GABA receptors are well known as the inhibitory receptors in the central nervous system [¹,²]. However, GABA receptors are also found in several peripheral tissues [^3-6]. The functions of GABA receptors in peripheral tissues are less studied. They may be involved in ion homeostasis [⁷], cell proliferation and differentiation [⁸], development [¹], and hormone secretion [⁵,⁹].

We have initially identified GABA receptor π-subunit as a specific alveolar epithelial type II cell marker through DNA microarray analysis [¹⁰]. The expression pattern of the GABA receptor π-subunit is regulated by various culture conditions and is consistent with the type II cell phenotypes [¹¹]. We have further identified 19 subunits of the ionotropic GABA receptors in alveolar epithelial cells [⁶]. Their expression is dynamically changed during lung development [¹²]. Functionally, GABA receptors play important roles in fluid homeostasis in the adult lung and fetal lung development [⁶,¹³].

GABA receptors can be classified into two major types: GABA_A and GABA_C as ligand-gated Cl^-channels, and GABA_B receptor as a metabotropic receptor coupled to a heterotrimeric G-protein. GABA_A and GABA_C receptors share a conserved structure that contains a long extracellular N-terminal region, 4 transmembrane domains (TM1-TM4), a large intracellular loop between TM3 and TM4, and a short extracellular C-terminus [¹,²,^14-16]. The N-terminal segment is responsible for ligand binding and subunit assembly. The TM2 domain forms the lining of the ion pore. The intracellular loop is the site for post-translational modifications and binding with other proteins. This loop harbors a number of consensus phosphorylation sites for protein kinase A and C (PKA and PKC) and tyrosine kinases [¹⁷].

MicroRNAs are small non-coding RNAs. They form a ribonucleoprotein complex, termed RISC that cleaves mRNA or represses protein translation. MicroRNAs regulate various biological processes [¹⁸,¹⁹]. Several microRNAs such as miR-17-92 cluster and miR-127 are involved in lung development [^20-22]. MicoRNAs have also been implicated in many lung diseases including lung inflammation, Chronic Obstructive Pulmonary Disease, Asthma and Idiopathic Pulmonary Fibrosis [^23-30]. Nevertheless, microRNAs that regulate GABA receptors have not yet been reported. In this study, we used online software, TargetScan (http://www.targetscan.org) [³¹] and mRanda (http://www.microrna.org) [³²] to predict the microRNAs that possibly target to GABA receptors and then selected some of them to verify experimentally using 3'-UTR reporter assays. We found that miR-181, miR-23 and miR-216 target the GABA receptor α1-subunit.

Both GABA_A and GABA_C receptors are ligand-gated Cl^-channels. However GABA_C receptors have very unique ligand binding characteristics in comparison with GABA_A and GABA_B receptors, including a high sensitivity to the physiological ligand, GABA, insensitivity to bicuculline, barbiturates, and bacofen, very weak de-sensitization, a smaller single-channel conductance, and a longer open time [^14-16]. Eight different subunits of GABA_A and GABA_C receptors (α1-6, β1-3, γ1-3, δ, θ, ε, π, and ρ1-3) have been identified. The assembly of a heteropentamer, with at least one α-, one β-, and one γ-subunit, forms functional GABA_A receptor channels. δ-, θ-, ε-, and π-subunits can substitute for the γ-subunit. However, GABA_C receptors are exclusively composed of ρ subunits in the form of homo- or hetero-pentamers.

To identify the microRNA that may potentially regulate GABA_A and GABA_C receptors, we used the online computer software, TargetScan (v 5.2) to predict the binding sites of microRNAs on the 3'UTR of rat GABA receptors. We chose rat GABA receptors because we use rats as most of our animal or cell models. Since our microRNA expression vectors use human sequences, we only queried conserved microRNA target sites among mammals based on conserved 8 mer and 7 mer sites that match the seed sequence of a microRNA. The results are listed in Table 3. Among α-subunits, we found that α1 had most of the microRNA binding sites. There were 6 binding sites for 6 microRNAs on α1 3'-UTR. For other α-subunits, we found miR-128, miR-27ab, let-7/miR-98 for α6. We did not find any microRNAs for α2, α4, and α5. There was no information available for α3.

For β-subunits, we found two binding sites for miR-30a/30a-5p/30b/30b-5p/30/384-5p and one binding site for miR-103/107. There were 15 binding sites for 15 microRNAs on β2 and 5 binding sites for 7 microRNAs on β3. For γ-subunits, we found two binding sites on γ2 and no binding sites on γ1 and γ3, There was only one binding site for ε and no binding sites on π-, δ-, θ-, ρ1- and ρ2-subunits. In general, the "common" subunits (α, β, and γ) had more miRNA target sites than the "rare" subunits (δ, θ, ε, π, and ρ). This is probably because these subunits had shorter 3'-UTRs, in particular for ρ-subunits.

We also used another software, miRanda to predict the microRNA that target to GABA receptors (Table 3). In general, miRanda predicted more microRNAs than TargetScan. There were some common microRNAs that were predicted by both software. For example, miR-137, miR-181, miR-203, and miR-216a for α1; miR-103, miR-107, miR-30, and miR-384-5p for β1; and miR-204, miR-211, miR-23, and miR-26 for β3.

We further utilized a recently developed software, miRWalk [³³], to predict the miRNAs targeting 3'-UTR and open reading frame (ORF) of GABA receptors. The results are presented in Table 4. Obviously, this method is less stringent compared to TargetScan and miRanda, since the miRWalk query yielded 79 miRNAs for GABA receptor α1 subunit in comparison with only 6 by TargetScan and 29 by miRanda.

We selected two subunits, α1 and γ2 for experimental verification of the predictions. For α1-subunit, we constructed a 3'-UTR reporter vector, in which the 3'-UTR of α1-subunit was placed after a Renilla luciferase reporter gene (Table 2). For γ2-subunit, we cloned the predicted binding site of miR-103/107 into the downstream of a Renilla luciferase reporter gene. We then co-transfected a microRNA expression vector with the reporter into HEK293T cells to see whether the microRNA depressed the reporter activity. The firefly luciferase pGL3 vector was used for normalization. As shown in Figure 1, among the predicted microRNAs tested, miR-181, miR-216, and miR-203 inhibited the reporter activity. Four miR-181 isoforms, a-2, b-1, c and d-1 generated the same mature miR-181 and all of them depressed the reporter activity. All other microRNAs tested had no effects. The results suggest that miR-181, miR-216, and miR-203 are the micoRNAs that regulates GABA receptor α1-subunit. It is noted that miR-15, miR-16, miR-146, miR-221, miR-424 and miR-497 were predicated to target the GABA receptor α1-subunit by an earlier version (v.4.2) of TargetScan. Some of the miRNAs such as miR-181 are expressed in astrocytes naturally expressing GABA receptors [³⁴]. For γ2-subunit, we did not find major effects of miR-103-1, miR-103-2, and miR-107 on the reporter activity (Figure 2).

It should be noted that we did not measure miRNA levels in the miRNA-overexpressed cells. Thus, there are possibilities that some of miRNAs may not be over-expressed in the experimental set-up; particularly for these miRNAs that had no effect on 3'-UTR reporter activity. However, the transfection efficiency is 90-100% under our experimental conditions based on the GFP reporter expression encoded in the same miRNA expression vector. Additionally, the effect of a miRNA on the luciferase activity does not necessarily mean that it was a direct effect on the binding of a miRNA to the 3'-UTR reporter construct. A miRNA could have indirect effects. The mutations of seed sequences in the miRNA binding sites are needed to exclude indirect effects. Further studies are also needed to see whether the overexpression of miR-181, miR-203, and miR-216 in a physiologically relevant cell type modifies GABA receptor expression, and whether these miRNAs are differentially regulated in diseased states.

It is also interesting to note that miR-15b and miR-146a/b actually increased the 3'-UTR reporter activity. It has been reported that miRNA increases translation [³⁵]. However, it is also possible that this is a result of indirect effects.

We have previously shown that the activation of GABA receptors promotes fetal lung development [¹³]. The inhibition of miRNAs that target GABA receptors may increase receptor density and thus sensitivity of GABA receptors, which may benefit the development of therapy in treating diseases related to developmental anomalies.

In summary, computational approaches predict many microRNA binding sites on the 3'-UTR of GABA_A receptors, but not on these of GABA_C receptors. 3'-UTR reporter assays only verified miR-181, miR-216, and miR-203 as the microRNAs that target GABA receptor α1-subunit among 10 microRNAs tested. These studies reinforce that micoRNA target prediction needs to be verified experimentally. The identification of microRNAs that target to GABA receptors provides a basis for further studies of post-transcriptional regulation of GABA receptors.

The authors declare that they have no competing interests.

CZ, TW and HM carried out experiments. CZ, CH and XX analyzed data and performed target predictions. LL conceived of the study, and participated in its design and coordination. LL and CZ drafted the manuscript. All authors read and approved the final manuscript.

We thank Ms. Tazia Cook for editorial assistance. This work was supported by the National Institutes of Health, HL-087884 and HL-095383 and OCAST, AR101-037.