Introduction The serendipitous discoveries of the monoamine oxidase inhibitor (MAOI) iproniazid and the tricyclic antidepressant (TCA) imipramine in the late 1950s were major breakthroughs in the treatment of depression. The discovery of imipramine initiated a search for new tricyclic antidepressants using analogue design. Among these analogues were amitriptyline (1), nortriptyline (2), and melitracen (3) (Figure 11.1). Lundbeck developed new patentable syntheses of these three drugs (at that time, product patents were not obtainable, only process patents) and entered the market with these products in the early 1960s. However, the use of TCAs is associated with disturbing and serious side effects, such as dryness of the mouth, constipation, confusion, dizziness, sedation, orthostatic hypotension, tachycardia, and/or arrhythmia. Relatively narrow therapeutic indices limit the dose range in which they can be used and pose a risk in the case of overdose. As increasing knowledge was gained about the mechanism of action of the TCAs and appropriate screening assays were developed, chemists started to look for ways to improve them. TCAs either inhibit the reuptake of serotonin (5-hydroxytryptamine, 5-HT) and norepinephrine (NE) (imipramine, amitriptyline), or predominantly NE (desipramine, nortriptyline). But they also block a number of postsynaptic receptors, notably those for NE, acetylcholine, and histamine. The latter effects were mainly associated with the side effects, whereas inhibition of the monoamine transporters was associated with the therapeutic effects. Two major hypothesis of depression emerged from this research, the 5- HT and the NE hypotheses, where lowered 5-HT levels in the brain were associated with lowered mood, and lowered NE levels were associated with lowered psychomotor drive. The discovery of the selective serotonin reuptake inhibitors (SSRIs) has already been described broadly in a previous volume of this series [1]. Thus, the following will focus on the discovery of citalopram and escitalopram [2–4]. 11.2 Discovery of Talopram The discoveries of both citalopram and escitalopram (S-citalopram) are good examples of how analogue design can lead to drugs with either totally different or greatly improved therapeutic profiles compared with the starting structure. The discovery of citalopram started with the discovery of talopram (6) (Figure 11.1) in 1965. In an attempt to make a trifluoromethyl-substituted derivative of melitracen (4) using a precursor molecule (5) and reaction conditions (ring closure in concentrated sulfuric acid) similar to those used in the production of melitracen, a chemist at Lundbeck ended up with a different product, which after meticulous structural elucidation proved to be the phthalane structure 7. This compound proved to be a surprisingly selective NE uptake inhibitor. Analogue design revealed that the unsubstituted N-des-methyl analogue 6 (talopram, Lu 03-010) was a highly selective and very potent NE uptake inhibitor. The sulfur analogue of talopram, called talsupram (8, Lu 05-003), was likewise a very potent, selective inhibitor of the NE transporter (NET). Both compounds were therefore major improvements with regard to selectivity compared with the nonselective TCAs, desipramine and nortriptyline. Both drugs entered early clinical testing in the late 1960s, but were stopped in phase II for various reasons, among which was a rather activating profile. This observation was supportive of the hypothesis proposed by Professor Arvid Carlsson that NE uptake inhibition would mainly increase psychomotor drive. Carlsson had noticed that the tertiary amine drugs, which were mixed 5-HT and NE reuptake inhibitors, were “mood elevating,” while the secondary amines, being primarily NE reuptake inhibitors, increased the psychomotor drive in the depressed patients [5]. Carlsson advocated the development of selective 5-HT uptake inhibitors to treat the lowered mood state and avoid the potential suicide risk associated with an increased psychomotor drive in a depressed patient. Carlsson presented his theory at Lundbeck, and in 1971 it was decided to start a search for a selective serotonin reuptake inhibitor.  Although it may seem paradoxical, it was decided to use talopram as template for designing an SSRI. However, the reason was that there were already a few analogues of talopram available in-house that had dual 5-HT/NE uptake inhibition [2]. In particular, derivatives lacking the 3,3-dimethyl substituents of talopram, but with a chlorine either at the 5-position or at the 40 -position (see the structure in Table 11.1), were more potent as 5-HT uptake inhibitors than NE uptake inhibitors. Like their tricyclic dual-acting counterparts, these compounds also had a dimethylamino group, instead of the monomethylamino group found in the selective NE uptake inhibitors. We, therefore, decided to use the structure at the top of Table 11.1 as the core structure for a program investigating the structure–activity relationships (SARs) of aromatic substitution in the two benzene rings. 11.3 Discovery of Citalopram We published our initial SAR study (actually a quantitative structure–activity relationship (QSAR) study) in 1977 [7]. At that time, in vitro assays measuring inhibition of neuronal 5-HT and NE uptake were not available at Lundbeck, so 5-HT and NE uptake inhibition was measured as inhibition of ½ 3H
-5-HT uptake into rabbit platelets and as inhibition of ½ 3H
-NE uptake into the mouse heart ex vivo, respectively. Although these models were not directly comparable, they were acceptable as long as the goal was to develop selective compounds. Later on, assays based on inhibition of ½ 3H
-5-HT and ½ 3H
-NE uptake into rat brain synaptosome preparations as well as recombinant cell-based assays expressing the cloned human serotonin transporter (hSERT) or NET were developed. Table 11.1 shows a number of key citalopram derivatives (9–16) with the original blood platelet results as well as data for inhibition of ½ 3H
-5-HT uptake in recombinant cells expressing the cloned hSERT. All citalopram derivatives are highly selective serotonin transporter (SERT) inhibitors compared to inhibition of the NET and the dopamine transporters (DATs). Thus, data for NET and DAT inhibition are not shown. Applying the observations mentioned above regarding monosubstituted chloro derivatives led to the synthesis of the 5,40 -dichloro derivative (10, Table 11.1) that proved to be a very potent SSRI. SAR studies of inhibitors of SERT, NET, and DAT very often show that optimal potency is found in 3,4- dichlorophenyl derivatives [8]. This derivative (11) also proved to be very potent, whereas the corresponding 30 ,50 -dichloro derivative (12) had 25–100 times weaker activity. Substitution with chlorine at the 6-, 4-, and 7-positions (compounds 13, 14, and 15) was also allowed. The most potent derivatives were generally found among the 40 ,5-disubstituted derivatives substituted with F, Cl, Br, or CF3. Electron-donating groups (30 -OCH3, 40 -OCH3, or 40 -isopropyl) had very low activity [7]. At that time, a cyano group was not considered an obvious choice in systematic aromatic substitution SAR investigations. One reason was that a nitrile may be metabolically labile to be transformed to an amide or carboxylic acid. However, one of the authors of this chapter (KPB) had previously worked on a project in which nitriles were key intermediates and had used the relatively new reaction where the aromatic halogen was reacted with cuprous cyanide to produce the nitrile. In August 1972, having already synthesized the potent 5- bromo-40 -fluoro-derivative 16, he used this reaction directly on 16 and made the first sample of 9 (citalopram). The compound proved to be a very potent SSRI. It was selected together with a few other potent SSRIs from the series for further preclinical studies, proved overall to have the best safety profile, and it was selected as a clinical development candidate. Interestingly, the cyano group proved to be totally metabolically stable. 11.4 Synthesis and Production of Citalopram The syntheses used for preparation of citalopram and derivatives are outlined in Scheme 11.1. The starting materials were phthalides (I). These were either made by methods published in the literature, or by improved or completely new procedures. The majority of compounds could then be produced by a “double Grignard” reaction, in which the phthalide was reacted with a substituted phenyl magnesium bromide to give the benzophenone intermediate II, which was further reacted with 3-(dimethylamino)propyl magnesium chloride to afford the “dicarbinol” III. The dicarbinols were then ring closed in strong acid to the final product VI. The 5-bromo derivative 16 was made by this synthesis (from 5-bromophthalide) and as mentioned above, citalopram (X ¼ 5-CN, Y ¼ 40 -F) was synthesized for the first time by reacting the Br in 16 with cuprous cyanide. However, it quickly turned out that this method was unsuited for the preparation of larger quantities, so an alternative method was developed. Starting again with 5-bromophthalide, the benzophenone II was reduced with LiAlH4 to the intermediate IV, which was ring closed in strong acid to the phenylphthalane V (X ¼ 5- Br, Y ¼ 40 -F). By reaction with cuprous cyanide, a crystalline 5-cyano-1-(40 - fluorophenyl)phthalane was obtained. Citalopram was then obtained by metallation with NaH in dimethyl sulfoxide and subsequent reaction with 3-(dimethylamino)propyl chloride. This method was used to produce the first 5 kg. Unfortunately, this method was also unsuited for large-scale production. This was a critical point in the development of citalopram, but we made a very surprising discovery. It turned out that the cyano group in 5-cyanophthalide did not react with the Grignard reagents, contrary to expectation, and did not hydrolyze in the strong acid used in the final ring closure. We also found that the second “side chain” Grignard reagent could be added directly after the 4- fluorophenyl magnesium bromide without isolating the intermediate benzophenone. This was a major improvement, and this synthesis proved to be an excellent production method. 11.5 The Pharmacological Profile of Citalopram The pharmacological characterization of citalopram showed that it potently inhibited the uptake of ½ 3 H
-5-HT in rabbit and rat thrombocytes and rat brain synaptosomes with IC50 values in the nanomolar range (14 and 1.7 nM, respectively) [9, 10]. Furthermore, citalopram turned out to be the most selective 5-HT uptake inhibitor among the SSRIs in clinical use at that time, with a selectivity ratio of 3400 relative to the inhibition of ½ 3 H
-NE uptake in rat brain synaptosomes compared to selectivity ratios of 840, 280, and 54 for sertraline, paroxetine, and fluoxetine, respectively, and the selectivity relative to the inhibition of ½ 3H
-dopamine (DA) uptake was even greater, that is 22 000 compared to selectivity ratios of 250, 18 000, and 740 for sertraline, paroxetine, and fluoxetine [11]. Citalopram had very low affinity for the receptors studied, the highest affinity being for the s1 and histamine H1 receptors (IC50 200 and 350 nM, respectively) [11]. Interestingly, early binding dissociation studies of ½ 3H
-imipramine, ½ 3 H
-paroxetine, and ½ 3H
-citalopram suggested that these drugs bind to different areas of the SERT [12], even though the implications of this were unclear at that time. The available mechanistic in vivo assays only provided indirect measures of uptake inhibition and selectivity. For example, citalopram inhibited the uptake of the 5-HT depleting agent, H75/12, into neurons with an ED50 of 0.80 mg/kg and failed to inhibit NE depletion by the depleting agent H77/77 at doses as high as 160 mg/kg [9]. Similarly, studies of amine turnover showed that an acute dose ofThe 5-bromo derivative 16 was made by this synthesis (from 5-bromophthalide) and as mentioned above, citalopram (X ¼ 5-CN, Y ¼ 40 -F) was synthesized for the first time by reacting the Br in 16 with cuprous cyanide. However, it quickly turned out that this method was unsuited for the preparation of larger quantities, so an alternative method was developed. Starting again with 5-bromophthalide, the benzophenone II was reduced with LiAlH4 to the intermediate IV, which was ring closed in strong acid to the phenylphthalane V (X ¼ 5- Br, Y ¼ 40 -F). By reaction with cuprous cyanide, a crystalline 5-cyano-1-(40 - fluorophenyl)phthalane was obtained. Citalopram was then obtained by metallation with NaH in dimethyl sulfoxide and subsequent reaction with 3-(dimethylamino)propyl chloride. This method was used to produce the first 5 kg. Unfortunately, this method was also unsuited for large-scale production. This was a critical point in the development of citalopram, but we made a very surprising discovery. It turned out that the cyano group in 5-cyanophthalide did not react with the Grignard reagents, contrary to expectation, and did not hydrolyze in the strong acid used in the final ring closure. We also found that the second “side chain” Grignard reagent could be added directly after the 4- fluorophenyl magnesium bromide without isolating the intermediate benzophenone. This was a major improvement, and this synthesis proved to be an excellent production method. 11.5 The Pharmacological Profile of Citalopram The pharmacological characterization of citalopram showed that it potently inhibited the uptake of ½ 3 H
-5-HT in rabbit and rat thrombocytes and rat brain synaptosomes with IC50 values in the nanomolar range (14 and 1.7 nM, respectively) [9, 10]. Furthermore, citalopram turned out to be the most selective 5-HT uptake inhibitor among the SSRIs in clinical use at that time, with a selectivity ratio of 3400 relative to the inhibition of ½ 3 H
-NE uptake in rat brain synaptosomes compared to selectivity ratios of 840, 280, and 54 for sertraline, paroxetine, and fluoxetine, respectively, and the selectivity relative to the inhibition of ½ 3H
-dopamine (DA) uptake was even greater, that is 22 000 compared to selectivity ratios of 250, 18 000, and 740 for sertraline, paroxetine, and fluoxetine [11]. Citalopram had very low affinity for the receptors studied, the highest affinity being for the s1 and histamine H1 receptors (IC50 200 and 350 nM, respectively) [11]. Interestingly, early binding dissociation studies of ½ 3H
-imipramine, ½ 3 H
-paroxetine, and ½ 3H
-citalopram suggested that these drugs bind to different areas of the SERT [12], even though the implications of this were unclear at that time. The available mechanistic in vivo assays only provided indirect measures of uptake inhibition and selectivity. For example, citalopram inhibited the uptake of the 5-HT depleting agent, H75/12, into neurons with an ED50 of 0.80 mg/kg and failed to inhibit NE depletion by the depleting agent H77/77 at doses as high as 160 mg/kg [9]. Similarly, studies of amine turnover showed that an acute dose of citalopram reduced 5-HT turnover in the brain (i.e., [5-HT] was unchanged and the metabolite, [5-hydroxyindoleacetic acid, 5-HIAA], was decreased) and had no effect on NE synthesis. This suggested that citalopram was the most selective of the compounds tested [13, 14]. Finally, simple behavioral screening assays, such as the potentiation of 5-hydroxytryptophan (5-HTP)-induced 5-HT syndrome (5-HT uptake inhibition), tetrabenazine-induced ptosis (NE uptake inhibition), and apomorphine-induced gnawing (DA uptake inhibition), supported a selective 5-HT uptake inhibition in vivo [15]. Citalopram, like other SSRIs, had limited and variable effects in validated behavioral models predictive of antidepressant effect, such as the forced swim test, which had been validated with TCAs and MAOIs [16]. It was not until after the development of citalopram had been completed that more advanced in vivo assays, such as microdialysis, in vivo electrophysiology, and more complex behavioral models, became available. 11.6 Clinical Efficacy of Citalopram Citalopram was first launched for the treatment of major depressive disorder in 1989 in Denmark as Cipramil1 and subsequently marketed worldwide, including the United States in 1998 under the trade name Celexa1. Following the approval for major depressive disorder, citalopram was also approved for the treatment of panic disorders. After only a few years on the market, the drug attained blockbuster status. A large number of clinical short- and long-term studies of patients with major depressive disorder have been conducted with citalopram over the years, and different subsets of the clinical data have been subjected to meta-analyses as well. In general, citalopram was found to have an efficacy comparable to that of other SSRIs and, in some studies, also like other SSRIs, to be slightly less efficacious than the TCAs [17, 18]. Citalopram has also been shown to be efficacious in the treatment of other conditions, such as social phobia, obsessive compulsive disorder, post-traumatic stress disorder, mixed anxiety and depression, and poststroke depression [17, 18]. Citalopram has a favorable pharmacokinetic profile with good bioavailability and linear kinetics and a low potential for interactions with concomitant medication [17]. It is generally well tolerated with a better tolerability than the TCAs [19, 20]. These favorable properties made citalopram a good choice, particularly for depressed patients who required continuation and long-term treatment, as well as for elderly patients [17]. Overall, citalopram was on par with the other SSRIs with respect to efficacy, but has more favorable drug metabolism and pharmacokinetics (DMPK) properties (e.g., approximately 80% bioavailability and 80% protein binding, dose-proportional linear pharmacokinetics, an elimination half-life of 1.5 days, negligible pharmacological activity of metabolites, and low drug–drug interaction potential), which are the likely reasons why citalopram became such a commercial success, even though it was the fifth SSRI introduced onto the US market. 11.7 Synthesis and Production of Escitalopram The preparation of the closest analogues of citalopram, its single enantiomers, proved to be a major challenge. Direct resolution via diastereomeric salts failed after many fruitless attempts. Different chiral acids, different solvents, and different stoichiometric ratios of citalopram and acid were tried, but the major problem was the lack of crystal formation in almost all cases. A crystalline (þ)-camphor sulfonate was obtained, but no separation of enantiomers could be accomplished. Chiral high-pressure liquid chromatography (HPLC) was in its infancy, and the available analytical columns were tried with negative results. We wanted to avoid acid ring closure of resolved intermediates due to the risk of racemization. Therefore, we spent some time on various asymmetric syntheses focusing on derivatives with a partially finished side chain that after potential resolution could be transformed to the citalopram enantiomers without risk of racemization. However, these attempts were also unsuccessful. Finally, we decided to try to resolve the intermediate diol III (Scheme 11.1), although we did not have a strategy for a subsequent stereoselective ring closure at hand. We made the diasteromeric esters with the enantiomers of a-methoxya-trifluoromethyl acetic acid (Mosher’s acid, known as a shift reagent for nuclear magnetic resonance (NMR)) and tried to separate them on preparative (nonchiral) HPLC. By repeated peak shaving, we obtained small samples of the pure diastereomers. Importantly, we had noticed a seemingly spontaneous slow ring closure to citalopram of the mixture of diastereomeric esters (in the presence of triethylamine) during their synthesis. This encouraged us to try a stronger base (potassium tert-butoxide), and to our great surprise, this resulted in a stereoselective ring closure of the pure diastereomers to afford the very first small sample of the pure citalopram enantiomers. Later we found that the diol III also could be resolved by diastereomeric salt formation with di-p-toluoyl tartaric acid (DTTA) and, in this way, it became possible to produce larger quantities of the enantiomers (using a basic ring closure with a mesylate of the diol and triethylamine). This method was possibly also suited for production scale, but as chromatographic separation of the diols using simulated moving bed (SMB) technology in the late 1990s proved very effective, two SMB plants (a pilot and a full scale) were built for escitalopram production. Later, production became even more cost-effective when development chemists discovered that acidic ring closure of the R-diol led to a mixture of approximately 30% R-citalopram and 70% escitalopram, which could subsequently be isolated as citalopram and pure escitalopram. In recent years, many chemists outside Lundbeck have worked with alternative syntheses (for a recent review, see Ref. [21]) of citalopram and escitalopram. Of special interest was a publication in 2007 by researchers from Dr. Reddy’s Laboratories who published a direct resolution of citalopram using DTTA [22]. At that point, we had completed a systematic study with a large number of chiral acids (including DTTA) without finding a single one that worked. So we were not surprised to find that in our hands the Dr. Reddy method did not work. Through a series of experiments, we showed that resolution of citalopram was not possible by Dr. Reddy’s method [23]. In a subsequent response, Dr. Reddy researchers admitted that the method did not work as described. They then claimed success with a highly modified procedure, but in our hands that did not work either [24, 25]. In conclusion, the preparation and production of citalopram and escitalopram have been major challenges, but we finally succeeded in developing highly innovative and effective syntheses that are still the best and most cost-effective production methods. 11.8 The Pharmacological Profile of the Citalopram Enantiomers Shortly after the citalopram enantiomers had been successfully produced in the laboratory in 1988, they were characterized in vitro in the rat brain synaptosome assay of ½ 3H
-5-HT uptake inhibition and in vivo in the mouse 5-HTP potentiation assay. Completely unexpectedly, the 5-HT uptake inhibition of citalopram turned out to reside in the S-enantiomer, and R-citalopram was found to be practically devoid of this activity in both the in vitro and in vivo assays. These initial findings were reproduced and substantiated by studies of NE and DA uptake inhibitory activity and published in 1992 [26]. Based on these studies and the fact that racemic citalopram is a highly selective inhibitor of the SERT, R-citalopram was thought to be pharmacologically inactive. Only very few pharmacological studies were conducted with escitalopram in the next few years. It was not until new production methods (see above) made it feasible to produce escitalopram at an industrial scale and it was decided in 1997 to develop escitalopram for major depressive disorder that there was a renewed interest in pharmacological studies of escitalopram. Based on the knowledge available then, the original expectation was that escitalopram would be comparable to citalopram with respect to efficacy, and that the tolerability would be improved by removing the presumably pharmacologically inactive Renantiomer and thereby minimizing the drug load in the body.