Application of the N-Dibenzyl Protective Group in the Preparation of β-Lactam Pseudopeptides

Despite the great importance of β-lactam antibiotics, there is still a limited number of synthetic approaches for the formation of β-lactam–containing dipeptides. In this study, we report upon the stereoselective preparation of β-lactam–containing pseudopeptides, where different reaction conditions and NH2 protective groups were tested to obtain compounds that contain 3-amino-azetidin-2-one. We demonstrate that the protective group is essential for the outcome of the reaction. Successful implementation of dibenzyl-protected serine-containing dipeptides through the Mitsunobu reaction can provide the desired products at high yields and stereoselectivity.


Introduction
The β-lactam heterocycle is the main building block of the most broadly used antibacterial agents such as the penams (penicillin), penems, carbapenems, cephems and monobactams. Moreover, monocyclic β-lactams are potent cholesterol absorption inhibitors, which act through the inhibition of the enzyme cholesterol acyl transferase [1]. They are also dopamine receptor antagonists [2] and vasopressin V1a antagonists [3]. Spirocyclic β-lactams have also been shown to be potent type II β-turn-inducing peptide mimetics [4]. In addition, these heterocycles also have several other biological activities, such as antitumor, antitubercular, hypoglycemic, and anti-inflammatory actions. Their pharmacological and chemical activities greatly depend on the types of substitution at the β-lactam ring. However, synthesis of compounds with an appropriately substituted β-lactam ring with the desired diastereoselectivity is a challenging task for medicinal chemists [5,6].
There are numerous approaches for the synthesis of the β-lactam ring. The most frequently applied synthetic strategies include cycloaddition reactions, such as the Staudinger or Gilman-Speeter reactions, the Kinugasa reaction, or different cyclization reactions [7,8]. These last reactions can occur through formation of N1-C2, N1-C4, or C3-C4 bonds. Interestingly, the Mitsunobu reaction, which is the subject of the present report, has rarely been applied for the formation of the N1-C4 bond [9,10].
In this study, we focused on N1-C4 β-lactam ring closure with less common reactants: dipeptides. There are only a few methodologies in the literature that describe the formation of β-lactam pseudopeptides without a substituent on the β-lactam ring [10][11][12][13][14][15][16][17][18][19]. Versatile and efficient synthetic routes for β-lactam rings that contain dipeptides are thus needed. Recently, a new pathway for the synthesis of these compounds was developed that involves N-alkylation of the β-lactam ring with a haloacetate electrophile followed by the introduction of a side chain in the α-position with respect to the ester carbonyl group (Scheme 1a) [15]. However, despite a number of different compounds being synthesized by this methodology, it was not possible to form the desired monoalkylated products stereoselectively. In addition, there were also limitations in the selection of the electrophiles used

Results and Discussion
During an ongoing project that is devoted to the preparation of pharmacologically interesting monocyclic β-lactam analogs (e.g., Figure 1, the nocardicins), we initially faced the task of synthesizing β-lactam phenylalanine derivatives that differ from nocardicins in their additional CH2 group between the phenyl ring and the exocyclic Cα'-H group.
The initial aim of this study was to find the optimal reaction conditions to prepare β-lactam pseudopeptides by the N1-C4 cyclization of serylphenylalanine derivatives, such as 1a (Scheme 2). The effects of various N-protecting groups on the reaction outcome, and the application of these findings to other chemically similar dipeptides under different reaction conditions, were first screened. A hypothesis was defined such that because of the steric and electronic properties of a protective group, it would be essential to the formation of both the β-lactam ring and the amounts of the dehydrated products.
Three very different protective groups were applied on the primary serine NH2 group, each with very distinctive electron-withdrawing properties that could potentially influence the reaction outcome: phthalimide (1a), trityl (1b) and dibenzyl (1c) protective groups. Both, phthalimide (1a) and dibenzyl (1c) protective groups were used to mask hydrogens of the primary NH2 in order to prevent the well-documented risk of aziridine formation in the cyclodehydration step [20]. In contrast, the application of the trityl protection was rationalized by the idea that its unmasked N-H group would not participate in the following reactions, because of its low acidity and the steric hindrance of the triphenylmethane moiety. In addition, both trityl (1b) and dibenzyl (1c) protective group also have lower electron withdrawing properties compared to the phthalimide group (1a).
Next, each of the three compounds 1a-1c was treated under different reaction conditions. The idea was also to avoid high reaction temperatures and strong bases, such as lithium diisopropylamide (LDA) or NaH, which have otherwise been used in other β-lactam ring preparations [11,21]. Firstly, the conversion of an alcoholic OH group of serine to a sulfonate group was tested because it is widely used in aziridine ring formation, in which carbamate or trityl nitrogen is used as the nitrogen source [22,23]. Furthermore, the activation of a serine OH group via conversion to an imidazolyl sulfonate has been reported to yield 3-aminonocardicinic acids [24,25]. It was therefore postulated that each of the derivatives 1a-1c will react with the amide nitrogen instead. For this purpose, three reagents were used: p-toluenesulfonyl chloride (condition a); sulfuryl chloride (condition b); and methanesulfonyl chloride (condition c). Unfortunately, all three reactions by which a sulfonate was formed (i.e., conditions a, b, c) as an intermediate were not suitable for -lactam ring formation. These conditions yielded only a product of ester hydrolysis 4, the aziridine 5 or a mixture of chlorides 6a and 6b in a ratio of 1:1 (which we were not able to separate) (Scheme 2).
Secondly, compound 1a was treated under Mitsunobu reaction conditions, which is commonly applied for the preparation of nocardicin derivatives. However, to the best of our knowledge, there are no previous reports on the preparation of β-lactam phenylalanine derivatives using the Mitsunobu reaction. Furthermore, attempts to prepare less reactive glycine derivatives using the Mitsunobu reaction have given the corresponding β-lactam ring at poor yields. In addition, the mixtures obtained were difficult to separate, which was explained by the lower acidity of the exocyclic Cα'-H group which consequently decreases the yield of a reaction and increases the number of side products [16]. It was therefore not surprising that our first attempts to cyclize 1a (Scheme 2) under the classical Mitsunobu conditions using PPh3/DIAD or DEAD gave only low yields of βlactam 3a and dehydropeptide 2, in nearly equimolar ratios. These are difficult to separate in larger For the successful reaction here, protection of both of the hydrogens of the primary amine functionality in the serine was demonstrated to be of vital importance. When the carboxybenzyl (Cbz) or tert-butyloxycarbonyl (Boc) protective groups were applied, the reaction resulted in only a mixture of aziridine and dehydropeptide (Scheme 1c, Cbz, Boc, respectively). The use of phthalimide (Pht) or 4,5-diphenyloxazoline-2-one (Ox) protective groups is thus essential for the successful reaction (Scheme 1b) [16,17,20].

Results and Discussion
During an ongoing project that is devoted to the preparation of pharmacologically interesting monocyclic β-lactam analogs (e.g., Figure 1, the nocardicins), we initially faced the task of synthesizing β-lactam phenylalanine derivatives that differ from nocardicins in their additional CH 2 group between the phenyl ring and the exocyclic Cα'-H group.
The initial aim of this study was to find the optimal reaction conditions to prepare β-lactam pseudopeptides by the N1-C4 cyclization of serylphenylalanine derivatives, such as 1a (Scheme 2). The effects of various N-protecting groups on the reaction outcome, and the application of these findings to other chemically similar dipeptides under different reaction conditions, were first screened. A hypothesis was defined such that because of the steric and electronic properties of a protective group, it would be essential to the formation of both the β-lactam ring and the amounts of the dehydrated products. quantities. In addition, the application of P(OEt)3 instead of PPh3 in the Mitsunobu reaction, which has been successfully applied in the synthesis of nocardicin analogs, resulted in no products whatsoever [13,18]. Our findings were therefore in agreement with the literature reports that have described the influence of the Cα'-H group acidity on the reaction outcome (Scheme 1) [16]. Further optimization of the reaction conditions through changing the solvent and temperature increased the yield of β-lactam 3a to 45%, but the ratio of dehydrodipeptide 2 to 3a remained at 1:1. A modified procedure that would give larger quantities of β-lactam using simple cyclization without the formation of dehydrodipeptide was therefore required. Compounds 1b and 1c with trityl and dibenzyl protection were then treated under the same classical Mitsunobu reaction conditions as 1a.
Even though the tritylated compound 1b gave no product whatsoever, a combination of the classical Mitsunobu reaction and dibenzyl protection, gave positive results, whereby dipeptide 1c was transformed into the desired SS diastereoisomer 3c at 76% isolated yield. Epimerization at the exocyclic Cα'-H stereogenic center was also seen, with the diastereoisomeric ratio (dr) of SS to SR of 7:1; these were easy to separate. Importantly, no dehydrodipeptide side products were seen, and the yields were much higher when compared to the results of the Mitsunobu reaction for phthalimide protected dipeptide 1a. Furthermore, the outcome of the reaction gave higher isolated yields when compared to the less reactive phthalimide-protected glycine [16] and cyclopropyl derivatives [17] that have been reported in the literature.
Three very different protective groups were applied on the primary serine NH 2 group, each with very distinctive electron-withdrawing properties that could potentially influence the reaction outcome: phthalimide (1a), trityl (1b) and dibenzyl (1c) protective groups. Both, phthalimide (1a) and dibenzyl (1c) protective groups were used to mask hydrogens of the primary NH 2 in order to prevent the well-documented risk of aziridine formation in the cyclodehydration step [20]. In contrast, the application of the trityl protection was rationalized by the idea that its unmasked N-H group would not participate in the following reactions, because of its low acidity and the steric hindrance of the triphenylmethane moiety. In addition, both trityl (1b) and dibenzyl (1c) protective group also have lower electron withdrawing properties compared to the phthalimide group (1a).
Next, each of the three compounds 1a-1c was treated under different reaction conditions. The idea was also to avoid high reaction temperatures and strong bases, such as lithium diisopropylamide (LDA) or NaH, which have otherwise been used in other β-lactam ring preparations [11,21]. Firstly, the conversion of an alcoholic OH group of serine to a sulfonate group was tested because it is widely used in aziridine ring formation, in which carbamate or trityl nitrogen is used as the nitrogen source [22,23]. Furthermore, the activation of a serine OH group via conversion to an imidazolyl sulfonate has been reported to yield 3-aminonocardicinic acids [24,25]. It was therefore postulated that each of the derivatives 1a-1c will react with the amide nitrogen instead. For this purpose, three reagents were used: p-toluenesulfonyl chloride (condition a); sulfuryl chloride (condition b); and methanesulfonyl chloride (condition c). Unfortunately, all three reactions by which a sulfonate was formed (i.e., conditions a, b, c) as an intermediate were not suitable for β-lactam ring formation. These conditions yielded only a product of ester hydrolysis 4, the aziridine 5 or a mixture of chlorides 6a and 6b in a ratio of 1:1 (which we were not able to separate) (Scheme 2).
Secondly, compound 1a was treated under Mitsunobu reaction conditions, which is commonly applied for the preparation of nocardicin derivatives. However, to the best of our knowledge, there are no previous reports on the preparation of β-lactam phenylalanine derivatives using the Mitsunobu reaction. Furthermore, attempts to prepare less reactive glycine derivatives using the Mitsunobu reaction have given the corresponding β-lactam ring at poor yields. In addition, the mixtures obtained were difficult to separate, which was explained by the lower acidity of the exocyclic Cα'-H group which consequently decreases the yield of a reaction and increases the number of side products [16]. It was therefore not surprising that our first attempts to cyclize 1a (Scheme 2) under the classical Mitsunobu conditions using PPh 3 /DIAD or DEAD gave only low yields of β-lactam 3a and dehydropeptide 2, in nearly equimolar ratios. These are difficult to separate in larger quantities. In addition, the application of P(OEt) 3 instead of PPh 3 in the Mitsunobu reaction, which has been successfully applied in the synthesis of nocardicin analogs, resulted in no products whatsoever [13,18]. Our findings were therefore in agreement with the literature reports that have described the influence of the Cα'-H group acidity on the reaction outcome (Scheme 1) [16].
Further optimization of the reaction conditions through changing the solvent and temperature increased the yield of β-lactam 3a to 45%, but the ratio of dehydrodipeptide 2 to 3a remained at 1:1. A modified procedure that would give larger quantities of β-lactam using simple cyclization without the formation of dehydrodipeptide was therefore required. Compounds 1b and 1c with trityl and dibenzyl protection were then treated under the same classical Mitsunobu reaction conditions as 1a. Even though the tritylated compound 1b gave no product whatsoever, a combination of the classical Mitsunobu reaction and dibenzyl protection, gave positive results, whereby dipeptide 1c was transformed into the desired SS diastereoisomer 3c at 76% isolated yield. Epimerization at the exocyclic Cα'-H stereogenic center was also seen, with the diastereoisomeric ratio (dr) of SS to SR of 7:1; these were easy to separate. Importantly, no dehydrodipeptide side products were seen, and the yields were much higher when compared to the results of the Mitsunobu reaction for phthalimide protected dipeptide 1a. Furthermore, the outcome of the reaction gave higher isolated yields when compared to the less reactive phthalimide-protected glycine [16] and cyclopropyl derivatives [17] that have been reported in the literature.
The reaction conditions for both of the reaction steps were then further tested on analogous serylphenylalanine dipeptides 1d, 1e, which both gave cyclized products 3d, 3e at 99% and 90% yields, respectively (Scheme 3). These demonstrated the utility of the dibenzyl protective group in the preparation of simple β-lactam phenylalanine derivatives. 7d, and 7e at 60%, 80%, and 78% isolated yields, respectively (Scheme 3). Many other side products were also observed, among which L-Phe-OCH3 was isolated at 10% yield when 3b was treated under these conditions. Unfortunately, the acetyl protective group on phenol 3f proved to be too labile to enable successful deprotection to 7f, giving instead only a mixture of unidentifiable acetylated products. However, it was easy to selectively remove acetyl protection from the tyrosine OH group prior to debenzylation of 3e. For this purpose, 0.1 M LiOH in methanol was used, to obtain phenol 8 at quantitative yields, with dr of 10:1. This product offered the possibility of future preparation of Oderivatized analogs. This compound was then easily debenzylated to 7d at 85% yield.
Because these compounds contain a primary amine and an ester functional group they have a potential to oligomerize after prolonged storage. We would therefore advise the reader to transform them into some form of a salt if stored for a prolonged period of time.  However, there are also limitations when the acidity of the Cα stereogenic center is higher, in terms of the dr and the optimal solvent. For example, phenylalanine 3c, benzyl 3d, and acetyl-protected tyrosine dipeptide 3f were all prepared at high dr in THF. In contrast, cyclization of the phenylglycine derivative 1e in which a direct link between Cα'-H and electron-withdrawing phenyl ring makes Cα'-H more acidic compared to the phenylalanine analogues 1c, 1d and 1f, proceeded only in DMF, to give 3e at high yields, but with almost equimolar dr of the SS and SR isomer. It appears that the basic nitrogen of the dibenzylamino moiety in combination with DMF as a solvent induces epimerization at the exocyclic Cα'-H when the basic dibenzylamino moiety is applied. This is in agreement with a report by Miller, who observed epimerization to thermodynamic equilibrium when traces of trimethylamine were added to pure phenylglycine derivatives [13].

Materials and Methods
Removal of the dibenzyl protecting group from 3b-3e preceded the next reaction step. Pd-C in a mixture of 5% EtOH in ethyl acetate was preferred to other combinations of catalysts and solvents because it gave significantly higher yields compared to the reactions in EtOH, MeOH or THF. Here, the percentage of EtOH and the traces of DIAD in the reaction mixtures had an enormous influence on the yield and the general outcome of the reactions. In the presence of DIAD as an impurity, higher amounts of EtOH had to be used to obtain full conversion of 3b-3f to yield the corresponding products 7b-7f. However, higher proportions of EtOH also resulted in lower yields, with higher amounts of side products. The purities of all of the compounds 3b-3f were thus of uttermost importance for the successful reaction (Scheme 3).
Compounds 3c-3e were deprotected with 100% conversion to yield debenzylated products 7b, 7d, and 7e at 60%, 80%, and 78% isolated yields, respectively (Scheme 3). Many other side products were also observed, among which L-Phe-OCH 3 was isolated at 10% yield when 3b was treated under these conditions. Unfortunately, the acetyl protective group on phenol 3f proved to be too labile to enable successful deprotection to 7f, giving instead only a mixture of unidentifiable acetylated products. However, it was easy to selectively remove acetyl protection from the tyrosine OH group prior to debenzylation of 3e. For this purpose, 0.1 M LiOH in methanol was used, to obtain phenol 8 at quantitative yields, with dr of 10:1. This product offered the possibility of future preparation of O-derivatized analogs. This compound was then easily debenzylated to 7d at 85% yield.
Because these compounds contain a primary amine and an ester functional group they have a potential to oligomerize after prolonged storage. We would therefore advise the reader to transform them into some form of a salt if stored for a prolonged period of time.

Procedure for Synthesis of Precursor 1c (Scheme 2)
Dibenzyl-L-serine (S6) [27]. Dibenzyl-L-serine (S6) was synthesized according to the modified literature procedure [27]. L-Ser (20.0 g, 190 mmol, 1 eq) and tetrabutylammonium iodide (7.0 g, 19 mmol, 0.1 eq) were dissolved in a solution of KOH (74.7 g, 1.3 mol, 7 eq, 3 M) in a mixture of water and ethanol (1:1, 440 mL). Benzyl chloride (131 mL, 1.14 mmol, 6 eq) was added dropwise and the mixture was stirred under reflux for 1 h. KOH (10.7 g, 190 mmol, 1 eq) was added again and the mixture was stirred further for another hour. After the mixture was cooled down, the white precipitate was dissolved by the addition of water. The solution was extracted with toluene (5 × 100 mL), combined organic fractions were washed with brine (50 mL) and dried over anhydrous Na 2 SO 4 . Organic solvent was evaporated under reduced pressure and the crude product was purified by flash column chromatography. The crude product was dissolved in water (150 mL), the pH value was adjusted to 5 with 1 M HCl and the mixture was left standing over 2 days in a refrigerator to obtain white crystals. Methyl dibenzyl-L-seryl-L-phenylalaninate (1c, Scheme 2). Dipeptide methyl dibenzyl-L-seryl-Lphenylalaninate (1b) was synthesized in a coupling procedure similar to the synthesis of S3 starting from dibenzyl-L-serine (S6) (1.10 g, 3.9 mmol, 1.0 eq). The crude product was purified by flash column chromatography on silica using a gradient of hexane/ethyl acetate starting from 3/1 to 1/1.