A Photoswitchable Inhibitor of the Human Serotonin Transporter

Abstract

The human serotonin transporter (hSERT) terminates serotonergic signaling through reuptake of neurotransmitter into presynaptic neurons and is a target for many antidepressant drugs. We describe here the development of a photoswitchable hSERT inhibitor, termed azo-escitalopram, that can be reversibly switched between trans and cis configurations using light of different wavelengths. The dark-adapted trans isomer, was found to be significantly less active than the cis isomer, formed upon irradiation.

Graphical Abstract

Serotonin (1) or 5-hydroxytryptamine (5-HT) is a monoamine neurotransmitter and plays a number of important roles in physiology and neurophysiology.^1,2 Most of the human body’s serotonin is located in the enterochromaffin cells in the gastrointestinal tract, where it regulates intestinal movements. The remainder is synthesized from tryptophan in serotonergic neurons of the central nervous system, and is released into the synaptic cleft, where it activates serotonin receptors of the postsynaptic neurons. Serotonergic signaling influences neurological processes including sleep, mood, cognition, pain, hunger and aggression behaviours.³ The serotonin transporter (SERT or 5-HTT) is a member of the neurotransmitter sodium symporter (NSS) family of transporters. It is responsible for the sodium- and chloride-dependent reuptake of serotonin from the synaptic cleft back to the presynaptic neuron, which terminates the serotonergic signaling and simultaneously enables the recycling of serotonin by the presynaptic neuron.⁴

Due to its importance in synaptic transmission and serotonin homeostasis, SERT has been subject to intense pharmacological studies, which gave rise to many clinically approved antidepressants.⁵ Selective serotonin reuptake inhibitors (SSRIs), including citalopram (2), are used clinically to treat anxiety and depression (Fig. 1).^6,7 Citalopram (2) was developed by the pharmaceutical company Lundbeck and first marketed in 1989 in Denmark. It binds with high affinity and selectivity to SERT relative to other monoamine transporters.⁸ Citalopram was originally used as a racemic mixture but Lundbeck later introduced the more potent (S)-enantiomer as escitalopram (3).

Figure 1.

Chemical structures of serotonin, citalopram, and escitalopram.

Using in silico docking and experimental validation with a combination of citalopram analogs and hSERT mutants it was shown in 2010 that escitalopram binds to the same site as 5-HT, the S1 site.^9,10 Escitalopram was proposed to reside in an orientation wherein the protonated aliphatic tertiary amine forms a salt bridge to the conserved Asp98 in subsite A, the fluorophenyl group utilizes subsite B lined by Ala173, Asn177 and Thr439, whereas the cyanophenyl group engages subsite C lined by Phe335 and Phe341. This orientation was confirmed by X-ray crystallographic structure of human SERT bound to the antidepressant escitalopram in 2016.^11,12 Escitalopram locks SERT in an outward-open conformation by lodging in the central binding site, directly blocking serotonin binding and preventing the binding site from closing by protruding the cyanophenyl moiety through the aromatic lid normally occluding the substrate.^9–11 This implies that the cyano group will be most forgiving to chemical modifications because substituents would extend up through the extracellular vestibule towards a second allosteric site,¹³ which is also supported by structure-activity relationship studies on citalopram analogs,¹⁴ in a similar fashion to what we have shown for substitutions on the 3-position of the tricyclic antidepressant imipramine.¹⁵

In recent years, our group and others have developed a range of synthetic azobenzene photoswitches, either as freely diffusible photochromic ligands (PCLs), photoswitchable tethered ligands (PTLs), or photoswitchable orthogonal remotely tethered ligands (PORTLs), that can be used to optically control the function of many biological systems.^16–20 These include ion channels^21–23, G-protein coupled receptors (GPCRs)^24–26, enzymes^27,28, lipids^22,26,29, nuclear hormone receptors³⁰, and recently transporters^31–33. We now describe the development of a photoswitchable azobenzene derivative of escitalopram that allows for the precise and reversible control of hSERT activity with light.

Extensive structure-activity studies at the 4- and 5-positions of the dihydroisobenzofuran moiety of escitalopram were conducted by the Newman group.³⁴ They found that replacement of the cyano group with a bulkier phenylvinyl (4) substituent retained most of the binding affinity with SERT (racemic K_i = 9.32 nM, (S)-enantiomer, K_i = 10.6 nM), and it showed higher binding affinity at SERT compared to its saturated phenylethyl analogue (racemic, K_i= 38.1 nM). Compound 4 is an azobenzene isostere (“azostere”) which allows for straightforward design of an azobenzene analog which largely retains the structure of the parent compound. This well-precedented azologization approach³⁵ yields a version of escitalopram that can undergo photoisomerization. We hypothesized that the two photoisomers would exhibit marked structural differences, resulting in different potencies, which would enable the optical control of SERT activity.

We termed the azostere of the escitalopram derivate 4 “azo-escitalopram” (5). The synthesis of 5 began from commercially available escitalopram oxalate (6). Neutralization of 6 afforded the free amine, which underwent a rhodium-catalyzed nitrile hydration reaction to afford the primary amide 7.³⁶ An NBS mediated Hofmann rearrangement and deprotection of the resulting carbamate 8 generated aniline 9.³⁷ A Baeyer-Mills reaction of aniline 9 with nitrosobenzene then afforded the desired product 5, which was isolated in its hydrochloride salt form (Scheme 1).

Scheme 1.

Design and synthesis of azo-escitalopram (5).

In its dark-adapted state, azo-escitalopram (5) exists in its thermally stable trans configuration with high absorption at 340 nm. It could be isomerized to its cis-form with UV-A (λ = 365 nm) light. Isomerization from cis to trans could be achieved with blue light (λ = 460 nm). Photoisomerization could be repeated over many cycles without obvious fatigue. In its cis form, 5 remained stable in the dark (Scheme 2 B).

Scheme 2.

Photoisomerization of azo-escitalopram (5) in 10% DMSO in PBS (50 μM). A) Absorption spectra of 5 in the dark-adapted, UV-A adapted, and blue-adapted state. B) Repeated cycles of trans-cis isomerization of 5 with 460 nm and 365 nm light.

With ample supplies of azo-escitalopram (5) in hand, we proceeded to test its action on the serotonin transporter. We performed uptake assays on HEK293MSR cells expressing hSERT to study the potency of 5 in inhibition of ³H-5-HT uptake. A dilution series of 5 was prepared in the dark, mixed with ³H-5-HT and then distributed to microwell plates where the compounds were either left in the dark or illuminated with 365 nm light from a handheld device for two seconds every minute in the time leading up to the experiment. Cells were first preincubated for 25 minutes with the inhibitor in either darkness or under illumination with 365 nm for two seconds every minute to achieve equilibrium. The uptake was initiated by addition of the inhibitor mixed with ³H-5-HT and was allowed to proceed for 10 minutes in either darkness or under illumination with 365 nm for two seconds every minute. The uptake was terminated by aspiration and washing.

We found that the potency of escitalopram was, as expected, unaffected by darkness or light at 365 nm whereas the potency of 5 was low in the dark (trans-state) but increased 43-fold (p=0.0087, paired t-test) in the cis-state after illumination with 365 nm light. The potency of 5 in the cis-state was found to be 18.9 nM as compared to 819 nM in the trans-state. To confirmed that this increase in activity was not due to rapid degradation to a more potent derivative, we exposed 5 to glutathione (10 mM) for 1h in the presence and absence of light. The azobenzene was found to be stable under these conditions (Supporting Fig. 1)

In order to show that the photoswitchable inhibitor could be used to control the transport activity of hSERT in real time, we investigated the ability of 5 to photoswitch during an activity assay for the serotonin transporter (Figure 3). We initially equilibrated the transporter with trans-5 at 460 nm for 20 minutes and then initiated uptake by addition of ³H-5-HT. The inhibitor was maintained in the trans configuration for 8 minutes by brief illumination with light at 460 nm for two seconds every minute, after which the cells were exposed to different illumination regimes either by continued illumination at 460 nm for two seconds every minute to maintain the trans configuration or 5 was switched to the cis configuration by illumination at 365 nm for two seconds every minute for 22 minutes (figure 3A). We observed that 5-HT uptake was reduced significantly at 365 nm at three subsequent time points (Figure 3A) but also by comparing the normalized uptake rates in the linear phase (linear regression for incubation time 8–24 minutes) of the uptake assay in four independent experiments (Figure 3B).

Figure 3.

The transport activity of hSERT can be controlled in real time through different illumination regimes of azo-escitalopram (5). (A) HEK293MSR cells were transiently transfected with hSERT in the pCDNA3 vector and challenged with 200 nM 5 prior to and during uptake of tritiated 5-HT. Cells were briefly illuminated with 460 nm for two seconds every minute during 20 minutes of pre-incubation and for the first 8 minutes of the uptake assay. After 8 minutes of incubation with tritiated 5-HT half of the cells were then subjected to brief illumination with 365 nm for two seconds every minute (the shift in illumination regime is marked by an arrow). Shown is the global fit of normalized data for four independent experiments where error bars on individual point represent SEM. Uptake at individual time points at different illumination regimes (12–30 minutes) were compared using the two-stage linear step-up procedure of Benjamini, Krieger and Yekutielia for paired t-test (p<0.001: ***).³⁸ (B) The uptake rates at 200 nM 5 was obtained from the slopes of the linear phase of the uptake assay at different illumination regimes (8–24 minutes in (A)) by linear regression analysis and then normalized to the uptake rates for cells challenged with 200 nM 5 illuminated with 460 nm. Comparison of the normalized uptake rates from four independent experiments exhibited highly statistically significant reduction (p<0.001, t-test) in uptake for hSERT exposed to cis-5 (365 nm) as opposed to trans-5 (460 nm).

Since the trans form of 5 sterically resembles the potent stilbene inhibitor 4, it was unexpected, albeit welcome, that cis-5 was more potent than trans-5. In order to better understand the cis- selectivity of 5, we performed induced fit docking simulations (Figure 4). Docking of cis-5 resulted in 100 poses which were clustered into 11 clusters, termed cis-C1–11 clusters (Supporting Table S1). Only one large cluster was identified: This cis-C2 cluster held 75 poses and contained the best scoring poses. Several elements in this cluster display the hallmarks of high-affinity inhibitors of hSERT identified earlier^10–12, most notably a salt bridge between the protonated amine of the ligand and Asp98 (55 of 75 poses), cation-π interaction between the protonated amine of the ligand and Tyr95 (73 of 75 poses) and a π-π interaction between the fluorophenyl of the ligand and Tyr176 (54 of 75 poses). Overall, cis-5 exhibited most similarity with the binding of the parent inhibitor, escitalopram (Figure 4A), whereas trans-5 produced a range of clusters, termed trans-C1–17 clusters, of which only two were of a significant size (27 and 32 out of 100 poses) (Supporting Table S1). For the trans-C1 cluster, the ligand is found more in the low-affinity allosteric S2 site rather than in S1^13,39,40, whereas the other major cluster (trans-C12) resembles the binding mode of the parent inhibitor, escitalopram (Figure 4B), albeit with some notable perturbations. For example, the important salt bridge between the protonated amine of the ligand and Asp98 is only found in less than one third of the trans-C12 poses (10 of 32 poses), the cation-π interaction between the protonated amine of the ligand and Tyr95 is found in only approximately half of the trans-C12 poses (17 of 32 poses) and a favourable π-π interaction between the fluorophenyl of the ligand and Tyr176 is found in less than half of the trans-C12 poses (15 of 32 poses). Overall, the significant similarity between cis-5 binding and escitalopram binding as well of the retainment of well-established key protein-ligand interactions responsible for the high affinity of escitalopram¹² supports that cis-5 should exhibit superior potency for inhibition of hSERT compared to trans-5 as the experimental data also shows (Figure 2–3).

Figure 4:

Induced fit docking of cis-5 (left) and trans-5 (right) demonstrates the molecular basis for hSERT selectivity for cis-5. (A) Out of a total of 100 docking poses for cis-5 (light grey), 75 poses (cis-C2 cluster) assume a binding mode highly similar to the high-affinity parent inhibitor, escitalopram (dark grey). (B) Out of a total of 100 docking poses for trans-5 (light grey), only 32 poses (trans-C12 cluster) assume a binding mode somewhat similar to the high-affinity parent inhibitor, escitalopram (dark grey), with notable disruption of the important salt bridge between the protonated amine of the inhibitor and the carboxylic acid side-chain of D98 as well as perturbations of the interactions between the inhibitor and the π-systems of Y95 and Y176.

Figure 2.

Azo-escitalopram (5) is a potent, photoswitchable inhibitor of the human serotonin transporter. (A) HEK293MSR cells were transiently transfected with hSERT in the pCDNA3 vector and exposed to the inhibitors either in the dark or illuminated at 365 nm for two seconds every minute prior to and during uptake of tritiated 5-HT. Shown is a representative example of normalized transport activity plotted as a function of drug concentration and fitted to a sigmoidal dose-response curve. Error bars represent standard error mean. (B) Mean K_i from three independent experiments show that the potency of the parent compound, escitalopram, is unaffected by irradiation at 365 nm whereas the inhibitory potency of 5 is significantly increased (P=0.0087, paired t-test) by photoswitching to the cis state by irradiation with light at 365 nm.

In conclusion, we have designed and synthesized azo-escitalopram (5) as the first photochromic inhibitor of hSERT and the first photopharmacological tool for serotonin biology at large. It can be reversibly switched between trans (low-affinity) and cis (high-affinity) configurations using light of different colors. We find that the cis configuration has nanomolar potency and is 43-fold more potent as an inhibitor than the trans configuration. Induced-fit docking calculations identify key protein-ligand interactions that are retained in cis-5 and perturbed in trans-5. As such 5 could be a useful tool for the spatiotemporal regulation of serotonin in neural networks with the precision that light affords.