Heterologous expression of concatenated nicotinic ACh receptors: Pros and cons of subunit concatenation and recommendations for construct designs

Abstract

Background and Purpose

Concatenation of Cys‐loop receptor subunits is a commonly used technique to ensure experimental control of receptor assembly. However, we recently demonstrated that widely used constructs did not lead to the expression of uniform pools of ternary and more complex receptors. The aim was therefore to identify viable strategies for designing concatenated constructs that would allow strict control of resultant receptor pools.

Experimental Approach

Concatenated dimeric, tetrameric, and pentameric α4β2‐containing nicotinic ACh (nACh) receptor constructs were designed with successively shorter linker lengths and expressed in Xenopus laevis oocytes. Resulting receptor stoichiometries were investigated by functional analysis in two‐electrode voltage‐clamp experiments. Molecular dynamics simulations were performed to investigate potential effects of linkers on the 3D structure of concatemers.

Key Results

Dimeric constructs were found to be unreliable and should be avoided for expression of ternary receptors. By introducing two short linkers, we obtained efficient expression of uniform receptor pools with tetrameric and pentameric constructs. However, linkers should not be excessively short as that introduces strain on the 3D structure of concatemers.

Conclusion and Implications

The data demonstrate that design of concatenated Cys‐loop receptors requires a compromise between the desire for control of assembly and avoiding introduction of strain on the resulting protein. The overall best strategy was found to be pentameric constructs with carefully optimised linker lengths. Our findings will advance studies of ternary or more complex Cys‐loop receptors as well as enabling detailed analysis of how pharmacological agents interact with stoichiometry‐specific binding sites.

Abbreviations

CRR

concentration–response relationship

nAChR

nicotinic ACh receptor

TLL

total linker length

What is already known

• Ternary or more complex Cys‐loop receptors are notoriously challenging to express in heterologous expression systems.

• Subunit concatenation can theoretically overcome these challenges; however, there are caveats associated with this technology.

What this study adds

• Expression of uniform ternary receptor pools was only obtained with optimised tetrameric and pentameric constructs.

• The guidelines presented here allow expression of complex Cys‐loop receptors with confidence in the stoichiometry.

What is the clinical significance

• Our results will advance studies of pharmacological agents that interact with Cys‐loop receptors.

• The technology additionally enables functional studies of disease‐causing variants of rare Cys‐loop receptor subtypes.

1. INTRODUCTION

Nicotinic ACh receptors (nAChRs) consist of five subunits encircling a central ion channel. While identical subunits can assemble to form homomeric receptors, it is more common for combinations of different subunits to form heteromeric receptors, referred to as subtypes. Evidence so far suggests the physiological existence of binary, ternary, and quaternary receptors, that is, receptors made up of two, three, or four different types of subunits respectively. Furthermore, some receptor subtypes, for example, α4 β2, can exist in two stoichiometric forms (α4)₃(β2)₂ and (α4)₂(β2)₃ with different subunit ratio (Harpsoe et al., 2011; Zhou et al., 2003). Considering that 16mammalian nAChR subunits (α1–7, α9–10, β1–4, γ, δ, ε) have been cloned to date, the number of potential receptor subtypes is staggering (Millar & Gotti, 2009).

Pinpointing the precise stoichiometry of heteromeric nAChRs is therefore a prerequisite for the accurate delineation of receptor function and the understanding of their pharmacology. This is elegantly exemplified by the unique selectivity of the compound NS9283, which acts to enhance submaximal ACh‐evoked currents at (α4)₃(β2)₂ receptors but shows no activity at the alternative (α4)₂(β2)₃ receptors (Olsen, Ahring, Kastrup, Gajhede, & Balle, 2014; Olsen et al., 2013; Timmermann et al., 2012). As many ligand binding sites reside in interfaces between subunits, the possibility for specific binding pockets increases along with the complexity of a receptor.

To define subunit stoichiometry and arrangement of heterologously expressed Cys‐loop receptors, researchers have relied on the concatenation of subunits. This technique fuses subunit cDNAs with a synthetic linker sequence such that multiple subunits are expressed as a fusion protein. Successful concatenation of nAChRs was first described by Im, Pregenzer, Binder, Dillon, and Alberts (1995), and systematic refinements for both nACh and GABA_A receptors emerged in the following decade (Baumann, Baur, & Sigel, 2001; Baur, Minier, & Sigel, 2006; Groot‐Kormelink, Broadbent, Beato, & Sivilotti, 2006; Zhou et al., 2003). As a result, concatenated constructs with anything from two to five subunits (dimeric to pentameric) became available. Based on a number of studies, the current consensus for construct design is to omit signal peptides from all downstream subunits and to use linkers that primarily consist of relatively small and inert amino acids such as alanine, serine, glycine, and glutamine. This approach avoids potential issues of protein degradation and strong direct interactions between linker and receptor. By following this approach, robust expression levels of stable concatemers in heterologous expression systems such as Xenopus laevis oocytes and HEK293 cells are readily obtained (Absalom et al., 2019; Carbone, Moroni, Groot‐Kormelink, & Bermudez, 2009; Jin & Steinbach, 2011; Kuryatov & Lindstrom, 2011; Prevost, Bouchenaki, Barilone, Gielen, & Corringer, 2020; Zhou et al., 2003).

While the use of concatenated constructs is a powerful technique, there are potential caveats which could affect experimental outcomes. This was investigated by Ericksen and Boileau (2007) for nAChRs and Sigel, Kaur, Luscher, and Baur (2009) for GABA_A receptors. However, the serious outcomes of these caveats have largely been overlooked. Recently, we performed an in‐depth analysis of commonly utilised concatenated nACh and GABA_A receptor constructs and found that these often do not lead to the expected receptors (Ahring, Liao, & Balle, 2018; Liao et al., 2019). This was due to a high degree of freedom for the concatemers, which enabled them to assemble in both the clockwise and counterclockwise orientations or fail in integrating all concatemer subunits in the final pentameric receptor.

In this study, we therefore aimed to develop a robust concatenation strategy that can serve as a guide to obtaining uniform receptor pools in future studies of ternary and more complex Cys‐loop receptors. Ternary α4β2‐containing receptors were used as a model system, and to enable flexibility in experimental design, new dimeric, tetrameric, and pentameric constructs were investigated. To enforce unidirectional counterclockwise assembly of concatemers, optimised linkers with assembly‐directing properties were introduced in either one or two construct positions. While robust expression of uniform ternary receptor pools has not yet been documented using concatenated constructs before, we discovered that this is achievable by use of two short assembly‐directing linkers in tetrameric and pentameric constructs. The success of these efforts paves the way for future studies of unique receptor subtypes but also highlights that many commonly used construct designs do not lead to the expected outcomes. Thus, great care should be taken when assessing historical data obtained using concatenated constructs with long linkers.

2. METHODS

2.1. Molecular biology

Human cDNA encoding monomeric α4, β2, and α4^VFL nAChR subunits were kind gifts from Saniona A/S, Copenhagen, Denmark. Design of existing concatenated constructs is described in detail in Ahring et al. (2018), and the new concatenated constructs were designed from wild‐type β2 and α4 sequences in an identical fashion. In brief, nAChR subunits were constructed to contain AGS‐based linker sequences with unique restriction sites. Signal peptide sequences were omitted for all but the first subunit within a concatemer. Standard PCR reactions with β2 or α4 as template were performed using Q5 polymerase, and amplified products were cloned into in‐house vectors based on pcDNA3 (Mirza et al., 2008). Translated C‐terminal, linker, and N‐terminal sequences were as follows: β2^A‐3a‐α4^B (…APSSKAGSAHAEE…), β2^A‐2a‐α4^B (…APSSKGSAHAEE…), β2^A‐1a‐α4^B (…SAPSSGSAHAEE…), β2^A‐0a‐α4^B (…HSAPSGSAHAEE…), α4^B‐33a‐β2^C (…LAGMI (AGS) ₅ LGS(AGS) ₅TDTEE…), β2^C‐18a‐α4^D (…APSSK (AGS) ₂ AGT(AGS) ₃AHAEE…), β2^C‐3a‐α4^D (…APSSKAGTAHAEE…), β2^C‐2a‐α4^D (…APSSKGTAHAEE…), β2^C‐1a‐α4^D (…SAPSSGTAHAEE…), and α4^D‐33a‐α4^E (…LAGMI (AGS) ₄ ATG(AGS) ₆AHAEE…). The superscript letters A–E signify the relative position of a given subunit within a concatemer. Linker numbers, for example, 0a–3a, denote the number of amino acids introduced during the linking process, which for 1a and 0a linkers entailed deleting existing C‐terminal β2 amino acids to allow introduction of a restriction site. The total linker length (TLL) is calculated based on the methodology described in Ahring et al. (2018). In brief, this method includes already existing as well as artificially introduced amino acids as parts of the total linker. TLL is calculated from the last amino acid in transmembrane segment 4 (TM4) of the upstream subunit through the artificially introduced parts to a leucine anchor in α‐helix‐1 segment in the downstream subunit (Leu⁴⁰ for α4 in Figure 5a). Based on such calculations, TLLs for the short 0a–3a linkers vary from 29 to 32 amino acids, whereas the TLLs for the remaining linkers are 47 amino acids. Correct introduction of linker sequences and fidelity of all coding sequences were verified by double stranded sequencing. Final concatemers were assembled by standard restriction digestion and ligation. Escherichia coli 10‐β bacteria were used as hosts for plasmid expansion, and plasmid purifications were performed with standard kits (Qiagen). cRNA was produced from linearised cDNA using the mMessage mMachine T7 Transcription kit according to the manufacturer's description (ThermoFisher Scientific, Massachusetts, USA) and stored at −20°C until use.

FIGURE 5.

Simulation models containing the β2‐9a‐α4 clockwise or counterclockwise concatemer were based on the (α4)₃(β2)₂ nAChR structure (Walsh et al., 2018) with full‐length sequence numbering for amino acids. The 9a linker was gradually shortened over numerous simulation windows, to a length of 7a, 5a, 3a, 2a, 1a, and 0a. (a) Sequences in the TM4 and N‐terminal helical regions where steric energy was analysed during the simulations are framed. (b) Overlay of initial clockwise and counterclockwise concatemer models. Subunits are coloured blue for β2, green for α4, and red for the flexible part of the total linker region (β2 C‐terminal tail and artificially introduced AGS linker sequence). All other subunits have been coloured transparent white for clarity. (c and d) RMSD plots for Cα atoms from extracellular sheets and TM helices (excluding the TM4 helix connected to the linker) where lighter grey and brown colours depict shorter linkers, respectively. The plots cover 150‐ns simulation at constant 310‐K temperature following simulated annealing. (e–g) Internal steric energy normalised by the number of residues in the TM4 and N‐terminal helical selections (23 and 12 residues, respectively) and their total are shown as means with 95% confidence intervals. (h–m) Structures from the final frame of each simulation window superimposed on the original 9a model constructs. Structures with clockwise linkers are positioned in the top panels (h–j) and the counterclockwise linkers in the bottom panels (k–m). Purple, teal, and orange colours depict the concatemer TM4, N‐terminus, and linker region in between, respectively, with lighter colours representing the shift from 9a to 0a. Note, for (i) and (l), a loop segment (α4 Lys⁹⁷‐Ser¹¹⁰) has been replaced with an outline for visual clarity

2.2. Electrophysiology

Xenopus laevis oocytes were obtained and prepared as previously described (Mirza et al., 2008). Briefly, ovarian lobes were removed from anaesthetised adult female Xenopus laevis frogs following a protocol approved by the Animal Ethics Committee of The University of Sydney (reference number: 2013/5915). To obtain isolated oocytes, lobes were sliced into small pieces using a surgical knife and defolliculated by collagenase treatment. Stage V and VI oocytes were injected with ~50 nl of a 0.5 μg·μl⁻¹ cRNA mixture encoding the desired nAChR subunits and incubated for 2–5 days (unless otherwise noted) at 18°C in modified Barth's solution (96 mM NaCl, 2.0 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, 5 mM HEPES, 2.5 mM sodium pyruvate, 0.5 mM theophylline, and 100 μg·ml⁻¹ of gentamicin, pH 7.4).

Electrophysiological recordings using the two‐electrode voltage‐clamp technique were performed as previously described (Ahring et al., 2018; Kowal et al., 2018; Mirza et al., 2008). Briefly, oocytes were placed in a custom‐designed recording chamber and continuously perfused with a saline solution termed ND96 (96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, and 10 mM HEPES, pH 7.4). Pipettes were backfilled with 3‐M KCl, and open pipette resistances ranged from 0.3 to 2 MΩ when submerged in ND96 solution. Oocytes were voltage clamped at a holding potential of −60 mV using an Axon Geneclamp 500B amplifier (Molecular Devices, LLC, Sunnyvale, CA, USA). Oocytes with initial leak currents exceeding 200 nA when clamped were discarded. Amplified currents were filtered at 10 Hz by a four‐pole low‐pass Bessel filter (Axon Geneclamp 500B), digitised by a Digidata 1440A (Molecular Devices), and sampled at 200 Hz as well as analysed on a PC using the pClamp 10.2 suite (Molecular Devices; pClamp, RRID:SCR_011323). Responses to individual applications were collected as episodic traces following triggering events. Each episodic trace contains recording of (i) initial baseline for ~15 s; (ii) application of ACh or co‐application of ACh with NS9283 for 25 s; and (iii) washout for ~30 s.

2.3. Compound preparation and application system

ACh was dissolved as a 3.16‐M stock solution in ultrapure water. NS9283 was dissolved as a 100 mM stock solution in DMSO, which upon final dilution gave a maximal concentration of 0.1%. This DMSO concentration did not evoke any measurable currents from wild‐type α4β2 receptors. Fresh ACh and NS9283 dilutions were prepared in ND96 on the day of the experiment. To obtain precise control of the solution exchange in the immediate oocyte vicinity, a 1.5 mM‐inner‐diameter capillary tube was placed approximately 2 mm from the oocyte. By way of this capillary, oocytes were continuously perfused with ND96 or test applications (ND96 with ACh or ACh mixed with NS9283) at a constant flow rate of 2.0 ml·min⁻¹. Low volume Teflon tubing (inner diameter of 0.5 mm) connected the capillary to a Gilson 231XL autosampler (Gilson, Middleton, WI, USA), and all applications as well as triggering events were controlled using the Gilson 735 software suite. Between applications, the 2.0 ml·min⁻¹ solution flow rate was maintained using a Gilson Minipuls 3 pump whereas test applications were applied by means of a Gilson 402 syringe pump. This application methodology has two advantages: (i) it ensures rapid solution exchange (order of a few seconds) in the immediate oocyte vicinity both when switching from ND96 to test application and when switching back and (ii) it ensures identical and constant flow rate during test applications. The washout time interval between triggered recording events was from 2 to 5 min depending on the applied ACh concentration.

2.4. Experimental protocols

A complete experiment from a single oocyte contains all initial control applications as well as the full planned concentration–response relationship (CRR) for ACh or NS9283. To ensure reproducibility of evoked current amplitudes, a set of initial control applications was performed prior to the actual CRR applications. These were three ACh_control (10 μM) applications, one ACh_max (3,160 μM) application, three ACh_control (10 μM) applications, and finally, a buffer (no ACh) application. Thereafter followed six to seven applications of NS9283 co‐applied with ACh_control (10 μM) or ACh alone in increasing concentrations. Final datasets for NS9283 and ACh were assembled from a minimum of n = 5 experiments conducted on a minimum of two batches of oocytes. A batch of oocytes originated from one Xenopus laevis frog and was used for maximally 1 week. As fresh compound serial dilutions were prepared each day, it follows that final datasets were obtained from experiments at independent oocytes from a minimum of two frogs and two independent compound preparations. Unless otherwise stated, single data points in a dataset were not excluded from calculations. However, all experiment data from an individual oocyte were omitted in cases where (i) the initial control applications did not result in reproducible ACh‐evoked responses or (ii) the full planned CRR was not successfully obtained (incomplete experiment).

2.5. Data analysis

Two‐electrode voltage‐clamp data were analysed using Clampex 10.2 (Molecular Devices). During analysis, all episodic traces from an experiment were overlain and baseline subtracted simultaneously. Responses to applications were next quantified from all traces simultaneously as peak current amplitudes between a cursor pair covering the application time. Hence, data were quantified from raw traces in a blinded manner in the sense that baseline subtraction and identification of the response to applications were performed automatically by Clampex 10.2 following simple user setting of a cursor pair.

For experiments with ACh, peak current amplitudes (I) of full CRRs were fitted to the Hill equation by non‐linear least‐sum‐of‐squares regression and normalised to the maximal fitted response (I_{max_fit_ACh}) for each individual oocyte (i.e., I/I_{max_fit_ACh}). For experiments with NS9283, the compound was co‐applied with ACh_control (10 μM). Differences between ACh_control‐evoked current amplitudes in absence or presence of NS9283 (I) were calculated as the percentage change from the ACh_control‐evoked current (i.e., ((I − I_{ACh_control}) × 100)/I_{ACh_control}). Data for all individual experiments were then averaged. The averaged CRRs were fitted by non‐linear regression in GraphPad Prism 8 (GraphPad Prism, RRID:SCR_002798) to a monophasic (first order) or biphasic (second order) equation with constrained Hill slope(s) of 1 and efficacy at infinitesimal compound concentrations set to 0, unless otherwise specified. Comparison of best approximation (monophasic vs. biphasic) was carried out using the F‐test in GraphPad Prism 8 with P < 0.01 as significance level. The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018).

2.6. Experimental strategy

In‐depth description of the rationale behind the experimental designs can be found in Ahring et al. (2018), and this section outlines only the key points.

Wild‐type (α4)₃(β2)₂ nAChRs display biphasic ACh CRRs, meaning that the data are best approximated by a second‐order equation as revealed by, for example, an F‐test (Harpsoe et al., 2011; Mazzaferro et al., 2011). The first component reflects ACh binding and activation via two high‐affinity α4(+)–β2(−) interface binding sites, and the second component reflects the additional activation of the same receptors via ACh binding to the low‐affinity α4(+)–α4(−) interface site (Indurthi et al., 2016).

NS9283 was originally identified as an allosteric modulator of α4β2 receptors but later found to have site‐selective agonistic actions (Olsen et al., 2014; Olsen et al., 2013; Timmermann et al., 2012). Essentially, NS9283 binds in the ACh binding pocket of the α4(+)–α4(−) interface where it largely interacts with the same amino acids as ACh. Thus, at non‐saturating ACh concentrations, NS9283 binding to the α4(+)–α4(−) interface increase receptor activation of (α4)₃(β2)₂ receptors (Figure 1, construct I). Given that the efficacy of NS9283 is measured by co‐application with a submaximal ACh_control (10 μM, ~EC₁₅) concentration, there is substantial natural variation between responses from randomly picked oocytes. We have chosen to define average values that exceed 300% as full efficacy responses (green territory in Figure 1).

FIGURE 1.

Summary of published NS9283 efficacy observations at nAChRs in various binary and ternary stoichiometries of α4, β2, and α4^VFL subunits. Xenopus laevis oocytes were injected with cRNA mixtures of α4 and β2 or α4^VFL and β2 subunits in a 10:1 ratio or pentameric concatenated constructs thereof and subjected to two‐electrode voltage‐clamp electrophysiology. Efficacy of NS9283 was determined by co‐applications with ACh_control (10 μM). Data were obtained from n = 9–16 experiments, and complete information for these experiments is found in Timmermann et al. (2012) and Ahring et al. (2018). In wild‐type (α4)₃(β2)₂ nAChRs (construct I), NS9283 (31.6 μM) display efficacy of ~800% which we termed as full efficacy. Given that efficacy values are calculated as ratios, natural variations in the ACh CRRs between oocytes can lead to substantial variation in the readout. Hence, we chose to define full efficacy as anything within a 301–1,200% range (light green range). At (α4^VFL)₃(β2)₂ receptors (construct II), NS9283 display no positive efficacy as the binding site is destroyed in the α4^VFL–α4^VFL site. We chose to define no‐responses as efficacy values below a noise threshold of 10% (light red range). For mixed 3α:2β stoichiometries of α4 and α4^VFL and β2 subunits, the three combinations that contain an α4^VFL subunit in the complementary position of the α4^?–α4^VFL interface likewise display no positive NS9283 efficacy (exemplified by construct III). Mixtures of NS9283‐sensitive and ‐insensitive receptors as well as 3α:2β stoichiometries of α4 and α4^VFL and β2 subunits that contain an α4 subunit in the complementary position of the α4^?–α4 site (exemplified by construct IV) display intermediate NS9283 responses. We chose to define intermediate efficacy as values within a 10–300% range (light blue range). Schematic illustrations of receptor stoichiometries that correspond to the data are shown on the right. α4, β2, and α4^VFL subunits are shown as green, blue, and red circles, respectively, when viewed from the extracellular space. Principal and complementary subunit faces are indicated with “+” and “−,” respectively. Triangular symbols signify the three different binding sites for ACh and NS9283: high potency ACh binding in the wild‐type α4–β2 interface (red); NS9283 and low potency ACh binding in the wild‐type α4–α4 interface (orange); and intermediate potency ACh binding in the single or double mutant α4^?–α4^VFL interface (black)

Potency and efficacy of NS9283 are dependent on the presence of three E‐loop amino acids (H142, Q150, and T152) in the complementary (−) face of the α4 subunit (Olsen et al., 2014). Point mutating these three amino acids to the corresponding amino acids in β2 (V136, F144, and L146) causes the agonist binding pocket of the α4(+)–α4^VFL(−) interface to resemble that of an α4(+)–β2(−) interface. Apart from increasing the sensitivity of ACh, these mutations additionally result in a complete loss of NS9283's ability to positively modulate at the α4(+)–α4^VFL(−) interface (Figure 1, exemplified by constructs II and III). Instead, NS9283 displayed inhibitory effects at the highest tested concentrations, which likely results from antagonising ACh binding at α4(+)–β2(−) interfaces. To allow natural variations, such as experimental run‐up of current amplitudes, we chose to define average values below 10% as no‐efficacy responses (red territory in Figure 1).

In a scenario where the receptor pool contained a mixture of wild‐type receptors and receptors that hold the α4^VFL subunit as the complementary (−) face, that is, of the α4(+)–α4^VFL(−), only the wild‐type receptor fraction will respond to NS9283 applications. Hence, the efficacy measured in percentage will decrease while the accompanying potency remain the same. This is an inherent feature of the methodology for calculating efficacy levels ((I − I_{ACh_control}) × 100)/I_{ACh_control} as the current amplitude of I_{ACh_control} is a sum of responses contributed by all expressed receptors irrespective of their NS9283 sensitivity (Figure 1). We have chosen to define average values between 10% and 300% as intermediate efficacy responses (blue territory in Figure 1).

For reasons that are currently unknown, the three HQT to VFL mutations cause increased sensitivity of the second component of the ACh CRR in all α4^VFL containing receptors even when there is a wild‐type α4 subunit in the complementary (−) face of the α4(+)–α4(−) interface (Ahring et al., 2018; Lucero et al., 2016). As a result, the response to applications of ACh_control (10 μM) increases from an ~EC₁₅ to an ~EC₃₀ level. Thus, despite the presence of NS9283‐sensitive interfaces, efficacy percentages dropped to become intermediate responses under the utilised experimental conditions (Figure 1, construct IV).

Finally, while full CRRs of NS9283 were obtained for all tested constructs, we use only the recorded value at the 31.6‐μM concentration for direct efficacy comparisons for two reasons: (i) in many instances, NS9283 (100 μM) was visibly observed to precipitate out of solution as time passes; hence, data for this concentration are not always consistent; and (ii) NS9283 (100 μM) significantly inhibits ACh_control‐evoked responses for receptors that lack a responding α4(+)–α4(−) interface; hence, the efficacy in mixed populations could be underestimated. For the same reason, data for the 100‐μM concentration are omitted from the fitting routine in cases where the measured efficacy is lower than that at the 31.6‐μM concentration.

2.7. Modelling

Concatemers were modelled using the Modeller program 9v23 (MODELLER, RRID:SCR_008395) (Sali & Blundell, 1993) based on the (α4)₃(β2)₂ nAChR structure (PDB ID 6CNK) (Walsh et al., 2018) with amino acids numbered according to the full‐length α4 and β2 sequences. This involved keeping resolved α4 residues Ala³⁴ to Trp⁶²² (Ala⁸ to Trp³⁸¹ in PDB ID 6CNK) and resolved β2 residues Asp²⁷ to Phe⁴⁷⁸ (Asp² to Phe³⁶⁹ in PDB ID 6CNK) and discarding all other atoms. The missing loop segment between the MX and TM4 helices in each subunit and a single missing Ala³⁴ in chain B were modelled in. Templates were fitted with a β2‐9a‐α4 linker by modelling unresolved post TM4 residues (⁴⁷⁹LQPLF‐APSSK⁵⁰²) in addition to three AGS repeats to bridge the β2 TM4 and the α4 N‐terminus (…APSSK (AGS)₃AHAEE…). Constructs with clockwise and counterclockwise linkers were created by modelling the linker from chain C to B and chain E to A, respectively. Modelling path for linker residues bridging the subunits in the concatemer models was guided by supplying initial models with the residues roughly aligned in the clockwise or counterclockwise directions. Model construction consisted of treating protein atoms rigidly while allowing flexibility for the modelled region in addition to residues flanking it. Top clockwise and counterclockwise models were each selected among 20 based on DOPE score (Shen & Sali, 2006) and visual examination.

2.8. Simulation set‐up and equilibration

Selected models were prepared by adding hydrogens and optimising their non‐bonded interactions using the Schrödinger protein preparation wizard (Schrödinger Release 2019‐4; Schrödinger) and then as molecular dynamics systems using the Desmond system builder (Bowers et al., 2006). This involved inserting models into a 1‐palmitoyl‐2‐oleoyl‐sn‐glycero‐3‐phosphocholine (POPC) membrane guided by the OPM database (Lomize, Pogozheva, Joo, Mosberg, & Lomize, 2012), adding water containing 0.15‐M NaCl and neutralising ions such that at least 12 Å separated the protein from the simulation box boundary. Simulations were carried out using the OpenMM program (OpenMM, RRID:SCR_000436) (Eastman et al., 2017). The TIP3P (Jorgensen, Chandrasekhar, Madura, Impey, & Klein, 1983), AMBER14SB (Maier et al., 2015), and LIPID17 (Gould, Skjevik, Dickson, Madej, & Walker, 2018) forcefields were used to parameterise water, protein, and lipid atoms, respectively. Simulation used a Langevin integrator with a collision frequency of 1 ps⁻¹ to control temperature; the Monte Carlo membrane barostat was used to apply 1 bar of pressure while scaling the x and y axes isotopically. Non‐bonded interactions were cut‐off at 10 Å; electrostatic interactions extending beyond this range were calculated using particle mesh Ewald summation. A script was adapted from Martin (2018) to scale the simulation box to an equilibrium size of ~150 × 100 × 100 Å. During scaling, a script by Rodrigues (2019) was used to tether protein heavy atoms to massless non‐interacting atoms. This enabled the protein to shift while maintaining its original geometry during scaling. Equilibration was performed at 310 K, constant pressure, and 2‐fs time steps following a brief energy minimisation. Positional restraints of 1 kcal·(mol·Å²)⁻¹ were applied to protein heavy atoms and lifted from all except Cα atoms after 1 ns. Cα atoms belonging to the loop segment between the concatenated β2 and α4 subunits were released after an additional 1 ns, and remaining restraints were gradually withdrawn over 20 ns.

2.9. Simulation protocol

Simulated annealing molecular dynamics simulations were performed to identify low energy conformations of linkers and study the effect shortening of linkers had on protein structure. The simulation started with the β2‐9a‐α4‐linked concatemers, and in a continuous protocol, linkers were shortened residue by residue with a ~35 ns simulation window designated to each unique receptor structure. Each simulation window consisted of (i) an equilibration, (ii) an annealing, and (iii) a post annealing simulation step. During these simulations, hydrogen mass repartitioning was implemented to achieve a 4‐fs time step (Hopkins, Le Grand, Walker, & Roitberg, 2015). The equilibration step consisted of a minimisation and 5‐ns simulation at constant 310‐K temperature. Simulated annealing was conducted in 11 steps with temperature scaling to 400 K and then back down to 310 K. Heating phases lasted for 0.5 ns each while the cooling phase was increased in 0.25‐ns increments such that the shortest and longest cooling times were 0.5 and 3 ns, respectively. Progressively longer cooling times aimed to efficiently sample the effects of shortening the linker on protein structure and arrive at loop conformations occupying low energy minima. The post annealing simulation was again a 5‐ns simulation at constant 310‐K temperature. To smoothly transition between simulation windows, the distance between backbone atoms flanking the residue to be deleted was gradually scaled down during the final 0.1 ns of the simulation window, and forcefield parameters were reassigned using the final particle positions excluding removed residues, and particle velocities were reset to 310 K. To counter structure degradation from repeated annealing cycles, pairwise distance restraints were applied to extracellular β‐sheets from each subunit to maintain their original cross distances. TM1–4 helices in each subunit (excluding the TM4 helix used for steric energy analysis) were restrained in the same manner. This was achieved by applying 50 kcal·(mol·Å²)⁻¹ of force to centroids encompassing the upper and lower halves of helices/sheets in each subunit; a flat bottom potential enabled the restrained distance to fluctuate by ±0.5 Å. For linker lengths 9a, 7a, 5a, 3a, 2a, 1a, 0a (TLL: 38, 36, 34, 32, 31, 30, 29), post annealing simulations from immediately before shortening were extended to 150 ns to establish stability and enable measurement of steric energies of N‐ and C‐terminal helices.

2.10. Simulation analysis

The root‐mean‐square deviation (RMSD) of structures over the 150‐ns simulations was obtained using MDTraj (McGibbon et al., 2015) after superimposing the protein in simulation frames spaced 0.1 ns apart on the original model and was smoothed over a window of 10 measurements. RMSDs of Cα atoms are calculated from extracellular sheets and transmembrane helices, excluding the TM4 helix used for steric energy analysis (Figure 5). The effects of shortening the linker were evidently seen in distortion of the N‐terminal and TM4 helices flanking it (Figure 5). Optimal helical residue packing correlates well with steric energy (Kilosanidze, Kutsenko, Esipova, & Tumanyan, 2004); thus, steric energies within the N‐terminal and TM4 helices were recorded to capture helical disruption. N‐terminal atom selection included all helical residues, Ala³⁴ to Phe⁴⁵, while TM4 atom selection spanned from Met⁴⁵⁶ to Phe⁴⁷⁸ to omit the solvent exposed lower helical region which exhibited high fluctuations. It is noteworthy that the counterclockwise linker region preceding the N‐terminal took two paths around a loop section (Pro⁴⁸‐Leu⁵⁵) from the neighbouring β2 subunit. By passing over the loop, the linker (Figure 5l) avoids being situated between the two subunits where it may interfere with assembly; hence, this path was chosen for analysis. Energy components were extracted using OpenMMTools (Rizzi et al., 2019) by setting up structures from each simulation window as “Alchemical Systems” with an “Alchemical Region” for each selection. Steric energies within each region were calculated using the same forcefield and non‐bonded cut‐off settings as above and normalised by the number of residues in each selection, 12 and 23 for the N‐terminal and TM4, respectively. Average energy was evaluated based on particle positions from the final 50 ns of simulation from each linker using 500 frames each spaced 0.1 ns apart. To estimate standard uncertainty, block averaging was used to partition energies into blocks of increasing size until the blocks became uncorrelated (Grossfield et al., 2018). At this point, standard uncertainty between the blocks was calculated and then multiplied by 1.96 to provide an estimate of the 95% confidence interval (CI).

2.11. Materials

3‐(3‐(Pyridine‐3‐yl)‐1,2,4‐oxadiazol‐5‐yl)benzonitrile (NS9283) was synthesised at Saniona A/S as described previously (Timmermann et al., 2012). The structure was confirmed using MS and NMR and is of >98% purity. ACh and all salts or other chemicals not specifically mentioned were purchased from Sigma‐Aldrich (St. Louis, Missouri, USA) and were of analytical grade. Oligonucleotides were purchased from Sigma‐Aldrich, and sequencing services were from Australian Genome Research Facility (AGRF). Restriction enzymes, Q5 polymerase, T4 DNA ligase, and competent E. coli 10‐β bacteria were from New England Biolabs Inc. (Ipswich, Massachusetts, USA).

2.12. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander et al., 2019).

3. RESULTS

Ternary α4β2‐containing receptors were used as a model system to investigate the influence of short linkers on the assembly order of concatenated Cys‐loop receptors. While wild‐type α4β2 represent binary receptors, (α4)₂(β2)₂(α4^VFL)₁ receptors are ternary, and the mutated α4^VFL subunit can assemble into three distinct positions. As described in Section 2.6, the response to NS9283 (31.6 μM) was used to determine the specific position of the α4^VFL subunit within populations of such ternary α4β2α4^VFL‐containing receptors (Figure 1). A NS9283 response is either in the full (>300%) or intermediary (10–300%) range (Figure 1, green and blue areas). A no‐response is identified when the average NS9283 efficacy value approached the noise (<10%) threshold (Figure 1, red area). Note that most experiments are designed such that a uniform receptor pool of α4^VFL‐containing receptors displays a no‐response to NS9283.

3.1. What is the apparent optimal linker length for dimeric β‐xa‐α constructs?

In Ahring et al. (2018), two dimeric constructs, β‐3a‐α and β‐0a‐α, were co‐expressed with free α4 or α4^VFL subunits to investigate resulting receptors (Table 1). However, they both did not achieve optimal outcomes with the β‐3a‐α concatemers failing to form a uniform receptor pool (NS9283 response of 22%) and the β‐0a‐α concatemers severely compromising ACh_max‐evoked current amplitudes. In attempt to identify an optimal intermediate linker length, two new constructs, β‐2a‐α and β‐1a‐α, were designed. These dimers have TLLs of 31 and 30 amino acids and were tested for their ability to express with dangling subunits by themselves and to co‐express with free α4 or α4^VFL subunits (Table 1).

TABLE 1.

Maximal fitted response and potency of ACh and NS9283 from α4β2‐containing nAChRs using wild‐type subunits and concatenated dimeric constructs

cRNA a	TLL b	ACh						NS9283
		Emax c	pEC50_1 c (M)	pEC50_2 c (M)	Frac c	AChmax c (nA)	n	E31.6 μM d (%)	Emax e (%)	pEC50 e (M)	n
α4 + β2 (10:1)		1.02 ± 0.01	5.7 ± 0.3	3.9 ± 0.04	0.11 ± 0.03	8,500 ± 1,200	14	760 ± 70	860 ± 60	5.3 ± 0.1	13
α4VFL + β2 (10:1)		1.04 ± 0.02	5.0 ± 0.1	3.9 ± 0.3	0.71 ± 0.12	1,800 ± 600	9	−9.9 ± 1.8	No fit		16
β‐3a‐α	32	n/a	n/a	n/a	n/a	83 ± 14	14	n/a	n/a	n/a	n/a
β‐3a‐α + α4	32	1.02 ± 0.01	5.7 ± 0.2	4.0 ± 0.03	0.11 ± 0.02	1,400 ± 200	17	610 ± 30	720 ± 30	5.2 ± 0.1	18
β‐3a‐α + α4VFL	32	1.04 ± 0.01	5.7 ± 0.1	4.4 ± 0.1	0.41 ± 0.06	680 ± 120	17	22 ± 3	27 ± 3	5.2 ± 0.2	24
β‐2a‐α	31	n/a	n/a	n/a	n/a	67 ± 12	27	n/a	n/a	n/a	n/a
β‐2a‐α + α4	31	1.01 ± 0.01	5.7 ± 0.5	4.0 ± 0.04	0.060 ± 0.027	2,300 ± 400	14	640 ± 50	730 ± 40	5.4 ± 0.1	12
β‐2a‐α + α4VFL	31	1.03 ± 0.01	5.4 ± 0.2	4.3 ± 0.2	0.45 ± 0.13	970 ± 270	13	−6.6 ± 2.2	No fit		19
β‐1a‐α	30	n/a	n/a	n/a	n/a	20 ± 5	27	n/a	n/a	n/a	n/a
β‐1a‐α + α4	30	1.00 ± 0.02	5.9 ± 0.9	3.8 ± 0.1	0.041 ± 0.028	1,300 ± 300	11	890 ± 60	1,100 ± 100	5.1 ± 0.1	10
β‐1a‐α + α4VFL	30	1.02 ± 0.01	5.3 ± 0.2	4.4 ± 0.2	0.45 ± 0.18	180 ± 40	13	−2.6 ± 2.1	No fit		17
β‐0a‐α	29	n/a	n/a	n/a	n/a	5.3 ± 1.9	24	n/a	n/a	n/a	n/a
β‐0a‐α + α4	29	1.03 ± 0.02	5.8 ± 0.7	3.8 ± 0.1	0.053 ± 0.028	130 ± 60	11	840 ± 90	1,100 ± 100	5.0 ± 0.1	9
β‐0a‐α + α4VFL	29	1.01 ± 0.01	5.4 ± 0.5	4.6 ± 0.2	0.28 ± 0.34	120 ± 30	14	−6.0 ± 2.4	No fit		15
β‐3a‐αVFL + α4	32	1.06 ± 0.01	5.5 ± 0.1	4.1 ± 0.1	0.39 ± 0.06	990 ± 190	13	190 ± 20	210 ± 10	5.1 ± 0.1	12
β‐2a‐αVFL + α4	31	1.06 ± 0.02	5.3 ± 0.1	3.7 ± 0.1	0.28 ± 0.04	1,600 ± 300	11	220 ± 10	270 ± 10	5.1 ± 0.1	11
β‐1a‐αVFL + α4	30	1.03 ± 0.02	5.4 ± 0.4	3.7 ± 0.1	0.16 ± 0.06	840 ± 180	10	550 ± 30	670 ± 30	5.1 ± 0.1	10
β‐0a‐αVFL + α4	29	1.01 ± 0.01	6.0 ± 0.4	3.9 ± 0.04	0.065 ± 0.019	520 ± 120	12	720 ± 30	940 ± 30	5.0 ± 0.1	12

^a Xenopus laevis oocytes were injected with the indicated cRNA mixtures in a 1:1 ratio (unless otherwise indicated), incubated for 2–3 days, and then subjected to two‐electrode voltage‐clamp electrophysiology as described in the methods; also see Figure 1.

^bTotal linker length (TLL) is calculated as the sum of existing amino acids and introduced linker sequence between two artificial fix points: (i) the last amino acid in TM4 of the upstream subunit and (ii) a leucine anchor in α‐helix‐1 segment in the downstream subunit.

^cNormalised ACh data points are fitted to a second‐order equation with bottom set to 0 and Hill slopes set to 1 by non‐linear regression. Biphasic fitting represented the preferred model for all datasets as determined by an F‐test. Fitted maximal response of ACh is presented as E_max ± SEM, and accompanying potencies are presented as pEC₅₀ ± SEM for each of the two fitted components. The fraction of the first component is presented as Frac. The average maximal current amplitude obtained with ACh (3,160 μM) is presented as ACh_max ± SEM in nA for the indicated number of n individual oocytes.

^dCalculated efficacy for NS9283 (31.6 μM) is presented as E_{31.6 μM} in %.

^eNS9283 data points were fitted to a first‐order equation with bottom set to 0 and Hill slope set to 1 by non‐linear regression. Fitted maximal responses are presented as E_max ± SEM in % with associated potencies presented as pEC₅₀ ± SEM for the indicated number of individual oocytes.

Injecting oocytes with cRNA for either β‐2a‐α or β‐1a‐α alone led to average ACh_max‐evoked peak current amplitudes ranging from 20 to 70 nA (Table 1). These values fall in between those observed with β‐3a‐α and β‐0a‐α indicating that shortening of the linker length reduces the ability of dimers to assemble with themselves. Next, co‐injecting β‐2a‐α or β‐1a‐α with free α4 subunits resulted in receptor pools that are virtually indistinguishable from wild‐type (α4)₃(β2)₂ receptors (Figure 2a). Finally, co‐injecting β‐2a‐α or β‐1a‐α with free α4^VFL subunits led to ACh_max‐evoked current amplitudes ranging from 200 to 1,000 nA (Table 1). The ACh CRRs were best approximated by a biphasic equation and resembled that of wild‐type (α4^VFL)₃(β2)₂ receptors (Figure 2a). No positive enhancement of ACh_control responses was observed with NS9283 (Figure 2b,c).

FIGURE 2.

ACh and NS9283 sensitivity and potential stoichiometry of receptors from concatenated dimeric constructs co‐expressed with free α4 or α4^VFL subunits. Xenopus laevis oocytes were injected with cRNA mixtures of β‐xa‐α constructs with free α4^VFL subunits or β‐xa‐α^VFL constructs with free α4 subunits in a 1:1 ratio and subjected to two‐electrode voltage‐clamp electrophysiology as described in the methods. The short 0a–3a linkers (xa) have TLLs of 29–32 amino acids (Table 1). Note that graphical representations are colour coded according to the observed NS9283 efficacy for each combination, that is, dark green = full efficacy, dark blue = intermediate efficacy, and dark red = no positive efficacy. (a and f) Baseline subtracted ACh‐evoked peak current amplitudes (I) for receptors from the indicated cRNA mixtures were fitted to the Hill equation by non‐linear regression and normalised to the maximal fitted value (I_{max_fit_ACh}). Normalised responses are shown as means ± SD as a function of the ACh concentration and fitted to a second‐order equation with a fixed bottom of 0 and a Hill slope of 1. Data were obtained from n = 10–14 experiments, and regression results are presented in Table 1. Note that data for cRNA mixtures of free subunits and mixtures containing dimeric β‐3a‐α and β‐0a‐α constructs are from Ahring et al. (2018). (b and g) NS9283 enhancement of ACh‐evoked currents was evaluated for the indicated receptors by co‐application with a submaximal control concentration of ACh (10 μM). Baseline subtracted peak current amplitudes (I) were expressed as % change from I_{ACh_control} [(I − I_C)/I_C] and are shown as means ± SD as a function of the NS9283 concentration. Data points were fitted by non‐linear regression to the Hill equation with a fixed bottom of 0 and a Hill slope of 1. In some instances, the fitting routine did not converge, and for these cases, data points are connected by lines. Data were obtained from n = 10–19 experiments, and regression results are presented in Table 1. Note that data for cRNA mixtures containing the dimeric β‐3a‐α and β‐0a‐α constructs are from Ahring et al. (2018). (c and h) Representative traces illustrating NS9283 CRRs at receptors from the indicated cRNA mixtures. Bars above the traces designate the ~25‐s application time and concentrations of applied compounds. (d, e, i, and j) Schematic representations of injected cRNA mixtures and the simplest receptor pools that explain the ACh and NS9283 observations. The constituents of cRNA mixtures are illustrated by colour‐coded bars with the identity of linked or free subunits indicated. Connecting grey bars illustrate linker lengths from 0a to 3a. Receptor complexes are illustrated as explained in Figure 1. Linkers are shown as purple arrows progressing from the C‐terminus of the first subunit to the N‐terminus of the second. Linker orientation(s) within a receptor complex is indicated below by orange (counterclockwise) or yellow (clockwise) arrows. Where NS9283 data indicate that one receptor stoichiometry is the dominant in a receptor pool, these are marked by “Preferred” or “Exclusive (?).” Note that only the simplest assembly possibilities are shown, and more options should not be excluded

Overall, both new constructs appear superior compared to the original β‐3a‐α and β‐0a‐α constructs for two reasons: (i) the new 2a‐ and 1a‐linked concatemers appear to give exclusive counterclockwise assembly (Figure 2d,e); and (ii) the new concatemers, in particular β‐2a‐α, does so without a substantial reduction of ACh_max‐evoked current amplitudes (Table 1). Hence, the optimal linker length for dimeric constructs appears to be a 2a linker corresponding to a TLL of 31 amino acids.

3.2. Are results from dimeric constructs trustworthy?

To investigate whether β‐xa‐α dimers assemble as expected (illustrated in Figure 2d,e), four new β‐xa‐α^VFL constructs were designed with linker lengths ranging from 0a to 3a (Table 1). These differ from the previous dimeric constructs by having the α4^VFL subunit integrated into the concatemer.

Injection of a β‐3a‐α^VFL + α4 cRNA mixture led to receptors that responded to ACh_max with robust current amplitudes (Table 1). While the ACh CRR resembled that of (α4^VFL)₃(β2)₂ receptors, the NS9283 efficacy was in the intermediate range at ~200% (Figure 2f–h). These data are consistent with a pool of receptors that contain β2–α4^VFL interfaces and a NS9283‐sensitive α4^VFL–α4 interface (Figure 1). Hence, this cRNA mixture probably formed receptors with dimers preferring the counterclockwise orientation mimicking that of the corresponding β‐3a‐α construct above (Figure 2d,i). In contrast, injection of the shortest linked β‐0a‐α^VFL + α4 cRNA mixture led to formation of receptors that resembled wild‐type (α4)₃(β2)₂ receptors with characteristic biphasic ACh CRR curve accompanying full NS9283 efficacy of ~700% (Figure 2f–h). This suggests the absence of an α4^VFL subunit in these receptors, and the most likely explanation is the integration of just the β2 subunits from two β‐0a‐α^VFL dimers into the pentameric complex leaving two dangling α4^VFL subunits (Figure 2j).

Given that β‐3a‐α^VFL and β‐0a‐α^VFL represent the longest and the shortest of the four new constructs, the linker in the 0a‐linked concatemer might be below a minimum threshold for possible assembly of both subunits into a pentameric receptor. In support of this, the sharp increase in NS9283 efficacy from 220% to 550% for the 2a‐ versus the 1a‐linked concatemer suggests a cut‐off for counterclockwise dimer assembly below a TLL of 31 amino acids (Figure 2g). However, while the 2a‐linked construct thus appears to be the optimal one, the potential existence of receptors with dangling linked subunits in the receptor pool cannot be excluded based on these data.

3.3. Tetrameric constructs with one or two short linkers

Next, a range of tetrameric (T) β‐xa‐α‐β‐xa‐α constructs were designed with either one, T(x), or two, T(x,x), short linkers. The same short 0a–3a linkers were used corresponding to TLLs in the 29–32 amino acid range while the remaining linkers were longer with TLLs of 47 amino acids (Figure 3a). For ease, only short linkers are indicated in the construct naming scheme. Depending on the assembly orientation, both T(x) and T(x,x) concatemers should result in NS9283‐sensitive or NS9283‐insensitive interfaces when co‐expressed with free α4 or α4^VFL subunits (Figure 3a).

FIGURE 3.

ACh and NS9283 sensitivity and potential stoichiometry of receptors from concatenated tetrameric constructs. Xenopus laevis oocytes were injected with cRNA mixtures of tetrameric constructs with free α4 or α4^VFL subunits in a 2:1 ratio and subjected to two‐electrode voltage‐clamp electrophysiology. Electrophysiological data were evaluated as described in the methods; also see Figure 2. (a) Schematic representations of injected cRNA mixtures and the simplest expected receptor pools. Tetrameric constructs contain either one T(x) or two T(x,x) short 0a–3a linkers (xa) with TLLs of 29–32 amino acids whereas remaining L linkers have TLLs of 47 amino acids. Depending on the free subunit and assembly orientation resulting receptors are either NS9283 sensitive or insensitive. For further schematic explanations, see Figure 2. (b and d) ACh CRRs are shown as means ± SD for receptors from the indicated cRNA mixtures. Data were obtained from n = 11–16 experiments, and regression results are presented in Table 2. (c and e) NS9283 (31.6 μM) efficacy values are depicted as means ± SEM for the indicated receptors from n = 10–15 experiments. Regression results for complete CRRs are presented in Table 2. (f and g) Schematic representations of T(3,3) + α4^VFL and T(2,2) + α4^VFL cRNA mixtures and the simplest receptor pools that explain the ACh and NS9283 observations. For explanation of schematics, see Figure 2

3.3.1. T(x) constructs

Injections of cRNA mixtures of T(x) constructs with free α4 subunits led to efficient expression of functional receptors that resembled wild‐type (α4)₃(β2)₂ receptors with respect to their ACh CRRs and NS9283 efficacies (Figure 3b,c). Injections with free α4^VFL subunits instead led to receptor pools with ACh CRRs resembling (α4^VFL)₃(β2)₂ receptors and accompanying NS9283 efficacies within the intermediate range at ~100% for the T(3) concatemer decreasing to ~30% for the shorter T(1) and T(0) concatemers. Hence, despite clear effects of shortening the linker, none of the four T(x) constructs appear able to fully restrict concatemer assembly to the counterclockwise orientation.

3.3.2. T(x,x) constructs

Injection of cRNA mixtures of T(x,x) constructs with free α4 subunits led to robust expression with observed ACh_max‐evoked current amplitudes in the μA range (Table 2). ACh CRRs and NS9283 efficacies resembled wild‐type (α4)₃(β2)₂ receptors demonstrating that two short linkers within a tetrameric concatemer do not impair normal receptor function (Figure 3d,e). As expected, co‐injection of T(x,x) constructs with free α4^VFL subunits led to ACh CRRs that resemble (α4^VFL)₃(β2)₂ receptors. A NS9283 efficacy of ~20% was narrowly within the intermediate range for the T(3,3) concatemer whereas no positive efficacy was observed with the T(2,2) and T(1,1) concatemers. Thus, introduction of a second short linker substantially increased control of concatemer assembly. Two short 3a linkers resulted in a preferred counterclockwise assembly, and shortening these further to 2a linkers gave exclusive counterclockwise assembly (Figure 3f,g). It is, however, important to note that all the tested tetrameric concatemers maintained an ability to express with dangling subunits by themselves. Irrespective of linker lengths, the current amplitudes for what is likely di‐tetramer receptors constituted approximately 1% of the amplitudes observed for co‐expression with free α4 subunits (Table 2).

TABLE 2.

Maximal fitted response and potency of ACh and NS9283 from α4β2‐containing nAChRs using tetrameric concatenated constructs

cRNA a	TLL b	ACh						NS9283
		Emax c	pEC50_1 c (M)	pEC50_2 c (M)	Frac c	AChmax c (nA)	n	E31.6 μM d (%)	Emax e (%)	pEC50 e (M)	n
β‐3a‐α‐β‐α	32	n/a	n/a	n/a	n/a	18 ± 7	25	n/a	n/a	n/a	n/a
β‐3a‐α‐β‐α + α4	32	1.03 ± 0.02	5.7 ± 0.2	3.8 ± 0.1	0.16 ± 0.03	2,900 ± 500	13	570 ± 50	720 ± 40	5.0 ± 0.1	14
β‐3a‐α‐β‐α + α4VFL	32	1.06 ± 0.01	5.5 ± 0.1	4.1 ± 0.1	0.35 ± 0.05	790 ± 150	16	110 ± 10	140 ± 10	5.2 ± 0.1	15
β‐2a‐α‐β‐α	31	n/a	n/a	n/a	n/a	31 ± 5	22	n/a	n/a	n/a	n/a
β‐2a‐α‐β‐α + α4	31	1.03 ± 0.02	5.6 ± 0.3	3.7 ± 0.1	0.15 ± 0.03	4,000 ± 400	14	630 ± 70	730 ± 40	5.2 ± 0.1	11
β‐2a‐α‐β‐α + α4VFL	31	1.05 ± 0.01	5.4 ± 0.1	4.1 ± 0.1	0.39 ± 0.05	1,500 ± 400	11	52 ± 5	60 ± 4	5.3 ± 0.1	10
β‐1a‐α‐β‐α	30	n/a	n/a	n/a	n/a	67 ± 10	22	n/a	n/a	n/a	n/a
β‐1a‐α‐β‐α + α4	30	1.03 ± 0.02	5.8 ± 0.3	3.6 ± 0.1	0.14 ± 0.03	4,400 ± 800	11	500 ± 60	610 ± 40	5.1 ± 0.1	10
β‐1a‐α‐β‐α + α4VFL	30	1.05 ± 0.02	5.5 ± 0.1	4.2 ± 0.1	0.49 ± 0.08	2,500 ± 700	12	32 ± 3	37 ± 3	5.4 ± 0.1	12
β2‐0a‐α‐β‐α	29	n/a	n/a	n/a	n/a	49 ± 7	23	n/a	n/a	n/a	n/a
β2‐0a‐α‐β‐α + α4	29	1.02 ± 0.02	5.8 ± 0.4	3.7 ± 0.1	0.12 ± 0.03	4,600 ± 800	11	720 ± 100	930 ± 70	5.0 ± 0.1	10
β2‐0a‐α‐β‐α + α4VFL	29	1.02 ± 0.01	5.4 ± 0.3	4.4 ± 0.1	0.34 ± 0.15	1,100 ± 400	16	32 ± 3	37 ± 3	5.1 ± 0.1	14
β‐3a‐α‐β‐3a‐α	32,32	n/a	n/a	n/a	n/a	32 ± 5	24	n/a	n/a	n/a	n/a
β‐3a‐α‐β‐3a‐α + α4	32,32	1.04 ± 0.01	5.9 ± 0.1	3.7 ± 0.04	0.18 ± 0.02	2,700 ± 600	13	450 ± 50	590 ± 50	5.0 ± 0.1	13
β‐3a‐α‐β‐3a‐α + α4VFL	32,32	1.04 ± 0.01	5.5 ± 0.2	4.3 ± 0.1	0.39 ± 0.10	1,300 ± 300	16	23 ± 4	26 ± 3	5.5 ± 0.2	13
β‐2a‐α‐β‐2a‐α	31,31	n/a	n/a	n/a	n/a	27 ± 5	27	n/a	n/a	n/a	n/a
β‐2a‐α‐β‐2a‐α + α4	31,31	1.04 ± 0.02	5.9 ± 0.2	3.8 ± 0.1	0.21 ± 0.03	1,500 ± 400	14	400 ± 30	490 ± 20	5.0 ± 0.1	10
β‐2a‐α‐β‐2a‐α + α4VFL	31,31	1.03 ± 0.01	5.7 ± 0.1	4.4 ± 0.1	0.40 ± 0.06	290 ± 70	12	2.3 ± 2.2	No fit		12
β‐1a‐α‐β‐1a‐α	30,30	n/a	n/a	n/a	n/a	7.5 ± 1.0	24	n/a	n/a	n/a	n/a
β‐1a‐α‐β‐1a‐α + α4	30,30	1.02 ± 0.02	5.8 ± 0.3	3.8 ± 0.1	0.14 ± 0.03	780 ± 160	11	510 ± 30	640 ± 20	5.0 ± 0.1	12
β‐1a‐α‐β‐1a‐α + α4VFL	30,30	1.03 ± 0.01	5.6 ± 0.2	4.5 ± 0.1	0.42 ± 0.10	100 ± 20	14	0.15 ± 2.7	No fit		13

^a Xenopus laevis oocytes were injected with the indicated cRNA mixtures in a 2:1 ratio, incubated for 4 days, and subjected to two‐electrode voltage‐clamp electrophysiology as described in the methods; also see Figure 1.

^bTotal linker length (TLL) is calculated as the sum of existing amino acids and introduced linker sequence between two artificial fix points: (i) the last amino acid in TM4 of the upstream subunit and (ii) a leucine anchor in α‐helix‐1 segment in the downstream subunit.

^cNormalised ACh data points are fitted to a second‐order equation with bottom set to 0 and Hill slopes set to 1 by non‐linear regression. Biphasic fitting represented the preferred model for all datasets as determined by an F‐test. Fitted maximal response of ACh is presented as E_max ± SEM, and accompanying potencies are presented as pEC₅₀ ± SEM for each of the two fitted components. The fraction of the first component is presented as Frac. The average maximal current amplitude obtained with ACh (3,160 μM) is presented as ACh_max ± SEM in nA for the indicated number of n individual oocytes.

^dCalculated efficacy for NS9283 (31.6 μM) is presented as E_{31.6 μM} in %.

^eNS9283 data points were fitted to a first‐order equation with bottom set to 0 and Hill slope set to 1 by non‐linear regression. Fitted maximal responses are presented as E_max ± SEM in % with associated potencies presented as pEC₅₀ ± SEM for the indicated number of individual oocytes.

3.4. Pentameric constructs with one or two short linkers

In Ahring et al. (2018), a pentameric construct with a single short 3a linker was used as a “good compromise” for obtaining the preferred counterclockwise assembly of expressed receptors; however, this construct did not lead to a fully uniform receptor pool. Therefore, five new pentameric β‐xa‐α‐β‐xa‐α‐α constructs were designed with either one, P(x), or two, P(x,x), short linkers (Figure 4a). The subunit in the last construct position was either an α4 or an α4^VFL subunit, and all TLLs were identical to those in the tetrameric constructs.

FIGURE 4.

ACh and NS9283 sensitivity and potential stoichiometry of receptors from concatenated pentameric constructs. Xenopus laevis oocytes were injected with cRNA for pentameric constructs and subjected to two‐electrode voltage‐clamp electrophysiology. Electrophysiological data were evaluated as described in the methods; also see Figure 2. (a) Schematic representations of injected cRNA and the simplest expected receptor pools. Pentameric constructs contain either a wild‐type α4 or a mutated α4^VFL subunit in the fifth construct position and either one P(x) or two P(x,x) short 0a–3a linkers (xa) with TLLs of 29–32 amino acids whereas remaining L linkers have TLLs of 47 amino acids. Depending on the assembly orientation resulting receptors are either NS9283 sensitive or insensitive. For further schematic explanations, see Figure 2. (b) ACh CRRs are depicted as means ± SD for receptors from the indicated cRNA mixtures. Data were obtained from n = 10–12 experiments, and regression results are presented in Table 3. Note that data for the β‐3a‐α‐β‐α‐α and β‐3a‐α‐β‐α‐α^VFL constructs are from Ahring et al. (2018). (c and d) NS9283 (31.6 μM) efficacy values are depicted as means ± SEM for the indicated receptors from n = 8–12 experiments. Regression results for complete CRRs are presented in Table 3. Note that data for the β‐3a‐α‐β‐α‐α and β‐3a‐α‐β‐α‐α^VFL constructs are from Ahring et al. (2018). (e) Representative traces illustrating NS9283 CRRs at receptors from the indicated cRNA mixtures. Bars above the traces designate the ~25‐s application time and concentrations of applied compounds. (f and g) Schematic representations of P(3) w. α4^VFL and P(3,3) w. α4^VFL cRNA and the simplest receptor pools that explain the ACh and NS9283 observations. For explanation of schematics, see Figure 2

3.4.1. P(x) constructs

Injection of cRNA for P(x) constructs containing only wild‐type α4 subunits resulted in receptor pools that were identical to wild‐type (α4)₃(β2)₂ receptors based on their ACh CRRs and NS9283 efficacies (Figure 4b,c). Similarly, P(x) constructs containing the α4^VFL subunit resulted in receptors that overall resembled (α4^VFL)₃(β2)₂ receptors. Interestingly, average NS9283 efficacies for these revealed a complex picture. For the shortest linked P(0) concatemer, the efficacy was clearly in the no‐response range suggesting exclusive counterclockwise assembly. NS9283 efficacies for the longer linked P(1) and P(2) concatemers, like the original P(3) concatemer, however, hover just below or above the 10% cut‐off for the intermediate efficacy range suggesting the occurrence of some assembly in the clockwise orientation (Figure 4c). The accompanying fitted pEC₅₀ values were in the 5.8–6.0 range, which is at least 0.5 log unit higher than the typical 5.0–5.3 range observed across the bulk of the experiments (Table 3). The reason for this peculiarity is that NS9283 only displays efficacy with the regular potency in some of the tested oocytes. Hence, the average values span observations of low efficacies (<25%) in responding oocytes mixed with noise observed in non‐responding oocytes.

TABLE 3.

Maximal fitted response and potency of ACh and NS9283 from α4β2‐containing nAChRs using pentameric concatenated constructs

cRNA a	TLL b	ACh						NS9283
		Emax c	pEC50_1 c (M)	pEC50_2 c (M)	Frac c	AChmax c (nA)	n	E31.6 μM d (%)	Emax e (%)	pEC50 e (M)	n
β‐3a‐α‐β‐α‐α	32	1.02 ± 0.01	6.0 ± 0.3	4.0 ± 0.04	0.11 ± 0.02	730 ± 130	9	640 ± 90	800 ± 90	5.1 ± 0.2	5
β‐3a‐α‐β‐α‐αVFL	32	1.04 ± 0.01	5.4 ± 0.1	4.3 ± 0.1	0.53 ± 0.08	290 ± 50	14	11 ± 2 f	13 ± 1 f	5.8 ± 0.2	15
β‐2a‐α‐β‐α‐α	31	1.02 ± 0.02	5.6 ± 0.4	3.9 ± 0.1	0.11 ± 0.05	5,400 ± 700	10	610 ± 90	740 ± 60	5.2 ± 0.1	12
β‐2a‐α‐β‐α‐αVFL	31	1.04 ± 0.01	5.5 ± 0.1	4.3 ± 0.1	0.42 ± 0.06	1,800 ± 400	12	16 ± 3 f	18 ± 2 f	5.9 ± 0.2	9
β‐1a‐α‐β‐α‐α	30	1.01 ± 0.01	5.6 ± 0.3	3.7 ± 0.03	0.088 ± 0.020	3,100 ± 600	12	660 ± 60	790 ± 40	5.2 ± 0.1	12
β‐1a‐α‐β‐α‐αVFL	30	1.04 ± 0.01	5.5 ± 0.1	4.3 ± 0.1	0.41 ± 0.08	760 ± 180	11	7.8 ± 3.1 f	12 ± 2 f	6.0 ± 0.3	12
β‐0a‐α‐β‐α‐α	29	1.00 ± 0.01	5.9 ± 0.6	3.7 ± 0.04	0.047 ± 0.022	1,200 ± 300	10	810 ± 70	1,000 ± 100	5.1 ± 0.1	10
β‐0a‐α‐β‐α‐αVFL	29	1.02 ± 0.01	5.4 ± 0.2	4.4 ± 0.2	0.43 ± 0.16	290 ± 80	12	−2.5 ± 2.1	No fit		12
β‐3a‐α‐β‐3a‐α‐α	32,32	1.02 ± 0.03	6.1 ± 0.4	3.9 ± 0.1	0.12 ± 0.04	720 ± 240	10	620 ± 30	740 ± 20	5.2 ± 0.1	11
β‐3a‐α‐β‐3a‐α‐αVFL	32,32	1.02 ± 0.01	5.6 ± 0.2	4.6 ± 0.2	0.51 ± 0.14	740 ± 170	11	−2.0 ± 3.5	No fit		10
β‐2a‐α‐β‐2a‐α‐α	31,31	1.03 ± 0.02	5.9 ± 0.3	3.9 ± 0.1	0.18 ± 0.04	530 ± 170	10	410 ± 20	490 ± 10	5.2 ± 0.04	8
β‐2a‐α‐β‐2a‐α‐αVFL	31,31	1.01 ± 0.01	5.5 ± 0.4	4.8 ± 0.4	0.49 ± 0.43	100 ± 20	11	−13 ± 2	No fit		10

^a Xenopus laevis oocytes were injected with cRNA for the indicated pentameric constructs, incubated for 4 days, and subjected to two‐electrode voltage‐clamp electrophysiology as described in the methods; also see Figure 1.

^bTotal linker length (TLL) is calculated as the sum of existing amino acids and introduced linker sequence between two artificial fix points: (i) the last amino acid in TM4 of the upstream subunit and (ii) a leucine anchor in α‐helix‐1 segment in the downstream subunit.

^cNormalised ACh data points are fitted to a second‐order equation with bottom set to 0 and Hill slopes set to 1 by non‐linear regression. Biphasic fitting represented the preferred model for all datasets as determined by an F‐test. Fitted maximal response of ACh is presented as E_max ± SEM, and accompanying potencies are presented as pEC₅₀ ± SEM for each of the two fitted components. The fraction of the first component is presented as Frac. The average maximal current amplitude obtained with ACh (3,160 μM) is presented as ACh_max ± SEM in nA for the indicated number of n individual oocytes.

^dCalculated efficacy for NS9283 (31.6 μM) is presented as E_{31.6 μM} in %.

^eNS9283 data points were fitted to a first‐order equation with bottom set to 0 and Hill slope set to 1 by non‐linear regression. Fitted maximal responses are presented as E_max ± SEM in % with associated potencies presented as pEC₅₀ ± SEM for the indicated number of individual oocytes.

^fOnly few oocytes show distinct NS9283 efficacy (<25%) with the expected fitted EC₅₀ value; hence, the average value primarily represents noise.

3.4.2. P(x,x) constructs

Injection of cRNA for P(x,x) constructs containing either wild‐type α4 subunits only or a single α4^VFL subunit resulted in receptor pools for which the ACh CRRs and NS9283 efficacies unsurprisingly resembled the (α4)₃(β2)₂ or the (α4^VFL)₃(β2)₂ receptors, respectively (Figure 4b,d,e). In contrast to the P(3) and P(2) concatemers, no NS9283 efficacy exceeding 10% was observed in any single oocyte with the P(3,3) and P(2,2) constructs.

Using a single short linker, a fully uniform receptor pool demonstrating no NS9283 efficacy was only obtained with the shortest P(0) concatemer (Figure 4f). Yet, the introduction of a second short linker increases the assembly control observed in tetrameric concatemers also holds true for pentameric concatemers as evidenced by the exclusive counterclockwise assembly of both the P(3,3) and P(2,2) concatemers (Figure 4g).

3.5. Modelled effects of gradually shortening linkers on receptor structure

From data with dimeric to pentameric concatemers, it is evident that a small change in linker length can have a surprisingly large effect on the assembly orientation. Therefore, a pentameric receptor incorporating a β2‐9a‐α4 dimer was modelled with the linker in both the clockwise and counterclockwise directions (Figure 5a,b). Next, molecular dynamics and simulated annealing were used to study how gradual shortening of the linker length affected the protein structure. Stability of the models was assessed from RMSD plots of extracellular β‐sheets and transmembrane α‐helices excluding the TM4 helix used for steric energy analysis (Figure 5c,d). β‐sheets were observed to be stable throughout the simulation while RMSD of Cα atoms in the TM‐helical regions converged beyond ~85 ns; hence, only the final 50 ns of the simulations were used for energy analysis.

Steric energy has been proposed as a determinant of helicity and, thus, internal steric energies normalised by the number of residues within the N‐terminus and TM4 helices were analysed to quantify linker‐induced strain. As linkers were shortened, energy within TM4 remained relatively constant but increased for the N‐terminus, which dominated the sum of the TM4 and N‐terminal energies (Figure 5e–g). The overall trend indicates that linker‐induced strain increases sharply when the linker in simulated constructs is below 5a for the clockwise and 3a for the counterclockwise orientations. Overall, the energy in simulated constructs with counterclockwise linkers trailed the energy of clockwise linkers by ~1–3 residues. The final frame in simulation windows corresponding to each clockwise and counterclockwise linker length enables direct comparison of linker‐induced effects on the protein structure (Figure 5h–m). In structures with gradually shorter clockwise linkers, TM4 dynamically tilts towards the left (Figure 5h) while the N‐terminus was downwardly distorted (Figure 5i). In contrast, for structures with gradually shorter counterclockwise linkers, TM4 dynamically tilts towards the right (Figure 5k), while the N‐terminus was laterally distorted (Figure 5l). The tilting of TM4 reflects the assembly directions and has little effect on the overall steric energy measurements (Figure 5e) as helicity is largely preserved.

Other noteworthy observations emerged from the simulation experiments. In Ahring et al. (2018), it was described how a series of hydrophobic residues (α4: LLxxLF⁴⁵) anchor the N‐terminal α‐helix into a hydrophobic pocket. The first leucine (α4: Leu⁴⁰) was then defined as an anchor from which to calculate TLLs, and for that to be meaningful, this leucine residue must remain anchored in concatemers. During simulations, the leucine anchor was indeed securely seated for all simulation constructs with counterclockwise linkers; however, it became dislocated when clockwise linkers were 2a or shorter indicating substantial strain. Finally, while counterclockwise linkers appear flexible and able to dynamically slide against the concatenated β2 subunit, clockwise linkers irrespective of length appear to follow a similar path between the N‐ and C‐termini (cf. Figure 5j,m). This fixed path appeared reinforced by β2 Leu⁴⁹² packing against the concatenated α4 subunit (Figure 5j), which indicates that specific residues and interactions may play a role in the characteristics of a linker.

4. DISCUSSION

Subunit concatenation is currently the only methodology that offers detailed control of Cys‐loop receptor assembly. Therefore, it has been widely used in attempts to determine the specific stoichiometry of complex nACh and GABA_A receptors (Baur, Kaur, & Sigel, 2010; Jain, Kuryatov, Wang, Kamenecka, & Lindstrom, 2016; Jin, Bermudez, & Steinbach, 2014; Kuryatov & Lindstrom, 2011). However, we recently demonstrated that there are two strong caveats associated with the concatenation technology: (i) dimeric or tetrameric concatemers have an inherent ability to express functional receptors by themselves; and (ii) long linkers allow concatemer assembly in both the clockwise and counterclockwise orientations (Ahring et al., 2018; Liao et al., 2019). Therefore, in this study, we aimed to identify the best strategies for designing concatenated constructs that can be used to reliably express uniform ternary receptor pools.

4.1. Modelling

Upon investigating static structures of the β2–α4 dimer interface, it was previously estimated that the difference in TLL required to connect subunits was approximately 20 Å between the longer clockwise and the shorter counterclockwise orientations (Ahring et al., 2018). For a short linker to exert control over the assembly orientation, it should therefore induce strain on the dimer interfaces when these assemble in the clockwise but not in the counterclockwise orientation. Achieving this appears feasible with an estimated difference between the two orientations equating to the length of six amino acids. However, one of the key findings from our experimental data is that a difference of a single amino acid in the linker sequence can result in significant changes to the assembly orientation of a concatemer. Our modelling analysis provides plausible explanations for why this is so. Two independent phenomena counteract the desired restrictions that a linker in a β‐xa‐α dimer is introducing during assembly. First, the N‐terminal amino acids in α‐helix 1 can be partly unwound and be pulled in either a lateral or downward direction. Second, part of the β2 TM4 helix can tilt towards the neighbouring α4 subunit in a direction respective to clockwise or counterclockwise assembly. Collectively, these phenomena mean that the actual difference in distance that a linker must traverse for clockwise or counterclockwise assemblies is marginal under experimental conditions. During simulations, a linker difference of one to three amino acids separated the rise in strain for the terminal helices in clockwise and counterclockwise linkers, which closely match the experimental observations.

Given the need for obtaining uniform receptor pools, it might seem logical to use linkers as short as possible. However, from the simulations, it is furthermore evident that counterclockwise linkers shorter than 3a, which correspond to a TLL of 32 amino acids, introduce increasing strain in the concatemer protein structure as the linker is shortened. Intuitively, it appears best to avoid such distortions and find a compromise between the need for a uniform receptor population and the possibility of incurring structural distortion.

4.2. Dimeric constructs

Dimeric constructs are the easiest to design, and a single construct can potentially be used to express a range of ternary receptors by mixing them with free subunits or other concatenated constructs, for example, trimeric constructs. In an attempt to identify the optimal linker length that would lead to efficient expression of a uniform receptor pool, we initially designed constructs with linker lengths in between the previously reported β‐0a‐α and β‐3a‐α constructs (Ahring et al., 2018). This seemed highly successful since co‐expression of β‐2a‐α and β‐1a‐α constructs with free α4^VFL subunits appeared to yield receptors that were exclusively assembled in the counterclockwise orientation as evidenced by no positive NS9283 efficacy. Furthermore, compared to β‐0a‐α concatemers, the expressed receptors resulted in higher maximal current amplitudes. However, by altering the experimental design and placing the reporter α4^VFL subunit within the dimeric constructs, confounding data emerged. Linkers with a TLL of 30 amino acids (i.e., 1a linker) and shorter did not only affect the assembly orientation, they also clearly resulted in assemblies in which only the dimer β2 subunit integrated into functional receptors. These observations corroborate modelling findings that short linkers introduce protein strain. Systematic shortening of the linker in dimeric concatemers from 3a towards 0a initially favours counterclockwise assembly, yet as linkers get shorter, the steric strain increases significantly, and full dimers fail to integrate.

Ironically, loop‐outs of the second dimer subunit were previously described for constructs with long linkers (Ahring et al., 2018; Groot‐Kormelink, Broadbent, Boorman, & Sivilotti, 2004). Thus, dimeric constructs inherently contain many degrees of freedom with respect to assembly, and it is impossible to fully exclude the existence of receptors with dangling subunits in a receptor pool. Therefore, dimeric constructs fundamentally represent an error‐prone strategy that cannot be recommended for demanding applications with ternary receptors. For less demanding applications or binary receptors, the β‐3a‐α and β‐2a‐α constructs represent reasonable compromises with different associated risk profiles: the former can lead to some clockwise assembly, and the latter can lead to a higher degree of loop‐outs of the α4 subunit (Figure 6).

FIGURE 6.

Guidelines for design of concatenated ternary nAChR constructs. Based on the collective information, this represents what we perceive as the overall best constructs from each of the investigated series. (α4)₂(β2)₂(α4^VFL)₁‐containing receptors are used as model constructs, but the findings are likely directly translatable to other nAChRs and translatable with minor modifications, that accommodate the difference in TM4 positions, to other Cys‐loop receptor classes. Translation of linker lengths to TLL calculations is found in Tables 1, 2, 3. Linkers for which no length is given should be longer (e.g., 36–45 amino acids). Preferred or exclusive counterclockwise assembly and key caveats of each construct are indicated. Recommended application is based on whether a given construct is suitable for expression of uniform pools of ternary receptors or not

4.3. Trimeric constructs

Concatemers based on trimeric constructs need two additional subunits to give a complete pentamer, which inherently limits the options for obtaining uniform ternary receptor pools from mixing with free subunits. Furthermore, given the overall design and assembly similarities, there is a high probability that trimeric constructs behave similarly to dimeric constructs with respect to bi‐directional assembly and potential loop‐outs, and we therefore avoided this strategy. While it might seem tantalising to use combinations of dimeric and trimeric constructs, our data with dimers suggest that this could give a staggering number of unexpected receptors. Yet there are examples of such experiments in the literature, so unless findings have been confirmed by other construct designs, data should be treated with utmost care (Botzolakis et al., 2016; Kaur, Baur, & Sigel, 2009).

4.4. Tetrameric constructs

An advantage of tetrameric constructs is that only one subunit is missing to complete the pentamer and such constructs could therefore be highly useful for studying the role of different free subunits. The tetrameric T(x) constructs with one short linker all led to efficient expression of functional receptors. While shortening the first linker in the T(x) concatemers increased the preference for counterclockwise assembly, none of them led to uniform receptor populations as shown by NS9283 efficacy values. When compared with the data for dimers, it appears that tetramers have a greater ability for overcoming the strain introduced by a single short linker and lower propensity for loop‐outs. Nonetheless, the T(x) constructs cannot be recommended for expression of ternary receptors.

Introducing a second short linker to give the T(x,x) constructs had profound effects. The T(3,3) concatemer resulted in a preferred counterclockwise assembly, and shortening the linkers to give T(2,2) resulted in a fully uniform receptor pool. This demonstrates that distributing the assembly‐directing properties between two short linkers in a tetrameric concatemer is a superior to relying on just one short linker. Furthermore, this strategy avoids having to use overly short linkers that introduce substantial strain even when assembled in the shorter counterclockwise orientation. Hence, tetrameric concatemers with two controlling linkers represent good strategies for expressing ternary receptors, and T(2,2) and T(3,3) constructs are suitable depending on whether a fully uniform receptor pool or higher current amplitudes are needed (Figure 6). It is important to note that while the assembly orientation caveat was addressed with two short linkers, all tetrameric concatemers retained the ability to express functional receptors with themselves to a low degree.

4.5. Pentameric constructs

Pentameric constructs intuitively appear advantageous for obtaining desired receptors since they consist of all subunits necessary to complete a receptor complex. Comparing with tetrameric T(x) constructs, a single short linker was observed to exert greater overall influence on the assembly orientation of pentameric P(x) concatemers. While α4^VFL‐containing P(3), P(2), and P(1) concatemers displayed NS9283 efficacy just within or below the intermediate range, only some oocytes showed positive response with the expected potency suggesting a preference for counterclockwise assembly. No robust NS9283 efficacy was observed in any of the tested oocytes with the shortest α4^VFL‐containing P(0) concatemer suggesting a uniform receptor pool. As expected from the data with tetrameric constructs, introducing a second short linker increased control of the assembly orientation, and exclusive counterclockwise assembly was observed for both α4^VFL‐containing P(3,3) and P(2,2) concatemers.

With pentameric constructs, it was thus possible to obtain a uniform receptor pool using either one short linker in the P(0) concatemer or two short linkers in the P(3,3) concatemer. Since the 0a linker in the P(0) concatemer is likely to introduce significant protein strain, this constitutes the less ideal option. Hence, for applications where a preferred counterclockwise assembly is sufficient, a construct such as P(3) is useful, but for demanding applications with ternary receptors, the P(3,3) construct is the best strategy (Figure 6).

In conclusion, this study has demonstrated that efficient expression of uniform pools of ternary nAChRs can be obtained with tetrameric and pentameric construct designs, by using two short assembly‐directing linkers. Due to many degrees of freedom during assembly, we found that strategies involving simpler dimeric constructs should be avoided for expression of complex receptors. While the experiments were performed using α4β2 nAChRs as a model system, the findings are likely to be applicable to all members of the Cys‐loop receptor family. Hence, the guidelines presented here will significantly aid future analysis of a wide range of unique receptor subtypes, which is important for several reasons. Recent genetic screening has, for example, revealed that variants in a range of Cys‐loop receptor subunits are involved in human diseases such as epilepsy and many Cys‐loop receptors are furthermore established targets for drug intervention. Thus, our findings not only enable precise functional analysis of clinically relevant variants but can also be utilised to develop new understanding of how pharmacological agents interact with specific binding sites.

AUTHOR CONTRIBUTIONS

P.K.A. designed the research and drafted the manuscript. V.W.Y.L., A.S.K., and P.K.A. performed the research. All authors analysed the data and approved the final version of the manuscript.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

DECLARATION OF TRANSPARENCY AND SCIENTIFIC RIGOUR

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Design and Analysis, and Animal Experimentation, and as recommended by funding agencies, publishers, and other organisations engaged with supporting research.

Liao VWY, Kusay AS, Balle T, Ahring PK. Heterologous expression of concatenated nicotinic ACh receptors: Pros and cons of subunit concatenation and recommendations for construct designs. Br J Pharmacol. 2020;177:4275–4295. 10.1111/bph.15188