Pentameric quaternary structure of the intracellular domain of serotonin type 3A receptors

In spite of extensive efforts over decades an experimentally-derived structure of full-length eukaryotic pentameric ligand-gated ion channels (pLGICs) is still lacking. These pharmaceutically highly-relevant channels contain structurally well-conserved and characterized extracellular and transmembrane domains. The intracellular domain (ICD), however, has been orphaned in structural studies based on the consensus assumption of being largely disordered. In the present study, we demonstrate for the first time that the serotonin type 3A (5-HT3A) ICD assembles into stable pentamers in solution in the absence of the other two domains, thought to be the drivers for oligomerization. Additionally, the soluble 5-HT3A-ICD construct interacted with the protein RIC-3 (resistance to inhibitors of cholinesterase). The interaction provides evidence that the 5-HT3A-ICD is not only required but also sufficient for interaction with RIC-3. Our results suggest the ICD constitutes an oligomerization domain. This novel role significantly adds to its known contributions in receptor trafficking, targeting, and functional fine-tuning. The innate diversity of the ICDs with sizes ranging from 50 to 280 amino acids indicates new methodologies need to be developed to determine the structures of these domains. The use of soluble ICD proteins that we report in the present study constitutes a useful approach to address this gap.

the transmembrane domain of full-length receptors 5 . The chimeric 5-HT 3A -ICD construct, hereafter referred to as 5-HT 3A -ICD, was expressed in E. coli and purified from the soluble fraction after cell disruption by affinity chromatography using amylose resin and size exclusion chromatography (SEC) (Fig. 1). The presence of the complete ICD was confirmed by mass spectrometry (MS/MS, MALDI-TOF/TOF 4800 mass spectrometer, Applied Biosystems). On SDS-PAGE gels the protein ran at 53 kDa, while higher molecular weight bands approximately every 50 kDa were discernible clearly with Western blot detection.
Using SDS-PAGE and denaturing microfluidic capillary gel electrophoresis (Bioanalyzer, Agilent) the weight of purified monomeric 5-HT 3A -ICD construct was determined to be 53 kDa (theoretical weight 54 kDa) (Figs 2 and 3). The peak fraction analyzed by SEC eluted as a single symmetric peak indicating assembly into a defined and stable oligomeric state. The elution volume (11.52 ± 0.02 mL, n = 5) of 5-HT 3A -ICD was significantly smaller than what would correspond to monomeric 5-HT 3A -ICD, indicating higher oligomeric assemblies. MBP elutes at 15.2 mL yielding a predicted weight of 44 kDa (theoretical weight 41 kDa), indicating that MBP is a monomeric protein under the experimental conditions tested 7,8 . Importantly, MBP has been extensively used to create fusion proteins for structural studies, including peptides of similar or even shorter lengths as studied here (Tom20, 61 amino acids 9 ; MCL1, 147 amino acids 10 ; PilA1, 133 amino acids 11 ; Norrin, 103 amino acids 12 , Menangle virus phosphoprotein, 49 amino acids 13 ; adenylate cyclase 1 Receptor-short N-terminal extracellular domain, 93 amino acids 14 ). If oligomeric assemblies were observed in X-ray structures of MBP fusion proteins, oligomerization was always mediated by the fusion protein of interest and not MBP. The elution volume of 5-HT 3A -ICD chimera during SEC corresponds to a weight of 265.1 ± 2.8 kDa (n = 5), indicative of a pentameric assembly. When the purified protein was treated with the cross-linker glutaraldehyde using a vapor diffusion method 15 and then separated by either SDS-PAGE or denaturing capillary gel electrophoresis, a ladder was observed, representing monomer, dimer, trimer, tetramer and pentamer (Fig. 3). Pretreatment of 5-HT 3A -ICD with SDS obliterated the higher oligomeric assemblies (Fig. 3). Dynamic light scattering (DLS) experiments yielded a particle size corresponding to 245 ± 19 kDa (n = 5). To further ascertain the results obtained by SEC, chemical cross-linking and DLS, we performed SEC coupled with multi-angle light scattering (SEC-MALS), a molecular shape-independent method, to directly establish the absolute molecular weight of purified protein in solution. Analysis of four separate experiments yielded a weight-average molecular mass of 250.8 ± 4.4 kDa (mean ± S.E.M., n = 4) with a dispersity term of 1.010 ± 0.004 (mean ± S.E.M., n = 4) (Fig. 2, Table 1), again confirming the pentameric assembly of 5HT 3A -ICD. The result that the 5-HT 3A -ICD chimera conferred pentameric assembly is surprising.
In pentameric ligand-gated ion channels (pLGICs), the three domains found along an axis perpendicular to the membrane are the extracellular (ECD), transmembrane (TMD), and intracellular domain (ICD). Structurally, at the level of each domain within a receptor, five-fold symmetry is found. Previously, ECDs and TMDs have been expressed individually and their oligomerization studied. The present study is the first to show pentameric assembly of the ICD alone. Originally, intersubunit-interactions between ECDs were thought to drive the assembly and stoichiometry of pLGICs [16][17][18][19] . This suggested that the ECD would be an autonomous folding and assembly domain. Indeed, AcetylCholine Binding Protein (AChBP) 20 , a soluble protein found in the snail Lymnaea stagnalis, that is homologous to the ECD of nAChR, forms pentameric assemblies in solution. Similarly, engineered ECD-only constructs of pLGICs that natively form functional homopentamers, including GlyRα 1 and nAChRα 7 subunits, assemble into pentamers in solution, that preserve high-affinity ligand-binding capability 21,22 . Conversely, and unsurprisingly, the ECD of nAChRα 1 that normally is found in heteropentamers does not assemble to higher homooligomeric states and in the X-ray structure nAChRα 1 ECD was monomeric 23 . In summary, these individual ECDs parallel the oligomerization properties of their full-length receptors.
Conversely, a construct of the prokaryotic Gloeobacter violaceus pLGIC (GLIC) that solely consists out of the ECD is monomeric in solution even at very high concentration and only forms hexa-but not pentameric structures in the crystallized form 24 . Full-length GLIC that consists of ECD and TMD has so far been only described as a pentamer 25,26 . It was postulated that the GLIC TMD is required for correct pentameric assembly 24 . Indeed, for diverse eukaryotic pLGICs, aromatic residues in non-channel-lining transmembrane helices (TM1, TM3, and TM4) network to determine α -helical packing and thus subunit stoichiometry 27 . Several pLGIC TMDs expressed individually assemble as pentamers in lipid micelles. For example, cation-conducting nAChRα 7 TMDs as well as nAChRα 4β 2 TMDs with short M3-M4 linkers, GGGEG and G 5 , respectively, assemble as pentamers in lauryldimethylamine-oxide (LDAO) micelles 28,29 , and anion-conducting GlyR TMDs form pentamers in lysophospholipid lyso-1-palmitoylphosphotidylglycerol (LPPG) 30 . Hydrophobic contacts, especially leucine-isoleucine, and leucine-leucine interactions, drive these assemblies. Interestingly, separate nAChR α 4 and β 2 TMDs alone also form homopentamers, whereas the complete subunits do not form such homopentamers, and instead only obligate heteropentamers, indicating that domains apart from the TMD control the defined stoichiometric assembly in full-length receptors. Only a few previous studies investigated expression of solely the ICD. No pentamers were observed and spectroscopic methods as well as limited digests lead to the conclusion that the ICD was unfolded 31,32 . In general, the ICD has largely been thought of as a region with disorganized structure 3,33 . Intriguingly, we show here, that the complete 5-HT 3A -ICD when fused to the soluble monomeric MBP confers pentameric architecture to the resulting soluble chimera. Only stable and defined pentamers were observed, without evidence for other oligomeric assembly states. The driving force behind this observed pentamerization must lie in sequences directly contributed by the ICD.
RIC-3 is a crucial chaperone protein for some pLGICs that influences their functional maturation 34 . For mouse 5-HT 3A receptors expressed in Xenopus laevis oocytes we have shown that RIC-3 co-expression attenuated serotonin-induced currents almost completely, as compared to 5-HT 3A expressed alone. When the 115-amino acid 5-HT 3A -ICD is replaced with the short heptapeptide M3-M4 linker of GLIC, this RIC-3 effect is abolished 6 . In reverse chimeras consisting of GLIC-ECD and TMD with the 5-HT 3A -ICD added to obtain a GLIC-5-HT 3A -ICD chimeras we showed that RIC-3 co-expression similarly only modulated the 5-HT 3A -ICD-containing chimeras, but not GLIC 35,36 . These results indicate that the 5-HT 3A -ICD is required for RIC-3 modulation. By using both proteins, GLIC-5-HT 3A -ICD and RIC-3, individually purified after expression in E. coli, in pull-down experiments, we showed that GLIC-5-HT 3A -ICD interacts with RIC-3 37 . Under the same conditions wild-type GLIC that does not contain any ICD is not pulled down. These experiments demonstrate that the interaction between the 5-HT 3A -ICD and RIC-3 is direct and not mediated by other proteins. Here, the soluble 5-HT 3A -ICD chimera was used in pull-down experiments using metal affinity resin directed towards a C-terminal His 6 tag in a RIC-3 construct (hRIC-3(His) 6 ). The soluble 5-HT 3A -ICD chimera was successfully pulled down in these experiments providing evidence that the 5-HT 3A -ICD is not only required but also sufficient for the interaction with RIC-3 (Fig. 4). These studies can be extended to the multitude of other intracellular proteins that interact with pLGIC ICDs. The characterization of such interaction sites may provide targets for future drug design.
It is astounding that the ICD that is still thought to be a vastly disordered domain drives pentamerization when expressed alone, and it is intriguing to propose that it is also a driver for pentamerization of pLGICs as a whole. Importantly, quaternary polymorphism as observed for many other oligomeric proteins including the GLIC ECD is absent for the 5-HT 3A -ICD. The soluble 5-HT 3A -ICD chimera construct is now a perfect starting point for determining its high-resolution structure and also for identification of molecular features inside the ICD that mediate RIC-3 as well as other cytosolic protein interactions.
The recent 5-HT 3A X-ray structure including the proteolyzed ICD showed that lateral windows identified as ion pathways in the Torpedo nAChR model connecting the intracellular channel cavity with the cytosol were entirely occluded by post-M3 loops 5,38 . Additionally, three MA-helix arginines previously thought to   electrostatically repel conducted cations and be therefore causative for the sub-pSiemens single-channel conductance were involved in inter-subunit salt bridges 5,38 . The structure does also not provide an intracellular pathway for cations, albeit the crystallized construct had a significantly increased single-channel conductance 5 . Clearly, further detailed studies of the 5-HT 3A -ICD are therefore needed to elucidate the intracellular ion pathway as well as opening mechanism, and to reconcile previous functional studies with structure. Due to the diversity of the ICD in this large ion-channel superfamily, more than 40 subunits in humans, investigations are needed for diverse pLGICs. Highly-specialized mechanisms for the fine-tuning of targeting, trafficking or even functional aspects have evolved to be increasingly complex in metazoans over unicellular organisms. Importantly, it is becoming increasingly apparent that the ICD is involved in all aspects of pLGIC from folding and structural assembly to modulation of agonist efficacy and desensitization.

Methods
Plasmids. The cytoplasmic domain of the mouse 5-HT 3A receptor fused to the C-terminus of maltose binding protein (MBP) was generated using a modified pMAl-c2x (New England Biolabs) vector, pMALX 7 . The 5-HT 3A -ICD (QDL… VGY) was amplified by PCR using appropriate NotI and BamHI forward and reverse primers, respectively. The BamHI reverse primer also included a stop codon. The final construct, MBP-5-HT 3A -ICD-pMALX, was obtained after double-digest of the PCR-amplified ICD as well as pMALX, and subsequent ligation.
where V e is the elution volume, V o is the column void volume (8.16 mL), and V c is the geometric column volume (24 mL). For each standard protein, the calculated K av value was plotted against the log of its known molecular weight (Log M r ). The standard curve analysis algorithm of SigmaPlot 13.0 (Systat Software Inc., San Jose, CA) was then employed to obtain a predicted value (x) of the molecular weight of the fusion protein using a linear equation: where a is the slope and y 0 is the y-intercept. The obtained molecular weight of the fusion protein represents the average of at least five separate experiments.
Chemical Cross-Linking. For investigating the oligomeric state of fusion protein, glutaraldehyde-mediated cross-linking experiments were performed at 30 °C by utilizing a vapor-diffusion technique. Briefly, 40 μL of 25 % (v/v) glutaraldehyde (Sigma-Aldrich) was placed in the bottom of a microfuge tube and acidified with 1 μL of 5 N HCl. 15 μL of buffer B containing 0.5 μg of purified fusion protein was placed on the inside of the cap of the same microfuge tube. Upon carefully closing the cap, the protein solution was allowed to form a hanging drop inside the tube, and separated at ≤1 cm from glutaraldehyde. The entire 'closed microfuge tube system' was then incubated in a mini dry-bath at the temperature indicated. The protein solution was then carefully transferred into a separate tube to quench the reaction by adding sodium dodecyl sulfate (SDS) sample buffer (62.5 mM Tris HCl, pH 6.8, 10% (v/v) glycerol, 2% (w/v) SDS, 100 mM dithiothreitol (DTT), 0.01% (w/v) bromophenol blue). All samples were boiled and then resolved by SDS-PAGE. Cross-linked proteins were transferred to polyvinyl difluoride (PVDF) membrane and immunoblotted as described later. The analysis was repeated at various intervals of time after the start of the reaction to reveal a ladder of the oligomeric states. Alternatively, in a similar experiment, protein was exposed to glutaraldehyde vapors for various time intervals as described above and the cross-linked proteins were resolved by 'On-Chip-Electrophoresis' under reducing Scientific RepoRts | 6:23921 | DOI: 10.1038/srep23921 conditions using the Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.). Protein chip and loading samples were prepared using the manufacturer provided kit, and according to instruction manual. Dynamic Light Scattering. Dynamic light scattering was used for particle size determination. Purified protein was diluted to a protein concentration of 1.0 mg/mL, filtered with a 0.22 um Millipore Millex-GV filter (Millipore, Billerica, MA) right before measurements, and subjected to the light scattering measurements at 90° using a Brookhaven Instruments (Brookhaven, CT) BI-200SM device with an avalanche photodetector. The molecular mass of the protein was calculated from the measured hydrodynamic radius and a calibration curve that included seven soluble proteins (ribonuclease A, carbonic anhydrase, bovine serum albumin, conalbumin, aldolase, ovalbumin, and IgG), and three membrane proteins (KcsA, connexin 26 hemichannels, and P-glycoprotein).

Size Exclusion Chromatography in tandem with Multi-angle Light Scattering (SEC-MALS).
SEC-MALS analysis was utilized to determine the weight-average molar mass (M w , g mole −1 ) and dispersity (Ð, M w /M n , where M n is the number-average molar mass) of purified protein independent of its shape. Chromatographic separation of protein samples was obtained under isocratic conditions using a Superdex TM 200 HR 10/300 GL column (GE Healthcare), pre-equilibrated with two column volumes of an SEC buffer (20 mM HEPES, 150 mM NaCl, 5 mM maltose, 0.01% NaN 3 , pH 7.4). The individual run was developed at an ambient temperature while a flow rate of 0.3 mL/min was maintained on an APLC system (LabAlliance, State College, PA) equipped with a solvent degasser. The chromatography system was in a tandem arrangement downstream with a miniDAWN-TREOS light scattering instrument equipped with three detectors positioned at angles 44 º , 90° and 136° (Wyatt Technologies, Santa Barbara, CA) and Optilab T-rEX dRI detector (Wyatt Technologies, Santa Barbara, CA). Both instruments were reading at a wavelength of 658 nm. Scattered angles were normalized using bovine serum albumin (BSA) following the manufacturer's instructions. For a typical experiment, the SEC buffer was injected for a blank run to establish stable baselines for the light scattering and differential refractive index measurements. All purified protein samples were filtered with a 0.22 μm low protein binding Millex-GV filter just before each experimental run. Subsequently, 0.1-1.0 mg of purified and filtered protein samples were loaded onto the column and their elution profiles were monitored at A 280 . The extinction coefficient of purified protein was calculated from the amino acid sequence by using the ExPASy-ProtParam tool. The online measurement of the Rayleigh scattering intensity was processed to determine the weight-average molar mass and dispersity of protein material contained in top one-third of the major A 280 chromatographic peak. Data was acquired with the ASTRA 5.3.4 software (Wyatt Technologies, Santa Barbara, CA) and processed using Debye model as per the manufacturer's guidelines.

SDS-PAGE and Western Blot Analysis. For SDS-electrophoresis, 4-15% precast gradient TGX
Stain-Free TM gels (Bio-Rad Laboratories, Inc.) were used and visualized by stain-free enabled imager (Gel Doc TM EZ Imager, Bio-Rad). After blotting onto PVDF membrane (Bio-Rad), the proteins were visualized by probing with a murine horseradish peroxidase (HRP) conjugated antibody (New England BioLabs) against MBP at 1:50,000 dilution. Bands were detected by chemiluminescence using enhanced chemiluminescent (ECL) substrate (Thermo Scientific) and a digital imaging system (ImageQuant TM LAS 4000, GE Healthcare).