|Home | About | Journals | Submit | Contact Us | Français|
About one third of human genetic diseases are caused by premature translation-termination codon (PTC)-generating mutations. These mutations in sodium channel and GABAA receptor genes have been associated with idiopathic generalized epilepsies, but the cellular consequences of the PTCs on the mutant channel subunit biogenesis and function are unknown. The PTCs could result in translation of a truncated subunit, or more likely, trigger mRNA degradation through nonsense-mediated mRNA decay (NMD), thus preventing or reducing production of mutant subunit at the transcriptional level. The GABAA receptor α1 subunit mutation, 975delC, S326fs328X, is an autosomal dominant mutation associated with childhood absence epilepsy that generates a PTC in exon 8 of the 9 exon GABRA1 gene that is 74 bp upstream of intron 8. Using an intron 8-inclusion minigene that supports NMD, we demonstrated that mutant mRNA was substantially reduced, but not absent. Loss of mutant transcripts was blocked by ribosome inhibition or by silencing the NMD-essential gene hUPF-1. In both neurons and nonneuronal cells, the PTC caused substantial loss of mutant α1(S326fs328X) subunit mRNA through NMD with a minor portion of the mRNA escaping NMD and producing a mutant protein. The translated mutant protein had reduced stability due to endoplasmic reticulum associated degradation (ERAD) and had enhanced association with molecular chaperones. This study suggests that loss of mRNA due to activation of NMD and activation of ERAD by the mutant protein may contribute to epileptogenesis. The molecular mechanisms outlined here delineate a model for the pathogenesis of many PTC-generating mutations.
Several GABAA receptor subunit missense and nonsense premature translation-termination codon (PTC)-generating mutations have been associated with idiopathic generalized epilepsies (IGEs) (Macdonald et al., 2006). Transcription of mutant genes produces mutant mRNA and translation of mutant mRNA produces mutant protein. Cellular mRNA surveillance mechanisms often degrade mutant PTC-containing mRNAs and endoplasmic reticulum (ER) protein quality control processes degrade misfolded or misrouted proteins. Trafficking deficient mutant subunits are often subject to ER retention and ER associated degradation (ERAD) after translation (Kang and Macdonald, 2004;Gallagher et al., 2005). The cellular fate of GABAA receptor subunits translated from PTC-containing mutant mRNAs is unknown.
PTC mutations are responsible for about one-third of inherited disorders including GABR subunit gene mutation associated epilepsies. In several of the genetic diseases caused by PTCs, mRNAs containing a PTC are often subject to degradation via a mechanism called nonsense mediated mRNA decay (NMD). NMD is a post-transcriptional, but translation-dependent, mRNA quality control mechanism that recognizes and selectively degrades mRNAs that contain a PTC that is 50–55 nucleotides upstream from an exon-exon junction (Isken and Maquat, 2007) or mRNAs with aberrantly configured 3′untraslated region (UTR) (Amrani et al., 2004). NMD is translation-dependent since translating ribosomes recognize PTC during the pioneering round of translation. NMD is also splicing-dependent and requires an exon junction complex (EJC) deposited during intron splicing at an exon-exon junction located downstream of the PTC. EJCs consist of mRNA decay factors including Upf-1 (rent1), which is an RNA helicase and essential factor to activate NMD. NMD rids cells of most transcripts containing PTCs and reduces intracellular levels of truncated and potentially deleterious proteins (Kuzmiak and Maquat, 2006). Most nonsense transcripts are reduced by cellular mRNA surveillance processes, including NMD, to ~5–25% of wild-type levels (Kuzmiak and Maquat, 2006), with mutant mRNAs escaping NMD resulting in translation of truncated proteins that are often trafficking deficient, misfolded and misrouted and consequently subject to ERAD (Stephenson and Maquat, 1996). The mechanisms by which ERAD targets misfolded proteins includes the ubiquitin-proteasomal system (UPS) (Turnbull et al., 2007) and the autophagy/lysosome pathway (Cuervo, 2004). Inhibition of translation of mRNA by cycloheximide, the protein synthesis inhibitor, or silencing the key mRNA decay factor, Upf-1, reverse the mRNA loss produced by NMD.
The GABAA receptor α1 subunit mutation, 975delC, S326fs328X, is a rare autosomal dominant mutation associated with childhood absence epilepsy (CAE) (Maljevic et al., 2006). The deletion, 975delC, should cause a frameshift at residue S326 and create a PTC at residue L328, which is in the second to last exon (exon 8) of the 9 exon gene and 74 nucleotides upstream of the last exon-exon junction. Based on either the “50 nucleotide boundary rule” (Maquat, 2005) or the “faux 3′-UTR model” (Amrani et al., 2004), the PTC likely would activate NMD, resulting in degradation of mutant transcripts (Shyu et al., 2008), but this has not been demonstrated. We used an intron-inclusion minigene approach to characterize the effect of the 975delC, S326fs328X mutation on α1 subunit mRNA and protein stability. This approach has been widely used to study aberrant splicing and NMD since it supports mRNA editing and splicing (Hefferon et al., 2002;Busi and Cresteil, 2005;Buhler et al., 2004). In addition to all 9 exons, the α1 minigene contained the entire intron 8, thus rendering mRNA produced from the minigene a candidate for NMD since splicing out intron 8 would create an exon-exon junction between exons 8 and 9. In contrast, mRNAs that contain PTCs derived from conventional intron-less cDNA constructs are not subject to NMD, since they do not contain an exon-exon junction. The α1 subunit mutation, 975delC, S326fs328X, is 74 nucleotides upstream of the splicing-generated exon-exon junction in the mRNA, and thus, should elicit NMD. Here we demonstrated that mutant α1 subunit mRNA was subject to substantial, but incomplete, NMD, and that the mutant subunit protein that was produced by the mRNA that escaped NMD was degraded by ERAD through the ubiquitin-proteasome system. The combination of mutant mRNA degradation by NMD and mutant protein degradation by ERAD may represent a molecular pathology model for many other PTC-generating mutations.
The cDNAs encoding human, GABAA receptor β2,γ2S and wild-type and mutant α1 subunits were subcloned into the expression vector pcDNA3.1(+). For the remainder of the paper, the mutant α1(975delC, S326fs328X) subunit will be referred to as the α1(S326fs328X) subunit. The FLAG (DYKDDDDK) epitope was inserted between amino acids 4 and 5 in the amino terminus of the mature peptide to create the α1FLAG and α1(S326fs328X)FLAG subunit cDNAs. The wild-typeα1 subunit minigene was generated by including the entire intron 8 of the GABAA receptor α1 subunit (nm_000806), which was extracted from human genomic DNA. The α1(S326fs328X),α1(A322D) and α1(Y368X) subunit mutations were generated using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) and confirmed by DNA sequencing.
Rat cortical neurons were dissected from the brains of embryonic 17–18 day old Sprague-Dawley rat embryos. Dissociation of cells and culturing and transfection procedures were carried out as described previously with minor modifications (Kang et al., 2006).
Cortical neurons were transfected by Fugene or nucleofection (Amaxa GmbH, Germany) with program 0–03 following the manufacturer’s protocols. HeLa cells were transfected with Fugene (Roche Molecular Systems, Inc Indianapolis, IN), and HEK 293T cells were transfected with the calcium phosphate precipitation (Calcium) method unless otherwise specified. Cells were cotransfected with 1–2 μg of each subunit plasmid for each 60 mm2 dish and 0.33 μg for each 35 mm2 dish. To silence the hUPF-1 gene, a factor essential for NMD, cells were transfected with siRNA targeting nucleotides of hUPF-1 (5′-AAGATGCAGTTCCGCTCCATTTT-3′) (Ambion) with Oligofectamine (Invitrogen) as described before (Mendell et al., 2002). The same cells were re-transfected with the wild-type and mutant α1 minigene constructs or in addition to β2 and γ2S subunits 48 hr later and harvested 2 additional days later for RT-PCR and western blot.
The procedures of cell surface receptor biotinylation and western blot were as described before (Kang and Macdonald, 2004). After SDS-PAGE, the membranes were incubated with specific primary antibody overnight at 4°C with gentle rotation. The monoclonal anti-human α1 antibody (BD24) was purchased from (Chemicon, Inc. Temecula, CA). The polyclonal anti-human Upf-1 (hUpf-1) antibody was from Abgent primary antibody company (San Diego, CA). Polyclonal rabbit anti-Bip/GRP78 and monoclonal anti-Bip(GRP79) were purchased from Abcam (Cambridge, MA) and BD Transduction Laboratories (San Jose, CA), respectively. After washing, the membranes were incubated with horseradish peroxidase conjugated secondary antibody (goat anti-mouse IgG or goat anti-rabbit IgG (1:7500) (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). The antibody-reactive bands were revealed by chemiluminescence. The western blots were quantified with ChemiImager AlphaEaseFC.
Rat cortical neurons were plated on poly-D-lysine-coated, glass-bottom imaging dishes at the density of 1–2 × 105 cells. Rat cortical neurons were transfected with wild-typeα1 minigene and β2 and γ2S subunits at a ratio of 1:1:1 (wt); mutantα1 minigene (S326fs328X) and β2 andγ2S subunits and mutantα1(Y368X) minigene and β2 and γ2S subunits for 6 days. The neurons were fixed, permeabilized and stained with anti monoclonal human α1 subunit antibody conjugated with fluorophore Alexa-647. The neurons were then visualized under confocal microscopy as described before (Kang and Macdonald, 2004).
HEK 293T cells were cotransfected with 2 μg of each subunit plasmid and 1 μg of the pHook-1 cDNA (Invitrogen, Carlsbad, CA) using a modified calcium phosphate precipitation method and selected 24 hours after transfection by magnetic hapten coated beads (Greenfield, Jr. et al., 1997). Recordings were obtained from the cells 2 days later using methods described previously (Kang and Macdonald, 2004).
Total RNAs were extracted by using Versagene RNA Cell kit (Gentra Systems, Minneapolis, Minnesota) or by RNAeasy Micro kit (Qiagen, Valencia, CA) following the manufacturer’s protocol. The transcribed cDNA products were probed with either human GABAA receptor α1 subunit probes or 18s with 1:10 dilution. Quantitative real-time PCR experiments were performed with these specific primers and Taqman probes (sequence available upon request) in a 5 μl final volume and normalized to endogenous 18s rDNA (acetyl-H3 CHIP). Relative RNA abundance was quantified by evaluating threshold cycle (Ct) values, according to the “relative standard curve” method (ABI Prism 7700 Manual).
The protocol was modified from our previously published protocol (Gallagher et al., 2007). Briefly, forty-eight hours after transfection, the cells were replenished with the starving medium which lacked methionine and cysteine (Invitrogen), and incubated at 37°C for 30 min. The starving medium was then replaced by 1.5 ml 35S radionuclide methionine [100–250 μCi/ml (1 Ci = 37GBq); PerkinElmer, Wellesley, MA] labeling medium for 20 min at 37°C. The labeling medium was then changed to the chase medium for a series of different time points. The FLAG-tagged α1 subunits were then immunoprecipitated from radio-labeled lysates with an anti-FLAG M2-agarose affinity gel by rotating at 4°C overnight. The immunoprecipitated products were then eluted from the beads with FLAG peptide (Sigma-Aldrich). The immunopurified subunits were then analyzed by 12.5% SDS-PAGE and exposed on a digital PhosphorImager (GE Healthcare, Piscataway, NJ).
Macroscopic currents were low pass filtered at 2 kHz, digitized at 10 kHz, and analyzed using pClamp9 software suite (Axon Instruments). Except for the pulse-chase assays, proteins were quantified by ChemiImager AlphaEaseFC software, and data were normalized to either wild-type subunits or loading controls. Data from pulse-chase experiments were quantified using Quantity One software (Bio-Rad, Hercules, CA). Numerical data were expressed as mean ± SEM. When wild-type data were arbitrarily taken as 1, column statistics were used. Statistical significance, using Student’s unpaired t test (GraphPad Prism), was taken as p < 0.05.
The minigene strategy has been used to study transcript alteration and NMD in a number of diseases, including cystic fibrosis, since it contains introns that support mRNA editing and splicing. We thus used this approach to study the intracellular processing of human wild-type and mutant α1 subunits. The α1 subunit mutation, 975delC, S326fs328X, is located in the middle of TM3, and the PTC produced by the deletion would result in a loss of the C-terminal 141 amino acids, thus producing a truncated α1(S326fs328X) subunit (Figure 1A). The human α1 subunit gene, GABRA1, has 9 exons and 8 introns. Minigenes were constructed by including the entire GABRA1 intron 8 (1242 bp) between exon 8 and 9 of the wild-type and mutant α1 subunit cDNA constructs (Figure 1Ba, b). The stop codon for the wild-type α1 subunit minigene is located at the end of exon 9 (Figure 1Ba), and the PTC for the mutant α1 subunit minigene α1(S326fs328X) subunit minigene is located in exon 8, which is 128 bp downstream from intron 7 and 74 bp upstream from intron 8 (Figure 1Bb). We also generated an α1(A322D) subunit minigene as a control (Figure 1A, Bc). The α1 subunit mutation, A322D, associated with juvenile myoclonic epilepsy, is a missense mutation four residues upstream of the 975delC, S326fs328X frameshift mutation and should not elicit NMD. In addition, we generated an artificial α1 subunit nonsense mutation, Y368X, in the middle of the last exon (exon 9) as an NMD-insensitive control (Figure 1Bd).
In HEK 293T cells, wild-type and mutant α1(S326fs328X) subunit minigenes were cotransfected with β2 and γ2 subunit cDNAs. In this and all other experiments, the Calcium transfection technique was used unless otherwise noted. Total RNAs were extracted 48 hrs after transfection, and equal amounts of total RNA were transcribed to cDNA. With flanking primers in exon 8 and 9, equal amounts of cDNAs from wild-type and mutant α1 subunit minigenes were used for a template in the PCR amplification, and both minigenes displayed a band at 260 bp, suggesting correct splicing in the reverse transcribed cDNA products from α1 and α1(S326fs328X) subunit minigene mRNAs (Figure 1C). The mutant α1 subunit minigene band integrated density values (IDVs), however, were only about ~20% of the wild-type α1 subunit minigene band IDVs, suggesting reduction of mutant transcript levels.
It is unknown how robust the endogenous NMD machinery is in different cell types, and if NMD in a given cell type can degrade mRNAs equally when transfection methods with different efficiencies are employed. We thus first used the less efficient Calcium transfection technique and then the more efficient Fugene transfection technique (see Methods). With Calcium transfection, we cotransfected HEK 293T cells with β2 and γ2S subunit and wild-type and/or mutant α1 subunit cDNAs (Figure 2A, B; left) or minigenes (Figure 2A, B; middle) at a cDNA ratio of 1:1:1 for wild-type α1 subunits and mutant α1(S326fs328X) subunits and at a cDNA ratio of 1:1:0.5:0.5 for mixed α1/α1(S326fs328X) subunits. Wild-type α1 subunits should migrate at about 50 KDa, and mutant α1(S326fs328X) subunits should migrate at about 37 KDa (Figure 2A; left, green arrows). Our data are consistent with the original report that mutant α1(S326fs328X) subunit protein was reduced in both mixed and mutant conditions, suggesting that the mutant protein was degraded (Figure 2B; left) (0.328 ± 0.063 mutant vs 0.765 ± 0.082 wt; p = 0.002). By flow cytometry, the total amount of mutant protein was less than 3% of wild-type protein (Supplementary Figure 1). Interestingly, the ratio of wild-type to mutant protein with mixed expression was larger than the ratio of wild-type to mutant protein with wt and mut expression, suggesting that the presence of the wild-type α1 subunit or a failure to compete with the wild-type subunit during assembly somehow facilitated degradation of the mutant protein.
Second, HEK 293T cells were cotransfected with β2 and γ2S subunit cDNAs and wildtype and mutant α1 subunit minigenes (wt, cDNA ratio of 1:1:1), mixed α1/α1(S326fs328X) subunit minigenes (mix, cDNA ratio of 1;1:0.5:0.5 or mutant α1(S326fs328X) subunit minigene (cDNA ratio of 1:1:1) (Figure 2A, B; middle). The wild-type α1 subunit minigene displayed a robust 50 KDa band in both wild-type and mixed conditions. With mixed or mutant expression of the minigene constructs, however, mutant protein (37 KDa) was substantially reduced (Figure 2A; middle, green arrows) (Figure 2B; middle), suggesting an additional mechanism for loss of mutant protein.
Finally, HEK 293T cells were cotransfected with β2 and γ2S subunit cDNAs and the wild-type α1 subunit minigene (wt, cDNA ratio of 1:1:1), mixed α1/α1(Y368X) subunit minigenes (mix, cDNA ratio of 1:1:0.5:0.5) or mutant α1(Y368X) subunit minigene (cDNA ratio of 1:1:1) minigenes (Figure 2A, B; right). Expression of the NMD-insensitive α1(Y368X) subunit minigene in both mixed and mutant conditions produced truncated proteins but at reduced levels that migrated at the predicted molecular mass of 43 KDa (Figure 2A, B; right).
We next extended our study to include other cell types including HeLa cells and rat cortical neurons. Since these cells are less efficiently transfected, we employed the more efficient Fugene transfection technique, and included Fugene transfection of HEK 293T cells for comparison. In HEK 293T cells with Fugene transfection of wild-type and/or mutant α1 subunit minigenes, mutant α1(S326fs328X) subunit protein was present, although substantially reduced compared to wild-type subunit protein in both mixed and mutant conditions (Figure 2C; left), despite being almost absent with Calcium transfection (Figure 2A; middle). However, the mutant protein was not detectable with Fugene transfection in either HeLa cells (Figure 2C; middle) or rat cortical neurons (Figure 2C; right), despite robust expression of wild-type α1 subunit protein.
Consistent with the biochemical data, similar data were obtained with immunocytochemistry and confocal microscopy for total human α1 subunit protein expression (Figure 2D). In rat cortical neurons expressing wild-type α1β2γ2S subunits six days following transfection of β2 and γ2S subunit cDNAs and wild-type α1 subunit minigenes (Figure 2D; left), there was robust signal in both the cell bodies and dendrites, but in neurons expressing α1(S326fs328X)β2γ2S subunits, there was only a minimal fluorescent signal in the soma (Figure 2D; middle). In contrast, a significant amount of fluorescence was detected in the cell somata of neurons expressing α1(Y368X)β2γ2S subunits (Figure 2D; right). The levels of mutant relative to wild-type subunit protein seemed to vary somewhat with different transfection reagents, different cell types and the presence or absence of the β2 and γ2 subunits (Supplementary Figure 1). It is likely that the more efficient transfection technique (Fugene) and use of efficiently transfected cell types (HEK 293T cells) were able to partially overwhelm the NMD machinery and ER quality control mechanisms due to high minigene copy number. Nonetheless, reduction of mutant protein following minigene expression occurred with all transfection techniques and in all cell types, suggesting common and consistent pathways for loss of mutant protein, likely NMD and ERAD.
We recorded currents evoked by 6 (Figure 3A, B) or 28 (Figure 3C, D) sec applications of 1 mM GABA from HEK 293T cells cotransfected with β2 and γ2S subunit cDNAs and the wild-type α1 subunit minigene (1:1:1 cDNA ratio; wt, black), half the amount of wild-type α1 subunit minigene as a haploinsufficiency control (1:1:0.5 cDNA ratio; hc, silver), mixed α1/α1(S326fs328X) subunit minigenes (1:1:0.5:0.5 cDNA ratio; mix, green) or mutant α1(S326fs328X) subunit minigene (1:1:1 cDNA ratio; mut, gray). Peak haploinsufficiency control and mixed receptor currents were smaller than wild-type currents, and there was no significant difference between them (p = 0.65) (Figure 3A, B). No currents were recorded from cells following mutant subunit expression. Compared with wild-type type peak currents (2490 ± 189.3 pA,), peak amplitudes of both haploinsufficiency control (1461 ± 294.7 pA; p = 0.01) and mixed receptor currents (1299 ± 158.2 pA; p = 0.0004) were significantly reduced (Figure 3A, B) without change in macroscopic kinetic properties. When normalized to the peak wild-type receptor current, there was no difference in the steady state/peak current ratio between wild-type and haploinsufficiency control or wild-type and the receptor from mixed expression of wild-type and mutant receptors (Figure 3C, D), suggesting that the functional receptors on the cell surface were likely wild-type receptors.
Results from cell surface receptor biotinylation and immunoblotting were consistent with the electrophysiological results. HEK 293T cells were cotransfected with β2 and γ2S subunit cDNAs and the wild-type α1 subunit minigene (1:1:1 cDNA ratio; wt), half the amount of wild-type α1 subunit minigene as a haploinsufficiency control (1:1:0.5 cDNA ratio; hc), mixed α1/α1(S326fs328X) subunit minigenes (1:1:0.5:0.5 cDNA ratio; mix) or mutant α1(S326fs328X) subunit minigene (1:1:1 cDNA ratio; mut) (Figure 3E). Compared with transfection of wild-type subunits, transfection of haploinsufficiency control or mixed subunits resulted in reduced wild-type α1 subunit surface levels. There was no mutant α1(S326fs328X) subunit on the surface with either mixed or mutant expression. When normalized to wild-type α1 subunit surface levels, α1 subunit levels were reduced with expression of either haploinsufficiency control (0.64 ± 0.04, p<0.0001) or mixed (0.62 ± 0.03, p<0.0001) subunits, but there was no significant difference in α1 subunit surface levels between the haploinsufficiency control and mixed subunits (p = 0.67) (Figure 3F). A similar pattern of surface α1 subunit protein expression was obtained by a more quantitative flow cytometry (Supplementary Figure 2A, B).
The mutant α1(S326fs328X) subunit that was detected was likely retained in the ER and degraded. To confirm this, following coexpression of β2 and γ2S subunits with either wild-type α1 subunit minigenes or mutant α1(S326fs328X) subunit cDNAs, total lysates from HEK 293T cells were either undigested (U) or digested with endoglycosidase H (Endo-H) (Figure 4A, H) or peptide-N-glycosidase F (PNGase-F) (Figure 4A, F). Since only small amounts of protein were produced with the mutant minigene expression, the mutant cDNA construct was used for this experiment. Endo-H removes N-linked high-mannose oligosaccharides attached in the ER, but not complex oligosaccharides attached in the trans-Golgi region. In contrast, PNGase-F removes N-linked oligosaccharides attached in both ER and trans-Golgi region. Therefore, protein retained inside the ER should have similar molecular mass with either Endo-H or PNGase-F digestion. Wild-type α1 subunits migrated at a higher molecular mass after Endo-H digestion and migrated at a lower molecular mass after PNGase-F digestion as we previously described (Gallagher et al., 2005). The mutant protein, however, migrated at a similar molecular mass after either Endo-H or PNGase-F digestion. The different shifts in molecular mass of wild-type and mutant α1 subunits after Endo-H or PNGase-F digestion suggested that the majority of the wild-type α1 subunits were trafficked beyond the ER, and that the mutant subunits were retained in the ER.
Loss of mutant protein could have been due to rapid degradation after protein synthesis, NMD of the PTC-containing transcripts before protein synthesis or both. If the mutant protein was mainly subject to ERAD after synthesis, proteasomal inhibition should “rescue” the mutant protein. Therefore, we applied lactacystin (Lac, 10 μM) and MG-132 (3 μM) (data not shown), two structurally different proteasome inhibitors, for 6 hours to HEK 293T cells cotransfected with β2 and γ2S subunit cDNAs and wild-type, haploinsufficiency control, mixed or mutant α1/α1(S326fs328X) subunit minigenes (Figure 4B). With either Lac or MG-132 (data not shown) treatment, the total mutant α1(S326fs328X) subunit proteins from both less efficient Calcium or more efficient Fugene transfections were increased, but still minimal, compared to the wild-type subunit (Figure 4B). Relative to the wild-type α1 subunits, the magnitude of the total protein reduction in the haploinsufficiency control, mixed or mutant condition was greater with Calcium transfection (Figure 4B top) than with Fugene transfection (Figure 4B bottom) but was significantly reduced with both transfection techniques (Figure 4C). The mutant α1(S326fs328X) subunit protein, but not wild-type, haploinsufficiency control or mixed α1 subunit proteins, were increased by lactacystin treatment when normalized to the subunits with lactacystin treatment, suggesting enhanced degradation of the mutant subunit (Figure 4D). There was no difference between the ratio of the total protein of the mixed and wild-type subunit proteins before or after lactacystin, probably because the mutant protein was minimal compared to the wild-type protein in the mixed condition (Figure 4B Fugene; Figure 4D), thus obscuring any increase of the mutant subunit. The NMD-insensitive control α1(Y368X) subunit proteins were all significantly increased by both proteasomal inhibitors (Supplementary Figure 3). The data suggest that the loss of the mutant α1(S326fs328X) subunit was mainly due to NMD while the loss of α1(Y368X) subunit was mainly due to ERAD. The mutant α1(S326fs328X) subunit, however, was partially rescued by blocking protein degradation, thus suggesting that any mutant subunit produced would be subject to degradation by ERAD.
We then used intronless cDNA constructs, which would not be subject to NMD, and Fugene transfection to produce more mutant protein (Figure 4E). Mutant α1(S326fs328X) subunit levels in both HEK 293T and HeLa cells were reduced compared to wild-typeα1 subunit levels but were increased by proteasomal inhibition to a greater extent than wild-typeα1 subunits (1.24 ± 0.12 vs 1.99 ± 0.11, p = 0.0007 for HEK 293T cells; 1.66 ± 0.20 vs 2.36 ± 0.23%, p = 0.046 for HeLa cells) (Figure 4F). The increased production of mutant α1(S326fs328X) subunit protein with transfection of intronless cDNA constructs and with Fugene transfection of minigene constructs (Figure 4B, D) also suggested that any mutant protein produced would be subject to rapid degradation by ERAD through the ubiquitin-proteasome system.
To further investigate the cellular fate of the mutant protein, we performed 35S methionine radio-labeling pulse-chase experiment to determine if the mutant protein had an altered rate of degradation. We used HEK 293T cells since they had highly efficient protein expression, which produced higher levels of mutant subunit. The cells were transfected with either wild-typeα1 or mutant α1(S326fs328X) subunits alone with Fugene and then used for study after 48 hours. After radionuclide labeling for 20 min, the cells were lysed immediately and analyzed as time 0 or as “chased” for 30, 60, 90, 120, 180 and 240 min (Figure 5A). Compared with the amount of protein at time 0, the remaining amounts of wild-type and mutant subunit protein were unchanged at 30 min, but the mutant subunit degraded faster than wild-type subunit from 60 min to 240 min. From 60 min to 240 min of chase, the mutant subunit was significantly reduced compared to wild-type α1 subunit. The half life of the mutant subunit was 64 min, and the half life of wild-typeα1 subunits was 194 min, which was more than 3 fold higher than that of the mutant subunit. To exclude the possibility that the increased mutant protein after proteasomal inhibition was due to increased synthesis, we applied cycloheximide (CHX) (100 μg/ml) to block protein synthesis (data not shown) and found that the mutant protein was degraded more rapidly either with or without CHX treatment.
The reduced mutant protein could also be due to the reduced synthesis. To determine the rate of synthesis of wild-type and mutant subunits, we pulse-labeled cells transfected with α1FLAG or α1(S326fs328X)FLAG subunits with 35S methionine for 0, 5, 10, 15 and 20 min. We used intronless α1 subunit cDNA constructs since they were used for degradation pulse chase experiments above. Normalized to the total wild-type or mutant protein labeled for 20 min, there were no significant differences in the amount of wild-type or mutant α1 subunits, demonstrating no difference in the rate of synthesis of these subunits.
Molecular chaperones are a group of proteins that mediate protein folding, assembly, trafficking and degradation (Hartl, 1996). Our data above suggested that the mutant protein was subject to ERAD, which is known to be mediated by chaperones. We thus tested the association of the mutant protein with Bip, which is an ER luminal chaperone that binds misfolded or unfolded, unassembled secretory proteins and ensures proper maturation and movement of proteins from the ER to the Golgi apparatus (Connolly et al., 1996). The cells were transfected with either wild-type α1FLAG or mutant α1(S326fs328X)FLAG subunits alone with Fugene and studied after 48 hours. The total lysates were purified with FLAG antibody conjugated beads and probed with either anti-FLAG antibody for α1FLAG subunits or anti-Bip antibody (Figure 6). The ratio of Bip to α1FLAG subunit IDV was then determined. Consistent with the accelerated degradation of the mutant subunit, there was less mutant than wild-type subunit protein. Also, more Bip protein was pulled down with the mutant subunit compared to the wild-type subunit. Thus the ratio of Bip to mutant subunit was enhanced substantially (7 fold) compared to the wild-type subunit (wt 0.12 ± 0.011 vs mut 0.8430 ± 0.053) (P<0.001, n = 4), consistent with an increased association of Bip with the mutant subunit.
Was loss of mutant α1(S326fs328X) subunit due also to loss of mRNA triggered by NMD? To determine this, we first quantified wild-type and mutant α1 subunit mRNAs in HeLa and HEK 293T cells and rat cortical neurons transfected with wild-type α1 and mutant α1(S326fs328X) subunit minigenes. Total RNAs were extracted 2–4 days after transfection, and mRNAs were analyzed by Real-time PCR. In HEK 293T and HeLa cells and in neurons, mutant α1(S326fs328X) transcripts were substantially reduced (Figure 7A). When normalized to wild-type α1 subunit mRNA levels, mutant α1(S326fs328X) subunit mRNA levels were reduced to 39.1 ± 7.3% (p = 0.0001) in HEK 293T cells, to 16.6 ± 4% (p = 0.006) in rat cortical neurons and to 23.8 ± 3.6% (p < 0.0001) in HeLa cells.
Treatment with the protein synthesis inhibitor cycloheximide (CHX) or silencing of the core component of the NMD machinery, Upf-1 (also known as Rent1), are widely used to determine if transcripts are subject to NMD (Montfort et al., 2006). CHX is an indirect NMD inhibitor since NMD activation is post-transcriptional, but translation-dependent, and ribosomal association is required during the pioneer round of translation. Upf-1 silencing is believed to produce direct suppression of NMD. The RNA helicase, Upf-1, is a central effector of NMD that links the translation-termination event to the assembly of a surveillance complex. Upf2 and Upf3 are believed to recruit and activate Upf1 on NMD substrates (Lykke-Andersen et al., 2000). Although the discrimination of PTCs from normal termination codons and the molecular links that trigger NMD are still unclear, it is well established that the core components of the NMD machinery are the conserved proteins, Upf-1, Upf-2 and Upf-3. PTC-containing mRNAs marked by the Upf complex are finally degraded by decay factors (Conti and Izaurralde, 2005). A recent study reported that Upf-1 phosphorylation triggers translation repression, which is a key step preceding mRNA decay (Isken et al., 2008). With CHX (100 μg/ml) treatment for 6 hrs, mutant α1(S326fs328X) subunit mRNA transcripts were increased in HeLa cells from 27.4% ± 5.2% to 84.5% ± 8.5% of wild-type α1 subunit mRNA after CHX treatment (Figure 7B). Addition of CHX reversed the down-regulation of the mutant transcripts, having no discernible effect on wild-type mRNA transcripts (Figure 7B).
We then investigated the role of Upf-1 in reducing α1(S326fs328X) subunit levels. We transfected cells with synthetic short-interfering RNA (siRNA) duplexes targeting the human Upf-1 (hUpf-1) gene (Chan et al., 2007;Mendell et al., 2002) and harvested the cells 48 hrs after transfection. Western blotting of whole cell lysates demonstrated significant loss of hUpf-1 protein in cells treated with hUpf-1 siRNA in HEK 293T (from 50 to 100 nM) and HeLa cells (from 10 to 100 nM) (Figure 7C). Treatment with 100 nM siRNA for either 2 or 4 days significantly reduced hUpf-1 protein levels in both HEK 293T and HeLa cells, but the siRNA appeared to be more effective in HeLa cells at both time points (Figure 7C). We thus used HeLa cells in these experiments. A similar effect, but to a lesser magnitude, was observed with HEK 293T cells (Supplementary Figure 4).
Wild-type or mutant α1 subunit minigene constructs were transfected into HeLa cells treated 48 hr earlier with siRNA. The cells were harvested after an additional 48 hr for protein (Figure 7D, E) and transcript (Figure 7F) assays. All results were normalized to the cells treated only with siRNA delivery reagents (mock). Cells treated with siRNA (100 nM) against hUpf-1 (Upf) had increased expression of the mutant α1(S326fs328X) subunit compared with cells treated with the negative control (neg) siRNA (Figure 7D, E) in both HEK 293T cells (Supplementary Figure 4) and HeLa cells (0.80 ± 0.35 vs 2.48 ± 0.45; p = 0.006). Cells treated with lactacystin, in addition to the hUpf-1 specific siRNA, had increased expression of the mutant protein compared to cells treated with lactacystin alone in both cell lines (3.93 ± 0.83 (Upf + lac) vs 1.46 ± 0.20 (lac) (p = 0.013). There was no significant change in expression of wild-typeα1 subunit proteins (Figure 7D) with hUpf-1 specific siRNAs treatment (100 nM). In contrast, the mutant α1(A322D) subunit protein was similarly increased when treated with lactacystin in addition to the hUpf-1 specific siRNA or with lactacystin alone (Figure 7D, E). The data suggest that the α1(S326fs328X) subunit protein was reduced mainly due to mRNA degradation, and that the α1(A322D) subunit protein was reduced mainly due to protein degradation by ERAD (Gallagher et al., 2007).
Similarly, normalized to mRNA levels in cells expressing wild-typeα1 subunits with mock treatment, expression of mutant α1(S326fs328X) subunit transcripts was increased in cells treated with siRNA against hUpf-1 (Figure 7F) compared to expression of mutantα1(S326fs328X) subunit transcripts in cells treated with negative control siRNA (11.3 ± 5 vs 91.4 ± 12; p = 0.002) in HeLa cells. There was no significant change in the wild-type (data not shown) or α1(A322D) (p = 0.28) subunit transcripts after hUpf-1 siRNA treatment (Figure 7F). Taken together, our data suggest that elimination of the mutant α1(S326fs328X) subunit was due primarily to loss of transcripts due to mRNA surveillance (NMD), but that ERAD also played a role in reducing mutant protein levels.
Although both GABRA1 epilepsy mutations, A322D and 975delC, S326fs328X, are located in the same domain of the α1 subunit (TM3) and both mutations produced haploinsufficiency, the molecular bases for the functional consequences are different. The α1(A322D) subunit mutation is a missense mutation, and the mutant mRNAs are translated and produce an unstable mutant protein, which is subject to rapid ERAD. The α1 subunit mutation, 975delC, S326fs328X, is a PTC-generating mutation and activated NMD. However, NMD is not complete; it degrades the majority of the mutant mRNA, but a small portion of mutant mRNA escapes NMD. The surviving mutant mRNA would result in translation of a mutant, truncated protein, which is not stable and would be subject to ERAD like the other mutant α1(A322D) subunit. The loss of mutant message and protein from both molecular pathways would directly contribute to the loss of the functional α1 subunit-containing GABAA receptors, and thus, the loss of inhibition that would result in epilepsy (Figure 8).
Several GABAA receptor subunit gene mutations that should generate PTCs have been associated with IGEs and include three nonsense γ2 subunit mutations, Q1X, Q351X and W390X, and a γ2 subunit intron splice donor site mutation, IVS 6+ 2T->G, in addition to the present α1 subunit mutation, 975delC, S326fs328X (Harkin et al., 2002; Hirose, 2006; Kananura et al., 2002; Sun et al., 2008). The epilepsy phenotypic spectrum produced by these mutations is wide and includes febrile seizures, GEFS+, Dravet syndrome and CAE. For unknown reasons there are also substantial intrafamilial variations in seizure phenotype for individual mutations such as the γ2 subunit mutations, Q1X and Q351X, (Macdonald et al., 2006). Studies of PTC-generating mutations in sodium channels indicated that mutation mosaicism was associated with the intrafamilial phenotypic variations (Marini et al., 2006; Morimoto et al., 2006). However, the bases for phenotypic variations in patients harboring GABAA receptor PTC-generating mutations have never been explored.
NMD is a conserved mRNA surveillance pathway in all eukaryotes, and our study indicated that PTC-generating mutations in a CNS ion channel gene would be subject to the same cellular fate as seen with other genes and in other cell types. The role of NMD has been confirmed in a number of human diseases such as Marfan’s syndrome and cystic fibrosis. NMD is rarely complete, however, with ~5–25% of the mutant mRNA escaping NMD (Kuzmiak and Maquat, 2006a). Similarly, our data demonstrated that the α1 subunit mutation, 975delC, S326fs328X, activated NMD, resulting a substantial loss of the mutant mRNA (~80% in HeLa cells and rat cortical neurons), but NMD was not complete, and small amounts of mutant protein were produced. The mutant protein was not stable and was likely subject to rapid proteasomal degradation, resulting in a reduced half life (64 min) compared with that of wild-typeα1 subunits (194 min).
Our data suggested that the functional defect caused by the α1 subunit mutation, 975delC, S326fs328X, was due to degradation of the mutant mRNA and protein. The mRNA reduction was reversed by treatment with either the indirect NMD inhibitor CHX or by suppression of the NMD machinery core factor, Upf-1. After suppression of Upf-1, mutant α1(S326fs328X) subunit protein was increased to over two fold of the mock treated cells and increased nearly fourfold in cells with Upf-1 suppression in addition to proteasomal inhibition. In contrast, proteasomal inhibition alone only increased mutant protein to 145% of the mock treated cells. This suggested that a substantial portion of mutant protein reduction was due to loss of mutant mRNA and that mutant protein produced was subject to rapid proteasomal degradation. In contrast, the mutant α1(A322D) subunit was degraded through the ubiquitin-proteasomal system (Gallagher et al., 2007). However, it is unclear why the α1 subunit mutation, A322D, was associated with juvenile myoclonic epilepsy while the α1 subunit mutation, S326fs328X, was associated with CAE, even though both mutations produced functional haploinsufficiency in vitro. The differential intracellular processing of these two different mutations may at least partially contribute to their phenotypic variations. For example, α1(S326fs328X) subunits were lost mainly through NMD at the transcriptional level while α1(A322D) subunits were lost through ERAD after translation. It is unclear if there are any negative effects imposed by activation of the quality control mechanisms, NMD or ERAD, themselves. In addition, our data suggested that both NMD and ERAD efficiency varies among cell types and with different mutant α1 subunit copy numbers. Future studies focusing on the efficiency of NMD and ERAD in different neuronal cell types and during different developmental time windows may further clarify this discrepancy.
The functional consequence of the α1subunit mutation, del975C, S326fs328X, is likely to be due, at least in part, to haploinsufficiency of the GABRA1 gene. Based on our data in HEK 293T cells, there was about 59% of total wild-type α1 subunit produced in the haploinsufficiency control condition and 56% of total wild-type α1 protein in the mixed α1/α1(S326fs328X) subunit condition. There was about half of the total wild-type current amplitude produced in the mixed condition, which was similar to the current produced in the haploinsufficiency control condition. There were no current kinetic differences among these three groups, suggesting that the functional receptors on the cell surface were recorded mainly from wild-typeα1β2γ2S receptors.
We cannot exclude, however, a small dominant negative effect of truncated α1(S326fs328X) subunits that escape NMD and have a dominant negative effect on GABAA receptor assembly. We demonstrated that the mutant protein was degraded rapidly, likely through the ubiquitin-proteasomal pathway. The stress of removal of mutant misfolded protein would cause an imbalance of the load of unfolded proteins that enter the ER and the folding capacity of the ER environment (Hartl, 1996). Given the fact that heterozygousα1(+/−) subunit knock-out mice do not develop epilepsy, there may be a subtle difference in functional and cellular consequences between α1(+/−) knockouts and mixedα1(S326fs328X) subunit mutants. For example, the enhanced chaperone activity of Bip and an increased association with mutant subunits in affected patients would not occur in α1(+/−) knockout mice. This subtle difference in vivo may be enough to cause a clinical phenotype during a restricted developmental period.
Activation of NMD can rid cells of most transcripts containing PTCs and reduce synthesis of the truncated proteins that have potentially deleterious effects inside cells. This may reduce the manifestation of some genetic diseases if the single wild-type allele is sufficient for physiological function or may only produce a mild phenotype compared to some C-terminal truncation mutations that have dominant-negative effects on wild-type proteins. Triggering NMD and escaping NMD may cause distinct disease phenotypes (Inoue et al., 2004). In osteogenesis imperfecta, Stickler syndrome and Marfan’s syndrome, it has been postulated that NMD moderates the phenotype compared with that produced by missense mutations. Marfan’s syndrome is an autosomal dominant systemic disorder of connective tissue caused by mutations in the extracellular matrix protein fibrillin 1 (Dietz et al., 1993). A genotype-phenotype difference has been noticed in Marfan’s syndrome; patients with low levels of mutant FBN1 due to NMD often exhibit milder phenotypes that fall outside the clinical criteria required for a Marfan’s syndrome diagnosis. It was also suggested that the minimal mutant transcript levels of about 7–10% of wild-type levels may produce sufficient amounts of truncated protein to produce dominant-negative interference with the wild-type allele leading to severe Marfan’s syndrome (Montgomery et al., 1998).
A molecular mechanism similar to that in Marfan’s syndrome may help explain why mutations in the same gene cause different clinical phenotypes and with different severities in GABAA receptor gene mutations associated with IGEs. Namely, how complete the NMD is in the individual may be associated with the intrafamilial phenotypic variation among family members carrying the same mutation. A similar discrepancy in genotype-phenotype correlation has been reported for the GABAA receptor γ2(+/−) knock-out mouse that displayed only a hyperanxiety phenotype, and in humans an early exon PTC-causing splice-donor site mutation produced CAE and febrile seizures (Kananura et al., 2002) and a C-terminal γ2 subunit truncation, Q351X, caused the more severe GEFS+ epilepsy syndrome (Harkin et al., 2002). The pathogenesis of the GABAA receptor α1 subunit mutation, 975delC, S326fs328X, is due mainly to NMD with a small portion of mutant transcripts escaping NMD and generating a truncated mutant protein that is degraded by ERAD. However, based on our observation in vitro and on previous reports, although the mRNA surveillance is conserved, the efficiency of NMD for a certain mutant transcript is cell-specific (Linde et al., 2007). The efficiency is likely dependent on the capacity of the NMD machinery, such as expression of the NMD core components such as UPF factors and the relative abundance of the mutant transcripts in a given cell. Thus, there may be a different balance of NMD reduction of the mutant mRNA or ERAD reduction of the mutant protein in neurons in different regions of the brain, during different developmental stages or among different individuals.
Our present work suggests that ion channel PTC mutations can cause NMD, and that neurons are likely to share common mRNA surveillance mechanisms present in other cell types, thus extending the molecular underpinning of and therapeutic strategy to develop treatment for ion channel diseases. For example, promotion of read-through with aminoglycoside treatment has been used in patients with cystic fibrosis (Linde et al., 2007). The same strategy may be useful for some patients with epilepsy who harbor PTC-generating mutations if the mutant protein is functional. From a therapeutic point of view, since theα1 subunit, 975delC, S326fs328X, mutation causes a frameshift, attempted promotion of read-through with aminoglycoside treatment to restore α1 subunit expression (Howard et al., 2004) would not be feasible. However, our data suggest that the pathology of this mutation is likely to be a combination of reduction of channel function and disturbance of cellular homeostasis due to the presence of small amounts of mutant protein. Thus, a therapeutic strategy to eliminate production of mutant protein using siRNA targeting of the mutant transcripts might be a useful approach (Rodriguez-Lebron and Paulson, 2006).
We thank Drs. Miles F. Wilkinson and Martin Gallagher for their constructive advice, Hannah Yan for technical assistance and Mengnan Tian for helpful discussions. This research was supported by NIH Grant R01 NS51590 to RLM and a Cure Research Grant to J-QK.