β Subunit M2–M3 Loop Conformational Changes Are Uncoupled from α1 β Glycine Receptor Channel Gating: Implications for Human Hereditary Hyperekplexia

Hereditary hyperekplexia, or startle disease, is a neuromotor disorder caused mainly by mutations that either prevent the surface expression of, or modify the function of, the human heteromeric α1 β glycine receptor (GlyR) chloride channel. There is as yet no explanation as to why hyperekplexia mutations that modify channel function are almost exclusively located in the α1 to the exclusion of β subunit. The majority of these mutations are identified in the M2–M3 loop of the α1 subunit. Here we demonstrate that α1 β GlyR channel function is less sensitive to hyperekplexia-mimicking mutations introduced into the M2–M3 loop of the β than into the α1 subunit. This suggests that the M2–M3 loop of the α subunit dominates the β subunit in gating the α1 β GlyR channel. A further attempt to determine the possible mechanism underlying this phenomenon by using the voltage-clamp fluorometry technique revealed that agonist-induced conformational changes in the β subunit M2–M3 loop were uncoupled from α1 β GlyR channel gating. This is in contrast to the α subunit, where the M2–M3 loop conformational changes were shown to be directly coupled to α1 β GlyR channel gating. Finally, based on analysis of α1 β chimeric receptors, we demonstrate that the structural components responsible for this are distributed throughout the β subunit, implying that the β subunit has evolved without the functional constraint of a normal gating pathway within it. Our study provides a possible explanation of why hereditary hyperekplexia-causing mutations that modify α1 β GlyR channel function are almost exclusively located in the α1 to the exclusion of the β subunit.


Introduction
Imbalance between the excitatory and inhibitory neurotransmission systems is the cause of many neurological disorders. Human hereditary hyperekplexia (startle disease), which is characterized by exaggerated startle reflexes and hypertonia in response to sudden, unexpected auditory or tactile stimuli, is a neuromotor disorder caused by dysfunction of inhibitory glycinergic neurotransmission in the spinal cord [1]. The majority of genetic mutations identified so far for this disorder have been mapped onto the postsynaptic neurotransmitter receptor, the glycine receptor (GlyR) chloride channel. The synaptic GlyR exists predominantly as the heteromeric a1 b form [2]. However the hereditary hyperekplexia-causing mutations of the GlyR are almost exclusively located in the a1 to the exclusion of b subunit, a fact which has puzzled the field for many years [1,3,4].
The GlyR, together with several other postsynaptic neurotransmitter receptors including the nicotinic acetylcholine receptor (nAChR), the type 3 5-hydroxytryptamine receptor (5HT 3 R), and the type A c-aminobutyric acid receptor (GABAAR), belong to the Cys-loop receptor ligand-gated ion channel superfamily, because they share common structural and functional characteristics [4,5,6]. The members of this superfamily exist as pentamers.
Each subunit is composed of an N-terminal extracellular domain (ECD) and a transmembrane domain (TMD). The TMD is comprised of four a-helical transmembrane segments (M1-M4) and a large intracellular domain between M3 and M4. Agonists bind to the receptor in a pocket that is formed by the principle (+) and complementary (-) sides of adjacent ECDs. Agonist binding, through a gating pathway, ultimately leads to the opening of a gate in the channel pore, which is formed by the M2 TMDs [7,8,9,10,11,12,13,14,15,16].
The putative stoichiometry of the a1 b GlyR is 2a1:3b [17], although other stoichiometries may also be possible. The a1 b GlyR has been shown to bind the agonist glycine at both the a+/ b2 and b+/a2 subunit interfaces, and agonist binding at either interface is sufficient to activate the channel [17]. Therefore, it seems that the a1 and b subunits play equivalent roles at the agonist-binding level. However, how the a1 and b subunits contribute to the downstream channel gating pathway is barely known. Hyperekplexia-causing GlyR a1 mutations can be classified into two groups: those that disrupt channel function and those that reduce surface expression. Interestingly, most mutations that disrupt GlyR channel function are concentrated within the a1 subunit gating pathway [1,3] (Fig. 1). Therefore, addressing the question of how the a1 and b subunits contribute to the channel gating pathway is key to solving the puzzle of the predominance of the GlyR a1 mutations in hereditary hyperekplexia.
To address this question, we concentrated on one of the essential structural components of the channel gating pathway, the M2-M3 loop ( Fig. 2A). Mutations in this region have been shown to cause drastic effects on channel function in many members of the Cys-loop receptor superfamily [18,19,20,21]. More importantly, this region of the GlyR a1 subunit hosts mutations responsible for most cases of hereditary hyperekplexia, such as R271(199)Q/L, K276(249)E, and Y279(279)C [1,4,20]. In addition, a systematic alanine-scanning of this region in the homomeric a1 GlyR further reveals that mutations of a few other residues, notably V277(259)A, also mimic the phenotype of hereditary hyperekplexia-causing mutations [20]. The primed numbers in brackets after the names of residues refer to the standard M2 domain numbering system that assigns 19 to the innermost M2 residue. This numbering system will be used henceforth as it enables residues from different subunits to be compared.
In this study, we compared effects of hyperekplexia-mimicking mutations in the M2-M3 loops of a1 and b subunits. We found that a1 b GlyR channel function is less sensitive to mutations introduced into the b than into the a1 subunit. We conclude that the b subunit M2-M3 loop plays a minor role in a1 b GlyR channel gating. A further attempt to identify the possible mechanism underlying this phenomenon indicates that the agonist-induced conformational changes in the b subunit M2-M3 loop are uncoupled from a1 b GlyR channel gating. In addition, we also discovered that the structural components responsible for this are distributed throughout the b subunit, implying that the b subunit has evolved without the functional constraint of a normal gating pathway within it. Our study provides a possible explanation of why hereditary hyperekplexiacausing mutations that modify a1 b GlyR channel function are almost exclusively located in the a1 to the exclusion of the b subunit.

a1 b
GlyR channel function is less sensitive to hyperekplexia-mimicking mutations introduced into the M2-M3 loop of the b than into the a1 subunit Previously we reported that a1 b GlyR channel function was not sensitive to cysteine mutations introduced into the M2-M3 loop of the b subunit. We also showed that the channel function of these cysteine-substituted a1 b GlyRs did not change in response to treatment with a cysteine-reactive compound, 2-trimethylammoniumethylmethane thiosulfonate (MTSET) [22]. However, there is a possibility that the cysteine mutations might cause gain-of-function: for example, a disulfide bond could form between the M2-M3 loops of adjacent subunits [23]. In addition, the lack of response to MTSET treatment might be due to the residue not being labeled by MTSET rather than not being sensitive to MTSET modification. We therefore introduced the functionally-inert Ala mutation to verify the results obtained from these experiments.
The following hyperekplexia-mimicking mutations in the M2-M3 loop were introduced, one at a time, to disrupt the channel function: K249A, V259A, and Y279A ( Fig. 2A). These mutations, when introduced into the a1 subunit, each cause a dramatic increase in glycine EC 50 of the homomeric a1 GlyR [20], which in principle could be due to compromised agonist binding, disrupted channel gating, or a mixture of both [24]. As the M2-M3 loop is spatially distant from the agonist binding site based on various structures of Cys-loop receptor members [10,12,13,16] and is temporally downstream from the agonist binding site in the channel gating pathway based on single-channel kinetic analysis [9], the increase in agonist EC 50 values caused by the K249A, V259A, and Y279A mutations can be attributed predominantly to a disrupted channel gating efficacy. This hypothesis is supported by the results of single-channel kinetic analyses on the M2-M3 loop mutants of both the GlyR and nAChR [25,26]. Here we use the agonist EC 50 as an index of channel gating efficacy changes to compare the effects of K249A, V259A, and Y279A mutations, when introduced into the a1 versus b subunits, on channel gating. Similar strategies (using either agonist EC 50 values or voltages of half activation) have been successfully employed to probe the gating mechanisms of both ligand-and voltage-gated channels [27,28]. Fig. 2B shows sample currents recorded in response to glycine of increasing concentrations in HEK293 cells expressing a1WT bWT, a1K249A bWT and a1WT bK249A GlyRs. The glycine concentration-response curve of the a1K249A bWT GlyR was dramatically right-shifted relative to that of the a1WT bWT GlyR (Fig. 2C). The a1K249A bWT GlyR exhibited an EC 50 of 500680 mM, which was much higher than the corresponding value, 1162 mM, recorded in the a1WT bWT GlyR (p,0.01, Table 1). In contrast, the same mutation introduced into the b subunit had no effect on the glycine EC 50 . The concentrationresponse curve of the a1WT bK249A GlyR almost overlapped that of the a1WT bWT GlyR ( Fig. 2B and C), and the EC 50 of the a1WT bK249A GlyR was not significantly different from that of the a1WT bWT GlyR (1361 mM versus 1162 mM, p.0.05, Table 1).
As noted above, the degree to which channel gating is disrupted is reflected by the increase in agonist EC 50 . Thus, the ratio of mutant to WT glycine EC 50 (R M/W ) provides an index of the extent to which channel gating has been disrupted. We will use R M/W in the following text to compare the degrees to which the respective subunits contribute to channel gating. K249A when introduced into the a1 subunit disrupted the a1 b GlyR channel gating by a factor 45612, which is the EC 50 of the a1K249A bWT GlyR divided by the EC 50 of the a1WT bWT GlyR (Table 1). On the other hand, K249A when introduced into the b subunit disrupted the a1 b GlyR channel gating by a factor 1.260.3, which is the EC 50 of the a1WT bK249A GlyR divided by the EC 50 of the a1WT bWT GlyR ( Table 1).
As summarized in Fig. 2D and Table 1, virtually identical results were obtained for V259A and Y279A. We thus conclude that the a1 b GlyR gating efficacy is more affected when the disruption of the gating pathway occurs in the a1 than in the b subunit.
One potential problem in drawing such a conclusion is that there is a possibility that the b subunit was not expressed and that the recorded currents may have arisen from homomeric a1 GlyRs. In such a case, no matter how the a1 and b subunits contribute to the a1 b GlyR channel gating, mutations introduced into the M2- M3 loop of a1 subunit would increase glycine EC 50 , as reported previously [20], and mutations introduced into the b subunit would not affect channel function at all, and give results similar to those we obtained. To eliminate this possibility and to maximize the expression of heteromeric a1 b versus homomeric a1 GlyRs, we transfected cells with a1 and b cDNAs in a ratio of 1:10. Moreover, we tested the sensitivity of the glycine-induced current to picrotoxin wherever there was a possibility that the recorded current may have arisen from homomeric a1 GlyRs. The heteromeric a1 b GlyR has been shown to be resistant to picrotoxin blockade compared to the homomeric a1 GlyR [29,30]. The magnitude of picrotoxin blockade can therefore reflect the degree to which the heteromeric a1 b GlyR versus the homomeric a1 GlyR has been expressed. In our experiments, only those cells showing significant picrotoxin resistance (Table 1) were used for further glycine concentration-response investigation. The picrotoxin sensitivity testing was applied only to receptors incorporating a1WT, a1K249A, and a1V259A subunits, as no glycine-induced current was detected for the homomeric a1Y279A GlyR (Table 1).
b subunit M2-M3 loop conformational changes are uncoupled from a1 b GlyR channel gating We next sought to determine the mechanism underlying the asymmetrical contributions of the a1 and b subunits to channel gating, i.e. how differently the M2-M3 loops of the a1 and b subunits responded during a1 b GlyR channel gating. To achieve this, we examined conformational changes that the M2-M3 loops of the a1 and b subunits experienced during channel gating by using voltage-clamp fluorometry (VCF). VCF correlates conformational changes occurring at the gate with those occurring in some other domain of interest in real-time [31,32]. A rhodamine fluorescent dye was used to label the M2-M3 loop, because rhodamine fluorescence exhibits an increase in quantum efficiency as the hydrophobicity of its environment is increased. Thus, rhodamine fluorescence intensity reports local conformational changes that cause a change in its immediate chemical microenvironment. These experiments were carried out in Xenopus oocytes as fluorescence detection is not routinely possible in HEK293 cell-expressed GlyRs [32]. Previously we reported that rhodamine methanethiosulfonate (MTSR), when attached to the cysteine-substituted 199 residue in the homomeric a1 GlyR via a disulfide bond, exhibited an increase in fluorescence intensity upon glycine binding [33]. As the current and fluorescence glycine concentration-response relationships overlapped, we concluded that the fluorophore reported M2-M3 loop conformational changes associated with channel gating.
Cysteine mutations were introduced into either the a1 or b subunits at the 199 position, and the mutant subunits were coexpressed with the respective WT b or a1 subunits. As shown in Fig. 3A and B, for the a1R199C b GlyR, where the fluorophore reports conformational changes of the a1 M2-M3 loop, the fluorescence intensity was increased upon glycine application. Moreover, the concentration-response curves of fluorescence and current overlapped and the respective glycine EC 50 value was not significantly different from each other (329657 mM and 396631 mM, respectively, p.0.05, Table 2). This implies that the conformational changes of the a1 M2-M3 loop are coupled to the channel gating in the a1 b GlyR, which is similar to the situation previously demonstrated in the homomeric a1R199C GlyR [33]. In contrast, in the a1 bA199C GlyR, where the fluorophore reports conformational changes of the b M2-M3 loop, although the fluorescence intensity was increased upon glycine application as in the a1R199C b GlyR, the concentrationresponse curve of the fluorescence was dramatically right-shifted relative to that of the current ( Fig. 3C and D). The fluorescence glycine EC 50 value was 8.861.9 times larger than that of current (40.768.3 versus 4.6160.26 mM, p,0.01, Table 2). These data imply that conformational changes of the b M2-M3 loop are uncoupled from channel gating in the a1 b GlyR. The degree of uncoupling is reflected by the ratio of glycine EC 50 values between fluorescence and current (R F/I ). These values were 1.260.2 and 8.861.9 ( Fig. 3E and Table 2) for the a1 and b subunits, respectively, which suggests that the gating signal takes the a1 subunit's gating pathway to activate the channel, whereas it bypasses the b subunit's gating pathway.
Structural basis of the lower sensitivity of a1 b GlyR channel function to hyperekplexia-mimicking mutations introduced into the M2-M3 loop of the b than into the a1 subunit We investigated whether the minor role of the b subunit in a1 b GlyR channel gating was due to structural differences in the ECD or TMD. To address this question, we constructed two chimeras of a1 and b subunits ( Fig. 4A and Fig. S1). Chimera a-b comprises the ECD of the a1 subunit and the TMD (including the M3-M4 domain) of the b subunit. Conversely, chimera b-a comprises the ECD of the b subunit and the TMD of the a1 subunit. We then investigated how these chimeras mimicked the b subunit to contribute to a1 b GlyR channel gating, by co-expressing each chimera with the a1 subunit (10:1 ratio) and examining the R M/W s of the hyperekplexia-mimicking mutations introduced into the a1 and chimeric subunits. It is worth noting that neither chimera, when transfected alone into HEK293 cells, induced any current upon the application of glycine at concentrations up to 100 mM (data not shown).
When the K249A mutation was introduced, the glycine EC 50 s of the a1K249A a-bWT and a1WT a-bK249A GlyRs were 83619 and 2564 mM, respectively. The relevant R M/W s of the a1 and ab subunits in the a1 a-b GlyR were 4.361.3 and 1.360.3 ( Fig. 4B and Table 1), respectively. On the other hand, the glycine EC 50 s of the a1K249A b-aWT and a1WT b-aK249A GlyRs were 6846157 and 36610 mM, respectively. The relevant R M/W s of the a1 and b-a subunits in the a1 b-a GlyR were 2167 and 1.160.4 ( Fig. 4B and Table 1), respectively. These data contrast dramatically with the corresponding values (45612 and 1.260.3) calculated for the a1 b GlyR (Fig. 4B and Table 1). The R M/W s of the a1 subunit in both the a1 a-b and a1 b-a GlyRs were significantly less than that of the a1 b GlyR (Fig. 4B, p,0.01 and p,0.05, respectively), suggesting that when the gating pathway is disrupted in the a1 subunit, both the a-b and b-a subunits partially restore channel gating efficacy to that of the WT a1 GlyR. Both the a-b and b-a subunits therefore behave less like the b subunit but more like the a1 subunit. This trend was also found when the other two mutations, V259A and Y279A, were investigated in the same way (Fig. 4C, D and Table 1).
It is worth noting that the a-b subunit behaves more like the a1 subunit than does the b-a subunit, based on their abilities to compensate the disrupted gating pathway in the accompanying a1 subunit (Fig. 4B-D). Indeed, the difference in the a1 subunit R M/W s between the a1 b-a and a1 b GlyRs was so minor that it was not even significant in the case of the Y279A mutation (Fig. 4D). Taken together, it seems that the ECD plays a more important role than the TMD in determining the minor role of the b subunit in a1 b GlyR channel gating.
We further dissected the ECD to determine which subdomains contributed to the b subunit's minor role in a1 b GlyR channel gating. The ECDs of the Cys-loop receptors comprise agonist binding sites at subunit interfaces and transition zones, which relay the agonist-binding information to the channel pore. The agonist binding site is formed by the loops A, B, and C from the (+) subunit interface and loops D, E and F from the (-) subunit interface, while the transition zone is formed by loop 2, the conserved Cys-loop and the pre-M1 linker [7,10,11,12,13,15,16]. The agonist binding site and transition zone have been shown to function as relatively independent modules [11]. We therefore investigated which domain might be responsible for the minor role of the b subunit in a1 b GlyR channel gating. To achieve this, two chimeras were constructed. Chimera a B -b comprised the a1 subunit agonist binding site and the b subunit transition zone and TMDs, while chimera a T -b comprised the a1 subunit transition zone and the b subunit agonist binding site and TMDs (Fig. 4A and Fig. S1). Both chimeras were co-expressed with the a1V259A subunit, and the ability of each to compensate the disrupted channel gating pathway of the accompanying a1 subunits was examined. The V259A mutation was investigated here because, among the K249A, V259A and Y279A mutations, the a1V259A subunit showed the largest difference in R M/W between the a1V259A b and a1V259A a-b GlyRs (Fig. 4B-D and Table 1). We therefore expected this mutation would most clearly distinguish the a B -b or a T -b subunits from the b or a-b subunits when comparing their abilities to compensate the disrupted gating pathway in the accompanying a1 subunits.
As shown in Fig. 4C and Table 1, the glycine EC 50 s of the a1V259A a B -bWT and a1V259A a T -bWT GlyRs were 186621 and 455611 mM, respectively. The relevant R M/W s of the a1V259A subunit were 11.662.7 and 3363, respectively, both of which are significantly less than that of the a1V259A b GlyR (71620, p,0.01 and p,0.05, respectively). This indicates that both chimeras compensate the disrupted channel gating pathway in the a1V259A subunit to some degree, but neither of them to the same extent as the a-b subunit (Fig. 4C and Table 1). It is thus evident that both the agonist binding domain and the transition zone of the ECD contribute to the b subunit's minor role in a1 b GlyR channel gating.
The structural basis of the uncoupling of b subunit M2-M3 loop conformational changes from a1 b GlyR channel gating As noted above, the reduced sensitivity of a1 b GlyR channel function to hyperekplexia-mimicking mutations introduced into the b subunit was mirrored by the uncoupling of the b M2-M3 loop conformational changes from the channel gating through VCF examination. We next investigated the structural basis for this uncoupling using the same chimera strategy as described in the previous section. We used VCF to monitor conformational changes experienced by the labeled 199C residues of the a-b, b-a, a B -b and a T -b subunits when they were co-expressed with the a1  subunit. It is noteworthy that neither glycine-induced current nor fluorescence changes could be detected from any of these chimeric subunits when expressed alone in oocytes (data not shown). Therefore, the fluorescence and current changes we detected when they were co-expressed with the a1 subunit must have arisen from heteromers formed with the a1 subunit. As shown in Fig. 5A and C, in the a1 a-b199C GlyR, the fluorescence and current EC 50 values were 6.1561.08 and 4.4660.43 mM, respectively, and the R F/I was 1.460.3 (Table 2). On the other hand, in the a1 b-a199C GlyR, the fluorescence and current EC 50 values were 92.165.5 and 22.261.7 mM, respectively, and the R F/I was 4.160.4 ( Fig. 5B and C and Table 2). The R F/I s of both the a-b and b-a subunits were significantly less than that of the b subunit, whose R F/I was 8.861.9 (p,0.01 and p,0.05, respectively). This implies that both the ECD and TMD contribute to the uncoupling of the b subunit's M2-M3 conformational changes from the channel gating. The ECD, however, might play a major role since the R F/I of the a-b subunit is not significantly different from that of the a subunit (p.0.05), while the b-a subunit R F/I is closer to that of the b subunit (Fig. 5C). This is consistent with the suggestion that the ECD dominates in determining the minor role of the b subunit in a1 b GlyR channel gating, obtained from the hyperekplexia-mimicking mutation experiments described above.
When further dissecting the ECD, in the a1 a B -b199C GlyR, the fluorescence and current EC 50 values were 18.061.9 and 3.860.3 mM, respectively, and the R F/I was 4.760.6 ( Fig. 5C and Table 2). On the other hand, in the a1 a T -b199C GlyR, the fluorescence and current EC 50 s were 14.061.0 and 2.160.2 mM, respectively, and the R F/I was 6.860.9 ( Fig. 5C and Table 2). The R F/I s of both chimeras lay between those of the a1 and b subunits (1.260.2 and 8.861.9, respectively), which implies that both the agonist binding site and transition zone contribute to the uncoupling of the b subunit's M2-M3 conformational changes from the channel gating. It is noteworthy that the transition zone might play a minor role, as the R F/I of the a1 a T -b199C GlyR was not statistically significantly different from that of the a1 b199C GlyR (p.0.05, Fig. 5C).

Discussion b subunit plays a minor role in a1 b GlyR channel gating
To investigate how the a1 and b subunits each contribute to the channel gating, we assumed that the more a certain subunit contributes to channel gating, the more channel function is compromised when the gating pathway is disrupted in this subunit. By introducing hyperekplexia-mimicking mutations to the M2-M3 loops of the a1 and b subunits, we found that disrupting the channel gating pathway within the a1 subunit had drastic effect on the overall a1 b GlyR function, whereas disrupting the channel gating pathway via the corresponding mutations within the b subunit had little effect. Thus, our results suggest that the a1 subunit dominates channel gating while the b subunit plays only a minor role. Asymmetrical contributions to Cys-loop receptor gating have previously been suggested in the nAChR. For example, cryo-electron microscopic structure analysis shows that the M2 pore-lining domains of the a subunits engage a rotation relative to those of the non-a subunits during channel gating [16], and single channel recording analysis shows a negligible coupling between the pre-M1 linker, the Cys-loop and the M2-M3 loop in the non-a subunits [14]. Our results imply that the a1 subunit of the GlyR behaves like the a subunit of the nAChR, while the b subunit of the GlyR behaves like the non-a subunits of the nAChR.  b199C, a a-b199C, a b-a199C, a a B -b199C and a a T - We employed VCF to investigate the mechanism underlying the asymmetrical contribution of the a1 and b subunits to channel gating. We found that the concentration of agonist required to induce a change in fluorescence of the fluorophore attached to the b subunit is higher than that required to activate the channel. In contrast, the corresponding fluorescence and current concentration-response curves are overlapping when the a1 subunit is labeled. Such an uncoupling of M2-M3 conformational changes from the agonist-induced channel gating has also been demonstrated in the b subunit of the nAChR [34]. A possible explanation for such an uncoupling is that when two or three agonist binding sites are occupied by a low concentration of agonist, the gating pathway is activated along the a1 subunit and the channel is activated. This channel activation might have reached its maximum, since it has been shown that two or three bound agonists are required for full activation of homomeric a1 [35,36,37,38] and heteromeric a1 b GlyRs [39]. In addition, a recent study has shown that in the homomeric a7nAChR-5HT3A receptor, three occupied agonist binding sites at nonconsecutive subunit interfaces are required to exhibit maximal mean channel open time [40]. Therefore, the binding of additional (4th and 5th) agonists when a high concentration of agonist is present, which leads to the gating pathway activation of the b subunit, would not further the channel opening. In other words, a functional b subunit is dispensable, and further disruption of this gating pathway would have no effect on the overall channel gating of the a1 b GlyR.
There is a possibility that fluorescence changes in the a1 b199C GlyR reflect conformational changes when the channel is desensitized, as both desensitization and fluorescence change appeared only when a high-concentration of glycine was applied (Fig. 3C). However, we consider this is unlikely since fluorescence changes occurred instantly while receptors accumulate in desensitized states with a much slower time course (Fig. 3C).
Structural basis of the minor role of the b subunit in a1 b GlyR channel gating By testing chimeras constructed from the a1 and b subunits, we found that both the ECD and TMD were responsible for the b subunit's minor role in a1 b GlyR channel gating ( Fig. 4 and 5), although it seems that the ECD plays a dominant role. We originally suspected that the 199 residue might be the cause since this residue is an Ala in the b subunit ( Fig. 2A) and, in the a1 subunit, the R199A mutation has been shown to drastically compromise a1 GlyR channel function and mimic the phenotype of hyperekplexia-causing mutations [20]. However, when we introduced the A199R mutation into the b subunit, the contribution of the b subunit to the a1 b GlyR channel gating was not changed (Table 1). Indeed, the natural existence of the 199A residue in the b subunit might be explained by the fact that, since the b subunit M2-M3 loop is not involved in channel gating and hence not sensitive to mutations, whether a gating-favorable 199R or gating-disfavorable 199A exists in the b subunit makes no difference to a1 b GlyR channel gating. Instead, the ECD, which is upstream from the M2-M3 loop in the gating pathway, plays a major role in limiting the contribution of the b subunit to channel gating.
When further dissecting the role of the ECD, our chimera studies indicate that motifs in both the agonist binding site and transition zone determine the contribution of the b subunit to overall receptor gating ( Fig. 4 and 5). Thus, no single domain is responsible for the minor role of the b subunit in a1 b GlyR channel gating. One possible explanation is that residues contributing to its minor role are distributed throughout the b subunit, including the agonist binding site, the transition zone and the transmembrane channel pore domains ( Fig. 4 and 5). As a result, no single domain from the a1 subunit is able to completely rescue the gating contribution of the b subunit to the level of the a1 subunit.
From an evolutionary perspective, we suggest that the reason why residues that disrupt the b subunit gating are distributed evenly throughout its coding region is that the gating pathway within the b subunit was not optimized to a ''normal level'' as in the a1 subunit, when the b subunit joined the a1 subunit to form the heteromeric a1 b GlyR. It has been speculated that ancestral Cys-loop receptors were homomers [41,42] and that the ancestral GlyR might exist in the homomeric a form. Alternatively, even if the gating pathway of the b subunit was equivalent to that of the a1 subunit when the heteromeric a1 b GlyR came into being, random mutations that compromise its channel gating pathway could have accumulated throughout the b subunit during evolution. This is because a normal gating pathway within the b subunit is dispensable for a1 b GlyR channel function and it would not serve as a constraint on the b subunit during evolution. Thus, no single domain from the a1 subunit could rescue b subunit gating efficacy.
The GlyR b subunit is reminiscent of the non-a muscle nAChR subunits, which have much higher ratios of the number of nonsynonymous substitutions to that of synonymous substitutions than the a muscle nAChR subunit, implying less functional constraint on the non-a than a nAChR subunits during evolution [43]. Another example is the AChBP, whose acetylcholine binding but not gating pathway is the function subjected to evolutionary pressure. When the AChBP is connected to the 5HT3R TMD, a functional channel can be formed only if the disabled gating pathway components in the AChBP are replaced by the corresponding ones from the 5HT3R [11].

Implications for the distribution of hereditary hyperekplexia-causing mutations in the a1 b GlyR
GlyR hereditary hyperekplexia-causing mutations have been mapped almost exclusively onto the gene of the a1 to the exclusion of the b subunit (Fig. 1). This is possibly because the gene of the a1 subunit is a hot spot, more amenable to genetic mutations than that of the b subunit, but it seems more likely because mutations occurring in the a1 subunit more drastically affect a1 b GlyR channel function than those occurring in the b subunit. This assumption is supported by our experiments showing that hyperekplexia-mimicking mutations introduced into the b subunit have much less effect on a1 b GlyR channel function than those introduced into the a1 subunit.
More interestingly, most mutations identified on the a1 subunit that affect channel function (rather than surface expression), cluster either in domains associated with the channel gating pathway (i.e., the M2-M3 loop, loop 2 and the pre-M1 linker) or along the pore-lining M2 domain, but rarely occur in the agonist binding sites (Fig. 1) [1,3,4]. This can be explained by the fact that the a1 b GlyR has five potential agonist binding sites and only two or three functional sites are required for efficient gating [17,39]. Thus, introducing mutations into the agonist binding sites of either the a1 or b subunit will have no effect on overall a1 b GlyR function [17]. In other words, the a1 and b subunits can compensate each other at the agonist binding level, but this compensation between the a1 and b subunits does not pass on to the downstream channel gating pathway. In summary, our experiments provide a possible explanation of why hereditary hyperekplexia-causing mutations concentrate in the channel gating pathway of the a1, to the exclusion of the b subunit, in the a1 b GlyR.

Mutagenesis and chimera construction of the GlyR cDNAs
The human GlyR a1 and b subunit cDNAs were subcloned into the pcDNA3.1zeo+ (Invitrogen) or pGEMHE [44] plasmid vectors for expression in HEK293 cells or Xenopus oocytes, respectively. Site-directed mutagenesis and chimera construction were performed using the QuickChange (Stratagene, La Jolla, CA, USA) mutagenesis and multiple-template-based sequential PCR protocols, respectively.
The multiple-template-based sequential PCR protocol for chimera construction was developed in our laboratory and has recently been described in detail elsewhere [45]. This procedure does not require the existence of restriction sites, or the purification of intermediate PCR products, and needs only two or three simple PCRs followed by general subcloning steps. Most importantly, the chimera joining sites are seamless and the success rate for construction is nearly 100%. The joining sites used in our experiment were chosen based on the following principles: (1) A site, based on the crystal structure of the AChBP [7], is to be near the boundary between the two flanking loops to minimize disturbance on the loop structures. (2) The pair of residues of a joining site is to be conserved between the GlyR a and b subunits, wherever possible. The joining sites used in our experiment are between the following pairs of residues: a L135-T136 and b I157-T158 for the N-terminus of the Cys-loop, a Q155-L156 and b Q178-L179 for the C-terminus of the Cys-loop, a T208-C209 and b T232-C233 for the N-terminus of the pre-M1 linker, and a R218-Q219 and b R242-Q243 for the C-terminus of the pre-M1 linker (Fig. S1). The a R218-Q219 and b R242-Q243 are also the joining sites for chimeras constructed between the ECD and TMD. The loop 2 transposition was achieved by incorporating either the aA52Q or bQ73A mutations, as the loop 2 sequences between the a1 and b subunits are otherwise conserved.
For the VCF experiments, both GlyR a1 and b subunit cDNAs in the pGEMHE vector were mutated to substitute non-essential background cysteines with alanines, including a1C41A and bC115AC291A [46,47].
For the b-a chimeras used for determining the effect of hyperekplexia-mimicking mutations on a1 b GlyR channel function, the Thr at the M2 69 position was replaced by a Cys. Thr to Cys mutation at this site in the homomeric a1 GlyR does not affect glycine activation, but does confer picrotoxin resistance on the channel [30]. Through such a modification, picrotoxin resistance was used to distinguish the heteromeric a1 b-a GlyR from the homomeric a1 GlyR when the a1 and b-a subunits were co-expressed.

HEK293 cell culture, expression and electrophysiological recording
The effects of the hyperekplexia-mimicking mutations, K249A, V259A and Y279A, were examined on GlyRs expressed in HEK293 cells (ATCC). Details of the HEK293 cell culture, GlyR expression and electrophysiological recording of the HEK293 cells are described elsewhere [30]. Briefly, HEK293 cells were maintained in DMEM supplemented with 10% fetal bovine serum. Cells were transfected using a calcium phosphate precipitation protocol. When co-transfecting the a1 together with the b or any other chimera subunits, their respective cDNAs were combined in a ratio of 1:10. In addition, the pEGFP-N1 (Clontech) was co-transfected to facilitate identifying the transfected cells. Glycine-induced currents were measured using the whole cell patch-clamp configuration. Cells were treated with external Ringer's solution and internal CsCl solution [30]. Cells were voltage-clamped at 240 mV. When the heteromeric GlyRs were expressed, the picrotoxin sensitivity was tested to confirm that the majority of receptors are heteromers [29,30]. A 10 mM or 100 mM concentration of picrotoxin was applied to the heteromeric GlyRs in the presence of glycine at the EC 50 concentration of their corresponding a1 homomers. Only the cells with significant picrotoxin resistance compared with their a1 homomers, e.g. where 100 mM picrotoxin inhibited the current by less than 50%, were used for further glycine-sensitivity examination.
Oocyte preparation, expression and VCF recording VCF experiments were performed on GlyRs expressed in Xenopus laevis oocytes. Female Xenopus laevis frogs were purchased from Xenopus Express, France. Details of oocyte preparation, GlyR expression and VCF recording are described elsewhere [33]. Briefly, the mMessage mMachine kit (Ambion, Austin, TX) was used to generate capped mRNA. The mRNA was injected into oocytes of the female Xenopus laevis frog with 10 ng (1 ng a1 and 9 ng b or any other chimeric subunits) per oocyte. After injection, the oocytes were incubated in ND96 solution [33] for 3-4 days at 18 uC before recording.
The sulfhydryl-reactive reagent, sulforhodamine methanethiosulfonate (MTSR, Toronto Research Chemicals, North York, Ontario, Canada), was used to label 199C residues. On the day of recording, the oocytes were labeled with 10 mM MTSR for 25 s, either in the absence or presence of glycine. The oocytes were then transferred to the recording chamber and perfused with ND96 solution. The current was recorded by the two-electrode voltage clamp configuration and the recording electrode was filled with 3 M KCl. Cells were voltage-clamped at 240 mV. The fluorescence was recorded using the PhotoMax 200 photodiode detection system (Dagan Corp., Minneapolis, MN).

Data analysis
Results are expressed as mean6standard error of the mean of three or more independent experiments. The empirical Hill equation, fitted by a non-linear least squares algorithm (SigmaPlot 9.0, Systat Software, Point Richmond, CA), was used to calculate the EC 50 and Hill coefficient (n H ) values for glycine-induced current and fluorescence change. Statistical significance was determined using the Student's t-test. Figure S1 Amino acid sequence alignment between the human GlyR a1 and b subunits. The joining sites for chimera construction are highlighted in blue. The K249, V259 and Y279 residues, where hyperekplexia-mimicking mutations were introduced, are highlighted in red. The a1R199 and bA199 residues, where the Cys mutation was introduced for VCF experiment, are highlighted in green. (DOC)