Carbamoyl radical-mediated synthesis and semipinacol rearrangement of β-lactam diols.

In an approach to the biologically important 6-azabicyclo[3.2.1]octane ring system, the scope of the tandem 4-exo-trig carbamoyl radical cyclization-dithiocarbamate group transfer reaction to ring-fused β-lactams is evaluated. β-Lactams fused to five-, six-, and seven-membered rings are prepared in good to excellent yield, and with moderate to complete control at the newly formed dithiocarbamate stereocentre. No cyclization is observed with an additional methyl substituent on the terminus of the double bond. Elimination of the dithiocarbamate group gives α,β- or β,γ-unsaturated lactams depending on both the methodology employed (base-mediated or thermal) and the nature of the carbocycle fused to the β-lactam. Fused β-lactam diols, obtained from catalytic OsO4-mediated dihydroxylation of α,β-unsaturated β-lactams, undergo semipinacol rearrangement via the corresponding cyclic sulfite or phosphorane to give keto-bridged bicyclic amides by exclusive N-acyl group migration. A monocyclic β-lactam diol undergoes Appel reaction at a primary alcohol in preference to semipinacol rearrangement. Preliminary investigations into the chemo- and stereoselective manipulation of the two carbonyl groups present in a representative 7,8-dioxo-6-azabicyclo[3.2.1]octane rearrangement product are also reported.

Introduction

β-Lactams, both naturally occurring and synthetic, play a pre-eminent role as medicinally important compounds, particularly as antibiotics.[1, 2] The inherent ring strain and resultant reactivity of the four-membered ring also renders β-lactams useful intermediates for organic synthesis, particularly through ring-opening reactions at the amide bond.[2, 3] As a consequence of these dual roles in medicinal chemistry and synthesis, a wide range of methodologies have been developed for the preparation and subsequent transformation of β-lactams, both monocyclic and fused to other ring systems.[2, 4]

We have previously reported a high yielding synthesis of fused β-lactam 2 through simple irradiation of readily prepared carbamoyl diethyldithiocarbamate 1 (Scheme 1).[5] 4-Exo-trig cyclization of carbamoyl radical 3 is followed by dithiocarbamate group transfer from 1 to the cyclohexyl radical 4 on the less hindered convex face of the bicyclic ring system.[6] In related research we have employed the regioselective cyclization of carbamoyl dithiocarbamate 5 to synthesize the bridged 6-azabicyclo[3.2.1]octane ring system 6 of aphanorphine, an alkaloid isolated from a blue–green algae.[7]

Scheme 1

Tandem carbamoyl radical cyclization—dithiocarbamate group transfer mediated synthesis of cis-fused β-lactams and 6-azabicyclo[3.2.1]octane ring system.

The 6-azabicyclo[3.2.1]octane ring system is found within a range of synthetic[8]–[10] and naturally occurring,[7, 11]–[18] biologically active compounds (Figure 1). The former include 7 (“azaprophen”), a synthetic muscarinic anatagonist,[9] and 8, a synthetic cocaine analogue and inhibitor of dopamine reuptake.[10] Representative natural products containing the 6-azabicyclo[3.2.1]octane ring system include aphanorphine,[7, 11] members of the Securinega alkaloids, for example securinine,[12] actinobolamine,[13] members of the hetisine alkaloids, for example nominine,[14] lyconadin A,[15] peduncularine,[16] calyciphilline D[17] and sarain A.[18] Hence this ring system has been the focus of intense synthetic interest, with a number of approaches reported in addition to those applied in target synthesis.[19]

Figure 1

Representative natural and non-natural products containing the 6-azabicyclo[3.2.1]octane ring system.

A challenge that must be addressed in the synthesis of many of the natural products shown in Figure 1 is the functionalization at C-8 of the 6-azabicyclo[3.2.1]octane ring system. Although potentially accessible through modification of our previous approach by incorporation of additional functionality in radical precursors, such as 5, we instead sought to exploit the simpler, high-yielding synthesis of β-lactam 2 (Scheme 1). In principle, the bridged bicyclic amide 11, bearing a ketone at C-8, could be prepared through a semipinacol rearrangement[20] of an oxygenated β-lactam 10, in turn derived from dithiocarbamate 9, the product of a group-transfer carbamoyl radical cyclization reaction (Scheme 2). Related semipinacol-like ring expansions of non-fused β-lactams to γ-lactams have been reported in the literature, and notably all occur with exclusive migration of the N-acyl group rather than the methylene or methine carbon atom (Scheme 3).[21]–[23] Our approach was also inspired by the X-ray crystal structure of β-lactam 2, in which the cyclohexane ring adopts a boat-like arrangement to accommodate the cis-ring fusion, placing the diethyldithiocarbamate group axial with the C–S bond approximately antiperiplanar with the C–C(O) bond of the β-lactam (C1-C2-C7-S1 torsion angle −173.2 (2)°) (Figure 2).[24] In as much as 2 can be regarded as a model for the proposed rearrangement precursor 10, migration of the N-acyl group of the β-lactam was expected to be preferred both electronically and stereoelectronically over migration of the N-alkyl group.

Scheme 2

Proposed semipinacol rearrangement approach to keto-bridged 6-azabicyclo[3.2.1]octane ring system 11. LG=leaving group.

Scheme 3

Semipinacol rearrangement of β-lactams with N-acyl group migration. PPTS=pyridinium p-toluenesulfonate.

Figure 2

Crystal structure of β-lactam 2 with ellipsoids drawn at the 50 % probability level. The group N2, C8–C12 is disordered over two positions. Only the major component has been shown for clarity.

In this paper, we report the successful implementation of this strategy through conversion of fused β-lactams of general structure 9 in three or four steps to keto-bridged bicyclic lactams 11. We also report studies on the scope and limitation of this methodology in our attempts to apply it to structurally related systems.[25]

Results and Discussion

Methodology development

Methodology development was carried on N-para-methoxyphenyl (PMP) substituted lactams (Scheme 4). The PMP derivative 13 was prepared through a similar sequence to that for the N-benzyl system 1.[5], 25a Treatment of p-methoxyaniline 12, readily prepared through alkylation of p-anisidine with 3-bromocyclohexene, with triphosgene gave an isolable carbamoyl chloride intermediate of sufficient purity to be carried through to the next step without the need for further purification. Chloride displacement with commercially available sodium diethyldithiocarbamate salt was found to require more forcing conditions for an N-PMP substituent than for the analogous N-benzyl system 1 (refluxing acetone rather than room temperature), but the radical precursor 13 could nevertheless be prepared in similarly high yield. Irradiation of 13 gave the fused β-lactam 14 as a single diastereoisomer in 84 % yield. The stereochemistry at the new dithiocarbamate stereocentre was assigned based on the close spectral similarity of 14 with 2, and the expected dithiocarbamate group transfer to the intermediate cyclohexyl radical (analogous to 4, Scheme 1). More generally, N-PMP systems were found to display favourable spectroscopic and practical features, including increased crystallinity, compared with N-alkylated systems.

Scheme 4

Preparation and dithiocarbamate group elimination of N-PMP β-lactam 14. mCPBA=meta-chloroperbenzoic acid; NMO=N-methylmorpholine N-oxide.

Our initial approach to convert β-lactam 14 into a suitable substrate for the proposed semipinacol rearrangement was based on incorporation of a hydroxyl group at the ring junction, and then employ the dithiocarbamate group as a latent leaving group.[26] Unfortunately attempts to deprotonate β-lactam 14 and quench the corresponding enolate with an electrophilic oxidant, a reaction that has precedent in non-fused systems,[27] met with failure.

Attention therefore turned to elimination of the dithiocarbamate group to form a ring-fused α,β-unsaturated β-lactam suitable for further oxidation. At the outset of this work we were unsure as to the feasibility of incorporating a double bond at the ring-junction of a [4.2.0] fused bicyclic β-lactam, although the analogous [5.2.0] bicyclic system had been previously prepared through a palladium-catalyzed carbonylation reaction.[28] Indeed, our previous studies on the thermal elimination of the dithiocarbamate group from 2 had shown exclusive elimination to the non-conjugaged alkene, and the same reaction conditions applied to the N-PMP β-lactam 14 gave alkene 16 regioselectively (Scheme 4).[29]

A screen of some common bases identified NaHMDS to be the most promising for further optimization, although avoidance of aqueous work-up proved necessary for the isolation of α,β-unsaturated β-lactam 15 (Scheme 4 and Table 1, entries 1–6). Increasing the equivalents of base or running the reaction at higher temperature resulted in inseparable mixtures of 15 and 16 (entries 7 and 8). Independent subjection of alkene 15 to 1.1 equivalents of NaHMDS in THF at room temperature for 6 h showed conversion to a 1:1 mixture of 15 and 16, whereas under the same conditions no change occurred starting from 16. This suggested that the formation of 16 in the base-mediated elimination reaction occurs through isomerization of 15 rather than a competing elimination pathway from 14.

Table 1

Base-mediated elimination of dithiocarbamate 14.[a]

Entry	Reagent[b] (equiv)	T [°C]	t [h]	Result
1	LDA (1.1)	−78 to RT	18	14
2	quinoline[c]	120	18	14
3	NaH (1.1)	RT	26	14
4	tBuOK (1.1)	RT	6	22 % 14, 37 % 15
5	NaHMDS (1)	−78	4.5	22 % 14, 46 % 15
6	NaHMDS (1.5)	−78	4.5	47 % 15
7	NaHMDS (3)	−78	1	42 % 15 + 16
8	NaHMDS (1.5)	−78 to RT	7.5 then 18	46 % 15 + 16
9	KHMDS (1.1)	−78	5.75	60 % 15
10	LHMDS (1.1)	−78	4	61 % 15
11	LHMDS (1.1)	0	2.75	16 % 15
12	LHMDS (1.1)	−40	2.5	39 % 15
13	LHMDS (1.05)+MeI (1.05)	−78	6.5	99 % 15

 All reactions were carried out in THF unless otherwise stated.

 LDA=lithium diisopropylamide; K/Na/LHMDS=potassium/sodium/lithium hexamethyldisilazide.

 Quinoline was used as solvent.

Increased yields of 15 were obtained using KHMDS and LHMDS at low temperatures (Table 1, entries 9 and 10), with higher temperatures resulting in reduced yields (entries 11 and 12). Addition of MeI to activate the dithiocarbamate group towards elimination[26] caused a dramatic increase in yield, with alkene 15 the exclusive product formed in nearly quantitative yield (entry 13).

The strain inherent in alkene 15 is evident in the X-ray crystal structure (Figure 3).[24] There is a notable deviation from planarity in the alkene, as evidenced in the C1-C2-C3-C4 torsion angle (−145.14(16)°). The C3-C2-C1 bond angle (138.30(13)°) is also larger than expected. This strain may account for the isomerisation of 15 to the non-conjugated alkene 16 under certain conditions.

Figure 3

Crystal structure of alkene 15 with ellipsoids drawn at the 50 % probability level. Selected bond lengths and angles: C3-C2-C1 138.30(13), C3-C2-C7 125.47(12), C2-C3-C4 120.17(13), C7-C2-C3-C4 −5.1(2), C1-C2-C3-C4 −145.14(16)°.

Treatment of alkene 15 with mCPBA gave epoxide 17 in reasonable yield as long as a non-aqueous work-up was employed.[30] Osmium tetraoxide-catalyzed dihydroxylation gave diol 18 stereoselectively (Scheme 4).[31] Both 17 and 18 were assigned as cis-fused β-lactams. The corresponding trans-fused β-lactams would be considerably more strained and hence unlikely to form under these conditions.[32]

The semipinacol rearrangement of epoxide 17 was first attempted. Treatment of 17 with BF3 gave no reaction at low temperature (−78 °C, CH2Cl2), and decomposition upon warming to room temperature. The use of TiCl4 resulted in ring-opening of the epoxide to chloroalcohol 19. However, treatment of 17 with PPTS in refluxing toluene gave tosylate 20 in 42 % yield after 2.25 h, and encouragingly the desired rearrangement product 21 when the reaction time was increased to 18 h, albeit in a moderate 33 % yield.

Structural assignment of the semipinacol rearrangement product as ketone 21 rather than the alternative 1,2-dicarbonyl 22 arising from methine carbon migration could not be made unambiguously by NMR spectroscopic analysis. However X-ray crystallography confirmed the formation of the 6-azabicyclo[3.2.1]octane ring system. Surprisingly the corresponding hydrate 23 crystallized rather than ketone 21 from acetone solution (Figure 4). Intermolecular hydrogen bonding is evident in the solid state, which presumably stabilizes the hydrate structure.[24]

Figure 4

Crystal structure of hydrate 23 with ellipsoids drawn at the 50 % probability level.

The more robust diol 18 offered a wider variety of potential conditions to affect the semipinacol rearrangement; however, selective activation of the secondary over the tertiary alcohol proved to be challenging. Attempted mesylation gave predominantly the dimesylate 25, alongside 24, the product of mesylation of the tertiary alcohol, with dimesylation favoured even in the presence of starting diol. Treatment of 18 with tosyl chloride, DMAP and triethylamine also showed the preference for reaction at the tertiary alcohol, with tosylate 26 isolated in 71 % yield. Switching to pyridine as base partially reversed this selectivity, with the desired secondary tosylate 27, clearly structurally distinct from its epimer 20, isolated in 50 % yield alongside ditosylate 28. Tosylate 27 underwent the desired rearrangement to 21 in refluxing toluene in the presence of pyridine, and 21 could be prepared in 58 % yield in one step from 18 without isolation of the tosylate (Scheme 5). Attempted direct pinacol rearrangement of diol 18 using PPTS in toluene at 80 °C gave only starting material, and recourse to a stronger acid (TsOH) gave either no reaction, or degradation at higher temperatures.

Scheme 5

Attempted semipinacol rearrangement of epoxide 17 and diol 18. DMAP=4-dimethylaminopyridine; py=pyridine.

The apparent higher reactivity of the tertiary over the secondary alcohol in 18 may be a consequence of the conformation adopted by the bicyclic ring system. Analogous to 2, the cyclohexane ring of 18 is forced to adopt a boat-like conformation to accommodate the cis-ring fusion of the β-lactam. As a consequence the secondary alcohol is in a more hindered flagstaff position and the tertiary alcohol is pseudo-equatorial and relatively exposed. Hence reaction at the tertiary alcohol, or migration of groups from secondary to tertiary, is feasible.

The moderate yields of 21 from epoxide 17 and diol 18, coupled with the difficulties in selectively activating the secondary alcohol of 18, led us to investigate cyclic systems. Attempted rearrangement via a cyclic orthoester through treatment of diol 18 with trimethylorthoformate and a Lewis acid gave the formate ester 29 in 41 % yield (Scheme 6).[33] Suspecting that the difficulty might lie in forming a five-membered ring intermediate with a carbon linking the two alcohols, attention turned to the use of a heteroatom linker offering a potential driving force for rearrangement.[34] Although less precedented in the literature, we were drawn to reports on the rearrangements of cyclopropyl diols to cyclobutanones by in situ formation of cyclic sulfites or sulfates, occurring at or below room temperature.[35, 36]

Scheme 6

Attempted semipinacol rearrangement through cyclic activation.

Treatment of diol 18 with thionyl chloride and pyridine gave a 1:1 mixture of cyclic sulfites 30 and 31 in 91 % yield (Scheme 6). The sulfites were separable by column chromatography, and the stereochemistry at the sulfinyl group initially assigned on the basis of the 1H NMR chemical shift of the proton in the cyclic sulfite ring, which appears at δ=5.41 ppm for 30 and 5.04 ppm for 31. The anisotropy of the sulfinyl (S=O) group results in a downfield shift for the appropriately aligned proton adjacent to oxygen in 30.[37] Subsequent X-ray crystallography of both 30 and 31 confirmed the stereochemical assignment (Figure 5).[24]

Figure 5

Crystal structures of 30 (top) and 31 (bottom) with ellipsoids drawn at the 50 % probability level. Selected bond lengths and torsion angles: 30: C2–C7: 1.542(2), C2–C3: 1.558(2), C1–O2: 1.4559(18) Å, O2-C1-C2-C7: 140.95(13), O2-C1-C2-C3: −117.46(14)°; 31: C2–C7: 1.5394(17), C2–C3: 1.5617(17), C1–O2 1.4670(15) Å; O2-C1-C2-C7 147.23(11), O2-C1-C2-C3 −108.61(12)°.

Heating a solution of cyclic sulfites 30 and 31 in diphenyl ether at 190 °C effected the desired semipinacol rearrangement to the target bridged bicyclic ketone 21, which was isolated in excellent yield by direct column chromatography of the reaction mixture (Scheme 6). Other solvents at comparable or lower temperatures were less effective, and attempts to catalyze the process with a Lewis acid resulted in lower yields (see the Supporting Information). Qualitatively, 31 required a longer reaction (135 min) time than 30 (45 min) for the semipinacol rearrangement to go to completion, with the yield of ketone 21 also lower (87 vs. 100 %). Although the crystal structures show that the migrating C2–C7 bond in 31 is shorter than in 30 (1.5394(17) vs. 1.542(2) Å), conversely the leaving group and the migrating group is better aligned in 31 than in 30 (O2-C1-C2-C7 torsion angle 147.23(11) vs. 140.95(13)°). Reduction in the overall dipole moment might also rationalize the faster and higher yielding rearrangement of 30 compared to 31: the S=O and C=O dipoles in 30 are more closely aligned than in 31. However, in the absence of additional examples and knowledge of the concertedness of the rearrangement with release of SO2 at relatively high temperatures, these rationales should be regarded as speculative at best.

Although the rearrangement of cyclic sulfites 30 and 31 overcame the need to selectively activate the secondary alcohol of 18, and provided 21 in much higher overall yield than previously achieved, we were keen to reduce both the temperature and the additional step required. To this end we investigated the use of cyclic phosphoranes. These can be prepared through reaction of diols with Ph3PCl2, conveniently generated in situ through reaction of Ph3P with a suitable chlorine source.[38] We[39] and others[40] have used in-situ-generated cyclic phosphoranes to achieve related rearrangements with 1,2-hydride migration (Meinwald-like rearrangement). In practice, treatment of diol 18 with 1.5 equivalents of Ph3P and C2Cl6 in refluxing acetonitrile gave the target ketone 21 in an excellent 94 % yield through the presumed rearrangement of an intermediate cyclic phosphorane 32 (Scheme 6). In contrast to the use of similar methodology for the rearrangement of diols to ketones with 1,2-hydride migration,[39, 40] the addition of a base such as iPr2NEt to neutralize the HCl formed in the cyclic phosphorane formation is not necessary, and in fact proved detrimental to the yield of 21. The formation of triphenylphosphine oxide as a byproduct did not prove problematic for the purification of 21 by column chromatography. This represents the first use of cyclic phosphoranes to affect semipinacol-like rearrangement with carbon–carbon bond migration.

Methodology scope and limitations

With a successful route from fused β-lactam 14 to bridged bicyclic ketone 21 by the semipinacol rearrangement of cyclic derivatives of diol 18 in hand, attention turned to determining the scope and limitations of the methodology.

Carbamoyl radical mediated synthesis of β-lactams: The 4-exo-trig carbamoyl radical cyclization has been previously investigated to a limited extent for the synthesis of β-lactams.[41]–[45] However, at the outset of this work we were aware of only one additional example in the literature, related to the cyclization of 3 to 4, for the synthesis of a ring-fused system.[42]

A range of carbamoyl radical precursors 33–39 were prepared in two steps (one purification) from the corresponding amine by using our previously described methodology: formation of a carbamoyl chloride using triphosgene and pyridine in toluene, followed by chloride displacement with sodium diethyldithiocarbamate in acetone (Table 2). Yields were generally >80 % over two steps.

Table 2

4-Exo-trig carbamoyl radical cyclization—dithiocarbamate group transfer mediated synthesis of β-lactams.[a]

Entry	Radical precursor	β-Lactam	Yield [%][b]
1			a: 80
2			b: 91
3			92
4[c]			39
			29
5			65
			16
6			55
7			86
8		no reaction	0
9[c]		degradation	0

 Conditions: hυ, 500 W Halogen lamp, Pyrex, cyclohexane (0.1 m), reflux, 2–5 h.

 Isolated yield after column chromatography.

 Chlorobenzene as the reaction solvent.

The bridged bicyclic amine 42 was prepared from the known dibromide 40 through a sequence of allylic substitution[46] followed by reductive lithiation to remove the vinyl bromide. Attempts to employ vinyl bromide 41 a in a palladium-mediated cyclocarbonylation to α,β-unsaturated β-lactam 43, a reaction reported for the corresponding non-bridged system,[28] were unsuccessful.

The novel 1,4-cyclohexadiene 44 was prepared as an inseparable 5:3 mixture with conjugated diene 45 through photoaddition of isopropylamine to benzene.[47] Treatment of the mixture of aminodienes with triphosgene gave a mixture of carbamoyl chlorides, from which the skipped diene 37 could be isolated in 48 % yield over two steps after reaction with sodium diethyldithiocarbamate (Scheme 7).

Scheme 7

Preparation of allylic amines through reductive debromination and amine–benzene photoaddition.

4-Exo-trig carbamoyl radical cyclization of carbamoyl dithiocarbamates 33–38 proceeded in reasonable to excellent yield under our standard conditions (irradiation with a 500 W halogen lamp; Table 2, entries 1–7). Although the cis-ring junction in the β-lactam products is ensured due to the constraint of the tether in the cyclization, the stereochemistry at the dithiocarbamate stereocentre depends on the facial selectivity in the group transfer to the intermediate carbon-centred radical. The cyclization of 33 a–b, 34, and 37 gave rise to single β-lactam products (Table 2, entries 1–3 and 6). As for 14, the stereochemistry of β-lactams 46 a and 46 b was assigned based on the close spectral similarity to the previously reported β-lactam 2.[24] The stereochemistry in 47 and 52 was assigned on the basis of the expected dithiocarbamate group transfer to the less-hindered face of the bicyclic radical intermediate, but has not been unambiguously determined.

The cyclization of the seven-membered carbocylic systems 35 and 36 gave rise to mixtures of diastereomers (Table 2, entries 4 and 5). The structure of the major diastereomer 48 from the cyclization of 35 was confirmed by X-ray crystallography (Figure 6).[24] The relative stereochemistry in the bridged systems 50 and 51 was confirmed by X-ray analysis of a subsequent derivative of the minor isomer 51 (vide infra). Hence the major isomer in the cyclization of both 35 and 36 results from group transfer syn to the hydrogen atoms at the β-lactam ring junction (as it does in the cyclization of 1, 13, 33, 34, and 37), although the diastereoselectivity is higher for 36, presumably due to the additional rigidity in the tricyclic ring system rendering the face syn to the one-carbon bridge less accessible.

Figure 6

Crystal structure of 48 with ellipsoids drawn at the 50 % probability level.

Carbamoyl dithiocarbamate 39 did not provide any of the expected cyclization product under our standard conditions, with extensive degradation occurring when the higher boiling chlorobenzene was instead used in place of cyclohexane as reaction solvent (Table 2, entries 8 and 9). Cyclization of 39, if it occurred, would generate tertiary alkyl radical 54 (Figure 7). We have previously observed lower yields in the 5-exo-trig carbamoyl radical cyclization onto a trisubstituted alkene to produce 55 b compared with a terminal alkene to produce 55 a.5b This situation is exacerbated in the case of the 4-exo-trig cyclization, with lactam 56 b only isolated in 5 % yield from a complex reaction mixture, compared to 56 a lacking the methyl groups.5a, [48] A combination of a slower cyclization[45] of the nucleophilic carbamoyl radical onto a more electron-rich double bond, and a slower group-transfer reaction,[49] may accounts for these trends, and the lack of product formation from 39.

Figure 7

Effect of substitution on carbamoyl radical cyclization—dithiocarbamate group transfer.

Dithiocarbamate elimination: Elimination of the dithiocarbamate group from β-lactams 46 a and 2 bearing N-benzylic groups to form α,β-unsaturated β-lactams 57 a and 57 c occurred in excellent yield (Table 3, entries 1 and 2). The N-octyl lactam 46 b required a larger excess of LHMDS and MeI to drive the reaction to completion, with α,β-unsaturated lactam 57 b obtained in lower yield (entry 2). Thermal elimination of 46 a gave alkene 58 a in a similar yield to 16 and that previously reported for 58 c.[29]

Table 3

Base-mediated and thermal elimination of the dithiocarbamate functional group.

Entry	β-Lactam	Base-mediated elimination[a]	Yield [%][b]	Thermal elimination[c]	Yield [%][b]
1			a: 88		a: 80
2			b: 67		b: not determined
3			c: 93		c: 79[29]
4			9[d]		77
5		no reaction	–		85
					10
6			93		89
7			78	degradation
8			82	no reaction, recovered starting material
9			2+98 rsm[e]		34
10			65+22 rsm[f]
11			89+11 rsm[g]
12			95	not determined

 Conditions: MeI (1.1–5 equiv), LHMDS (1.1–5 equiv), THF, −78 °C, 5–8 h.

 Isolated yield after column chromatography.

 Conditions: Ph2O, reflux, 1–7 h.

 Conditions: MeI (5 equiv), LHMDS (5 equiv), THF, −78 °C to RT, 18 h.

 Conditions: MeI (10 equiv), LHMDS (10 equiv), −78 °C to RT, 18 h.

 Conditions: MeI (1.1 equiv), LDA (1.1 equiv), THF, −78 °C to RT, 18 h.

 Conditions: LHMDS (1.5 equiv), Davis oxaziridine (1.5 equiv), 0 °C to RT, 18 h.

Attempted base-mediated elimination of the cyclopentyldithiocarbamate 47 was unsuccessful, with only starting material returned. The use of five equivalents each of LHMDS and MeI gave mainly starting material alongside small quantities of the C-methylated β-lactam 59, which suggests that deprotonation occurs but presumably the α,β-unsaturated β-lactam is too strained to form. Thermal elimination of 47 gave the expected alkene 60 in good yield (Table 3, entry 4).

The epimeric dithiocarbamates 48 and 49 showed divergent behaviour towards thermal and basic elimination conditions (Table 3, entries 5 and 6). Whereas 48 proved surprisingly resistant to base-mediated elimination, thermal elimination gave predominantly the conjugated alkene 61, presumably the larger ring size better accommodating the double bond at the ring junction. In contrast, thermal elimination of 49 gave non-conjugated alkene 62, consistent with the concerted Chugaev-like mechanism of the process.[29] Previous studies in our group have shown that, despite the high temperatures required for thermal elimination of the dithiocarbamate group, equilibration between alkene regioisomers does not occur under the reaction conditions. Product outcome is determined by the availability of a suitable β-H syn to the dithiocarbamate group, hence only 62 forms from thermolysis of 49. Base-mediated elimination was successful in the case of 49, providing the target alkene 61 in an excellent 93 % yield. Hence both 48 and 49 could be converged to the desired alkene 61 under appropriate conditions.

Tricyclic dithiocarbamate 50 was successfully converted to alkene 63 under basic conditions, but degraded upon attempted thermolysis (Table 3, entry 7). The minor epimer 51 could not be utilized—it underwent clean C-methylation to provide 64 in high yield under basic conditions, and unsurprisingly could not be eliminated under thermal conditions given that this would generate an anti-Bredt alkene (entry 8). The structure of 64 was proven by X-ray crystallography (Figure 8),[24] which, given that the deprotonation and resulting methylation of the β-lactam does not affect the dithiocarbamate stereocentre, also confirmed that the major diastereoisomer formed in the cyclization of 36 was 50 (Table 2, entry 5).

Figure 8

Crystal structure of 64 with ellipsoids drawn at the 50 % probability level.

Attempted base-mediated elimination of 52 gave mainly starting material and small amounts of N-isopropylbenzamide (65) (Table 3, entry 9). The yield of 65 increased slightly with a change of base to LDA (entry 10), and dramatically when MeI was replaced with the Davis oxaziridine in an attempt to trap out the putative deprotonated β-lactam with an oxygen source (entry 11). Although the role of the oxaziridine in this process is not known,[50] the formation of 65 in all cases is consistent with a presumably facile base-mediated fragmentation of the target diene 68, generated in situ (Scheme 8).

Scheme 8

Base-mediated and thermal elimination of dithiocarbamate 52.

Thermolysis of 52 also generated a surprising result. In this case the urea 66 was isolated from the reaction mixture in 34 % yield (Table 3, entry 9). The formation of 66 can be rationalized according to the sequence shown in Scheme 8. At the high temperatures involved, conjugated diene 69, generated in situ, undergoes pyrolytic ring fission to benzene and isopropylisocyanate.[51] The dithiocarbamic acid byproduct of the dithiocarbamate group elimination fragments to diethylamine and carbon disulfide,[29] and whereas normally these are lost at high temperature, the amine reacts with the isocyanate to form urea 66.

Base-mediated elimination of the dithiocarbamate group in the simple monocyclic β-lactam 53 gave alkene 67 in excellent yield (Table 3, entry 12). We have previously found the analogous monocyclic N-Bn β-lactam degrades under thermal elimination conditions,[29] and hence overall this route provides efficient access to exo-methylene β-lactams in combination with our high-yielding radical cyclization methodology. Alkene 67 has previously been synthesized in a palladium-catalyzed carbonylation of a 2-bromoallylamine, albeit in low yield.[28]

mCPBA-mediated epoxidation of alkene 16 provided a means to introduce additional functionality on the cyclohexanone ring of bicyclic lactam 21 (Scheme 9). The stereochemistry of epoxide 70 was assigned based on the expected preferential attack on the convex face of the bicyclic ring system. Regioselective base-mediated ring-opening of epoxide 70 gave α,β-unsaturated β-lactam 71. The liberated alcohol was protected as the benzoate ester 72.

Scheme 9

Functionalization of alkene 16.

Dihydroxylation and semipinacol rearrangement: The dihydroxylation of α,β-unsaturated β-lactams 57 a–c, 72, 61 and 67 gave diols 73 a–c, 74, 75 and 78, respectively, in reasonable to good yield (Table 4, yields of diols in parentheses). Treatment of alkene 63 under the same conditions gave the expected diol 76 as the major product as an inseparable mixture along with the α-hydroxyketone 77. Attempts to minimize the formation of the unwanted byproduct 77 were unsuccessful.[52] In all cases dihydroxylation is completely diastereoselective, re-establishing the cis-ring fusion of the β-lactam.

Table 4

Dihydroxylation and semipinacol rearrangement of β-lactams.

Entry	Diol (% yield from alkene dihydroxylation)[a]	Method[b]	Product(s)	Yield [%]
1		A		79 a: 77
2				79 b: 77
3				79 c: 83
4		B		79 a: 80
5				79 b: 80
6				79 c: 97
7		A		75
8		B		77
9		A		82
10		B		98
11		A[c]		64[d]
12		B		28[e]
13		A	no reaction[f]	–
14		B		85

 Dihydroxylation conditions: cat. OsO4, NMO (2.4 equiv), 5:5:2 acetone/H2O/tBuOH, 40 °C, 18 h.

 Method A: i) SOCl2, pyridine, 0 °C to RT; ii) Ph2O, 190 °C, 2–5 h. Method B: PPh3, C2Cl6, CH3CN, reflux, 18 h.

 Reaction conditions: ii) Ph2O, reflux, 2 h.

 Yield over three steps from alkene 63.

 Yield over two steps from alkene 63.

 No reaction at 190 °C in Ph2O. Decomposition in refluxing Ph2O.

Semipinacol rearrangement of diols 73 a–c, 74, 75 and 78 was attempted via both the corresponding cyclic sulfites in two steps (Table 4, method A), and in one step via the cyclic phosphorane (method B). As for 30 and 31, cyclic sulfites were obtained as approximately 1:1 mixtures of diastereomers. For comparison purposes yields in Table 4 for method A are over two steps, formation of the cyclic sulfite and subsequent thermolysis of the mixture. In general yields are comparable or slightly better using method B, but in some cases this method fails, despite the milder conditions.

β-Lactams fused to six-membered rings rearranged under both conditions (Table 4, entries 1–8). Notably the epoxide stereochemistry, established in 70, is translated into the axially-orientated benzoate in 80. The lack of a large axial–axial coupling for the proton adjacent to oxygen in 80 is consistent with the axial orientation, and confirms the expected stereoselectivity of the epoxidation step (Scheme 9).

Rearrangement of the seven-membered ring-fused β-lactam diol 75 gave the keto-bridged bicyclic lactam 81, again with complete selectivity for N-acyl group migration, despite the larger ring size (Table 4, entries 9 and 10). The 7-azabicyclo[4.2.1]nonane ring system 81 is found in members of the Gelsemium alkaloids, which have been the subject of some synthetic interest.[53]

Treatment of the mixture of diol 76 and ketoalcohol 77 with thionyl chloride and pyridine allowed for the separation of 77 from the cyclic sulfites derived from 76. Rearrangement of the cyclic sulfites did not occur at 190 °C, but in refluxing diphenyl ether (b.p. 259 °C) conversion to the doubly bridged ring system 82 occurred (Table 4, entry 11). In contrast, direct subjection of the mixture of 76 and 77 to Ph3P and C2Cl6 in refluxing MeCN (method B) did not give any of the semipinacol rearrangement product 82. Instead, chloroalcohol 83 was isolated in low yield. The conversion of alcohols to chlorides by using Ph3P and electrophilic chlorine sources (Appel conditions) is known in the literature.[54] The stereochemistry in 83 was assigned on the basis of the expected SN2 displacement by chloride, and also suggested by the lack of coupling between the proton adjacent to chlorine with the adjacent bridgehead proton. Molecular models showed that the dihedral angle between these protons is close to 90° when the chlorine is syn to the one carbon bridge, as in 83. The lack of rearrangement of 76 under these conditions may suggest that the cyclic phosphorane does not form, although we do not have any evidence for this. The higher temperature required to rearrange the corresponding sulfite suggests that the barrier to rearrangement is higher for the diol 76 compared to diol 75, which lacks the constraint imposed by the additional one-carbon bridge, allowing other reaction pathways to compete. Raising the temperature of the Ph3P/C2Cl6 reaction by running the reaction in a microwave up to 150 °C over 3 h still only provided 83 in low yield, with no evidence of formation of 82.

Semipinacol rearrangement of the monocyclic β-lactam diol 78 could not be achieved under either set of conditions. Although the cyclic sulfite could be prepared in 68 % yield, no rearrangement occurred upon heating at 190 °C, and the reaction mixture underwent decomposition in refluxing Ph2O. High-yielding conversion to the chloroalcohol 84 occurred upon treatment of 78 with Ph3P and C2Cl6 in refluxing acetonitrile. Clearly the competing SN2 substitution pathway is particularly favourable at the primary alcohol of 78. More generally, the preference for ring-fused systems to undergo semipinacol rearrangement rather than Appel reactions can, therefore, be ascribed to a combination of factors: the slower SN2 reaction at a secondary rather than a primary alcohol, the enforcement of a favourable orbital alignment for bond migration, and increased ring strain offering a greater driving force for rearrangement.

Functional-group transformations: The 7,8-dioxo-6-azabicyclo[3.2.1]octane ring system is a potentially versatile intermediate for organic synthesis. Preliminary studies have shown that the two carbonyl groups in 21 can be chemo- and stereoselectively functionalized (Scheme 10). Treatment with l-selectride gave the axial alcohol 85 stereoselectively. Wittig methylenation gave terminal alkene 86 in excellent yield. Carbon–carbon bond formation at the amide carbonyl proceeded through formation of thioamide 87, activation as the methyl sulfonium salt 88, and subsequent treatment with allyl Grignard followed by sodium cyanoborohydride.16a The resulting diene 89 was isolated as a single diastereomer, presumed to be the result of reduction from the less-hindered exo-face of the intermediate imminium. Tentative assignment of the C-7 stereocentre was also based on the absence of an nOe signal between the axial hydrogen at C-3 and the newly installed hydrogen at C-7.

Scheme 10

Reactions of lactams 21 and 57 c.

In the course of investigating potential activation pathways for the transformation of α,β-unsaturated β-lactam 57 c into bridged bicyclic amide 79 c, we also investigated a 1,3-dipolar cycloaddition reaction with the isolable nitrile oxide 90.[55] The double bond in 57 c was shown to be a competent dipolarophile, providing 2-isoxazoline 91 as a single stereoisomer in good yield (83 %). Notably, the regioselectivity is opposite to that reported for the 1,3-dipolar cycloaddition of a nitrile oxide with a monocyclic exo-methylene β-lactam.[56]

Conclusions

The 4-exo-trig carbamoyl radical cyclization—dithiocarbamate group transfer reaction has been shown to be an efficient and practical methodology for the synthesis of ring-fused β-lactams. Good yields of β-lactams are achieved with the exception of a system carrying double substitution at the alkene terminus. Novel conditions for the base-mediated elimination of the dithiocarbamate group have been developed, which provide access to alkene regioisomers unavailable through thermolysis. Dihydroxylation of α,β-unsaturated β-lactams provides substrates which undergo semipinanol rearrangement with exclusive N-acyl group migration under all conditions. This selectivity is consistent with prior examples of semipinacol rearrangement of non-fused β-lactams in the literature and the expected better alignment of the migrating bond with the breaking C–O bond, even when constrained within a heterocyclic ring system. In situ generated cyclic phosphoranes have been shown to undergo semipinacol rearrangement with C–C bond migration for the first time, and provide milder and shorter routes to target compounds over the use of non-cyclic systems and of cyclic sulfites, unless chloroalcohol formation competes. The resulting keto-bridged bicyclic lactams are versatile intermediates in target synthesis.