A triad of residues is functionally transferrable between 5-HT3 serotonin receptors and nicotinic acetylcholine receptors

Cys-loop receptors are pentameric ligand-gated ion channels that facilitate communication within the nervous system. Upon neurotransmitter binding, these receptors undergo an allosteric activation mechanism connecting the binding event to the membrane-spanning channel pore, which expands to conduct ions. Some of the earliest steps in this activation mechanism are carried out by residues proximal to the binding site, the relative positioning of which may reflect functional differences among members of the Cys-loop family of receptors. Herein, we investigated key side-chain interactions near the binding site via mutagenesis and two-electrode voltage-clamp electrophysiology in serotonin-gated 5-HT3A receptors (5-HT3ARs) and nicotinic acetylcholine receptors (nAChRs) expressed in Xenopus laevis oocytes. We found that a triad of residues aligning to Thr-152, Glu-209, and Lys-211 in the 5-HT3AR can be exchanged between the homomeric 5-HT3AR and the muscle-type nAChR α-subunit with small functional consequences. Via triple mutant cycle analysis, we demonstrated that this triad forms an interdependent network in the muscle-type nAChR. Furthermore, nAChR-type mutations of the 5-HT3AR affect the affinity of nicotine, a competitive antagonist of 5-HT3ARs, in a cooperative manner. Using mutant cycle analyses between the 5-HT3A triad, loop A residues Asn-101 and Glu-102, β9 residue Lys-197, and the channel gate at Thr-257, we observed that residues in this region are energetically linked to the channel gate and are particularly sensitive to mutations that introduce a net positive charge. This study expands our understanding of the differences and similarities in the activation mechanisms of Cys-loop receptors.

Cys-loop receptors are pentameric ligand-gated ion channels that facilitate communication within the nervous system. Upon neurotransmitter binding, these receptors undergo an allosteric activation mechanism connecting the binding event to the membrane-spanning channel pore, which expands to conduct ions. Some of the earliest steps in this activation mechanism are carried out by residues proximal to the binding site, the relative positioning of which may reflect functional differences among members of the Cys-loop family of receptors. Herein, we investigated key side-chain interactions near the binding site via mutagenesis and two-electrode voltage-clamp electrophysiology in serotonin-gated 5-HT 3A receptors (5-HT 3A Rs) and nicotinic acetylcholine receptors (nAChRs) expressed in Xenopus laevis oocytes. We found that a triad of residues aligning to Thr-152, Glu-209, and Lys-211 in the 5-HT 3A R can be exchanged between the homomeric 5-HT 3A R and the muscle-type nAChR ␣-subunit with small functional consequences. Via triple mutant cycle analysis, we demonstrated that this triad forms an interdependent network in the muscle-type nAChR. Furthermore, nAChR-type mutations of the 5-HT 3A R affect the affinity of nicotine, a competitive antagonist of 5-HT 3A Rs, in a cooperative manner. Using mutant cycle analyses between the 5-HT 3A triad, loop A residues Asn-101 and Glu-102, ␤9 residue Lys-197, and the channel gate at Thr-257, we observed that residues in this region are energetically linked to the channel gate and are particularly sensitive to mutations that introduce a net positive charge. This study expands our understanding of the differences and similarities in the activation mechanisms of Cysloop receptors.
The overall structure of Cys-loop receptors is highly conserved (8). Each subunit is composed of an extracellular domain (ECD) ␤ sandwich, a transmembrane domain formed by four ␣-helices (numbered M1-M4), and a variable intracellular domain defined primarily by the loop between the M3 and M4 helices. Five subunits assemble in a concentric manner to form a single receptor. Ligand-binding sites are formed at subunit interfaces in the ECD by three primary (ϩ) face "loops" A-C, and three complementary (Ϫ) face loops D-F. An overall picture of the interface between two subunits, as exemplified in the crystal structure of the homopentameric 5-HT 3A receptor, is shown in Fig. 1A (PDB code 4PIR) (9).
The advent of complete structures of eukaryotic Cys-loop receptors, whereas generally at fairly low resolution, has provided valuable guidance for comparing members of the Cysloop receptor family. The present work was motivated by the recognition of a triad of residues that are, in a global sense, conserved across several members of the family. Our primary focus has been on the homopentameric 5-HT 3A R, for which the residues of interest are Thr-152, Glu-209, and Lys-211 ( Fig. 1A; numbering as in the crystal structure; these have been referred to elsewhere at Thr-179, Glu-236, and Lys-238). The structurally aligning residues in the ␣-subunits of the recently reported heteromeric ␣4␤2 nAChR (PDB code 5KXI) (10) are Lys-145, Asp-200, and Thr-202, whereas those in the ␤3 GABA A homopentamer are Glu-153, Arg-207, and Ser-209 (PDB code 4COF) (11). In each triad, there is a cationic residue (Lys/Arg), an anionic residue (Asp/Glu), and a residue with an alcohol side chain (Ser/Thr). However, the relative positioning of these functionalities varies among these receptors, which led us to ask whether the arrangement reflects differences among receptors. To facilitate comparison, we have designated the three sites as A, B, and C (Fig. 1A). A sequence alignment (Fig. 1B) illustrates the conservation of these arrangements within individual receptor subfamilies in humans. Note that there are exceptions in the ␣5 and ␣10 nAChR subunits, however, these particular ␣-subunits do not contribute to the primary binding face of nAChRs (1). The ␣8 nAChR subunit was excluded because it is not found in mammals (1).
Herein we evaluate the functional role of this triad in two excitatory Cys-loop receptors, the 5-HT 3A R and the muscletype ((␣ 1 ) 2 ␤ 1 ␥␦) nAChR, using two-electrode voltage-clamp electrophysiology of mutant receptors expressed in Xenopus laevis oocytes. We find that the 5HT 3A R and nAChR triads can be swapped between these two receptors with minimal functional consequences. Additionally, substituting the nAChR triad into the 5-HT 3A R increases the affinity for nicotine, demonstrating that this triad influences ligand binding. In contrast, swapping the triad from the inhibitory GABA A R into the 5HT 3A R does not result in a functional receptor.
We note from the start that several of the residues considered here have been evaluated by other investigators, and we have built off that work. In muscle-type nAChRs, state-dependent interactions have been proposed between Lys-145, . The former two residues are part of the above-described "triad," and Tyr-190 sits on loop C and is part of the aromatic box that contributes to the agonist-binding site (18,19). The key hydrogen bonding role proposed for Tyr-190, however, is not possible in the 5-HT 3A R, as the aligning residue is Phe-199. As described below, a more general interaction between the triad of interest and loop A of the agonist-binding site is seen. Overall, these results highlight similarities and variations across members of the Cys-loop family.

Effects of exchanging triads between 5-HT 3A and the muscle-type nAChR
To evaluate the role of the triad of residues, we used sitedirected mutagenesis to introduce mutations into receptors expressed in X. laevis oocytes and measured functional responses via dose-response relationships using two-electrode voltage-clamp electrophysiology, as described under "Experimental procedures." We generated the 5-HT 3A R single mutants T152K, E209D, and K211T, (corresponding to the nAChR residues at sites A, B, and C, respectively) as well at the double mutants AB, BC, and AC and the triple mutant ABC. The dose-response curves measured for all of these mutant receptors are compared in Fig. 2A, and the full pharmacological characterization in response to the native agonist serotonin (5-hydroxytryptamine, or 5-HT) is provided in Table 1. Representative two-electrode voltage-clamp traces are provided in Fig. S1. Relative to the wildtype 5-HT 3A R, the single mutant A, view of two adjacent receptor subunits from the 5-HT 3A R crystal structure (PDB code 4PIR). The inset shows residues mutated in this study highlighted in yellow and binding-site residues of the aromatic box shown in green for context; red ϭ oxygen, blue ϭ nitrogen. B, sequence alignment of subunits contributing to the primary binding face of human-type 5-HT 3A Rs, nAChRs, and GABA A Rs. Cells are colored according to side-chain chemistry. Unfilled cells indicate non-conservation with regard to the triad in that receptor subtype. TMD, transmembrane domain; ICD, intracellular domain. *, Residue numbers corresponding to the 5-HT 3A R crystal structure (PDB code 4PIR). See "Experimental procedures" for sequence accession numbers.

Transferrable residues in serotonin and nicotinic receptors
receptors had EC 50 values for serotonin that were 39-, 10-, and 1.2-fold greater than wildtype for A, B, and C, respectively. Although the loss of function for receptors mutated at the A and B sites is clearly meaningful, the consequence of mutation at C is not. Note that structurally, however, the mutation at site B (Glu to Asp) would be expected to be the least perturbing.
Next, double mutants incorporating each of the above three mutations in pairs were generated to evaluate energetic coupling between the mutations via mutant cycle analyses. The extent of energetic coupling in a mutant cycle analysis is defined as the ⌬⌬G 0 value for the double mutant, and quantifies the extent to which the individual mutations cooperate in the activation mechanism of the receptor (20). If residues are close enough in space that a direct interaction is possible, the ⌬⌬G 0 value can also be used as a readout of the strength of this interaction.
The results of the double mutant cycles are provided in Table  1. Fig. 2B illustrates the non-additive effects of the mutations as a scatterplot. The AC double mutant had an EC 50 of only 2.5 Ϯ 0.05 M 5-HT, a mere 2-fold loss of function relative to wildtype (1.3 Ϯ 0.03 M) and a 19-fold gain of function relative to the single mutant at A (EC 50 ϭ 49 Ϯ 1.03 M). Thus, these two mutations are strongly energetically coupled, with ⌬⌬G 0 ϭ 1.8 kcal mol Ϫ1 . This and their close orientation in space suggest a

Transferrable residues in serotonin and nicotinic receptors
direct interaction between these two residues. The double mutant BC showed a much lower coupling energy of ⌬⌬G 0 ϭ 0.49 kcal mol Ϫ1 . We consider a 2-fold deviation from additivity (⌬⌬G 0 Ն 0.40 kcal mol Ϫ1 ) to be meaningful, suggesting that the B and C mutations do affect one another, albeit much less so than the AC double mutant. Fig. 2B compares the loss of function for each combination of mutations to wildtype. The double mutant AB was nonfunctional up to 800 M 5-HT. To meet our criterion for a meaningful coupling energy of Ն0.40 kcal mol Ϫ1 , the EC 50 would either have to be less than 200 M if the mutations are meaningfully subadditive or greater than 810 M 5-HT if they are superadditive. The effects of the A and B mutations are clearly not subadditive, as we observed no response to concentrations of 5-HT up to 800 M. It remains possible that the effects of these mutations are superadditive. If this were the case, we unfortunately could not measure a doseresponse relationship due to open channel block of 5-HT 3A Rs at high concentrations of 5-HT. It is also possible that combining the A and B mutations impairs proper folding or trafficking of receptors to the cell surface.
Having generated the single mutants and double mutants, we next generated the triple mutant 5-HT 3A R T152K/E209D/ K211T, "ABC," incorporating all three residues of the nicotinic triad. This receptor had only an 8.4-fold increase in 5-HT EC 50 relative to wildtype. Thus, the full three-residue swap of nAChR residues into the 5-HT 3A R produces functional receptors. In fact, EC 50 for the triple mutant is comparable with that for the single mutation at B and much less than the single mutation at A. The increase in EC 50 for the triple mutant is much less than would be expected if the effects of the A, B, and C mutations were additive (Fig. 2B).
Having the triple mutant enables a double mutant cycle between the A and B sites in the background of the C mutation. This mutant cycle has a ⌬⌬G 0 ϭ 0.04 kcal mol Ϫ1 ( Table 1), suggesting that A (T152K) and B (E209D) affect the receptor in distinct ways, at least in the background of K211T. The triple mutant also had somewhat different apparent activation/deactivation kinetics than wildtype receptors, and a greater extent of open channel block at concentrations of 5-HT above those that caused a maximal response (Fig. S2).
We next tested to see how a prototypical nAChR, the muscletype ((␣1) 2 ␤1␥␦) nAChR, would respond when the 5-HT 3A R triad was introduced. Although the most recent nAChR structural data are for the ␣4␤2 neuronal subtype (10), we were unable to perform the desired experiments in that subtype due to poor expression levels. This has been addressed previously (21) by introducing the channel pore mutation L9ЈA, however, this would complicate our analysis, as the residues of interest are involved in receptor gating. However, the muscle-type nAChR expresses robustly in Xenopus oocytes. The relevant residues are the same in the ␣1 (muscle-type) and ␣4 subunits, and they are positioned similarly in the cryo-EM structure of the highly homologous Torpedo receptor (PDB code 2BG9) (22). We generated single mutants ␣1 K145T, D200E, and T202K (A, B, and C), and the pharmacological responses of these mutants to the native agonist acetylcholine (ACh) are summarized in Table 2. Representative two-electrode voltageclamp traces are provided in Fig. S3, and comparisons of the dose-response relationships between the mutant receptors and wildtype can be found in Fig. S4. All mutations caused modest losses of function: 4.9-, 2.9-, and 9.0-fold for A, B, and C, respectively.
We next generated double mutant pairs in the nAChR ( Table  2) as described previously for the 5-HT 3A R. Interestingly, the ⌬⌬G 0 values for each of the double mutants AB, BC, and AC were all near 1.0 kcal mol Ϫ1 , indicating a similar degree of cooperativity between any two mutations. It is not surprising that mutations to Lys-145 and Asp-200 should be coupled, as previous work has proposed a salt bridge between these residues in the muscle-type nAChR (12). Indeed, analysis via patch clamp electrophysiology of the steady-state dissociation of ACh from the double mutant K145Q/D200N has been previously observed to give a very similar ⌬⌬G 0 value of 1.0 kcal mol Ϫ1 , and other measures suggest a strong interaction between the two (12). Our observation that both residues also couple to Thr-202 suggests interactions among all three.
The next logical step was to evaluate the functional response to ACh of the triple mutant: ␣1 K145T/D200E/T202K (ABC). The triple mutant receptor was functional, with EC 50 ϭ 67 Ϯ 5.3 M ACh, a 4.3-fold increase from wildtype ( Table 2). The data from the single, double, and triple mutants enables a triple mutant cycle analysis, which can be represented as a cube (Fig.  3). The difference between each opposite face gives a ⌬⌬⌬G 0 , which quantifies the interaction energy among all three residues as a unit (23). We find in each case that ⌬⌬⌬G 0 ϭ 0.95 kcal mol Ϫ1 . That is, any pair of mutations gives ⌬⌬G 0 Ϸ 0.95 kcal mol Ϫ1 , but in the background of any single mutation a double mutant cycle of the other two mutations yields ⌬⌬G 0 Ϸ 0 kcal mol Ϫ1 . Thus these residues not only interact with one another in the activation mechanism, but operate as a co-dependent unit in the mouse muscle-type nAChR. Each pair of residues is energetically dependent on the third residue of the triad.
To further evaluate the role of this triad, we mutated residues in the 5-HT 3A R to their equivalents in GABA A R ␤-subunits, which form the primary binding face in GABA A Rs. The results are summarized in Table S1. The single mutant at the A site (T152E) showed a 10-fold loss of function, whereas mutation at the C site (K211S) had no functional consequence. The single mutant E209R (B site), however, was nonfunctional up to 800 M 5-HT. Attempts to transplant the entire GABA A R triad into the 5-HT 3A R, T152E/E209R/K211S, yielded no observable response to concentrations of 5-HT as high as 800 M. Thus, the inhibitory GABA A R appears to behave differently than the excitatory 5-HT 3A R and nAChRs. It is also possible, however,

Transferrable residues in serotonin and nicotinic receptors
that the receptor mutants, which did not appear to respond to 5-HT, were not expressing or not reaching the cell surface.

Mutating 5-HT 3A R triad to that of the nAChR increases the binding affinity of nicotine
Having established that 5-HT 3A Rs incorporating the nAChRtype triad are functional, we next evaluated whether residues in the 5-HT 3A R triad contribute to the pharmacological properties of nicotine, an agonist of most neuronal nAChRs and a competitive antagonist of 5HT 3A Rs (1,24). To accomplish this, we again expressed mutant 5-HT 3A Rs incorporating the nicotinic-type mutations T152K (A), E209D (B), and K211T (C) as single, double, and triple mutants in X. laevis oocytes, and generated Schild plots for inhibition by nicotine for each receptor variant (25,26). In this procedure, dose-response relationships of each receptor variant were measured in the presence of different concentrations of nicotine. Higher concentrations of nicotine progressively increase the EC 50 of 5-HT as nicotine displaces 5-HT in the orthosteric binding site (Fig. 4). In a Schild plot, the x intercept gives a readout of the equilibrium dissociation constant (K d ) for the antagonist (25,26).
The Schild plots for each variant are compared with that of the wildtype 5-HT 3A R in Fig. 5, the parameters of the linear fits are provided in Table S2, and the full pharmacological characterization of the nicotine inhibition experiments is provided in Table S3. The single mutants T152K (A), E209D (B), and K211T (C) increased the K d of nicotine by 1.5-, 13-, and 2.5-fold, respectively.
Meaningful non-additive effects are observed when the mutations are combined in the double and triple mutants. Despite A and C each producing small increases in the K d of nicotine, the double mutant AC is 2.2-fold more sensitive to nicotine than wildtype. The double mutant BC is more sensitive to inhibition by nicotine than B on its own, even though C increases the K d of nicotine on its own. The triple mutant is the most nicotine-sensitive of all the variants, with a K d 5-fold lower than that of wildtype. Thus these three residues influence the binding of nicotine to the 5-HT 3A R. In particular, rearranging of Thr-152 (A-site) and Lys-211 (C-site) to their respective positions in nAChRs generates variants that are more sensitive to nicotine inhibition than wildtype. The mutation B, whereas causing a sizable decrease in nicotine affinity on its own, gives a further increase in nicotine sensitivity when combined with the A and C mutations.
This together suggests intimate cooperativity between these three residues in regard to nicotine binding. Mutant cycles of nicotine K d values (Fig. S5) illustrate meaningful coupling energies between A and C (1.1 kcal mol Ϫ1 ), B and C (0.60 kcal mol Ϫ1 ), and A and B in the background of C (1.2 kcal mol Ϫ1 ).
We attempted to measure inhibition by 5-HT of muscle-type nAChRs containing the 5-HT 3A R-type triad. However, the results suggested that 5-HT inhibits both wildtype and triple mutant muscle-type nAChRs via a noncompetitive mechanism, consistent with previous work (27). Maximal currents in the presence of 5-HT were decreased, but EC 50 values of ACh were unchanged (Fig. S6).

The 5-HT 3A R triad is functionally coupled to nearby residues and the channel gate
Recent structures of Cys-loop receptors show that the triad investigated here is near the agonist-binding site, but none of the residues can directly contact a bound agonist (9 -11, 15, 17, 28). The structures do suggest strong interactions between the triad and the canonical loops that position key residues that directly contribute to agonist binding (Fig. S7). It has been previously proposed in the muscle-type nAChR that a strong interaction exists between Lys-145 and Asp-200 (A-site and B-site) in the closed state of the receptor. In the open state this interaction was proposed to be weakened by an interaction between Lys-145 and the OH of Tyr-190, a loop C residue (often referred to as TyrC1) and a completely conserved member of the agonist-binding site aromatic box (12)(13)(14)(15)(16)(17). In the ␣4␤2 nAChR crystal structure (10), Asp-200 interacts with both Lys-145 and Thr-202 (the triad considered here), and the OH of Tyr-190 interacts with both Lys-145 and Asp-200. Recall the crystal structure most likely represents a desensitized state. Very similar interactions are seen in the crystal structure of the nAChR ␣2 ECD (28), crystal structures of acetylcholinebinding protein (17), and in a structure of the ␣1 subunit ECD (15).
In the 5-HT 3A R the residue that aligns with Tyr-190 is Phe-199, which is unable to make the hydrogen bonding interaction proposed for Tyr-190 of the nAChR. Instead, in the crystal structure the triad of the 5-HT 3A R makes no apparent contact with loop C, but interacts extensively with Asn-101 and Glu-102 (9), residues that are on the A loop of the agonist-binding site and that have been probed previously. Note that Asn-101 aligns with Tyr-93 (TyrA) of the nAChR, another contributor to the aromatic box (18,19). In particular, Thr-152 (A site) interacts with the side chain of Glu-102; Glu-209 (B site) interacts with the side chain of Asn-101; and there is also an inter-

Transferrable residues in serotonin and nicotinic receptors
action between the backbone carbonyl of the A site and the backbone NH of Asn-101. The backbone interactions between the A site and the A loop are also present in structures of the nAChR (10) and GABA A R (11), suggesting an interaction between the triad considered here and the A loop is a universal feature of Cys-loop receptors.

Transferrable residues in serotonin and nicotinic receptors
We have probed the interactions suggested by the structural work using mutagenesis. For these studies we typically used more conservative mutations than were used in the "triad swapping" experiments (Table 3). For residues Glu-102 and Glu-209, we employ mutations to glutamine and/or aspartate; the E102D mutation was not examined as it has been observed previously that this mutation has little effect on the response of receptors to 5-HT (23). We employ a more drastic mutation at Asn-101, N101K. This was motivated by previous work that demonstrated a high loss of function for the N101K mutation (29). For mutant cycles with Thr-152, we use the conservative T152V mutation. Table 3 provides the full characterization of the pharmacological response of these receptors to 5-HT. There is extensive cross-talk between the triad and the residues on loop A. All triad residues couple to Asn-101, and Lys-211 couples to Glu-102. Fig. 6 provides a visual depiction of the largest coupling energies observed for pairs of mutations at the sites shown, including those discussed previously and those discussed hereafter. The ⌬⌬G 0 values for different mutant cycles in Fig. 6 are grouped into five categories, each representing an additional 2-fold deviation from additivity.
Previous work in the 5-HT 3A R on loop A residues Asn-101 and Glu-102 showed that mutations to these residues that introduce a net positive charge (N101K and E102Q) generate very large losses of function (23,29). This parallels our observation on the T152K mutation, which is the most disruptive of the single mutations in the triad swap. We also observed a large

Transferrable residues in serotonin and nicotinic receptors
loss of function for E209Q, which likewise introduces a net positive charge. In contrast, both K211T and K211M, mutations that remove a positive charge, have minimal impacts on function, but K211T rescued function of receptors with T152K.
We also generated the double mutants around the mutation K211M, the reasoning being that the nonpolar side chain of methionine is sterically similar to the lysine side chain, and does not introduce any other functionalities as in the case of the threonine hydroxyl in K211T. The ⌬⌬G 0 values observed for the double mutants N101K/K211M, E102Q/K211M, T152K/ K211M, and E209Q/K211M were 0.91, 1.7, 0.61, and 1.1 kcal mol Ϫ1 , respectively (Table 3). We also generated the double mutant T152V/K211M, for which the ⌬⌬G 0 value was only 0.05 kcal mol Ϫ1 (Table 3), demonstrating that a positive charge must be introduced at Thr-152 to couple to Lys-211. Full dose-response curves for mutant cycles examining the role of positive charge among these residues are provided in Fig. S8.
Interestingly, each of the mutations that introduce a net positive charge couple comparably to K211T and K211M, with the exception of T152K. This mutation couples strongly to K211T (⌬⌬G 0 ϭ 1.8 kcal/mol) but more weakly to K211M (⌬⌬G 0 ϭ 0.61 kcal/mol). That is to say, the K211T mutation rescues function of T152K more than eight times better than K211M does (Fig. 7D). We therefore argue that there is a direct interaction between Thr-152 and Lys-211 in the 5-HT 3A R, the likes of which cannot be accounted for by merely eliminating the nearby positive charge of the Lys-211 side chain. On the other hand, the deleterious effects of introducing a net positive charge either by introducing N101K, E102Q, or E209Q, are attenuated whenever the basic amine on Lys-211 is removed, regardless of the ability of the residue a position 211 to hydrogen bond.
Previously, we proposed a strategy for evaluating the role of a particular residue in channel function by evaluating its coupling to a residue in the pore-lining M2 helix that is clearly involved in channel gating (31). In the nAChR, that residue is termed Leu-9Ј (the 9th residue from the bottom of the M2 helix). In the Table 3 Functional   Tables  1 and 3.

Transferrable residues in serotonin and nicotinic receptors
nAChR we found that D200N couples strongly to L9ЈA (⌬⌬G ϭ 1.3 kcal/mol), establishing a role in gating for Asp-200 (32). However, K145Q and T202Tah (Tah ϭ ␣-hydroxy threonine, a backbone mutation) do not show strong coupling to L9ЈA (32). In the 5-HT 3A R, the appropriate mutation is T6ЈS (T257S in our numbering scheme), which has been used extensively to modulate serotonin receptor currents (33). We looked for coupling to T6ЈS at Asn-101, Glu-102, Thr-152, and Glu-209. The results are summarized in Table 3. We find that N101K, E102Q, T152V, and T152K are all meaningfully coupled to T6ЈS, suggesting a role in gating for these residues (32). Interestingly, E209D and E209Q do not show meaningful coupling to T6ЈS. Thus, the nAChR and 5-HT 3A R show different coupling behaviors to the pore-forming M2 helix. In the nAChR triad, site B couples but sites A and C do not. In the 5-HT 3A R site A couples and site B does not. We are unable to perform the same analysis for mutations to site C, as none of the mutations we have studied generates an interpretable loss of function relative to wild-type. However, we do note that mutations to site C couple energetically to all of the residues that couple to T6ЈS.
Attempted expression of the 5-HT 3A R double mutants E102Q/T152V, E102Q/E209Q, N101K/E209Q, and T152K/ N101K yielded no measurable response to high concentrations of 5-HT. This may result either from nonfunctional receptors or a lack of surface expression. Especially for the less-conservative mutations we employ, individual mutations likely affect multiple interactions involved in receptor function, the deleterious effects of which may sometimes add together even if other pairs of mutations at the same sites generate functional receptors.

Discussion
We have identified a triad of residues with one cationic, one anionic, and one hydroxyl-containing side chain that is rearranged between the 5-HT 3A R, nAChRs, and GABA A Rs, and behaves differently in different receptors. In the muscle-type nAChR, the three triad residues cooperate as a unit, as demon-  Table 3. Dose-response curves provided in Fig. S8. Error bars represent S.D.

Transferrable residues in serotonin and nicotinic receptors
strated via triple mutant cycle analysis. This corroborates previous functional and crystallographic work that suggested a salt bridge between two members of the triad is involved in receptor gating (10,12,15,17,22,28). We show here that the salt bridge between Lys-145 and Asp-200 depends on Thr-202, which may coordinate this interaction. Mutant cycles of the aligning residues in the 5-HT 3A R indicate that Thr-152, Glu-209, and Lys-211 cooperate in the activation mechanism of the 5-HT 3A R, but not in a perfectly codependent manner. Interestingly, the triads of the 5-HT 3A R and the nAChR can be swapped and still produce functional receptors with minimal loss of function.
The triad of residues in the 5-HT 3A R affects the competitive binding of nicotine for the orthosteric site. Each mutation on its own causes a decrease in nicotine binding affinity, although the shift for T152K is very small. Interestingly, however, receptor variants with both the T152K (A) and K211T (C) mutations saw increases in the ability of nicotine to compete with 5-HT relative to wildtype receptors. Moreover, E209D (B), which causes a large loss of nicotine binding affinity on its own, increases the binding affinity for the triple mutant ABC relative to double mutant AC. These three residues thus cooperate in the binding of nicotine, and likely represent a characteristic network of nAChRs. Worth noting is that the only deviations from the cationic/anion/hydroxyl triad of nAChRs are seen in the ␣9 and ␣10 nAChRs, which are inhibited rather than activated by nicotine (34). It has been demonstrated elsewhere, however, that mutating the ␣9 Thr to a Lys does not produce receptors that are activated by nicotine (35), so there is more to it than this one residue.
The triad from the GABA A R does not appear to function in the 5-HT 3A R, suggesting this network functions differently in GABA A Rs. Other work has proposed a functionally important salt bridge between residues in the GABA A R aligning to Lys-197 and Thr-152 in the 5-HT 3A R (30), however, we do not observe meaningful functional coupling between mutations at these sites in the 5-HT 3A R. We also note that the pattern of functionalities in the triad we have studied is not conserved in primary-face ␣-subunits of inhibitory GlyRs. All this together suggests that the triad of residues behaves similarly in excitatory receptors, but is functionally distinct in inhibitory Cysloop receptors.
Many amino acid residues that we have considered in this study have been examined previously in regard to ligand binding. Based on radioligand binding studies of the 5-HT 3A R antagonist [ 3 H]granisetron, it has been proposed that Thr-152 helps to shape the binding site (36). Mutations to 5-HT 3A R residues Glu-209 and Lys-211 have been shown previously to have minimal effect on binding of [ 3 H]granisetron (29,36). Our results show a minimal effect on nicotine binding for T152K, a modest effect for K211T, and a sizable effect for E209D. Thus these results highlight different binding modes for different competitive antagonists.
We further investigated the involvement of residues in the gating mechanism via mutant cycle analyses to the channel gate mutation T6ЈS. Loop A residue Asn-101 couples to T6ЈS, supporting a previous proposition that this residue is involved in receptor gating (29). The mutation E102Q couples to T6ЈS, in line with a previously published result (23). We observe that the T152K mutation affects ligand binding and also couples to the T6ЈS; thus we propose that this residue is involved in both binding and gating. Quizzically, mutant cycles between Glu-209 and T6ЈS do not reveal functional coupling between these residues, despite the previously published observation that Glu-209 does not participate in the binding of [ 3 H]granisetron (36). It is possible that Glu-209 plays different roles in response to 5-HT versus granisetron binding, or that Glu-209 participates in a functionally relevant pathway that is not linked to the channel gate at Thr-6Ј. We have not identified any single mutation to Lys-211 that generates interpretable shifts in 5-HT EC 50 , so we were unable to perform the same analysis on this residue. However, spatially proximal residues that couple to the channel gate (Asn-101, Glu-102, and Thr-152) also couple to Lys-211.
Mutations that introduce a net positive charge in the region of the 5-HT 3A R probed here generally increase the 5-HT EC 50 by a substantial amount. Most of these large losses of function could be rescued by mutations that remove the side chain amine of Lys-211. An exception is the highly deleterious T152K mutation, which can be nearly fully rescued by the K211T mutation, corresponding to the nAChR swap, but to a much lesser extent by the K211M mutation. This argues for a highly specific interaction between Thr-152 and Lys-211 in the native receptor.
We have demonstrated via mutant cycle analyses that triad residues and loop A residues are functionally coupled in the 5-HT 3A R. Structural studies suggest similar interactions between the triad and loop A in nAChR. However, a previously established interaction between triad residues and loop C in nAChRs does not carry over to the 5-HT 3A R. These results highlight the fact that there is considerable variation in the degree to which structurally/functionally important features are or are not conserved across members of the Cys-loop family.

Molecular biology
The cDNA for the mouse 5-HT 3 R A subunit (m5HT 3A ) was in the pGEMhe vector. The cDNA for the mouse nAChR ␣1, ␤1, ␥, and ␦ were in the pAMV vector. The ␣1 subunit contains a hemagglutinin epitope tag in the M3-M4 loop, which does not alter the EC 50 of ACh. Site-directed mutagenesis was performed using a QuikChange mutagenesis kit (Stratagene). cDNA was linearized using XhoI (New England Biolabs) for m5-HT 3A , XbaI (New England Biolabs) for m␣1, or NotI (New England Biolabs) for m␤1, m␥, and m␦ subunits. Linearized DNA was purified using a Qiaquick PCR Purification Kit (Qiagen) before performing an in vitro runoff transcription using a T7 mMessage mMachine kit (Ambion). The mRNA was purified using an RNEasy Mini Kit (Qiagen) and concentrations were quantified via UV-visible spectroscopy (Nanodrop 2000, ThermoFisher Scientific). cDNA was stored at Ϫ20°C and mRNA was stored at Ϫ80°C.

Protein expression in X. laevis oocytes
Stage V-VI X. laevis oocytes were prepared as previously described (37). Oocytes were injected with 50 nl of mRNA in nuclease-free water, and expression levels of different mutants

Transferrable residues in serotonin and nicotinic receptors
were optimized by varying mRNA concentrations and incubation times. For experiments involving 5-HT 3A R variants, oocytes were injected with 5-50 ng of mRNA once at 24 h before, once at 48 h before, or twice at 24 and 48 h before electrophysiological recording was performed. For experiments involving muscle-type nAChRs, oocytes were injected with a mixture of ␣1, ␤1, ␥, and ␦ mRNAs in a 10:5:5:5 ng ratio (per 50 nl) once at 24 h before, once at 48 h before, or twice at 24 and 48 h before electrophysiological recording was performed. Cells were incubated at 18°C in ND96 (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 , and 5 mM HEPES at pH 7.5) with 0.05 mg/ml of gentamycin (Sigma), 2.5 mM sodium pyruvate (Acros Organics), and 0.67 mM theophylline (Sigma).

Whole cell electrophysiological recording
Electrophysiology and drug perfusion were performed at ambient room temperature (20 -25°C) using the OpusXpress 6000A (Axon Instruments) in two-electrode voltage-clamp mode. Oocytes were impaled with borosilicate glass pipettes filled with 3 M KCl (r ϭ 0.3-3.0 megaohm) and clamped at a holding potential of Ϫ60 mV. Data were sampled at 125 Hz. The running buffer was Ca 2ϩ -free ND96 (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES at pH 7.5). Solutions of serotonin hydrochloride (Sigma), acetylcholine chloride (Sigma), and (Ϫ)-nicotine tartrate (Sigma) were prepared 0 -24 h before recording from stock solutions stored at Ϫ80°C.
For dose-response experiments, 1 ml of each concentration of drug in Ca 2ϩ -free ND96 was applied for 15 s followed by a 116-s wash using Ca 2ϩ -free ND96 before subsequent doses. For slow-activating variants, 1 ml of drug solution was applied for 15 s and cells were incubated in the drug solution for an additional 15 s before washout. For slow-deactivating variants, washout times were increased to 176 s between doses. The different procedure for slow activating/deactivating variants did not meaningfully change the overall dose-response relationships recorded versus the standard 15-s application/116-s wash procedure.
For Schild analysis experiments, cells were continuously perfused with Ca 2ϩ -free ND96 containing (Ϫ)-nicotine, and 1-ml solutions of 5-HT in Ca 2ϩ -free ND96 with the requisite concentration of (Ϫ)-nicotine were applied for 15 s/dose, followed by 116-s washes between doses. As in the dose-response experiments, application times and wash times were adjusted for variants displaying slow activation/deactivation.

Data analysis
Raw two-electrode voltage-clamp traces were prepared and analyzed in Clampfit 10.3 (Axon Instruments). Raw data were first filtered using a low pass Gaussian filter at 1 Hz. A 30-s baseline was established prior to each drug application during which cells were continuously perfused with Ca 2ϩ -free ND96. The averaged current of the baseline was subtracted from the peak amplitude following each drug application to generate dose-response data.
Dose-response data for individual concentrations were averaged, plotted, and fit to the Hill equation I/I max ϭ 1/(1 ϩ (EC 50 / [agonist] nH ) in Kaleidagraph 3.6 (Synergy), where EC 50 is the concentration for a half-maximal response, n H is the Hill coef-ficient, and I/I max is the normalized response at a given drug concentration. For dose-response experiments, data were normalized to the maximum current observed.
Dose-response curves shown in figures were generated in Prism 7 (GraphPad). The linear fit for the Schild plots were determined in Prism 7, whereas the figure itself was made in Microsoft Excel 2013. The values for the slopes are reported in the text as slope fit Ϯ S.E.
Coupling energies for double-mutant cycles were calculated using the formula ⌬⌬G ϭ ϪR ϫ T ϫ ln((EC 50wt ϫ EC 50AB )/ (EC 50A ϫ EC 50B )), where R is the gas constant, T is temperature, A and B denote individual mutations, and AB denotes double mutant receptors. The value used for T was 293.15 K.
Author contributions-R. M. and D. A. D. designed the study and prepared the manuscript. All laboratory work was conducted by R. M. Both R. M. and D. A. D. analyzed the results and approved the final version of the manuscript.