs are substantial contributors to circulating miRNA concentrations, which makes whole blood a suboptimal specimen type [19], [52]. Comparisons of plasma versus serum for miRNA detection have yielded conflicting results as to which is the superior specimen type [5], [20], [30], [35]. It has been suggested that the observed increased plasma-derived miRNA yields are due to the presence of cellular contaminants [33]. Serum, in contrast, has less baseline platelet contamination than plasma; however, clot formation may result in variable release of confounding miRNAs from blood cells [20], [22]. As there is not overwhelming evidence for one specimen type, the choice of serum versus plasma should be based on the specific study goals with miRNAs of interest being cross-referenced to published literature of differential expression in serum versus plasma [30]. For standardization, it is important to use the same specimen type consistently throughout a study and to explicitly define both the specimen type and handling.

Fluid specimens likely contain miRNAs from different sources in different forms: excreted freely by cells, packaged in micro, or exosomes, or passively leaked from damaged cells. Although they exist in circulation primarily bound to Argonaute-containing protein complexes, the relative abundance of each miRNA phase within a sample can vary biologically, be influenced by specimen processing conditions, and/or be differentially detected. It is therefore necessary to address whether distinguishing mature miRNAs from precursor forms is feasible in study design. For example, some extraction methods do not capture exosomes. Some suggest that focusing on specific carriers in study design, such as extracellular vesicles, may be a means to not only generate more reproducible miRNA profiles, but also more clinically relevant profiles. Studies suggest that exosomal miRNA expression patterns correlate with tumor characteristics and could serve well as tumor markers, however evaluation has been suboptimal due limited available methods for isolation [53]. In addition, implementation into clinical practice will not be practical until isolation methods that require less time, labor and specimen volume have been optimized [6], [51].

Specimen collection

Multiple aspects of specimen collection can influence miRNA concentrations. Simple venipuncture can contaminate the specimen with epithelial cells and traumatic phlebotomy can cause hemolysis. Current recommendations are to use, at minimum, a 22-gage needle and to discard the first 1–2 mL of blood drawn [3], [54]. Tourniquet use and differences in posture at sampling can cause hemoconcentration and should be standardized [13]. Various anticoagulants and blood stabilizers commonly used for collection of plasma and serum can also affect miRNA concentrations. A direct comparison of four different anticoagulants [EDTA, heparin, sodium citrate, or sodium fluoride/potassium oxalate (NaF/KOx)] demonstrated the most reproducible quantitation using NaF/KOx [55]. As EDTA and citrate are more commonly used in clinical practice, EDTA is recommended because citrate may trigger hemolysis [5]. Heparin should generally be avoided due to its ability to inhibit enzymes used in amplification-based quantitation assays. If, however, specimens have already been collected in heparin tubes, at least one study demonstrated that adequate treatment with heparinase facilitated miRNA detection [5].

Overall, efforts should be made during collection and processing to minimize hemolysis, release of miRNAs from platelets, and contamination from skin or WBCs to eliminate potential systematic bias from blood cell contamination.

Specimen composition

Blood cell counts, plasma volume, and plasma components can significantly alter the profile of circulating miRNAs [2]. Multiple studies have correlated blood cell counts and measured miRNA concentrations, even without hemolysis present [19], [30], [52]. It is therefore important to minimize in vitro hemolysis and blood cell contamination and also to document the extent of blood cell contamination in each specimen.

For study of miRNAs that are know to be expressed in blood cells, measurement of a concurrent CBC should be considered to allow for appropriate interpretation of the miRNA levels and potential exclusion of patients with blood cell counts outside of a predefined range (Table 2) [19]. For all specimens a visual, spectrophotometric, or alternative hemolysis cutoff should be established in the specimen processing inclusion criteria for circulating miRNA studies (Table 2). Based on variability in measured concentrations of miR-451 and miR-16 (both highly expressed in red blood cells), some authors suggest measuring free hemoglobin at an absorbance of 414 nm with a cutoff of 0.2 or higher to identify hemolyzed specimens [56]. Since lipemia can falsely elevate absorbance at 414 nm, additional spectrophotometric assessment at 385 nm can be a lipemia indicator [5]. In cases where only purified RNA is available, a ratio of hemolysis-affected miRNAs to non-hemolysis affected miRNAs can determine the likelihood of erythrocyte miRNA contamination in the original specimen. For example, a delta quantification cycle value (Cq) (miR-23a – miR-451) of more than five can indicate possible red cell contamination whereas a delta Cq of >7 indicates a high risk of hemolysis [28]. Although it would be ideal to completely exclude all hemolyzed specimens, this may not be feasible or practical. A modified workflow for hemolyzed specimens has been proposed that excludes hemolyzed specimens from initial biomarker discovery, but that then conditionally allows for their inclusion in the final analysis if the miRNAs of interest are not affected by hemolysis [5].

The evaluation of hemolysis alone, however, is inadequate to account for blood cell-derived changes in miRNA concentrations. A recent study demonstrated hemolysis-independent release of primarily vesicle-associated blood cell miRNAs miR-16 and miR-21. This release occurred during the first 5 h of room temperature storage of whole blood, prior to separation of plasma or serum from cellular components [52]. Notably, increased concentrations were absent for miRNAs not associated with blood cell expression, such as liver-specific miR-122. Thus, there is a need for meticulous specimen processing and storage documentation as well as evaluation of blood cell associated miRNAs in each specimen.

Another underappreciated variable that can affect miRNA quantification is plasma volume. Plasma inherently contains polymerase inhibitors such as hemoglobin, lactoferrin, and IgG, which can co-purify with nucleic acids. A comprehensive analysis demonstrated that optimizing plasma volume is critical for accurate quantification: too little plasma will have insufficient miRNA for quantification and too much plasma can increase polymerase inhibitors. One study suggested 50 μL of serum or plasma optimized miRNA detection using SYBR Green or TaqMan qRT-PCR (11-fold and three-fold, respectively) compared to 200 and 10 μL [55]. Another study demonstrated better recovery with 200 μL of plasma compared to higher volumes of up to 500 μL [37]. Whether putative polymerase inhibitors interfere with miRNA detection depends on multiple factors including both extraction and quantification methods. Additional processing, such as enriching small RNAs and silica adsorption can be used to effectively remove inhibitors [55]. Depending on miRNA detection methods, a combination of polymerases resistant to blood-borne inhibitors could be an alternative workaround.

Specimen processing

Blood sample processing requires centrifugation to remove cellular components. Although plasma is considered the cell-free fluid portion of blood, it typically contains residual platelets and microparticles. It is essential to recognize that even trace amounts of contaminating platelets or microparticles will artificially increase miRNAs, and potentially obscure disease-related miRNA expression profiles because cellular miRNA concentrations are so much higher than fluid concentrations. Centrifugation speed can significantly impact miRNA concentrations either by inadequately removing cellular components (slow spin) or by causing cell lysis (fast spin). The influence of platelets and microparticles on miRNAs in matched fresh plasma and serum was systematically evaluated by measuring miRNAs after two-step centrifugation (slower 1940 g with no brake for 10 min followed by faster 3400 g with high brake for 10 min) and 0.22 μm filtration [30]. Significant differences were identified in 72% of the measured miRNAs due to processing (4× to >1000× variation in expression). Notably, the miRNAs most affected were those with the highest expression in platelets. Another study examined the effects of two-step centrifugation (1000 g followed by either 1000 g, 2000 g, or 10,000 g) and found that much faster spins of 10,000 g substantially reduced platelet-associated miRNAs [36]. A separate study showed that two-step centrifugation effectively removed platelets from frozen plasma and serum specimens stored for up to 6 years [30]. Incorporation of an additional centrifugation step during specimen processing is highly recommended to remove confounding platelet miRNAs [3], [5], [30]. Alternatively, filtration can remove platelets and cell debris. Although platelet-derived miRNAs accounted for the majority of differences observed, filtration additionally removed microparticle miRNAs [30]. The use of filtration should be considered based on the relevance of microparticle miRNAs to the study objectives. A post-processing platelet count with rejection of specimens above a predefined threshold can also be incorporated to further improve standardization [30].

In laboratory medicine, it is well established that delayed processing can cause significant changes in many measured analytes [13]. It is therefore not unexpected that delayed specimen processing affects miRNAs. For example, a delay in centrifugation of 6 h compared to 2 h caused significant variation in six miRNA profiles, particularly miR-15b and miR-191 [36]. An additional study investigating the effect of delayed processing showed an initial selective increase of vesicle-associated miRNAs at 1–3 h, followed by a steady decline in all miRNAs measured up to 24 h [52]. They also demonstrated that the decrease in miRNA concentrations could be ameliorated by the addition of an RNaseA inhibitor. These studies illustrate the need for rapid specimen processing that may include freezing separated plasma/serum or using a stabilizing agent such as an RNase inhibitor to maximize miRNA yield.

Specimen storage

Circulating miRNAs are typically preserved by storage at temperatures low enough to significantly reduce RNase activity rather than by chemical fixation. Fortunately, studies to date suggest that samples can be stored as processed/separated plasma/serum for at least 24 h either at room temperature or 4 °C [2], [33], [37], [47]. Furthermore, miRNAs are relatively stable and can be stored for at least 1 year at –20 °C or –70/80 °C [37]. A recent study compared multiple storage conditions on miR-134 and miR-346 concentrations in whole blood specimens measured by qRT-PCR [57]. There was no significant change due to processing delay, storage conditions, or storage duration; however, increased freeze-thaw cycles significantly decreased miRNA concentrations. Another storage study at −80 °C and −20 °C showed good stability of miRNAs for up to 6 years [58]. Interestingly, they found overall similar amounts of miRNAs when comparing storage at −80 °C and −20 °C but differential expression of miRNAs depending on the storage temperature, suggesting that storage temperatures should be consistent throughout a study. A separate study demonstrated similar miRNA profiles and expression following storage at −80 °C for 12 years [36]. Since there is minimal and occasionally conflicting data about the effects of long-term storage (multiple years), control specimens matched for storage time should be considered [35].

Freeze-thaw cycles can potentially affect multiple analytes and conflicting results have been reported on the impact of freeze-thaw cycles on miRNAs [47], [57], [58], [59]. Thus, minimizing freeze-thaw cycles is recommended.

Ischemic time after sample procurement

A notable variable in clinical tissue samples is time the tissue spends at either room temperature or at 4 °C following excision and the resulting loss of blood supply prior to freezing or fixing. The warm and cold ischemic time, respectively, vary significantly from sample to sample and both have profound and deleterious effects on downstream recovery of nucleic acids and protein antigens [62], [63]. In addition, hypoxic/ischemic injury is a major biological stress that can induce significant transcriptional changes; the degree to which hypoxia alters miRNA expression profiles is an open area of research. The degree of ischemic effects, particularly tissue autolysis, varies significantly with tissue type. For example, pancreatic tissue is particularly sensitive given the high concentration of intracellular peptidases [64]. Unfortunately, the ischemic time is not readily controllable for most clinical samples and is generally not routinely recorded. Few studies have systematically examined the effect of ischemic time on miRNA recovery, further confounding attempts to establish best practices.

A recent study carefully assessed the effects of ischemic time preceding fixation or freezing on miRNA expression of autopsy cardiac tissue following myocardial infarct (also known as the post mortem interval) [60]. In this study, the post mortem interval ranged from 10 h to 154 h; cadavers were stored at room temperature for 3 h–24 h prior to cold storage at 4 °C. High quality miRNAs were detectable from cardiac tissue up to a week following death and the concentration of miRNAs varied less than that of other small RNA species. Despite the ability of miRNAs to withstand harsh treatment, the authors suggest prompt transfer to 4 °C to better preserve miRNA integrity. They further note that electrophoretic profiling of total RNA likely underestimates the population of intact miRNA since longer RNA species are more readily degraded than shorter molecules, which can be visualized as a smear on gel electrophoresis.

Multiple studies have documented changes in miRNA expression profiles following even short ischemic events. One study demonstrated time-dependent changes in the expression of 56 miRNAs and 1788 mRNAs during warm ischemic time [65]. The most prominent effects (particularly for miRNAs) occurred after 1 h of warm ischemic time; thus, the authors recommend freezing tissue at 30 min of warm ischemic time [65]. A more recent study using a gerbil model of cerebral ischemia examined the effects of 2 min carotid artery occlusion on miRNA expression in the hippocampus [66]. They identified seven miRNAs upregulated following transient carotid artery ligation and concluded that these changes are a neuroprotective mechanism against ischemic injury.

Although miRNAs tend to remain stable after tissue is removed from a patient, ischemic time can nonetheless potentially change miRNA expression or, to a lesser degree, cause sample degradation. Confounding effects are that time dependent and warm ischemia are more potent than cold ischemia. Uncertainty about total ischemic time is of particular concern in retrospective analysis of archived tissue samples, whereas prospective studies may incorporate time and/or recording standards. In addition to recording and recognizing warm and cold ischemic times as sources of variability, we offer specific recommendations for prospective studies in Table 3. For studies of archived tissue, we recommend collecting ischemic time if available. If unavailable, a common miRNA control whose expression is not affected by ischemia is critical.

Fixative, fixation time and tissue storage

For clinical samples, neutral buffered formalin is the fixative of choice for optimal evaluation by pathologists. Fixation is a biochemically complex process, with significant variation over time, temperature, and tissue thickness and has notable implications for tissue histology, antigen and nucleic acid recovery [61], [62], [63]. Tissue samples may also be frozen without fixative or occasionally fixed in an alternative agent for a specific purpose (e.g. immunofluorescence or electron microscopy). Following fixation, specimens are “processed” and undergo a series of buffer exchanges through alcohols of increasing hydrophobicity, xylene, and ultimately paraffin wax (or less commonly plastic). The resulting FFPE “tissue blocks” are stored for many years, serving as reference material in the case of recurrent disease and as biospecimens for research. Therefore the choice of fixative, effects of freezing, and the total fixation time prior to processing are additional variables worth considering.

Specimen processing

Once specimens are fixed and submitted for histology, they are “processed” through a series of incubations in increasingly hydrophobic solutions (alcohols) and finally into xylene prior to being embedded in paraffin. Larger resections are typically processed for 10 h, while biopsies and rush specimens may go through a 2 or 4 h processing time. The question of whether different processing times/protocols influences miRNA recovery or downstream analysis has not been rigorously studied. Given the general robustness of miRNAs following formalin fixation, we hypothesize that a large difference is unlikely, but this remains a potential confounding variable. Thus, we reiterate the importance of standardizing sample types and selection of an appropriate control miRNA species.

Tissue variability

An clear advantage of tissue-based miRNA analysis is the ability to define specific cell populations for miRNA analysis and confirm the biological of interest. However, tissue sections are inherently heterogenous with widely variable amounts of pathologic tissue. Thus, careful cell population selection is a vital pre-analytical variable in tissue-based miRNA analysis. Further, tissue-specificity of miRNAs is an active area of research that requires correlation with histomorphology or cell lineage markers. If tissue is analyzed without careful selection of cell population, study results may be unduly biased due to varying density of the cells of interest admixing with other cell types. Companion histological analysis by an experienced pathologist is thus recommended prior to miRNA tissue analysis. A routine hematoxylin & eosin (H&E) stained section allows rapid confirmation/identification of diseased tissue, reactive tissue around the disease process including the presence of blood elements, and potential contaminating tissue from a different specimen.

Selected cell populations of interest

In addition to cells of interest, neoplastic tissue sections invariably contain reactive cells as well as uninvolved normal tissue. For tissue-based miRNA analysis, FFPE blocks are selected based on arbitrary concentrations of diseased tissue. Thresholds are variable and may be qualitative [49] or semi-quantitative, ranging from at least 70%–80% neoplastic cells [73], [74] to as low as >30% neoplastic cells [75]. Tissue sections of neoplasms are complex and contain multiple cell types such that an estimate of 80% diseased cells likely includes a substantial population of stromal/mesenchymal cells, including activated stromal cells responding to the disease process. Contamination has the potential to severely bias results, as demonstrated by recent work where two miRNAs, miR-143/145, proposed to function as colorectal adenocarcinoma tumor suppressors, are not expressed by colonic epithelial cells at all [76]. Rather, the authors discovered that mesenchymal cells (fibroblasts and smooth muscle cells) express these miRNAs, which are expressed 33-fold less in non-neoplastic epithelial cells.

Several solutions to this sampling problem are regularly employed to enrich for cells of interest. One approach is to macrodissect regions containing apparently homogenous populations of the cell type of interest based on histopathological evaluation of H&E stained slides [77]. A second approach adopted in multiple studies is to use laser capture microdissection to isolate cells of interest [69], [78] Additionally, flow cytometry has been successfully used to select specific cell subtypes, such as epithelial cells, in samples of dissociated, intact cells [76]. Finally, post hoc validation can be performed by detection of candidate miRNAs of interest in neoplastic cells by fluorescence in situ hybridization (FISH) [79], [80].

It must be noted that there could be biological significance to altered miRNA profiles in reactive/adjacent cells as neighboring cells may have profound roles in the pathologic process; however, we stress that determining the cells from which the miRNA signal originates is a critical factor. We emphasize that each of the above validation/selection mechanisms requires histopathologic examination to confirm the cell population examined and that while the mechanism to select/validate the cells studied may vary by disease process and available tissue, it should be explicitly stated. Finally, we suggest that erratic, suspect, or unexpected results be analyzed to confirm whether the miRNA signals detected originate with pathologic cells or bystander/reactive cells.

Contamination by blood components

As discussed above, multiple blood components can confound miRNA analysis [29], [30], underscoring the importance of carefully selecting regions of interest by macro- or microdissection and, ideally, determining what cell types express the candidate miRNA(s) by in situ visualization. Both tissue resident immune cells and blood components are always present in tissue samples, sometimes with prognostic significance (i.e. tumor-infiltrating lymphocytes in colonic adenocarcinoma and melanoma) [81], [82], [83], [84] The challenge of avoiding confounding cells from any hematopoietic lineage may be particularly acute in disease processes with a robust immunological response, bloody specimens, friable tissue, highly vascularized tissue, or neoplasms with a prominent vascular component.

Variation in miRNA extraction

As previously reviewed and discussed above, multiple miRNA extraction protocols are available but do not always yield identical results [2]. Unfortunately, many studies extract miRNA from tissue using different methods depending upon the specimen type (i.e. FFPE tissue, cytology specimens, cell culture, or frozen tissue) [60], [61], [68], [69], [74]. A full review of the performance of different miRNA isolation techniques is beyond the scope. However, we emphasize that this is a notable and frequently ill-considered pre-analytical variable, particularly when specimen type is changed simultaneously.

Pre-analytical variables associated with in situ analysis

As highlighted above, it is often both informative and a helpful ancillary validation step to correlate miRNA expression with specific cell types through dissection, FISH/chromogenic in situ hybridization, or immunohistochemical (IHC) phenotype of cells from which RNA is extracted. Already, miRNAs are being advocated to distinguish diseased (typically neoplastic) tissue or characterize cell differentiation. For example, recent work has suggested that miR-211 may be aid in distinguishing invasive melanoma from dysplastic nevi, a potentially challenging diagnostic decision in anatomic pathology [85]. Whether for biomarker discovery or clinical diagnostic assays, such in situ techniques are becoming more commonplace and several important pre-analytic variables are worth noting.

For laser capture microdissection, one recent study reported enhanced miRNA yields from microdissected tissues by using crystal violet for histologic staining (as opposed to the more routine H&E), silicon carbide (SiC) matrix in RNA-binding columns, and overnight treatment with proteinase K [86]. Storing stained slides at room temperature prior to laser capture microdissection for up to 7 days did not adversely impact miRNA recovery, nor did storage at room temperature for up to a day between dissection and RNA extraction. These findings may simplify workflow in the laboratory.

IHC staining is frequently performed for pathological classification and is useful in identifying specific cells for microdissection. One study demonstrated sensitive quantitation of miRNAs by qRT-PCR in as few as 20 cells isolated from FFPE tissue and tested the impact of temperature in IHC staining protocols on miRNA recovery [78]. Relative expression of five miRNAs analyzed within each specimen remained consistent regardless of whether the IHC protocol avoided heating the tissue (hemalaun) or required heating to 98 °C or 90 °C.

Detection of miRNAs by FISH poses several technical challenges, including diffusion of miRNAs out of tissue and suboptimal hybridization efficiency. It is feasible and potentially useful to perform multicolor miRNA FISH and several solutions to these challenges have been demonstrated [79], [87]. To minimize diffusion of miRNAs, tissue sections may be incubated with 1-ethyl-3-[3-dimethylaminopropyl] carbodiimide (EDC) during or post-fixation [79], [87]. A longer linker region between a fluorophore-binding hapten (biotin) and a target-specific locked nucleic acid (LNA) probe has been demonstrated to improve probe-target hybridization efficiency [79]. Multiple mechanisms have been identified to boost fluorescence signal, particularly tyramide amplification [79], [87]. In addition, not all FISH protocols are amenable to subsequent or concurrent IHC [79], [87]. Methods aimed at combined detection of miRNAs and protein targets are an area of active research [87].