Aberrant calcium signaling by transglutaminase-mediated posttranslational modification of inositol 1,4,5-trisphosphate receptors

Significance

Reversible and repetitive structural changes are essential for ligand-gated ion channels to mediate biological signaling. The inositol 1,4,5-trisphosphate receptor (IP₃R) assembles ligand-gated ion channels that mediate calcium signaling. IP₃ activates channels at a distance by reversible allosteric changes in the IP₃R tetramer. Here we show a new mode of posttranslational modification that irreversibly blocks allosteric changes in the IP₃R. We identified an IP₃R-modifying enzyme as tissue transglutaminase that inhibits IP₃R function by locking subunit configurations. This modification chronically impaired calcium signaling and autophagy regulation in living cells, and up-regulated modification was observed in Huntington disease models. To our knowledge, this is the first demonstration of transglutaminase-catalyzed posttranslational modification in ligand-gated channel allostery and provides a new framework for enzymatic regulation of allostery.

Abstract

The inositol 1,4,5-trisphosphate receptor (IP₃R) in the endoplasmic reticulum mediates calcium signaling that impinges on intracellular processes. IP₃Rs are allosteric proteins comprising four subunits that form an ion channel activated by binding of IP₃ at a distance. Defective allostery in IP₃R is considered crucial to cellular dysfunction, but the specific mechanism remains unknown. Here we demonstrate that a pleiotropic enzyme transglutaminase type 2 targets the allosteric coupling domain of IP₃R type 1 (IP₃R1) and negatively regulates IP₃R1-mediated calcium signaling and autophagy by locking the subunit configurations. The control point of this regulation is the covalent posttranslational modification of the Gln2746 residue that transglutaminase type 2 tethers to the adjacent subunit. Modification of Gln2746 and IP₃R1 function was observed in Huntington disease models, suggesting a pathological role of this modification in the neurodegenerative disease. Our study reveals that cellular signaling is regulated by a new mode of posttranslational modification that chronically and enzymatically blocks allosteric changes in the ligand-gated channels that relate to disease states.

Ligand-gated ion channels function by allostery that is the regulation at a distance; the allosteric coupling of ligand binding with channel gating requires reversible changes in subunit configurations and conformations (1). Inositol 1,4,5-trisphosphate receptors (IP₃Rs) are ligand-gated ion channels that release calcium ions (Ca²⁺) from the endoplasmic reticulum (ER) (2, 3). IP₃Rs are allosteric proteins comprising four subunits that assemble a calcium channel with fourfold symmetry about an axis perpendicular to the ER membrane. The subunits of three IP₃R isoforms (IP₃R1, IP₃R2, and IP₃R3) are structurally divided into three domains: the IP₃-binding domain (IBD), the regulatory domain, and the channel domain (3–6). Fitting of the IBD X-ray structures (7, 8) to a cryo-EM map (9) indicates that the IBD activates a remote Ca²⁺ channel by allostery (8); however, the current X-ray structure only spans 5% of each tetramer, such that the mechanism underlying allosteric coupling of the IBD to channel gating remains unknown.

The IP₃R in the ER mediates intracellular calcium signaling that impinges on homeostatic control in various subsequent intracellular processes. Deletion of the genes encoding the type 1 IP₃R (IP₃R1) leads to perturbations in long-term potentiation/depression (3, 10, 11) and spinogenesis (12), and the human genetic disease spinocerebellar ataxia 15 is caused by haploinsufficiency of the IP₃R1 gene (13–15). Dysregulation of IP₃R1 is also implicated in neurodegenerative diseases including Huntington disease (HD) (16–18) and Alzheimer’s disease (AD) (19–22). IP₃Rs also control fundamental cellular processes—for example, mitochondrial energy production (23, 24), autophagy regulation (24–27), ER stress (28), hepatic gluconeogenesis (29), pancreatic exocytosis (30), and macrophage inflammasomes (31). On the other hand, excessive IP₃R function promotes cell death processes including apoptosis by activating mitochondrial or calpain pathways (2, 17). Considering these versatile roles of IP₃Rs, appropriate IP₃R structure and function are essential for living systems, and aberrant regulation of IP₃R closely relates to various diseases.

Several factors such as cytosolic molecules, interacting proteins, and posttranslational modifications control the IP₃-induced Ca²⁺ release (IICR) through allosteric sites in IP₃Rs. Cytosolic Ca²⁺ concentrations strictly control IICR in a biphasic manner with activation at low concentrations and inhibition at higher concentrations. The critical Ca²⁺ sensor for activation is conserved among the three isoforms of IP₃ and ryanodine receptors, and this sensor is located in the regulatory domain outside the IBD and the channel domain (32). A putative ATP regulatory region is deleted in opisthotonos mice, and IICR is also regulated by this mutation in the regulatory domain (33). Various interacting proteins, such as cytochrome c, Bcl-2-family proteins, ataxin-3, huntingtin (Htt) protein, Htt-associated protein 1A (HAP1A), and G-protein–coupled receptor kinase-interacting protein 1 (GIT1), target allosteric sites in the carboxyl-terminal tail (3–5). The regulatory domain and the carboxyl-terminal tail also undergo phosphorylation by the protein kinases A/G and B/Akt and contain the apoptotic cleavage sites for the protease caspase-3 (4, 5). These factors allosterically regulate IP₃R structure and function to control cellular fates; therefore, understanding the allosteric coupling of the IBD to channel gating will elucidate the regulatory mechanism of these factors.

Transglutaminase (TG) catalyses protein cross-linking between a glutamine (Gln) residue and a lysine (Lys) residue via an N^ε-(γ-glutamyl)lysine isopeptide bond (34, 35). TG type 2 (TG2) is a Ca²⁺-dependent enzyme with widespread distribution and is highly inducible by various stimulations such as oxidative stress, cytokines, growth factors, and retinoic acid (RA) (34, 35). TG2 is considered a significant disease-modifying factor in neurodegenerative diseases including HD, AD, and Parkinson’s diseases (PD) (34, 36–45) because TG2 might enzymatically stabilize aberrant aggregates of proteins implicated in these diseases—that is, mutant Htt, β-amyloid, and α-synuclein; however, the causal role of TG2 in Ca²⁺ signaling in brain pathogenesis has been unclear. Ablation of TG2 in HD mouse models is associated with increased lifespan and improved motor function (46, 47). However, TG2 knockout mice do not show impaired Htt aggregation, suggesting that TG2 may play a causal role in these disorders rather than TG2-dependent cross-links in aberrant protein aggregates (47, 48).

In this study, we discovered a new mode of chronic and irreversible allosteric regulation in IP₃R1 in which covalent modification of the receptor at Gln2746 is catalyzed by TG2. We demonstrate that up-regulation of TG2 modifies IP₃R1 structure and function in HD models and propose an etiologic role of this modification in the reduction of neuronal signaling and subsequent processes during the prodromal state of the neurodegenerative disease.

Results

Identification of an IP₃R1-Modifying Enzyme as TG.

In exploring a new allosteric mechanism of IP₃R tetramers, we developed an agarose gel electrophoresis (AGE) system that can evaluate the large quaternary structure of IP₃R1 (>1 MDa) (Fig. 1 A and B). We found that the incubation of IP₃R1 with brain cytosol resulted in cross-linking of the IP₃R1subunits (Fig. 1C), indicating the existence of an endogenous cross-linking enzyme in the brain. We fractionated the cross-linking activity using chromatography (Fig. S1 A–C). The Ca²⁺ dependence, elution profile of the activity, and inhibition with cystamine led us to postulate that the cross-linking might be catalyzed by a TG, most likely TG2 (34, 49).

Fig. 1.

Identification of an IP₃R1-modifying enzyme as TG. (A) Detection of cross-linked IP₃R1 by AGE. IP₃R1 subunits were cross-linked by the chemical cross-linker bis[sulfosuccinimidyl] suberate (BS). Roman numerals indicate cross-linked subunits observed by AGE; IP₃R1 sizes were compared with the thyroglobulin monomer (335 kDa) and dimer (670 kDa). (B) IP₃R1 (left lane) pretreated with Triton X-100 (1), Zwittergent 3–14 (2), SDS (3), or no detergent (4) was cross-linked with BS (right lane), indicating the oligomeric status (97). (C) Endogenous activity of Ca²⁺-dependent cross-linking between IP₃R1 subunits in the brain cytosol (Cyt). (D) Cross-linking activity in HeLa cells pretreated with control siRNA or TG2 siRNA for 48 h. Western blots (WBs) with αTG2 and amido black (AB) staining show the cellular levels of TG2 and total proteins, respectively. (E) TG2 expression levels in SH-SY5Y or U87-MG cells treated with RA or DMSO. The Right panels show cross-linking of IP₃R1 by 10 μM ionomycin (ION) in human SH-SY5Y or U87-MG cells, as revealed by AGE.

The direct cross-linking of IP₃R1 by TG2 was confirmed using purified IP₃R1 and TG2 (Fig. S1D). To examine the role of TG2 in the cross-linking of IP₃R1, we silenced TG2 expression using siRNA in HeLa cells. TG2 knockdown completely abolished the in vitro cross-linking activity in HeLa cell lysates (Fig. 1D), indicating that TG2 is essential for the cross-linking of IP₃R1. In contrast, induction of TG2 expression by RA enhanced the cross-linking activity in HeLa cells (Fig. S1E). Similar cross-linking was observed in living cells including human neuroblastoma (SH-SY5Y) and human glioblastoma (U87-MG) cells (Fig. 1E). These results support the conclusion that IP₃R1 is an endogenous substrate of TG2 in living cells.

TG2 Inhibits IP₃R-Mediated Ca²⁺ Release.

The AGE analysis revealed that TG2 modifies the IP₃R1 structure; therefore, we examined whether TG2 regulates IP₃R1 function. We measured IICR using P2/P3 vesicles and found that purified TG2 inhibited IICR in a dose-dependent manner, without altering ATP-dependent Ca²⁺ uptake or the total pool of Ca²⁺ (Fig. 2A). The inhibitory effects were not observed using boiled TG2 or N-ethylmaleimide (NEM)-treated TG2, indicating that the catalytic activity of TG2 was required for inhibition (Fig. 2A).

Fig. 2.

TG2 inhibits IP₃R1-mediated Ca²⁺ release. (A) Effects of purified TG2 on IICR from P2/P3 vesicles after Ca²⁺ loading by adding 1 mM Mg–ATP. IICRs were evoked with 0.1 μM IP₃, and the total Ca²⁺ pool was estimated using 10 mM Triton X-100 (TX). The bar graph summarizes the mean + SD from three independent experiments using P2/P3 vesicles pretreated with TG2 (0, 1, 2, 5, 10, 20, and 50 μg/mL), 10 μg/mL boiled TG2, or 10 μg/mL NEM–TG2. The IC₅₀ value of TG2 is 34 nM, calculated by a nonlinear regression analysis using a four-parameter logistic function. (B) The BP was incorporated into the IP₃R1 by TG2-catalyzed reactions. IP₃R1 was pretreated with both TG2 and BP (1, 2, 5, 10, and 20 mM), and the oligomeric status was confirmed using AGE. (C) The IICR from TG2-pretreated P2/P3 vesicles prepared in the presence of 10 mM BP or 10 mM cystamine (CA). (D) The BP-containing (BP + TG) or BP-lacking (−TG) P2/P3 vesicles were pretreated with 10 or 0 μg/mL TG2. The IICR was measured in the presence of 0, 31, 63, 94, 125, and 156 μg/mL (slope) or 156 μg/mL (−TG) of avidin. (E) Fura-2 ratio images of living HeLa cells transfected with control (black lines) or TG2 (red lines) siRNA for 48 h were obtained under a nominally Ca²⁺-free condition. The bar graph represents the mean + SD of peak amplitudes (control siRNA, n = 117; TG2 siRNA, n = 119). ΔR indicates changes in the fura-2 ratio in response to 3 μM ATP. ***P = 8 × 10⁻¹⁸.

To elucidate the mechanism by which TG2 negatively regulates channel activity through covalent modification, we used a monoamine compound 5-biotinamido-pentylamine (BP) that contains both primary amine and biotin moieties. BP mimics a Lys residue, and TG2 can enzymaticaly incorporate BP into the Gln residue, which can be traced by horseradish peroxidase (HRP)-conjugated streptavidin. Therefore, we prepared BP-containing IP₃R1 using purified TG2 and P2/P3 vesicles (Fig. 2B). The IICR assay revealed that the BP-containing IP₃R1 has intact activity (Fig. 2C), indicating that the structures of both IP₃-binding and Ca²⁺-permeating domains are not affected by the covalent incorporation of BP molecules. Removal of TG2 and BP followed by the addition of avidin proteins with a neutralized isoelectric point (NeutrAvidin) to the BP-containing IP₃R1 inhibited IICR in a dose-dependent manner (Fig. 2D). These results indicate that steric hindrance near the BP-containing Gln residue is important for the allosteric inhibition of IICR.

To examine the inhibitory role of endogenous TG2 in living cells, Ca²⁺ imaging was performed using HeLa cells. TG2 silencing significantly enhanced ATP-evoked Ca²⁺ release in the Ca²⁺-free condition (Fig. 2E), indicating that endogenous TG2 inhibits IICR activity in living cells.

TG2 Negatively Regulates IP₃R-Mediated Autophagy.

Our data indicate that TG2 inhibits IP₃R1 function. To examine the physiological role of this regulation in the cellular functions of IP₃R1, we studied the role of TG2 and IP₃R in autophagy. IP₃R1 regulates autophagy (24–27), and TG2 is also thought to regulate autophagy (50, 51); however, the interplay between TG2 and IP₃R1 in autophagy regulation has not been studied. We hypothesized that TG2 affects IP₃R1-mediated autophagy regulation. To examine autophagy, we used HeLa cells in which both IP₃R1 and TG2 were expressed dominantly. IP₃R1 knockdown increases LC3-II (lapidated microtubule-associated protein light chain 3, LC3, an index of autophagy) content without changing the amount of actin (Fig. 3A), consistent with previous reports (24–27). Furthermore, we found that TG2 knockdown had the opposite effect (Fig. 3A).

Fig. 3.

TG2 negatively regulates IP₃R1-mediated autophagy processes. (A) LC3-I, LC3-II, IP₃R1, TG2, and actin levels in HeLa cells were evaluated 48 h after transfection with control, IP₃R, or TG2. We added 2 μM Xest or 100 μM 2-APB to HeLa cells 6 h after siRNA transfection. The bar chart represents the mean + SEM of LC3-II (LC3-II/actin ratio). ***P = 0.0011 (control versus IP₃R1), ***P = 0.0031 (control versus TG2). (B) Loss of IP₃R1 or TG2 affects autophagy as measured by GFP and RFP signals of an autophagosome indicator with tandem fluorophores (tfLC3) (52) in HeLa cells. We adjusted the sensitivity of microscopy for green and red signals using 10 nM BafA1- and amino-acid–deprived HeLa cells as negative and positive controls, respectively.

To clarify whether IP₃R1 mediated the TG2-dependent regulation of autophagy, we examined the IP₃R1 inhibitors xestospongin C (Xest) and 2-aminoethoxydiphenyl borate (2-APB). Xest and 2-APB block the TG2-dependent effect on autophagy (Fig. 3A) without affecting TG2 levels (Fig. S2A), suggesting that TG2 regulates autophagy through IP₃R1. To confirm this possibility in living cells, we observed the maturation of autophagosomes using a tandem fluorescent LC3 indicator containing both red (RFP) and green (GFP) fluorescent proteins (tfLC3) that distinguishes between autophagosomes and autolysosomes by quenching GFP fluorescence at lysosomal pH (52). In control siRNA-transfected cells, GFP and RFP puncta were independently distributed, indicating the presence of both autophagosomes and autolysosomes in the cells (Fig. 3B). However, IP₃R1 knockdown caused significant accumulation of autophagosomes that were colabeled with GFP and RFP fluorescence, whereas TG2 knockdown yielded predominantly RFP signals, as observed in amino acid deprivation (Fig. 3B). These results confirm that TG2 and IP₃R1 have opposing effects in the control of autophagy.

To verify the opposing effects of TG2 and IP₃R1 on autophagy, we monitored the dynamics of autophagosomes in living cells. Time-lapse imaging of tfLC3 revealed that the autophagosomes in control and TG2 siRNA-transfected cells actively moved inside cells, whereas the altered trafficking of autophagosomes was observed in IP₃R1 siRNA-transfected cells (Fig. S2B).

To confirm these observations, we monitored the distribution of endogenous autophagosomes and lysosomes in siRNA-transfected HeLa cells. In control and TG2–siRNA-transfected cells, LC3-labeled autophagosomes and lysosomal-associated membrane protein 1 (LAMP1)-labeled lysosomes were widely distributed in the cells and stochastically overlapped with each other (Fig. S2C). However, autophagosomes accumulated near the nucleus in cells transfected with IP₃R1 siRNA (Fig. S2C). These data confirmed the observations using tfLC3 indicators that autophagy was regulated by the opposing effects of TG2 and IP₃R1.

The IP₃R1 Gln2746 Residue Is a Potent TG2 Substrate.

To elucidate the mechanism by which TG2 inhibited IP₃R1-mediated signaling, we performed limited digestion experiments to identify the site of TG2 modification in IP₃R1. Calpain digestion of IP₃R1 pretreated with TG2 revealed a marked decrease in the C-terminal 130-kDa (C130) and 85-kDa (C85) fragments but not the N-terminal 180-kDa (N180) or 100-kDa fragments (N100) (Fig. S3 A and B). Consistent with these observations, cross-linking by TG2 after limited digestion also resulted in a marked decrease in C130 recovery, whereas the amount of N180 was not altered by TG2 (Fig. S3C). These results indicate that cross-linking occurs between the C130 fragments but not the residual N180 fragments. To confirm this, we performed limited digestion of BP-containing IP₃R1 (Fig. S3 D and E). The apparent sizes of BP-labeled bands corresponded to the C130 and C85 fragments. Further digestion confirmed that the BP-labeled bands corresponded to the C-terminal fragment. Together, these results indicated that subunit cross-linking occurred within the C130 fragments.

To precisely determine the Gln (Q) donor site for IP₃R1 cross-linking, we performed mass spectrometric analyses. We pulled down endogenous substrates of TG2 from mouse brains by incorporation with BP. A band of ∼300 kDa was identified as mouse IP₃R1 by in-gel digestion and mass spectrometry (MS), indicating that IP₃R1 is a preferred substrate of TG2 (Fig. S4 A and B). Calculating the mass of BP, we found that the carboxyl-terminal peptide was exclusively modified among 63 Gln residues assigned by mass analysis (Fig. S4C). We determined the precise modified residue in IP₃R1 using liquid chromatography–MS (LC–MS), followed by tandem MS (MS/MS). This analysis identified one transamidation site of the Gln residue at position 2746 (Gln2746) (Fig. 4A). Compared with the wild-type peptide, a clear mass shift corresponding to BP was detected in the modified peptide (Fig. 4A). The substrate specificity of TG2 was determined using the sequence motif QxPl (x, any; l, aliphatic amino acids) (53). The sequence containing Gln2746 is consistent with this motif, indicating that Gln2746, but not Gln2747, is a target of TG2. Peptides containing QxP motifs at Gln524, Gln1137, and Gln1219 in mouse IP₃R1 were detected by MS, but no BP incorporation was observed at these sites. Therefore, these Gln residues are not accessible to TG2 in the native conformation of tetrameric IP₃R1.

Fig. 4.

Gln2746 of IP₃R1 is modified by TG2. (A) IP₃R1 was modified at Gln2746 by TG2. MS/MS scanned the C-terminal peptide of IP₃R1 with ion m/z 1,935.99 (Upper) and the biotinylated peptide with m/z 2,247.15 (Lower). The b and y ions are indicated on the MS/MS spectra. The asterisk denotes a modified Gln residue (see also Fig. S4). (B) A scheme shows the primary structure of IP₃R1 and recombinant proteins corresponding to C-terminal 130-kDa fragments generated by calpain cleavage (Fig. S3). WT, a mutant in which four C-terminal amino acids were deleted (ΔQQPA), and a point mutant containing a Gln2746 to Asn substitution (Q2746N) are shown in the scheme. (C) The lower gels show intersubunit cross-linking of WT and ΔQQPA expressed in COS-7 cells. (D) Intersubunit cross-linking of WT and Q2746N expressed in COS-7 cells. Black and white arrows indicate the cross-linked dimer and its corresponding size, respectively.

The QxP motif at Gln1705 might be modified by TG2 modification; therefore, we examined the cross-linking of mutants. We generated deletion mutants of the C-terminal domain of mouse IP₃R1 (calculated mass, 132.4 kDa, corresponding to the C130 fragment) and expressed them in COS-7 cells (Fig. 4B). We found that the recombinant wild-type (WT) channel was cross-linked by TG2, but the deletion mutant ΔQQPA (Fig. 4C) and the Q2746N mutant (Fig. 4D) were not cross-linked. These results indicate that the C-terminal QxP motif at Gln2746 was essential for subunit cross-linking. This motif is conserved among vertebrates including fish, birds, and mammals but not among invertebrates, suggesting that this regulation might be specific to vertebrates. The Gln2746-containing QxP motif is not conserved among other isoforms, but mouse IP₃R2 (Gln897, Gln1221, and Gln1997) and mouse IP₃R3 (Gln896, Gln1211, and Gln1948) also contain three QxP motifs. If these motifs are exposed and accessible to TG2, IP₃R2/3 might also be TG2 targets.

Next, we examined whether the Gln2746-containg motif is sufficient for cross-linking by TG2. The NPQQPA peptide, but not NPNQPA, reacted with tetramethylrhodamine cadaverine (TMR-cad) to produce a cross-linked product with TG2 (Fig. S4D). We confirmed the structure of this purified product using MS and subsequently subjected it to a de–cross-linking reaction with TG2. Incubation of the product with both TG2 and CaCl₂ formed a product that was consistent with TMR-cad (Fig. S4E). These data indicate that the Gln2746-containg motif is essential and sufficient for cross-linking by TG2.

Either isopeptide hydrolysis and/or deamidation by TG2 can generate a glutamate residue (E) at the position of the intrinsic Gln residue (Q) (34). To test this possibility, we generated the antibody αE2746, which specifically recognizes deamidated IP₃R1 (Fig. S5A). The site-specific αE2746 antibody detected TG2-catalyzed deamidation of tetrameric IP₃R1 in P2/P3 vesicles (Fig. S5B). The αE2746 antibody revealed the existence of deamidated IP₃R1 in cerebellar IP₃R1 (Fig. S5C), and LC–MS analysis also confirmed the deamidated IP₃R1 (Fig. S5 D and E). The specificity of the αE2746 antibody was further validated using the full-length IP₃R1 with Q2746E mutation (Fig. 5A and Fig. S5F).

Fig. 5.

IP₃R1 structure and function are modified in HD models. (A) The 4C11 monoclonal antibody (Left) detected both full-length IP₃R1 and Q2746E mutant expressed in COS-7 cells, whereas our newly developed αE2746 antibody (Right) detected only the Q2746E mutant. (B) Gln2746 modification in endogenous IP₃R1 in HD models. Blotted endogenous IP₃R1 in mouse P2/P3 fractions was prepared from the cerebrum (cerebr) containing the cortex and the striatum (Upper) or the CB (Lower) of 4-wk-old WT or Tg male R6/2 mice. The loading volumes were adjusted based on the total IP₃R1 content. The bar charts represent the mean + SEM of normalized Gln2746 modification in three-litter pairs of 4-wk-old and 12-wk-old R6/2 mice. Relative amounts of Gln2746 modification were calculated from the ratio of the αE2746 signal to the 4C11 signal. Four weeks, *P = 0.025, ***P = 0.0099; 12 wk, P = 0.051 (not significant, n.s), *P = 0.049. (C) Rescue of Ca²⁺ signaling in HD patient cells by inhibition of TG2. Black and red lines show IP₃R-mediated Ca²⁺ release by anti-IgM (1.2 μg/mL) stimulation in B-lymphocytes obtained from an HD patient (HD1). Bar charts represent the mean + SEM of peak amplitudes of anti–IgM-induced Ca²⁺ release in lymphocytes from three HD patients [HD1, ***P = 0.0078 (control, n = 21; ZDON, n = 38); HD2, ***P = 7.3 × 10⁻⁵ (control, n = 34; ZDON, n = 54); HD3, ***P = 5.9 × 10⁻⁵ (control, n = 15; ZDON, n = 48)] pretreated with DMSO (control) or a site-directed TG inhibitory peptide ZDON (100 μM) for 30 min. HD1, HD2, and HD3 correspond to GM5560, GM6773, and GM4830, respectively. The peak amplitude was calculated as the difference between the mean value before stimulation and the peak value during stimulation.

Modification of IP₃R1 in HD Models.

Our data indicated that the modification by TG2 impaired IP₃R1-mediated cellular processes; therefore, we hypothesized that this regulation might lead to cellular dysregulation under pathological conditions. TG2 is up-regulated in the brain in HD model mice, including R6/2 and YAC128 mice (54–57), and HD patients (49, 58–61). Therefore, we tested whether IP₃R was modified by TG2 in these HD models.

We investigated TG2-mediated modification of IP₃R1 in an animal model of HD R6/2, which is transgenic (Tg) for exon 1 of the human Htt gene with 120 polyQ repeats, using the specific antibody that recognizes modified Gln2746 in IP₃R1 (Fig. 5A). The brain P2/P3 vesicles were obtained from the cerebrum including the striatum and the cortex or the cerebellum (CB) of three litter pairs of 4-wk-old and 12-wk-old HD model mice (R6/2 Tg or WT); we observed endogenous IP₃R1 using αE2746 and 4C11 that detects both modified and unmodified IP₃R1. The IP₃R1 expression levels were suppressed particularly in 12-wk-old HD mice (16, 28); thus, we normalized the αE2746 signal to the 4C11 signal. The relative level of the αE2746 signal was significantly higher in the brains of both 4-wk-old and 12-wk-old R6/2 Tg mice compared with WT mice (Fig. 5B).

TG2 up-regulation has been observed in the brain and lymphocytes of patients with HD (49, 58–61); therefore, we examined a role of TG on IP₃R activity in B-lymphocytes from HD patients. Pretreatment with a specific TG blocker (Z-DON–Val–Pro–Leu–OMe, ZDON) (56, 62) enhanced IP₃R-mediated Ca²⁺ release during B-cell antigen receptor stimulation by the anti-IgM antibody, which evoked IICR (63) in lymphocytes from a patient with HD (Fig. 5C). Notably, this effect was consistently observed in three individuals with HD. This indicates that in addition to the modification in the structure of IP₃R in HD model mice (Fig. 5B), TG modifies the function of IP₃R in B-lymphocytes from HD patients (Fig. 5C).

We observed the effects of the human Htt mutant on Ca²⁺ signaling in other HD models. The induced expression of the full-length human Htt including 145 polyQ in PC12 cells impaired acetylcholine (ACh)-evoked Ca²⁺ release under Ca²⁺-free conditions, whereas the enhanced ACh-evoked Ca²⁺ increase (Fig. S6A) was observed in the presence of extracellular Ca²⁺ in line with the sensitized Ca²⁺ increase in YAC128 neurons (64–66). To verify the inhibitory effect of the Htt mutant, we observed Ca²⁺ release in Neuro2a cells transiently expressing the myc-tagged full-length Htt including 23 polyQ (Htt23Q–myc) or 145 polyQ (Htt145Q–myc). The Ca²⁺ release upon stimulation by bradykinin (BK) in Neuro2a cells expressing with Htt145Q–myc was less than Htt23Q–myc under Ca²⁺-free conditions (Fig. S6B). Transient expression of Htt exon1 including 60 polyQ (eHtt60Q–EGFP) or 150 polyQ (eHtt150Q–EGFP) also impaired Ca²⁺ release in Neuro2a cells compared with the control (eHtt18Q–EGFP) (Fig. S6C), and the 3,5-dihydroxy phenylglycine (DHPG)-evoked Ca²⁺ increase in eHtt150Q–EGFP-expressing neurons was lower compared with eHtt18Q–EGFP-expressing neurons (Fig. S6D). To examine the involvement of TG2-dependent modification, we transfected Neuro2a cells coexpressing eHtt150Q–EGFP with either a Gln2746 mutant fragment (AQPA) or a Gln2746-containing fragment (QQPA). Overexpression of QQPA enhanced Ca²⁺ release compared with AQPA (Fig. S6D), suggesting the involvement of TG2-dependent modification. We also examined the Gln2746 modification in Neuro2a cells inducibly expressing Htt exon1 including 16 polyQ (eHtt16Q–EGFP) or 150 polyQ (eHtt150Q–EGFP) using the αE2746 antibody. Western blotting of endogenous IP₃R1 (Fig. 6E) and immunostaining of the overexpressed IP₃R1 C terminus (Fig. 6F) demonstrate that the αE2746-detectable modification was enhanced in the Neuro2 cells with eHtt150Q–EGFP after 48-h induction compared to those with eHtt16Q–EGFP.

Fig. 6.

Model for TG2-mediated regulation of IP₃R1. Schematic diagrams indicating the subunit arrangement of the allosteric coupling domain within the C-terminal 130-kDa region (C130) of IP₃R1. The Q-donor at the C terminus is located in close proximity to the K-donor region of the neighboring subunit. We propose reversible configuration changes between the closed and open states in tetrameric IP₃R1. TG-catalyzed transamidation causes covalent intersubunit cross-linking, which locks the closed configuration. A large molecule such as avidin but not a small moiety such as BP inhibits the motion by amine incorporation. Gln2746 of IP₃R1 is deamidated by TG2 and converted into a glutamate residue that is resistant to TG2-mediated inhibition.

Discussion

The role of defective allostery in ER-resident calcium channels in subsequent cellular defects in neurological disorders remains unclear. The current study addresses this issue, demonstrating a novel allosteric mechanism—the covalent modification of IP₃R structure and function—as well as its implications for the etiology of neurological diseases. Specifically, we discovered that TG2 in the brain regulates IP₃R1 by catalyzing a posttranslational modification at the Gln side chain of the Gln2746 residue. Up-regulation of TG2 enhanced the Gln modification of IP₃R1, and TG2 suppressed IP₃R1-mediated Ca²⁺ signaling and subsequent cellular processes. Modification of Gln2746 and IP₃R1 function was observed in HD models and B-lymphocytes from HD patients; these results therefore suggest a novel mechanism linking the activation of brain TG to the control of IP₃R-mediated Ca²⁺ signaling in HD. Based on these results, we propose a general model for the progressive brain dysfunction observed in brain pathogenesis that involves aberrant IP₃R signaling.

An important point in this study is the identification of TG2 as a chronic regulator of IP₃R1 structure and function. This finding has broad implications for pathological neuronal function and degeneration because it links TG2 functions in cellular differentiation, cellular stress, and inflammation (34, 35) to the chronic defects in Ca²⁺ signaling. Numerous literatures have demonstrated that TG2 is up-regulated in various diseases including HD (49, 54, 55, 59–61), AD (37, 38, 40, 42, 43, 45), and PD (39, 44). The splicing mechanism (67, 68) and stress-dependent inductions (34, 35, 69–72) are considered to elucidate the up-regulation, and the increased Ca²⁺ levels observed in many diseases (4, 17–22) are also competent to trigger TG2 activation. The Ca²⁺-dependent activation of TG2 inhibits the IP₃R1-mediated Ca²⁺ release, as revealed in the current study; thus, this enzymatically allosteric modulation should primarily contribute to a negative feedback mechanism to maintain homeostatic control of intracellular Ca²⁺ levels. However, this modification subsequently would have a chronically negative effect on Ca²⁺ signaling, because it is much more static than noncovalent allosteric modulators required for the control of rapid and dynamic Ca²⁺ signaling such as Ca²⁺ oscillation. Therefore, the TG2-dependent allosteric modulation should chronically set in motion cascades of defective signaling related to pathological states; up-regulation of TG2 may disrupt the IP₃R1-mediated plasticity of neuronal signaling (3, 10–12) and the homeostatic control of cellular processes including mitochondria energy production (23), autophagy regulation (24–27), ER stress (28), and others (29–31) and ultimately increases cellular stress, resulting in further TG2 up-regulation proceeding a vicious cycle. Given the widespread brain expression profiles of TG2 and IP₃R1, we propose that the mechanism described in this study might potentially serve as a general principle for many other diseases of the brain or other tissues in which TG2 is up-regulated by a disease-initiating agent. In this study, we assume that the disease-initiating agent is the mutant Htt; future studies are required to examine whether other disease-initiating agents such amyloid proteins in AD and α-synuclein in PD also would cause chronic TG2 activation, resulting in IP₃R modification and aberrant calcium signaling, and suggesting its presumable role in brain dysfunctions including cognitive declines occurs frequently during the prodromal states of these neurodegenerative diseases (73–77).

This study revealed that the covalent modification by TG2 results in three alternative states, as shown in Fig. 6; thus, the other roles of TG2 in Ca²⁺ signaling are also conceivable. TG2 inhibits IICR activity by both cross-linking and steric hindrance around Gln2746 (Fig. 2), whereas functional analysis of Q2746E indicates that deamidation of IP₃R1 suppresses TG2-mediated inhibition (Fig. S6G). All potential modifications are shown in the schematic diagram, but the overall activity of a single channel would depend significantly on the relative number of cross-linked isopeptide bonds, amine incorporations, and deamidation in the tetramer. Acute increases in Ca²⁺ levels cause subunit cross-linking in neuroblastoma and glioblastoma cells (Fig. 1E), whereas amine incorporation and deamidation were observed in resting cells and brain tissues without stimulation by increasing Ca²⁺ concentrations (Fig. 5B and Figs. S5 C–E and S6 H and I). The deamidation reaction by TG2 is dependent on pH (78) and substrate affinity (79); therefore, the relative abundance of these three modifications is presumably determined by the cellular state and the conformation around Gln2746. These irreversible processes would contribute to chronic changes in homeostasis of Ca²⁺ signaling and subsequent cellular processes, whereas the TG2-resistant states including the small amine-containing IP₃R1, such as BP incorporation shown in Fig. 2C and/or the deamidated IP₃R1 shown in Fig. S6G, might contribute to Ca²⁺-dependent cell death, including apoptosis (2, 17), due to the lack of negative feedback regulation by Ca²⁺. Although deletion of the C-terminal 12 amino acids has no effect on IICR (80), binding of the specific monoclonal antibody 18A10 (81) and GIT1 to this region inhibits IICR (82), similar to the BP–avidin complex in this study. Because both IgG and GIT1 form bulky architectures, these observations are consistent with our conclusion that large molecules such as avidin but not small moieties such as BP interfere with configuration changes by steric hindrance. IICR is regulated by binding of the C-terminal region near Gln2746 by various proteins governing cellular fate—for example, Bcl-X_L, Bcl-2, cytochrome c, HAP1, and Htt (4, 5, 17, 18); thus, we speculate that the IP₃R1 C terminus might act as a signaling hub whose structure affects cellular processes by regulating the subunit configuration by steric hindrance. Considering the direct activation of IP₃R1 by the Htt mutant (18), the net activity of IP₃R1 may be comprehensively determined by the Htt mutant and TG2 in HD. The impairment of Ca²⁺ release by eHtt150Q–EGFP was more potent than Htt145Q–myc (Fig. S6 B and C), consistent with higher TG2 activities in R6/2 than YAC128 mice expressing the full-length Htt mutant (55, 57) and the sensitized Ca²⁺ increase in YAC128 mice more than shortstop mice expressing Htt exon1/2 (64). The poly-Gln repeat of the Htt mutant is a favorable substrate of TG2 (41); thus, the possibility of substrate competition and steric hindrance around Gln2746 is also interesting. Thereby, we suggest that the prevalence of full-length Htt mutant and TG2 modulations on IP₃R1 may determine the total net activity of Ca²⁺ signaling from the ER and the pathological states in HD—that is, the prodromal decline of neural signaling or the degeneration by apoptosis.

The first insight into the mechanism underlying the coupling between the IBD and the channel domain by IP₃-evoked structural changes was described more than two decades ago (83). From our early EM studies of negatively stained IP₃R1 using heparin-gold labeling, we proposed a long-range coupling of IBD to the channel domain by Ca²⁺-dependent structural changes (84–86). However, direct binding of the IBD to the channel domain was postulated, and a model was proposed in which an N-terminal suppressor domain directly binds a loop between putative transmembrane helices (87–90). Recently, highly improved cryo-EM reconstruction was reported (9), and by fitting calculations using X-ray structures (7, 8), it was concluded that direct interaction of the IBD with the channel domain is unlikely (8). However, because the X-ray structure of the IBD is limited to only 5% of each tetramer, which is equivalent to 20% of each subunit, the role of the other domains in channel gating remains unclear (91). Here, we show new evidence demonstrating that configuration changes outside the IBD are required for channel activation (Fig. 6). Our data are consistent with the conclusion that the Gln2746 in the C-terminal tail is proximal to the neighboring IP₃R1 subunit, and TG2 covalently tethers Gln2746 to the adjacent subunit by the Gln–Lys isopeptide bond that locks the channel in a closed configuration (Fig. 6). Notably, our BP incorporation experiments indicate that the isopeptide bond occurs outside the IBDs without impairing the intrinsic structures required for IP₃ binding and Ca²⁺ permeation (Fig. 2 B–D). IP₃R1 function was not inhibited by transamidation of a small molecule such as BP; however, the subsequent addition of a bulky molecule such as avidin completely blocked receptor function (Fig. 2 B–D), indicating that steric hindrance near Gln2746 is crucial. We propose that an “allosteric coupling domain” in the C-terminal 130-kDa region (C130) exists between the IBD and the transmembrane domains to couple the IBD to the channel and that configuration changes in this domain are required for channel activation (Fig. 6). The EcoRI fragment-encoded short domain (ES; amino acids 2216–2749) of IP₃R1 is sufficient for tetramer formation (92), but the channel in the ES tetramer is always open (93). Therefore, the allosteric coupling domain presumably keeps the channel closed, and the transition to the open configurations could revert to the default channel state that constitutively permeates Ca²⁺—for instance, the leaky channel assembled from ES domains.

What is the key mechanism regulating the allosteric transitions during channel activation? Two intrinsic determinants stabilizing the subunit configurations can be considered a priori—IP₃ and Ca²⁺—because channel activation requires both IP₃ and Ca²⁺. In fact, the recent X-ray structures of the IBD indicate that conformational changes occur upon IP₃ binding (7, 8); our group and others have demonstrated Ca²⁺-induced structural changes (84, 85, 94, 95). Fluorescence resonance energy transfer (FRET) analysis of recombinant IP₃Rs also showed reversible structural changes induced by physiological concentrations of IP₃ and Ca²⁺ (96). Considering these observations and our current data, we propose that structural changes by IP₃ and Ca²⁺ determine the configuration of the allosteric coupling domain. Given the Ca²⁺-sensor site at E2100 (32) and the deletion site in opisthotonos mice (33) are located in this domain, these allosteric sites presumably control subunit configurations within this domain. These ligands—that is, IP₃, Ca²⁺, and ATP—might confer a predominantly open configuration by noncovalent binding, but the covalent Gln–Lys isopeptide bonds and steric hindrance around Gln2746 are more energetically stable; therefore, these modifications can robustly and chronically lock the closed subunit configuration.

In summary, we identified TG2 as an IP₃R1-modifying enzyme that irreversibly controls the essential function of IP₃R1 in Ca²⁺-signaling and subsequent cellular processes whose homeostasis is crucial for cellular fate and uncovered the allosteric mechanism by which IP₃R channel activation is controlled by ligand binding. Many functional proteins undergo allosteric regulation; therefore, understanding allosteric mechanisms can elucidate the mechanisms underlying functions of these proteins and pave the way for drug design (1). This is the first demonstration, to our knowledge, of IP₃R1 allosteric transitions locked by TG2-catalyzed cross-linking, and these observations might provide a new framework for TG2-dependent enzymatic regulation of ligand-gated ion channels and allosteric proteins similar to IP₃R1.

Materials and Methods

Cell Culture.

SH-SY5Y, U87-MG, HeLa, and COS-7 cells were cultured in DMEM containing 10% (vol/vol) heat-inactivated FBS, 50 U/mL penicillin, and 50 μg/mL streptomycin. Human B-lymphocytes from patients with HD were obtained from the Coriell Institute (GM5560, GM6773, and GM4830) and cultured in RPMI medium 1640 containing 10% (vol/vol) heat-inactivated FBS, 50 U/mL penicillin, and 50 μg/mL streptomycin. All cells were maintained in 5% (vol/vol) CO₂ at 37 °C. The differentiation medium supplemented with 20 μM RA was replaced every 48 h for the SH-SY5Y and U87-MG cells. All of the animals were ethically treated according to the guidelines of the Animal Experiments Committee of the Institute of Physical and Chemical Research (RIKEN) Brain Science Institute.

Transfection of siRNAs or Plasmids.

The siRNAs and human negative controls were purchased from B-Bridge. The nucleotide sequence of the TG2 siRNA was ggucaaugccgacguggua. IP₃R1 knockdown was performed as described previously (28). The siRNA and ptfLC3 were transfected in HeLa cells using Lipofectamine 2000 (Invitrogen) and TransIT (Mirus) transfection reagents, respectively.

Preparation of P2/P3 Fractions.

Mouse brains were washed extensively with chilled PBS and were then homogenized in a homogenate buffer (pH 7.5) composed of 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (Hepes)/sodium hydroxide (NaOH), 0.32 M sucrose, 1 mM ethylene diamine tetraacetic acid (EDTA), 2 mM Tris(2-carboxyethyl)phosphine (TCEP), and protease inhibitors (0.2 mM phenylmethylsulfonyl fluoride, 10 μM leupeptin, 10 μM pepstatin A, and 10 μM E-64). The homogenates were centrifuged at 800 × g for 5 min at 4 °C, and the resultant supernatants were subjected to ultracentrifugation at 105,000 × g for 60 min at 2–4 °C as previously described (97). The precipitates were suspended in resuspension buffer ResB (pH 7.5), containing 50 mM Hepes–NaOH, 150 mM NaCl, 1 mM EDTA, 2 mM TCEP, and protease inhibitors. The supernatant obtained after ultracentrifugation was used as the cytosolic fraction. The protein concentrations in the solution were estimated, adjusted to 3 mg/mL using ResB, and the samples were then quickly frozen in liquid nitrogen for storage until use in various experiments.

Analysis of Intersubunit Cross-Linking Within IP₃R1 by Electrophoresis.

We estimated the cross-linking between IP₃R1 subunits using 1% AGE. The samples containing mouse cerebellar P2/P3 proteins (0.75–2.25 μg) were loaded in individual wells. Agarose was dissolved in a buffer containing 0.1% SDS using a microwave device. The agarose was heated and then gently mixed by manual rotation. The 1% agarose solution was poured into an acrylic tray (Nihon Eido) in which a 26-lane comb was placed. After 30 min at room temperature, the apparatus was filled with the electrolyte buffer, and the comb was carefully removed. Electrophoresis was performed at 100 V for 1–2 h. Immediately following electrophoresis, the agarose gels were placed in transfer buffer for 30 min. The separated proteins were transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore). IP₃R1 was detected using the 4C11 monoclonal antibody as described previously (28, 97). A molecular marker was prepared by cross-linking bovine thyroglobulin with 1 mM 1,8-bis-maleimido-diethyleneglycol overnight at 4 °C.

Measurement of Ca²⁺ Release Activity of IP₃R1.

We loaded fura-2AM (Dojindo) into HeLa cells and B-lymphocytes by incubation with 5 μM fura-2AM in balanced salt solution (pH 7.4) containing 115 mM NaCl, 5.4 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 20 mM Hepes, and 10 mM glucose for 30 min at room temperature, and performed Ca²⁺ imaging using inverted fluorescence microscopy with a 20× objective lens (S Fluor 20× N.A. 0.75) (Nikon) and AQUACOSMOS (Hamamatsu Photonics) to obtain F340/F380 ratio images as described previously (28, 98). For measurements of IICR, the P2/P3 membranes (3 mg/mL protein) of mouse cerebella were washed three times with a buffer (pH 7.4) containing 50 mM Hepes–KOH, 110 mM KCl, 10 mM NaCl, and 5 mM KH₂PO₄. The washed 10-μL P2/P3 membranes (2 mg/mL protein) of the cerebella were added to 287 μL of a buffer containing 1 mM 2-mercaptoethanol, 1 μM CaCl₂, and 0.2 μM fura-2 in a cuvette. To introduce Ca²⁺ into P2/P3 membrane vesicles, we added 3 μL of 100 mM Mg–ATP to the cuvette with stirring. Ratiometric measurement of fura-2 was performed at 30 °C using a CAF-110 (Jasco) or an F-2500 spectrofluorometer (Hitachi).

Analysis of Autophagosome Dynamics by Fluorescence Microscopy.

We used ptfLC3 plasmids (52) expressing mRFP–GFP–LC3 to monitor autophagic processes in living cells. Cells plated in a 35-mm glass-bottomed dish were transfected with siRNA in Lipofectamine 2000 transfection reagent (Invitrogen) and then with 1 μg ptfLC3 by TransIT transfection reagent (Mirus) 6 h after the siRNA transfection. We observed GFP and RFP fluorescence signals in living cells under an FV1000D confocal microscope (Olympus) with a stage top incubator in 5% CO₂ at 37 °C.

Affinity Purification of TG2 Substrates.

We incubated the reaction solution containing the P2/P3 fractions (1 mg/mL protein), 50 mM Hepes–NaOH (pH 7.5), 150 mM NaCl, 10 mM BP, 1 mM EGTA, 2 mM CaCl₂, and 30 ng/mL TG for 30 min at 37 °C. The mixture was centrifuged at 20,000 × g for 30 min at 4 °C, and the pellets were washed with 50 mM Hepes–NaOH (pH 7.5). The proteins were solubilized with 1% (wt/vol) 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) and mixed with avidin–agarose. After washing three times with the 1% CHAPS-containing Hepes buffer, the biotin-containing proteins bound to avidin–agarose were subjected to SDS/PAGE and were detected by Coomassie brilliant blue (CBB) staining or Western blotting analysis. The CBB-stained band was excised, digested using an in-gel digestion protocol, and analyzed by MS using MALDI-TOF MS (ABI).

Specific Antibody Against Modified IP₃R1.

To examine the status of Gln2746 in mouse IP₃R1, a deamidated peptide (PPHMNVNPEQPA) corresponding to amino acids 2738–2749 was synthesized and conjugated to keyhole limpet hemocyanin via a cysteine added to the N terminus of the peptide. Japanese white rabbits were immunized with these peptides. The resulting polyclonal antibody was first applied to an affinity resin immobilized with a control peptide (PPHMNVNPQQPA) via cysteine added to the amino terminus, and the flow-through was applied onto a resin immobilized with the deamidated peptides. The specific antibody against the deamidated peptide was eluted with 100 mM glycine–HCl buffer (pH 2.5). To verify the specificity of the antibodies, we prepared recombinant IP₃R1, which was expressed as an EGFP–fusion protein (28). The PCR product was inserted into EGFP–IP₃R1-containing pcDNA3.1 vector with EcoRI and XhoI. Mutation of the Gln residue at the amino acid position of 2746 was confirmed by sequence analysis using an ABI 3730xl automatic sequencer. To confirm the specificity of the E2746 antibody, the C-terminal fragment of IP₃R1 (66 amino acids) was subcloned into the bacterial expression vector pGEX-6p (GE Healthcare) using EcoRI and SalI. Quantification of modified or native IP₃R1 was performed by densitometric analysis of Western blots using Image J software (National Institute of Health).

Statistical Analysis.

All probability calculations were performed using unpaired and two-tailed Student t tests. The significance after multiple comparisons was estimated by the Bonferroni method or by Dunnett’s method. For all data, no statistics were used for predetermination of the sample size, randomization, or blinding.