Identification of critical functional determinants of kainate receptor modulation by auxiliary protein Neto2

Abstract

Key points

• Kainate receptors (KARs) are ionotropic glutamate receptors (iGluRs) that modulate synaptic transmission and intrinsic neuronal excitability.

• KARs associate with the auxiliary proteins neuropilin‐ and tolloid‐like 1 and 2 (Neto1 and Neto2), which act as allosteric modulators of receptor function impacting all biophysical properties of these receptors studied to date.

• M3–S2 linkers play a critical role in KAR gating; we found that individual residues in these linkers bidirectionally influence Neto2 modulation of KAR desensitization in an agonist specific manner.

• We also identify the D1 dimer interface as a novel site of Neto2 modulation and functionally correlate the actions of Neto2 modulation of desensitization with modulation of cation sensitivity.

• We identify these domains as determinants of Neto2 modulation. Thus, our work contributes to the understanding of auxiliary subunit modulation of KAR function and could aid the development of KAR‐specific modulators to alter receptor function.

Abstract

Kainate receptors (KARs) are important modulators of synaptic transmission and intrinsic neuronal excitability in the CNS. Their activity is shaped by the auxiliary proteins Neto1 and Neto2, which impact KAR gating in a receptor subunit‐ and Neto isoform‐specific manner. The structural basis for Neto modulation of KAR gating is unknown. Here we identify the M3–S2 gating linker as a critical determinant contributing to Neto2 modulation of KARs. M3–S2 linkers control both the valence and magnitude of Neto2 modulation of homomeric GluK2 receptors. Furthermore, a single mutation in this domain abolishes Neto2 modulation of heteromeric receptor desensitization. Additionally, we found that cation sensitivity of KAR gating is altered by Neto2 association, suggesting that stability of the D1 dimer interface in the ligand‐binding domain (LBD) is an important determinant of Neto2 actions. Moreover, modulation of cation sensitivity was eliminated by mutations in the M3–S2 linkers, thereby correlating the action of Neto2 at these structurally discrete sites on receptor subunits. These results demonstrate that the KAR M3–S2 linkers and LBD dimer interface are critical determinants for Neto2 modulation of receptor function and identify these domains as potential sites of action for the targeted development of KAR‐specific modulators that alter the function of auxiliary proteins in native receptors.

Abbreviations

ATD

amino terminal domain

AMPAR

AMPA receptor

CUB

C1r/C1s‐Uegf‐BMP

eGFP

enhanced green florescent protein

iGluR

ionotropic glutamate receptor

KAR

kainate receptor

LBD

ligand binding domain

NMDAR

NMDA receptor

Neto

neuropilin‐ and tolloid‐like

TARP

transmembrane AMPA receptor regulatory protein

Introduction

Ionotropic glutamate receptors (iGluRs) are critical components of excitatory synaptic machinery; while similar structurally in their modular organization, physiologically the three primary types of iGluRs execute discrete roles in information processing (Traynelis et al. 2010). AMPA receptors (AMPARs) and NMDA receptors (NMDARs) are required for fast synaptic transmission and learning and memory, respectively (Huganir & Nicoll, 2013; Nicoll & Roche, 2013), whereas kainate receptors (KARs) are predominantly modulatory and serve in part to balance excitatory and inhibitory transmission (Contractor et al. 2011). These dissimilar functional roles are accompanied by subtle but distinctive biophysical properties that are further influenced by co‐assembly with auxiliary proteins, imparting yet another level of functional diversity.

KAR function and neuronal localization are altered by association with two auxiliary proteins known as neuropilin‐ and tolloid‐like (Neto) proteins. Neto proteins are single‐pass integral membrane proteins containing two extracellular C1r/C1s‐Uegf‐BMP (CUB) domains followed by a low density lipoprotein class A (LDLa) module, and an intracellular C‐terminus (Stöhr et al. 2002). Incorporation of Neto1 or Neto2 into KAR complexes changes every parameter of receptor function examined to date, including rates of desensitization, deactivation and recovery from desensitization, receptor open probability, pharmacological profiles, and the subcellular and synaptic localization of receptors (Zhang et al. 2009; Copits et al. 2011; Straub et al. 2011 a, b ; Tang et al. 2011 b; Wyeth et al. 2014). However, many aspects of the structural basis for auxiliary protein modulation of iGluRs remain obscure. Recent studies suggest that AMPAR auxiliary subunits alter the structure of the ligand‐binding domain (LBD) and that multiple domains, including the amino‐terminal domain (ATD), are sites of interaction between auxiliary and receptor subunits (Cais et al. 2014; MacLean et al. 2014; Shanks et al. 2014). Yet Neto proteins and the various AMPAR auxiliary proteins are unlike in structure, and thus their respective modulatory actions may arise from differential mechanistic activities or distinct modes of interactions with their cognate receptors.

In this study, we identified the M3–S2 linker and the LBD dimer interface as structural determinants of Neto2 modulation. We previously demonstrated that mutations in the KAR M3–S2 gating linker bidirectionally alter receptor desensitization in a manner reminiscent of auxiliary subunit modulation (Vivithanaporn et al. 2007). Indeed, two charged residues in these linkers were shown to impact several stages of the iGluR gating cycle, including activation, deactivation, and the rate and extent of desensitization (Yelshansky et al. 2004; Vivithanaporn et al. 2007; Harms et al. 2014; Kazi et al. 2014). Here we demonstrate that a number of other amino acids in the M3–S2 linker are critical determinants of desensitization and that this domain in part determines both the magnitude and the directionality of the Neto2 effects on KAR gating. Additionally, Neto2 association alters KAR cation sensitivity, consistent with a Neto2‐induced structural rearrangement at the level of the LBD. Neto2 modulation of these two domains is functionally correlated in that mutations in the M3–S2 linkers that impact Neto2 modulation of desensitization also alter Neto2 modulation of cation sensitivity. Taken together, these results provide key insights into the structural basis of KAR modulation by auxiliary subunits.

Methods

Constructs and materials

Rat Neto2, myc‐GluK2a, and myc‐GluK5 were gifts from Susumu Tomita (Yale University School of Medicine, New Haven, CT, USA), Christophe Mulle (Université Bordeaux II, France) and John Marshall (Brown University, Providence, RI, USA), respectively. All mutants were created using a PCR‐based mutation protocol previously described (Swanson et al. 1997). All mutant cDNAs were sequenced at the Northwestern University Genomics Core Facility. The following antibodies were used: mouse anti‐HA (05‐904; Millipore) and rabbit anti‐myc (06‐549; Millipore).

Electrophysiology

Cell culture, transfection, whole‐cell patch‐clamp recordings and fast drug application to HEK293‐T/17 cells expressing recombinant receptors were performed as previously described (Copits et al. 2011). The external solution contained the following (in mM): 150 NaCl, 2.8 KCl, 2 CaCl₂, 1 MgCl₂, 10 glucose and 10 Hepes, adjusted to pH 7.3. The internal solution contained the following (in mM): 110 CsF, 30 CsCl, 10 Cs‐Hepes, 5 EGTA, 4 NaCl, 0.5 CaCl₂, adjusted to pH 7.3. For homomeric receptors, the transfection ratio was 0.5:1:3 for enhanced green florescent protein (eGFP), receptor and Neto cDNAs, respectively. For heteromeric receptors, the transfection ratio was 0.25:0.25:1:1 for eGFP, GluK2, GluK5 and Neto cDNAs, respectively. All cells were held at −70 mV. Rise times (10–90%) for whole‐cell currents evoked by fast application of glutamate (10 mM) or kainate (1 mM) ranged from 1 to 2 ms for whole‐cell recordings and 0.3 to 1 ms for outside‐out patches. Desensitization rates during 1 s application of glutamate (10 mM) or kainate (1 mM) were analysed by fitting currents averaged from four to six traces with single‐ or double‐exponential functions offline using Clampfit 10 software (Molecular Devices, Union City, CA, USA). For experiments with Cs⁺ substitution, external recording solutions contained equimolar concentrations (150 mM) of CsCl instead of NaCl. For low sodium experiments, solutions contained 50 mM NaCl and 100 mM sucrose. In Cs⁺ substitution and low Na⁺ experiments, desensitization of currents during 100 ms applications of glutamate were fitted with single exponential functions.

Mutant cycle analysis

Mutant cycle analysis was performed on macroscopic desensitization rates. Coupling energy (ΔΔG) was calculated as

$$\Delta \Delta G=RT\ln \left(\Omega \right)$$ (1)

where R is the ideal gas constant, T is the absolute temperature and Ω is defined as

$$\Omega =\frac{{{\tau }_{\mathrm{des}}}_{(\mathrm{wt}\mathrm{receptor})}\times {\tau }_{\mathrm{des}}{}_{(\mathrm{mutant}\mathrm{receptor}+\mathrm{Neto}2)}}{{{\tau }_{\mathrm{des}}}_{(\mathrm{wt}\mathrm{receptor}+\mathrm{Neto}2)}\times {{\tau }_{\mathrm{des}}}_{(\mathrm{mutant}\mathrm{receptor})}}.$$ (2)

If changes in desensitization caused by mutations in the GluK2 M3–S2 linker are functionally independent from Neto2 modulation of the desensitization time course, the coupling energy will be close to 0 kcal mol⁻¹. Conversely, energetic coupling between M3–S2 linker residues and Neto2 will yield a value that deviates from zero. In previous studies a ΔΔG of ∼0.5 kcal mol⁻¹ was interpreted as indicative of non‐additivity and therefore coupled interactions (Laha & Wagner, 2011). We have adopted the same approximate cut‐off value for analysis of ΔΔG in our studies.

Immunoprecipitation

Forty‐eight hours after transfection, HEK293/T‐17 cells in 100 mm dishes cells were washed once with phosphate‐buffered saline (PBS) and subsequently lysed in 500 μl of lysis buffer containing protease inhibitors (50 mM Tris, 150 mM NaCl, 2 mM 4‐(2‐aminoethyl) benzenesulfonyl fluoride (AEBSF), 0.3 μM aprotinin, 116 μM bestatin, 14 μM E‐64, 1 μM leupeptin and 1 mM EDTA). Crude protein lysates were centrifuged at 21,000 g for 25 min and supernatants containing solubilized membrane proteins of interest were collected. Lysates were incubated with 50 μl of 50% protein A/G‐agarose beads (Fisher Scientific, Pittsburgh, PA, USA) at 4°C for 1 h. Two micrograms of rabbit anti‐myc antibody and 50 μl of 50% protein A/G‐agarose beads were added to precleared supernatants and incubated at 4°C overnight. After three washes in lysis buffer, bound proteins were eluted from beads by boiling in 2× Laemmli (Bio‐Rad, Hercules, CA, USA) sample buffer for 5 min and separated by SDS‐PAGE. Proteins were electrotransferred onto PVDF membranes using a semidry process and were probed with a mouse anti‐HA antibody. Immunoreactive bands were visualized using horseradish peroxidase (HRP)‐conjugated anti‐mouse (Fisher Scientific) secondary antibody.

Statistical analysis

Summary data are presented as means ± SEM from n cells or patches. All electrophysiological data were assessed for statistical difference using a Kruskal–Wallis non‐parametric test with Dunn's multiple comparisons where data did not conform to a Gaussian distribution or had different standard deviations. Fold change data were log‐transformed and analysed using a one‐way ANOVA with Dunnett's post hoc test. Desensitization kinetics in Figs 6 and 7 were analysed using a one‐way ANOVA with Bonferroni's post hoc test. Unpaired Student's t tests were used on a small subset of data as noted. Statistical significance in each case was denoted as follows: *P < 0.05; **P < 0.01 and ***P < 0.001, and ****P < 0.0001. Statistical tests were performed using Prism 6.0 (GraphPad Software, La Jolla, CA, USA).

Figure 6. AMPAR reciprocal exchanges in the GluK2 M3–S2 linker of homomeric and heteromeric receptors .

A, left, representative, whole‐cell glutamate‐evoked peak amplitude‐scaled current traces from HEK293/T‐17 cells expressing GluK2 in the absence or presence of Neto2 (top), GluK2(E665V)+/−Neto2 (middle) and GluK2(E665V,D669E)+/−Neto2 (bottom). Right, quantification of desensitization rates measured from currents evoked by 1 s applications of glutamate to wild‐type and mutant receptors alone (top) and in the presence of Neto2 (middle), and quantification of average fold difference in Neto2 slowing of desensitization of wild‐type and mutant receptors in response to glutamate (bottom). Black bars indicate wild‐type GluK2, green bars indicate significant attenuation of desensitization or Neto2 modulation, and grey bars indicate mutants with no significant change, as compared to wild‐type. Circles represent individual data points. B, same as A except that receptors are heteromeric and also contain wild‐type or mutant GluK5. C, left, confirmation of expression of HA‐Neto2 in the transfected cells. Right, co‐immunoprecipitation of HA‐Neto2, wild‐type or mutant myc‐GluK2, and myc‐GluK5 from transfected HEK293/T‐17 cells. Values shown are means ± SEM. Statistical significance is denoted as follows: *P < 0.05, ***P < 0.001.

Figure 7. Neto2 reduces cation sensitivity of homomeric and heteromeric KARs .

Left, quantification of homomeric (A) and heteromeric (B) receptor glutamate‐evoked current response amplitudes in 150 mM external CsCl normalized to control currents evoked from the same patch or cell, respectively, in standard NaCl‐containing external solution in the absence (black bars) and presence (blue bars) of Neto2. Middle, quantification of desensitization rates of receptor currents in 150 mM NaCl and 150 mM CsCl external solution. Representative glutamate‐evoked current traces from outside‐out patches or whole‐cell expressing GluK2 (left, top), GluK2/Neto2 (right, top), GluK2/GluK5 (bottom, left), and GluK2/GluK5/Neto2 (bottom right). Black traces represent standard external solution and red traces represent external with equimolar substitution of CsCl. Responses in each set of traces are from the same patch. Arrows highlight the decrease in current amplitude in extracellular CsCl. Values shown are means ± SEM. Statistical significance is denoted as follows: *P < 0.05, ***P < 0.001, ****P < 0.0001.

Results

The extracellular domains of Neto proteins are predicted to consist largely of compact, tandem β‐sandwiches connected by a short linker, and thus are likely to influence KAR function through interactions with more membrane‐proximal components of receptor subunits. We hypothesized that the M3–S2 gating linker represented one such target of Neto2 modulation in the GluK2 KAR subunit. This linker is a critical component of the gating process in that it transduces agonist‐dependent conformational changes in LBD structure into movement of the M3 membrane domain required for pore opening; desensitization occurs as a function of rupture of the LBD dimer interface (and other structural changes), linker rearrangements and pore closing (Meyerson et al. 2014). The physical composition of the M3–S2 linker shapes gating phenomena (Yelshansky et al. 2004; Vivithanaporn et al. 2007; Kazi et al. 2014) and therefore could represent a potential site of action of auxiliary subunits like Neto proteins. To test this possibility, we altered the structure of the M3–S2 linker domain in the GluK2 KAR subunit and determined the impact on Neto2 modulation of homomeric and heteromeric receptor desensitization.

Alanine scanning mutation of the M3–S2 linker impacts homomeric GluK2 receptor desensitization

In order to test the role of the entire M3–S2 linker in Neto modulatory function, it was first necessary to determine how structural alteration impacts homomeric GluK2a receptor desensitization. In both KARs and AMPARs, mutations at E662 and R663 (E627 and R628 in AMPARs) have bidirectional effects on receptor desensitization (Yelshansky et al. 2004; Vivithanaporn et al. 2007; Dong & Zhou, 2011; Harms et al. 2014), yet the impact of mutations at other residues within the linker domain is not well defined, nor is it known if mutations have agonist‐specific effects. We therefore extended this analysis by substituting alanine for each residue in the GluK2 subunit corresponding to the extended M3–S2 linker found to the B/D subunits of the resolved structure of GluA2 AMPA receptor (Fig. 1), which are longer than those found in A/C‐positioned subunits (Sobolevsky et al. 2009). Recombinant receptors were transiently expressed in HEK293‐T/17 cells and activated rapidly by application of the full agonist glutamate (10 mM) or the partial agonist kainate (1 mM) to cells that had been lifted into laminar solution streams. Tau values derived from best fits of desensitizing currents for all GluK2 receptors (with and without the Neto2 auxiliary protein) can be found in Table 1.

Figure 1. M3–S2 linker structure .

A, left, crystal structure of the GluA2 receptor (Sobolevsky et al. 2009) showing proximal subunits in blue and distal subunits in green. M3–S2 linkers are shown in red. Side (middle) and top‐down (right) views of the M3–S2 linkers attached to the M3 helices highlighting the asymmetry of linkers in distal and proximal subunits. Red spheres indicate the position of the aspartic acid at the 634 position (D669 in GluK2 and E653 in GluK5). B, comparison of M3–S2 linker residues between GluK2, GluK5 and AMPARs. Number corresponds to GluK2 residue numbering. Blue and green boxes indicate sequence divergence between GluK2 and GluK5 and GluK2 and AMPARs, respectively.

Table 1.

Neto2 modulation of GluK2 homomeric receptor M3–S2 linker mutants

	Glutamate		Kainate
Receptor	i peak (−pA)	τdes(wtd) (ms)	i peak (−pA)	τdes(wtd) (ms)
GluK2	8549 ± 1021 (25)	4.5 ± 0.3	2166 ± 421 (8)	7.6 ± 1.9
GluK2+Neto2	6700 ± 633 (38)	19.2 ± 2.9	6150 ± 828 (12)	72.2 ± 21.0
Alanine mutants
GluK2(E662A)	7491 ± 1119 (13)	1.8 ± 0.2	1651 ± 449 (10)	1.8 ± 0.1
GluK2(E662A)+Neto2	7406 ± 1243 (15)	5.9 ± 0.3	3430 ± 538 (14)	4.0 ± 0.5
GluK2(R663A)	85 ± 18 (4)	7.8 ± 1.0	48 ± 5.5 (5)	210.3 ± 75.4
GluK2(R663A)+Neto2	727 ± 483 (5)	77.8 ± 11.3	205 ± 94 (5)	246.0 ± 16.8
GluK2(M664A)	1907 ± 882 (10)	17.5 ± 4.3	2884 ± 830 (10)	29.7 ± 10.4
GluK2(M664A)+Neto2	1751 ± 983 (10)	55.0 ± 7.1	305 ± 106 (8)	259.0 ± 39.4
GluK2(E665A)	4162 ± 927 (11)	5.0 ± 0.2	3077 ± 666 (9)	4.4 ± 0.3
GluK2(E665A)+Neto2	6726 ± 1223 (12)	15.2 ± 1.3	4353 ± 1086 (10)	32.6 ± 7.8
GluK2(S666A)	10626 ± 2307 (7)	4.7 ± 0.2	5284 ± 646 (10)	5.0 ± 1.0
GluK2(S666A)+Neto2	10327 ± 1372 (15)	19.1 ± 1.9	6755 ± 813 (15)	59.3 ± 9.0
GluK2(P667A)	3342 ± 1517 (7)	4.4 ± 0.4	1650 ± 517 (6)	3.3 ± 0.4
GluK2(P667A)+Neto2	5193 ± 1106 (15)	8.3 ± 0.2	1968 ± 584 (10)	10.9 ± 1.6
GluK2(I668A)	217 ± 36 (13)	2.6 ± 0.1	99 ± 15 (10)	1.7 ± 0.2
GluK2(I668A)+Neto2	4926 ± 1523 (11)	5.1 ± 0.3	2664 ± 584 (12)	9.0 ± 1.5
GluK2(D669A)	2629 ± 623 (11)	7.2 ± 0.8	3452 ± 572 (13)	9.1 ± 1.4
GluK2(D669A)+Neto2	7778 ± 2068 (10)	13.4 ± 2.4	5602 ± 712 (16)	90.0 ± 19.8
GluK2(P667A,I668A)	121 ± 22 (5)	1.6 ± 0.1	—	—
GluK2(P667A,I668A)+Neto2	2044 ± 743 (5)	3.5 ± 0.0	—	—
GluK2(P667A,I668A,D669A)	134 ± 22 (5)	1.5 ± 0.2	—	—
GluK2(P667A,I668A,D669A)+Neto2	2846 ± 504 (7)	3.1 ± 0.2	—	—
Dipole mutants
GluK2(E662R)	3347 ± 426 (8)	1.3 ± 0.0	—	—
GluK2(E662R)+Neto2	6121 ± 1108 (10)	5.2 ± 0.3	—	—
GluK2(R663E)	2766 ± 652 (4)	13.9 ± 2.3	—	—
GluK2(R663E)+Neto2	657 ± 248 (7)	146.3 ± 21.5	—	—
GluK2(E662R,R663E)	5389 ± 1313 (10)	3.8 ± 0.5	—	—
GluK2(E662R,R663E)	7959 ± 1880 (10)	40.8 ± 9.5	—	—
AMPAR mutants
GluK2(E665V)	663 ± 180 (10)	3.9 ± 0.5	—	—
GluK2(E665V)+Neto2	506 ± 84 (14)	8.8 ± 0.8	—	—
GluK2(D669E)	5077 ± 1113 (10)	5.0 ± 0.4	—	—
GluK2(D669E)+Neto2	5467 ± 683 (10)	14.8 ± 1.2	—	—
GluK2(E665V,D669E)	7428 ± 1080 (10)	3.5 ± 0.1	—	—
GluK2(E665V,D669E)+Neto2	1244 ± 261 (10)	10.7 ± 0.4	—	—

Wild‐type and M3–S2 linker mutant GluK2 homomeric receptors with and without the auxiliary subunit Neto2 were expressed in HEK293/T‐17 cells and recorded using whole‐cell patch clamp. Summary of peak current amplitude and mean weighted desensitization rates from are shown for 1 s applications of glutamate and kainate. For each receptor and agonist application, the mean ± SEM peak current amplitude value is shown, with the number of cells recorded in parentheses.

Several alanine‐substituted homomeric GluK2 KARs exhibited altered desensitization kinetics. In line with previous results, alanine substitution at E662 and R663 sped and slowed GluK2 receptor desensitization, respectively, in response to both glutamate and kainate application (Vivithanaporn et al. 2007). GluK2(E662A) receptors responded identically to glutamate and kainate application, desensitizing at a rate of 1.8 ± 0.1 ms (n = 14, P < 0.01 compared to wild‐type GluK2, Kruskal–Wallis test with Dunn's multiple comparison test) and 1.8 ± 0.1 ms (n = 10, P < 0.05; Fig. 2), respectively. In response to glutamate, GluK2(R663A) receptors desensitized at a rate of 8.9 ± 1.4 ms (n = 5); however, kainate‐evoked currents from this receptor were very slow, with an average desensitization time constant of 210.3 ± 75.4 ms (n = 5; Fig. 2). Alanine substitution at I668, like E662, also sped receptor desensitization of glutamate‐evoked currents (2.8 ± 0.1 ms, n = 9), similar to what was observed in AMPARs at the analogous site (Chen et al. 2014), in addition to speeding kainate‐evoked responses (1.8 ± 0.2 ms, n = 10, P < 0.05). The impact of alanine substitution at M664 and D669 was similar to R663, slowing receptor desensitization in response to glutamate to 17.5 ± 4.3 ms (n = 10, P < 0.01) and 7.5 ± 0.9 ms (n = 11), respectively (Fig. 2 A). Kainate‐evoked currents from these two mutants were also slowed to similar degrees (GluK2(M664A), 29.7 ± 10.4 ms, n = 10; GluK2(D669A), 9.1 ± 1.4 ms, n = 13; Fig. 2 B). Alanine substitution at E665, S666 and P667 did not impact basal receptor desensitization in response to either glutamate or kainate application. Taken together, these data confirm that the some (but not all) residues in M3–S2 linkers bidirectionally influence KAR desensitization.

Figure 2. Agonist dependent contributions of M3–S2 linker residues to Neto2 modulation of GluK2 desensitization .

A, left, representative whole‐cell, peak amplitude‐scaled current traces from HEK293/T‐17 cells expressing GluK2 in the absence and presence of Neto2 (top), GluK2(R663A)+/−Neto2 (middle) or GluK2(P667A)+/−Neto2 (bottom). Glutamate (10 mM) was applied for 100 ms. Black traces represent receptor alone; blue traces represent receptor co‐assembled with Neto2. Right, quantification of desensitization rates measured from currents evoked by 1 s applications of glutamate to wild‐type and M3–S2 linker alanine mutant GluK2 receptors alone (top) and in the presence of Neto2 (bottom). Bottom, Quantification of average fold difference in Neto2 slowing of desensitization of wild‐type and mutant GluK2 receptors in response to glutamate and kainate. Black bars indicate wild‐type GluK2, red bars indicate significant enhancement of desensitization or Neto2 modulation, green bars indicate significant attenuation of desensitization or Neto2 modulation, grey bars indicate mutants with no significant change as compared to wild‐type. Circles represent individual data points. B, same as A except that 1 mM kainate was applied to cells. Values shown are means ± SEM. Statistical significance is denoted as follows: *P < 0.05, **P < 0.01, ***P < 0.001.

M3–S2 linker residues impact Neto2 modulation of GluK2a receptors in an agonist‐dependent manner

Mutations in the M3–S2 linker also had bidirectional effects on Neto2 modulation of glutamate and kainate‐evoked current desensitization. Neto2 co‐expression slowed the time course of desensitization of wild‐type GluK2 receptors in response to glutamate application to by ∼5‐fold (from 4.5 ± 0.2 ms for GluK2 to 21.6 ± 3.8 ms with GluK2/Neto2; n = 28 and 38, respectively; Fig. 2 A) (as first reported by Zhang et al. 2009). Four of the eight M3–S2 linker alanine mutants deviated from this 5‐fold effect (Fig. 2 A). Alanine substitution of R663 significantly enhanced Neto2 slowing of desensitization (GluK2(R663A)/Neto2, 78.0 ± 11.3, 8.7 ± 1.3‐fold change, n = 9, P < 0.001; Fig. 2 A). Conversely, alanine substitution of P667, I668 and D669 attenuated Neto2 modulation of desensitization (GluK2(P667A)/Neto2 8.0 ± 0.2 ms, 1.7 ± 0.1‐fold change, n = 15; GluK2(I668A), 5.1 ± 0.4 ms, 1.8 ± 0.1‐fold change, n = 13; GluK2(D669A), 13.4 ± 2.4 ms, 1.8 ± 0.3‐fold change, n = 10, all P < 0.001; Fig. 2 A). Introduction of multiple alanine substitutions in GluK2(P667A,I668A) and GluK2(P667A,I668A,D669A) had no greater impact on Neto2 modulation of desensitization than their single mutant counterparts (Table 1). Alanine substitution at E662, M664, E665 and S666 did not have a significant impact on Neto2 modulation of GluK2 receptor desensitization in response to glutamate.

To determine if the diverse changes in receptor desensitization introduced by mutation of linker residues and co‐assembly with Neto2 were energetically coupled or independent, we applied mutant cycle analysis (using eqns (1) and (2)) on receptor desensitization rates from wild‐type GluK2, wild‐type GluK2/Neto2, mutant GluK2, and mutant GluK2/Neto2 receptors (Fig. 3 A); macroscopic rate constants were used previously to test for energetic coupling in a similar fashion (Carbone & Plested, 2012). A coupling constant (ΔΔG) of zero is indicative of independent, and therefore additive, effects of the mutation and co‐assembly with Neto2, whereas divergence from zero quantifies the energetic coupling between that residue and Neto2 modulation (as described in principle by Hidalgo & MacKinnon (1995)). Alanine substitution at P667, I668 and D669 resulted in negative energetic coupling values (GluK2(P667A), ΔΔG = −0.56 kcal mol⁻¹; GluK2(I668A), ΔΔG = −0.54 kcal mol⁻¹; GluK2(D669A), ΔΔG = −0.57 kcal mol⁻¹; Table 2). GluK2(P667A) is shown as an example in Fig. 3 A. Thus in wild‐type receptors, P667, I668 and D669 stabilize Neto2 modulation of GluK2 receptor glutamate‐evoked currents. We obtained positive energetic coupling between GluK2(R663A) and Neto2 (ΔΔG = 0.44 kcal mol⁻¹, Fig. 3 A and Table 2). No energetic coupling was observed for E662A, M664A, E665A and S666A, supporting the lack of an effect on Neto2 modulation observed for these mutant receptors (Table 2).

Figure 3. Homomeric wild‐type and mutant GluK2 desensitization rates do not correlate with the magnitude of Neto2 modulation .

A and B, diagrams of thermodynamic cycles used in mutant cycle analysis of glutamate‐ (A) and kainate‐ (B) evoked current desensitization rates for wild‐type and mutant receptors with and without Neto2. C, current desensitization rates for wild‐type and mutant receptors plotted against the fold difference in Neto2 modulation for glutamate‐ (top) and kainate‐ (bottom) evoked currents. Black circle is wild‐type GluK2, red circle is a mutant with enhanced Neto2 modulation, green circles are mutants with attenuated Neto2 modulation, and grey circles are mutants with no significant change in Neto2 modulation. D, left, confirmation of expression of HA‐Neto2 in the transfected cells. Right, co‐immunoprecipitation of HA‐Neto2 and wild‐type myc‐GluK2 or alanine mutants with attenuated Neto2 modulation from transfected HEK293/T‐17 cells.

Table 2.

Mutant cycle analysis of M3–S2 linker mutant KARs showing values of ΔΔG (kcal mol⁻¹)

Receptor	Glutamate	Kainate
Homomeric receptors
E662A	−0.22	−0.87
R663A	0.44	−1.25
M664A	−0.25	−0.05
E665A	−0.27	−0.15
S666A	−0.1	0.13
P667A	−0.56	−0.61
I668A	−0.54	−0.35
D669A	−0.57	0.02
P667A,I668A	−0.47	—
P667A,I668A,D669A	−0.5	—
E662R	−0.11	—
R663E	0.47	—
E662R,R663E	0.48	—
E665V	−0.45	—
D669E	−0.29	—
E665V,D669E	−0.27	—
Heteromeric receptors
E662A	−0.31	—
R663A	0.04	—
M664A	0.32	—
E665A	−0.4	—
S666A	−0.1	—
P667A	−0.41	—
I668A	−0.59	—
D669A	−0.1	—
P651A,P667A	−0.68	—
E665V	−1.01	—
D669E	−0.53	—
E665V,D669E	−0.05	—

Mutant cycle analysis was performed on macroscopic desensitization rates from mutant homomeric and heteromeric KARs. A ΔΔG value ∼0.5 kcal mol⁻¹ was set as the cutoff for non‐additivity indicative of energetic coupling.

Neto2 association slows desensitization of homomeric GluK2 receptor currents evoked by the partial agonist kainate (1 mM) by nearly 10‐fold (GluK2: 7.6 ± 1.9 ms, n = 8; GluK2/Neto2: 72.1 ± 21.0 ms, n = 12; fold change 9.5 ± 2.8, Fig. 2 B). We therefore investigated whether the difference in magnitude of Neto2 modulation of glutamate‐ and kainate‐evoked currents is accompanied by a differential engagement of M3–S2 linker residues. Similar to glutamate‐evoked currents, alanine substitution of M664, E665 and S666 had no impact on Neto2 modulation of kainate‐evoked currents (Fig. 2 B and Table 2). GluK2(E662A) receptors, however, exhibited reduced Neto2 modulation of kainate‐evoked currents (GluK2(E662A), 1.8 ± 0.1 ms, n = 10; GluK2(E662A)/Neto2, 4.0 ± 0.5 ms, n = 14; fold change 2.2 ± 0.3, P < 0.001) and negative energetic coupling was observed (ΔΔG = −0.87 kcal mol⁻¹). Thus, though E662 has little impact on Neto2 modulation of glutamate‐evoked currents (Fig. 2 A) it stabilizes Neto2 modulation of kainate‐evoked currents. Furthermore, while alanine substitution at 663 had a stabilizing effect on Neto2 modulation of glutamate‐evoked currents, a destabilizing effect on Neto2 modulation was observed for kainate‐evoked currents (ΔΔG = −1.25 kcal mol⁻¹), which desensitized at similar rates in the presence and absence of the auxiliary protein (GluK2(R663A), 210 ± 75 ms, n = 5; GluK2(R663A)/Neto2, 246 ± 17 ms, n = 5; fold change 1.2 ± 0.1, P < 0.001; Figs 2 B and 3 A). Alanine substitution at P667 resulted in attenuated Neto2 modulation of kainate‐evoked currents (fold change = 3.3 ± 0.5, n = 10, P < 0.5) and negative energetic coupling (ΔΔG = −0.61 kcal mol⁻¹; Fig. 3 B and Table 2) and therefore had the same impact on Neto2 modulation for currents evoked by both agonists. GluK2(I668A) and GluK2(D669A) receptors also exhibited agonist‐dependent sensitivity to Neto2; we found no impact of these mutations on Neto2 modulation of kainate‐evoked currents despite their relevance to Neto2 slowing of glutamate currents (Fig. 2 B and Tables 1 and 2). Taken together these results indicate Neto2 modulation of GluK2 desensitization engages individual M3–S2 linker residues in an agonist‐dependent fashion, which may contribute to the differences in magnitude of Neto2 modulation observed for full versus partial agonists.

These experiments demonstrated that alanine mutation of M3–S2 linker residues can enhance, attenuate, or have no effect on the normal slowing of GluK2 KAR desensitization that results from co‐assembly with Neto2. We also examined the relationship between the Neto2‐dependent fold slowing of GluK2 wild‐type and mutant receptor desensitization and the time constant for homomeric GluK2 receptors lacking Neto2. As shown in Fig. 3 C, there is no evidence of a correlation between these parameters, again suggesting that the impact of M3–S2 linker mutations resulted from specific alteration of the efficacy of modulation by Neto2. This conclusion is underscored by data from two specific examples, the GluK2(P667A) and GluK2(M664A) mutant receptors. GluK2(P667A) desensitizes with a similar time course as wild‐type receptors but GluK2(P667A)/Neto2 desensitizes faster than GluK2/Neto2, resulting in a reduced fold modulation of desensitization by Neto2 and a negative coupling constant. Conversely, GluK2(M664A) receptors exhibit a slower desensitization rate than GluK2 receptors, but the fold slowing of desensitization induced by Neto2 was equivalent to wild‐type receptors and no energetic coupling was observed. Finally, the reduction in Neto2 modulation with the three most markedly attenuated alanine mutants occurred despite biochemical association of the myc‐tagged GluK2 mutant subunits and HA‐tagged Neto2 protein in equilibrium co‐immunoprecipitation experiments from whole‐cell lysates (Fig. 3 D), with the caveats that these experiments do not rule out the possibility of altered rates of oligomerization and the proteins analysed were not located exclusively on the plasma membrane. In summary, we conclude that alteration of the structure of M3–S2 linker profoundly impacts Neto2 modulation of homomeric GluK2 KARs.

Role of electrostatic interactions in Neto2 modulation

Previous studies with both AMPARs and KARs suggest that electrostatic interactions involving the charged linker residues at E662 and R663 influence receptor gating (Yelshansky et al. 2004; Vivithanaporn et al. 2007; Dong & Zhou, 2011; Harms et al. 2014). To determine if Neto2 modulation is impacted by the electrostatic dipole created by these two residues, we measured Neto2 modulation of GluK2(E662R), GluK2(R663E), or GluK2(E662R,R663E) receptor desensitization (Fig. 4 and Table 1). Similar to alanine substitutions at these residues, charge reversal at E662 sped receptor desensitization (1.34 ± 0.04 ms, n = 8, P < 0.001), whereas reversal of the positive charge at R633 slowed receptor desensitization (13.9 ± 2.3 ms, n = 4). Furthermore, Neto2 modulation was intact for GluK2(E662R) receptors (3.7 ± 0.2‐fold change, n = 10) and no energetic coupling was observed (ΔΔG = −0.11 kcal mol⁻¹), whereas it was significantly enhanced for GluK2(R663E) receptors (12.3 ± 2.4‐fold change, n = 5, P < 0.001; ΔΔG = 0.47 kcal mol⁻¹; Table 2). The double mutant GluK2(E662R,R663E) also had enhanced Neto2 modulation (40.8 ± 9.5 ms, 10.7 ± 2.5‐fold change, n = 10, P < 0.001) and the coupling between E663 and Neto2 modulation was maintained (ΔΔG = 0.48 kcal mol⁻¹) without alteration of homomeric receptor desensitization kinetics (3.8 ± 0.5 ms, n = 10); thus, in addition to alanine substitution, insertion of a negative charge at R663 potentiates Neto2 modulation of KAR desensitization. The charge balance between R662 and E663, however, does not seem to be a critical factor in Neto2 modulation.

Figure 4. Charge inversion at R663 enhances Neto2 modulation .

Left, representative whole‐cell, glutamate‐evoked (10 mM) peak amplitude‐scaled current traces from HEK293/T‐17 cells expressing GluK2 in the absence or presence of Neto2 (top), GluK2(E662R)+/−Neto2 (upper middle), GluK2(R663E)+/−Neto2 (lower middle) and GluK2(E662R,R663E)+/−Neto2 (bottom) receptors. Black traces represent receptor alone, blue traces represent receptor co‐assembled with Neto2. Glutamate was applied for 100 ms. Right, quantification of desensitization rates measured from currents evoked by 1 s applications of glutamate to wild‐type and mutant GluK2 receptors alone (top) and in the presence of Neto2 (middle), and quantification of average fold difference in Neto2 slowing of desensitization of wild‐type and mutant GluK2 receptors in response to glutamate (bottom). Black bars indicate wild‐type GluK2, red bars indicate significant enhancement of desensitization or Neto2 modulation, green bars indicate significant attenuation of desensitization or Neto2 modulation, grey bars indicate mutants with no significant change, as compared to wild‐type. Circles represent individual data points. Values shown are means ± SEM. Statistical significance is denoted as follows: ***P < 0.001.

GluK2 M3–S2 linker mutants attenuate Neto2 modulation of heteromeric KARs

In many types of heteromeric KARs, the component subunits have distinct pharmacological and biophysical properties that produce unusual gating behaviour, such as partial activation of discrete subunits, profound concentration dependence of gating, or tail currents upon deactivation (Swanson et al. 2002; Barberis et al. 2008; Mott et al. 2010; Fisher & Mott, 2013). To determine the importance of the M3–S2 linker in desensitization of heteromeric KARs containing the GluK2 and GluK5 subunits, we recorded glutamate‐evoked currents from GluK2/GluK5 wild‐type and M3–S2 linker alanine mutants alone and with Neto2 (Fig. 5 and Table 3).

Figure 5. Neto2 modulation of GluK2 M3–S2 linker alanine mutants in heteromeric receptors .

Left, representative whole‐cell, glutamate‐evoked (10 mM) peak amplitude‐scaled current traces from HEK293/T‐17 cells expressing GluK2/GluK5 in the absence or presence of Neto2 (top), GluK2(E665A)/GluK5+/−Neto2 (upper middle), GluK2(I668A)/GluK5+/−Neto2 (lower middle) and GluK2(P667A)GluK5(P651A)+/−Neto2 (bottom) receptors. Black traces represent receptor alone, blue traces represent receptor co‐assembled with Neto2. Glutamate was applied for 100 ms. Right, quantification of desensitization rates measured from currents evoked by 1 s applications of glutamate to wild‐type and mutant receptors alone (top) and in the presence of Neto2 (middle), and quantification of average fold difference in Neto2 slowing of desensitization of wild‐type and mutant receptors in response to glutamate (bottom). Black bars indicate wild‐type GluK2, red bars indicate significant enhancement of desensitization or Neto2 modulation, green bars indicate significant attenuation of desensitization or Neto2 modulation, grey bars indicate mutants with no significant change, as compared to wild‐type. Circles represent individual data points. Values shown are means ± SEM. Statistical significance is denoted as follows: ***P < 0.001.

Table 3.

Neto2 modulation of GluK2/GluK5 heteromeric receptor M3–S2 linker mutants

	Glutamate
Receptor	i peak (−pA)	τdes(wtd) (ms)	τfast	τslow	% fast	% slow
GluK2‐5	428 ± 86 (18)	2.0 ± 0.1	2.0 ± 0.1	—	100	—
GluK2‐GluK5+Neto2	1227 ± 313 (19)	11.3 ± 2.1	3.6 ± 0.6	75.9 ± 8.0	88.8 ± 1.2	11.2 ± 1.2
Alanine mutants
GluK2(E662A)/GluK5	294 ± 49 (5)	1.8 ± 0.7	1.8	—	100	—
GluK2(E662A)/GluK5+Neto2	895 ± 191 (8)	6.1 ± 1.1	2.3 ± 0.2	78.7 ± 5.2	91.4 ± 1.0	8.6 ± 1.0
GluK2(R663A)/GluK5	213 ± 89 (6)	2.4 ± 0.4	2.4 ± 0.4	—	100	—
GluK2(R663A)/GluK5+Neto2	1341 ± 308 (6)	14.4 ± 2.8	2.8 ± 0.2	132.4 ± 15.6	88.8 ± 1.0	11.2 ± 1.0
GluK2(M664A)/GluK5	293 ± 66 (6)	2.2 ± 0.3	2.2 ± 0.3	—	100	—
GluK2(M664A)/GluK5+Neto2	682 ± 209 (9)	21.1 ± 5.4	2.1 ± 0.1	150.7 ± 14.6	84.1 ± 2.3	15.9 ± 2.3
GluK2(E665A)/GluK5	599 ± 67 (8)	5.9 ± 1.4	2.5 ± 0.2	52.7 ± 9.5	85.6 ± 0.8	14.4 ± 0.8
GluK2(E665A)/GluK5+Neto2	1938 ± 419 (9)	17.2 ± 1.4	4.3 ± 0.7	52.5 ± 3.6	73.1 ± 1.9	26.9 ± 1.9
GluK2(S666A)/GluK5	2093 ± 486 (5)	3.2 ± 0.3	3.2 ± 0.3	—	100	—
GluK2(S666A)/GluK5+Neto2	2603 ± 1223 (7)	15.3 ± 1.1	4.1 ± 1.2	76.2 ± 5.5	84.2 ± 1.7	15.8 ± 1.7
GluK2(P667A)/GluK5	289 ± 81 (9)	2.5 ± 0.6	1.9 ± 0.2	65.6*	99.1 ± 0.9	0.9 ± 0.9
GluK2(P667A)/GluK5+Neto2	649 ± 145 (8)	7.1 ± 1.1	2.5 ± 0.2	56.5 ± 9.9	85.7 ± 2.2	14.3 ± 2.2
GluK2(I668A)/GluK5	205 ± 41 (9)	2.4 ± 0.2	2.4 ± 0.2	—	100	—
GluK2(I668A)/GluK5+Neto2	320 ± 100 (8)	5.1 ± 0.6	3.2 ± 0.2	36.3 ± 4.8	87.0 ± 1.5	13.0 ± 1.5
GluK2(D669A)/GluK5	426 ± 129 (8)	2.3 ± 0.4	2.0 ± 0.2	51.4	94	6
GluK2(D669A)/GluK5+Neto2	1909 ± 467 (7)	11.1 ± 1.7	4.1 ± 0.6	75.3 ± 11.9	87.6 ± 1.3	12.4 ± 1.3
GluK2(P667A)/GluK5(P651A)	313 ± 61 (8)	2.1 ± 0.1	2.1 ± 0.1	—	100	—
GluK2(P667A)/GluK5(P651A)+Neto2	622 ± 153 (10)	3.8 ± 0.4	2.5 ± 0.2	31.6 ± 7.7	88.1 ± 1.9	11.9 ± 1.9
AMPAR mutants
GluK2(E665V)/GluK5	663 ± 180 (10)	2.3 ± 0.3	2.3 ± 0.3	—	100	—
GluK2(E665V)/GluK5+Neto2	506 ± 84 (13)	2.4 ± 0.2	2.4 ± 0.2	—	100	—
GluK2(D669E)/GluK5	139 ± 64 (4)	2.8 ± 0.3	3.1 ± 0.1	—	100	—
GluK2(D669E)/GluK5+Neto2	413 ± 221 (7)	6.5 ± 0.9	2.0 ± 0.2	56.7 ± 8.3	88.9 ± 1.1	11.2 ± 1.1
GluK2(E665V,D669E)/GluK5	405 ± 110 (8)	1.8 ± 0.1	1.8 ± 0.1	—	100	—
GluK2(E665V,D669E)/GluK5+Neto2	1244 ± 261 (10)	9.3 ± 2.2	2.8 ± 0.3	63.5 ± 7.7	85.8 ± 3.1	14.2 ± 3.1

Wild‐type and M3–S2 linker mutant GluK2/GluK5 heteromeric receptors with and without the auxiliary subunit Neto2 were expressed in HEK293/T‐17 cells and recorded using whole‐cell patch clamp. Summary of peak current amplitude and mean weighted desensitization rates from are shown for 1 s applications of glutamate. Individual fast and slow component desensitization rates and contributions are also shown. For each receptor, the mean ± SEM peak current amplitude value is shown, with the number of cells recorded in parentheses.

As with GluK2 homomeric receptors, Neto2 slows the weighted desensitization rate of GluK2/GluK5 heteromeric receptors (GluK2/GluK5, 2.0 ± 0.1 ms, n = 18; GluK2/GluK5/Neto2, 11.3 ± 2.1 ms, n = 19; fold change 5.7 ± 1.1) (Straub et al. 2011 b); this effect primarily arises from an increase in a slowly desensitizing component that is not detectable in the absence of Neto2 (GluK2/Neto2 τ_des fast, 3.6 ± 0.6, 88.8 ± 1.2%; τ_des slow = 75.9 ± 8.0 ms, 11.2 ± 1.2%, slow component detected in 16 of 19 recordings; Table 3). M3–S2 linker alanine mutations did not impact heteromeric GluK2/GluK5 receptor glutamate‐evoked desensitization except in the case of GluK2(E665A)/GluK5 receptors. These receptors desensitized at a rate of 5.9 ± 1.4 ms (n = 8, P < 0.05; Fig. 5 and Table 3) due to the presence of a slowly desensitizing component detected in half of the cells recorded (τ_des fast, 2.5 ± 0.2 ms, 85.6 ± 0.8%, n = 8; τ_des slow, 52.7 ± 9.5 ms, 14.4 ± 0.8%, detected in 4 of 8 recordings).

Neto2 modulation of heteromeric KAR desensitization was largely unaffected by GluK2 M3–S2 linker alanine substitutions (Fig. 5); only alanine substitution of I668 caused a significant attenuation of Neto2 slowing of the weighted τ_des (GluK2(I6668A)/GluK5, 2.4 ± 0.2 ms, n = 9; GluK2(I668A)/GluK5/Neto2, 5.1 ± 0.6 ms, n = 8; 2.2 ± 0.3‐fold change, P < 0.05). Accordingly, a negative coupling energy of −0.59 kcal mol⁻¹ was obtained in the mutant cycle analysis (Table 2). Though the magnitude of Neto2 modulation of GluK2(M664A)/GluK5 receptors was enhanced (GluK2(M664A)/GluK5, 2.2 ± 0.3 ms, n = 6; GluK2(M664A)/GluK5/Neto2, 21.1 ± 5.4 ms, fold change = 9.6 ± 2.4, n = 9), this effect was not significant when compared statistically to Neto2 modulation of wild‐type GluK2/GluK5 receptors. What variability occurs with this mutant appears to arise from inconstant contributions of a slowly desensitizing component to the net time course of current decay (GluK2/GluK5/Neto2, τ_des slow 75.9 ± 8.0 ms, 11.2 ± 1.2%, detected in 16 of 19 recordings; GluK2(M664)/GluK5/Neto2, 150.7 ± 14.6 ms, 15.9 ± 2.3%, n = detected in 7 of 9 recordings, P < 0.001).

In these receptors, wild‐type or mutant GluK2 subunits comprise two of the four component subunits, based on current models of 2:2 subunit stoichiometry of heteromeric KARs (Kumar et al. 2011; Reiner et al. 2012). GluK2(P667A)/GluK5 receptors show a 50% reduction in Neto2 modulation (Fig. 5). We introduced an analogous mutation in GluK5, P651A, to test if an assembly with all four linkers mutated at this position would exhibit a more profound impact on Neto2 modulation. GluK2(P667A)/GluK5(P651A) receptors desensitized with a time course of 2.1 ± 0.1 ms (n = 8), and GluK2(P667A)/GluK5(P651)/Neto2 receptors indeed showed attenuated Neto2 modulation (3.8 ± 0.4 ms, n = 10) (Fig. 5), desensitizing 1.8 ± 0.2‐fold slower than receptors lacking Neto2 (P < 0.01). Furthermore, the slowly desensitizing component of the current decay was significantly faster than wild‐type receptors, with a desensitization rate of 32 ± 8 ms (P < 0.05). A concomitant lower ΔΔG for the double mutant as compared to the single mutant was observed (ΔΔG = −0.41 kcal mol⁻¹ and −0.68 kcal mol⁻¹, respectively; Table 2). Thus, GluK5 M3–S2 linkers also influence Neto2 modulation and contribute to the slowly desensitizing component of the current decay; however, generation of a fully substituted alanine mutant, at least at the 667 position, does not completely abrogate Neto2 actions on heteromeric KARs.

Substitution of a residue found in AMPA receptors abolishes Neto2 modulation of heteromeric receptors

The M3–S2 linker in AMPAR subunits diverges from that of KARs at two positions, V630 and E634, which are E665 and D669 in the GluK2 subunit (Fig. 1 B). To test if the two divergent residues in the M3–S2 linker contribute to the receptor‐type specificity of Neto2 modulation, we recorded from GluK2(E665V), GluK2(D669E), and GluK2(E665V,D669E) mutant receptors in the absence and presence of Neto2 (Fig. 6 A and Table 1). Valine substitution at E665 did not impact KAR desensitization (3.9 ± 0.5 ms, n = 10) but partially reduced Neto2 modulation (GluK2(E665V)/Neto2, 8.8 ± 0.8 ms, n = 13; fold change 2.2 ± 0.2, P < 0.001; Fig. 6 A) and resulted in energetic coupling in the mutant cycle analysis (ΔΔG = −0.45 kcal mol⁻¹; Table 2). In contrast, desensitization of GluK2(D669E) mutant receptors was similar to wild‐type GluK2 in both the absence (5.0 ± 0.4 ms, n = 10) and the presence of Neto2 (14.8 ± 1.2 ms, n = 10, fold change 3.0 ± 0.2), as was the double mutant receptor (GluK2(E665V,D669E), 3.5 ± 0.1 ms, n = 10; GluK2(E665V,D669E)/Neto2: 10.8 ± 0.4 ms, n = 10, fold change 3.1 ± 0.1; Fig. 6 A). Thus, generation of a more ‘AMPAR‐like’ M3–S2 linker either had a greater effect than alanine substitution (E665V) on Neto2 modulation or had no effect on receptor function with a more conservative replacement (D669E).

We also examined Neto2 modulation of these AMPAR substitution mutants in heteromeric receptors (Fig. 6 B and Table 3). Surprisingly, Neto2 failed to change desensitization kinetics of the GluK2(E665V)/GluK5 mutant receptor despite intact biochemical association of the myc‐tagged GluK2 mutant subunits, GluK5 and HA‐tagged Neto2 protein in whole‐cell lysates (Fig. 6 C). GluK2(E665V)/GluK5 receptors desensitized with a time course similar to wild‐type receptors (2.3 ± 0.3 ms, n = 10), and this was unchanged by addition of the Neto2 auxiliary subunit (2.4 ± 0.2 ms, n = 13, fold change 1.0 ± 0.1, P < 0.001). A large ΔΔG of −1.01 kcal mol⁻¹ was obtained for V665, supporting the destabilization of Neto2 modulation by this mutation (Table 2). The conservative substitution of glutamate at D669 had a modest effect on desensitization in the absence or presence of Neto2 (GluK2(D669E)/GluK2, 2.8 ± 0.3 ms, n = 4; GluK2(D669E)/GluK5/Neto2, 6.5 ± 0.9 ms, n = 7; fold change 2.3 ± 0.3), but did not reach statistical significance. However, a ΔΔG value of −0.53 kcal mol⁻¹ was obtained, indicative of coupling between this mutation and Neto2 modulation (Table 2). The heteromeric GluK2(E665V,D669E)/GluK5 receptor exhibited wild‐type desensitization kinetics in the absence (1.8 ± 0.3 ms, n = 8) and presence of Neto2 (9.3 ± 2.1 ms, n = 11, fold change 5.2 ± 1.2; Fig. 6 B and Table 2) and the negative energetic coupling observed for single mutants was absent (ΔΔG = −0.05 kcal mol⁻¹). Thus, we have identified a mutation in the GluK2 M3–S2 linker that abolishes Neto2 modulation of heteromeric GluK2/GluK5 receptors, but compensation for this loss of modulatory activity can occur through additional structural modification in this critical linker domain.

Neto2 modulates the KAR cation sensitivity

Desensitization of AMPARs and KARs arises in part from the dissolution of the interface between D1 domains of the LBD pairs (Meyerson et al. 2014). KARs are unusual within the iGluR family in that normal gating function is profoundly influenced by cations and anions acting at discrete sites formed by the D1 domains (Bowie, 2002; Paternain et al. 2003; Wong et al. 2006; Chaudhry et al. 2009). In homomeric GluK1 and GluK2 receptors, each LBD pair contains one anion and two cation binding sites (Plested & Mayer, 2007; Plested et al. 2008); occupancy of these sites stabilizes the dimer interface, is permissive for activation, and slows entry into the desensitized state (reviewed in Bowie, 2010). Conversely, alteration of the LBD dimer stability through other manipulations, such as disulfide bridge crosslinking, eliminates the cation dependence of GluK2 KAR gating (Plested et al. 2008). We therefore used cation sensitivity as an indirect assay for the impact of Neto2 co‐assembly on stability of the LBD dimer interface.

We first compared the effect of equimolar substitution of extracellular Cs⁺ for Na⁺ on the amplitude of currents evoked by glutamate from GluK2 or GluK2/Neto2 receptors. In outside‐out patches from cells expressing homomeric GluK2 receptors, glutamate current amplitudes recorded in extracellular Cs⁺ were only 21.8 ± 1.5% (n = 6) of the initial peak recorded in standard sodium‐based external solution (Fig. 7 A), similar to previous reports (Wong et al. 2007). GluK2/Neto2 receptors were less sensitive to Na⁺ replacement by Cs⁺. Receptor current amplitudes in external Cs⁺ were 37.0 ± 5.8% (n = 7) of the current amplitude in the presence of external sodium, and this was a more modest reduction than observed with GluK2 receptors (P < 0.05, Student's t test). These data indicate that Neto2 impacts cation‐dependent gating of GluK2 receptors; this could conceivably occur through structural rearrangements in the LBD that stabilize the dimer interface, reducing the critical influence of the cation binding on the interface, and thereby shifting the cation concentration dependence.

Heteromeric GluK2/GluK5 receptors are also sensitive to Na⁺ replacement with Cs⁺ (Paternain et al. 2003), albeit to a lesser degree than homomeric GluK2 receptors, and we therefore tested if Neto2 alters heteromeric KAR sodium‐dependent gating. Whole‐cell glutamate‐evoked currents from GluK2/GluK5 receptors decreased to 28.5 ± 3.4% (n = 8) of the peak amplitude in extracellular Cs⁺ as compared to standard sodium external (Fig. 7 B), similar to what was previously reported (Paternain et al. 2003). Co‐assembly with Neto2 was more permissive for activation in the presence of extracellular Cs⁺, because the peak amplitude of glutamate‐evoked currents was only reduced by 54.9 ± 4.9% (n = 8, P < 0.001 compared to GluK2 receptors) upon ionic substitution. We conclude from these ion substitution data that Neto2 likely stabilizes the LBD dimer interface and thereby reduces cation dependence of gating for both homomeric and heteromeric KARs.

Consistent with these effects on receptor activation, we also found that Neto2 modulation of homomeric GluK2 receptor desensitization was altered in extracellular Cs⁺. That is, GluK2/Neto2 desensitization was significantly attenuated in extracellular Cs⁺ (GluK2/Neto2_(Na), 13.0 ± 1.6 ms vs. GluK2/Neto2_(Cs), 2.0 ± 0.3 ms, n = 6, P < 0.0001, one‐way ANOVA with Bonferroni's post hoc comparison; Fig. 7 A) and only 1.8 ± 0.3‐fold slower than GluK2_(Cs) KAR currents (GluK2_(Cs), 1.1 ± 0.1 ms, n = 6). Neto2 slowing of heteromeric GluK2/GluK5 receptor desensitization was also attenuated, albeit to a lesser extent: GluK2/GluK5/Neto2 receptors desensitized at a rate of 4.2 ± 0.8 ms in extracellular Cs⁺ compared to 6.9 ± 1.3 ms in standard external solution (n = 7, Fig. 7 B). These results further indicate that integrity of the LBD D1 dimer interface, which varies dependent upon the nature of the extracellular cation (Chaudhry et al. 2009), greatly impacts the efficacy of Neto2 modulation of GluK2 receptor desensitization.

To more directly test this hypothesis, we measured desensitization rates of GluK2 and GluK2/Neto2 receptors under low sodium conditions (50 mM NaCl; Fig. 8). Both GluK2 and GluK2/Neto2 KARs desensitized at similar rates in 50 mM NaCl (1.7 ± 0.2 ms (n = 5) and 2.8 ± 0.3 ms (n = 5), respectively), consistent with results obtained in 150 mM CsCl. Importantly, GluA4 receptor desensitization was unaffected by low extracellular sodium (150 mM NaCl, 3.7 ± 0.2 ms, n = 4; 50 mM NaCl, 4.4 ± 0.1 ms, n = 4; Fig. 8) consistent with the lack of modulation of these receptors by extracellular ions. Taken together, these results support the hypothesis that the integrity of the D1 dimer interface is required for Neto2 modulation and that, conversely, Neto2 strengthens the dimer interface to reduce cation sensitivity.

Figure 8. Destabilization of the dimer interface with low sodium prevents Neto2 modulation .

A, representative glutamate‐evoked peak amplitude‐scaled current traces from outside‐out patches expressing GluK2 (left), GluK2/Neto2 (middle), and GluA4 (right). Responses in each set of traces are from the same patch. Black traces represent external solution containing 150 mM NaCl and red traces represent external containing 50 mM NaCl. B, left, quantification of glutamate‐evoked current response amplitudes in 50 mM external NaCl normalized to control currents evoked from the same patch in standard 150 mM NaCl external for GluK2 receptors in the presence (blue bars) and absence (black bars) of Neto2 and GluA4 receptors (grey bars). Right, quantification of desensitization rates of receptor currents in 150 mM NaCl and 150 mM CsCl external. Values shown are means ± SEM. Statistical significance is denoted as follows: ***P < 0.001, ****P < 0.0001.

M3–S2 linker mutations that impact Neto2 modulation of desensitization also attenuate modulation of cation sensitivity

To determine if M3–S2 linker mutations that attenuated Neto2 modulation of KAR gating also impact the apparent effect of Neto2 on the LBD dimer interface, we measured cation sensitivity of two M3–S2 linker mutants (Fig. 9). Homomeric GluK2(P667A) and heteromeric GluK2(E665V)/GluK5 receptors desensitize at rates comparable to wild‐type GluK2 and GluK2/GluK5 receptors, respectively (GluK2(P667A), 4.6 ± 0.6 ms, n = 15; GluK2(E665V)/GluK5, 2.3 ± 0.3 ms, n = 10), and therefore are particularly useful for determining how these mutations specifically impact Neto2 function. Glutamate‐evoked currents from GluK2(P667A) and GluK2(E665V)/GluK5 receptors in 150 mM CsCl external solution had peak amplitudes that were 20.1 ± 2.7% (n = 9) and 36.4 ± 4.1% (n = 9), respectively, of their Na⁺ external control currents (Fig. 9). However, the relative peak amplitudes of glutamate‐evoked currents from GluK2(P667A)/Neto2 and GluK2(E665V)/GluK5/Neto2 receptors in 150 mM CsCl (as compared to Na⁺ external) were not different from receptors lacking the auxiliary subunit (GluK2(P667A)/Neto2_(Cs), 23.8 ± 2.7%, n = 6; GluK2(E665V)/GluK5/Neto2_(Cs), 47.11 ± 5.3%, n = 9; Student's t test, P = 0.73 and 0.52, respectively; Fig. 9). Thus, Neto2 modulation of cation sensitivity of GluK2 KARs was absent in these mutants harbouring mutations in the M3–S2 linker. Residues in the GluK2 M3–S2 linkers therefore impact not only Neto2 modulation of desensitization but also the conformational changes that underlie Neto2 alteration of cation sensitivity arising from sites in the D1 domain of the LBD.

Figure 9. M3–S2 linker mutations prevent Neto2 modulation of KAR cation sensitivity .

Quantification of mutant homomeric (A) and mutant heteromeric (B) receptor glutamate‐evoked current response amplitudes in 150 mM external CsCl normalized to control currents evoked from the same patch or cell, respectively, in standard NaCl‐containing external solution in the absence (black bars) and presence (blue bars) of Neto2. Representative glutamate‐evoked current traces from outside‐out patches or whole‐cell expressing GluK2(P667A) (left, top), GluK2(P667A)/Neto2 (right, top), GluK2(E665V)/GluK5 (bottom, left), and GluK2(E665V)/GluK5/Neto2 (bottom, right). Black traces represent standard external solution and red traces represent external solution with equimolar substitution of CsCl. Arrows highlight the decrease in current amplitude in extracellular CsCl. Responses in each set of traces are from the same patch or cell. Values shown are means ± SEM.

Discussion

Here we identify the KAR M3–S2 linkers and the LBD dimer interface as structural determinants central to functional modulation by the auxiliary protein Neto2. Mutations in the GluK2 M3–S2 linker reveal that this region influences the magnitude of Neto2 modulation in an agonist‐dependent manner. Indeed, a single mutation in this domain completely abolished Neto2 modulation of heteromeric GluK2/GluK5 receptors. Destabilization of the LBD dimer interface by Cs⁺ substitution or reduction in external Na⁺ prevents Neto2 modulation of GluK2 desensitization. Finally, M3–S2 linker mutations that impact Neto2 modulation of desensitization also abrogate modulation of KAR sodium‐dependent gating, correlating modulatory effects of Neto2 between these two regions. Our results therefore reveal sites of allosteric modulation in KARs by auxiliary proteins.

Domains critical for auxiliary proteins modulation of iGluR function

Structural determinants within iGluRs important for modulation by auxiliary proteins only recently have been identified. Transmembrane AMPA receptor regulatory protein (TARP) γ‐2 association with AMPARs promotes relatively closed conformations of the LBD, which potentially favours the activated state, slowing entry into desensitization (MacLean et al. 2014). Additionally, TARPs interact with the ATD and the ATD–LBD linker to mediate changes in AMPAR gating (Cais et al. 2014). The relevance of these domains to Neto‐dependent modulation of KARs is unknown, although we found the ATD to be unnecessary for Neto2 slowing of GluK2 KAR desensitization (data not shown). Instead, our data reveal a critical role for the M3–S2 linkers and D1 dimer interface as sites of Neto2 modulation within KARs. This juxtamembrane site of functional modulation is consistent with the critical physical association with KARs mediated by the extracellular CUB domains (Tang et al. 2011 a) and the alteration in modulation observed with mutations to the LDLa module of Neto2 (Zhang et al. 2009; Fisher & Mott, 2012). By comparison, in AMPA receptors the first extracellular loop (Ex1) of stargazin (TARP γ‐2) mediates modulation of GluA1 receptor deactivation, desensitization, decay kinetics of mEPSCs in CA3 hippocampal neurons (Tomita et al. 2005), and underlies auxiliary protein‐specific effects (Milstein et al. 2007); the TARP carboxy‐terminus also influences various aspects of AMPAR function (Milstein & Nicoll, 2009; Sager et al. 2009). The carboxy‐terminal tails of Neto proteins do not have analogous roles in KAR modulation. It remains to be resolved where Neto extracellular CUB domains contact KAR subunits and how that interaction is influenced by the structure of the M3–S2 linker.

A critical role for the M3–S2 linker in Neto2 modulation

The coordinated movements of linkers connecting the LBD to transmembrane segments are critical for transmitting the energy of agonist binding to the pore during receptor activation and accommodating rupture of the LBD dimer interface and pore closure during desensitization (Dong & Zhou, 2011). Alteration of the initial two charged residues closest to the M3 domain impacts desensitization of both AMPARs and KARs (Yelshansky et al. 2004; Vivithanaporn et al. 2007), and we show that other, non‐charged residues are critical to receptor gating. In NMDARs, M3–S2 linkers are critical for mechanical coupling between the LBD and receptor pore (Kazi et al. 2014) and a de novo mutation in the human GluN2A subunit S2–M4 linkers leads to early‐onset epileptic encephalopathy (Yuan et al. 2014).

During desensitization, the M3–S2 linkers adopt a conformation unique to the desensitized state (Meyerson et al. 2014). It is possible that auxiliary proteins modulate KAR desensitization rates by influencing linker movement during the gating cycle, such that altering the composition of residues in these linkers, and consequently properties such as hydrophobicity, polarity and linker structure, will impact the magnitude of Neto2 effects on biophysical parameters such as desensitization. Examination of the specific residues that produced changes in Neto2 modulation of KARs did not reveal a clear correlation with any of the particular properties of the key residues, including hydrophobicity, charge or polarity. In a general sense, structural alteration of GluK2 near the M3 side of the linker enhanced Neto2 slowing of desensitization (e.g. R663A), mutations on the S2 side of the linker reduced the impact of Neto2 association (e.g. P667), and those in the centre of the linker tended not to have much effect. There were clear exceptions to this generalization, however. One residue, I668, appears to associate with the S2 domain in AMPARs and in part mediate receptor activation (Chen et al. 2014). Alanine substitution at this site in the GluK2 receptor produced currents with small amplitudes that desensitized rapidly, similar to the analogous mutation in GluA2 (Chen et al. 2014). Neto2 modulation of desensitization therefore might be compromised when the active state of the receptor is inherently unstable, a hypothesis supported by the nearly complete loss of Neto2 modulation in low extracellular Na⁺. Finally, we think it unlikely that physical interactions at the M3–S2 linker are essential for assembly of Neto2 with KARs, because mutations in this region that attenuated or eliminated Neto modulation did not preclude equilibrium association between auxiliary and receptor subunit proteins.

Alteration of a single residue in the GluK2 M3–S2 linkers occludes Neto2 modulation of heteromeric KARs

KAR and AMPAR M3–S2 linkers differ at two residues. While reciprocal exchange of both sites in GluK2 did not significantly impact homomeric or heteromeric KAR modulation by Neto2, valine substitution at E665 in GluK2 led to a significant attenuation of Neto2 modulation of homomeric receptor desensitization and a complete loss of modulation of heteromeric receptor desensitization, without impacting basal receptor kinetics. In heteromeric receptors, proximal and distal subunits are occupied by GluK2 and GluK5, respectively (Kumar et al. 2011). During KAR desensitization, the distal M3–S2 linkers undergo little movement relative to proximal linkers (Meyerson et al. 2014); thus introduction of this highly hydrophobic residue in the GluK2 subunits of heteromeric receptors could induce a conformational change in the proximal linkers, specifically impacting Neto modulation. Indeed, substitution of the considerably less hydrophobic alanine at E665 had little effect on Neto2 modulation of KAR desensitization.

Neto2 modulation of KAR cation sensitivity

A unique gating requirement of KARs that distinguishes them from AMPARs and NMDARs is binding of two Na⁺ and a single Cl⁻ at discrete sites located at the apex of the LBD at the dimer interface (Wong et al. 2006; Plested & Mayer, 2007). Association of these ions with the receptor increases LBD dimer affinity; conversely, decreased occupancy of these sites destabilizes dimer association (Chaudhry et al. 2009). GluK1 crystal structures suggest that occupancy of the cation binding site is reduced in extracellular Cs⁺, which destabilizes Cl⁻ binding to the anion binding pocket (Plested et al. 2008); consequently, fewer receptors are available for activation. Furthermore, KAR desensitization is thought to be mediated in part by unbinding of sodium prior to that of glutamate, leading to LBD dimer interface rupture (Dawe et al. 2013).

We identified the D1 dimer interface as a target for Neto2 modulation. Neto2 reduces the cation sensitivity of homomeric GluK2 and heteromeric GluK2/GluK5, increasing the amplitude of glutamate‐evoked currents in extracellular Cs⁺ relative to control currents. This could occur through a stabilizing effect of Neto2 on the dimer interface as mutations that prevent interface rearrangements (and therefore receptor desensitization) abrogate KAR cation sensitivity (Plested et al. 2008). Therefore, under physiological conditions Neto2 co‐assembly may slow KAR desensitization by inducing conformational changes at sites in the dimer interface that favour the active state, a mechanism also proposed for TARPs (MacLean et al. 2014). However, under conditions where the active state is inherently unstable and dimer interface integrity is compromised, such as low extracellular Na⁺, Neto2 modulation is reduced. This idea is also consistent with our observation that Neto2 modulation of GluK2(I668A) receptor desensitization is attenuated (Chen et al. 2014). Alternatively, Neto2 may reduce the sensitivity of KARs to cation and anion binding for receptor activation, thereby increasing the number of receptors available for activation.

Correlated actions of Neto2 on the M3–S2 linkers and dimer interfaces

Mutations in the M3–S2 linkers eliminated Neto2 modulation of KAR cation sensitivity. In NMDARs, a mutation in the S2–M4 linker similarly abrogated Zn⁺ inhibition of these receptors despite localization of the Zn²⁺ binding site in the ATD (Yuan et al. 2014) and remote effects were described for TARP alteration of the AMPAR ATD orientation via the ATD–LBD linker (Cais et al. 2014). Our results could be accounted for by an allosteric coupling between M3–S2 linkers and the LBD dimer interface via Neto2 proteins. Alternatively, the modal gating behaviour of individual GluK2/Neto2 receptor‐channels suggests that distinct conformational states of the receptor exist and underlie the macroscopic receptor desensitization kinetics; mutations to the M3–S2 linker could alter the equilibrium between gating modes or produce conformational states in which Neto2 no longer alters D1 interface stability (Zhang et al. 2014). Thus, the nature and causal relationship between Neto modulation of desensitization and cation sensitivity requires further investigation.

In conclusion, we elucidated critical structural determinants within the KAR that contribute to modulation by Neto2. Auxiliary subunits act as allosteric modulators of receptor function, and thus our work contributes to the understanding of auxiliary subunit modulation of KAR function, which could have therapeutic implications for disease states in which these receptors have been implicated, such as epilepsy, pain and stroke.

Additional information

Competing interests

The authors declare that they have no conflicts of interest in any aspects of this study.

Author contributions

Both authors contributed to the conception and design of the experiments. T.N.G collected and analysed data at Northwestern University Feinberg School of Medicine. T.N.G drafted the article and both authors revised it critically for content and read and approved the final version of the manuscript.

Funding

This study was supported by grants R01NS071952 from the National Institute of Neurological Diseases and Stroke to G.T.S., 5T32MH06754‐09 from National Institute of Mental Health to T.N.G. and PRE16820014 from the American Heart Association to T.N.G.