Packing of the Extracellular Domain Hydrophobic Core Has Evolved to Facilitate Pentameric Ligand-gated Ion Channel Function

Protein function depends on conformational flexibility and folding stability. Loose packing of hydrophobic cores is not infrequent in proteins, as the enhanced flexibility likely contributes to their biological function. Here, using experimental and computational approaches, we show that eukaryotic pentameric ligand-gated ion channels are characterized by loose packing of their extracellular domain β-sandwich cores, and that loose packing contributes to their ability to rapidly switch from closed to open channel states in the presence of ligand. Functional analyses of GABAA receptors show that increasing the β-core packing disrupted GABA-mediated currents, with impaired GABA efficacy and slowed GABA current activation and desensitization. We propose that loose packing of the hydrophobic β-core developed as an evolutionary strategy aimed to facilitate the allosteric mechanisms of eukaryotic pentameric ligand-gated ion channels.

Proteins are conformationally dynamic and require thermodynamic stability and structural flexibility to function (1)(2)(3)(4)(5). Stability and flexibility are interdependent, with higher stability increasing structural rigidity, which in many cases reduces protein activity (6). Proteins are only marginally stable, with ⌬G folding values mostly between Ϫ3 and Ϫ15 kcal/ mol (7). Trade-offs between thermodynamic stability and conformational flexibility are observed during protein evolution, with novel or specialized functions often acquired at the expense of stability (1,2,8,9). New-function mutations, however, rarely involve active site residues (10, 11) but occur with high frequencies in hydrophobic cores (12). Core mutations, although, are highly demanding in terms of stability, because a well packed hydrophobic core is the major factor that drives protein folding and maintains a stably folded structure through entropic ("hydrophobic effect") and enthalpic (van der Waals interactions) forces (13)(14)(15)(16).
Protein cores are normally packed with amino acid residues with bulky hydrophobic side chains that efficiently fill the space (17,18). Core mutations tend to loosen the packing and create "defects" (empty or water-filled cavities) that decrease thermodynamic stability up to several kcal/moles (3, 19 -21). Numerous proteins (e.g. p53 (22), serpins (23), and hemagglutinin (24,25)) have evolved localized core packing defects that enhance their biological activity at the cost of their stability.
Recently, the crystal structure of the extracellular domain of the nicotinic acetylcholine receptor (nAChR) 3 ␣1 subunit revealed a hydrophilic water-filled cavity buried in the hydrophobic core of the protein and it was hypothesized that loose core packing is important for nAChR activity (26). nAChRs are members of the pentameric ligand-gated ion channel (pLGIC) superfamily, which includes ␥-aminobutyric acid (GABA A Rs), serotonin (5-HT 3 Rs), and glycine (GlyRs) receptors. For these receptors, binding of neurotransmitter to the extracellular ligand-binding domain triggers rapid structural rearrangements (in the submillisecond to millisecond time frame) that ultimately lead to the opening of an intrinsic ionconducting channel. Signaling in the brain and neuromuscular junction depends on their activity. pLGICs are composed of five identical or homologous subunits, each consisting of an extracellular ligand-binding domain (ECD) made of ␤-strands forming a ␤-sandwich core, a transmembrane ion pore-forming domain made of four ␣-helices, and a large cytoplasmic loop (27,28). Recently, prokaryotic pLGICs were identified (29), and two were crystallized in presumed closed (Erwinia chrysanthemi ELIC) and open (Gloeobacter violaceus GLIC) channel states (30 -33).
In the crystal structure of the nAChR-␣1 ECD, two polar amino acids, Thr-52 and Ser-126, buried in the ␤-sandwich core coordinate an ordered water molecule (26) (Fig. 2). Mutation of these residues to bulky hydrophobic amino acids substantially reduced acetylcholine-activated channel current suggesting loose packing of the ␤-sandwich core is important for nAChR function (26). Molecular dynamics studies indicate that loose packing enhances the flexibility of nAChR-␣1 ECD (34). In crystal structures of related non-channel acetylcholine-binding proteins (AChBPs), which are structural homologues of pLGICs ECDs (27), tight packing is observed with bulky valine/leucine and phenylalanine at aligned positions. Here, we examined the generality and importance of loose ␤-core packing for pLGIC function. Using computational approaches, we defined the residue composition of the ␤-cores in pLGICs and AChBPs and determined in silico how changes in packing affect their stability. Inserting less bulky or hydrophilic residues in AChBPs and GLIC destabilized their cores. Conversely, mutating nAChR-␣1 Thr-52 and Ser-126 to bulkier hydrophobic residues stabilized its core. We show that loose packing of the ECD ␤-core is a distinctive feature of eukaryotic pLGICs. We also examined the functional consequences of altering the ECD ␤-core packing in related eukaryotic GABA A Rs. We show that increasing the ␤-core packing significantly decreased GABA A R activation. We argue that loose packing of eukaryotic pLGIC hydrophobic ␤-cores likely developed as an evolutionary strategy aimed to facilitate the allosteric transitions required for rapid switching between closed and open channel states in the presence of ligand.
Thermodynamic stability changes of in silico mutants were computed based on the PDB files above using FOLDX (37). Because structural studies have extensively demonstrated that the pLGICs ECD or AChBP in monomeric form constitutes an independent folding unit (26,27,38), monomeric nAChR-␣1 ECD and monomers of GLIC ECD and AChBPs were used. Briefly, each structure was first optimized using the repair function of FOLDX. Single amino acid mutations were then introduced with the repair position scan function. Finally, the energies (⌬G folding ) of the non-mutated and mutated structures were computed with the energy calculation function, and for each mutation the resulting ⌬⌬G folding ϭ ⌬G(mutant) Ϫ ⌬G(non-mutant) was obtained. The reported ⌬⌬G folding values were then corrected using the normalization function ⌬⌬G exp ϭ (⌬⌬G FoldX ϩ 0.078)/1.14 as described (7).
Two-electrode Voltage Clamp-Oocytes were perfused continuously with ND96 at a flow rate of 5 ml/min, while being held under two-electrode voltage clamp at Ϫ80 mV in a bath volume of 200 l. Borosilicate glass electrodes (Warner Instruments) used for recordings were filled with 3 M KCl and had resistances of 0.4 to 1.0 M⍀. Electrophysiological data were collected using Oocyte Clamp OC-725C (Warner Instruments) interfaced to a computer with an ITC-16 A/D device (Instrutech) and were recorded using the Whole Cell Program, version 4.0.9 (kindly provided by J. Dempster, University of Strathclyde, Glasgow, UK). Stock solutions of GABA (Sigma), pentobarbital (PB) (Sigma), and piperidine-4sulfonic acid (P4S) (Sigma) were prepared fresh daily in ND96.
Concentration-Response Analysis-GABA, PB, and P4S concentration responses were scaled to a low, non-desensitizing concentration of agonist (EC 2 to EC 10 ) applied just before the test concentration to correct for any slow drift in current responsiveness over the course of the experiment. Currents elicited by each test concentration were normalized to the corresponding low concentration current before curve fitting. At high micromolar concentrations and above, PB blocks GABA A R. Relief of channel block upon drug wash yields a characteristic tail current. For PB concentration-response curves, currents measured at high micromolar concentrations and above included tail current measurements. For currents recorded on the same oocyte with saturating concentrations of GABA and PB or P4S, extensive washes (6 -10 min) were performed before each application of the different drugs. GraphPad Prism 4 software was used for data analysis and fitting. Concentration-response data were fit to the equation: where I is the peak response to a given concentration of agonist, I max is the maximum amplitude of current, [A] is the agonist concentration, and n is the Hill coefficient.
Statistical Analysis-Log(EC 50 ) values for GABA and PB concentration responses, max current ratios, and GABA-elicited current activation and desensitization rates were analyzed using one-way analysis of variance, followed by a post hoc Dunnett's test to determine the level of significance between WT and mutant receptors. Natural logarithm (ln) transformed values of WT and mutant EC 50
We used FOLDX (37) to analyze and compare in silico the folding stability of the ECD ␤-core of nAChR-␣1, GLIC, and AChBP. In particular, we aimed to determine how the loose ␤-core packing in nAChR affects its thermodynamic stability compared with AChBPs and GLIC. FOLDX estimates the energetic contribution of amino acids to the stability of a protein based on a computer algorithm that introduces a mutation at FIGURE 1. Sequence alignment of AChBPs and representative pLGIC ECDs. The core residues, in crystal structures, with accessible surface area (ASA) Յ5 and Յ25% are highlighted in red and yellow, respectively. Based on sequence alignments, putative core residues in other pLGICs are highlighted in turquoise. Black downward arrows mark the positions of core residues aligned with nAChR-␣1 Thr-52 (site 1) and Ser-126 (site 2). Amino acid numbering refers to the mouse nAChR-␣1 sequence. Secondary structure elements are indicated below the sequences and are labeled in Greek letters based on the crystal structure of nAChR-␣1 ECD (26). ␤-Strands are blue (inner sheet) or green (outer sheet) arrows, helices are rectangles (the N-terminal ␣-helix is labeled in uppercase, the 3 10 -helices in lowercase), and connecting loops are black lines. a given site and calculates the ⌬⌬G folding of mutant relative to wild-type. Namely, each core-buried residue in nAChR-␣1 (PDB code 2QC1), GLIC (PDB 3EAM), and AChBP (L. stagnalis Ls-AChBP, PDB 1UV6, and A. californica Ac-AChBP, PDB 2BYN) was individually mutated to alanine and the effects on ⌬⌬G folding calculated (supplemental Table S2). Because the mutations to alanine (the residue with the smallest hydrophobic side chain of all natural amino acids) involve removal of hydrophobic side chains, no steric clash is expected and major protein reorganization is unlikely to occur (19,40). Therefore, the calculated ⌬⌬G folding energy differences can be considered approximately the contribution of the wild-type residue to the folding stability of the protein.
Hence, large positive changes in ⌬⌬G folding would indicate that introduction of the alanine mutations were destabilizing at specific sites and the wild-type residues are highly stabilizing to the protein fold.
Mutating the conserved bulky, hydrophobic ␤-core residues of each of the proteins into alanines was highly destabilizing, with ⌬⌬G folding values Ͼ ϩ2 kcal/mol and in most cases exceeding ϩ3 kcal/mol (supplemental Table S2). This observation is consistent with the idea that introducing nonbulky residues into the ␤-core loosens packing and decreases stability (19, 20, 40 -42). Moreover, the data indicate that residues contributing to ␤-core packing stability are remarkably similar in nAChR, AChBPs, and GLIC. The major differences between nAChR-␣1, AChBPs, and GLIC in the effects of the alanine mutations were observed at sites 1 and 2 (supplemen-tal Table S2). In nAChR-␣1, introducing an alanine at site 1 (T52A) had little effect (ϩ0.68 kcal/mol) and the alanine at site 2 (S126A) was even stabilizing (Ϫ0.95 kcal/mol), whereas the alanines at sites 1 and 2 in AChBPs and GLIC were highly destabilizing (⌬⌬G folding from ϩ2 kcal/mol to ϩ4.5 kcal/mol). This indicates that loose packing at sites 1 and 2 in nAChR-␣1 results in an intrinsically less stable fold compared with the high fold stability conferred to GLIC and AChBP by bulky hydrophobic residues at the corresponding sites.
We then used FOLDX to explore how changing the packing at sites 1 and 2 would affect folding stability. In each structure, we mutated sites 1 and 2 to aligned residues found in other pLGICs and AChBPs, and computed the ⌬⌬G folding (Fig. 3). Inserting less bulky or hydrophilic residues in the AChBP or GLIC destabilized the cores (most ⌬⌬G folding Ն 3 kcal/mol). Conversely, mutating nAChR-␣1 Thr-52 and Ser-126 to bulkier hydrophobic residues stabilized the core, with ⌬⌬G folding values from Ϫ0.95 to Ϫ2.70 kcal/mol. Overall, the FOLDX analysis suggests that the contribution of core-forming amino acids to the folding stability of pLGIC ECDs and AChBP is very similar, except for sites 1 and 2 where marked differences in the thermodynamic stability were estimated between nAChR-␣1, GLIC, and AChBP. Namely, tight core packing makes AChBPs very stable, whereas the loose packing in the nAChR ECD locally decreases its stability likely resulting in increased flexibility of the core (34).
Using structure based sequence alignment we scrutinized other eukaryotic pLGIC subunits for potential ECD ␤-core loose packing residues (Fig. 1). All nAChR subunits and 5-HT 3 R subunits contain a threonine at site 1 and a serine at site 2 indicating that loose packing at these positions may be a general feature for these receptors. Interestingly, the GABA A R subunits contain a hydrophobic leucine or isoleucine at site 1 (as observed in tightly packed cores) but a strictly conserved alanine at site 2 ( Fig. 1). Alanines (100 Å 3 smaller than a phenylalanine) buried in hydrophobic cores tend to create loose packing through cavities that decrease protein stability (19,40). Thus, for GABA A Rs, a potential loose packing defect is also present but a conserved alanine at site 2 may be responsible (Fig. 2, bottom). GlyR subunits appear to be the only eukaryotic pLGICs without potential packing defects at sites 1 or 2, as bulky hydrophobic residues (valine and leucine) are found at these positions. Whether GlyR subunits contain other unique packing defects will need to await direct structural analysis. In prokaryotic GLIC, packing at sites 1 and 2 is similar to AChBPs as both residues are hy-drophobic valines. However, in the GLIC ECD ␤-core, an alanine at position 108 (two residues away from site 2) may contribute to a loose packing defect. Finally, it should be noted that in limited subsets of eukaryotic pLGICs a handful of other residues at different positions from sites 1 or 2 could contribute to loose packing. These residues include threonine in strand ␤6Ј of GABA A R ␤-subunits, and alanine (GABA A R-␣2, GlyR-␣1) or threonine (GABA A R-␣1,3) in strand ␤10. However, differently from sites 1 and 2 these potentially coreloosening residues are not absolutely conserved in pLGIC families.
Altering the Packing at Site 1 and Site 2 Affects GABA A R Function-To further test the generality of the idea that loose packing of the ␤-core contributes to pLGIC function, we examined how altering GABA A R ␤-core packing at sites 1 and 2 affects its function. We mutated the alanine at site 2 in the GABA A R ␣1, ␤2, and ␥2 subunits to bulkier hydrophobic residues (valine, as in GLIC, or phenylalanine, as in AChBP). In In available crystal structures of AChBPs (top) and pLGIC ECDs (bottom), ␤-core residues at sites 1 and 2 were mutated and ⌬⌬G folding values were computed with FOLDX. In AChBPs or GLIC, substituting less bulky or hydrophilic residues at sites 1 and 2 resulted in significant thermodynamic destabilization, with values of ⌬⌬G Ն 3 kcal/mol particularly at site 2. In contrast, mutations in nAChR-␣1 that replaced the hydrophilic Thr-52 and Ser-126 with bulkier hydrophobic residues made its core more stable. Overall, the data suggest that the tight core packing makes AChBPs stable, whereas loose packing in the nAChR ECD decreases its folding stability. addition, we replaced site 1 leucine/isoleucine with smaller residues (valine or alanine). Double mutants with valine introduced at sites 1 and 2 (GLIC-like) were also engineered. Wildtype and mutant subunits were expressed in X. laevis oocytes to form ␣1␤2␥2 GABA A Rs and functionally characterized using two-electrode voltage clamp.
All of the mutant subunits assembled into receptors that responded to GABA. In general, the mutations shifted the GABA dose-response curves to the right. Representative current traces (top panel) and GABA dose-response curves (middle panels) are shown in Fig. 4. Replacing site 2 alanine in the ␣1 and ␤2 subunits with bulky residues (valine and phenylalanine) increased GABA EC 50 values 3-30-fold (Fig. 4, bottom, and Table 1). In the ␤2 subunit, as the size of the substituting residue increased, the effect on GABA EC 50 increased. The largest increase in GABA EC 50 values occurred when ␣1 and ␤2 site 2 phenylalanine mutant subunits were co-expressed (111-fold). Interestingly, co-expressing the site 2 mutant ␥2 subunit with mutant ␣1 and ␤2 subunits reduced the magnitude of the GABA EC 50 changes observed. At site 1, replacing the bulky leucine/isoleucine with smaller residues (alanine and valine) also significantly increased GABA EC 50 values, with the largest increase (21-fold) occurring when the alanine mutation was present in each of the subunits, ␣I61A␤L59A␥I74A. Overall, the data suggest that the combination of a bulky residue at site 1 (leucine/isoleucine) and a small alanine at site 2 is required for efficient GABA-induced channel activation. At site 2, replacing alanine with bulky side chains (like in AChBP) likely fills a cavity, which alters ␤-core packing, reduces flexibility, and impairs GABA A R activation. At site 1, reducing side chain volume also results in impaired GABA-mediated channel activation, likely due to destabilized ␤-core packing and/or local changes in the arrangement of nearby functionally important parts of the protein.
We also swapped the residues at sites 1 and 2 in the GABA A R ␤2 subunit. The swap ␣␤(L59A,A134F)␥ mutant did not rescue function, but rather showed cumulative effects of the individual mutations (low GABA-activated current amplitudes with I GABAmax range 0.48 -0.07 A, GABA EC 50 Ͼ100 mM). Our data are consistent with previous studies on a model system, the bacteriophage f1 gene V protein (16,18), which showed that the contribution of amino acids to the folding stability of a protein strongly depends on their position within the core, and that the effect of swapping buried residues is usually cumulative. Our results suggest that the effect of the putative cavity created by alanine strongly depends on its position within the core, being functionally favorable only at site 2. We postulate that only when alanine is at site 2 is there an appropriate balance between loss of thermodynamic stability and gain of molecular function (through structural flexibility).
Increasing Packing at Site 2 Reduces GABA Efficacy-The shifts in GABA EC 50 values observed following mutation of ␤-core sites 1 and 2 could be due to changes in GABA binding, GABA efficacy (channel gating), or both (45). The maximal GABA current amplitudes (I GABAmax ) for several of the mutant GABA A Rs, particularly the site 2 Phe mutants and double Val mutants, were smaller when compared with WT receptors (Table 1) suggesting changes in GABA efficacy and/or GABA A R expression. To test if the mutations affected GABA efficacy, we measured and compared currents induced by a saturating GABA concentration to those induced by a saturating concentration of PB in the same oocyte. PB is an allosteric modulator that binds to a different site than GABA (46) and at high concentrations can directly gate the channel with a similar conductance as GABA (47). For WT receptors, the ratio of currents elicited by saturating concentrations of PB and GABA (I PBmax /I GABAmax ) was 0.57 Ϯ 0.06 (Fig. 6, Table 2). Because at saturating concentrations of ligand, all the receptors are fully bound, a change in the I PBmax /I GABAmax ratio would provide evidence that the mutation altered channel open probability (gating). If the mutations only affected protein folding and expression levels, we should observe comparable effects on I PBmax and I GABAmax . In oocytes expressing ␣␤A134F␥, ␣A136F␤A134F␥, ␣A136V␤A134V␥, and ␣'VV'␤'VV'␥'VV' receptors, the currents elicited by saturating concentration of GABA were smaller than currents induced by PB indicating a reduction in GABA efficacy (Fig. 6, bottom  left).
The site 2 phenylalanine mutations and the double Val mutations also significantly increased PB EC 50 values compared with WT (2-6-fold, Table 2, Fig. 6, top) suggesting that increasing side chain volume at site 2 not only affects GABA EC 50 values but also adversely alters the ability of PB to activate the receptor. The differences in the magnitude of the mutations effects on GABA EC 50 and PB EC 50 values (e.g. for ␣A136F␤A134F␥ receptors, ⌬ in GABA EC 50 ϭ 111-fold, PB EC 50 5-fold, Tables 1 and 3) likely reflect the fact that GABA and PB bind to different sites in the receptor and trigger different activation pathways and movements (48). Some of the mutations with larger effects on GABA EC 50 may also be due to changes in GABA binding.
If the mutations adversely altered the gating equilibrium, then one would also expect the efficacy and maximum response to a partial agonist would decrease. Changes in efficacy are much more readily detected using partial agonists FEBRUARY 4, 2011 • VOLUME 286 • NUMBER 5  Table 1. At site 1, replacing the bulky hydrophobic (leucine/isoleucine) in GABA A R subunits with smaller residues (alanine and valine) right-shifted the GABA dose-response curves. At site 2, replacing the conserved alanine, which likely creates a cavity in the hydrophobic ␤-core, with larger residues (valine and phenylalanine) also significantly right-shifted the GABA dose-response curves.

Loose Packing of Cys-Loop Receptors ECD Hydrophobic Core
than full agonists (45). Thus, we looked at the effect of ␤-core mutations at site 2 on the relative efficacy of the partial agonist P4S macroscopic currents (49). We measured and com-pared currents induced by a saturating P4S concentration (10 mM for WT, 300 mM for mutants) to those induced by a saturating concentration of GABA (3 mM for WT, 300 mM for mutants) in the same oocyte for WT, ␣␤A134F␥, and ␣A136V␤A134V␥ receptors (Fig. 6, supplemental Table S4). For WT GABA A R, P4S elicited a maximum response only a third of that elicited by GABA, I P4Smax /I GABAmax ϭ 0.35 Ϯ 0.03. For ␣␤A134F␥ and ␣A136V␤A134V␥ receptors, the I P4Smax /I GABAmax ratio decreased to 0.19 Ϯ 0.03 and 0.24 Ϯ 0.05, respectively (Fig. 6, bottom right). The efficacy of P4S relative to PB was also measured and was significantly decreased for both mutants (0.12 Ϯ 0.03 for ␣␤A134F␥, and 0.18 Ϯ 0.06 for ␣A136V␤A134V␥) compared with WT (0.61 Ϯ 0.11) (Fig. 6, bottom right). Taken together, the data indicate that increasing the packing at site 2 reduces agonist efficacy.
Packing at Sites 1 and 2 Influences GABA A R Current Activation and Macroscopic Desensitization-We also measured the current 10 -90% apparent rise times at saturating GABA concentrations for mutant and WT receptors ( Fig. 7 and Table 3). Although measuring current onset is limited by the slow solution-exchange times when recording from oocytes (Ϸ300 ms), a slowed GABA current apparent rise time compared with WT would suggest a change in channel gating. GABA rise times for ␣I61A␤L59A␥, ␣I61A␤L59A␥I74A, ␣I61V␤L59V␥, ␣A136V␤A134V␥, ␣␤A134F␥, ␣A136F␤A134F␥, ␣'VV'␤'VV'␥ and ␣'VV'␤'VV'␥'VV' receptors were significantly slower than WT receptors (0.28 Ϯ 0.1 s for WT, Fig. 7, Table 3) with the double Val mutations having the largest effect (11-fold) suggesting that disrupting the packing at ␤-core sites 1 and 2 reduced channel opening. The mutations also decreased the extent of macroscopic current desensitization measured after a 20-s application of saturating GABA (Fig. 7, Table 3).  Table 1. The insets depict mutant cycles, changes in free energies due to mutations (⌬G), and the overall interaction energies (⌬⌬G).

DISCUSSION
Although packing defects (cavities) in hydrophobic cores account for as little as 1% of the volume of all globular proteins, they are often functionally relevant (50). Here, we provide evidence that a localized loose packing defect in the ECD ␤-cores of eukaryotic pLGICs (namely nAChRs, 5-HT 3 Rs, and GABA A Rs) is a key structural element for receptor function, which likely facilitates ligand-induced activation.
pLGICs are allosteric proteins that exist in at least three distinct, interconvertible states: resting (unliganded, closed channel), activated (liganded, open channel), and desensitized (liganded, closed channel). These receptors respond to neurotransmitter binding with allosteric transitions that switch the protein from a non-conducting closed conformation to an ion-conducting open conformation rapidly and with high probability. Our understanding of the protein movements underlying these conformational transitions is limited. Receptor activation is initiated by agonist binding in the ECD, and propagates ϳ60 Å to eventually gate the membrane-embed-ded ion-conducting pore. One model of activation suggests that neurotransmitter binding induces movements of the inner and outer ␤-sheets of the ECD relative to each other (51,52). This, in turn, changes the orientation and interactions of the flexible loops at the ECD-transmembrane ion pore-forming domain interface resulting in a repositioning of the M2 channel-lining helices, such that the channel pore opens. A second model suggests that a concerted quaternary twist of the ECD drives channel activation (32,53). Experiments using rate-equilibrium free energy relationships suggest that transduction of binding to gating occurs as a conformational wave from the transmitter binding site to the channel (54). Regardless of the model, any conformational change required for channel activation would be facilitated by enhanced flexibility of the ECD, as this would decrease the energy barrier necessary for the conformational switch to take place (3)(4)(5).
Flexibility of hydrophobic cores is usually the result of buried polar or non-bulky residues that loosen the core packing  Table 2. Top right, representative GABA and PB currents from oocytes expressing WT and ␣'VV'␤'VV'␥'VV' GABA A Rs. Bottom, currents elicited by sequential applications of saturating concentrations of GABA and PB (left) or P4S (right) measured from the same oocyte expressing either WT or mutant GABA A Rs. Maximum current ratios are plotted and represent mean Ϯ S.E. from Ͼ3 independent experiments. Black bars represent ratios significantly different from WT (p Ͻ 0.05).

TABLE 2 PB EC 50 values from wild-type (WT) and mutant GABA A Rs
Values are mean Ϯ S.E. from n experiments. Ranges in I PBmax and I GABAmax currents elicited from oocytes expressing WT or mutant GABA A Rs are indicated. I PBmax / I GABAmax ratios from the same oocyte are mean Ϯ S.E. from n experiments. Mutants indicated with 'VV' have a valine mutation at both sites 1 and 2.
at the expense of thermodynamic stability. Here, we identify a localized loose packing defect in the ECD ␤-cores of eukaryotic pLGICs that decreases the stability of the hydrophobic core. Moreover, we demonstrate that loose packing facilitates ligand-induced receptor activation supporting the idea that local ␤-core flexibility helps drive pLGIC activation.
Structural and computational studies show that pLGIC ECDs and AChBPs share the same fold, with the main chain traces of their ␤-sandwich cores nearly superimposable (26,(31)(32)(33). Our computational analyses demonstrate that pLGICs ECDs and AChBP ␤-cores are tightly packed with bulky hydrophobic amino acids at most positions (Figs. 1 and 2, supplemental Figs. S1 and S2), which confer thermodynamic stability (supplemental Tables S1 and S2). The ␤-core packing of the nAChR, 5-HT 3 R, and GABA A R, however, each contains a localized packing defect not observed in AChBPs or GLIC (Figs. 1 and 2). In nAChRs and 5-HT 3 Rs, two conserved polar residues corresponding to Thr-52 and Ser-126 in the nAChR ␣1 subunit (sites 1 and 2, respectively) are present that loosen the ␤-core packing (Figs. 1 and 2). In GABA A Rs, a strictly conserved alanine at site 2 creates a potential packing defect. These residues are unusual in hydrophobic cores as they decrease thermodynamic stability (4, 18 -20).
Using FOLDX, we determined that the ECD ␤-core of nAChR-␣1 is less stable at sites 1 and 2 compared with AChBP and GLIC (supplemental Tables S1 and S2). Consistent with our results, in the crystal structure of nAChR-␣1 ECD, the side chains of Thr-52 and Ser-126 are buried in the ␤-core and form a water-filled cavity (26). This packing defect is energetically costly because placing two polar side chains and a water molecule into a hydrophobic environment has high entropic penalty (ϳ20 kcal/mol), which is only partly compensated by the formation of water-protein hydrogen bonds (55,56). In AChBP and GLIC, the cores are conventionally packed at both sites with bulky hydrophobic residues (valine, leucine, and phenylalanine). As expected, introducing residues that potentially create packing defects (threonine, serine, and alanine) at these sites was highly destabilizing, with ⌬⌬G folding increasing 3-5 kcal/mol (Fig. 3).
Mutation of Thr-52 and Ser-126 in the nAChR ␣1 subunit to bulky hydrophobic amino acids substantially reduces acetylcholine-activated channel current (26). When we used FOLDX and mutated nAChR-␣1 Thr-52 and Ser-126 to bulkier hydrophobic residues to pack the core, it stabilized the core (Fig. 3). Previous work has shown that mutating Ser-126 to alanine or valine had little effect on gating equilibrium and is consistent with our results, as the smaller side chain of alanine and valine would likely still leave a cavity (57). Together, the data indicate that loose packing of the ␤-sandwich core is important for nAChR function. Because 5-HT 3 R subunits have the same loose packing defect as the nAChR-␣1 (threonine at site 1, and threonine or serine at site 2), we reason that it also plays a critical role in 5-HT 3 R channel activation.
We show that loose ␤-core packing is also important for GABA A R function. In GABA A R subunits, a hydrophobic leucine or isoleucine is at site 1 (as observed in tightly packed cores) but a strictly conserved alanine is at site 2, which may create loose packing. Leucine and isoleucine are canonical side chains for a hydrophobic core. In our FOLDX analysis, mutating any ␤-core leucine or isoleucine to alanine in pLGIC or AChBP structures caused destabilization, with ⌬G folding increments of 3-5 kcal/mol (supplemental Table S2). Thus, we expected that mutating leucine/isoleucine at site 1 in GABA A R to a smaller residue would decrease the stability of the ␤-core even further (which is constitutively less stable than AChBP because of the site 2 alanine) and exert an adverse effect on receptor function. As anticipated, replacing site 1 leucine/isoleucine with smaller residues (valine or alanine) significantly increased GABA EC 50 values 4 -20-fold ( Fig. 4, left, and Table 1). The effects on GABA EC 50 were inversely proportional to the size of the side chain substitution, and increased with co-expression of multiple mutant subunits.
Replacing site 2 alanine with bulky residues (valine or phenylalanine) to stabilize the core resulted in significantly larger disruptions in GABA A R activation with increased GABA EC 50 values (3-111-fold, Fig. 4, right, and Table 1), and decreased GABA efficacy, GABA current apparent rise times, and macroscopic current desensitization (Fig. 7, Tables 2 and 3). Depending on the subunit, the mutations had different effects. The valine and phenylalanine mutations in the ␣1 and ␤2 subunits had the largest effects, which increased with co-expression. On the other hand, increasing the packing of the ␥2 subunit ␤-core at site 2 had little effects on GABA EC 50 , GABA current apparent rise time, and macroscopic current desensitization (Table 1, Fig. 7) suggesting that GABA A R activation has distinct subunit requirements for ECD flexibility. The data are consistent with the fact that agonist binding occurs at the interfaces between GABA A R ␣ and ␤ subunit ECDs and that movements in the ECD triggered by agonist binding appear asymmetrical (51,58). Interestingly, when the ␥ mutant subunits were co-expressed with the ␣ and ␤ mutant subunits the magnitude of the effects on GABA A R activation were reduced (Table 1) suggesting that decreasing the flexibility of the ␥2 subunit can act as a positive modulator of GABA A R activation.
The Ala 3 Phe mutation, which replicates in GABA A R the packing at sites 1 and 2 of Ac-AChBP (leucine and phenylalanine), had the largest effects on GABA A R function (Tables 1  and 3). Compared with WT GABA A R, the ␣A136F␤A134F␥ mutant displayed a 111-fold increase in GABA EC 50 , a ϳ4fold slower current rise time, and a ϳ8-fold slower current desensitization rate (Fig. 7). It is unlikely that these effects are due to improper ECD folding. First, the IgG-fold common to pLGICs ECD and AChBP can accommodate at sites 1 and 2 a variety of residue pairs (from threonine-serine to leucinephenylalanine) with little adjustments of protein backbone, because the distance (C ␣ -C ␣ and C ␤ -C ␤ ) between sites 1 and 2 residues changes very little from structure to structure, either pLGIC or AChBP (supplemental Table S3). The GABA maximal elicited currents from oocytes expressing ␣A136F␤A134F␥ were smaller compared with oocytes expressing WT receptors (Table 1), which could be explained by decreased expression of surface receptors. However, when we compared the currents elicited by a saturating concentration FIGURE 7. Altering packing at sites 1 and 2 slows GABA current rise time and desensitization. Representative saturating GABA current traces from oocytes expressing WT and mutant GABA A Rs are peak normalized to highlight the slowing in GABA 10 -90% rise times (left) and current desensitization (right) for mutant receptors. Values are reported in Table 3. Bars represent mean Ϯ S.E. of 10 -90% rise times (s) and percent current desensitization after 20 s (((I max Ϫ I 20 sec )/I max ) ϫ 100) from WT and mutant receptors (n Ͼ 3 individual experiments). Black bars indicate values significantly different from WT (p Ͻ 0.05).
of GABA to those induced by a saturating concentration of PB in the same oocyte, the mutations had much larger effects of I GABAmax demonstrating that the mutations reduced GABA efficacy (Fig. 6, bottom left, and Table 2). The Ala 3 Phe mutation also decreased the efficacy of partial agonist P4S in the ␣␤A134F␥ mutant (Fig. 6, bottom right, and supplemental Table S4), further supporting the idea that tight packing at site 2 adversely affects receptor activation.
Similar results were seen for the double Val mutants, which mimic GLIC ␤-core packing with valines at sites 1 and 2. Along with increasing GABA EC 50 , the ␤-core packing reconfiguration slowed current activation and desensitization rates (Fig. 7). The ␣'VV'␤'VV'␥ and ␣'VV'␤'VV'␥'VV' mutant receptors displayed current profiles reminiscent of the slow activating, slow desensitizing GLIC (30), suggesting ␤-core packing influences current response properties. Again, for 'VV' mutant receptors, I PBmax were decreased to a much less extent than I GABAmax confirming that amino acid composition at sites 1 and 2 affects GABA efficacy.
It is important to note that increasing the rigidity of the ␤-core packing in GABA A Rs or nAChRs does not abolish function but rather hinders the ability of agonist to efficiently activate the receptor. In a similar fashion, whereas the common scaffold of AChBP apparently possesses the basic ability to trigger ion pore gating when fused to the 5-HT 3A R channel transmembrane domain and its loops 2, 7, and 9 are replaced with corresponding 5-HT 3A R loops, it is not as efficient as an eukaryotic pLGIC ECD in coupling binding to channel activation (59,60). Channel gating of the prokaryotic pLGIC homologue, GLIC, is also different compared with eukaryotic heteromeric pLGICs, with slow channel activation kinetics and little apparent desensitization (30). Taken together, the data suggest that loose packing was evolutionarily developed to increase ECD flexibility, which facilitates agonist-induced channel activation in nAChR, 5-HT 3 R, and GABA A R. Support for this idea comes from a recent study on glutamate receptors (GluRs). Like pLGICs, eukaryotic GluRs have much faster activation and desensitization rates than prokaryotic GluRs. This is likely due to loose packing of the hydrophobic core of Lobe 2, which confers to the ligand-binding domain of eukaryotic GluRs greater flexibility that speeds up the receptor response to agonist (61). Furthermore, a recent study on the GlyR proposes channel activation is accompanied by reorganization of the hydrophobic core (62).
Enhanced ECD flexibility due to loose packing of the ␤-core is also supported by the structural information available. In the crystal structures of AChBP in complex with agonists, no significant ligand-dependent conformational changes are observed other than movement of loop C to cap the ligand-binding pocket (63,64). Similarly, comparing the recent crystal structures of GLIC at pH 7 (corresponding to closed state (38)) and pH 4.6 (presumably open state (32)) limited reorganization of the two core-forming ␤-sheets is observed. Larger conformational changes are observed in eukaryotic receptors. In the low-resolution electron microscopy analysis of the Torpedo nAChR, agonist-induced rotations of the ␤-core inner and outer ␤-sheets about an axis normal to the membrane plane were detected in the ␣-subunits (51). Molecular dynamics simulations also find larger flexibility of the nAChR-␣1 ECD at sites 1 and 2 compared with the prokaryotic receptor ELIC, which has valine and phenylalanine at the two sites like Ls-AChBP (34).
This structural information correlates well with our computational analysis. The packing of GLIC at sites 1 and 2 (valine and valine, for a total volume of 160Å 3 of buried side chains) is more rigid than that of nAChR-␣1 (threonine and serine, 85 Å 3 ), but less rigid than that of AChBP (valine/ leucine and phenylalanine, 210 -235 Å 3 ) (Fig. 2, bottom). In our FOLDX simulations, mutations that reduce side chain volume at both sites in GLIC are less energetically costly compared with AChBP (about 3 versus 5 kcal/mol, see Fig. 3), suggesting a slightly higher propensity to structural flexibility of the ECD ␤-core, in agreement with the crystallographic evidence above.
Overall, the efficiency of allosteric activation seems to depend on the flexiblity of the ␤-core. Taken together, our results support the idea that during the evolution of pLGIC loose packing of the ECD ␤-sandwich core was acquired to increase flexibility to facilitate the allosteric activation of nAChRs, GABA A Rs, and 5-HT 3 Rs. This provides novel insight into the activation mechanism of pLGICs and other related allosteric proteins.