Mechanisms of photoswitch conjugation and light activation of an ionotropic glutamate receptor

Abstract

The analysis of cell signaling requires the rapid and selective manipulation of protein function. We have synthesized photoswitches that covalently modify target proteins and reversibly present and withdraw a ligand from its binding site due to photoisomerization of an azobenzene linker. We describe here the properties of a glutamate photoswitch that controls an ion channel in cells. Affinity labeling and geometric constraints ensure that the photoswitch controls only the targeted channel, and enables spatial patterns of light to favor labeling in one location over another. Photoswitching to the activating state places a tethered glutamate at a high (millimolar) effective local concentration near the binding site. The fraction of active channels can be set in an analog manner by altering the photostationary state with different wavelengths. The bistable photoswitch can be turned on with millisecond-long pulses at one wavelength, remain on in the dark for minutes, and turned off with millisecond long pulses at the other wavelength, yielding sustained activation with minimal irradiation. The system provides rapid, reversible remote control of protein function that is selective without orthogonal chemistry.

Much progress has been made recently in the real-time noninvasive detection of protein function (1), but the development of approaches for remote protein manipulation within the complex environment of the cell has lagged behind. A major advance has been the development of photolysable cages for soluble ligands (2). The caged ligand is allowed to slowly diffuse into tissue in its inert form, and a powerful light flash cleaves a photolabile protecting group, releasing the active ligand rapidly (within microseconds) (3, 4). This provides fast on-rates that are complemented with reasonably fast off-rates, which depend on native binding affinity, diffusion, sequestration, and breakdown. However, because native ligands often act on multiple proteins, this approach has limited selectivity. Introducing foreign receptors that bind nonnative ligands, on the other hand, enables cellular stimulation without the activation of endogenous proteins (5, 6) and can even be used in a photolysable form for rapid release (7).

Photoisomerizable moieties have also found use in the remote and selective control of native protein function. In this case, reversible photochemically induced changes in the shape or electronic character of functionally important amino acids have been used to control the function of proteins in response to light (8–10) or to alter the backbone structure of peptides (11), thereby controlling their interaction with other biological macromolecules (12). In an alternative strategy, photoisomerization of a tethered ligand can be used to reversibly present, and withdraw, a ligand from a binding site. To date, several ion channel photoswitches have been reported wherein a ligand is tethered to the channel via a linker containing a photoisomerizable azobenzene moiety. The ligand in these photoswitches can operate either as an active site (i.e., pore) blocker (10, 13), or as an allosteric ligand for an ionotropic receptor (14, 15).

To tune the properties of tethered ligand photoswitches, optimize their efficacy, and generalize their application to other proteins, it is important to understand their basic physical and chemical properties. We set out here to characterize the optical gating of the ionotropic glutamate receptor subtype 6 (iGluR6) with a photoisomerizable tethered agonist termed MAG (Maleimide–Azobenzene–Glutamate) (MAG-1 and MAG-2, Fig. 1b). We address several key issues that aim to characterize both the photochemical and functional properties of the modified channel protein. Investigations into the photochemistry of the azobenzene photoswitch, including the half-life of thermal relaxation and the wavelength dependence of activation and deactivation, are discussed. We investigate the estimated local concentration and efficacy of different length photoswitches under a variety of labeling conditions in an attempt to optimize the degree of activation of the light-activated receptor. Lastly, the elucidation of other key properties, including affinity labeling at concentrations that are approximately three orders of magnitude below the K_d, and thus will not activate receptors appreciably, and selective targeting of MAG to individual cells within a culture using patterned illumination, provide a basis for the development of light-activated ion channels as a powerful tool in neurobiology.

Fig. 1.

Modular photoswitchable tethered ligands. (a) The light-gated glutamate receptor operates by reversibly binding of the photoswitchable agonist MAG (14), which is attached covalently to a cysteine introduced in the ligand binding domain of the receptor. The ribbon structure of apo-iGluR2 (Protein Data Bank ID code 1FTO) (16) is shown on the left, together with the ball-and-stick structure of MAG in the extended (trans) and unbound conformation. Under 380-nm illumination, MAG can activate the receptor as is shown on the right with cis-MAG docked on the structure of iGluR6 in complex with methylglutamate (Protein Data Bank ID code 1SD3) (17). Photoswitching is reversible with 500-nm illumination. (b) MAG-1 (14) can be elongated by introducing an additional glycine unit (MAG-2). Compound 3 is a nonphotoswitchable MAG-1 analog and an iGluR6 agonist termed the “tether model.”

Results

Further Details.

For further details, see supporting information (SI) Text, SI Appendix, and SI Figs. 8–11.

Modular Photoswitchable Tethered Ligands.

The photoswitchable tethered ligand was designed to possess a maleimide for conjugation to a cysteine residue on the exterior of the ligand-binding domain (LBD), a glutamate analog, and an azobenzene linker enabling reversible state-dependent control over the reach of the glutamate analog (Fig. 1a) (15). The glutamate analog was chosen based on previously established structure–activity relationships of the selective iGluR agonists (2S,4R)-4-allyl-glutamate (LY310214) and (2S,4R)-4-methyl-glutamate (SYM 2081) (18–20), and on our iGluR6 agonist, termed the “tether model” (3; Fig. 1b) (15). The modularity of the design allows for the introduction of additional glycine units in the tether with minimal synthetic investment. Initial studies were based upon models of docking MAG-1 in the iGluR6-SYM 2081 crystal structure (19), though the exact tether length required for optimal activation remained unknown. After the synthesis of MAG-1, the elongated MAG-2 was synthesized by using chemistry analogous to that described in ref. 15. MAGs of different lengths allow for the study of tether length dependence on channel activation and agonist binding using readily accessible and minimally disruptive amino acid building blocks. The success of the MAG design, and the ease with which it can be modified, opens the possibility of replacement of the glutamate moiety for other iGluR agonists or antagonists, or application to other similarly functioning allosteric proteins with well defined ligand binding modes.

Photostationary State Determination by NMR.

In the thermally relaxed state, azobenzene exists entirely in the trans configuration. Upon illumination, a steady state mixture is generated, with a fraction of the azobenzene in the cis configuration and the rest in trans. The balance between cis and trans (the photostationary state) depends on the wavelength of irradiation. The cis population is maximally populated in the near UV, and trans population is maximally populated in the visible range of the light spectrum (21). Usually UV/visible spectroscopy is used to determine the fraction of azobenzene in the two states (22). Here we used a novel approach of NMR spectroscopy to distinguish between the two isomers (23). NMR was used to determine the ratio of cis- to trans-MAG-1 conjugated to 2-mercaptoethanol between 340 and 500 nm, at 20 nm increments (Fig. 2a). Optimal wavelengths for cis and trans populations were found to be 380 and 500 nm, respectively. At 380 nm, 92.7 ± 0.3% of MAG-1 is in the cis state (n = 3, note that values throughout are mean ± SEM) and at 500 nm 83.3 ± 0.3% of MAG-1 is in the trans state (n = 3).

Fig. 2.

Photostationary state determination by NMR and spectral sensitivity of photoresponses. (a) Fraction of MAG-1 in the cis form determined from NMR spectroscopy. Maximal wavelengths for cis and trans populations are 380 and 500 nm, respectively. Values are given as mean ± SEM (n = 3). (b) Wavelength dependence of photoresponses of iGluR6-L439C channels expressed in HEK293 cells and conjugated to MAG-1. Currents were measured by whole-cell patch clamping in the voltage clamp mode at V_h = −60 mV, with desensitization blocked by Con A. The current vs. time traces and corresponding wavelength step protocol used to record action spectra are indicated. The first set of steps (activation spectrum) starts at the wavelength of maximal deactivation (500 nm) and spans UV illuminations of increasing wavelength. The second set of steps (deactivation spectrum) starts at the wavelength of maximal activation (380 nm) and spans visible illuminations of increasing wavelength. (c) Each temporal trace can be fitted with a single exponential function whose amplitude and time constant is used to build the action spectra. (d) The activation spectrum (“ON”) is centered on 380 nm and falls off rather steeply at higher and lower wavelengths. The deactivation spectrum (“OFF”) is wider, with maximal amplitude between 460 and 560 nm. Values are mean ± SEM (n = 5). (e) Wavelength dependence of photoswitch rate. Time constants τ_ON and τ_OFF from fits of traces in b are represented as switch rates (1/τ). Values are mean ± SEM (n = 5).

Spectral Sensitivity of Photoresponse Produces an Analog Output.

To quantify the relationship between the photostationary state of MAG-1 in solution and after conjugation to iGluR6-L439C, the current amplitude and switching kinetics were measured as a function of wavelength. Experiments were carried out in HEK293 cells expressing CMV-iGluR6-L439C and recorded 1–2 days after transfection by whole cell patch clamping. Recordings were in voltage clamp mode with a holding potential of −60 mV, and desensitization was blocked with Con A (see SI Text). We examined activation by stepping wavelengths from maximal steady state deactivation (500 nm) to a series of shorter wavelengths. The step duration was selected to be 10 s, long enough for currents to reach steady state. We then examined deactivation by starting at the wavelength of maximal steady state activation (380 nm) and stepping to longer wavelengths (Fig. 2b). The activation and deactivation components were each well fit with a single exponential (Fig. 2c). The activation spectrum is centered at 380 nm (Fig. 2d “On”), falling off steeply at higher and lower wavelengths, whereas the deactivation spectrum is broadly centered at ∼500 nm (Fig. 2d “Off”). The NMR-based determination of the photostationary states of MAG-1 in solution, over the wavelength range of 320–500 nm, closely match the action spectrum of channel activation when MAG-1 is conjugated to the channel protein (Fig. 2a). Furthermore, the on and off rates were fastest at wavelengths between 380 nm and 500 nm [k_{on-380 nm} = 2.8 ± 0.1 s⁻¹ (n = 5) and k_{off-500 nm} = 2.6 ± 0.2 s⁻¹ (n = 5), see Fig. 2d].

Thermal Relaxation of MAG.

Experimentally, it may be advantageous to control channel opening without continuous irradiation. In such situations, a single activating pulse of UV light would be used to initiate activation for extended periods of time (minutes). As such, the MAGs were designed with a 4,4′-azodianiline scaffold modified with amide linkages to the glutamate and maleimide moeities of the molecule. These amide-based azobenzene cores are known to possess half-lives of minutes for the rate of thermal relaxation from the cis state to the lower energy trans state in the dark (24).

We measured the rate of thermal relaxation from cis to trans in the dark for free MAG-1 in solution. This was done by following the accumulation of the trans form by measuring absorbance at 360 nm, near the maximum absorbance of the trans form at 380 nm. We obtained a half-life for the spontaneous cis to trans isomerization of free MAG-1 of 17.65 ± 0.03 min (n = 3) (Fig. 3a). This behavior of free MAG was compared with the spontaneous deactivation of iGluR6-L439C channels in the dark using whole cell patch clamp in HEK293 cells. Because it was not possible to maintain seals for long enough to observe full deactivation, we followed activation induced by a 5-s pulse of illumination at 380 nm with an observation period of 2 min in the dark. We observed little decay in the current during these times, with the amplitude being 0.98 ± 0.1 of its initial value after 2 min in the dark (n = 7) (Fig. 3b). These findings show that the slow thermal cis to trans isomerization of azobenzene is preserved in MAG attached to the functioning receptor. Indeed, it appears that binding of the glutamate end of MAG in the binding pocket even stabilizes the cis state of the azobenzene. The significance of persistent channel activity in the dark after a brief pulse of illumination is that long-lasting currents can be maintained in the absence of irradiation, thus reducing photo-bleaching of the azobenzene, photo-damage to the protein, and photo-toxicity to cells. Even if some stabilization of the cis state occurs, 500-nm light is still able to rapidly turn the current off.

Fig. 3.

Slow thermal relaxation of MAG allows for persistent activation in dark. (a) Rate of thermal relaxation in the dark of free MAG-1 from cis to trans, measured by absorbance at 360 nm. Traces are exponential and display a half-life of 17.65 ± 0.03 min (n = 3). (b) Minimal spontaneous deactivation of iGluR6-L439C channels conjugated to MAG-1 over a 2-min period after activation with a 5-s pulse at 380 nm. Photocurrent at end of 2 min in dark is as large as photocurrent evoked by new pulse at 380 nm after light-induced deactivation at 500 nm.

MAG Conjugation to iGluR6-L439C Occurs by Affinity Labeling.

In our first study (15), a model of MAG-1 in the cis state was docked onto the crystal structure of iGluR6 in complex with SYM 2081. When the glutamate moiety was fit in the agonist binding site, the maleimide end of MAG-1 was able to reach amino acid 439, where an introduced cysteine permits conjugation and yields a light-gated channel (Fig. 1a). This provided a vivid picture of the photoactivated state, and raised the question of whether occupancy of the binding site by MAG-1 would enhance the conjugation efficiency of the maleimide to the cysteine at position 439 by affinity labeling. Affinity labeling has been observed in a variety of systems (25), including in the conjugation of tethered blockers to the Shaker K⁺ channel (26), which served as a basis for the development of the photoswitchable SPARK channel (13).

To investigate the nature of MAG conjugation, we designed two experiments that would test the effect of interfering with affinity labeling. In a first experiment, we asked whether we could hinder labeling by using visible light to favor the trans state of MAG-1 conformation, which is expected to extend the maleimide away from cysteine 439 when the glutamate end of the molecule is docked in the binding pocket (Fig. 4a). We evaluated the efficiency of MAG-1 conjugation from the amplitude of photo-responses using calcium imaging to detect the activation of the calcium permeant iGluR6 channels, as described previously (15). Incubation with 100 nM MAG-1 under 380-nm light (favoring the cis state) produced larger subsequent photo-responses (36.1 ± 3.6%, n = 22, of the 300 μM glutamate response) than did incubation under 500-nm light, favoring the trans state (10.7 ± 0.9%, n = 22, of the 300 μM glutamate response) (Fig. 4c). This ∼3-fold difference is consistent with state-dependent affinity labeling, which is expected to better position the maleimide near the engineered cysteine when cis-MAG-1 is bound.

Fig. 4.

MAG-1 conjugation to iGluR6-L439C occurs by affinity labeling. MAG-1 conjugation at 100 nM can be interfered by two means: (a) favoring the trans conformation with 500-nm illumination, which puts the maleimide group away from cysteine 439 when the glutamate is bound to the LBD, and (b) occupying the binding site with a competing concentration of glutamate, thus preventing docking of MAG-1 in a conformation that favors conjugation of the maleimide to the introduced cysteine. (c) Photoresponses obtained by calcium imaging (Fura2 fluorescence ratio at 350 vs. 380 nm excitation) in HEK293 cells transfected with CMV-iGluR6(L439C) and conjugated to MAG-1 under the conditions shown in a. Weak responses are obtained after MAG-1 conjugation at 100 nM under visible illumination (trans-MAG, maleimide group facing away from cysteine 439), but a substantial increase in photoresponses (by a factor 3.5 ± 0.3, n = 22) is observed after conjugation under UV (cis-MAG; maleimide group facing cysteine 439). (d) Weak responses are obtained after MAG-1 conjugation at 100 nM in the presence of 300 μM glutamate (ligand-binding pocket occupied), but they are increased a factor 3.2 ± 0.2 (n = 10) after MAG-1 conjugation at 100 nM in absence of glutamate (binding pocket free to dock glutamate end of MAG-1 and place maleimide near cysteine 439).

In a second experiment, we attempted to interfere with affinity labeling of MAG-1 with a competing concentration of free glutamate (300 μM) during the incubation period (Fig. 4b). Incubation was carried under 380-nm light as shown above. Incubation of iGluR6-L439C with 100 nM MAG-1 in the absence of glutamate for 15 min produced significantly larger subsequent photo-responses (53.1 ± 7.7% of the 300 μM glutamate response, n = 10) than did incubation in the presence of 300 μM glutamate (17.9 ± 3.3% of the 300 μM glutamate response, n = 10) (Fig. 4d). Again, this ∼3-fold difference is consistent with the disruption of affinity labeling by MAG-1 by glutamate competition for the ligand binding site.

Spatially Controlled MAG Conjugation.

The above experiments demonstrate that, at low concentrations, MAG conjugation operates by affinity labeling. Furthermore, the ability to control photoswitch conjugation with light opens the possibility of selective labeling only in regions of a sample illuminated at shorter wavelengths. We tested this idea using a simple light pattern (Fig. 5a). We washed 100 nM MAG-1 into a dish containing HEK293 cells expressing iGluR6-L439C and illuminated a small region of cells through the objective of the inverted microscope with 374-nm light, while illuminating a neighboring region of cells with 500-nm light delivered by a fiber optic cable from above. After a 10-min incubation, MAG was washed from the chamber. We then patched cells in the two regions and compared the amplitude of photocurrent of each cell normalized by the amplitude of the current evoked by 300 μM glutamate. The normalized photocurrents were ∼4.5-fold larger (P < 0.001) in cells located within the 374-nm spot [normalized values = 0.227 ± 0.015 within 374 nm light spot (n = 3), versus 0.050 ± 0.016 within 500 nm light spot (n = 4)] (Fig. 5b). This finding demonstrates that, in addition to selecting cells for optical control by targeting iGluR6-L493C expression with cell-specific promoters, further subselection can be achieved by using light to preferentially label regions illuminated by UV light. One can imagine generating complex patterns of optical responsiveness in a tissue using patterned illumination.

Fig. 5.

Spatial patterning of MAG conjugation with patterned illumination. (a) Illustration of coverslip (12 mm in diameter) on which HEK293 cells are adhered showing illumination pattern used during a 10-min MAG exposure. Simultaneous illumination at two wavelengths was with 374-nm laser light through the ×40 objective, yielding a small (530 μm in diameter) spot (violet area) and with 500-nm light through a fiber light guide on the remaining area of the coverslip (cyan area). After a 10-min exposure to MAG and the designated light pattern, MAG was washed away before subsequent patch clamp recording. (b) Bar graph showing amplitude (mean ± SEM) of inward currents evoked by illumination at 380 nm normalized to amplitude of current evoked by 300 μM glutamate. Photocurrents are ≈4.5-fold larger in cells illuminated at the shorter wavelength during MAG exposure, indicating that affinity labeling could be biased spatially with patterns of light.

Concentration Dependence of MAG Conjugation.

Previous work had demonstrated that the photo-currents of iGluR6-L439C conjugated to MAG-1 were smaller than the saturating glutamate response. We asked whether the partial activation by iGluR6-L439C-MAG-1 is due to incomplete MAG conjugation. Channels were labeled with MAG-1 for 1 h under 380-nm illumination, using concentrations of 0.1, 10, and 200 μM, with the final concentration being the solubility limit of MAG-1. Using whole-cell patch clamp in HEK293 cells, the average photo-current was found to increase with increasing MAG-1 concentration. Relative to currents evoked by 300 μM glutamate, the optical activation of iGluR6-L439C-MAG-1 at 380 nm was 21 ± 3% (n = 7), 54 ± 9% (n = 5), and 71 ± 6% (n = 9) at 0.1, 10, and 200 μM, respectively.

Under optimal excitation at 380 nm to maximize the activating state, MAG-1 will have a photostationary state with 93% of the molecules in the cis state (Fig. 2a). Assuming complete conjugation, a tetrameric channel possessing four LBDs will then be fully activated 75% (0.93⁴) of the time by MAG-1. Thus, our results suggested that, at higher concentrations, we are close to complete labeling under the assumption that MAG-1 functions with similar efficacy to glutamate.

MAG Functions as a Full Agonist.

We asked whether MAG-1 operates as a full agonist, or if it is a strong, but partial agonist. As shown earlier, when partial agonists bind they allow partial closure of the LBD and thus only partial channel activation (27). Thus, If MAG-1 is a partial agonist then in the presence of 300 μM glutamate, photoswitching to cis-MAG-1 should compete with glutamate and reduce the observed current (i.e., act as an antagonist).

We found that iGluR6-L439C labeled with 100 μM MAG-1 for 15 min (i.e., expected to yield substantial, but likely incomplete conjugation, see above) did not show a sign of partial agonism. Rather than decrease currents, photo-activation (isomerization to the cis state at 380 nm) of iGluR6-L439C-MAG-1 in the presence of glutamate slightly increased the current (SI Fig. 11). This observation argues that MAG-1 functions with similar efficacy to glutamate, i.e., is a full agonist.

High Effective Local Concentration of MAGs.

We have shown that agonist 3, a MAG analogue lacking a maleimide and full-length azobenzene, has an EC₅₀ of 180 μM (15) at iGluR6. Although compound 3 possesses relatively weak affinity, the local concentration of the glutamate end of cis-MAG-1 when conjugated to iGluR6-L439C is expected to be very high based on its short tether.

To test this idea, we estimated the effective concentration of the glutamate end of MAG-1 using the competitive antagonist DNQX (26). DNQX inhibits iGluR activation by occupying the glutamate binding site and stabilizing an open conformation of the LBD (Fig. 6a) (16). We examined the ability of DNQX to competitively inhibit the responses of iGluR6-L439C to light-activation with MAG-1, and to perfusion with compound 3 or glutamate. DNQX inhibited the response to MAG-1 at 380 nm illumination in a concentration dependent manner and was completely reversible upon washout (Fig. 6b). The inhibition curve had a 50% inhibition (IC₅₀) of the cis state light response at 220 ± 65 μM DNQX (n = 8) (Fig. 6c). However, even at the DNQX solubility limit of 4 mM, the block of the photo-current was incomplete.

Fig. 6.

Effective local concentration of MAG-1 is in the millimolar range. (a) The competitive antagonist DNQX inhibits iGluR activation by occupying the glutamate binding site without allowing LBD closure. (b) Patch clamp current traces of iGluR6–L439C conjugated to MAG-1 show responses to perfusion of 300 μM glutamate and to illumination. The corresponding wavelength–time traces are shown below. Perfusion of DNQX partially inhibits photoresponses to 380 nm illumination and reveals a basal activation under 500-nm illumination. Inhibition by DNQX is reversible on washout after each DNQX perfusion. (c) Quantification of DNQX inhibition of photoresponses and comparison to its effect on free tether model compound 3. Current under 380-nm light (filled circles) is inhibited by DNQX to 36% of total photoresponse (IC₅₀ = 220 μM ± 65 μM DNQX, n = 8) and the current under 500-nm (open circles) is completely blocked, which reveals a basal activation ≈20% of total photoresponse, IC₅₀ = 7 ± 2 μM DNQX (n = 8). For comparison, DNQX blocks responses to 10 mM tether model 3 (filled squares, IC₅₀ = 202 ± 26 μM DNQX, n = 5) and 3 mM tether model 3 (open squares, IC₅₀ = 39 ± 15 μM DNQX (n = 5). (d) Determination of the effective concentration of MAG-1 as a function of the DNQX IC₅₀ values. The IC₅₀ values for DNQX/tether model 3 are used to calibrate the local concentration axis assuming a linear relationship (straight line), and yield 12.5 mM and 0.5 mM for MAG-1 under UV and visible, respectively. Values are mean ± SEM.

To calculate the local concentration of the glutamate end of MAG-1 we examined DNQX competition versus compound 3, the closest soluble MAG analogue. We measured the concentration dependence of DNQX inhibition using two known concentrations of the tether model (3 mM and 10 mM, the latter being the solubility limit) to extrapolate effective MAG concentrations from their DNQX IC₅₀. At 3 mM and 10 mM concentrations of compound 3 we obtained DNQX IC₅₀ values of 39 ± 15 μM (n = 5) and 202 ± 26 μM (n = 5), respectively (Fig. 6c). Thus, inhibition by DNQX reveals that in the cis state the glutamate moiety of MAG-1 has an effective concentration of 12.5 mM (Fig. 6d). Such a high effective concentration (50-fold greater than the EC₅₀ of compound 3) suggests that the photo-switched tethered ligand functions as designed on the channel, generating a very high effective local concentration in the cis state.

The antagonist competition experiment revealed the existence of a basal current of ∼20% at 500 nm that was blocked by DNQX. We quantified the IC₅₀ value of block of this basal current by DNQX and found it to be 7 ± 2 μM (n = 8) (Fig. 6c). This value indicates an effective glutamate concentration of 0.5 mM, which is ∼30-fold lower than the value measured at 380 nm (Fig. 3d). This finding supports the model that light activates the channel by changing the local concentration of the ligand upon cis–trans photoisomerization.

Tether Length Dependence on Channel Activation.

We next investigated the dependence of light-gating on tether length using an elongated tethered ligand, MAG-2 (Fig. 7a). We found that for iGluR6-L439C conjugated to MAG-2 at 10 μM for 1h the amplitude of the photoresponse at 380 nm was 24 ± 2% (n = 12) that of the current evoked by 300 μM glutamate, about half that measured for iGluR6-L439C-MAG-1 (Fig. 7b). Competition studies on iGluR6-L439C-MAG-2 using DNQX yielded an IC₅₀ under 380 illumination of 80 ± 20 μM (n = 7), indicating an effective concentration that was three-fold lower than that observed for MAG-1. Consistent with the lower effective concentration of MAG-2 in the cis state, the basal activation was lower than for MAG-1 (Fig. 7c) and high concentrations of DNQX were able to completely block the photo-current at 380 nm for iGluR6-L439C-MAG-2 (Fig. 7d).

Fig. 7.

Tether length dependence on channel activation. (a) DNQX titrations of iGluR6-L439C conjugated to MAG-1 (UV, filled circles; visible, open circles) and MAG-2 (UV, filled triangles; visible, open triangles). (b) Amplitude of MAG-1 and MAG-2 photoresponses after conjugation at 10 μM for 1 h, compared with 300 μM glutamate responses. (c) Basal activation of iGluR6 L439C conjugated to MAG-1 and MAG-2, as obtained from low-DNQX plateaux under visible illumination in a. Residual photoresponse for MAG-1 and MAG-2 obtained from high-DNQX plateau under UV illumination in a. Values are mean ± SEM.

Discussion

We have characterized the optical control of a glutamate receptor by a photoswitchable tethered glutamate that is covalently attached to the exterior of the receptor's LBD. Properties of MAG-1 when free in solution, such as the wavelength dependence of isomerization, are found to match those of the fully conjugated, light-activated receptor. NMR experiments elucidating the photostationary state ratios of cis- to trans-MAG-1 provide a greater understanding of the activation of this light-actuated system. For example, at 380 nm, despite the fact that MAG-1 is a full agonist, the photocurrent is submaximal. This is likely due to the fact that even at the optimal wavelength of 380 nm ∼7% of MAG-1 remains in the trans state. By the same token, the basal current at 500 nm can be attributed in part to the ∼17% of azobenzene that is in the cis state at that wavelength. Further optimization could be achieved by building photoswitches with different properties. For instance, fast-relaxing urea-based azobenzene cores (24) could reduce basal activation by quickly isomerizing from the higher energy cis state to the more stable trans conformation in the dark. Furthermore, azobenzene photoswitches with more complete conversion to the cis state should improve the photo-efficacy of the tethered ligand.

Using a soluble mimic of the tethered ligand, the effective local concentration was estimated to be very high (>10 mM), consistent with the short tethers to which the ligand is attached. The increased concentration of the ligand in the cis state is a function of the increased proximity of the glutamate moiety to its binding site, illustrating the geometric constraints imposed by the azobenzene. Furthermore, MAG functions as a full agonist, indicating that the tether does not prevent full closure of the ligand binding domain on the glutamate end of MAG. This is something that cannot be taken for granted because the pathway to the binding pocket is narrow, presenting potential steric clashes. For applications in neuroscience, the high effective concentration allows for selective block of activation by synaptically released glutamate with 25 μM DNQX (29) while maintaining the near maximal photoresponses at 380 nm (Fig. 7c). The length of the tether, here studied using an additional glycine building block, is essential to the function of the photoswitch. Although the increased length of the tether decreased the basal current under 500 nm illumination, the ease with which it is displaced by competing DNQX and the low efficacy of the photoresponse reveals the importance of optimizing tether length to best suit an individual protein under photocontrol.

The high local concentration of the glutamate moiety when the maleimide end is attached makes the reverse geometry possible too, so that binding of the glutamate end places the maleimide in a high effective concentration near the introduced cysteine, resulting in affinity labeling. This enables conjugation to the target receptor selectively, and to do so at concentrations that are far below the EC₅₀ for activation, thus avoiding significant activation and possible cytotoxicity. Moreover, we can use the state-dependence of affinity labeling to pattern conjugation with patterned illumination. Regions illuminated at ∼380 nm are preferentially labeled, a valuable way of constraining where in a tissue light gating will take place.

Our findings represent advances in the remote control of protein function and new insights into the mechanism of light-gating glutamate receptors. We find that, although MAG-1 maintains similar action spectra when it is conjugated to the protein compared with free in solution, binding of the glutamate moiety within the binding pocket may stabilize the cis state of the azobenzene and slow thermal relaxation. As a result of the slow relaxation of cis-MAG-1, the channels display a memory for a short light pulse at 380 nm, which activates the receptors and keeps them open in the dark for an extended period. This can be useful in experimental setups where light cannot be delivered continuously, and it makes it possible to reduce illumination time and thus to minimize photo-toxicity to cells and photo-damage to other elements in a system, including to the azobenzene itself.

Light switching provides the ability to temporally and spatially control protein function in cells, and makes it possible to set switching kinetics and the degree of activation. Future work will exploit these properties in complex cellular environments and could be a powerful tool for the production of artificial biochemical circuits and artificial cells or reengineered cells for synthetic biology. The properties of the MAG photoswitches suggest that structure-based design could, in principle, provide rapid, reversible, and remote control of any protein for which tethered ligands can operate as active site blockers or as allosteric agonists or antagonists.

Methods

Synthesis of iGluR6 Tethered Agonist MAG-2.

MAG-2 was synthesized by using chemistry similar to that described for MAG-1 (15). See SI Text.

Introduction of Cysteine into Glutamate-Binding Domain.

Cysteine point mutations were introduced to the iGluR6 DNA, containing Q at the position 621 RNA editing site (30) using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The following PCR profile was used: one cycle (95°C for 30 s); 20 cycles (95°C for 30 s, 55°C for 1 min, 68°C for 12 min). The forward and reverse oligonucleotide sequences designed for the L439C mutant were 5′-GATTGTTACCACCATTTGCGAAGAACCGTATGTTCTG-3′ and 5′-CAGAACATACGGTTCTTCGCAAAATGGTGGTAACAATC-3′, respectively.

Abbreviations

MAG

maleimide–azobenzene–glutamate

LBD

ligand-binding domain

iGluR6

ionotropic glutamate receptor subtype 6.