Clickable Photoaffinity Ligands for the Human Serotonin Transporter Based on the Selective Serotonin Reuptake Inhibitor (S)-Citalopram

Abstract

To date, the development of photoaffinity ligands targeting the human serotonin transporter (hSERT), a key protein involved in disease states such as depression and anxiety, have been radioisotope-based (i.e., ³H or ¹²⁵I). This letter instead highlights three derivatives of the selective serotonin reuptake inhibitor (SSRI) (S)-citalopram that were rationally designed and synthesized to contain a photoreactive benzophenone or an aryl azide for protein target capture via photoaffinity labeling and a terminal alkyne or an aliphatic azide for click chemistry-based proteomics. Specifically, clickable benzophenone-based (S)-citalopram photoprobe 6 (hSERT K_i = 0.16 nM) displayed 11-fold higher binding affinity at hSERT when compared to (S)-citalopram (hSERT K_i = 1.77 nM), and was subsequently shown to successfully undergo tandem photoaffinity labeling-biorthogonal conjugation using purified hSERT. Given clickable photoprobes can be used for various applications depending on which reporter is attached by click chemistry subsequent to photoaffinity labeling, photoprobe 6 is expected to find value in structure-function studies and other research applications involving hSERT (e.g., imaging).

Table of Contents Graphic

The human serotonin transporter (hSERT) is a clinically significant neurotransmitter-sodium symporter (NSS) protein targeted by therapeutic medicines (e.g., selective serotonin reuptake inhibitors (SSRIs) and tricyclic antidepressants (TCAs) for treatment of anxiety and depression) and addictive drugs (e.g., cocaine and amphetamines), towards affecting serotonergic neurotransmission.¹ At this point in time, the molecular pharmacology and number of ligand-binding sites in hSERT and other NSS proteins by crystal structure analysis has been determined with contentious results.^2–10 In brief, some studies provide corroboration for two binding sites within hSERT: the S2 site located within the protein’s extracellular-facing vestibule, and the central S1 site that holds the substrate serotonin in an occluded state upon closure of the S1 site outer gate.^11–17 However, other studies indicate the S1 site by itself accounts for high-affinity binding of serotonin and inhibitors.^{4–5, 18–19} Additionally, another unique binding site located outside the serotonin translocation pathway has been reported and utilized in the discovery of a hSERT allosteric modulator with a novel mechanism of action.²⁰

Towards addressing these controversies regarding the function and number of ligand-binding sites in hSERT, and alternative to crystal structure analysis or singlemolecule force spectroscopy,²¹ photoaffinity ligands (PALs) can be used as molecular tools to directly determine ligand-protein interactions in cells.²² Over the past 10 years, clickable photoaffinity ligands have dominated and augmented the value of photoaffinity labeling for a number of disparate research applications, including 1) target identification of hit compounds emanating from phenotypic screening, 2) compound mechanism of action studies, 3) affinity-based protein profiling (AfBPP; e.g., confirmation of target engagement or selectivity profiling), 4) imaging applications, 5) binding site location and mapping, and 6) elucidation of ligand-target interactions.²³

Traditionally, photoprobes have contained radiolabeled atoms (e.g., ³H, ¹²⁵I), serving as a detection method. Several radioactive hSERT photoprobes have been reported based on serotonin,²⁴ tropanes,^25–26 the tricyclic antidepressant (TCA) imipramine,^27–28 and the selective serotonin reuptake inhibitors (SSRIs) paroxetine²⁹ and (S)-citalopram;³⁰ however, none of these photoprobes have elucidated the structural basis for non-covalent interactions with hSERT. Using radioactive isotopes as reporter tags for detection purposes after photoaffinity labeling has several drawbacks, including practical concerns of special handling, the lack of an affinity handle for enrichment of probe-labeled biological macromolecules, and potentially short half-lives due to chemical degradation. The preferable alternative is to develop non-radioactive photoprobes containing a click chemistry handle.

The experimental strategy known as “tandem photoaffinity labeling-bioorthogonal conjugation” is particularly versatile and highly advantageous.³¹ This is namely because the same clickable photoprobe can be used for various applications depending on which reporter is attached by click chemistry subsequent to photoaffinity labeling (e.g., attachment of a fluorophore for in-gel fluorescence detection, biotin for identification/enrichment, etc.). As a result, clickable photoprobes are expected to be highly desirable compounds for hSERT structure-function studies that are not only capable of addressing the disadvantages associated with previously reported radioisotope-based probes, but also to aid in the rational design of compounds capable of modulating hSERT function in a controllable and desirable therapeutic manner.³²

To our knowledge, only one clickable, covalent hSERT photophobe has been reported: azidobupramine, a terminal alkyne-containing derivative of imipramine that also contains a photoreactive aryl azide functional group for protein target capture.³³ Therefore, to expand the collection of multifunctional tools derived from clinically approved antidepressants, we report herein our initial development of clickable photoprobes 5 – 7 based on (S)-citalopram (1) (Figure 1). (S)-citalopram was chosen as our first clickable hSERT photoprobe target for two reasons. First, (S)-citalopram uniquely displays high affinity for both the S1 and S2 hSERT binding sites, and is thus capable of modulating its own pharmacological activity.³⁴ Second, in vivo studies involving mice treated with (R)-citalopram (i.e., the stereochemical enantiomer of 1) in combination with different SSRIs indicated that (R)-citalopram inhibits the serotonin signal-enhancing effects of SSRIs via an in vitro allosteric self-potentiating effect, as observed in dissociation studies involving paroxetine and escitalopram, but does not antagonize the serotonin signal-enhancing effect of fluoxetine as a non-allosteric SSRI.³⁵ In particular, the latter observations potentially indicate elaborate interactions between serotonin and SERT inhibitors at the S2 site, which has led to a call for better tool compounds to provide direct experimental evidence for an allosteric in vivo mechanism.³⁶

Figure 1.

Chemical structures of (S)-citalopram (1), radioactive azido-iodo (S)-citalopram PALs 2 - 4, and novel clickable (S)-citalopram PALs 5 – 7.

To begin our work, we postulated that the C-1 and C-5 positions of (S)-citalopram could tolerate incorporation of an “all-in-one moiety”^{23, 31} (i.e., a structural motif that contains a photoreactive functional group for protein capture via photoaffinity labeling and a click chemistry functional group as a latent affinity handle) without significant loss in hSERT binding affinity. This hypothesis, used in the design of radioactive azido-iodo (S)-citalopram PALs 2 – 4³⁰ (Figure 1), was based on previously described structureactivity relationships for citalopram analogs^37–39 that suggested these two positions could accommodate the significant steric bulk associated with known “all-in-one moieties” without an appreciable decrease in hSERT binding affinity. To test this, novel clickable (S)-citalopram PALs 5 – 7 (Figure 1) were chemically synthesized and pharmacologically evaluated for binding affinity to hSERT as described below.

First, we designed photoprobe 5 to contain a benzophenone photoreactive functional group and a propargyl ether click chemistry handle attached as an “all-in-one moiety” to the C-1 position of (S)-citalopram (Scheme 1). To synthesize this compound, we envisioned N-alkylation of N-desmethyl (S)-citalopram (13)³⁸ with a benzophenone propargyl ether containing a benzylic leaving group (12). Specifically, mesylate 12 was synthesized by first protecting the ketone of the known benzophenone ester 8⁴⁰ as a ketal, and then employing a sequence of methyl ester reduction, ketal deprotection, and conversion of the 1° alcohol to the corresponding mesylate. Final alkylation of 2° amine 13 with mesylate 12 provided target photoprobe 5 in 31% yield. It should be noted that concomitant to this work, the synthesis of a benzyl bromide derivative of alcohol 11 was reported, employing six steps and proceeding in 19% overall yield,⁴¹ whereas mesylate 12 was produced here in seven steps and 15% overall yield. However, benzophenone 5 showed a 275-fold loss in binding affinity (K_i = 487 nM) compared to (S)-citalopram (K_i = 1.77 nM) at hSERT-expressing HEK293 cells (Table 1), thus effectively eliminating this compound as a candidate for future hSERT photoaffinity labeling experiments. The outcome was consistent with the report of C-1-substituted azido-iodo (S)-citalopram photoprobe 2 sustaining a ~100-fold loss in serotonin reuptake inhibition potency compared to (S)-citalopram.³⁰

Scheme 1.

Synthesis of clickable (S)-citalopram PAL 5.^a

Table 1.

Inhibition of [¹²⁵I]-RTI-55 binding of (S)-citalopram (1) and clickable (S)-citalopram PALs 5 – 7 at hSERT-HEK293 cells.

	hSERT binding affinitya
Compound #	(Ki,nM)
1,(S)-ciatalopram	1.77±1.14
5	487±97
6	0.16±0.04
7	10.7±7.5

^aEach Ki value represents data from at least three independent experiments with each data point on the curve performed in duplicate.

Next, we turned our attention to the design of photoprobe 6, which also contains a photoreactive benzophenone for protein capture and a propargyl ether as a click chemistry handle, but this time attached as an “all-in-one moiety” to the C-5 position of (S)-citalopram via an amide functional group (Scheme 2). Specifically, photoprobe 6 was readily synthesized in good yield by EDC coupling the nitrile-reduced form of (S)-citalopram (16)³⁰ with known benzophenone-propargyl ether-carboxylic acid 14.⁴⁰ In sharp contrast to C-1-subsituted probe 5 (hSERT K_i = 487 nM), C-5-subsituted probe 6 (hSERT K_i = 0.16 nM) displayed an 11-fold improvement in hSERT binding affinity relative to (S)-citalopram (hSERT K_i = 1.77 nM) (Table 1). To our knowledge, benzophenone 6 displays the highest hSERT binding affinity of any reported photoprobe to date. To maximize chances of photoaffinity labeling success, it is generally advisable not to limit photoprobe design to one type of photoreactive group, but rather to include at least one nitrene- or carbene-based photoprobe in addition to one benzophenone-based photoprobe.²³ With this in mind, photoprobe 7 was designed and synthesized, once again by EDC coupling 1° amine 16³⁰ with known carboxylic acid 15,⁴² to contain a wellknown diazide motif⁴³ attached to the C-5 position of (S)-citalopram via an amide linkage. Specifically, diazide 7 contains (S)-citalopram conjugated to a phenyl ring bearing an aryl azide photoreactive group for protein capture and an aliphatic alkyl azide as a latent affinity handle, the latter of which survives specific photolysis conditions and can be derivatized via Staudinger-Bertozzi ligation, traditional copper-catalyzed Huisgen 1,3-dipolar cycloaddition, or strain-promoted click chemistry reactions after photoaffinity labeling. While comparing unfavorably to benzophenone-based probe 6 (hSERT K_i = 0.16 nM), diazide-based probe 7 retained strong hSERT binding affinity (hSERT K_i = 10.7 nM), only 6-fold lower than (S)-citalopram (hSERT K_i = 1.77 nM).

Scheme 2.

Synthesis of clickable (S)-citalopram PALs 6 and 7.

Finally, as a way of demonstrating the functional utility of these compounds, we performed tandem photoaffinity labeling-bioorthogonal conjugation using benzophenonebased probe 6 and purified hSERT⁴⁴ expressed in tetracycline-induced HEK293 cells (Figure 2). Briefly, hSERT was purified using anti-FLAG affinity beads by single-step immunoaffinity chromatography, incubated with 1 μM of photoprobe 6 in the presence or absence of 100 μM (S)-citalopram (1) as a competitor, and then irradiated with 350–450 nm UV light for 20 minutes. Copper-catalyzed click chemistry with IRDye 800CW azide was then carried out at room temperature for 1 hour analogous to that previously described,⁴⁵ followed by hSERT visualization using FLAG antibody detection on a Western blot. As expected, hSERT-selective photolabeling was only detected when the photoprobe 6 plus hSERT combination was UV-irradiated (Figure 2, Lane 3). The addition of 100 μM (S)-citalopram (1) competitor eliminated the hSERT crosslinking signal (Figure 2, Lane 4). Control experiments lacking UV light (Lane 1) or photoprobe 6 registered negligible hSERT photoaffinity labeling.

Figure 2.

Tandem photoaffinity labeling-biorthogonal conjugation of purified hSERT. Lane 1 = 1 μM of photoprobe 6 in the presence of purified hSERT, but not exposed to UV light conditions for hSERT photoaffinity labeling; Lane 2 = a DMSO solution of purified hSERT without photoprobe 6 present and exposed to UV light conditions for hSERT photoaffinity labeling; Lane 3 = 1 μM of photoprobe 6 in the presence of purified hSERT and exposed to UV light conditions for hSERT photoaffinity labeling; Lane 4 = 1 μM of photoprobe 6 plus 100 μM of (S)-citalopram (1) in the presence of purified hSERT and exposed to UV light conditions for hSERT photoaffinity labeling. Panel A = detection of hSERT (red) via FLAG antibody and IRDye 800CW azide conjugated to photoprobe 6 (green); Panel B = detection of hSERT (red) via FLAG antibody; Panel C = detection of IRDye 800CW azide conjugated to photoprobe 6 (green); Panel D = % the relative intensity of staining observed for citalopram crosslinking in the monomeric band in Panel C. Staining intensity was limited to the monomeric bands to avoid any artifacts due to potentially reduced staining in immunoblots of dimer and oligomer bands due to limited access to epitopes. For less-magnified gel images that show the monomeric, dimeric, and oligomeric hSERT bands, and the absence of staining outside of these bands, please see Figure S2 in the Supporting information.

In conclusion, the most important outcome of the work briefly described in this letter is a high-affinity hSERT clickable photoaffinity ligand (i.e., 6) that is capable of undergoing tandem photoaffinity labeling-biorthogonal conjugation using purified hSERT. The motivation for this work was to expand the collection of multifunctional tools derived from clinically approved antidepressants, namely by focusing on the most pharmacologically relevant SSRI, citalopram, and by addressing the disadvantages associated with radioisotope-based hSERT photoprobes. To our knowledge, photoprobe 6 is the highest-affinity hSERT photoprobe reported, as well as the first clickable photoprobe based on a SSRI. We anticipate this photoprobe will find value in structure-function studies and other research applications involving hSERT (e.g., imaging) or other clinically-relevant members of the neurotransmitter-sodium symporter family.