Concerted Synthesis of a Spirobicyclic Type-VI β-Turn Mimic of Pro-Pro-Pro-NH₂

Abstract

A highly concerted strategy for the synthesis of symmetrical type-VI β-turn mimics was formulated. A proof of concept is presented in the synthesis of a spirobicyclic peptidomimetic of Pro-Pro-Pro-NH₂; compound 6. The formation of an unusual adduct that was encountered in the process, also is reported. This approach is potentially general for type-VI β-turn mimics where the i+1 and i+2 residues are identical.

In our ongoing investigations into the conformation-bioactivity relationships of peptidomimetics of the allosteric dopamine receptor modulator Pro-Leu-Gly-NH₂, we were faced with the need to constrain Pro-Pro-Pro-NH₂ in a type-VI β-turn conformation. In the past, theconformational mimicry of this cis-amide containing turn has been achieved through the utilization of either the (rac) cis-3-amino-6-carboxypiperidone (1)¹, indolizidinone (2)² or the trans- 3 - amino- 8 -carboxyazocanone (3)³ skeletons. Alternatively, induction of the cis-amide isomer has been achieved through the use of 5-tert-butyl proline (4).⁴

In our case, we deemed the indolizidinone (piperidone)-based mimic of Germanas to be the most appropriate due to its minimal steric deviation from the parent peptide. Incorporation of this type of mimic into the tripeptide Pro-Pro-Pro-NH₂ (5) gives the novel spirobicycle 6 (Figure 1).

Figure 1.

Design of the spirobicyclic type-VI β-turn mimic

We initially envisioned an approach to the spirobicyclic framework of 6, that would employ a but-3-enyl substituted system like 8. Such a system could be prepared from 7 according to Germanas’ methodology (Scheme 1).² Functionalization at the terminal alkene would give alcohol 9, that in turn could be cyclized by an intramolecular Mitsonubu reaction⁵ to potentially afford the desired spirobicycle. This linear approach (approximately 18 steps) plus the fact that it hinged on the sequential creation of two stereocenters seemed daunting. We thus explored a potential concerted approach to 6.

Scheme 1.

Retro-synthetic approach to spirobicycle 6 via Germanas’ methodology

One of the features in 6 that sets it apart from the other known type-VI β-turn mimics is the fact that the i+1 and i+2 amino acids are identical. They are in essence, substituted prolyl residues. This presents an opportunity to disconnect the lactam bond in 10, thereby leading to a symmetrical 2,5-diaminoadipic acid-like derivative 11 that essentially is two prolines linked enantioselectively with a 2-carbon linkage at their α-carbons (Scheme 2). The reported propensity of 2,5-diaminoadipate esters to spontaneously form a monolactam,⁶ and the stability of such monolactams over the diketopiperazine led us to believe that 11 could be a suitable entry into the spirobicyclic framework of 6.

Scheme 2.

Retro-synthetic concerted approach to spirobicycle 6 involving a C2-linked biproline.

For such an approach, the lithium enolate of Seebach’s oxazolidinone (12)⁷ was seen as a suitable system for enantiocontrol due to the complete diastereoselectivity of its alkylation and the cleanliness of reactions that it undergoes. Our initial efforts with vicinal dihalides, both symmetrical and asymmetrical, failed to afford the required dimer. However, we recently reported the stereochemical details of a dehydrodimerization that occurred when 1,2-dibromoethane was used in such a reaction.⁸ We hence concluded that the “onium” character of the dihalo electrophiles, which is a result of the dipole distortion induced by the second halo function, was responsible for their unusual reactivity (1,2-dibromoethane) or their unreactivity (1-bromo-2-chloroethane). This prompted us to turn our attention to other leaving groups; the sulfonate esters being the next obvious choice.

Glycol bis-tosylate was insoluble/unreactive in THF below −30 °C (the stability limit of 12) while glycol bis-mesylate failed to react. However, the bis-triflate 13 reacted with 12 almost instantaneously, giving the highly crystalline dimer 14 in excellent yields ranging from 76–96% over multiple runs (Scheme 3). A cursory attempt at producing the potentially very useful monosubstituted product 15 by adding enolate 12 to a solution containing a two-fold excess of the bis-triflate was unsuccessful. The reaction led only to a modest yield of 14 with no 15 being detected. The reaction did contain a number of side-products, but none corresponded to any of the possible solvolysis or ring-opened products of 15. This suggested that the disturbance in normal dipole moment is relevant even in the highly reactive bis-triflate.

Scheme 3.

Reaction of glycol bis-triflate with 12

In our initial attempts at forming 14 when we were rather unsure about the reactivity of glycol bis-triflate, we used 4 equivalents of 12. In these runs we also isolated a crystalline by-product as a single diastereomer, whose structure was elucidated by NMR to be 16. This structure and the stereochemical assignment was confirmed by X-ray crystallography (Figure 2). The backbone of 16 appears to have been formed through a nucleophilic attack of 12 on an electrophilic iminium species derived from proline and pivalaldehyde (see Scheme A in the Supporting Information for a postulated pathway to the formation of 16).

Figure 2.

X-ray crystal structure of 16

Acid hydrolysis of the dimer 14 afforded 11 in a quantitative yield. Fischer esterification led to the dimethyl ester 17, which could be cyclized (even in its hydrochloride form!) to give predominantly the monolactam 10 either when 17 was heated at reflux in toluene for 8 h or when it was heated neat under aspirator vacuum at about 60 °C for 30 min (Scheme 4).

Scheme 4.

Formation of spirobicycle 10 and tetracyclic diketopiperazine 18.

Exposure of 10 to 1 equiv of NMM for approximately 10 min or exposure of 17 to 2 equiv of NMM for 1–2 hours led to the formation of the tetracyclic diketopiperazine 1 8 . This is in conflict with the observations of Lyssenko et al.⁶ in the case of the simple 2,5-diamino dimethyladipate, wherein a strong base was required to induce formation of a diketopiperazine, while the monolactam formed readily. We attribute this difference to the fact that the basic nitrogen in the monolactam in our case is positioned axial due to the boat-conformation of the 6-membered lactam component of the indolizidine, which in turn is induced by the cisoid fusion with the 5-membered pyrrolidine ring. Acidic methanolysis of 18 gave rise to a mixture of 10 and 17 with 10 being the predominant component.

We next attempted to trap 10 with an activated ester of Boc-proline. As expected the yields of 20 were generally abysmal with weaker coupling reagents like EDC and DCC, with or without DMAP. In these cases, the predominant by-product was 18. Mukaiyama’s reagent, 2-chloro-N-methyl pyridinium iodide (CMPI), only gave yields of 5–10%.

We serendipitously found that when 17 or mixtures of 10 and 17, where 17 was the predominant component, were exposed to these coupling conditions, the yields of 20 increased dramatically. The most reproducible yields(~30%) were obtained when CMPI was used as the coupling reagent along with 2 equiv of Boc-proline (Scheme 5). It appeared from these observations that the acylation reaction with the formation of 19 takes place before the lactamization. True to this possibility, diluting the reaction mixture to 10× its volume and refluxing it for an extra day increased the yields of 20 to a range of 50–62%.

Scheme 5.

Synthesis of 6.

The methyl ester function of 20 resisted direct amidation under a variety of conditions (cyanide catalysis, heating and Weinreb’s amidation). It could however be broken down very easily to the acid with LiOH and then coupled to ammonia using HATU (2-(1H-7-azabenzotriazol-1-yl)-1,1,3,3-tetramethyl uronium hexafluorophosphate methanaminium) as the coupling reagent to obtain the primary carboxamide (21), which was deprotected to afford the target mimic 6.

The ability of the indolizidinone skeleton to mimic the type-VI β-turn has been well-studied by Germanas.² A hydrogen bond between the i carbonyl and the i+4 NH₂ hydrogens is typical of the type-VIa1 subclass. In the case of the Boc-protected precursor 21, the terminal carboxamide hydrogens appear in a typical H-bonded conformation, with the H-bonded hydrogen being downfield at 8.62 and 8.51 ppm (rotameric signals) and the non-H-bonded hydrogen being unusually upfield at 5.56 and 5.51 ppm (rotameric signals). This is mirrored in spirocycle 6 as well, with the hydrogens appearing at 7.65 and 6.31 ppm. The bias of 6 toward the type-VIa1 β-turn subclass is likely to be enhanced by the natural φ dihedral angle of −60° of the i+1 prolyl residue, the ideal value of type-VIa1 β-turn.

In summary, we have successfully explored a more convergent strategy towards the required triproline type-VI β-turn mimic. A salient feature of our route is that the core mimic can be synthesized in only 7 steps instead of 18 steps as would be required in the linear approach. Our approach is independent of the identity of the i+2 residue since formation of the i+1 α-center is not dependent on the cisoid fusion of the indolizidinone skeleton. It is however, at this time, limited to cases where the i+2 and i+1 residues are the same. Further investigations aimed at exploring the generality of this approach and extension to cases where i+1 ≠ i+2 are in progress.

Supplementary Material

si20060106_023. Supporting Information Available.

Experimental procedures and spectroscopic data for 6, 11, 14, 16, 18, 20, and 21, X-ray crystallographic data for 16, in CIF format, and a scheme depicting a postulated pathway for the formation of 16. This material is available free of charge via the Internet at http://pubs.acs.org.