Photochemical control of drug efficacy – a comparison of uncaging and photoswitching ifenprodil on NMDA receptors.

Abstract

Ifenprodil is an important negative allosteric modulator of the N-methyl-D-aspartate (NMDA) receptors. We have synthesized caged and photoswitchable derivatives of this small molecule drug. Caged ifenprodil was biologically inert before photolysis, UV irradiation efficiently released the drug allowing selective inhibition of GluN2B-containing NMDA receptors. Azobenzene-modified ifenprodil, on the other hand, is inert in both its trans and cis configurations, although in silico modeling predicted the trans form to be able to bind to the receptor. The disparity in effectiveness between the two compounds reflects, in part, the inherent ability of each method in manipulating the binding properties of drugs. With appropriate structure-activity relationship uncaging enables binary control of effector binding, whereas photoswitching using feely diffusable chromophores shifts the dose-response curve of drug-receptor interaction. Our data suggest that the efficacy of pharmacophores having a confined binding site such as ifenprodil can be controlled more easily by uncaging in comparison to photoswitching.

Two photochemical methods were introduced in the 1970s for the control of cellular physiology. First, Bartels, et al. developed photoswitchable ion channels blockers in 1971^[1]. With this approach they showed that an azobenzene attached to a drug could potentiate acetylcholine receptors photoreversibly. In 1978 Kaplan et al. used a photochemical protecting group to inactivate ATP^[2]. Covalent attachment of an ortho-nitrobenzyl chromophore^[3] to the gamma phosphate of ATP prevented hydrolysis by cellular ATPases. Irradiation of the optical probe cut the covalent bond, liberating ATP in its bio-active form. This transformation was called “uncaging”^[4] and the approach gained great popularity in many areas of physiology and biochemistry^[5–11]. The reason for this arose from the simplicity of the structure-activity relationship of most important small biomolecules, allowing covalent attachment of the photochemical protecting group at a single site to inactivate completely the biological properties of the signaling molecule^[12]. Further, since irradiation results in recovery of the effector, uncaging produces clean, binary effects. In other words full off to full on is inherent to such an approach, similar to a unidirectional switch^[13]. Thus, most important organic biological molecules have been caged and used to study a myriad of physiological processes^[12].

However, the very strength and simplicity of the uncaging technique can be seen as methodologically limited. Since most biological processes are bidirectional, the intellectual appeal of reversible optical probes is obvious. Thus, starting in 2000, the approach of Erlanger^{[1, 14–16]} was revived for biological studies by Woolley and co-workers^[17]. While the chemistry of both systems is fascinating in itself^[18], it is important to point out that the true goal must be to explore and develop enabling technologies for biological studies^[19]. Here we describe the synthesis, photochemistry and biological properties of optical probes (1 and 2, Fig. 1) designed to modulate the efficacy of a small molecule drug (ifenprodil, Fig. 1) towards its target bioreceptor (the N-methyl-D-aspartate (NMDA) receptor).

Figure 1.

Structures of ifenprodil and the caged (1) and photoswitchable (2) derivatives.

NMDA receptors are ligand-gated ion channels that are central to neuronal plasticity and thus are involved in several neurological pathologies^[20]. Ifenprodil is the lead compound of a large class of compounds acting as negative allosteric modulators selective for NMDA receptors containing the GluN2B subunit^[21]. We rendered this compound photosensitive, in two fundamentally different ways in order to compare the ability of nitrobenzyl uncaging of 1 with azobenzene photoswitching of 2 to control small molecule drug binding to this important ion channel^[20]. We would note there are alternative photoswitches^[22] but chose the most widely used one for this comparison. Ifenprodil is a useful test bed for such studies as it is widely used in physiological studies, and the X-ray structures of drug-receptor complexes are known at atomic resolution^{[23, 24]}. We find that attachment of chromophores to ifenprodil reduces significantly high affinity binding of the drug to the NMDA receptor, however only uncaging can enable significant photoregulation of receptor activity.

Results.

Modeling photoprobe binding to NMDA receptors.

The NMDA receptor plays a central role in neuronal plasticity, as it is the major ligand-gated ion channel through which Ca²⁺ enters neurons^{[20, 25]}. It is a tetramer, each subunit consisting of four domains: a C-terminal domain, a transmembrane domain (TMD); an agonist binding domain (ABD), a N-terminal (NTD, sometimes called the amino-terminal domain) (Fig. 2a). Several high-resolution crystal structures of the NMDA receptor^{[23, 26]}, and the isolated NTDs^{[24, 27]}, have been published with and without ifenprodil (review^[25]), allowing a molecular model to be proposed for the NTD-driven negative allosteric modulation^[27–29]. This drug is known to bind specifically to the interface between the upper lobes of the NTDs of the GluN1 and GluN2B subunits (Fig. 2a,b).

Figure 2. Docking models of ifenprodil derivatives.

(a) Left: global architecture of the NMDA receptor. Right: crystal structure of the NTD of the NMDA receptor subunits GluN1 (dark gray) and GluN2B (light gray) in complex with ifenprodil (green, with red O atoms). PDB code 3QEL^[24]. (b) Close up of the ifenprodil binding pocket showing the accessible volume of the binding site as a green surface. (c) Docking pose of compound 3 showing that the benzyl group lies outside the ifenprodil binding surface. (d) Docking pose of compound 5 showing the 4-carbon sidechain (orange, with blue N atoms) occupying a “tunnel” adjacent to the ifenprodil binding pocket. (e) Docking pose of compound 2 (trans configuration) with the 4-carbon linker shown in orange and the chromophore in yellow. (f) Docking pose of compound 2 (cis configuration) with the 4-carbon linker shown in orange and the chromophore in yellow.

We predicted the binding mode of various compounds to the NTDs by using Glide (Schrödinger) to dock the ligands into the crystal structure of the NTD/ifenprodil complex (PDB code 3QEL^[24]). First, we analyzed the ifenprodil binding pocket area to evaluate the presence of unoccupied cavities adjacent to it that could be used to modify the ligand with an aromatic chromophore. This analysis revealed two cavities flanking the central piperidine moiety of the molecule with little space around the aromatic ends (Fig. 2a,b). This docking suggested that there was probably insufficient space in the ifenprodil binding site to accommodate modifications at the OH of the phenyl group. Induced Fit Docking using Prime/Glide (Schrödinger) with compound 3 confirmed that, even allowing for flexibility in the target structure, it is not possible to accommodate such modifications (Fig. 2c - benzyl-ifenprodil can be considered the simplest non-photolabile model of caged ifenprodil). A result confirmed by in vitro electrophysiology in oocytes^[30] (Fig. S1). These data suggested that caging ifenprodil on the phenol might produce a biologically inert, photolabile derivative of the drug. In contrast, alkyl derivatives^[31] of ifenprodil, such as compound 5^[31] are predicted to be tolerated because they access a narrow “tunnel” that connects the ifenprodil binding pocket to the solvent exposed surface in the closed conformation of the NTD clamshell (Fig. 2d, and Supplemental movie 1). Since it has been shown previously that analogs of ifenprodil having substituents at the isopropyl core of ifenprodil of various linear lengths (4–7 atoms) bind with high affinity in vitro^{[30, 31]}, we expected such drugs would be accommodated when docked into the ifenprodil binding site. Such was the case with 4, 5 and 6-carbon side chains which can all access the “tunnel” when docked onto the ifenprodil binding pocket (Fig. 2d, and Supplemental Movie 1). Consequently, we modeled if a trans and cis azobenzene (AB) derivative 2 might bind to the NTD. In silico docking suggested the trans configuration might fit into the drug binding pocket, with the chromophore filling a vestibule in a relatively relaxed configuration of the drug without clashing with the protein (Fig. 2e). In contrast, the cis configuration is predicted to bind in a strained conformation that tends to clash with the protein, which would result in a much lower binding affinity (Fig. 2f).

Synthesis of photochemical probes.

Encouraged by these in silico data, we synthesized caged and photoswitchable derivatives of ifenprodil as shown in Figure 3. Caged compound 1 was made by base-coupling of our di-PEG-nitrobenzyl^[32] (DPNB) chromophore 4 to ifenprodil to give 1 (DPNB-ifenprodil, or “DPNB-ifen”) in 12% yield. Following previous work^{[33, 34]}, the dimethyl-AB photoswitch 6 was derivatized with PEG-Br using base coupling to give acid 7. This was then activated with N-hydroxysuccinimide to give 8, which was coupled to 5 to yield the target dimethyl-(diMe)-AB-ifenprodil 2.

Figure 3. Synthesis of the photoprobes.

Reagents and conditions: (a) K₂CO₃; 12%; (b) 1-bromo-3,6,9-trioxadecane, K₂CO₃; 77%; (c) N-hydroxysuccinimide, N,N’-dicyclohexylcarbodiimide, 78%; (d) N, N-diisopropylethylamine in N-methyl-2-pyrrolidinone, 44%.

Uncaging photochemistry.

Photolysis of DPNB-ifen (Fig. 4a) with 365-nm light in oocyte extracellular recording solution resulted in its rapid disappearance and the clean release of ifenprodil, as shown by HPLC analysis (Fig. 4b). Further, co-photolysis with DPNB-ifen with our previously quantified caged compound DPNB-ABT594^[32] at 405 nm in HEPES revealed that DPNB-ifen was photolyzed with a quantum yield of 0.1 (Fig. 4c). Note that due to the nature of covalent bond used to attach the DPNB chromophore to ifenprodil, DPNB-ifen is completely stable in physiological buffer.

Figure 4. Uncaging photochemistry and quantum yield.

(a) Photolysis of DPNB-ifen with UV light releases ifenprodil and the spent caging chromophore 9. (b) HPLC chromatograms monitored at 220 nm of pure ifenprodil (black trace, the peaks at 5 min are from DMSO), DPNB-ifen (orange trace), and 365-nm irradiation of DPNB-ifen showing the release of ifenprodil (pink trace, the additional, small peaks are side products). (c) Time-course of comparative photolysis of equal concentration of DPNB-ifen and DPNB-ABT594. Both caged compounds have the same chromophore, and DPNB-ABT594 has a quantum yield of photolysis of 0.2^[32].

Uncaging photopharmacology: selective inhibition of NMDA receptors containing GluN2B subunits.

We tested if DPNB-ifen could be used for photopharmacological modulation of the NMDA receptor containing the GluN2B subunit by two-electrode voltage-clamp on Xenopus oocytes^[30]. Thus, we expressed the wild type (wt) GluN1 and GluN2B subunits in oocytes, and used two-electrode voltage clamp to monitor the modulation of NMDA receptor-mediated currents when saturating concentrations (0.1 mM) of glutamate and glycine were bath-applied, as previously described^[30]. When DPNB-ifen was co-applied at 1 μM, irradiation of the perfusion chamber induced significant inhibition of these currents (56 ± 2 %, n = 12, Fig. 5a). When the concentration of DPNB-ifen was increased 10-fold, irradiation under the same conditions produced a more robust effect (83 ± 2 % inhibition, n = 5, Fig. 5b). In contrast, ifenprodil inhibition was not UV-dependent (Fig. 5c and S1c). By comparing the dose-response curves from simple bath-application of ifenprodil and photo-evoked release, we could estimate that the photochemical yield in our uncaging system was about 22 % (Fig. 5c). Interestingly, when DPNB-ifen (0.1 – 10 μM) was applied there was a small decrease in the neurotransmitter-induced currents, which was only significant at the highest concentration tested (Fig. 5). We found that the dose-response curve of DPNB-ifen prior to photolysis overlapped with ifenprodil dose-response curve on an NMDA receptor construct in which the entire NTD of the GluN2B subunit was deleted^[35] (dashed line, Fig. 5c). This result indicates that the residual inhibition induced by DPNB-ifen at high concentrations is not mediated by the GluN2B specific ifenprodil binding site, but probably by an non-specific receptor pore blockade^[36]. Finally, we confirmed the NMDA receptor subunit-specific effects of photolysis of DPNB-ifen. Bath-application of DPNB-ifen (10 μM) had almost no effects on NMDA receptors consisting of GluN1/GluN2A (3 ± 1 % inhibition, n = 6, Figure S3a,b). And photolysis did not induce further inhibition of GluN2A-containing NMDA receptors (4 ± 1 % inhibition, n = 6, Fig. S3a,b).

Figure 5. Uncaging DPNB-ifen induces inhibition of GluN1/GluN2B NMDA receptors.

a,b) Representative current traces from oocytes expressing wt GluN2B-NMDA receptors, following co-application of glutamate and glycine (100 μM each). DPNB-ifen was bath-applied at 1 μM (a) or 10 μM (b), then photolyzed with 365-nm light (violet bar). (c) Dose-response curves of DPNB-ifen (circles) and ifenprodil (triangles) in the dark (black traces) and under 365 nm light (in UV, violet traces) on wt GluN2B-NMDARs. IC₅₀, Hill coefficient and maximum inhibition are, respectively: 0.12 ± 0.01 μM, 1.6 ± 0.2 and 0.91 ± 0.02, n = 4 for ifenprodil in the dark; 0.13 ± 0.01 μM, 1.7 ± 0.3 and 0.91 ± 0.02, n = 4 for ifenprodil under UV light; and 0.57 ± 0.04 μM, 1.18 ± 0.08 and 0.86 ± 0.02, n = 5 to 12 for DPNB-ifen under UV light. Data points of DPNB-ifen in the dark (n = 5 (0.1 μM), 8 (0.3 μM), 12 (1 μM), 12 (10 μM)) were not fitted due to the very small effects of the compound in this condition. The dashed line represents the ifenprodil dose-response curve of GluN1wt/GluN2B-ΔNTD receptors (from^[37]).

Photostationary states and thermal stability of diMe-AB-ifenprodil.

Irradiation of all trans-2, the material produced from the synthesis, with 365-nm light caused transformation (Fig. 6a) almost completely to the cis configuration of 2 (spectra shown in Fig. 6b). We detected 95% cis-2 by HPLC (Fig. S2a). Irradiation of this mixture with 532 nm generated 52% trans-2. These values are slightly different from the chromophore without ifenprodil (6), with the UV photostationary state (PSS) being 90%, and green PSS being 64% (Fig. S2b). The cis configuration of azobenzenes is normally thermodynamically unstable, so we studied the time course of thermal relaxation of cis-2 by HPLC. We found that at RT this configuration had a half-life of about 1.7 days (Fig. 6c).

Figure 6. Photoswitching photostationary state spectra and thermal half-live of cis configuration.

(a) Irradiation of all trans-2 in HEPES (1 μM, pH 7.2, + 1% DMSO) with 365-nm light drives a change in configuration to a photostationary state (PSS) consisting of 95% cis-2, as determined by HPLC (Fig. S2). Irradiation of this reaction mixture with a 532 nm laser creates a green PSS of 52% trans-2 (R = PEG); (b) Absorption spectra of trans-2, and the UV and green PSS; (c) HPLC analysis of the thermal reversion of the cis PSS revealed cis-2 had a thermal half-life of 1.7 d at RT in HEPES buffer (pH 7.4).

Testing of diMeAB-ifen on NMDA receptors.

We assessed the effectiveness of our photoswitchable ifenprodil in the same manner as the caged ifenprodil. After purification diMe-AB-ifen (2) was obtained in the 100 % trans configuration. This was irradiated for 20 min with 365-nm light to obtain the UV PSS (i.e. 95 % cis configuration), then kept in the dark to avoid cis-to-trans conversion. Given the thermodynamic stability of the cis configuration in the dark, this allowed us to bath apply almost pure cis-diMeAB-ifen (called “cis-2” in Fig. 7). At the concentration of 1 μM, neither trans-2, nor cis-2 induced inhibition of GluN2B-containing NMDA receptors (2.4 ± 5 % inhibition for both trans-2 and cis-2, n = 5; Fig. 7). On the contrary, at this concentration, ifenprodil induced near-complete inhibition of GluN2B-NMDARs (84 ± 2 % inhibition, Fig. 7). Introduction of the azobenzene moiety therefore strongly decreased ifenprodil antagonist activity. It is quite common that attachment of an AB photochrome to an effector significantly decreases the efficacy of the effector, nevertheless conditions can often be found for one the configurations to be moderately active^[38]. However, attempts to do this with our photoswitch were thwarted by the fact that 2 was not soluble above ca. 1 μM in 1% DMSO-HEPES.

Figure 7. Addition of a photochrome to ifenprodil strongly decreases its bio-activity.

(a) Representative current traces from oocytes expressing wt GluN2B-NMDA receptors, following co-application of glutamate and glycine (100 μM each), and 1 μM of either trans-2, cis-2 (95% cis in the UV PSS), or ifenprodil. (b) Summary of the amount of inhibition induced by trans-2, cis-2 and ifenprodil (n = 5 for trans-2 and cis-2, and n = 20 for ifenprodil). n.s., p = 1; ***, p < 0.001; one-way Anova followed by Tuckey test.

Discussion.

In this study we compared the ability of two highly versatile photochemical methods for control of effector-receptor binding using the widely used and well-characterized drug ifenprodil as a test bed. Guided by in silico modeling of possible photoprobes interactions with the NTDs of the GluN1-GluN2B subunits of the NMDA receptor we synthesized 1 and 2 and assayed the photopharmacological efficacy of our compounds. Our modeling (Fig. 2) suggested that the caged ifenprodil 1 was unlikely to be accommodated by the binding pocket (Fig. 2c). Similarly modeling predicted it was possible that the extended (i.e. trans) configuration of the photoswitchable ifenprodil might fit into the binding site (Fig. 2e), whereas the more compact (cis) configuration would be less likely to do so (Fig. 2f). In reality only the caged compound conformed to these predictions (Figs. 5 & 7). The reason for this may lie in the methodological differences of each photochemical technique (Fig. 8), and in the particularities related to the ifenprodil molecule and its binding site. Uncaging relies on unidirectional photorelease of native effectors from bio-inert precursors no matter what concentration is used. Therefore, it is binary in terms of biological activity: it induces release of a fully active molecule from a fully inert compound. On the contrary, freely diffusible photoswitches cycle between two different light absorbing molecular configurations having different affinities for target receptors^{[19, 39]}. Thus, photoswitching of soluble ligands is more challenging from a design point of view. Indeed, the AB-photoswitchable chromophore is often attached to the ligand of interest, consequently the drug binding site must be large enough to accommodate the added photoswitchable moiety, but not too large to accommodate it only in one configuration (cis or trans). As a result, photoswitchable compounds are often less active than their native effector^{[22, 40–43]}. Further, this approach typically only shifts the dose-response curve of the photodrugs (Fig. 8b). This is potentially a strong limitation, as the concentration window within which the photoswitch is active in one configuration and inactive in the other one can be very narrow. Accordingly, efficient photoswitching on most neuronal receptors was so far obtained with tethered photoswitches^[41]. Indeed, once the ligand is directly attached to the target receptor, concentration is not an issue anymore and a binary effect can be obtained. However, the necessity to modify the target receptor to allow such ligand-receptor coupling can be a severe drawback to in vivo and clinical applications.

Figure 8. Idealized comparison of the results from uncaging and photoswitching small molecule drugs.

Uncaging and photoswitching have distinct strengths and weaknesses as photochemical methods for controlling drug action. Uncaging destroys the bond connecting the photochemical protecting group to the drug, and therefore is unidirectional, enabling complete recovery of effector efficacy after photolysis. In contrast, photoswitching does not consume the optical probe, so offers scope for long term control of receptor stimulation without the need to replenish probe supply. (a) Complete photolysis of any caged compound allows full recovery of the bio-active effector. Thus, the relative response of uncaging, when normalized, is one. (b) In contrast to uncaging, photoswitching using freely diffusible chromophores shifts the dose-response curve of the small molecule drug. Typically the cis configuration is less active than the trans, and neither configuration is as active as the drug itself, hence the label “relative response”. The concentration window for efficient photoswitching is thus narrow. Though in a small number of cases there are exceptions to both of these rules^{[44, 45]}.

Both techniques have also been used in many biological studies. The simplicity of uncaging biological signaling molecules has allowed caged compounds to be used to study cells from many different organs such the brain^[46], heart^[47], liver^[48], smooth muscle^[49], erythrocytes^[50], pancreas^[51], skeletal muscle^[52], etc. Uncaging of drugs, or “photopharmacology”, has been much less widely used. The exact reasons for this are unclear because every small organic molecule has one or two functionalities that are key for the binding of the effector to its receptor. Blocking such functionality is normally easy with a photochemical protecting group, resulting in a pre-photolysis inert compound (Fig. 8). In the case of the ortho-nitrobenzyl photochemical protecting groups any type of C-X bond may be cleaved, where X = acids, amines, alcohols, thiols, amides, carbamates, etc^[53]. We took advantage of this versatility with the development of caged ifenprodil by direct attachment of photochemical protecting group to the phenolic functionality of ifenprodil (Fig. 1). Caged ifenprodil 1 worked as expected, from being fully inert in the dark to active after UV illumination (Fig. 5 and 8) and could prove useful to study the physiology of the specific population of GluN2B-NMDA receptors.

Similar to uncaging, photoswitchable biological signaling molecules have been developed intensely^[54]. It is highly attractive, as a method as, unlike uncaging, it is does not consume the photoprobe. Much developmental effort has been focused on drugs as substrates, and typically these are either competitive agonists or antagonists^[40]. Relatively few photoswitchable allosteric modulators of neuronal ion channels have been reported^{[22, 41]}. The first examples of such appeared in 2012 using propofol, a positive allosteric modulator of GABA-A receptors, as a substrate for photoswitching^{[44, 55]}. Pepperberg and colleagues attached an AB chromophore to propofol via a linker of similar lengthy to the one we used for ifenprodil^[44]. This drug could be applied up to 50 μM, much higher concentrations than we could use for 2. But even at 1 μM their drug was effective as a positive allosteric modulator. In an alternative approach, Trauner and colleagues used the aromatic ring of propofol itself to be part of the AB chromophore (so-called “azo-extension”), a strategy which also yielded excellent photopharmacology^[55]. While we considered this approach for ifenprodil in our initial planning, modeling suggested success was unlikely due to the structural constraints of the binding site around the molecule aromatic rings (“azologized ifenprodil” behaved like a caged compound, data not shown). Recently, Trauner and colleagues used the former strategy very effectively for diltiazem to control L-type calcium channels^[56]. Again, this drug could be used at concentrations of at least 25 μM. Since diltiazem is similar in size and character to ifenprodil, the significantly lower solubility of 2 was quite surprising, as this may explain why 2 was not effective as we predicted (Fig. 2).

Finally, even though the photopharmacological properties of 1 are more efficacious than 2, the chemical properties of the diMeAB photochrome^{[33, 34]} are noteworthy in that UV irradiation generates a composition of configurations that is almost pure cis, and the dimethyl substituents confer excellent thermodynamic stability of the cis configuration at RT in physiological buffer (Fig. 5). The diMeAB chromophore should therefore confer photosensitivity to biological molecules with a better photochemistry than “classical”, unsubstituted azobenzenes.

Conclusions.

Uncaging and photoswitching of freely diffusible ligands are complementary ways to control the activity of neuronal receptors with light-driven chemical tools. The choice between the two approaches depends on multiple factors including size and solvent accessibility of the binding site, tolerance of the molecule to chemical modifications, as well as the nature of the biological system, and its amenability for control of the in situ ligand concentration. Our work is the first direct comparison of a directly caged small molecule drug and photoswitchable small molecule drug. We show that to control photochemically the well-known GluN2B-NMDA receptor antagonist ifenprodil, the balance is clearly in favor of the uncaging approach.

Methods.

Synthetic methods

All chemicals were purchased from commercial sources and used as received unless otherwise noted. Reactions were monitored by thin-layer chromatography (TLC) on Merck KGaA glass silica gel plates (60 F254) and were visualized with UV light or Iodine staining. Column chromatography was performed using Agela Technologies industrial grade silica (200–300 mesh, 40–60 μm). NMR spectra were recorded on an Varian 300 MHz NMR spectrometer and the chemical shifts are reported in ppm using the solvent peak as the internal standard (CDCl₃ or CD₃OD). Peaks are reported as: s = singlet, d = doublet, t = triplet, q = quartet, dd = doublet of doublets, m = multiplet. High-resolution mass spectral data were obtained using an Agilent G1969A ToF LC-MS.

DPNB-ifenprodil (1).

To a solution of bromide 4 (60 mg, 0.11 mmol) in dry DMF (2 mL), K₂CO₃ (60 mg, 0.43 mmol) and threo-ifenprodil (42 mg, 0.13 mmol) were added under N₂, then reaction mixture was heated at 50 ºC for 8 h. After cooling to RT, the reaction mixture was diluted and extracted with CH₂Cl₂ (30 mL x 2). The combined organic extract was dried over Na₂SO₄ and concentrated. The residue was purified by silica gel column chromatography (Gradient: 80% EtOAc/n-Hexane to 3% MeOH/EtOAc) and further purified by HPLC (C-18, 5 micron, 22mm x 250 mm, isocratic elution with 58% MeCN/water with 0.1%TFA) to yield 1 DPNB-ifen as a yellow liquid (10 mg, 12 %). ¹H NMR: (300 MHz, CDCl₃) δ7.70 (d, J = 4.4 Hz, 1H), 7.34–7.25 (m, 3H), 7.22 (d, J = 7.0 Hz, 2H), 7.17 – 7.07 (m, 3H), 6.74 (bs, 2H), 6.12 (q, 8Hz 1H), 5.04 (s, 1H), 4.25–4.0 (m, 4H), 3.95–3.48 (m, 22H), 3.33 (s, 6H), 2.87 (m, 2H), 2.61 (s, 1H), 2.64 (d, J = 8 Hz, 2H), 2.08–1.70 (m, 5H), 1.67 (d, J = 6 Hz, 3H), 0.98 (bs, 3H). ¹³C NMR (75 MHz, CDCl₃) δ 157.56, 153.55, 147.12, 139.55, 138.89, 134.71, 128.97, 128.64, 128.52, 126.43, 115.90, 110.06, 71.71, 70.60, 70.26, 69.22, 68.78, 68.47, 58.93, 52.33, 46.49, 41.97, 36.64, 29.14, 23.34. ESI-MS (m/z) for C₄₃H₆₃N₂O₁₂. Calcd. 799.4381. Found: m/z 799.4422 [M+H] ⁺.

Compound 7.

A mixture of 6 (300 mg, 1.11 mmol), 1-bromo-3,6,9-trioxadecane (667 mg, 2.10 mmol), potassium iodide (129 mg, 0.77 mmol), and potassium carbonate (337 mg, 2.20 mmol) in DMSO (5 mL) was heated at 100 ºC for 7 h. The reaction mixture was cooled to RT and extracted with EtOAc (30 mL x 3). The organic extracts were washed with brine, dried over Na₂SO₄ and concentrated. The residue was dissolved in water (5 mL) and THF (5 mL), and KOH (77 mg, 1.38 mmol) was added, and then heated at 70 ºC for 3 h. The reaction mixture was cooled acidified with 1N HCl, and the precipitate was filtered, dried and purified by silica gel column chromatography (50% EtOAc/n-hexane) to yield 7 as an orange solid (320 mg, 77%). ¹H NMR: (300 MHz, CDCl3) δ 8.21 (d, J = 9 Hz, 2H), 7.86 (d, J = 9 Hz, 2H), 6.68 (s, 2H), 4.22 – 4.16 (m, 2H), 3.89 (dd, J = 5.4, 4.2 Hz, 2H), 3.80 – 3.75 (m, 2H), 3.74–3.65(m, 4H), 3.61–3.55(m, 2H), 3.39 (s, 3H), 2.52 (s, 6H). ¹³C NMR (75 MHz, CDCl₃) δ. 170.9, 159.6, 156.2, 144.1, 136.2, 131.2, 130.1, 122.1, 115.3, 71.9, 70.8, 70.7, 70.6, 69.7, 67.4, 59.0, 20.8. ESI-MS (m/z) for C₂₂H₂₉N2O₆. Calcd. 417.2026. Found: m/z 417.2009 [M+H] ⁺.

Compound 8.

A solution of N,N’-dicyclohexylcarbodiimide (74.4 mg, 0.36 mmol) in THF (2 mL) was added dropwise to a solution of 7 (150 mg, 0.36 mmol) and N-hydroxysuccinimide, (45.6 mg, 0.40 mmol) at 0 ºC. The reaction mixture was stirred at RT for 24 h. The resulting precipitate was filtered and the filtrate evaporated under reduced pressure. The residue was purified by silica gel column chromatography (gradient: 40 to 60% EtOAc/n-hexane) to yield 8 as an viscous orange liquid (145 mg, 78%).¹H NMR: (300 MHz, CDCl₃) δ 8.25 (d, J = 9 Hz, 2H), 7.90 (d, J =9 Hz, 2H), 6.70 (s, 2H), 4.25 – 4.11 (m, 2H), 3.86 (dd, J = 12.3, 7.7 Hz, 2H), 3.78 – 3.73 (m, 2H), 3.72 – 3.64 (m, 4H), 3.56 (dd, J = 5.6, 3.4 Hz, 2H), 3.38 (s, 3H), 2.92 (s, 4H), 2.55 (s, 6H). ¹³C NMR (75 MHz, CDCl₃) δ169.3, 161.7, 160.1, 157.1, 144.2, 136.8, 131.8, 125.7, 122.5, 115.5, 72.1, 71.0, 70.8, 70.7, 69.8, 67.6, 59.2, 25.8, 21.1. ESI-MS (m/z) for C₂₆H₃₂N₃O₈. Calcd. 514.2189. Found: m/z 514.2178 [M+H] ⁺.

diMe-AB-ifenprodil (2).

A solution of amine 5 (38.0 mg, 0.10 mmol) and N, N-diisopropylethylamine (DIEA) (0.038 mL, 0.59 mmol) in 0.8 mL of N-methyl-2-pyrrolidinone (NMP) was added to a solution of 8 (44.8 mg, 0.09 mmol) in NMP (0.8 mL). The reaction mixture was stirred at RT for 22 h, then diluted with 50 mL water and extracted with EtOAc (30 mL × 3). The combined organic fractions were washed with brine, concentrated under reduced pressure. The residue was purified by silica gel column chromatography (gradient: 100% CH₂Cl₂ to 5% MeOH/CH₂Cl₂) to get 2 diMe-AB-ifenprodil as an orange solid (30 mg, 44%). 1H NMR: (300 MHz, CDCl3) δ7.92 (d, J = 8.4 Hz, 2H), 7.81 (d, J = 8.5 Hz, 2H), 7.46 (s, 1H), 7.25 – 7.13 (m, 4H), 7.06 (d, J = 6.9 Hz, 2H), 6.86 (d, J = 8.0 Hz, 2H), 6.65 (s, 2H), 6.18–5.12 (bs, 3H), 4.63 (bs, 1H), 4.43 (bs, 1H), 4.24 – 4.08 (m, 2H), 3.95 – 3.80 (m, 2H), 3.78–3.73 (m, 2H), 3.72 – 3.64 (m, 5H), 3.58–3.53 (m, 3H), 3.38 (s, 3H) 3.36–3.29 (m, 2H), 3.15 (dd, J = 7.3, 4.7 Hz, 1H), 3.06 – 2.73 (m, 2H), 2.53 (d, J = 5.3 Hz, 2H), 2.47 (s, 6H), 1.94–1.62 (m, 7H), 1.55 (bs, 2H), 1.32 (dd, J = 8.7, 6.0 Hz, 1H), 0.97 (bs, 3H).¹³C NMR (75 MHz, CDCl3) δ 144.2, 138.6, 135.8, 134.5, 128.9, 128.5, 128.2, 126.4, 122.2, 120.7, 117.1, 115.2, 71.8, 70.8, 70.5, 70.4, 69.6, 67.4, 64.2, 62.2, 59.0, 46.0, 41.8, 39.1, 35.9, 25.9, 23.0, 20.6, 11.5, 8.5.ESI-MS (m/z) for C47H64N5O6. Calcd. 794.4857. Found: m/z 794.4863 [M+H] ⁺.

Photolysis of DPNB-ifenprodil: quantum yield of photolysis and ifenprodil photorelease.

DPNB-ifen was photolyzed at 365 nm or 405 nm (Laserglow Technologies, LRD-0405TSR-000050–10). Liberation of ifenprodil (Fig. 4b) was followed using HPLC analysis of the reaction on an Agilent 1200 using a Proto 200 C-18 (100×4.6 mm, 3 μm. Higgins Analytical Inc) monitored at 220 nm. Elution at 1 mL/min used a linear gradient (5–100% MeCN, in water for 10 min). Both solvent contained 0.1 % TFA. The time course of photolysis (Fig. 4c) of DPNB-ifen using 405-nm light was compared to DPNB-ABT594^[32] using UPLC using a Waters Acuity Arc using Cortecs C-18 column (4.6 × 50 mm, 2.7 μm) monitored at 250 nm. Elution at 1 mL/min used a linear gradient (0–65% acetonitrile, in water for 2 min), then 65% acetonitrile/35% water for 5 min. Both solvents contained 0.1 % TFA. DPNB-ABT594 eluted at 2.68 min and DPNB-ifen at 3.37 min. Both compounds (30 μM each) gave a total absorption at 405 nm 0.07 in a 1 mL cuvette) in HEPES buffer (pH 7.4). Inosine was included as a photochemically inert standard.

Spectroscopic analysis

UV-Vis spectra were recorded using a Cary 50 spectrophotometer (Agilent, Santa Clara, CA, USA). Photostationary states of cis and trans isomers of diMe-AB-ifenprodil was determined in HEPES (pH 7.2) with 1% DMSO by continuous illumination with either UV or green LED (M365LP1, M530L3, Thorlabs, NJ, USA) until no further change in the absorption spectra. The cis-trans composition was determined by HPLC using the relative integrated areas of the cis and trans peaks at the isosbestic wavelength, considering the sum of integrated area for two peaks is 100%.

Molecular modeling.

The crystal structure of the NTDs of the NMDA receptor subunit GluN1 and GluN2B in complex with ifenprodil (PDB code 3QEL) was prepared for docking using the Protein Preparation wizard in Maestro (Schrödinger) to add missing atoms, optimize hydrogen bonds, and minimize the structure (heavy atom convergence to RMSD 0.3 Å). Structures of ifenprodil derivatives were prepared using ligprep (Schrödinger) with the OPLS3e force field and Epik ionization. A docking grid was generated centered on the ifenprodil location in the prepared NTD crystal structure. The compounds were docked into the grid using Glide (Schrödinger) with XP precision and with a core constraint using ifenprodil as the reference position with a tolerance of 0.1 Å. Induced Fit Docking was carried out with Glide/Prime (Schrödinger) using the standard protocol and docking box centered on the ifenprodil position in the crystal structure. Figures and animations were generated using PyMOL (Schrödinger).

Molecular Biology.

The pcDNA3-based expression plasmids for rat GluN1–1a (named GluN1 herein), rat GluN2A and mouse ε2 (named GluN2B herein) and mutant GluN2B-ΔNTD subunits have been described previously^[35].

Two-electrode voltage clamp (TEVC) on Xenopus oocytes.

Female Xenopus laevis were housed and oocytes harvested according to the European Union guidelines (husbandry authorization #C75–05-31; project authorization #05137.02). Xenopus laevis oocytes were harvested, prepared, injected and perfused according to previously published procedures^[30]. Recombinant NMDA receptors were expressed in Xenopus oocytes by co-injection of 32 or 50 nL of a 1:1 mixture of cDNAs coding for the wt GluN1–1a subunit and the GluN2 subunit of interest (at 10 or 30 ng/μL, nuclear injection). Oocytes were then kept at 18 °C in a Barth solution (in mM: 88 NaCl, 1 KCl, 0.33 Ca(NO₃)₂, 0.41 CaCl₂, 0.82 MgSO₄, 2.4 NaHCO₃, 7.5 HEPES, pH adjusted to 7.6 with NaOH) supplemented with gentamicin (50 μg/mL) and D-(−)-2-Amino-5-phosphonopentanoic acid (APV, 50 μM), and recorded 24–72 h after injection.

Data were collected and analyzed using pClamp 10.5 (Molecular Devices) and fitted using Sigmaplot 11.0 (Systat Software Inc). The standard external recording solution used for recordings at pH 7.3 contained (in mM): 100 NaCl, 0.3 BaCl₂, 5 HEPES and 2.5 KOH. The pH was adjusted to 7.3 with HCl. NMDAR-mediated currents were induced by simultaneous application of saturating concentrations of L-glutamate and glycine (100 μM each). UV uncaging was performed by illuminating from the top the oocyte and the perfusion chamber with a 365 nm LED (pE-2, CoolLED, UK) by the means of a liquid light guide. Unless notified, recordings were performed at a holding potential of −60 mV. All experiments were performed at room temperature.

Compounds 1, 2 and 3 were diluted as stock solutions of 20 (1 and 3) or 10 mM (2) in DMSO. The day of the experiment, they were diluted to the appropriate concentration in the recording solution and kept in the dark during the whole duration of the experiment to avoid photolysis or photoconversion. Due to the poor solubility of 2, DMSO was added to Compound 2 solutions to the final concentration of 1 %. Since DMSO itself induced a small inhibition of NMDAR currents, control and agonist solutions were also supplemented with 1 % DMSO. cis-2 solution was obtained by irradiating from the top 25 mL of trans-2 solution with 365 nm light for 20 min in a graduated cylinder covered with aluminum foil. The UV PSS was checked by UV/vis spectroscopy. The cis-2 solution was then kept in the dark during the whole course of the experiment.

Voltage-sensitivity was measured by performing 10 s voltage ramps from −100 to +40 mV. The oocyte membrane was equilibrated during 2 s at −100 mV before performing the voltage ramp. NMDAR I-V curves were obtained by subtracting offline the current measured in absence of agonists (i.e. leak current) to the current measured in presence of agonists (plus or minus Compound 3). Values are shown as mean ± standard error of the mean.