https://orcid.org/0000-0003-3270-5526
Abstract
Human disease-associated genetic variations often map to long non-coding RNA (lncRNA) genes; however, elucidation of their functional impact is challenging. We previously identified a new genetic variant rs4454083 (A/G) residing in exon of an uncharacterized lncRNA ARBAG that strongly associates with plasma levels of C-peptide, a hormone that regulates insulin bioavailability. On the opposite strand, rs4454083 also corresponds to an intron of a cerebellum-specific GABA receptor subunit gene GABRA6 that mediates strengthening of inhibitory synapses by insulin. Here, we show that alleles of rs4454083 modulate transcript levels of the antisense gene, ARBAG, which then controls the expression of the sense gene, GABRA6. Predisposing to low C-peptide, GG (a minor allele genotype across ethnicities) stabilizes ARBAG lncRNA causing higher transcript levels in cerebellum. ARBAG lncRNA abundance leads to cleavage of GABRA6 mRNA at the complementary region, resulting in a dysfunctional GABRA6 protein that would not be recruited for synapse strengthening. Together, our findings in human cerebellar cell-line and induced Pluripotent Stem Cells (iPSCs) demonstrate biological role of a novel lncRNA in determining the ratio of mRNA isoforms of a protein-coding gene and the ability of an embedded variant in modulating lncRNA stability leading to inter-individual differences in protein expression.
Introduction
C-peptide, which is a marker of insulin secretion, is tested as a measure of pancreatic β-cell reserve. Once considered a junk byproduct of insulin synthesis, C-peptide is now recognized as a hormone that functions in a variety of tissue systems (1–5). Recent studies show that its circulatory levels serve as an early indicator of multiple metabolic diseases in humans, e.g. type 2 diabetes (T2D), cancer, non-alcoholic fatty liver disease, etc. (6–9). Despite its important biological role, genetics of C-peptide is unknown. In an endeavor to study T2D heritability, our group earlier did a genome-wide association study (GWAS) of C-peptide in Indians (10), a population genetically predisposed to T2D (11). In that study, we identified a novel variant rs4454083 (A/G) in the intron of a GABA receptor gene GABRA6 that was strongly associated with fasting C-peptide levels measured in the plasma samples of 877 healthy Indians (P = 8.25 × 10−6; effect = −0.25); the finding was later replicated in an independent set of 1829 Indians. The identified GABRA6 intronic variant simultaneously resided in an exon of a previously unannotated long non-coding RNA (lncRNA; ENST00000521984) that maps to the same location on the opposite strand antisense to GABRA6; henceforth the lncRNA is referred as ARBAG. In the current study, rs4454083 was fine-mapped and functionally characterized to determine the biology underlying its genetic association with C-peptide. We show that the presence of minor allele genotype (GG) that associates with lower C-peptide levels, induces overexpression of ARBAG that further results in a non-functional GABRA6 protein in human cerebellar cell-line and induced Pluripotent Stem Cells (iPSCs). At a molecular level, we illustrate that the ARBAG lncRNA is involved in post-transcriptional regulation of GABRA6 mRNA, cleaving it at the junction between the regions that code for ligand binding and trans-membrane domains. By affecting the stability of lncRNA, rs4454083 results in altered GABRA6 mRNA isoforms and consequently inter-individual variability in protein levels.
Results
C-peptide variant rs4454083 influences ARBAG expression in human cerebellum
Single nucleotide polymorphism (SNP) rs4454083, which is exonic to ARBAG and intronic to GABRA6, resides in an evolutionary conserved genomic region (Fig. 1A). This variant is very common with a minor allele (G) frequency of 0.27 in Indians, which is comparable with other ethnicities in the world (Fig. 1B). We did imputation to identify additional variants in the associated locus correlated by linkage disequilibrium (r2 > 0.80) (Fig. 1C). Association analysis of 5265 imputed variants with C-peptide revealed 11 variants flanking ARBAG that exhibited similar association like rs4454083 (Fig. 1D, Supplementary Material, Fig. S2 and Supplementary Material, Table S2). No new variant was observed in the ARBAG gene. Detected variants were examined for gene regulatory signatures using ENCODE database (12). Barring rs4454083 that nested in a highly conserved genomic region characterized by gene enhancer mark H3K4Me1, all other SNPs resided in desert regions devoid of any regulatory signatures (evolutionary conservation, histone marks, DNase-I hypersensitivity, TF binding, miRNA sites, etc.). Therefore, rs4454083 was selected for a functional follow-up.
Fine-mapping the C-peptide associated locus harboring variant rs4454083. (A) SNP simultaneously resides in an intron of GABRA6 and an exon of ARBAG on chromosome 5 in humans and forms a highly conserved element of human genome (UCSC genome browser). lod: transformed log-odds score equal to log probability of an element under the conserved model minus its log probability under the non-conserved model. (B) Frequency of A and G alleles of rs4454083 in different populations in the world. IND: Indians (current study); EUR: Europeans (1000Genomes); AFR: Africans (1000Genomes); AMR: Americans (1000Genomes); EAS: East Asians (1000Genomes). (C) Study design of the imputation of rs4454083 locus. (D) Schematic showing the genomic location of rs4454083 with respect to GABRA6 and ARBAG genes. Gray pinheads mark the position of the imputed SNPs that showed association with C-peptide similar to rs4454083 in Indians.
To verify our imputation results, we sequenced ARBAG exons in 37 Indians (ARBAG has two exons). For ARBAG exon 1, no variation was observed among rs4454083 –AA genotyped people (N = 21) and GG genotyped people (N = 16) and, for exon 2, these groups differed only at rs4454083 and a downstream indel that constituted deletion in all GG individuals except one person (Fig. 2A, Supplementary Material, Fig. S3). People carrying the GG genotype showed lower plasma C-peptide levels than people who had AA genotype (Fig. 2B). To identify genes that are affected by rs4454083, we examined the expression quantitative trait locus (eQTL) data for various human tissues provided in the GTEx database (13). Rs4454083 was found to be a significant cis-eQTL in human cerebellum for only two genes—GABRA6 and ARBAG (Fig. 2C, D); however, the association was much stronger for ARBAG where the individuals with GG genotype presented higher ARBAG expression than AA individuals (P = 7.0 × 10−18). This eQTL association was not seen in any other human tissue (Supplementary Material, Fig. S4).
SNP 4454083 influences GABRA6 and ARBAG expression in human cerebellum. (A) Validation of imputation results by Sanger sequencing in Indians displaying a region of sequenced exon 2 of ARBAG that was different in 21 AA and 16 GG homozygous individuals. Ref: reference sequence of human genome. (B) Levels of C-peptide in plasma samples of the sequenced AA and GG individuals showing low C-peptide in GG homozygotes. The whiskers in boxplot indicate the range of data and the bar denotes the median C-peptide levels. (C) eQTL association data from GTEx database (version 8), indicating that alternate alleles of rs4454083 are associated with variable GABRA6 transcript levels in human cerebellum. GG, minor allele genotype, AA, major allele genotype in Indians. The violin plot depicts the density of gene expression levels in each genotype and the bar denotes the median values. N = number of tissue samples for each genotype. P = nominal P-value generated for variant-gene pair by testing whether that the slope of a linear regression model between genotype and expression deviates from 0. (D) Rs4454083 is a stronger eQTL for ARBAG as opposed to GABRA6 in human cerebellum (GTEx portal v8).
ARBAG is a stable lncRNA likely involved in gene expression regulation
ARBAG is a novel lncRNA with unknown function. To understand its biological consequence on GABRA6 expression, ARBAG lncRNA was functionally characterized in the human cerebellar cell-line DAOY, chosen to correspond the brain region that expresses ARBAG. We checked its binding with the cellular translation machinery by immuno-precipitation using an antibody against the 40S ribosomal subunit –RPS12. Unlike β-actin, ARBAG was not detected in the pulled-down fraction though it was present in the input fraction (Fig. 3A), implying that ARBAG lncRNA is truly non-coding and does not translate to any peptide. To check its cellular stability, we first examined polyadenylation (Poly-A tail protects RNA from degradation) and, ARBAG was polyadenylated as determined by its amplification after oligo dT-primed synthesis of cDNA from human cerebellum tissue RNA (Fig. 3B). Next, we calculated its half-life following global inhibition of transcription using actinomycin D. The controls, β-actin and GABRA6 mRNAs revealed half-lives of nearly 10 and 7 h, respectively (Supplementary Material, Fig. S5). In comparison to lncRNAs in general, ARBAG was a stable lncRNA with a half-life of 3 h 44 min (Fig. 3C). Later, we studied cellular localization using RNA-FISH and, ARBAG RNA puncta were clearly seen in cytoplasm encircling the nucleus of human cerebellar cells (Fig. 3D), denoting that ARBAG is a cytoplasmic lncRNA and thus might regulate the expression of its target gene by influencing mRNA stability, translation or competing for miRNA binding. Localization results were validated by ARBAG transcript amplification in the cytoplasmic fraction following cellular fractionation.
Functional characteristics of ARBAG lncRNA. (A) ARBAG-ribosome complex pulldown assay done by using antibody against RPS12 (a component of 40S ribosomal subunit) in human DAOY cells revealed absence of ARBAG in the pulled-down fraction. β-Actin was used as a control. Input was taken as 10% of the pulldown fraction. Data shown for two independent experiments. (B) ARBAG is polyadenylated as determined by amplification following oligo-dT primed synthesized cDNA from human cerebellar tissue RNA. M: 10- bps DNA ladder; NTC: no template control. (C) Decay assay of ARBAG lncRNA in DAOY cells showing abundance of ARBAG transcripts at 2 and 4 h relative to 0 h after global transcription inhibition using 10-μg actinomycin D. Experiment was repeated on two independent days, each with three replicates. (D) RNA-FISH images indicating the localization of ARBAG lncRNA (red) in the cytoplasm of human DAOY cells. Nuclei were counterstained using DAPI (blue). Scale bar = 50 μm. The localization results were confirmed by amplification of ARBAG in RNA obtained from the cellular fractionation of DAOY cells.
ARBAG lncRNA co-expresses and co-localizes with GABRA6 mRNA in cerebellum
Investigation of gene expression in different human tissues showed that both GABRA6 and ARBAG are strongly expressed in brain, especially cerebellum (Fig. 4A). The results were quantified by realtime-PCR in various brain tissues. Relative to blood, GABRA6 showed more than 700-fold higher expression in cerebellum, whereas ARBAG expression was nearly 10-folds higher (Fig. 4B), which is the highest expression of the two genes in humans. This cerebellum-specific enrichment of gene expression is consistent with the data reported in the GTEx portal (13) that also contains samples from different regions of brain (Supplementary Material, Fig. S6). In a tissue sample of human cerebellum (the deceased was Indian), we noticed that ARBAG is expressed in both layers of the cerebellar cortex- the outer molecular and the inner granular layer (Fig. 4C). Further, we examined localization of GABRA6 and ARBAG transcripts in cerebellar cells using RNA-FISH and both the molecules were seen to co-localize in cytoplasm (Fig. 4D). To confirm co-localization, we did a 3D analysis of the co-localized GABRA6 and ARBAG puncta using stacked 2D images obtained from RNA-FISH experiments and the union was observed to be > 70% (Supplementary Material, Fig. S7).
ARBAG correlates with the presence of GABRA6 transcripts. (A) Analysis of ARBAG and GABRA6 gene expression in 21 different human tissues revealed co-expression in human cerebellum. The gel shows RT-PCR results. Tested tissues were purchased from a commercial source. GAPDH mRNA levels are shown as a control for tissue panel validation. M: 100-bps DNA ladder. (B) Real-time PCR in different tissues of human brain showed significant enrichment of ARBAG and GABRA6 expression in cerebellum. Gene expression is shown relative to whole blood. Data is from three replicates, presented as mean + SD. (C) RNA-FISH in human cerebellum tissue section indicated prominent expression of ARBAG (red) in both the molecular and the granular layers. Nuclei were counterstained using DAPI (blue). Scale bar = 50 μm. (D) RNA-FISH images presenting the co-localization of ARBAG lncRNA (red) and GABRA6 mRNA (cyan) in the cytoplasm of human cerebellar cells. Nuclei were counterstained using DAPI (blue). Scale bar = 10 μm.
Loss of ARBAG expression decreases GABRA6 mRNA levels
SiRNA mediated knockdown of ARBAG expression in cerebellar cells resulted in a strong fall in GABRA6 mRNA levels to 70% (Fig. 5A). To explore the dynamics of this effect further, we monitored GABRA6 mRNA at various times after ARBAG knockdown. Levels of GABRA6 mRNA closely followed ARBAG lncRNA levels at different timepoints (Fig. 5B). These findings indicated a possible link between the two RNAs. To our surprise, when ARBAG was knocked down, expression of GABRA6 isoforms was altered. Full-length GABRA6 isoform, monitored using transcripts across exons 5, 6 and 7 (i.e. isoform 5 to 6–7) was considerably high in ARBAG knocked-down cells compared to the control set (Fig. 5C), whereas other isoforms spanning only exon 4, exons 4 and 5 (isoform 4–5) or exons 6 and 7 (isoform 6–7) were detected in both sets.
ARBAG knockdown reduces GABRA6 transcript levels. (A) siRNA mediated knockdown of ARBAG lncRNA in human DAOY cells decreases GABRA6 transcript abundance. Experiment was repeated on three independent days where each experiment was repeated three times. Data are shown as mean + SD. P-value was calculated using Student’s t-test. *P < 0.05. (B) Time-dependent analysis of gene expression following ARBAG knockdown in DAOY cells. GABRA6 mRNA levels closely follow ARBAG levels at various time-points. Data are from two independent experiments. The whiskers in boxplot indicate the range of data and the bar denotes the median fold change. The P-value was calculated using Student’s t-test. *P < 0.05; **P < 0.002. (C) Schematic shows GABRA6 exons and amplicons corresponding to various transcript isoforms. RT-PCR gel shows the presence/absence of different amplicons in ARBAG knocked-down and control samples 24 h post-transfection. Full-length GABRA6 transcripts (5 to 6–7 amplicons) were seen in cells where ARBAG expression was knocked down compared to cells transfected with scrambled siRNA. Amplicons from other regions were observed in both sets. Data are shown for three experiments done on independent days.
Overexpression of ARBAG cleaves full-length GABRA6 mRNA
To validate the findings of our knockdown study, we transiently overexpressed ARBAG in human cerebellar cells (Fig. 6A) and searched for various isoforms of GABRA6 mRNA. When ARBAG was overexpressed, only shorter GABRA6 isoforms spanning exons 4–5 or 6–7 were detected (Fig. 6B). No full-length GABRA6 isoform carrying exons 4–5 to 6–7 or 5 to 6–7 was seen in the overexpression samples though the control set showed both full-length and shorter isoforms of GABRA6 transcripts. Investigation at the protein level also confirmed a cleavage of GABRA6 protein in ARBAG overexpression samples (Fig. 6C). Overexpression studies described here were done using a construct that harbored A-allele of variant rs4454083. We constructed a similar plasmid with G-allele of the SNP, overexpressed in DAOY cells (Fig. 6D) and checked various splice-isoforms of GABRA6 mRNA. Irrespective of the allele overexpressed, the lncRNA always led to the cleavage of GABRA6 mRNA (Fig. 6E). This implied that the SNP does not affect cleavage of GABRA6 mRNA directly.
ARBAG overexpression cleaves full-length GABRA6 mRNA. (A) Transient overexpression (OE) of ARBAG lncRNA construct harboring A-allele in human cerebellar cells. Data are shown as mean + SD. (B) ARBAG OE results in cleavage of GABRA6 mRNA. RT-PCR gel demonstrates the absence of full-length GABRA6 transcript (4–5 to 6–7 amplicons) in ARBAG OE samples in comparison with the empty-vector transfected cells. Amplicons 5 to 6–7 were very weakly expressed in ARBAG OE samples. Other isoforms (exons 4–7) were detected in both sets. Results from three independent experiments have been shown. (C) Western blot following ARBAG OE detects two GABRA6 isoforms at protein level in DAOY cells. β-Actin was used as a loading control. (D) Transient OE of ARBAG lncRNA construct harboring G-allele in human cerebellar cells. Data are shown as mean + SD. (E) RT-PCR gel denoting cleavage of full-length GABRA6 mRNA after overexpressing ARBAG construct carrying the G-allele of rs4454083 in cerebellar cells (like A-allele ARBAG OE results). Results are from two experiments done on independent days. (F) Analysis of GABRA6 protein domains was performed before and after cleavage by ARBAG using NCBI Conserved Domain tools. Full-length GABRA6 protein (453 amino acids) has two conserved domains—a ligand binding and a transmembrane. Cleavage separates the ligand binding domain and the trans-membrane domain. (G) DAOY cells feature presence of both A and G-alleles at rs4454083 locus at DNA level (genomic PCR gel). Alleles were detected through ARMS-PCR using allele-specific primers. Results from 3 independent experiments (expt. 1, 2, and 3) have been shown. Quantification by realtime PCR indicated higher abundance of A-allele than G-allele, reflecting a greater number of AA cells than GG cells. M: 100-bps DNA ladder; NTC, no template control. (H) At the RNA level, only G-allele ARBAG transcripts are seen endogenously. No A-allele transcripts were detected, suggesting decreased cellular stability of A-allele transcripts though both A and G alleles exist at the DNA level. The data 1, 2, 3 represent three independent experiments. M: 100-bps DNA ladder; NTC, no template control.
Putting together the results from overexpression and knockdown of ARBAG lncRNA, we found that the lncRNA could drive the cleavage of GABRA6 mRNA into two transcripts encompassing exons 1–4 and exons 5–9. Presence of multiple GABRA6 transcript isoforms was verified in human cerebellum tissue (Supplementary Material, Fig. S8). ARBAG lncRNA is complementary to a stretch of 77 nucleotides within exon 5 of GABRA6 mRNA (Supplementary Material, Fig. S9) and the cleavage site appears to be at/around the region of complementarity between the two RNAs. GABRA6 protein comprises two key domains—a ligand-binding domain encoded by exons 1–4 and a trans-membrane domain spanning exons 5–9. Cleavage of GABRA6 mRNA at the observed location separates ligand binding domain from trans-membrane domain (Fig. 6F), resulting in a non-functional GABRA6 protein.
G-allele leads to greater ARBAG overexpression
In our overexpression studies, we noticed that G-allele plasmids resulted in higher ARBAG expression than A-allele plasmids (Fig. 6A and D, Supplementary Material, Fig. S10). To rule out any possibility of varying concentrations of plasmids being transfected in vitro, we performed a dose-dependent overexpression (Supplementary Material, Fig. S11). A-allele plasmids consistently featured lesser ARBAG expression irrespective of the amount of transfected plasmid (transfection efficiency of both two plasmids was identical as checked by eGFP co-expression). The results thereby indicate that A-allele ARBAG lncRNAs are relatively less stable in-vitro than G-allele transcripts. Here, it is worthy to mention that G-allele of rs4454083 (a minor allele in Indians and other populations) possibly leads to an alteration in secondary structure of ARBAG lncRNA (Supplementary Material, Fig. S12).
We verified this allelic specificity in ARBAG expression endogenously in DAOY cells. At the DNA level, DAOY cells were observed to possess both alleles of rs4454083—A and G (Fig. 6G). At the RNA level, only G-allele transcripts of ARBAG were present (Fig. 6H). Consistent with our allele-specific over-expression studies (Fig. 6A and D, Supplementary Material, Fig. S11), these findings imply that transcripts carrying G allele are significantly more stable (survive better) than A-allele transcripts and, an abundance of lncRNA transcripts perhaps, promotes cleavage of GABRA6 mRNA. These results explain the direction of the observed eQTL association of this variant with ARBAG levels in human cerebellum (Fig. 2D).
Next, we looked for commercial human iPSC lines with different genotypes of rs4454083 and found that the DYR iPSC line is homozygous for the major allele (AA genotype, Fig. 7A). Like DAOY cells, the DYR line endogenously expressed ARBAG and GABRA6. The GABRA6 mRNA included exons 4–5 and exons 6–7 but the product containing exons 4–5 to 6–7 could not be detected (data not shown). We overexpressed A- and G-allele ARBAG in DYR iPSCs and examined the presence of various GABRA6 transcript isoforms. In agreement with DAOY cells, the expression of G-allele and A-allele ARBAG constructs in iPSCs differed significantly (Fig. 7B), implying that the instability conferred by rs4454083 to the ARBAG A-transcript expression is independent of cell type. Further, amplicons spanning GABRA6 exons 4, 4–5 were detected in all the samples that were transfected with A-, G-allele ARBAG or the empty vector, however the amplicons constituting exons 6–7, 4–5 to 6–7 and 5 to 6–7 were undetected (Fig. 7C). We speculate that rs4454083 allele-specific GABRA6 cleavage is restricted to cerebellar cells, whereas iPSCs express a truncated isoform of GABRA6 mRNA.
ARBAG instability conferred by rs4454083 exists in human DYR iPSCs. (A) Chromatogram demonstrates Sanger sequencing peaks indicating the presence of homozygous AA genotype for rs4454083 in the DNA of human DYR iPSCs. (B) Transient OE of ARBAG lncRNA construct carrying rs4454083 alleles in human iPSCs confirms existence of higher levels of G-allele ARBAG transcripts than A-allele transcripts upon transfection. Data are shown as mean + SD. Results from four independent experiments have been shown. (C) RT-PCR gel shows the presence of exon 4 (marked by an arrow) and the amplicon 4–5 endogenously and upon ARBAG OE (A- or G-allele). The other isoforms are inconsistently detected indicating a truncated GABRA6 mRNA in both sets (transfected and control). The experiment was repeated on four independent days. The data have been shown for two experiments (1 and 2). M: 100-bps DNA ladder.
Discussion
Though numerous GWASs have been conducted so far but only a handful of the identified associations are understood functionally. Here, we established the biological role of a genetic variant for C-peptide (rs4454083), previously identified by our group (10), residing in an uncharacterized novel lncRNA gene ARBAG whose alternate alleles modulate the stability of the transcribed lncRNA leading to an altered regulation of a key protein gene—GABRA6.
GABRA6 is the α6 subunit of A-type receptors (GABAA) of the major inhibitory neurotransmitter GABA (14). Its expression is highly restricted to human cerebellum where it pairs with other subunits at synapse to causes a transient, rapidly desensitizing GABAergic conductance (phasic inhibition), or extra-synaptically to set a persistent GABAergic conductance (tonic inhibition) (14). Targeted disruption of GABRA6 in mice causes loss of tonic conductance (15), which, in turn, triggers a form of homeostatic plasticity altering the magnitude of a voltage-independent potassium conductance that preserves normal neuronal behavior (16). Genetic studies in humans reveal significant association of GABRA6 with mental illness and neuropsychiatric diseases including stress, anxiety, depression, autism (17–20), supported by functional studies (21–23). For metabolic disorders, so far, a sole study in humans suggestively linked GABRA6 with obesity (24); however, GABRA6 has never been implicated in T2D or related traits.
The mechanistic link between GABRA6 function and alteration in C-peptide levels is unknown. Case studies in children with autism and behavioral disorders have shown that intranasal administration of C-peptide improves social skills, cognition, and betters anger management (these phenotypes are linked to GABRA6) (25). C-peptide dictates the bioavailability of monomeric insulin needed to activate insulin receptor (26,27). In diabetic rats, intranasal administration of C-peptide stimulates insulin signaling in hypothalamus (28). Insulin is known to strengthen inhibitory GABAergic synapses in mouse cerebellar cells by recruiting GABRA6 receptors to postsynaptic sites (29). C-peptide concentration in human brain is greater than insulin (30) and as C-peptide potentiates insulin regulation (27), we believe that variable C-peptide levels might hamper insulin mediated recruitment of GABRA6 receptors in cerebellum, a mechanistic basis for observed genetic association of a C-peptide variant with GABRA6 gene.
In the present study, eQTL data was instrumental in speculating a credible role of the novel lncRNA ARBAG to propagate the functional effects of SNP rs4454083 in regulating GABRA6 expression. Individuals with minor allele genotype (GG) showed significantly low C-peptide levels in plasma and elevated ARBAG expression in cerebellum relative to the AA people. We validated the eQTL results in-vitro in a human cerebellar cell line (DAOY) and an iPSC line (DYR) and noticed that the G-allele disrupts the functional GABRA6 protein needed to maintain tonic conductance in cerebellum. A genetic study in Finns has associated rs4454083 with neuropathy in the advanced stages of T2D (P = 0.002, odds-ratio = 1.23) where G allele predisposes to a higher risk of neuropathy (31). Comparing the findings in DAOY and DYR cells, we think that the cleavage of GABRA6 transcript is unique to cerebellum, however the allele-specific influence of variant rs4454083 on GABRA6 cleavage might be universal to ARBAG, though the effect is particularly relevant in cerebellar cells. Rs4454083 exhibits similar allele frequency across ethnicities, we suspect strong homogeneity in allelic effect at ARBAG locus. Till date, only few previous studies have suggested an involvement of SNPs in regulating lncRNA stability (32,33). SNPs in lncRNA gene or promoter can influence the turnover and stability of the transcribed lncRNA by modifying its native miRNA binding sites, creating an endonuclease site, damaging splice sites or altering the binding of transcription factors/proteins that determine stability (34); we suspect that the ARBAG destabilizing A-allele might lead to one of such outcomes. As this instability was seen in cerebellar cells and iPSCs, the underlying mechanism may not require a cerebellum specific factor, although GABRA6 and ARBAG are specifically expressed in the cerebellum.
LncRNAs regulate gene expression by diverse mechanisms (34–36), and we identified an lncRNA that presides over the transcript isoforms of an important protein coding gene. RNA cleavage is determined by RNA–protein interactions of splicing factors with gene regulatory elements, RNA–RNA base-pairing interactions, or chromatin architecture that governs splicing patterns (37). Some earlier studies have elucidated a direct regulation of alternative splicing by non-coding RNAs. The brain specific snoRNA HBII-52 regulates alternative splicing of the serotonin receptor 2C mRNA by binding to a splicing silencer element in Prader Willi Syndrome (38). A conserved nuclear antisense lncRNA from within human FGFR2 locus establishes splicing specific chromatin histone marks to govern alternative splicing of FGFR2 and hence regulate epithelial-to-mesenchymal transition (39). Insertion of an Alu element in human 5S rRNA overlapping lncRNA (5S-OT) leads to primate specific modulation of alternative splicing of multiple target genes in trans via Alu/anti-Alu pairing (40). The ratio of two overlapping isoforms (α1, a truncated isoform, and α2, a full-length isoform) of rat c-erbA mRNA was observed to depend upon expression of antisense protein coding RNA that complements base-pairing with α2 isoform and inhibit its splicing (41). Here, we propose that the newly discovered lncRNA ARBAG may play such a role in cerebellar cells, altering the ratio of the ligand binding and full-length isoforms of GABRA6.
To summarize, in this study exclusively performed on human subjects, iPSCs and human cerebellar cell-line, we demonstrate the functional role of the variant rs4454083 genetically associated with C-peptide in deploying antisense lncRNA ARBAG to regulate a crucial neurotransmitter receptor subunit gene GABRA6. Going from genetics to function, the current study provides a systematic framework to translate disease associations to clinically actionable gene sets.
Materials and Methods
Imputing variant rs4454083
The locus harboring SNP rs4454083 was imputed by engaging the default parameters of Impute2 program (42) using 1000Genomes Phase3 data as reference (43). A region of 1 Mb on either side of the SNP was prephased into haplotypes using Shapeit (44). Investigated region included the complete LD block of rs4454083 and the ARBAG gene. Imputation accuracy was checked by masking the genotyped SNPs and, the concordance was > 98%. Plasma C-peptide levels were measured by electrochemiluminescence immunoassay using Elecsys 2010 (Roche Diagnostics, IN, USA) as described previously (10). Imputed variants were tested for association with C-peptide using PLINK (http://pngu.mgh.harvard.edu/purcell/plink/) adjusting for age, sex, body mass index and first three principal components. Imputation results were validated by sequencing of ARBAG exons in 37 Indians (rs4454083 AA genotyped: 21, GG genotyped: 16) using ABI Biosystems 3730/3730xl DNA Analyzer (ABI Biosystems, CA, USA). Sequences of all the primers used in the study are given in Supplementary Material, Table S1.
Cell culture
Human DAOY cell-line was obtained from American Type Culture Collection (ATCC, VA, USA) and maintained in MEM medium (Gibco, ThermoFisher Scientific, MA, USA) supplemented with 10% Fetal Bovine Serum (Gibco) at 37 °C in a humidified incubator (5% CO2). The cell-line was authenticated by STR profiling (Lifecode Technologies, New Delhi, India) and tested to be mycoplasma-free by PCR (ATCC). Human DYR iPSCs (ATCC) were a kind gift from Dr Sivaprakash Ramalingam (CSIR-Institute of Genomics and Integrative Biology, New Delhi, India). The iPSCs were cultured in StemFlex medium (Gibco) at 37 °C.
ARBAG lncRNA pulldown
One million DAOY cells were cultured in a T-25 flask in serum-free medium. Next day, to inhibit protein synthesis, cells were treated with 100-μg cycloheximide (Sigma-Aldrich, MO, USA) per ml of medium at 37 °C. After 10 min, cells were scrapped in ice-cold DPBS (Gibco), pelleted and resuspended in cold lysis buffer (100-mM KCl, 5-mM MgCl2, 10-mM HEPES {pH 7.0}, 0.5% NP-40, 1-mM DTT, 100 U/ml RNase OUT, Protease Inhibitor Cocktail {Roche}) and then spun at 16 000 g (4 °C; 15 min) to collect lysate. For immunoprecipitation, 1.5-mg Dynabeads Protein G (Novex, UA, USA) were incubated with 1.5-μg anti-RPS-12 antibody (Abcam, Cambridge, UK) in NT2 buffer (50-mM Tris {pH 7.4}, 150-mM NaCl, 1-mM MgCl2, 0.05% NP-40) at 4 °C for 2 h and then incubated with 900 μL of cell lysate overnight at 4 °C. 10% of lysate (100 μL) was used as input fraction. Next day, the complex was digested with 10 U DNase I (37 °C; 10 min) (NEB, MA, USA) and eluted with 0.1-M glycine (pH 2.9). Beads were discarded and supernatant was treated with 50-μg proteinase K (55 °C; 45 min). RNA was purified from the pulled-down fraction and the input using TRIzol (Ambion, TX, USA), cDNA was synthesized using random hexamers (Invitrogen, ThermoFisher Scientific) and lncRNA presence was analyzed by PCR.
Polyadenylation
cDNA was synthesized from 1-μg human cerebellar tissue RNA using Oligo(dT)20 primer (Invitrogen) and M-MuLV reverse transcriptase at 42 °C for 1 h. ARBAG was detected by PCR.
RNA decay
0.2 million DAOY cells were treated with 10 μg Actinomycin D (Sigma-Aldrich) and harvested at 0, 2, 4, 6, 8, 10 and 12 h post-treatment. Control-set was treated with DMSO (Sigma-Aldrich). Transcript levels were measured by real-time PCR using SYBR Green (Roche). Half-life of an RNA (T1/2) is inversely proportional to its decay rate constant (Kdecay). Kdecay was obtained from the slope of a semilogarithmic plot of RNA concentration as a function of time.
RNA-FISH
FISH probes for GABRA6 mRNA (CAL Flour Red-610) and ARBAG lncRNA (Quasar-570) were synthesized commercially (Biosearch Technologies, WI, USA). Probes spanned entire RNA length except their complementary region (77 bases). Experiment was conducted as per manufacturer’s protocol. Confocal images were acquired in sequential mode at 63× magnification using Leica TCS SP5 system (Leica Microsystems, IL, USA). For ARBAG: excitation (ex) 543 nm, emission (em) 550–616 nm; GABRA6: ex 594 nm, em 607–701 nm; DAPI: ex 405 nm, em 415–504 nm.
Cellular localization
At a 80–90% confluency, DAOY cells were washed with cold DPBS, pelleted and treated with three volumes of cold Buffer A (10-mM KCl, 1.5-mM MgCl2, 20-mM Tris–HCl {pH 7.5}) and incubated in ice for 10 min. Cellular fractionation was done using dounce homogenizer. Homogenate was constituted to be 0.1% Triton-X-100 (Sigma-Aldrich). Nuclei were pelleted by centrifugation at 1500 g for 5 min. Cytosolic fraction (supernatant) was separated. RNA was extracted by TRIzol method for the nuclear fraction and by phenol-chloroform method for the cytosolic fraction.
Human tissue expression
From an RNA panel of various human tissues (Clonetech, CA, USA), 1-μg RNA was reverse transcribed to cDNA. Desired transcripts were analyzed by PCR. Gene expression in different regions of human brain was quantified by real-time PCR.
ARBAG knockdown
0.2 million DAOY cells were seeded per well in a six-well plate. Following day, 40 nM ARBAG siRNA (Invitrogen) was transfected using FUGENE HD (Promega, WI, USA) in Opti-MEM (Gibco); transfection medium was replaced with complete growth medium after 12 h. Cells were harvested 24 h post-transfection and RNA was extracted. For time-dependent knockdown assay, cells were harvested at 6, 12 and 24 h post-transfection. The experiment was done on two independent days (each repeated three times). SiRNA sequences are provided in Supplementary Material, Table S1.
ARBAG overexpression
ARBAG constructs harboring rs4454083 alleles (A/G) were synthesized by gene-assembly PCR using pairs of overlapping primers (alleles were confirmed by sequencing, Supplementary Material, Fig. S1) and were cloned downstream of CMV promoter in pcDNA 3.1(+) vector (the constructs differed only at rs4454083, rest of their sequence was the same). To overexpress, each well of a six-well plate was seeded with 0.2 million DAOY cells followed by transfection with 2-μg plasmid (ARBAG or empty vector) the next day. Transfection medium was replaced with growth medium after 5 h, and cells were harvested 24 h post-transfection. RNA was extracted, cDNA synthesized and different exons of GABRA6 mRNA were detected by PCR. Experiment was performed on three independent days (each repeated three times).
The iPSCs were harvested using accutase solution (Gibco) and 1.5 million cells were transfected with 1.5-μg allele-specific ARBAG constructs using The Neon™ Transfection System 10-μL Kit (Invitrogen) engaging the pulse settings 1400 V, 20-ms pulse width and 1 pulse. Following transfection, cells were seeded into a 35-mm dish for 48 h and subsequently analyzed for GABRA6 isoforms. RNA was isolated using RNAiso Plus (Takara) and cDNA was synthesized from 1 μg of total RNA followed by RT-PCR using SYBR Green master mix (Genetix).
Western blotting
Total protein was harvested from DAOY cells using RIPA buffer (10-mM Tris-Cl (pH 8.0), 1-mM EDTA, 0.5-mM EGTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140-mM NaCl, 1-mM PMSF) and the concentration was estimated by Bradford method. Proteins were separated by SDS-PAGE and probed with respective antibodies. Detection was done using chemiluminescence.
Detection of allele-specific ARBAG DNA and RNA
In DAOY cells, DNA was isolated using the standard salt precipitation method. Alleles in ARBAG DNA or RNA were detected by ARMS PCR using allele-specific forward primers and a common reverse primer for amplification. In iPSCs, DNA was isolated using a commercial kit (Qiagen) and Sanger sequencing was done to examine rs4454083 genotype.
Acknowledgements
We acknowledge the core imaging facility (funded through CSIR project BSC0403) of CSIR-Institute of Genomics and Integrative Biology. We thank Dr Sivaprakash Ramalingam (CSIR-Institute of Genomics and Integrative Biology, New Delhi, India) for DYR cells and guidance in iPSC experiments. We appreciate the help of Dr Yasmeen Kauser in the ARBAG ribosome complex pull-down experiment. We also thank Dr Sivaram V S Mylavarapu, Associate Professor, Regional Centre for Biotechnology, for enabling confocal imaging for certain RNA-FISH experiments.
Conflict of interest statement. None declared.
Funding
Indian Council of Medical Research (ICMR) grant entitled ‘Non-coding Variants and Fine Tuning of Childhood Obesity Genes’; Council of Scientific and Industrial Research (CSIR), Government of India (grant projects BSC0122 (16) and MLP2008); Department of Science and Technology, Government of India (PURSE II CDST/SR/PURSE PHASE II/11 to Jawaharlal Nehru University, New Delhi, India).
Author contributions
Khushdeep Bandesh: Conceptualization, Study design, Investigation, Methodology, Analysis, Writing—original draft. Muneesh Pal: Data acquisition. Abitha Balakrishnan: Data acquisition. Pradeep Gautam: Data acquisition. Punam Jha: Data acquisition. Nikhil Tandon: Intellectual inputs. Beena Pillai: Conceptualization, Study design, Supervision, Resources, Writing—original draft. Dwaipayan Bharadwaj: Conceptualization, Study design, Supervision, Resources, Funding acquisition, Writing—review and editing. All authors approved the final version of the paper.
D.B. and B.P. are the guarantors of this work and, as such, had full access to all the data in the study.
Data availability
The data generated in the study are available from the corresponding authors upon reasonable request.
