Influence of ring size on the reduction of vinylogous urethanes. Applications to the synthesis of lupinine and epilupinine

.


Introduction
The general approach to alkaloid and antibiotic synthesis explored in these laboratories over a number of years is based on the use of β-acylated enamines (enaminones) and related compounds as scaffolds upon which to build the more complex architectures found in a representative range of alkaloidal families. 1The focus of our investigations has hitherto largely been on using 2-(acylmethylidene)pyrrolidines derived from pyrrolidine-2-thiones as versatile intermediates en route to target systems containing isolated or fused pyrrolidine rings, 2 for example, the 1-azabicyclo [4.3.0]nonane(indolizidine) nucleus. 3Complementary explorations of piperidine-based precursors for the synthesis of the homologous 1-azabicyclo [4.4.0]decane (quinolizidine) ring system are under-represented in our work. 4However, our recently reported observations of intriguing discrepancies in the behavior of five-membered and six-membered thiolactam precursors 5 have prompted us to examine the reactions of the latter class of compounds and their 2-alkylidene derivatives more closely, and to pay more attention to their use in the synthesis of quinolizidine alkaloids.Some relevant results are reported in this article.
The quinolizidine ring system is well represented among alkaloids isolated from both plant and animal sources, 6 and novel strategies for the stereoselective synthesis of compounds containing this important structural motif continue to receive considerable attention. 7The diastereomeric alkaloids lupinine 1 and epilupinine 2 (numbering scheme shown), the simplest members of the large group of alkaloidal metabolites of the Leguminosae (Fabaceae), 8 are exceptionally popular targets for synthesis, and are frequently used to exemplify emerging methodologies. 9In fact, an early communication from these laboratories disclosed a simple synthesis of (±)-lupinine rac-1 from the piperidine-2-thione 3 via the enaminone intermediates 4-6 (vinylogous urethanes) (Scheme 1). 10 However, full experimental details were never reported, the relatively unsophisticated spectroscopic techniques available at the time made the characterisation of compounds somewhat tentative, and the apparently stereoselective conversion of 6 into 1, although precedented, 11 was not fully probed.We thus chose to re-investigate aspects of this synthesis, using the route as a framework within which to contextualise a further exploration of behavioral differences between five-and six-membered ring intermediates.

Results and Discussion
The vinylogous urethanes on which this study is based were prepared in two steps from pyrrolidine-2-thione 12 7a or piperidine-2-thione 12 7b, conjugate addition of which to esters of acrylic acid was readily achieved by stirring the reactants together in dry tetrahydrofuran at 40 °C in the presence of a catalytic quantity of sodium hydride 13 or sodium hydroxide (Scheme 2).The N-alkyl products 3 and 8a-c were isolated in yields of 92-99%.Conversion of these thiolactams into the vinylogous urethanes 4 and 9a-c was accomplished in 75-85% overall yields by alkylation on sulfur with ethyl bromoacetate followed by Eschenmoser sulfide contraction 14 in the presence of triphenylphosphine and triethylamine.The (E)-geometry of the double bond in these products was inferred from the chemical shift of the hydrogen atoms at C-3 in the ring (δ ca.3.1), the downfield shift of about 0.6 ppm relative to (Z)-analogues 14a arising from the anisotropic deshielding effect of the carbonyl group.

75-85%
Scheme 2 A critical feature of the previously reported synthesis of lupinine shown in Scheme 1 was the chemoselective reduction of the saturated ester group of intermediate 4 while leaving the vinylogous urethane untouched.3b Indeed, treatment of the pyrrolidinylidene compounds 9a and 9b with lithium aluminium hydride (1.1 equiv.) in tetrahydrofuran at room temperature was complete within 2 h, affording the same saturated alcohol 10 in yields of 73% and 79%, respectively; as expected, the vinylogous urethane was unaffected.However, it was disconcerting to discover that the piperidinylidene analogues 4 and 9c were substantially less inert under the same conditions.The former, for instance, gave only a 3% yield of the expected product 5, while the major product 11 (65%) resulted from reduction of both the saturated ester and the double bond of the vinylogous urethane.Even at a lower temperature (0 °C) and with a shorter reaction time (45 min), the yield of 5 was only 66%, while the over-reduced product 11 was still significant (26%).Interestingly, in the original lupinine synthesis reported from this laboratory 10 (Scheme 1), the reactivity of the reductant was tempered by the addition of an equivalent of ethanol 15 to give a 71% yield of 5. Over-reduction was not mentioned, or, more probably, was not recognised.In the present work, attempts to moderate the reaction in other ways, e.g. by using Superhydride,® lithium pyrrolidineborohydride 16 or sodium borohydride in combination with lithium iodide, 17 failed to reduce either functional group.However, we were able to minimise over-reduction with lithium aluminium hydride by changing the solvent to a mixture of toluene and diethyl ether.With a 2:1 ratio of these solvents, reduction of 4 at 0 °C for 45 min afforded 5 (58%) and 11 (27%); with a 4:1 ratio of solvents, the isolated yields were 62% and 13%, respectively.
The difference in reactivity of the exocyclic C=C bond in the five-and six-membered vinylogous urethanes 4 and 9a-c is in line with Brown's observations about the stability and reactivity of double bonds that are exocyclic or endocyclic to rings of varying size. 18Brown's carefully worded generalisation included the following statement: "Reactions which involve the loss of an exo double bond will be favored in the 6-ring as compared to the corresponding 5-ring derivative".Thus the greater lability of the piperidinylidene vinylogous urethanes towards conjugate reduction, though troublesome for our purposes, should perhaps not have surprised us.However, other factors may well make even this feature of enaminone reactivity unpredictable.For example, the vinylogous cyanamide 12a, the reduction of which to the alcohol 13 proceeds in 74% yield, 4 is clearly far less susceptible to reduction of the C=C bond than the corresponding vinylogous urethane.To confirm this observation and complete our series of comparisons, we prepared the tert-butyl ester analogue 12b in 75% yield by sulfide contraction between the piperidine-2-thione 8c and bromoacetonitrile.Reduction of the ester group with lithium aluminium hydride in THF at room temperature, although affording only a 60% yield of 13, gave no observable formation of the over-reduced product.As originally performed, the cyclisation of alcohol 5 to give the quinolizidine ring system 6 (Scheme 1) entailed activation of the leaving group by initial conversion into a toluenesulfonate.We have subsequently found that this is not an ideal approach; since mixtures of toluenesulfonate and chloride products are formed, the latter in general not being sufficiently reactive to give efficient ring closure.Our preferred mode of activation is now via an alkyl iodide, which is formed in situ and immediately cyclised under appropriate conditions.In the present investigation, the alcohol 5 was treated with iodine, triphenylphosphine and imidazole, according to the method developed by Garegg and Samuelson 19 for preparing alkyl iodides from alcohols.When the reactants were heated in a 2:1 mixture of toluene and acetonitrile, the intermediate iodide cyclised spontaneously to give the 3,4,6,7,8,9-hexahydro-2H-quinolizine 6 in 74% yield.Although other workers have prepared compound 6 by different routes, 11,20,21 characterisation has hitherto been sketchy and 13 C NMR data have not previously been reported.
The selective conversion of 6 into the target quinolizidine alkaloids 1 and 2 demands that the reduction of the C=C bond proceed with reliable stereocontrol.Goldberg and Ragade reported that stereocontrolled reduction of 6 with sodium borodeuteride gave the 9a-deuterated isotopomer of (±)-ethyl lupinoate 14 exclusively, a somewhat surprising result that was rationalised in terms of conformational effects and the preferred trajectory of hydride approach. 11Lhommet and coworkers prepared 14 by catalytic hydrogenation of 6 over Raney nickel under harsh conditions (150 atm., 100 °C).21b Significantly, at a reaction temperature of 200 °C, they found that the sole product was (±)-ethyl epilupinoate 15, in which the ester group occupies the thermodynamically favoured equatorial position.Furthermore, 14 could be epimerised to 15 merely upon heating at 200 °C.Less drastically, Goldberg and Lipkin accomplished epimerisation in unspecified yield by heating the lupinate ester 14 with sodium ethoxide in ethanol. 20e found that stereoselective cis-hydrogenation of the bicyclic vinylogous urethane 6 could be accomplished under far milder conditions than those employed by Lhommet's group.Hydrogenation over Adams catalyst in absolute ethanol at a hydrogen pressure of 5 atm afforded (±)-ethyl lupinoate 14 as the sole detectable isomer in a yield of 83% (Scheme 3).Moreover, in our hands the base-catalysed epimerisation with a catalytic quantity of sodium ethoxide in boiling ethanol gave a quantitative yield of the epilupinoate ester 15.Although both isomers have been known for decades, reliable spectroscopic data are rare.Our spectroscopic data agreed well those of Lhommet 21b and also Hua et al., 22 who reported data for the enantiomerically pure compounds (-)-14 and (+)-15.In the IR spectra, Bohlmann bands 23 in the region ca.2750-2800 cm -1 indicated trans-fusion of the quinolizidine ring system.
Reduction of (±)-14 and (±)-15 to give (±)-lupinine 1 and (±)-epilupinine 2, respectively, was most successfully achieved by the dropwise addition of the precursors in dry diethyl ether to a suspension of lithium aluminium hydride in ether at 0 °C, a "reverse addition" procedure recommended by Davies and Smyth. 24The isolated yields of the two alkaloids were 95% and 88%, respectively.Spectroscopic data for these products, recorded in deuterated chloroform at 300 MHz, were in excellent agreement with those recorded at 400 MHz by Hua et al. 22 In addition, spectra recorded for lupinine in deuterated benzene accorded well with those reported at 600 MHz by Rycroft and Robins. 25

Scheme 3
The methodology described in this article is conceptually and experimentally straightforward and the transformations are efficient.We are currently applying the principles elaborated above to the enantioselective synthesis of a suite of 1,4-disubstituted quinolizidine alkaloids recently isolated from the skins of poison-dart frogs. 26

Experimental Section
General Procedures.All solvents used for reactions and chromatography were distilled before use.Tetrahydrofuran (THF) and diethyl ether were distilled from Na/benzophenone, dichloromethane, acetonitrile and triethylamine from CaH 2 , and benzene and toluene from Na. Commercially available chemicals were used as received.Melting points, recorded on a Reichert hot-stage microscope apparatus, are uncorrected.TLC was performed on aluminium-backed Alugram Sil G/UV 254 plates pre-coated with 0.25 mm silica gel 60.Column chromatography was carried out on silica gel 60, particle size 0.063-0.200mm (conventional columns) or Whatman Partisil Prep 40, particle size 0.040-0.063mm (flash columns).FTIR spectra were recorded on a Bruker Vector 22 spectrometer.NMR spectra were recorded on a Bruker AC-200 (200.13MHz for 1 H, 50.32 MHz for 13 C), Bruker AVANCE 300 (300.132MHz for 1 H, 75.473MHz for 13 C) or Bruker DRX 400 (400.132MHz for 1 H, 100.625 MHz for 13 C) spectrometers.CDCl 3 was used as solvent and TMS as internal standard.DEPT and CH-correlated spectra were routinely used for assignment of signals.J values are given in Hz.High-resolution mass spectra were recorded at 70 eV on a VG 70 SEQ mass spectrometer with a MASPEC II data system.