MicroRNAExpressionProfilingbyBeadArray4

Bead-array-based microRNA detection technology, including the bio-statistic analysis, is currently not well established or widely used and we have applied a commercial PCR-based assay to confirm the array data for some microRNAs that cover different expression levels and change factors. In contrast to mRNA profiling, where RT-PCR-based assays are considered as gold standard for data validation, new generation deep sequencing is considered as the method of choice for microRNA quantification but is not available in our research institute. For the microRNAs let-7 a/b, miR-19 a/b, and miR-203, the PCR-based quantification method (Fig. 2b) confirmed the direction of change found with microarray technology (Table 2a). Expression of miR-130b and miR-455 was at similar levels in both assays. The correlation calculated for the eight tested microRNAs was acceptable: multiple r² from f test of mean relative cycling times (ΔCT) to mean log2 microarray expression values was 0.9279. Differences of absolute levels between the microRNA targets probably results from different hybridization properties of the microarray probes and variation in the performance of Taqman primers for the specific microRNA on the other side.

Assuming that any IFNα relevant microRNA will have the same kinetics as the mRNA for PRGs, we looked at the regulation of microRNA genes in our experiment. These IRmiRs should respond to IFNα stimulation preferentially in both cell lines, because this would be a good indication of a general mechanism in the IFNα response. Within the 25 most significantly regulated genes (Table 2b), only one gene (HS_250) is downregulated. A general upregulation of transcripts is consistent with classic IFNα signaling seen for mRNAs. However, the maximal observed change factor with high significance was 1.84 (miR-33b in Table 2b) which is clearly lower than the values seen for protein coding mRNAs (2). We also included an expression analysis 24 h after IFNα stimulation in order to detect microRNA genes that show either delayed induction or remain activated at comparable levels to the 4 h stimulus. Based on our data set, the majority of the microRNA response genes show no further induction, but rather moderate downregulation 24 h after induction. This finding is not surprising as we expected immediate early impact of IFNα-mediated primary signaling.

We also measured the IFNα response in the same experiment and for the same microRNAs (Table 2c). When we analyze the IFNα effect at the early time point in both cell lines we find all the validated microRNAs to be upregulated (Fig. 3a). The magnitude of upregulation and the basal expression levels of the microRNA-19a and 19b are similar in both cell lines (Fig. 3b, top). This and the finding that miR-19 regulates SOCS1 (4) may be relevant for the regulation of cytokine signaling. let-7a and let-7b had higher levels in the melanoma-cell-line-derived samples compared HuH7, but the induction by IFNα in ME-15 could not be reproduced by RT-PCR (Fig. 3b, bottom). In both assays accurate fold changes are difficult to calculate, if the baseline expression level is close to background noise or the detection limit. An example of a gene at the detection limit is miR-203, which is not detectable without IFNα treatment in HuH7 cells (Table 2c). Upon IFNα stimulation (24 h in HuH7) the microRNA is detectable above background suggesting minimal induction. Consequently a solid change factor cannot be calculated, which is consistent with the high variance obtained by qPCR (ME-15). This result is in fact not surprising, because both technologies rely on logarithmic PCR amplification of microRNA templates. At low expression levels, both technologies show relatively high variation in biological replicates, which should be considered for data interpretation. Interestingly, miR-203 has a putative binding site for ISGF3 in the promoter region, which would enable IFNα-dependent upregulation. miR-30 has been reported to be IFNβ inducible, although the subclass measured was not specified by the authors (9). We decided to analyze the most promising candidate (miR-30e-5p) present in our microarray dataset (Fig. 3a in gray). Detection of miR-30e failed in ME-15 cells due to technical problems, but induction in HuH7 was similar to miR-19a/b.

Fig. 3 IFNα-dependent modulation of microRNA expression. a Volcano plot display of IFNα induced microRNA upregulation 4 h after induction. The change factor values (2^log factor change −1) are plotted on the X-axis against the p value in logarithmic scale on the Y-axis. Top-rated microRNAs are annotated together with the let-7 family members. b Confirmation of IFNα effect for selected microarray data by qPCR. The CT-values are the average of three technical and three biological replicates and changes were calculated with 2^ΔMNE (mean normalized expression values). Error bars show 2^ΔMNE ± Δx (average standard error of treated and untreated MNE) from biological triplicates. miR-30e failed to amplify in ME-15 and miR-203 was below the detection limit in HuH7. Expression values were normalized against endogenous snoRNA RNU48 levels.

Some technology-related questions remain open. The microRNA assay measures essentially the number of amplicons generated by RT-PCR for each transcript. Thus the signal is an indirect measurement of transcript abundance as compared to classical mRNA microarray platforms, where the target mRNA is directly labeled during linear amplification by in vitro transcription. As a consequence, change factor calculations for amplicon-based assays are ambiguous.

In summary, Illumina’s bead array technology is well suited for multi-parallel profiling of microRNAs expressed in different cell types or tissues. We were also able to detect IFNα-inducible microRNA genes although the changes observed were moderate and biological significance remains to be proven. Like most microarray-based detection technologies the technical variability among identical samples is low compared to biological variations of individual cell cultures. At this point it is important to note that variation among biological samples occurs and is independent of the parameters that are measured. Consistent with IFNα-dependent induction of mRNAs we find that virtually all modulated microRNA genes are upregulated. However, the IFNα-induced changes detected in our study are relatively small compared to the changes induced by IFNβ in HuH7 cells (9). Finally, it is noteworthy that IRmiRs have similar kinetic properties to their mRNA counterparts. miR-10b for instance is induced early in ME-15 and remains upregulated, while miR-19 abundance ceases after 24 h. In general, the majority of IRmiR genes were reset to basal levels after 24 h and further studies are needed for kinetic classification. Thus, our study adds another level of complexity to the dynamic regulation IFNα signaling and other mechanisms like epigenetic promoter methylation are currently under intense investigation in our laboratories.

Acknowledgements  We thank Dr. Guido Steiner and Andreas Buness (F. Hoffmann-La Roche Ltd.) for their support in bioinformatics and statistics, Dr. Martin Ebeling (F. Hoffmann-La Roche Ltd.) for the comparative genomic evaluation of microRNA-targeted transcripts and Prof. Dr. Giulio Spagnoli (University Hospital Basel) for the gift of the ME-15 melanoma cell line. Finally, we are grateful to Heather Hinton (F. Hoffmann-La Roche Ltd.) for critical reading of the manuscript and to Dr. Laura Suter-Dick (F. Hoffmann-La Roche Ltd.) for sharing lab space and introduction into GCP sampling.