Little is known about the response of microRNAs as a result of specific receptor signaling. Microarray studies have shown regulated expression of hippocampal microRNAs in response to the induction of mGluR-dependent LTD (Park and Tang, 2008
) or LTP (Wibrand, et al., 2010
). Thus, we examined the direct effects of the group I mGluR-specific agonist, DHPG on microRNA expression by ICV injection in adult mouse brain, a well-established model for examining in vivo
responses. We profiled microRNA expression using Ambion and Invitrogen microarrays, as well as a SABiosciences qPCR array which detects mature mouse microRNA sequences. Our studies show that DHPG injection leads to the regulation of microRNA expression in mouse cerebral cortex at 4, 8 and 24 hours after treatment, with the greatest number of microRNA changes detected at 8 hours after DHPG in all array platforms. The mechanisms for regulation of microRNA expression by DHPG are not yet known. However, potential mechanisms might include changes in gene transcription or post-translation modification of microRNA processing proteins or RNA-induced silencing complexes, or a combination of both mechanisms mediated by group I mGluR activation.
Our understanding of microRNAs, their sequences, and physiological roles is continuously expanding, as evidenced by the rapid changes in mirBase version numbers (www.mirBase.org
). Changes include newly identified microRNAs, nomenclature changes, 3′ and 5′ strand distinctions, and deletions of sequences. To ensure that our analysis reflects the most recent understanding of microRNAs in mouse brain, we used the probe sequences provided by Ambion and Invitrogen to reannotate the microarray identifiers. We used the most stringent condition, 100% sequence match with no deletions or gaps in mouse specific mature sequences, to ensure only high confidence annotations were used. In this process, we found good correspondence among reannotated genes (Supplemental File S1
). The growing repositories of microarray data provide a wealth of knowledge; however, to fully exploit this resource, it will be necessary to reannotate archived data to the most current curated version of mirBase (Git, et al., 2010
The analysis of qPCR array data relies on a “housekeeping” gene that does not change across experimental conditions to normalize data sets. As with other quantifiable methods, the identification of a reliable unchanging reference is a continuing challenge. We examined several methods for normalization including normalizing to the average of all probes, analogous to the microarray method, by unchanging targets as recommended by the vendor, and by small nucleolar and nuclear RNAs present on all plates in the qPCR array. The housekeeping genes provided by the array manufacturer (Snord85, Snord66, Snord68, and Rnu6) did not change uniformly across experimental groups. Regardless of the normalization method selected, at 4 and 24 hours after DHPG, the ΔCt values followed a Gaussian distribution, and the ΔΔCt values were centered around zero, suggesting that relatively few of the microRNAs changed significantly. However, at 8 hours, while the ΔCt distribution was still Gaussian (Supplemental File S5
), the ΔΔCt distribution was skewed significantly depending on the method employed, making the use of a ΔΔCt threshold for significant increase or decrease difficult to establish. Therefore, we chose to determine significance based on the difference from the average ΔΔCt for each time point, which was also insensitive to normalization method at each time point examined. These results highlight that, while qPCR is a quantifiable method, the selection of housekeeping genes can have a profound effect on the interpretation of “up” or “down” regulation of a specific microRNA.
We found little overlap in the microRNAs identified by microarray and qPCR array methods. This likely reflects technology differences such as the distinct probe designs used by each vendor and the evolution of mirBase. This finding is in accord with a recent rigorous study which revealed that such outcomes represent inherent problems within and between the different assays (Git, et al., 2010
). The study examined identical RNA samples on 6 distinct microRNA microarrays obtained from different vendors, and found that there was little correlation between the datasets. Only 1 of 6 microarray vendors (Agilent) used probes specifically targeted to the mature microRNA sequence, while the others used probes that can detect the mature microRNA sequence but can also detect microRNA sequences in the primary and precursor transcripts. Our samples were analyzed by an array comprised of both Ambion and Invitrogen probe sets, which showed high levels of false positives and negatives in the Git study.
Perhaps the most significant difference between the two platforms is that qPCR selectively amplifies mature microRNAs and the microarrays are qualitative indicators of primary, precursor, and mature microRNA transcripts. The selection of either method must be based on specific experimental needs. For example, the cost of array processing makes the use of experimental replicates feasible in most cases. Multiple replicates makes the data amenable to familiar statistical techniques, and can reveal relatively small magnitude changes from control. The qPCR is a more precise, quantifiable method, but significantly increases the cost of for experimental or technical replicates, and requires the use of a threshold to establish significance; for difficult-to-obtain tissue, this may be a rational trade off. MicroRNA microarray and qPCR array technologies will continue to evolve, as they have for mRNA analysis. Git and colleagues also found inconsistencies between data obtained from microarrays and qPCR arrays, as well as from NextGen sequencing (Git, et al., 2010
). However, even with discrepancies between the Ambion, Invitrogen, and SABiosciences qPCR array datasets, the most compelling results from these studies was that DHPG consistently induced significant changes in microRNA expression at each time point examined, with the greatest changes seen 8 hours after DHPG injection. In addition, the KEGG pathway analysis (discussed below) identified specific validated group I mGluR-associated pathways, giving credence to the specificity of the reported microRNA response to DHPG.
We examined the anatomic distribution of select microRNAs using in situ hybridization, and quantified microRNA expression by RNA Blot. As with the microarrays, the probes for these experiments can recognize primary, precursor, and mature forms of the microRNA. The tissue we used for both the microarrays and qPCR array consisted of the cerebral cortex, which includes both the cortex and hippocampus. The in situ hybridization revealed regional differences in the effects of DHPG. For example, while miR-132 was not predicted to change in the microarray platforms, there is a slight increase in the pyramidal neurons of the hippocampus, which was likely too small to be detected in the cerebral cortex preparations used in these experiments. This was further confirmed by the RNA blots, which demonstrate no significant change in the mature miR-132 bands. The diffuse distribution of the mir-19a probe follows the pattern of reduced expression at 8 hours relative to the 4 and 24 hour time point, and may reflect a dendritic localization, though additional experiments would be necessary to confirm this. While miR-709 was predicted to increase significantly by the qPCR array, and the in situ hybridization confirms this increased expression, there is also an increase in the saline injected brain suggesting that while miR-709 may have a DHPG specific response, it is also likely to have a role in the endogenous stress response from either the handling or the surgery.
Previous studies used deep sequencing to identify novel microRNAs in mouse embryo, including the mature miR-709 sequence (Mineno, et al., 2006
). Target validation studies using luciferase reporter assays were used to examine the effect of the precursor form (mir-709), not the mature sequence, on a predicted target (Tamminga, et al., 2008
). These studies suggest that a mature miR-709 might exist, or that a unidentified mature microRNA resides within the precursor mir-709 sequence. However, the RNA blot studies from our lab and others (Tang, et al., 2007
) show abundant expression of precursor mir-709, but not mature miR-709, suggesting that a mature miR-709 as currently annotated in MirBase Version 16 does not exist. It is interesting to note that the annotated mature miR-709 sequence is located within 1 nucleotide of the 3′ end of the precursor sequence, which might have enabled erroneous detection of a ‘mature’ sequence by the qPCR array primers. Thus, further studies will be necessary to validate that a mature microRNA sequence is located within the precursor mir-709 sequence in adult mouse brain, and/or to examine the possibility that precursor microRNAs such as mir-709 can serve a functional role without further processing into a mature form.
We used the bioinformatics tool mirPath by DIANA (Papadopoulos, et al., 2009
) to perform in silico
validation of the microRNA predictions for each of the probe sets used in these experiments, separating out the microRNAs that increased from those that decreased. Here, we present KEGG pathways predicted by all three of the probe sets (Ambion, Invitrogen, and SABiosciences qPCR) at 8 hours after DHPG injection. In addition to the pathways familiar to neuroscientists, there were several cancer and pancreas/insulin/diabetes related pathways. A PubMed search revealed that mGluRs have been identified both in cancers within the brain (de Groot and Sontheimer, 2011
) and unassociated with nervous tissue (Li, et al., 2005
, Namkoong, et al., 2007
, Stepulak, et al., 2009
). Also unexpected was the role of mGluRs in the regulation of insulin secretion in the pancreas (Brice, et al., 2002
, Storto, et al., 2006
). These results illustrate the growing potential of using in silico
methods to validate the specificity of global assays. While this study is not translational, these results further illustrate the role of bioinformatics tools to anticipate potential unexpected drug interactions in the development of clinical therapeutics.
In conclusion, these studies show that DHPG regulates microRNAs in mouse brain, and support a potential role for mGluRs in the regulation of microRNA expression. These results warrant further studies to identify the pathways and mechanisms regulating changes in microRNA expression.