Synthesis and conformational analysis of locked carbocyclic analogues of 1,3-diazepinone riboside, a high-affinity cytidine deaminase inhibitor.

Cytidine deaminase (CDA) catalyzes the deamination of cytidine via a hydrated transition-state intermediate that results from the nucleophilic attack of zinc-bound water at the active site. Nucleoside analogues where the leaving NH(3) group is replaced by a proton and prevent conversion of the transition state to product are very potent inhibitors of the enzyme. However, stable carbocyclic versions of these analogues are less effective as the role of the ribose in facilitating formation of hydrated species is abolished. The discovery that a 1,3-diazepinone riboside (4) operated as a tight-binding inhibitor of CDA independent of hydration provided the opportunity to study novel inhibitors built as conformationally locked, carbocyclic 1,3-diazepinone nucleosides to determine the enzyme's conformational preference for a specific form of sugar pucker. This work describes the synthesis of two target bicyclo[3.1.0]hexane nucleosides, locked as north (5) and south (6) conformers, as well as a flexible analogue (7) built with a cyclopentane ring. The seven-membered 1,3-diazepinone ring in all the three targets was built from the corresponding benzoyl-protected carbocyclic bis-allyl ureas by ring-closing metathesis. The results demonstrate CDA's binding preference for a south sugar pucker in agreement with the high-resolution crystal structures of other CDA inhibitors bound at the active site.

Introduction

Cytidine deaminase (CDA; EC 3.5.4.5) is a zinc-dependent enzyme that catalyzes the deamination of cytidine to uridine via the formation of an unstable, hydrated transition-state (ts) intermediate (Figure 1).(1) The spontaneous deamination of cytidine is very slow and proceeds with a rate constant around 10−10 s−1; however, the enzyme is able to accelerate the rate of deamination by an impressive 12−14 orders of magnitude.(2) This tremendous enhancement reflects an extraordinary affinity for the hydrated intermediate (Kts ∼10−16 M).(3a) Formation of such a hydrated intermediate has been confirmed biochemically[3b,3c] and by the crystal structure of the inhibitor zebularine (1) bound to CDA, which revealed the formation of the 3,4-double-bond hydrate (2) at the active site.(3d) Inhibition of CDA is of particular interest therapeutically to enhance the activity of readily deaminated antitumor nucleosides such as cytosine arabinoside and 5-aza-2′-deoxycytidine.(4)

Figure 1

Transition state for the hydrolytic deamination of cytidine.

Another well-known inhibitor of CDA is tetrahydrouridine (3, THU), which despite lacking the 5,6-double bond of the hydrated zebularine ring possesses a hemiaminal structure that embodies the properties of a powerful transition-state mimic. Indeed, a recent, high-resolution crystal structure of mouse CDA complexed with THU shows the active diastereoisomer with the 4-OH group spatially oriented toward the zinc ion.(5)

An intriguing molecule, also displaying a strong inhibition of CDA (Ki = 25 nM), but for which the mechanism of action remained unexplained for many years, is the diazepinone ribose 4.(6) This compound possesses a critical double bond, but as opposed to the hydrated 3,4-double of zebularine and the hemiaminal functionality of THU, it lacks any possibility of interacting with CDA via zinc coordination. Fourteen years later, after solving the crystal structure of human CDA bound to the diazepinone ribose 4, Verdine et al. demonstrated that the mode of action of this tight-binding inhibitor was indeed wholly independent from a hydration mechanism and showed instead that the strong binding at the active site resulted from the edge-on π−π interaction between the diazepinone ring’s double bond and Phe137.(7)

For these three molecules, zebularine (1), THU (3), and diazepinone riboside (4), it is clear that the interactions of the nucleobase with the active site are critical for strong binding, and not surprisingly, the majority of the structure−activity-relationship (SAR) studies reported to date for CDA deal almost exclusively with changes in the heterocyclic base. In terms of the role of the sugar moiety, however, fewer studies have been conducted. Even though the sugar ring has a small effect on the rate of the spontaneous deamination reaction of cytidine, it greatly enhances the reactivity of the enzymatic reaction by increasing the effective concentration of the altered substrate at the active site.(8) Such an entropic advantage is the direct result of the parts (base and ribose) being properly connected to fit the active site.

The critical parameter of the ribose (or 2′-deoxyribose) that ensures that the interactions between the parts are optimal is the sugar pucker, which is defined by the pseudorotational phase angle P.(9) Normally, conventional nucleosides in solution undergo rapid equilibration between ranges of P defined by two main twist furanose puckering domains centered around a 3T2 (north-type) and a 2T3 (south-type) conformation. Despite the small difference in energy between these two conformations, an emerging picture from recent observations is that the majority of nucleoside(tide) target enzymes, whether anabolic or catabolic, appear to have strict conformational requirements for substrate binding, accepting the furanose ring only in a particular, well-define shape.(10) Because the value of P and the corresponding glycosyl torsion angle χ are interdependent, with the latter responding in a concerted manner to minimize steric clashes, the value of χ is also of critical importance for the two connecting pieces to provide optimal recognition.

In the specific case of CDA, the crystal structures of the enzyme bound to zebularine hydrate (2), THU (3), and diazepinone riboside (4) show the conformation of the sugar ring in the south hemisphere, close to a 2′-endo conformation. In particular, the 1.48 Å resolution structure by Kumakasa et al.(5) shows the ribose ring of THU clearly in a south conformation with the 3′-OH hydrogen bonded to the side chains of highly conserved Asn54 and Glu56. Using our interactive tool, PROSIT,(11) which calculates the pseudorotational parameters of nucleosides and nucleotides bound to proteins, we calculated a value of P = 158.55° (2′-endo) for THU which is clearly in the south hemisphere. Other parameters calculated included the maximum puckering amplitude (νmax = 36.68°) and the glycosyl torsion angle of the tetrahydrouracil ring, which is in the anti range (χ = −136.51°). Similarly, the conformational parameters for the bound diazepinone riboside (4)(7) calculated by PROSIT matched very closely those of THU (P = 158.18°, νmax = 32.10°, and χ = −149.66°).

The unique preference of CDA for the south (2′-endo) conformation stands in sharp contrast with the related purine deaminase, adenosine deaminase (ADA), which prefers to bind both substrate and inhibitors in the antipodal north (3′-endo) conformation.(12) Despite the similarity of the mechanisms of deamination, which in both enzymes proceed via analogous tetrahedral intermediates, the secondary and tertiary structural motifs of the two enzymes are completely unrelated and lack any evolutionary homology.(13)

With the aim of studying the impact that the sugar conformation has on CDA, and how it might impinge on the other critical interactions with important residues at the active site that are independent of zinc binding, and thus not controlled by the chemical generation of an unstable, chiral hydrated intermediate, we chose the diazepinone riboside 4 as a prototype. This diazepinone riboside preserves all the other critical interactions with conserved amino acids, such as Glu56 and Asn64, plus the CONH portion of the diazepinone ring forms a similar set of hydrogen bonds with Ala66 and Glu67 at the active site that are seen with the nucleobases of the hydrate of zebularine and with THU. The only difference is the pronounced wrinkle of the seven-membered ring in 4, which juts the double bond out of plane and allows it to interact with the π-face of Phe137 via a canonical π/π-interaction.(7)

In the present work, we report on the construction of conformationally locked north- (5) and south-diazepinone (6) riboside analogues of 4 built on a stable carbocyclic scaffold: the bicyclo[3.1.0]hexane template.(14) Because CDA accepts and deaminates efficiently both ribo and 2′-deoxyribonucleosides, we chose for simplicity to prepare 2′-deoxy versions of these carbocyclic nucleosides. For comparison, the plain and flexible cyclopentane analogue (7) was also synthesized; all of the compounds were evaluated as CDA inhibitors and compared with the parent riboside (4).

Results

Synthesis

To accomplish the syntheses of these compounds, we sought a linear strategy that could be adapted to all three targets. This approach differs from the convergent approach used for the synthesis of 4 in order to avoid problems with the likely formation N- and O-alkylated products.(6) The retrosynthetic analysis used in the current approach shows how the 1,3-diazepin-2-one moiety is to be engendered by a ring-closing metathesis (RCM) reaction on the bis-allyl ureas (I) (Figure 2). These in turn could be accessed by the reaction of allyl isocyanate with the monoalkylated amines (II) derived from the appropriate carbocyclic primary amines (III).

Figure 2

Retrosynthetic analysis of the target compounds 5−7 (PG = protecting group).

Retrosynthetic analysis of the target compounds 5−7 (n class="Chemical">PG = protecting group).

The synthesis of the locked north-diazepinone nucleoside started with the known bicyclo[3.10]hexanol 8.(15) As shown in Scheme 1, conversion of 8 to its methanosulfonyl ester 9 was followed by nucleophilic attack with sodium azide to give 10 with inversion of stereochemistry. Alternatively, alcohol 8 was reacted with diphenylphosphoryl azide (DPPA) in DMF to give directly the desired azide 10.(16) After hydrogenolysis over Lindlar’s catalyst, the azide was quantitatively reduced to the key intermediate amine 11, and alkylation of the primary amine with allyl bromide in DMF, with either K2CO3 or Cs2CO3 as catalysts, gave the monoalkylated amine 12 in 35 and 33% yield, respectively, showing no clear advantage with the use of Cs2CO3.(17)

Scheme 1

A change in strategy involving the reduction of the azide 11 in the presence of di-tert-butyl dicarbonate (Scheme 2) led to the Boc-protected amine 14,(18) which was selectively alkylated with allyl bromide and potassium hexamethyldisilazide in DMF to give 15. Removal of the Boc group with TFA led to the monoallylated amine 12, which reacted with allylisocyanate to form the desired diallylurea 16. Unfortunately, all attempts to form the seven-membered cyclic urea 17 via RCM(19) with Grubb’s second-generation catalysts at room temperature or reflux in either CH2Cl2 or toluene failed and only starting material was recovered.

Scheme 2

We surmised that the failure of the metathesis reaction was due to the unfavorable orientation of the allyl groups in 16, which prevented the intramolecular condensation.(19c) At 6 mM substrate concentration no cross-metathesis products were observed either. After considering the equilibrium between the cis- and trans-configurations of the amide, we realized that the expected, more stable trans-configuration would indeed place the allyl groups away from each other making ring closure impossible (Figure 3).

Figure 3

Equilibrium between the cis- and trans-conformations of the amide bond in 16.

Since Kaválec et al. have demonstrated that the cis/trans-isomerization barrier can be significantly lowered by alkylation of the amide,(20) we considered that such equilibrium could be tunable by using a suitably protected amide group. As a first attempt, the amide moiety was silylated with hexamethyldisilazane and ammonium sulfate as catalyst to give the aza-enol ether 18 in quantitative yield, where the bulky trimethylsily group was expected to favor a desirable conformational change (Scheme 3). Because of the moisture sensitivity expected for this class of silylated compounds, the RCM reaction was performed immediately with Grubbs second-generation catalyst under reflux in methylene chloride to give, according to TLC analysis, a single product after 30 min. Unfortunately, the expected seven-membered ring was not formed and instead the six-membered urea 20 was obtained in 75% yield. Although the formation of 20 was disappointing, a result that most likely happened due to the isomerization of the double bond to form the s-cis conjugated π-system,(21) the isolation of a cyclic product proved that a cis-geometry was essential for the success of the RCM reaction.

Scheme 3

Encouraged by these results, we then investigated the effect of a simple acyl group for the RCM reaction, which would allow for the easy removal of all protecting groups in one step. Therefore, urea 16 was acylated with benzoyl chloride in pyridine to form the UV-active urea 21 in 89% yield. Interestingly, the introduction of the benzoyl group led to significant broadening of some signals in the proton and carbon NMR spectra. The RCM reaction, performed under the same conditions as with 18, smoothly afforded the desired seven-membered ring in 92%, thus confirming the role of the stereochemistry of the urea in the intramolecular cyclization. Finally, after global deprotection of the acyl groups, achieved with 1% NaOH solution in MeOH, the target structure 5 was obtained in excellent yield (Scheme 4).

Scheme 4

The optimized reaction conditions for the synthesis of the conformationally locked north-diazepinone 5 were applied to the syntheses of both south and flexible targets (6 and 7), respectively (Scheme 5). Starting from enantiomerically pure 23(22) for the flexible analogue, mesylation and subsequent nucleophilic substitution with azide afforded compound 24 in 78% yield (two steps). Reduction of the azido group with Lindlar’s catalyst(23) in the presence of Boc-anhydride gave the Boc-protected amine 26b. Similarly, reaction of the enantiomerically pure amine 25(24) with Boc-anhydride gave the protected amine 26a. From this point onward, the two approaches converge and the reactions and yields obtained for the penultimate intermediates 32a and 32b were essentially identical to those obtained before. The final removal of the benzyl protecting groups proceeded uneventfully in the presence of BCl3 to give the target compounds 6 and 7.

Scheme 5

Kinetic Analysis of Carbocyclic Diazepinones as Inhibitors of Cytidine Deaminase

Cytidine deaminase activity was measured spectrophotometrically by following the loss of absorbance of cytidine at 282 nm (Δε = −3600) in 0.1 M phosphate buffer (pH 6.8) at 25 °C. Concentrations of stock solutions of the three carbocyclic diazepinones were determined by proton NMR (see the Figures 1−3) in reference to a known concentration of pyrazine (10−3 M) added as an integration standard. Initial rates for the deamination of cytidine in the presence of all three inhibitors were determined and plotted as a function of [S] vs [S]/vi (Hanes plot, Figures 4 and 5). Inhibition constants (Ki) for the three inhibitors (Table 1) were calculated using the following equationwhere KMapp is the apparent value of the Michaelis constant (KM) when measured in the presence of an inhibitor. Values for KMapp at various concentrations of inhibitor were estimated from the Hanes plots (Figures 4 and 5), which have −KM as the intercept on the [S] axis.

Figure 4

Hanes plot for cytidine deaminase activity at various concentrations (defined in the figure) of south/north carbocyclic diazepinone inhibitors. The parallel lines represent linear fits to the data and are indicative of competitive inhibition: (A) south-diazepinone 6 (●); (B) north-diazepinone 5 (◼).

Figure 5

Hanes plot for cytidine deaminase activity at various concentrations (defined in figure) of the flexible diazepinone inhibitor 7 (▲). The parallel lines represent linear fits to the data and are indicative of competitive inhibition.

Table 1

Inhibition Constants for Carbocyclic Diazepinones

compd	Ki (μM)
north-diazepinone (5)	70 ± 10
south-diazepinone (6)	40 ± 10
flexible-diazepinone (7)	0.4 ± 0.1

Figure 4 shows that both south (6) and north (5) carbocyclic diazepinones are competitive inhibitors of cytidine deaminase (parallel lines) and that the south derivative 6 is more potent than its north counterpart 5 (by a factor of ∼2). Figure 5 shows that the flexible, carbocyclic diazepinone is also a competitive inhibitor, which surprisingly, as indicated in Table 1, is significantly more potent than either one of the locked derivatives. This compound, however, is still 16-fold less potent than the corresponding riboside 4.(6)

Molecular Modeling

Molecular dynamics calculations and docking of the new carbocyclic diazepinone nucleosides (5−7) at the active site of CDA were based on the available crystal structure of human CDA bound to the diazepinone riboside 4 (PDB ID: 1MQ0).(7) The structures of the locked north- and south-diazepinone nucleosides (5 and 6) were constructed by importing the bicyclo[3.1.0]hexane templates from the crystal structures of north- (LETJEZ) and south- (YESMIS) bicyclo[3.1.0]hexane thymidine nucleosides which were downloaded from the Cambrige structural database (CCDC ConQuest version 1.5). The diazepinone ring as it appeared in the crystal structure of the CDA complexed with the diazepinone riboside 4 was mounted on the two bicyclo[3.1.0]hexane templates (see Figure 1 of the ).

The diazepinone riboside 4 (Figure 6A) forms six n class="Chemical">hydrogen bonds (yellow dashed lines). Two hydrogen-bonding interactions are between the diazepinone ring and Glu67 (subunit C) and Ala66 (subunit C). The ribose ring forms the rest of the hydrogen bonds with Glu56 (subunit C), Asn54 (subunit C), Ala58 (subunit D), and Tyr60 (subunit D). In addition, there is the important canonical π/π interaction with the π face of Phe137 (subunit A). The worst ligand, the north-diazepinone 5 (Figure 6B), maintains the hydrogen-bonding interactions of the diazepinone ring to Glu67 and Ala66, but the north-pseudosugar moiety forces the equatorially disposed 3′-OH to form a weaker hydrogen bond to Glu56. The 5′-OH cannot bind to Ala 58 or Tyr60 and reaches to Leu142 to form a new weak hydrogen bond. In addition, the interaction with Asn 54 is missing. The flexible diazepinone 7 (Figure 6C) and the south-diazepinone 6 (Figure 6D) form almost identical interactions. However, the hydrogen-bonding interactions of the 3′-OH in the flexible compound 7 (1.682 and 1.795 Å) appear to be stronger than those in the locked south-diazepinone 6 (1.751 and 1.930 Å).

Figure 6

Monte Carlo/Stochastic Dynamics simulations in MacroModel 9.5 of four diazepinone compounds in human cytidine deaminase: (A) diazepinone riboside (4, conformation in the crystal structure); (B) north-diazepinone (5); (C) flexible diazepinone (7); (D) south-diazepinone (6). Hydrogen bonds appear as yellow dashed lines, and the canonical π/π interaction with the π-face of Phe137 is represented by a group of purple dotted lines. As a reference point, the atom of zinc appears bonded to Cys65.

The major discrepancies in the hydrogen-bonding network between these compounds (4−7) can be better understood by viewing the superimposed structures and observing the orientation of the critical OH groups (Figure 7). With the exception of the equatorial 3′-OH in north-diazepinone 5 (yellow), the axial orientations of the 3′-OH in the rest of the molecules remains in close proximity. There is good freedom of rotation for the CH2OH group in 5, which allows it to reach a new amino acid, Leu142, to compensate for the lack of interaction with Ala58 and Tyr60 seen with the rest of the compounds. The diazepinone ring appears perfectly superimposed in all four structures. The 2′-OH is present only in the parent riboside 4 but this group appears not to be engaged in any important interaction, thus supporting the equivalent affinity that CDA displays for both ribo and 2′-deoxyribonucleosides (substrates or inhibitors).

Figure 7

Overlapping of CDA-bound conformation of 4 (khaki), 5 (yellow), 6 (cyan), and 7 (green).

Discussion

The selection of the diazepinone riboside 4 as a prototype to study the preferred conformation of CDA for a particular sugar pucker was based on the premise that the newly revealed edge-on π−π interaction between the diazepinone ring’s double bond and Phe137(7) provided a mechanism of inhibition independent of the zinc-promoted hydration for which the role of the sugar O4′ oxygen would not be considered critical. In the(25) case of the mechanistically related adenosine deaminase (ADA)(26) and also for CDA,(27) studies with carbocyclic nucleosides in our hands have revealed that the sugar moiety plays a significant role in facilitating zinc-promoted hydration due to the more electronegative nature of the sugar moiety and important orbital interactions between the oxygen’s (O4′) lone pair with the highest p orbital component and the adjacent C−N glycosyl bond through the anomeric effect. Thus, replacement of the ribose or 2′-deoxyribose ring of adenosine and cytidine with carbocyclic pseudosugars removes any communication between the sugar and the nucleobase and the rates of deamination decrease.(27) The diazepinone-based inhibitors (4−7), which lack any possibility of interacting with CDA via zinc coordination and work independently of hydration, are therefore ideal candidates to study the effect of sugar (or pseudosugar) ring puckering on the mechanism of CDA.

Because all the crystal structures of bound inhibitors to CDA have shown the conformation of the sugar ring to be in the south hemisphere,[3c,3d,5,7] close to a 2′-endo conformation, we hypothesized that the conformationally south-locked diazepinone nucleoside 6 would be the most potent inhibitor in comparison with the antipodal north-locked 5, and the flexible diazepinone nucleoside 7. As mentioned before, the preference of CDA for the south (2′-endo) conformation is in sharp contrast with that of ADA, which prefers to bind both substrate and inhibitors in the antipodal north (3′-endo) conformation.(12) This unique preference for an antipodal sugar puckering by CDA is another important element of differentiation from the seemingly similar ADA.

The biochemical results from this investigation indeed confirm CDA’s preference for a south sugar conformation since the south inhibitor 6 was more potent than its north counterpart 5 albeit only by a factor of ∼2 (Table 1). Surprisingly, the flexible cyclopentane diazepinone nucleoside 7 was more potent than the south-diazepinone 6 by a factor of 100 (Table 1). However, relative to the normal riboside (4), the flexible diazepinone 7 was only 16-fold less potent. A likely explanation for the less than expected performance of the south-diazepinone nucleoside 6 stems from the fusion of the cyclopropane ring adjacent to the glycosyl C−N bond. As the crystal structures of CDA in complex with THU and diazepinone riboside 4 show, the glycosyl torsion angles (χ = −136.51° and χ = −149.66°) are clearly in the anti range. Unfortunately, as we have observed with all pyrimidine nucleosides built on a south-bicyclo[3.1.0]hexane pseudosugar template, both single-crystal X-ray structures and ab initio calculations have shown that the glycosyl torsion angle is more stable in the syn region.(28) This means that the enzyme has to pay a significant penalty to accommodate the inhibitor in the less stable anti conformation. Despite this shortcoming, the south-diazepinone 6 was still able to inhibit CDA more effectively than its north counterpart. However, the penalty incurred in changing the conformation from syn to anti inside a binding pocket that is completely sequestered from bulk solvent diminishes the capacity of CDA to discriminate effectively between the two inhibitors on the basis of ring pucker.

The same argument can explain the 100-fold difference between the flexible diazepinone 7 and the south-diazepinone 6, which as seen in parts C and D of Figure 6 fit almost identically. The flexible diazepinone does not have a constrained syn glycosyl torsion angle since the plastic cyclopentane ring can easily reach the P value observed for the ribose ring in 4 (P = 158.18°) at the active site of CDA. If we take as a reference the approximate 1′-exo (P = 118.6°) conformation found in the crystal structure of plain carbocyclic thymidine,(29) the cyclopentane ring in 7 has to change its puckering 40° (pseudorotational degrees) to reach the 2′-endo conformation. This change is very much achievable in solution and with a much lower energy penalty when compared to that required to rotate the glycosyl bond in 6. This small penalty in changing ring puckering could very well reflect the 16-fold lower potent inhibition of CDA by 7 relative to the parent ribose 4.

To strengthen our argument that the effective north/south discrimination by CDA was weakened by the syn-preference of the south-locked diazepinone 6, we performed a conformational analysis by sampling the entire conformational space of χ from 0° to 360° (Figure 8) as we have done previously for the locked north- and south-thymidine analogues.(28) For the locked south-thymidine analogue, which has the same bicyclo[3.1.0]hexane scaffold of 6, a relative energy barrier of ∼17 kcal/mol separated the most stable syn conformer (χ = ∼60°) from the less stable anti conformer (χ = ∼−120°). The energy barrier for the larger, bulkier diazepinone ring was somewhat higher (∼18 kcal/mol kcal/mol), reflecting a greater difficulty in achieving the anti conformation demanded by the enzyme. Furthermore, the anti conformer for 6 was significantly less stable than the corresponding south-locked thymidine anti conformer. As in the case of the south-locked thymidine analogue, the integrity of the bicyclo[3.1.0]hexane system remained constant during the calculations as indicated by the small variance of P across the entire range (not shown).

Figure 8

Ab initio potential energy scan (PES) of south-locked diazepinone 6 and south-locked thymidine (six-membered ring).

Ab initio potential energy scan (PES) of south-locked diazepinone 6 and south-locked n class="Chemical">thymidine (six-membered ring).

In conclusion, the observed order of inhibitory potency 4 > 7 ≫ 6 > 5 can be easily explained in terms of two important factors: (1) the preferred sugar puckering of CDA for the south conformation and (2) the preferred anti disposition of the base bound at the active site. The small ∼2-fold difference between south-locked 6 and north-locked 5 suggests that the most efficient network of hydrogen bonds provided by the south template allows CDA to overcome the penalty incurred in rotating χ from syn to anti and differentiate between the two. However, this penalty becomes an important issue when comparing the south-locked 6 with the flexible-diazepinone 7, for which such a penalty no longer exists, allowing the ring to reach the required 2′-endo pucker with relative ease from its preferred 1′-exo conformation. These conformational issues could explain the 100-fold separation in potency observed between these two compounds. Finally, the diazepinone-riboside 4 also does not incur the penalty of rotating χ, and by its very nature, the ribose ring has a very small energy barrier separating the 3′-endo from the precise 2′-endo conformation required for binding. We can therefore conclude that in accordance to the structural information gathered from X-ray crystallography our set of inhibitors prove that CDA indeed has a penchant for the south conformation, which is opposite to that of ADA. One final point of interest in cancer chemotherapy regarding the combined use of antitumor nucleosides that are readily deaminated by CDA is that the very stable and quite potent carbocyclic diazepinone nucleoside (7) could overcome many of the instability problems associated with the use of tetrahydrouridine (THU, 3).(30)

Experimental Section

Molecular dynamics calculations were performed using the Monte Carlo/Stochastic Dynamics (MC/SD) program available in MacroModel 9.5(25) to simulate molecular movement over time using Newton’s equation of motion. This program performs constant temperature calculations that take advantage of the strengths of the Monte Carlo method for quickly introducing large changes in a few degrees of freedom and stochastic dynamic for an effective local sampling of collective motions. For these calculations, the OPLS_2005 force field was employed and water was as used as a solvent. To force the system to meet the desired geometrical conditions and to reduce the cost of computer time by eliminating calculations expected to have little influence on the results, all residues that do not have any atoms within 6 Å of the ligands and the seven atoms of the diazepinone ring were fixed. The maximum iterations and convergence threshold were set to 5000 and 0.00, respectively. The number of structures to sample was set to 10, and the simulation time was 100 ps.

Ab Initio Potential Energy Scans (PES)

The structure of the south-locked thymidine analogue (entry code: YESMIS(31)) was taken from the Cambridge Structural Database.[32,33] Since YESMIS is a dimer containing two conformers, one of them was taken for further ab initio calculations. The structure of compound 6 was built from two segments, the sugar ring extracted from YESMIS and the diazepinone ring extracted from the PDB file 1MQ0.(7) The structure of 6 was first minimized, and then a conformational search was performed employing the Systematic Search method in MOE 2007.09.(34) The identified global energy minimum conformation was then used for ab initio calculations. Both the south-locked thymidine and 6 were preoptimized at the HF/3-21G level of theory and then reoptimized at the HF/6-311G level in the program Gaussian 03, revision E01 (Gaussian, Inc.). The two structures, which were assumed to be local or global energy minima, were investigated by performing a PES of the dihedral angle χ by rotating the base in steps of 15°. Using the Gaussian HF method at the same 6-311G level of theory, each of the geometries obtained were optimized for all internal coordinates, except the dihedral angle χ.

Cytidine Deaminase Inhibition

Cytidine deaminase from E. coli was purified as previously described.(3a) Enzymatic activity (∼1 nM CDA) was measured spectrophotometrically in 1 mL cuvettes by following the loss of absorbance of cytidine at 282 nm (Δε = −3600) in 0.1 M phosphate buffer (pH 6.8) at 25 °C. Concentrations of stock solutions of the three carbocyclic diazepinones (5−7) were determined by proton NMR (see Figures 2−4 in the ) in reference to a known concentration of pyrazine (10−3 M) added as an integration standard.

((1R,2S,4S,5S)-2-Acetoxy-4-azidobicyclo[3.1.0]hexan-1-yl)methyl Acetate (10) (Method A)

To a stirred solution of alcohol 8 (850 mg, 3.72 mmol) in dry DMF (25 mL) were added diphenylphosphoryl azide (DPPA, 2.40 mL, 11.2 mmol) and triethylamine (1.80 mL, 13.0 mmol) under argon at room temperature. The mixture was warmed to 60 °C and stirred for 20 h. The solvent was evaporated, and the residue was partitioned between water (50 mL) and CH2Cl2 (50 mL). After phase separation, the aqueous phase was extracted again with CH2Cl2 (2 × 50 mL), and the combined extracts were dried (Na2SO4) and concentrated. The crude material was purified by flash chromatography on silica gel (EtOAc in hexanes 25−45%) to yield the azide 10 (706 mg, 75%) as a colorless oil: [α]20D −57.71 (c 0.78, CHCl3); IR (neat) 3014, 2098, 1733, 1631, 1370, 1230, 1025, 973, 905, 835 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.49 (app t, 1H, J ≈ 8.8 Hz, H-2), 4.22 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.97 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.81 (d, 1H, J = 6.2 Hz, H-4), 2.25 (ddd, 1H, J = 14.8, 8.1, 1.0 Hz, H-3a), 2.09 (s, 3H, OAc), 1.99 (s, 3H, OAc), 1.53 (dd, 1H, J = 8.6, 4.3 Hz, H-5), 1.41 (ddd, 1H, J = 14.8, 8.4, 6.2 Hz, H-3b), 0.84 (dd, 1H, J = 6.2, 4.3 Hz, H-6a), 0.79−0.75 (m, 1H, H-6b); 13C NMR (100 MHz, CDCl3) δ 171.0, 70.8 (2 × C=O, Ac), 74.7 (C-2), 65.1 (CH2-OAc), 61.5 (C-4), 34.5 (C-3), 30.8 (C-1), 26.5 (C-5), 20.9, 20.8 (2 × CH3, Ac), 11.1 (C-6); FAB-MS m/z (relative intensity) 254 (45), 194 (100). Anal. Calcd for C11H15N3O4: C, 52.17; H, 5.97; N, 16.59. Found: C, 52.26; H, 6.01; N, 16.35.

((1R,2S,4S,5S)-2-Acetoxy-4-azidobicyclo[3.1.0]hexan-1-yl)methyl Acetate (10) (Method B)

The bicyclic hexanol 8 (600 mg, 2.36 mmol) was dissolved in dry CH2Cl2 (20 mL) under argon at 0 °C, and triethylamine (1.10 mL, 7.89 mmol) and methansulfonyl chloride (MsCl, 305 μL, 3.94 mmol) were slowly added. The reaction mixture was stirred for 1 h at 0 °C and then poured into a mixture of ice-cold phosphate buffer (pH, 7.2, 150 mL) and Et2O (150 mL). The aqueous phase was extracted with Et2O (2 × 100 mL), and the combined extracts were dried (MgSO4) and concentrated. The crude mesylate 9 was used in the next substitution reaction without any further purification (decomposition occurred during chromatography on silica gel): IR (neat) 3019, 2941, 1733, 1446, 1351, 1237, 1171, 1032, 970, 922, 846 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.20 − 5.11 (m, 2H, H-2, H-4), 4.20 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.78 (d, 1H, J = 12.0 Hz, CHH-OAc), 2.94 (s, 3H, CH-SO3), 2.63 (ddd, 1H, J = 14.1, 7.8, 7.8 Hz, H-3a), 1.99 (s, 3H, OAc), 1.98 (s, 3H, OAc), 1.83 − 1.78 (m, 1H, H-3b), 1.25 (dd, 1H, J = 6.0, 4.3 Hz, H-6a), 0.81 (app t, 1H, J ≈ 7.0 Hz, H-6b); FAB-MS m/z (relative intensity) 307 (5), 247 (24), 211 (100).

The mesylate 9 (810 mg, 2.36 mmol) was dissolved in dry DMF (20 mL), and n class="Chemical">sodium azide (1.70 g, 23.6 mmol) was added in one portion under argon at room temperature. The mixture was heated to 90 °C for 1 h and allowed to cool to room temperature. The reaction was poured into a mixture of phosphate buffer (pH 7.2, 100 mL) and Et2O (100 mL), and the aqueous phase was extracted with Et2O (2 × 100 mL). The combined organic extracts were dried (MgSO4) and concentrated. Flash chromatography on silica gel (EtOAc in hexanes 25−45%) yielded the azide 10 (436 mg, 73%) as a colorless oil. The spectral data were identical to those reported above.

((1R,2S,4S,5S)-2-Acetoxy-4-aminobicyclo[3.1.0]hexan-1-yl)methyl Acetate (11)

Azide 10 (690 mg, 2.72 mmol) was dissolved in MeOH (25 mL) and CH2Cl2 (25 mL) under argon at room temperature, and Lindlar’s catalyst (100 mg) was added. The mixture was flushed with hydrogen and stirred at room temperature overnight. The mixture was filtered over a short pad of Celite, and the solvent was evaporated off. The crude amine 11 (618 mg, 100%) was obtained as a colorless wax, which was sufficiently pure to be used in the next step without any further purification. An analytical sample was prepared by flash chromatography on silica gel (MeOH in EtOAc 10−40%): [α]20D −19.21 (c 0.36, CHCl3); IR (neat) 3332, 2947, 1729, 1654, 1548, 1370, 1238, 1025, 733 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.45 (app t, 1H, J ≈ 8.3 Hz, H-2), 4.17 (d, 1H, J = 11.6 Hz, CHH-OAc), 3.92 (d, 1H, J = 11.6 Hz, CHH-OAc), 3.15 (d, 1H, J = 6.0 Hz, H-4), 2.40−2.10 (bs, 2H, NH2), 1.94 (s, 3H, OAc), 1.92 (s, 3H, OAc), 1.74 (dd, 1H, J = 13.5, 8.0 Hz, H-3a), 1.30−1.22 (m, 2H, H-3b, H-5), 0.72 (app t, 1H, J ≈ 4.8 Hz, H-6a), 0.62 (dd, 1H, J = 8.2, 5.5 Hz, H-6b); 13C NMR (100 MHz, CDCl3) δ 171.5, 171.4 (2 × C=O, Ac), 75.9 (C-2), 66.4 (CH2-OAc), 51.2 (C-4), 36.2 (C-3), 30.2 (C-1), 29.5 (C-5), 19.5, 19.4 (2 × CH3, Ac), 10.4 (C-6); FAB-MS m/z (relative intensity) 228 (100), 186 (38). Anal. Calcd for C11H17NO4·0.2H2O: C, 57.23; H, 7.60; N, 6.07. Found: C, 57.08; H, 7.59; N, 5.86.

((1R,2S,4S,5S)-2-Acetoxy-4-allylaminobicyclo[3.1.0]hexan-1-yl)methyl Acetate (12) and ((1R,2S,4S,5S)-2-Acetoxy-4-diallylaminobicyclo[3.1.0]hexan-1-yl)methyl Acetate (13)

The amine 11 (100 mg, 0.440 mmol) was dissolved in n class="Disease">dry DMF (2.0 mL) at room temperature, and potassium carbonate (anhydrous, 166 mg, 1.20 mmol) was added in one portion under argon. Allyl bromide was slowly added at room temperature and the mixture was stirred overnight. The solvent was evaporated off, and the residue was chromatographed on silica gel (MeOH in EtOAc 0−10%) to yield the title compound 12 (41.0 mg, 35%) as a colorless oil, together with the dialkylated derivative 13 (33.8 mg, 25%).

Compound 12: [α]20D 3.54 (c 0.58, CHCl3); IR (neat) 3323, 2946, 1731, 1643, 1457, 1370, 1238, 1111, 1024, 916, 759 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.84 (m, 1H, CH=CH2), 5.49 (app t, 1H, J ≈ 8.3 Hz, H-2), 5.13 (ddd, 1H, J = 17.1, 3.0, 1.6 Hz, CH=CHH), 5.04 (ddd, 1H, J = 10.2, 3.0, 1.6 Hz, CH=CHH), 4.18 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.99 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.25−3.15 (m, 2H, CH2-CH=CH2), 3.14 (d, 1H, J = 6.2 Hz, H-4), 2.09 (dd, 1H, J = 14.3, 8.3 Hz, H-3a), 2.00 (s, 3H, OAc); 1.99 (s, 3H, OAc), 1.64−1.59 (bs, 1H, NH), 1.40 (dd, 1H, J = 8.4, 4.4 Hz, H-5), 1.26 (m, 1H, H-3b), 0.82 (dd, 1H, J = 5.6, 4.4 Hz, H-6a), 0.68 (dd, 1H, J = 8.4, 5.6 Hz, H-6b); 13C NMR (100 MHz, CDCl3) δ 171.1 (2 × C = O, OAc), 136.6 (CH=CH2), 115.8 (CH=CH2), 76.0 (C-2), 66.2 (CH2-OAc), 57.5 (CH2-CH=CH2), 49.6 (C-4), 34.2 (C-3), 30.1 (C-1), 28.2 (C-5), 21.1, 20.9 (2 × CH3, OAc), 11.2 (C-6); FAB-MS m/z (relative intensity) 268 (100), 208 (30). Anal. Calcd for C14H21NO4: C, 62.90; H, 7.92; N, 5.24. Found: C, 62.81; H, 7.96; N, 5.02.

Compound 13: 1H NMR (400 MHz, CDCl3) δ 5.73−5.78 (m, 1H, CH=CH2-A), 5.72−5.68 (m, 1H, CH=CH2-B), 5.39 (app t, 1H, J ≈ 8.3 Hz, H-2), 5.13 (dd, 1H, J = 3.0, 1.7 Hz, CH=CHH-A), 5.08 (dd, 1H, J = 3.0, 1.6 Hz, CH=CHH-A), 5.05 (br dd, 1H, CH=CHH-B), 5.02 (br dd, 1H, CH=CHH-B), 4.26 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.89 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.38 (d, 1H, J = 7.9 Hz, H-4), 3.20 (ddd, 2H, J = 14.0, 5.4, 1.5 Hz, CH2-allyl-A), 2.80 (dd, 2H, J = 14.0, 7.3 Hz, CH2-allyl-B), 2.24 (dd, 1H, J = 15.1, 8.5 Hz, H-3a), 1.98 (s, 3H, OAc), 1.95 (s, 3H, OAc), 1.34 (dd, 1H, J = 7.5, 5.4 Hz, H-5), 1.22−1.14 (m, 1H, H-3b), 0.69−0.64 (m, 2H, 6-CH2); 13C NMR (100 MHz, CDCl3) δ 170.9 (2 × C=O, OAc), 136.6 (2 × CH=CH2), 117.0 (2 × CH=CH2), 76.4 (C-2), 66.1 (CH2-OAc), 60.8 (2 × CH2-CH=CH2), 53.1 (C-4), 31.7 (C-3), 30.8 (C-1), 26.5 (C-5), 21.1, 20.8 (2 × CH3, Ac), 11.4 (C-6); ESI-MS (m/z) 308.2 (M + H+).

((1R,2S,4S,5S)-2-Acetoxy-4-(tert-butoxycarbonylaminobicyclo[3.1.0]hexan-1-yl)methyl Acetate (14)

Azide 11 (1.20 g, 4.74 mmol) was dissolved in CH2Cl2, and di-tert-butyl dicarbonate ((Boc)2O, 1.24 g, 5.69 mmol) was added at room temperature under argon. The hydrogenation catalyst (Pd/C, 10%, 150 mg) was added, and the mixture was flushed with H2 and stirred at room temperature until all starting material was consumed by TLC analysis. The solvent was filtered through a short pad of Celite, and the filtrate was concentrated. The crude was purified by flash chromatography on silica gel (EtOAc in hexanes 20−40%) to yield the carbamate 14 (1.46 g, 94%) as a colorless oil: [α]20D 7.59 (c 0.83, CHCl3); IR (neat) 3345, 2976, 1707, 1518, 1454, 1364, 1237, 1166, 1038, 974, 905, 853, 783 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.47 (app t, 1H, J ≈ 8.4 Hz, H-2), 4.57 (br s, 1H, NH), 4.27 (d, 1H, J = 11.8 Hz, CHH-OAc), 4.04−3.98 (m, 1H, H-4), 3.92 (d, 1H, J = 11.8 Hz, CHH-OAc); 2.11 (dd, 1H, J = 14.4, 8.0 Hz, H-3a); 2.05 (s, 3H, OAc); 2.02 (s, 3H, OAc), 1.47−1.37 (m, 11H, H-3b, H-5, t-Bu, Boc), 0.89 (dd, 1H, J = 6.0, 4.4 Hz, H-6a), 0.72 (dd, 1H, J = 8.4, 6.0 Hz, H-6b); 13C NMR (100 MHz, CDCl3) δ 171.1, 170.9 (2 × C=O, OAc), 154.8 (C=O, Boc), 79.6 (C(CH3)3, Boc), 74.9 (C-2), 65.8 (CH2-OAc), 51.1 (C-4), 35.4 (C-1), 30.2 (C-3), 28.4 (C(CH3)3, Boc), 27.7 (C-5), 21.0, 20.9 (2 × CH3, OAc), 11.0 (C-6); FAB-MS m/z (relative intensity) 328 (3.7), 212 (100). Anal. Calcd for C16H25NO6: C, 58.70; H, 7.70; N, 4.28. Found: C, 58.25; H, 7.70; N, 4.19.

((1R,2S,4S,5S)-2-Acetoxy-4-(allyl-tert-butoxycarbonylamino)bicyclo[3.1.0]hexan-1-yl)methyl Acetate (15)

To a stirred solution of carbamate 14 (350 mg, 1.07 mmol) in dry DMF (14.0 mL) were added KHMDS (0.5 M in toluene, 2.57 mL, 1.28 mmol) and allyl bromide (120 μL, 1.39 mmol) at 0 °C under argon. The mixture was continued to stir for 5 h at 0 °C and then poured into a mixture of phosphate buffer (100 mL, pH 7.2) and Et2O (100 mL). The aqueous phase was further extracted with Et2O (2 × 100 mL), and the combined organic phases were dried (Na2SO4) and concentrated. The crude residue was purified by flash chromatography (EtOAc in hexanes 20−40%) to yield the title compound 15 (252 mg, 64%) as a colorless oil: [α]20D 3.31 (c 1.16, CHCl3); IR (neat) 2976, 1736, 1687, 1454, 1364, 1330, 1237, 1170, 1143, 1023, 910, 776 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.81 (m, 1H, CH=CH2), 5.54 (app t, 1H, J ≈ 8.4 Hz, H-2), 5.16−5.06 (m, 2H, CH=CH2), 4.16 (br s, 1H, H-4), 4.35 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.89 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.81 (br s, 2H, CH2-CH=CH2), 2.16 (dd, 1H, J = 15.3, 8.5 Hz, H-3a), 2.03 (s, 3H, OAc), 2.00 (s, 3H, OAc), 1.62−1.54 (m, 1H, H-3b), 1.41 (s, 9H, OtBu), 1.37 (dd, 1H, J = 8.8, 4.3 Hz, H-5), 0.77 (dd, 1H, J = 5.8, 4.3 Hz, H-6a), 0.71 (dd, 1H, J = 8.8, 5.8 Hz, H-6b); 13C NMR (100 MHz, CDCl3) δ 171.1, 170.8 (2 × C=O, Ac), 155.1 (C=O, BOC), 135.9 (CH=CH2), 115.2 (CH=CH2), 79.8 (O-C(CH3)), 75.5 (C-2), 65.9 (CH2-OAc), 56.3 (CH2-CH=CH2), 45.6 (C-4), 36.1 (C-3), 32.2 (C-1), 28.4 (O-C(CH3), 26.8 (C-5), 21.0, 20.8 (2 × CH3, Ac), 11.2 (C-6); FAB-MS m/z (relative intensity) 368 (5), 252 (100). Anal. Calcd for C19H29NO6: C, 62.11; H, 7.96; N, 3.81. Found: C, 62.36; H, 7.89; N, 3.79.

((1R,2S,4S,5S)-2-Acetoxy-4-allylaminobicyclo[3.1.0]hexan-1-yl)methyl Acetate (12)

The Boc-protected amine 15 (250 mg, 0.680 mmol) was dissolved in anhydrous CH2Cl2 (15 mL), and TFA (2.5 mL) was added at 0 °C under argon. After the addition, the mixture was continued to stir at 0 °C and frequently monitored by TLC (CH2Cl2−MeOH 9:1). After the mixture was stirred for 3 h at 0 °C, all starting material was consumed and satd NaHCO3 (50 mL) was carefully added while the mixture was vigorously stirred. The layers were separated, and the aqueous phase was extracted with CH2Cl2 (2 × 50 mL). The combined extracts were dried (MgSO4) and concentrated. The crude product was sufficiently pure (by TLC, 1H NMR) to be used without any further purification in the next step. The spectroscopic data were identical to those reported above.

((1R,2S,4S,5S)-2-Acetoxy-4-(1,3-diallylureido)bicyclo[3.1.0]hexan-1-yl)methyl Acetate (16)

The allyl amine 12 (260 mg, 0.973 mmol) was dissolved in anhydrous CH2Cl2 (20 mL) and cooled to −20 °C under argon. Allyl isocyanate (162 mg, 1.38 mmol) was slowly added, and the mixture was gradually warmed to room temperature and stirred overnight. The solvent was evaporated off under reduced pressure, and the residue was purified by flash chromatography (silica gel, EtOAc in hexanes 60−80%) to yield the urea 16 (307 mg, 93%) as a colorless oil: [α]20D 4.94 (c 0.89, CHCl3); IR (neat) 3430, 3366, 3079, 2982, 1734, 1637, 1522, 1370, 1237, 1024, 915, 850 cm−1; 1H NMR (400 MHz, CDCl3) δ 5.85−5.75 (m, 2H, 2 × CH=CH2), 5.49 (app t, 1H, J ≈ 8.5 Hz, H-2), 5.30−5.20 (m, 2H, 2 × CH=CHH), 5.10−5.00 (m, 2H, 2 × CH=CHH), 4.87 (d, 1H, J = 8.1 Hz, H-4), 4.49 (t, 1H, J = 5.4 Hz, NH), 4.35 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.84−3.75 (m, 5H, CHH-OAc, 2 × CH2-CH=CH2), 2.09 (dd, 1H, J = 15.5, 8.8 Hz, H-3a), 1.98 (s, 6H, 2 × OAc), 1.66−1.58 (m, 1H, H-3b), 1.30 (dd, 1H, J = 8.8, 4.4 Hz, H-5), 0.77 (dd, 1H, J = 5.9, 4.4 Hz, H-6a), 0.70 (dd, 1H, J = 8.8, 5.9 Hz, H-6b); 13C NMR (100 MHz, CDCl3) δ 171.1, 170.7 (2 × C=O, Ac), 157.9 (C=O, urea), 135.7, 135.4 (2 × CH=CH2), 116.8, 115.4 (2 × CH=CH2), 75.3 (C-2), 65.9 (CH2-OAc), 55.7 (C-4), 45.6, 43.2 (2 × CH2-allyl), 36.3 (C-3), 32.2 (C-1),26.9 (C-5), 21.0, 20.8 (2 × CH3, Ac), 11.3 (C-6); FAB-MS m/z (relative intensity) 351 (15), 291 (100). Anal. Calcd for for C18H26N2O5: C, 61.70; H, 7.48; N, 7.99. Found: C, 61.60; H, 7.68; N, 7.86.

((1′R,2′S,4′S,5′S)-2′-Acetoxy-4′-(2-oxo-2,3-dihydropyrimidin-1(6H)-yl)bicyclo[3.1.0]hexan-1′-yl)methyl Acetate (20)

Urea 16 (20.0 mg, 0.057 mmol) was dissolved in hexamethyldisilazane (HMDS, 1.0 mL), and a catalytic amount of ammonium sulfate was added. The mixture was heated to reflux for 2 h and cooled to room temperature. The solvent was removed in vacuo, and the residue was dissolved in anhydrous and deoxygenated CH2Cl2 (5 mL). Grubbs second-generation catalyst (4.8 mg, 5.65 μmol) was added. The mixture was heated to reflux for 1 h and cooled to room temperature. The solvent was evaporated off, and the residue was purified by flash chromatography on silica gel (EtOAc in hexanes, 60−80%) to yield the six-membered urea 20 (12.5 mg, 71%) as a brown foam. The sample was still contaminated with tricyclohexylphosphineoxide resulting from catalyst decomposition: 1H NMR (400 MHz, CDCl3) δ 6.10−6.00 (br s, 1H, NH), 6.00−6.95 (m, 1H, H-4), 5.48 (app t, 1H, J ≈ 8.4 Hz, H-2′), 4.99 (d, 1H, J = 8.2 Hz, H-4′), 4.69−4.66 (m, 1H, H-5), 4.30 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.98 (ddd, 1H, J = 14.3, 3.2, 1.3 Hz, H-6a), 3.90 (d, 1H, J = 12.0 Hz, CHH-OAc), 3.79 (ddd, 1H, J = 14.3, 3.2, 1.7 Hz, H-6b), 2.14 (dd, 1H, J = 15.6, 8.4 Hz, H-3′a), 1.99 (s, 3H, CH3, OAc), 1.98 (s, 3H, CH3, OAc), 1.64−1.56 (m, 1H, H-3′b), 1.30 (dd, 1H, J = 8.6, 4.4 Hz, H-5′), 0.77−0.70 (m, 2H, H-6′a,b); 13C NMR (100 MHz, CDCl3) δ 171.2, 170.7 (2 × C=O, Ac), 153.4 (C-2), 124.8 (C-4), 96.6 (C-5), 75.8 (C-2′), 65.8 (CH2-OAc), 54.1 (C-4′), 41.2 (C-6), 34.1 (C-3′), 32.0 (C-1′), 25.8 (C-5′), 21.0, 20.8 (2 × CH3, Ac), 11.0 (C-6′); FAB-MS m/z (relative intensity) 309 (32), 249 (100), 297 (tricyclohexylphonsphineoxide, 71).

((1R,2S,4S,5S)-2-Acetoxy-4-(1,3-diallyl-3-benzoylureido)bicyclo[3.1.0]hexan-1-yl)methyl Acetate (21)

Urea 16 (140 mg, 0.400 mmol) was dissolved in dry pyridine (3.5 mL) at 0 °C under argon, and triethylamine (168 μL, 1.20 mmol) and benzoyl chloride (140 μL, 1.20 mmol) were added dropwise. The mixture was allowed to warm to room temperature and was stirred for 3 h. The solvent was evaporated off, and the residue was partitioned between CH2Cl2 (50 mL) and 1N HCl (50 mL). The aqueous phase was further extracted with CH2Cl2 (2 × 50 mL), and the combined extracts were dried (MgSO4) and concentrated. The crude was purified by flash chromatography on silica gel (EtOAc in hexanes 50−70%) to yield the N-acylurea 21 (162 mg, 89%) as a colorless syrup: [α]20D −6.57 (c 1.06, CHCl3); IR (neat) 3088, 2948, 2252, 1736, 1674, 1648, 1578 1412, 1369, 1238, 1195, 1025, 915, 795 cm−1; 1H NMR (400 MHz, CDCl3) δ 7.62−7.58 (m, 2H, CH-arom), 7.47−7.42 (m, 1H, CH-arom), 7.38−7.32 (m, 2H, CH-arom), 5.94−5.84 (m, 1H, CH=CH2-A), 5.66−5.56 (m, 1H, CH=CH2-B), 5.41 (app t, 1H, J ≈ 8.3 Hz, H-2), 5.30−5.10 (m, 4H, CH=CH2-A, CH=CH2-B), 4.45−4.15 (m, 4H, CH2-CH-A, CHH-OAc, H-4), 3.90−3.60 (m, 3H, CH2-CH-B, CHH-OAc), 1.98 (s, 3H, CH3, OAc), 1.96 (s, 3H, CH3, OAc), 1.50−1.30 (m, 1H, H-5), 0.65−0.59 (m, 2H, H-6a,b); 13C NMR (100 MHz, CDCl3) δ 170.8, 170.6 (2 × C=O, Ac), 169.4 (C=O, Bz), 157.8 (C=O, urea), 134.9, 133.4 (2 × CH=CH2), 132.4, 131.6, 128.3, 128.1 (C-arom), 119.2, 117.7 (2 × CH=CH2), 75.0 (C-3), 67.9 (CH2-OAc), 65.4 (CH2-CH-a), 59.2 (C-1), 49.3 (CH2-CH-b), 35.6 (C-4), 32.8 (C-2), 25.5 (C-5), 20.9, 20.7 (2 × CH3, Ac), 11.1 (C-6); FAB-MS m/z (relative intensity) 455 (14), 105 (100). Anal. Calcd for C25H30N2O6: C, 66.06; H, 6.65; N, 6.16. Found: C, 65.95; H, 6.33; N, 6.08.

((1′R,2′S,4′S,5′S)-2′-Acetoxy-4′-(N-benzoyl-2-oxo-2,3,4,7-tetrahydro-1H-1,3-diazepin-1-yl)bicyclo[3.1.0]hexan-1′-yl)methyl Acetate (22)

The N-benzoylurea 21 (150 mg, 0.330 mmol) was dissolved in dry and deoxygenated CH2Cl2 (50 mL) under argon at room temperature, and second-generation Grubbs catalyst (28.0 mg, 0.033 mmol) was added in one portion. The mixture was heated to reflux for 1 h and cooled to room temperature again. The solvent was evaporated off, and the crude was directly subjected to flash chromatography on silica gel (EtOAc in hexanes 45−65%) to yield the diazepinone 22 (130 mg, 92%) as a colorless foam: [α]20D 39.21 (c 1.00, CHCl3); IR (neat) 3037, 2960, 2883, 1732, 1677, 1461, 1364, 1235, 1192, 1024, 990, 848 cm−1; 1H NMR (400 MHz, CDCl3) δ 7.50−7.46 (m, 2H, CH-arom), 7.43−7.38 (m, 1H, CH-arom), 7.36−7.31 (m, 2H, CH-arom), 5.90−5.82 (m, 1H, H-6), 5.81−5.70 (m, 1H, H-5), 5.54 (app t, 1H, J ≈ 8.4 Hz, H-2′), 4.61 (d, 1H, J = 8.1 Hz, H-4′), 4.54 (br d, 1H, J = 18.5 Hz, H-4a), 4.44 (d, 1H, J = 12.0 Hz, CHH-OAc), 4.29 (br d, 1H, J = 18.5 Hz, H-4b), 4.11−3.92 (m, 2H, H-7a,b), 3.79 (d, 1H, J = 12.0 Hz, CHH-OAc), 2.06−2.01 (m, 1H, H-3′a), 2.05 (s, 3H, OAc), 2.00 (s, 3H, OAc), 1.58 (m, 1H, H-3′b), 1.27 (dd, 1H, J = 8.3, 4.8 Hz, H-5′), 0.77−0.70 (m, 2H, H-6′a,b); 13C NMR (100 MHz, CDCl3) δ 171.1 (C=O, Bz), 170.8, 170.6 (2 × C=O, Ac), 158.8 (C-2), 135.2, 131.4, 128.93, 128.4 (C-arom), 127.1 (C-6), 123.8 (C-5), 74.8 (C-2′),65.6 (CH2-OAc), 57.0 (C-4′), 43.0 (C-4), 40.8 (C-7), 35.7 (C-1′), 32.4 (C-3′), 26.5 (C-5′), 21.0, 20.9 (2 × CH3, Ac), 11.1 (C-6′); HRMS-FAB m/z calcd for C23H27N2O6 (M + H+) 427.1869, found 427.1856.

1-((1′S,2′S,4′S,5′R)-4′-Hydroxy-5′-hydroxymethylbicyclo[3.1.0]hexan-2′-yl)-3,4-dihydro-1H-1,3-diazepin-2(7H)-one (5)

The fully acylated nucleoside analogue 22 (50.0 mg, 0.117 mmol) was dissolved in a 1% solution of NaOH in methanol (5.0 mL) and stirred at room temperature for 10 min. The mixture was heated to reflux for 1 h and cooled to room temperature again. After neutralization with 2N HCl, the solvent was evaporated off and the crude was purified by flash chromatography on silica gel (MeOH in EtOAc 5−15%) to yield the diazepinone nucleoside 5 (24.5 mg, 88%) as a colorless foam. The foam was dissolved in acetonitrile/water (1:1, 3 mL) and was lyophyllized to yield a hygroscopic and colorless amorphous semisolid: [α]20D 16.23 (c 1.00, H2O); IR (neat) 3318, 2869, 1633, 1485, 1428, 1373, 1309, 1190, 1149, 1046, 1007, 846, 786 cm−1; 1H NMR (400 MHz, D2O) δ 5.83−5.76 (m, 1H, H-6), 5.74−5.68 (m, 1H, H-5), 4.65 (t, 1H, J = 8.6 Hz, H-4′), 4.27 (d, 1H, J = 8.3 Hz, H-2′), 3.85 (d, 1H, J = 12.3 Hz, CHH-OH), 3.76 (m, 1H, H-7a), 3.66−3.51 (m, 3H, H-7b, H-6a,b), 3.29 (d, 1H, J = 12.3 Hz, CHH-OH), 1.86 (dd, 1H, J = 15.3, 8.6 Hz, H-3′a), 1.52 (m, 1H, H-3′b), 1.17 (dd, 1H, J = 7.8, 5.1 Hz, H-1′), 0.56−0.52 (m, 2H, H-6′a,b); 13C NMR (100 MHz, D2O) δ 166.5 (C-2), 128.0 (C-6), 126.3 (C-5), 72.5 (C-4′), 62.9 (CH2-OH), 57.7 (C-2′), 42.2 (C-7), 41.1 (C-5′), 37.4 (C-4), 36.3 (C-3′), 25.6 (C-1′), 9.3 (C-6′); ESI-MS m/z 239 (M + H+), 261 (M + Na+), 277 (M + K+). Anal. Calcd for C12H18N2O3·1.0 H2O: C, 56.23; H, 7.87; N, 10.93. Found: C, 56.49; H, 7.51; N, 10.59.