The Minimum M3-M4 Loop Length of Neurotransmitter-activated Pentameric Receptors Is Critical for the Structural Integrity of Cytoplasmic Portals*

Background: Residue 436 within 5-HT3A cytoplasmic portals determines single channel conductance (γ) and rectification. Results: M3-M4 loop truncation to 75 residues spared inward rectification; truncation to 70 abolished rectification and increased γ. Human Cys-loop neurotransmitter receptor M3-M4 loops all exceed 70 residues. Conclusion: M3-M4 lengths of >70 are critical. Significance: Cytoplasmic portals may be essential for normal function of Cys-loop neurotransmitter receptors. The 5-HT3A receptor homology model, based on the partial structure of the nicotinic acetylcholine receptor from Torpedo marmorata, reveals an asymmetric ion channel with five portals framed by adjacent helical amphipathic (HA) stretches within the 114-residue loop between the M3 and M4 membrane-spanning domains. The positive charge of Arg-436, located within the HA stretch, is a rate-limiting determinant of single channel conductance (γ). Further analysis reveals that positive charge and volume of residue 436 are determinants of 5-HT3A receptor inward rectification, exposing an additional role for portals. A structurally unresolved stretch of 85 residues constitutes the bulk of the M3-M4 loop, leaving a >45-Å gap in the model between M3 and the HA stretch. There are no additional structural data for this loop, which is vestigial in bacterial pentameric ligand-gated ion channels and was largely removed for crystallization of the Caenorhabditis elegans glutamate-activated pentameric ligand-gated ion channels. We created 5-HT3A subunit loop truncation mutants, in which sequences framing the putative portals were retained, to determine the minimum number of residues required to maintain their functional integrity. Truncation to between 90 and 75 amino acids produced 5-HT3A receptors with unaltered rectification. Truncation to 70 residues abolished rectification and increased γ. These findings reveal a critical M3-M4 loop length required for functions attributable to cytoplasmic portals. Examination of all 44 subunits of the human neurotransmitter-activated Cys-loop receptors reveals that, despite considerable variability in their sequences and lengths, all M3-M4 loops exceed 70 residues, suggesting a fundamental requirement for portal integrity.

The 5-HT 3 A receptor homology model, based on the partial structure of the nicotinic acetylcholine receptor from Torpedo marmorata, reveals an asymmetric ion channel with five portals framed by adjacent helical amphipathic (HA) stretches within the 114-residue loop between the M3 and M4 membrane-spanning domains. The positive charge of Arg-436, located within the HA stretch, is a rate-limiting determinant of single channel conductance (␥). Further analysis reveals that positive charge and volume of residue 436 are determinants of 5-HT 3 A receptor inward rectification, exposing an additional role for portals. A structurally unresolved stretch of 85 residues constitutes the bulk of the M3-M4 loop, leaving a >45-Å gap in the model between M3 and the HA stretch. There are no additional structural data for this loop, which is vestigial in bacterial pentameric ligand-gated ion channels and was largely removed for crystallization of the Caenorhabditis elegans glutamate-activated pentameric ligand-gated ion channels. We created 5-HT3A subunit loop truncation mutants, in which sequences framing the putative portals were retained, to determine the minimum number of residues required to maintain their functional integrity. Truncation to between 90 and 75 amino acids produced 5-HT 3 A receptors with unaltered rectification. Truncation to 70 residues abolished rectification and increased ␥. These findings reveal a critical M3-M4 loop length required for functions attributable to cytoplasmic portals. Examination of all 44 subunits of the human neurotransmitter-activated Cys-loop receptors reveals that, despite considerable variability in their sequences and lengths, all M3-M4 loops exceed 70 residues, suggesting a fundamental requirement for portal integrity.
5-Hydroxytryptamine type 3 (5-HT 3 ) 2 receptors are members of a family of pentameric ligand-gated ion channels (pLGICs), which includes the eukaryotic Cys-loop receptors and the bacterial pLGICs. Other members of the Cys-loop receptor family include the nicotinic acetylcholine (nACh), glycine, and ␥-aminobutyric acid type A (GABA A ) receptors. There are five genes that encode 5-HT3 subunits (A-E), which have features common to all Cys-loop subunits: an extracellular N-terminal domain containing the Cys-loop motif, four membrane-spanning domains (M1-M4), and a large cytoplasmic loop between M3 and M4. Homologous bacterial pLGICs lack both the N-terminal Cys-loop motif and the large M3-M4 loop (1,2). 5-HT3A subunits combine to form functional homomeric receptors with a sub-pS single channel conductance (␥), unique among all Cys-loop receptors examined to date. Unitary microscopic currents are too small to be directly resolved using conventional electrophysiological recording techniques (3). 5-HT 3 A receptors also exhibit inwardly rectifying macroscopic and single channel current-voltage relationships, slow desensitization kinetics, and substantial Ca 2ϩ permeability (3). 5-HT3B, 5-HT3C, 5-HT3D, and 5-HT3E subunits are unable to form functional homomers (4). However, incorporation of the 5-HT3B subunit into heteromeric 5-HT 3 AB receptors has profound effects on the biophysics of receptor function (5). Human 5-HT 3 AB receptors have a linear current-voltage relationship, fast desensitization, negligible Ca 2ϩ permeability, and single channel ␥ of 16 pS (5).
A chimeric strategy, replacing segments of 5-HT3A subunit amino acid sequences with equivalent residues in the 5-HT3B subunit, revealed that amino acid determinants of Ca 2ϩ permeability lie within the M2 domain (6). By contrast, a similar approach revealed that the molecular determinants for ␥ lie within the helical amphipathic (HA) stretch in the M3-M4 * This work was supported by grants from the National Science Foundation (0818741) and Tenovus Scotland. 1  loop. Three Arg residues at positions 432, 436, and 440, when replaced by equivalent residues (Gln, Asp, and Ala) in the 5-HT3B subunit, produced 5-HT 3 A(QDA) receptors with greatly increased ␥ (ϳ40-fold). The R436D substitution had the largest individual impact (7). Methanethiosulfonate modifications of the substituted Cys residues in mutant 5-HT 3 A(QCA) receptors revealed that the introduction of positive or negative charge decreased or increased ␥, respectively, whereas increasing molecular volume without charge reduced ␥ (8). Such data provide further evidence that 5-HT3A residue 436 is a ratelimiting determinant of conductance. In the 4-Å resolution partial structural model of the T. marmorata nACh receptor, HA stretches from adjacent subunits frame narrow intracellular portals through which ions must pass to enter or exit the cytoplasm (9). The five cytoplasmic portals enclose a vestibule that is contiguous with the transmembrane pore. The homology model of the 5-HT 3 A receptor, based on the structure of the nACh receptor, indicates that Arg-436 lies at the mouth of each portal (see Fig. 2A).
Although the nACh receptor model (9) provides insight into the structure of cytoplasmic portals framed by adjacent HA stretches, the bulk of the M3-M4 loop remains structurally unresolved, probably due to a disorganized structure (10), and its role in Cys-loop receptor remains unclear. X-ray structures of the homologous pLGICs of Gloeobacter violaceus (GLIC) (1) and Erwinia chrysanthemi (ELIC) (2), as well as the glutamateactivated Cl Ϫ channel (GluCl) from Caenorhabditis elegans (11), have shed no further light on the structure of the M3-M4 loop, which is vestigial in the bacterial channels and was largely removed from GluCl to enable crystallography. The M3-M4 loop is not required for functional expression of recombinant 5-HT 3 A receptors in Xenopus oocytes. However, its replacement by the heptapeptide M3-M4 loop of GLIC resulted in receptors with an increased single channel conductance, comparable with that of the 5-HT 3 A(QDA) receptor (11).
In this study, we demonstrate using mutagenesis that, in addition to its critical role in determining ␥, residue 436 governs 5-HT 3 A receptor inward rectification, a property that correlates with the volume of the amino acid. Therefore, the distinctive 5-HT 3 A receptor properties of ␥ and inward rectification can both be used to assay the rate-limiting nature of residues located within cytoplasmic portals. Systematic truncation of the 5-HT3A M3-M4 loop proximal to the HA stretch revealed that an M3-M4 loop length of Ͼ70 amino acids is required to maintain rate-limiting portals.

MATERIALS AND METHODS
Cell Culture and Transfection-Human embryonic kidney 293 (HEK-293) cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 100 g ml Ϫ1 penicillin, and 100 units ml Ϫ1 streptomycin at 37°C and 5% CO 2 . Cells were seeded at low density in 35-mm dishes for electrophysiological recording, and transfections were performed by calcium phosphate precipitation, using 1 g of total cDNA per dish, as described previously (6). cDNAs encoding wild-type and mutant 5-HT3 subunits were cloned into the pCDM8 mammalian expression vector using HindIII (5Ј) and XhoI (3Ј) sites. Unless stated otherwise, for heteromeric expression of 5-HT3 subunits, a 1:1 transfection ratio was used. cDNA encoding enhanced green fluorescence protein (in pEGFP vector, 0.1 g) was included to identify successfully transfected cells using fluorescence microscopy. Cells were washed with media 16 h after transfection and used after 48 -72 h. Tissue culture reagents were obtained from Invitrogen (Paisley, UK).
Mutagenesis of 5-HT3A Subunits-Point mutations and truncations of the M3-M4 loop were performed by overlap extension polymerase chain reaction (PCR) (12). Single point mutations and deletions were generated by chimeric PCR primers. PCR products were digested using HindIII and XhoI restriction endonucleases and ligated into pCDM8 vector. All mutagenesis reactions and ligations were verified using agarose gel electrophoresis, and constructs were sequenced to confirm fidelity prior to functional testing (Genetics Core Services, University of Dundee). All PCR and cloning reagents were from Fermentas (Thermo Fisher, Loughborough, UK).
Electrophysiology-The whole cell configuration of the patch clamp technique was used to record 5-HT-evoked currents from HEK-293 cells expressing wild-type (WT) or mutant 5-HT 3 receptors. Recording electrodes were fabricated from borosilicate glass capillaries, and when filled with intracellular solution, had resistances of 1-3 M⍀. The electrode solution contained (in mM): 140 CsCl, 2 MgCl 2 , 0.1 CaCl 2 , 1.1 EGTA, 10 HEPES (pH 7.4 with CsOH). The extracellular solution contained (in mM): 140 NaCl, 2.8 KCl, 2 MgCl 2 , 1 CaCl 2 , 10 HEPES, 10 glucose (pH 7.4 with NaOH). Unless otherwise stated, cells were voltage-clamped at an electrode potential of Ϫ60 mV. Macroscopic current-voltage (I-V) relationships were determined by adjusting the electrode potential from Ϫ60 mV to ϩ60 mV in 20-mV increments. To determine the linearity of I-V relationships, the ratio of peak currents evoked at ϩ60 mV to Ϫ60 mV (the rectification index) was used. A unitary rectification index indicates a linear I-V relationship. A unitary rectification index was used as a marker of the successful incorporation of the 5-HT3B subunit, which causes linearization of the I-V relationship (5). Unless otherwise stated, currents were evoked by rapid application of 100 M 5-HT, using the threepipe Perfusion Fast Step system (Warner Instruments), as described previously (6). The outside-out patch configuration was used to record single channel currents from excised patches from transfected cells as detailed previously (5). Electrodes were coated with Sylgard (Corning, NY) to minimize capacitance and noise. Currents were evoked by transient pressure application (Picospreizer II, General Valve Corp., Fairfield, NJ) of 5-HT (3 M). Single channel I-V relationships were determined by adjusting the electrode potential between Ϫ100 and Ϫ40 mV, in 20-mV increments.
All electrophysiological data were recorded using an Axopatch 200B amplifier. Data were low pass-filtered at 1 kHz for single channel data and 2 kHz for whole cell data, digitized at 20 kHz using a Digidata 1320A interface, and acquired using pClamp8 software (all from Molecular Devices).
Data Analysis-The peak amplitudes of 5-HT-evoked currents were measured using pClamp10 software. Desensitization kinetics were measured by fitting a double exponential function to the decaying phase of a 5-HT-evoked current. Sin-gle channel current amplitudes were determined from Gaussian fits of all points histograms. Single channel conductance (␥) values were expressed as slope conductance, estimated from linear fits to I-V relationships plotted from single channel data.
Structural Modeling-The homology model of the 5-HT 3 A receptor was created with Deepview Swiss-Pdb Viewer using the Torpedo marmorata 2BG9 structure (9), as described previously (12). Spartan Ј04 (Wavefunction Inc., Irvine, CA) was used to determine the volumes of amino acids substituted into the HA 436 position. Structures were energy-minimized in the extended chain conformation.
Statistics-Data are presented as mean Ϯ S.E. Correlations between amino acid charge or volume and the rectification index were analyzed using a two-tailed Pearson correlation test. Differences in rectification indices, or ␥ values, mediated by WT and mutant 5-HT 3 receptors were compared using oneway analysis of variance (ANOVA), with a post hoc Tukey's test. Individual comparisons between WT and a mutant data were made, where stated, using Student's t test. In all cases, p Ͻ 0.05 was considered statistically significant.

The HA Stretch Determines 5-HT 3 Receptor Rectification-
The ␥ of homomeric 5-HT 3 A receptors is Ͻ1 pS, and 5-HTevoked currents exhibit an inwardly rectifying I-V relationship (3). Incorporation of the 5-HT 3 B subunit into heteromeric 5-HT 3 AB receptors increases ␥ to 16 pS and abolishes rectification (5). The use of a chimeric strategy, in which regions of the 5-HT3B subunit were introduced into the 5-HT3A subunit, isolated determinants of ␥ within the HA stretch of the large cytoplasmic M3-M4 loop (6 -8). We adopted the same chimeric strategy here to identify the location of residues that determine rectification (Fig. 1). Consistent with previous work, WT 5-HT 3 A receptors displayed inward rectification and the incorporation of the 5-HT3B subunit increased the rectification index to near unity (5). Replacement of the entire HA stretch in the 5-HT3A subunit with that of the 5-HT3B subunit (chimera C4) conferred a linear IV relationship and a near unitary rectification index.
The Charge and Volume of the 0Ј Residue within the HA Stretch Influences 5-HT 3 Receptor Rectification-Our previous studies demonstrate that Arg-432, Arg-436, and Arg-440, at Ϫ4Ј, 0Ј, and 4Ј positions within the HA stretch, are responsible for the sub-pS single channel conductance of 5-HT 3 A receptors, with Arg-436 having the largest individual impact on ␥, giving rise to its designation as HA 0Ј (13). The homology model of the 5-HT 3 A receptor, based on the structure of the T. marmorata nACh receptor (9), indicates that these Arg residues lie near the cytoplasmic entrances of putative ion-conducting portals ( Fig. 2A). In this study, we investigated the role in rectification of Arg residues within the HA stretch of the 5-HT 3 A receptor. Residues were systematically substituted, individually and in combination, by their amino acid counterparts within the 5-HT3B subunit ( Fig. 2A). WT and mutant constructs were transfected into HEK-293 cells, and the amplitudes of currents, evoked by 5-HT (100 M) at ϩ60 mV, were expressed as a ratio of those evoked at Ϫ60 mV (I ϩ60mV/Ϫ60mV ), providing a rectification index (see "Materials and Methods"). The average rectification index data for WT and mutant receptors are shown in Fig. 2B. The 5-HT3A(R426S) construct, in which Arg at the HA Ϫ10Ј position was replaced by Ser (the HA Ϫ10Ј residue in 5-HT3B), had a rectification index similar to that of WT 5-HT 3 A receptors (Fig. 2B). Likewise 5-HT3A R432Q and R440A (HA Ϫ4Ј and HA 4Ј positions, respectively) substitutions were also without significant effect on the rectification index when introduced into the 5-HT3A subunit either individually or in combination. However, the 5-HT3A(R436D) mutation (at the HA 0Ј position) significantly increased the rectification index close to unity when incorporated into the 5-HT3A subunit alone or in combination with either R432Q or R440A. These data demonstrate that the HA 0Ј residue is the key determinant of rectification in 5-HT 3 receptors, the same HA stretch residue that plays a predominant role in determining ␥ (13).
The charge of the HA 0Ј residue is the principal determinant of ␥, with residue volume playing a more minor role through steric hindrance (8). We replaced Arg-436 by alternative amino acids with differing properties and examined the rectification The HA stretch is the determinant of 5-HT 3 receptor rectification. A, diagrams representing the 5-HT3A/B chimeras, which were made by substituting different regions of the 5-HT3A (white) subunit with that of the 5-HT3B (gray) subunit. Chimeras were expressed in HEK cells, and currents were recorded at potentials between Ϫ60 and ϩ60 mV in 20-mV increments. B, graph of the relationship between holding potential (V) in mV and current amplitude expressed relative to that of the current recorded at Ϫ60 mV. A near linear current-voltage relationship was exhibited by receptors that contained the HA stretch of the 5-HT3B subunit within the M3-M4 loop (5-HT3B and C3 and C4 constructs). As reported previously, the 5-HT3B subunit, C1, C2, and C3 required co-expression with the 5-HT3A subunit for function (5,7).
index of each of the mutant receptors, represented in Fig. 3 by the single letter abbreviation of the substituent amino acid. The change in charge of the 5-HT3A subunit HA 0Јresidue caused by the substitution of the Arg by each amino acid was plotted against the rectification index (Fig. 3A). Changing the charge at HA 0Ј produced a positive correlation with the rectification index (Pearson coefficient ϭ 0.61; p ϭ 0.006; Fig. 3A). Linear regression through the data points had an R 2 value of 0.37. There is considerable heterogeneity in the rectification index for nonpolar residues. We therefore repeated the analysis with only charged (positive and negative) and polar residues at HA 0Ј (Fig. 3B). Excluding the nonpolar residues from the analysis improved the correlation coefficient (Pearson coefficient ϭ 0.82; p ϭ 0.004), as well as the linear fit (R 2 ϭ 0.68). We also examined the relationship between the volume of the residue at HA 0Ј and the rectification index because volume also plays a role in determining ␥ (8). The volume of the residue at HA 0Ј correlated with the rectification index (Pearson coefficient ϭ Ϫ0.62; p ϭ 0.005; Fig. 3C). Linear regression produced an R 2 value of 0.38. Similar to the relationship between charge and the rectification index, we repeated the analysis, excluding nonpolar residues. This also improved the correlation (Pearson coefficient ϭ Ϫ0.72; p ϭ 0.02) and R 2 value (0.51; Fig. 3D). These data indicate that both the charge and the volume of the HA 0Ј residue influence rectification in 5-HT 3 The HA stretch contains the determinants of inward rectification in 5-HT 3 A receptors. A, ribbon diagram (left panel) based on the T. marmorata structure and alignment of human, mouse, and rat 5-HT3A and 5-HT3B subunit HA stretch residues (right panel). The three conserved arginine residues (432 at HA Ϫ4Ј, 436 at HA 0Ј, and 440 at HA 4Ј) in the human 5-HT3A subunit previously shown to affect single channel conductance are highlighted, along with the corresponding residues in the 5-HT3B subunit (13). These HA stretch Arg residues frame the cytoplasmic portals in a 5-HT 3 A receptor homology model (13) based on the T. marmorata nACh receptor structure (9). Also highlighted in the alignment is Arg-426 (at Ϫ10Ј). B, systematic replacement of arginines within the 5-HT3A subunit HA stretch by corresponding residues in the 5-HT3B subunit reveals that the HA stretch contains the determinants for rectification. Data are represented as the ratio of 5-HT-evoked currents at ϩ60 mV and Ϫ60 mV (I ϩ60mV/Ϫ60mV ) in WT 5-HT 3 A and 5-HT 3 AB (black bars) and mutant 5-HT 3 A receptors (open bars for single mutants and gray bars for double mutants). 5-HT 3 A receptors display inward rectification as revealed by the below unitary I ϩ60mV/Ϫ60mV . 5-HT 3 AB receptors on the other hand display a linear I-V relationship. Amino acid substitutions in the 426, 432, and 440 positions alone, or in combination, did not alter I ϩ60mV/Ϫ60mV . However, R436D produced a significant increase in I ϩ60mV/Ϫ60mV (p Ͻ 0.05, one-way ANOVA). R436D in combination with R432Q and/or R440A also produced receptors with a unitary I ϩ60mV/Ϫ60mV (*, p Ͻ 0.05, one-way ANOVA). Taken together, these data demonstrate that residues within the HA stretch govern I-V relationship properties of 5-HT 3 A receptors.

M3-M4 Loop Length Affects 5-HT 3 A Receptor Function
M3-M4 intracellular loops of any of the Cys-loop receptors (9). The structure of the M3-M4 loop between the end of M3 and the start of the HA stretch remains unresolved in the T. marmorata ACh receptor model (9). This omission is represented by a gap of Ͼ45 Å in the model. In the case of the 5-HT 3 A receptor, the gap between M3 and the HA stretch is bridged by 85 residues. We reasoned that systematic removal of amino acids within this region would reveal the minimum number required to connect the 5-HT3A subunit M3 with the HA stretch without compromising the functional integrity of the portals, assessed by the rectification index. The sequences of the structurally unresolved sections of the M3-M4 loop of the human 5-HT3A subunit and truncation constructs are shown in Fig. 4A. We initially truncated the M3-M4 loop close to M3. However, when expressed in HEK-293 cells and challenged with 5-HT (100 M), neither of these shortened constructs produced functional receptors (n ϭ 4 and 3 for L-10(1) and L-10(2), respectively), suggesting that residues in the region removed are essential for functional expression of 5-HT 3 A receptors. We therefore adopted an alternative strat-egy of truncation starting closer to the HA stretch (Fig. 4A). Truncations of 24 (n ϭ 7), 34 (n ϭ 7), 39 (n ϭ 14), and 44 (n ϭ 6) amino acids from the M3-M4 loop (termed: L-24, L-34, L-39, and L-44) produced functional homomeric constructs (Fig. 4B). The construct truncated by 55 amino acids (L-55) did not form functional homomeric receptors (n ϭ 3), but combined with WT 5-HT3A subunits to form heteromeric receptors with properties distinct from those of the WT 5-HT 3

4, D and E)
. Desensitization kinetics were analyzed by fitting exponential functions to the decaying phases of 5-HT-evoked currents up to the point of agonist removal. The sum of two exponentials was required to obtain the best fit to the data (Fig.  4D), and weighted ( w ) values (Fig. 4E) were calculated for 5-HT-evoked current desensitization as described under "Materials and Methods." The fast ( fast ) and slow ( slow ) time constants for desensitization and their relative contributions (Fig. 4D) did not differ (one-way ANOVA), leading to similar W values for 5-HT 3 A L-24 ( w ϭ 204 Ϯ 48 ms, n ϭ 7), L-34 ( w ϭ 187 Ϯ 58 ms, n ϭ 7), L-44 ( w ϭ 156 Ϯ 12 ms, n ϭ 6), and WT ( w ϭ 229 Ϯ 19 ms, n ϭ 32) receptors (Fig. 4E). By contrast, 5-HT 3 A L-39 receptors exhibited significantly slower fast and slow values than did the other homomeric receptor constructs (p Ͻ 0.05; Fig. 4D). Furthermore, the L-39 truncation caused an increase in the relative proportion of the slow component of desensitization. Together these changes led to a substantial increase in the W of desensitization for currents mediated by 5-HT 3 A L-39 receptors (648 Ϯ 57 ms, n ϭ 14) when compared with the other homomeric receptors (Fig. 4E). Receptors in HEK-293 cells expressing both the WT 5-HT 3 A subunit and the 5-HT 3 A L-55 construct desensitized significantly slower ( w ϭ 495 Ϯ 52 ms, n ϭ 7, p Ͻ 0.05 Student's t test) than did cells transfected with WT 5-HT 3 A receptors alone (Fig. 4C).
Desensitization kinetics could be influenced by the phosphorylation status of the M3-M4 loop. Indeed, phosphorylation has been implicated in the slow desensitization of currents mediated by 5-HT 3 A receptors (14). There are Ser and Thr residues in the sections of the M3-M4 loop removed to produce the truncation constructs (Fig. 5A). A loss of potential sites for phosphorylation could account for slow desensitization kinetics of currents mediated by 5-HT 3 A L-39 receptors. The L-39 truncation removes an additional Ser and a Thr residue when compared with L-34. The L-44 truncation removes another Ser and a Thr residue. There may be combinatorial effects on desensitization kinetics of different phosphorylation sites. However, according to a phosphorylation consensus site recognition algorithm (15), several of the residues in question are unlikely to be phosphorylated. The putative consensus sites are highlighted in Fig. 5A. Interestingly, this approach reveals that removal of 39 amino acids in the L-39 construct resulted in the artificial de novo introduction of a phosphorylation consensus site into the L-39 construct at Thr-369, which is not found in the L-24, L-34, or L-44 constructs.  JULY 26, 2013 • VOLUME 288 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 21563
We examined the possible contribution of phosphorylation at Thr-369 in the 5-HT 3 A L-39 construct by making a phosphonull T369A substitution. A representative recording of a 5-HT (100 M)-evoked current mediated by the 5-HT 3 A L-39(T369A) construct is shown superimposed on a 5-HT 3 A L-39-mediated current (Fig. 5B, left panel). The desensitization ( w ϭ 428 Ϯ 49 ms; n ϭ 11) of L-39(T369A) was significantly (p Ͻ 0.05, Student's t test) faster than that of L-39 (Fig. 5C), consistent with the idea that phosphorylation at Thr-369 in the L-39 construct contributes to the slower desensitization rate when compared with the L-24, L-34, and L-44 constructs, which all lack this putative phosphorylation consensus site. We tested this hypothesis further by introducing a phosphomimetic T369E substitution into the L-34 truncation construct to produce L-34(T369E). The desensitization kinetics of 5-HT (100 M)-evoked currents ( w ϭ 640 Ϯ 48 ms, n ϭ 7) mediated by 5-HT 3 A L-34(T369E) receptors (Fig. 5B, right panel) were significantly slower (p Ͻ 0.05, Student's t test) than those of the L-34 construct that lacked the phosphomimetic substitution ( w ϭ 187 Ϯ 58 ms, n ϭ 7, Fig. 5C). The rate of desensitization of currents mediated by L-34(T369E) was similar to those mediated by L-39. Collectively, these data are suggestive that phosphorylation of Thr-369 in L-39 is responsible for the slower desensitization kinetics of this construct when compared with the L-34 truncation construct.
Systematic M3-M4 Loop Truncation Disrupts Rectification and ␥-Having determined that the 5-HT3A truncation constructs, L-24, L-34, L-39, and L-44, produce functional homomeric 5-HT 3 receptors and that expression of L-55 with the wild-type 5-HT3A subunit produces functional receptors, we next explored whether inward rectification was affected by shortening the M3-M4 loop. Perhaps truncation beyond a certain critical number of amino acids would abolish rectification by disrupting the orientation of HA stretches that frame cytoplasmic portals (Fig. 6A). We measured the rectification indexes for L-24, L-34, L-39, and L-44 expressed as homomeric receptors and L-55 expressed with the WT 5-HT3A subunit. 5-HT (100 M) was applied, and currents were recorded at Ϫ60 and ϩ60 mV to determine the rectification index for each construct (Fig. 6B). L-24, L-34, and L-39 constructs mediated 5-HT-evoked currents that were consistently smaller in peak amplitude when recorded at ϩ60 mV than when recorded at Ϫ60 mV. Thus, these truncation constructs, like the WT 5-HT 3 A receptor, exhibited inward rectification. However, the L-44 construct mediated 5-HT-evoked currents with similar amplitudes at ϩ60 and Ϫ60 mV, indicating that there is a linear I-V relationship. The L-55 truncation mutant, when expressed together with the 5-HT3A subunit, produced receptors that also mediated 5-HT-evoked currents with similar peak amplitudes at ϩ60 and Ϫ60 mV, again indicative of a linear I-V relationship (Fig. 6B). WT 5-HT 3  The effects of truncation suggest that a 5-HT3A subunit M3-M4 loop length of 70 residues is insufficient to support inward rectification (Fig. 6B), a property conferred by residues located within putative cytoplasmic portals (Fig. 2). The loss of inward rectification in receptors with loop lengths of Յ70 res- idues is consistent with the idea that portals have become disrupted and no longer limit ion conduction (Fig. 6A). We performed recordings of 5-HT-evoked currents mediated by 5-HT 3 A L-44 receptors in excised outside-out patches to determine whether the loss of 44 residues also affected ␥, a property of 5-HT 3 A receptors that is governed by residues that lie within the putative cytoplasmic portals (7,13). 5-HT-evoked currents recorded from outside-out patches excised from HEK-293 cells expressing WT 5-HT 3 A receptors resemble miniature whole cell currents with no discernible unitary events (3,5,7). By contrast, 5-HT-activated single channel events were observed in recordings from patches containing 5-HT 3 A L-44 receptors (Fig. 6C). The chord conductance of single channels mediated by 5-HT 3 A L-44 receptors was 16.3 Ϯ 1.3 pS (n ϭ 7).

M3-M4 Loop Length of All Neurotransmitter-activated Human pLGICs Exceeds 70
Residues-The M3-M4 loops of human pLGICs are highly heterogeneous in both amino acid composition and number. Having established that Ͼ70 residues are required to maintain functional characteristics consistent with integrity of cytoplasmic portals of 5-HT 3 A recep-tors, we examined how this relates to the M3-M4 loop lengths of the other human Cys-loop receptor subunits. We determined the length of the large cytoplasmic loop of the 5-HT3A subunit to be 114 residues from the homology model of the 5-HT 3 A receptor based on the structure of the T. marmorata nACh receptor (9,13). The M3-M4 loop lengths of the other human Cys-loop receptors were then established by alignment with the 5-HT3A amino acid sequence. With the exception of the Zn 2ϩ -activated ion channel (ZAC), which has an atypically short intracellular loop (16) estimated to be 43 amino acids, all the other Cys-loop receptor subunits (i.e. subunits of the neurotransmitter-activated human pLGICs) have loop lengths of Ն73 residues (Fig. 7). The M3-M4 loops of the neurotransmitter-activated cation channels range from 80 to 262 amino acids in length (Fig. 7A). The large cytoplasmic loops of the human pentameric anion channels range from 73 to 247 (Fig. 7B).

DISCUSSION
Using a chimeric strategy followed by site-directed mutagenesis, we determined that Arg-436 (in the HA 0Ј position) is  JULY 26, 2013 • VOLUME 288 • NUMBER 30 critical for the characteristic inward rectification of 5-HT 3 A receptor-mediated macroscopic currents. This is the residue we previously demonstrated to be the principal determinant of ␥ for unitary events mediated by 5-HT 3 A receptors, a property that is primarily dictated by the charge of the HA 0Ј residue (8). In this study, substitution of numerous alternative amino acids into the HA 0Ј position revealed that both the volume of the residue and its charge correlate with inward rectification. These findings add to prior evidence for the existence of cytoplasmic portals that form rate-limiting barriers to ion conduction in some Cys-loop receptors (7)(8)(9)13).

M3-M4 Loop Length Affects 5-HT 3 A Receptor Function
Previous studies demonstrate that rectification of 5-HT 3 receptor-mediated macroscopic currents occurs at the level of ␥ (17)(18)(19). Furthermore, residues in the HA stretch of the 5-HT3A subunit impose a rate-limiting barrier to both inward and outward currents (7,8,13,18), and the present study demonstrates that macroscopic current rectification is governed by the same residues. Taken together these findings imply that 5-HT 3 A receptor portals impose a greater impediment to outward current than to inward current. Membrane potential is unlikely to directly influence portals, which lie outside the electrical field. It is also unlikely that charged cytosolic molecules interact with portals in a voltage-dependent manner to initiate inward rectification because this property persists in excised outside-out membrane patches (20). Instead, it is possible that rectification is caused by the asymmetric nature of the ion conduction pathway. The large extracellular vestibule contains rings of negatively charged residues, which augment single channel conductance in 5-HT 3 A(QDA) mutant receptors in which portals are no longer rate-limiting (21). The negative charges likely assist the diffusion of cations into the vestibule and ion channel pore, where they will be subjected to the driving force imposed by membrane potential. By contrast, the arginine residues of the 5-HT 3 A receptor portals may restrict diffusion of ions into the intracellular vestibule as well as constraining the rate at which ions can traverse in either direction through the channel. Collectively, the data suggest that the asymmetric pathway formed by cytoplasmic portals exerts a greater impediment to ion conduction in the outward direction, giving rise to inward rectification. This is, to our knowledge, the first example of intrinsic inward rectification in a cationic neurotransmitter-gated ion channel. Inward rectification of cation channels, including some nACh receptors, is often caused by voltage-dependent blockade by intracellular polyamines (22). This study demonstrates that the different amino acid sequences of 5-HT3A and 5-HT3B subunits, and in particular the presence of an HA 0Ј Arg or Asp in the former and latter, account for the inward rectification and linear I-V relationships that are characteristic of recombinant human 5-HT 3 A and 5-HT 3 AB receptors, respectively. However, the subunit composition of 5-HT 3 receptors in vivo remains largely unknown, particularly in humans. Even for rodents, there is little definitive information available regarding the roles of homomeric and heteromeric 5-HT 3 receptors in central and peripheral neurons (reviewed in 23). There are conflicting reports as to the prevalence of the 5-HT3B subunit in the rodent brain. The situation in rodent neuroblastoma cells is simpler. The functional properties of their 5-HT 3 receptors (strong inward rectification and sub-pS single channel conductance) are those of homomeric 5-HT 3 A receptors (20). The introduction of the 5-HT3B subunit changes the properties of neuroblastoma 5-HT 3 receptors consistent with the idea that these "native" receptors are indeed 5-HT3A homomers (24).
There is no structural information for the M3-M4 loop in Cys-loop receptors lying outside the HA stretch. The recently solved pLGIC structural models shed no light on the structure of the large intracellular loop of Cys-loop receptors. The GLIC and ELIC bacterial pLGICs have vestigial M3-M4 loops (8 and 11 amino acids, respectively) and add no additional insights (1,2). Of necessity, the M3-M4 loop of the C. elegans GluCl Ϫ receptor ␣1 subunit was largely removed to aid protein crystallization (11). The solitary pLGIC model that provides any structural data for the M3-M4 loop of Cys-loop receptors is the cryoelectron microscopy model of the T. marmorata nACh receptor, which resolves cytoplasmic portals formed by adjacent HA stretches, but lacks resolution for the larger section of the loop between M3 and the start of the HA stretch (9). The Ͼ45-Å gap is occupied by 85 residues in the case of the human 5-HT3A subunit, which has a predicted total M3-M4 loop length of 114 residues. The large cytoplasmic loops of human Cys-loop receptor subunits exhibit diversity in length. All the subunits of the neurotransmitter-activated human Cys-loop receptors have a loop length Ն73 residues. It is interesting to speculate why there is not a normal distribution of loop lengths from the minimal lengths of the bacterial pLGICs up to the considerable length (262 amino acids) of the ␣4 nACh subunit M3-M4 loop. Our data suggest that the minimum loop length is constrained by the requirement to connect the M3 to the HA stretch without disrupting portal-associated function.
We systematically truncated the unresolved stretch of the 5-HT3A subunit to reveal the minimum number of residues required to connect the M3 with the HA stretch without compromising the functional integrity of the cytoplasmic portals. Our initial attempts to truncate at the M3 end of the loop revealed a series of residues that were essential for the expression of functional 5-HT 3 A receptors in HEK-293 cells. Interestingly, the rules governing 5-HT 3 A receptor expression in HEK-293 cells and Xenopus oocytes appear to differ because replacement of the entire intracellular loop by a series of Ala residues had no effect on peak 5-HT-evoked current amplitude measured in the latter (25). Alternatively, the truncations may cause the remainder of the loop to adopt a conformation that prevents functional expression. By contrast to our initial approach, which removed residues near M3, truncations starting close to the HA stretch and progressing toward M3 were well tolerated and did not significantly impact on the density of 5-HT-evoked currents. The M3-M4 loops of Cys-loop receptors have been implicated in trafficking, either through intrinsic mechanisms or by interacting with other cytosolic proteins and, in agreement with our findings, mutagenesis or deletion can have profound effects on expression (26,27). It is possible that a series of 5-HT3A cytoplasmic residues adjacent to the M3 interacts with trafficking proteins. The PVPXWXR amino acid motif located within this critical region of the 5-HT3A subunit is conserved in nACh ␣9 and ␣10 subunits, potentially implicating its involvement in the trafficking of other Cys-loop receptors.
Perturbations of individual amino acids within the M3-M4 loop influence the kinetics of currents mediated by some Cysloop receptors, whereas other receptors appear unaltered by removal of the entire loop (28,29). Mutations associated with myasthenic syndrome that cause insertions, or substitutions, in the large cytoplasmic loops of the human muscle nACh receptor ⑀ subunit alter gating kinetics (29). Furthermore, differences in the kinetics of nACh receptors containing ␥ or ⑀ subunits can be accounted for by specific amino acids that differ within their M3-M4 loops. Additionally, phosphorylation of a specific residue within the M3-M4 loop of the GABA A receptor ␤1 subunit influences GABA-evoked current desensitization (30). There is also evidence that phosphorylation regulates 5-HT 3 receptor function. Casein kinase II modulates the kinetics of 5-HT 3 receptor-mediated currents recorded from NG108-15 neuroblastoma cells (14). The 5-HT3A L-39 truncation construct formed receptors that mediated 5-HT-evoked currents that desensitized slower than homomeric receptors formed by the other truncation constructs and by WT 5-HT 3 A receptors. This could be caused by the truncation introducing a phosphorylation consensus sequence into 5-HT3A L-39, not found in the other constructs, enabling phosphorylation of Thr-369. Con-sistent with this idea, substitution of Thr-369 by Ala accelerated the rate of 5-HT 3 A L-39 (T369A) receptor desensitization to rates similar to those of WT receptors. Furthermore, a phosphomimetic mutation in 5-HT 3 A L-34 (T369E) caused desensitization kinetics of the L-34 truncation construct to slow to a level similar to that of the L-39 construct.
Our findings with mutant constructs suggest that phosphorylation of the M3-M4 loop can produce changes in 5-HT 3 A receptor kinetics. It remains to be determined whether there is a physiological role for 5-HT 3 A receptor phosphorylation. Our recent unpublished observations with the casein kinase II inhibitor 4,5,6,7-tetrabromobenzotriazole support previous data implicating the kinase in slowing 5-HT 3 A receptor desensitization (14). 3 4,5,6,7-Tetrabromobenzotriazole increased the rate of desensitization of the WT and M3-M4 loop truncation constructs, suggesting that the site of action of casein kinase II may lie elsewhere, possibly in the M1-M2 loop where there is a casein kinase II consensus sequence. 3 We will explore this possibility in future studies.
Having determined that inward rectification requires the key HA 0Ј Arg located at the mouth of the putative portals, we used rectification as an assay of portal function. Removal of up to 39 residues from the loop had no effect on inward rectification, suggesting that 46 residues are sufficient to bridge the gap between the end of M3 and the start of the HA stretch. By contrast, removal of five additional residues (L-44) abolished inward rectification, suggesting that 41 residues are insufficient to bridge the gap without disrupting portal function. In keeping with this interpretation, 5-HT 3 A L-44 receptors exhibited clearly resolvable unitary events with 16 pS-␥ substantially greater than the Ͻ1 pS-␥ of wild-type 5-HT 3 A receptors (3,5).
It is unlikely that the absence of the five specific (SQATK) residues accounts for the loss of portal function when the loop is truncated by more than 39 residues (i.e. L-44 or L-55). The SQATK motif is not found in any of the other human pLGIC subunits. More importantly, it is also not conserved in rat or mouse 5-HT3A subunit variants (in which the equivalent sequence is FQANK) despite a high level of overall 5-HT3A amino acid sequence identity in human, mouse, and rat, even in the M3-M4 loop. Like the human 5-HT 3 A receptor, rodent 5-HT 3 A receptors exhibit strong inward rectification and a single channel conductance that is below resolution in outside-out patch recordings (17). Therefore, it is unlikely that the SQATK sequence is required for the functional integrity of 5-HT 3 A portals. 5-HT 3 A L-39 receptors, which did not have disrupted portal function, have a total loop length of 75 residues, shorter than all the M3-M4 loops of cation-permeable Cys-loop neurotransmitter receptors and only two residues longer than the shortest anionic receptor loop (that of the GABA A receptor ␥1 subunit). Therefore, a loop length of 75 residues, which supports portal integrity, is within the range of lengths tolerated by neurotransmitter-activated human pLGIC subunits. By contrast, the loop length of 70 residues, which does not support portal integrity, is below the length of all the neurotransmitter-activated pLGIC subunit M3-M4 loops. Taken together these findings suggest that there is a selective advantage to the cytoplasmic loops being long enough to support the structural integrity of portals. Because there is little variability between the amino acid sequences of orthologous mammalian Cys-loop subunits, this observation can be extended beyond humans.
It is interesting that pLGICs of bacteria have vestigial M3-M4 loops, whereas neurotransmitter-activated pLGICs have M3-M4 loop lengths that all exceed 72 amino acids. It is possible that the cytoplasmic portal structure has evolved to enable interactions with other intracellular proteins, enabling modulation of trafficking and function in the complex environment of a nervous system. Additionally, portals of limited dimensions may exclude molecules within the cytoplasm from entering the intracellular aspect of the channel pore, potentially causing channel blockade. Photo-affinity labeling experiments using [ 3 H]azietomidate suggest that cytoplasmic portals of nACh receptors also serve as drug binding sites (31). As evidence accumulates for the importance of the M3-M4 loop, the value of resolving its structure at the atomic level becomes increasingly clear.