MicroRNAs (miRNAs) are small, siRNA-like molecules, encoded in the genomes of plants and animals, that regulate the expression of genes by binding to and modulating the translation of specific mRNAs. Published reports indicate that the expression levels of miRNAs vary between tissues and developmental stages, and that several miRNAs appear to be down-regulated in patients with chronic lymphocytic leukemia, colonic adenocarcinoma, and Burkitt's lymphoma [1-3]. Viral miRNAs also appear to play roles in infection .
Ambion scientists have developed a microarray analysis process that makes it possible to compare the miRNA expression profiles of different tissue and cell samples. This process has been used to analyze samples from different human organs as well as tumor and normal adjacent tissues from patients with lung, colon, breast, prostate, bladder, thyroid, and pancreatic cancers. The miRNA array procedure uses the mirVana miRNA Probe Set, to create miRNA arrays, and the mirVana miRNA Labeling Kit, to label samples for subsequent miRNA array hybridization. This procedure is highly sensitive, accurate, and reproducible, providing an effective tool for evaluating miRNA expression during development, differentiation, viral infection, or ned for each microRNA. Average correlation was 98%. Distribution of Log Ratio for the 6 replicates was centered and normal-like (data not shown).
miRNAs in Normal Human Tissues
miRNA from 26 normal human tissues was purified. A pool of one half of each sample was placed in a tube. The miRNAs from the pooled sample as well as each of the 26 single-tissue samples were fluorescently labeled with Cy3 or Cy5 using the mirVana miRNA Labeling Kit. Each tissue-specific miRNA sample was mixed with part of a pooled sample and the mixtures were hybridized to miRNA arrays generated from the mirVana miRNA Probe Set. The relative signal intensities of the single and pooled samples were compared for each miRNA. Each of the 26 normal human tissues that were analyzed had a unique pattern of miRNA expression. Hierarchical clustering shows that miRNA profiles are similar between related tissues and distinct between unrelated tissues--heart and skeletal muscle profiles are very similar, digestive tract tissues cluster, and reproductive organ tissues are similar (Figure 4). Interestingly, the brain miRNA profile is clearly distinct from the other tissues that were analyzed. Approximately 30 miRNAs are expressed primarily in 1-3 related tissue types. Interestingly, many of the putative target genes for these miRNAs associate with the same tissue wherein the miRNAs are expressed (data not shown).
Figure 4. Relative miRNA Expression inNormal Human Tissues. The miRNA expression profiles (y axis) of 26 different human normal tissues (x axis) was compared to a pool of all 26 samples. Green in the heat map shows miRNAs t hat are down-regulated in the sample relative to the pool, and red shows miRNAs that are up-regulated in the sample relative to the pool. Clustering of the 26 samples based on the miRNA expression profiles is shown above the heat map.
The mirVana miRNA Detection Kit, which relies on a solution hybridization detection scheme optimized for detection of miRNA and other small RNAs, was used to measure miRNA expression levels of three of the miRNAs in total RNA samples from ten of the human tissues. As seen in Figure 5, the quantitative data from the miRNA array procedure is very similar to what was generated using the mirVana miRNA Detection Kit or Northern analysis. This indicated that the miRNA Array procedure is generating reliable quantitative data for miRNA analysis.
Figure 5. Independent Verification of Data from the miRNA Array Procedure. The expression of two different miRNAs was measured in ten different human total RNA samples using the mirVana miRNA Detection Kit or Northern analysis. 5S and 5.8S rRNAs were used as loading controls. The relative expression levels of the miRNAs in the specified tissue relative to the pooled sample from the miRNA Array experiment in Figure 4 are shown graphically below each gel. Panel A: let-7c. Panel B: miR-16.
miRNAs in Cancer Samples
The miRNA array procedure was also used to compare the miRNA expression profiles of tumor and normal adjacent tissues from patients with lung, colon, breast, bladder, and thyroid cancers. Cancer samples have miRNA expression profiles that are clearly distinct from normal tissues (Figure 6) . A number of miRNAs appear to be routinely under- or over-expressed in tumors. For instance, miR-126, miR-143, and miR-145 were expressed at significantly lower levels in more than 80% of the tumor samples compared to their associated normal tissues. miR-21 was found to be over-expressed in 80% of the tumor samples. These miRNAs likely represent biomolecules that directly or indirectly influence oncogenesis. In addition, several miRNAs were found to be differentially expressed in specific types of cancers, suggesting that there are disease-specific miRNAs.
Figure 6. miRNA Expression Profiles of Tumor vs Normal Adjacent Tissue from Cancer Patients. The miRNA expression profiles (y axis) in tumor vs normal adjacent human tissues (x axis) were compared for 19 cancer patients. Green in the heat map shows miRNAs that are down-regulated in the tumor sample relative to the normal adjacent tissue sample, and red shows miRNAs that are up-regulated in the tumor sample relative to the normal adjacent tissue sample.
As with the normal tissue samples described above, we used a secondary method of detection to confirm the miRNA array results for a few of the more interesting miRNAs (Figure 7). The samples from the two lung cancer patients that provided the most total RNA were analyzed for expression of miR-16, miR-21, miR-143, miR-145, and let-7. As was observed in the comparisons of the miRNA expression profiles in the normal samples, the quantitative data from the miRNA array procedure were very similar to the secondary analysis, again, validating the procedure.
Figure 7. Independent Verification of miRNA Arra y Procedure Expression Data. Total RNA samples from two lung cancer patients were analyzed for expression of miR-16, miR-21, miR-143, miR-145, and let-7 using a secondary method of detection (Northern blotting or mirVana miRNA Detection Kit). The graphs show the relative abundance of each miRNA (ratio of tumor:NAT) from the array analysis and the secondary detection method after phosphoimager analysis.
Using novel miRNA array technology to simultaneously analyze the expression of all known mammalian miRNAs, we identified distinct miRNA expression profiles between tumor and normal adjacent tissue samples in cancer patients. The expression of a handful of miRNAs appears to correlate with cancer, suggesting that perturbing the expression of some miRNAs could contribute to oncogenesis.
Additionally, we verified that adult human tissues can be clearly distinguished by their miRNA expression profiles. This suggests that miRNAs are functioning in adults to regulate gene expression. The miRNA array procedure used for this investigation represents the first commercially available system for miRNA profiling.
Although the first published description of an miRNA occurred more than ten years ago , only recently has the breadth and importance of this class of small, regulatory RNAs been appreciated. Hundreds of unique miRNAs are encoded in animal, plant, and fungal genomes and are expressed in a regulated manner [reviewed in 6]. miRNAs are transcribed as parts of longer RNA molecules  that are processed in the nucleus into hairpin RNAs of 70-100 nucleotides by the dsRNA-specific ribonuclease Drosha  (Figure 1). The hairpin RNAs are transported to the cytoplasm and digested by a second, double-strand specific ribonuclease, Dicer. The resulting 19-23mer miRNA is bound by a complex that is similar to or identical to the RNA-Induced Silencing Complex (RISC), which participates in RNA interference . The complex-bound, single-stranded miRNA binds mRNAs with sequences that are often significantly, though not completely, complementary to the mRNA. In animals, the bound mRNA typically remains intact but is not translated, resulting in reduced expression of the corresponding gene.
Figure 1. miRNA Processing and Activity.
miRNAs as Regulators of Global Gene Expression
Several hundred miRNAs have been cloned and sequenced from mouse, human, Drosophila, Caenorhabditis elegans, and Arabidopsis samples. Estimates suggest that 200-300 unique miRNA genes are present in the genomes of both humans and mice . The sequences of many of t he miRNAs are homologous between species, suggesting that miRNAs represent an evolutionally conserved and critical regulatory pathway .
The best understood miRNA, lin-4, was identified in a C. elegans mutant screen designed to identify genes involved in developmental timing [reviewed in 12]. Cloning and analysis revealed that lin-4 is a small RNA that does not encode a protein . The lin-4 miRNA accumulates during the first and second larval stages and triggers passage to the third larval stage by repressing the translation of at least two genes, lin-14 and lin-28 . The activity of lin-4 depends on the partial homology of the miRNA to specific regions of the 3' untranslated regions (UTRs) of lin-14 and lin-28 mRNAs [5, 14]. Additional miRNAs have been functionally characterized (Figure 2). The range of miRNA functions affect early stage development, cell differentiation, and cell death, illustrating the critical roles that miRNAs play in cellular biology.
Figure 2. Function of Known miRNAs in Plants, Mouse, Fly, and Worm.
In animals, miRNAs are partially homologous to their mRNA targets. Using similar target selection criteria, a number of labs have predicted the mRNAs that miRNAs bind and regulate [15, 16]. Some of these putative miRNA target sites have been verified by monitoring target protein expression and using reporter constructs [15, 16].
Three critical points about miRNA function are:
(1) Each miRNA appears to regulate the translation of multiple gene s, and many genes appear to be regulated by multiple miRNAs . This could explain why at least some miRNAs have such broad functionality and might also point to complex translational control of some genes.
(2) It has been predicted that the expression of as many as 10% of mammalian genes are regulated by miRNAs .
(3) If miRNAs indeed regulate the translation of, but not the stability of, mRNAs, this might at least partially explain why gene expression profiles based on mRNA analysis do not always correlate with protein expression data . As more is learned about the mRNA targets of the different miRNAs, it will be possible to more accurately assess gene expression for a given sample by combining the profiles of mRNA and miRNA expression.
Role of miRNAs as Disease Markers
Given the role of miRNAs in regulating development and differentiation, multiple researchers have investigated the relationship between miRNA and oncogenesis. Several publications report a correlation between aberrant miRNA expression and miRNA gene sites with cancer, including the following:
(1) Chronic lymphocytic leukemia (CLL) patients commonly exhibit a chromosomal abnormality at 13q14. No known tumor suppressor genes exist at this locus and no consistent involvement of any of the genes in the deleted region has been demonstrated. Interestingly, the genes for miR-15 and miR-16 are located at this locu s and appear to be deleted in the majority of CLL cases .
(2) Twenty-eight different miRNAs were identified in human colorectal mucosa; two of these proved to be significantly down-regulated in twelve adenocarcinoma samples compared to matched, normal tissues . The two miRNAs were also down-regulated in the precancerous adenomatous polyps that were tested.
(3) Sequence analysis of tumor cells from eleven Burkitt lymphoma patients revealed that every one had a chromosome rearrangement in the miR-155 gene . None of the twenty-one normal (negative control) individuals tested had the rearrangement.
(4) A comparison of the chromosomal locations of 186 miRNA genes and cancer-associated genomic regions revealed that more than half of the miRNAs (98 out of 186) are in common break-point regions, fragile sites, minimal regions of loss of heterozygosity, and minimal regions of amplification .
Each of these reports suggest that miRNAs play a role in oncogenesis. We have developed and used an array procedure to measure miRNA expression in a variety of cancerous samples. The miRNA profiles from these samples suggest that a handful of miRNAs are routinely up- and down-regulated in tumors. The miRNA array procedure is now available and can be used not only for cancer research, but also for the analysis of development, tissue differentiation, and viral regulation.
Materia ls and Methods
miRNA Array Analysis
Total RNA from tumor and normal adjacent tissue (NAT) samples was isolated using the mirVana miRNA Isolation Kit (Ambion). Each total RNA sample (20 g) was gel fractionated using an electrophoresis unit specifically designed to speed up small RNA size fractionation and isolation. miRNA fractions for each sample were recovered and fluorescently labeled with Cy3 or Cy5 (Amersham) using the mirVana miRNA Labeling Kit (Ambion). The labeled miRNAs were hybridized (14 hours) with slides arrayed with 167 distinct miRNA probes, representing all known human and mouse miRNAs. The microarrays were washed 3 x 2 min in Wash Solution (Ambion) and scanned using a GenePix 4000B (Axon). Fluorescence intensities for the Cy3- and Cy5-labeled samples for each element were normalized by total Cy3 and Cy5 signal on the arrays. The normalized signal intensity for each element was compared between the tumor and NAT samples from each pair of patient samples and expressed as a log ratio of the tumor to normal adjacent sample.
Validation of miRNA Array Results
Northern analysis--Total RNA (1 g) was fractionated using a 15% denaturing polyacrylamide gel. The RNA was transferred to a positively charged nylon membrane by electroblotting at 200 mA in 0.5X TBE for 2 hr. After drying, the Northern blot was incubated overnight in 10 ml ULTRAhyb-Oligo (Ambion) with 107 cpm of a radiolabeled probe complementary to let-7c prepared with the mirVana Probe & Marker Kit (Ambion). The blot was washed 3 x 10 min at room temperature in 2X SSC, 0.5% SDS, and then 1 x 15 min at 42C in 2X SSC, 0.5% SDS. Overnight phosphorimaging using the Storm system (Amersham) revealed the indicated mature miRNAs. The process was repeated using a radiolabeled probe for 5S or 5.8S rRNA as a loading control.
Solution hybridization detection - The mirVana miRNA Detection Kit (Ambion) was used to measure the expression levels of several other miRNAs across multiple samples. Total RNA (1 g) was incubated with radiolabeled probes for several different miRNAs. Following digestion to remove probe that was not bound by target miRNA, the radiolabeled products were fractionated by denaturing polyacrylamide gel electrophoresis and quantitated using the Storm system (Amersham). Probes were prepared by in vitro transcription with the mirVana Probe Construction Kit (Ambion).
Results and Discussion
miRNA Expression Profiling by Microarray Analysis
To analyze the global expression patterns of the nearly 200 miRNA genes identified so far, we employed a novel nonisotopic detection method. To reduce the opportunity for nonspecific cross hybridization of longer mRNA, or rRNA or precursor miRNA species to individual miRNA-specific oligonucleotide probe, Ambion scient ists have developed a microarray analysis procedure that interrogates only the mature and functionally active miRNAs (See Novel MicroRNA Array Technology for Sensitive miRNA Profiling). The procedure comprises four steps and relies on many of the processes and instrumentation that are used for standard mRNA microarray analysis:
Step 1 - Total RNA that includes the small RNA fraction is isolated from a tissue or cell sample using the mirVana miRNA Isolation Kit (Ambion).
Step 2 - The RNA in the sample is fractionated using a specialized electrophoresis unit and the miRNA fraction is recovered.
Step 3 - The miRNA sample is tailed by poly(A) polymerase. The tail includes an amine-modified nucleotide that reacts with activated Cy3 or Cy5 to fluorescently label the miRNAs. The mirVana miRNA Labeling Kit (Ambion) makes the tailing and subsequent purification steps simple, and more importantly, reproducible.
Step 4 - The fluorescent miRNAs are hybridized to a glass slide upon which specially designed miRNA-specific probes have been arrayed. miRNA probes for printing glass miRNA Arrays are available from Ambion as the mirVana miRNA Probe Set (Ambion). The resulting miRNA arrays are then processed using standard array scanners.
miRNA Array Rep roducibility
For any method that is used to compare the abundance of RNAs in two different samples, it is critical that the process be highly reproducible. We tested the reproducibility of the miRNA array procedure by repeatedly comparing the miRNA profiles of human prostate and colon samples. Total RNA from a single human prostate sample and a single human colon sample were isolated. The total RNA from the prostate sample was split into six samples, and miRNA from each sample was purified and fluorescently labeled with Cy3 using the mirVana miRNA Labeling Kit. The miRNA from the human colon sample was purified and fluorescently labeled with Cy5.
Each prostate miRNA sample was mixed with one-sixth of the Cy5-labeled colon sample. The six miRNA mixtures were analyzed using a miRNA array containing a complete set of all known human and mouse miRNAs. The signal from the hybridized miRNAs was quantitated, and the signals from the colon and prostate samples were compared. Signal ratios at each element are expressed as a log ratio in Figure 3. The average correlation between the six independent reactions was 98%, indicating that the miRNA array process is highly reproducible.
Figure 3. The
miRNA Array Procedure is Reproducible. The
reproducibility of the miRNA array procedure was tested by repeatedly
comparing the miRNA profiles of human prostate and colon samples. Signal
ratios (colon:prostate) at each element are expressed as a log ratio.
Standard deviation of Log Ratio between the 6 replicates was determi