Synthetic approaches towards a new class of strained “lactenediynes”

The paper described two alternative synthetic approaches towards a new class of strained “lactenediynes”, compounds where a 10-membered enediyne ring is fused with a β -lactam. Although the two alternative syntheses were successful up to the last step, the cyclization to give the desired products failed in both cases, probably because of excessive steric strain in the products. In one of the two approaches (which was the most efficient one in terms of overall yield) a mixture of diastereoisomeric cyclodimers was isolated in moderate yield. A certain degree of stereoselectivity was observed. These compounds may be interesting as new supramolecular systems.


Introduction
The natural enediyne antibiotics are among the most potent anti-cancer substances known to date. 1 They act through a unique mechanism involving direct radical attack on cellular DNA that provokes single or double cleavage of the DNA chain. 2The high potency of these compounds (e.g., Calicheamicin) 3 is counterbalanced by their poor selectivity.This problem has recently been solved partly by conjugation with a selective antibody.The resulting conjugate (Mylotarg™) has shown very promising activity towards some types of previously intractable tumors. 4However, the quest for simpler analogues of the highly complex (and often unstable) natural enediynes, endowed with higher selectivity, is still very active. 5,6he design of "artificial" enediynes 7 is based on the special mechanism of action of these compounds, which involves a triggering event that converts a stable prodrug into a reactive drug.The reactive moiety is typically a 3-ene-1,5-diyne moiety embedded in a 10-membered ring.
While it has been demonstrated that this system can easily undergo the Bergman rearrangement, 8 producing a diradical, the presence of suitable additional steric elements may prevent this reaction.These elements may therefore be regarded as a "safety-catch".In order for the drug to work, they should be easily removable under physiological conditions.Along these lines, about ten years ago we introduced 9 a new class of bicyclic enediynes, characterized by the fusion of the 10-membered enediyne ring with a β-lactam.The latter acts as the "safety-catch", preventing cycloaromatization.On the other hand, opening of the β-lactam (that is removal of the safetycatch) represents the needed "triggering event" that converts the stable prodrug into a drug.
In our previous papers we reported the synthesis of two types of lactenediynes, represented by the general formulas 1 [10][11][12] and 2 13 (Scheme 1).In particular, compounds 1 have proved to be very promising.They are very stable prodrugs (even in the dry state), but readily undergo cycloaromatization after β-lactam opening.While normally the β-lactam is too stable to undergo spontaneous opening under physiological conditions, the incorporation of appropriate activating substituents affords compounds that undergo the cascade of reactions leading to DNA damage under mild conditions.Taking advantage of this fact, and using the "handles" present on the starting lactenediyne for attaching DNA-complexing substructures, we have recently prepared a series of lactenediynes 14 characterized by their ability to induce single-and double strand DNA cleavage at concentrations as low as 10 -7 M.This kind of activation is however unselective, thus bringing the same drawbacks as the natural enediynes.Thus, we are now trying to develop a controlled mechanism for β-lactam opening, by tethering an amino group to the nitrogen in order to induce an intramolecular transamidation reaction. 15,16After a thorough study of this acyl transfer process 16 we have been able to demonstrate the feasibility of this approach, by preparing the protected amines 3 that are able to cleave DNA only after removal of the protecting group. 15Coupled with the use of enzymatically or photochemically removable protecting groups, this result might lead to the development of lactenediyne prodrugs which could be made active at will.However, the DNA-cleaving activity of the free amines 4 turned out to be disappointingly lower than that of unselectively activated lactenediynes.Further studies have shown that the main problem is the slowness of the intramolecular transamidation process that converts the stable lactenediyne-amines of general formula 4 into enlarged lactams that then undergo fast cycloaromatization.Surprisingly, in these compounds the β-lactam ring was found to be even more stable towards hydrolysis than in the simple monocyclic β-lactam models used in our preliminary studies. 16actenediynes of type 1 were not expected to be more strained than simple monocyclic βlactams.As evidenced by force-field calculations carried out with CSC Chem3D, trans fusion with the cyclodeca-3-ene-1,5-diyne produces only minor changes in the conformation of the 10membered ring (Figure 1).Thus, the β-lactam may preserve its planarity.
In seeking analogues of 1 endowed with higher reactivity at the β-lactam site, we were attracted to the isomeric lactenediynes of type 5 (Scheme 1), where the azetidinone is trans fused to carbons 7 and 8 of the parent enediyne ring.In this case, it is clear that the fusion must bring about a greater change in the conformation of the 10-membered ring.Figure 1 shows the minimized (Chem3D) conformations of unsubstituted cyclodeca-3-ene-1,5-diyne compared to those of simple representatives of the general structures 1 and 5.While in 1 the conformation of the 10-membered ring is quite similar to that of the parent monocyclic compound, in the case of 5, there is a remarkable deviation from the preferred conformation of the parent compound.Especially noteworthy is the modification of the indicated dihedral angle.This distortion also causes a deviation from planarity for the β-lactam ring (the characteristic N-C-C-C=O dihedral angle becomes 8.5°).As a result, 5 experiences an increase of calculated steric strain of about 6.5 Kcal/mole, compared to the isomeric 1.We argued that the increase in steric strain in this new class of lactenediynes could overcome the low reactivity in the intramolecular transamidation of the β-lactam experienced for compounds 4 and therefore permit the development of more efficient selectively activated lactenediyne prodrugs.Therefore, we embarked on the synthesis of these new types of strained lactenediynes.

Results and Discussion
In the total synthesis of lactenediynes it is mandatory to assemble the enediyne ring after the azetidinone: the latter is necessary as a "safety-catch" in order to stabilize the former.There are two main general strategies: the conjugated enediyne may be constructed during cyclization of the 10-membered ring or, alternatively, it may be assembled in a previous step.We have studied both approaches.Scheme 2 summarizes the designed retrosyntheses.The final formation of the 10-membered ring was expected to be the most difficult step, in view of the steric strain of the bicyclic system.Therefore, we selected the two methodologies that, in our personal experience and according to literature data, seemed most likely to meet success.In both cases, the first stages entail the preparation of a highly functionalized β-lactam precursor.For this task, the Staudinger reaction is probably the most useful method thanks to its high convergence.For the preparation of 5, the presence of a triple bond directly attached to carbon-4 of the azetidinone is an advantage, since Staudinger reactions are known to give good results when non-enolizable imines are employed.

Scheme 2. Retrosyntheses.
For route A, we needed a diyne 6, in order to exploit the Stille-type coupling developed by Danishefsky. 17,18It should be noted that this is the cross-coupling method of choice for cyclic enediyne synthesis, whereas other types of cross-coupling reactions (Sonogashira, Suzuki, metathesis) have so far met no success in the final cyclization to 10-membered enediynes.1a,17 This compound could derive from manipulation of the enyne 9. Our plan was to prepare it by steroselective allylation of β-lactam 10, in turn prepared by Staudinger reaction of the imine 11.We decided to use the common intermediate 9 also for the second approach (route B).In this case, the allyl group should be manipulated by removing one carbon atom.After construction of the acyclic enediyne, we planned to accomplish the ring formation through the Nozaki reaction, 19 which is probably the most general and efficient reaction employed in enediyne synthesis for the crucial closure of the mesocyclic ring.
Before starting the synthesis we had to select the R 1 and R 3 groups.Our choice was dictated by the compatibility with the planned reaction sequences.Moreover, we wanted a group R 3 that could be removed at the end of the synthesis.Finally, R 1 should preferably be a protected alcohol for two reasons: (a) because the Staudinger reaction is particularly efficient with alkoxyacetic acids; (b) in order to have an additional handle to be exploited for attaching activating or DNAcomplexing substituents. 12Thus we selected two orthogonal oxidatively removable protecting groups: the p-methoxyphenyl (PMP) and the p-methoxybenzyl (PMB).
The required alkoxyacetic acid 12 was prepared straightforwardly on a multigram scale starting from p-methoxybenzyl alcohol (Scheme 3). 20Although the aldehyde 15, 21 as well as the alcohol 14, 22 were known compounds, the reported preparations turned out to be, in our hands, unsatisfactory for large scale synthesis.We preferred a slightly longer route that was, however, well suited for our purposes thanks to the high yields and the possibility of purifying the intermediates by distillation.Thus, propargyl alcohol 13 was converted, according to a literature procedure, into the THP derivative, 23 silylated, and deblocked to give 14 in excellent yields.After oxidation, the volatile aldehyde 15 was converted directly into the crude imine 11 which, without purification, was subjected to a Staudinger condensation with 12, under the conditions developed by Palomo. 24The yield was rather good for this kind of reaction, while the selectivity favoring the cisisomer 16a over the transone 16b was slightly lower than usual.The cisand transisomers could easily be separated by chromatography and/or crystallization.Obviously, these β-lactams were obtained in racemic form.Thus, all the compounds described in this paper were racemic, although, for the sake of clarity, just one enantiomer is shown.
The next step turned out to be one of the most troublesome.We had previously alkylated, with high stereoselectivity, azetidinones similar to 16 by reaction of the corresponding lithium enolate with a propargyl bromide.We expected the same behavior in this case.However, while the diastereoselectivity was also complete in this case, the yield was lowered by the formation of considerable amounts of self-condensation products.The main difference between 16 and the analogues previously used is the triple bond (instead of a styryl unit) positioned at carbon-4.The low steric requirements of the alkyne are most likely the cause of this unexpected behavior.After careful optimization we could raise the yield to 59%, but were unable completely to suppress the self-condensation by-process.Interestingly, by replacing the PMBO group with a methoxy group an even worse result was obtained: we could isolate only 20-25% of the expected allylated product.Even more surprisingly, starting from the transepimer of 16, no adduct 17 was obtained, but only self-condensation products.Since the transisomer should obviously form the same enolate, this striking behavior reflects the fact that the balance between the rate of enolization and self-condensation is crucial.The transisomer is probably enolized more slowly because the proton is more encumbered.It is interesting to note that the obtained self-condensed adducts are not allylated, so selfcondensation must take place only before allylation.In order to improve the yield we tried to use other N-protecting groups (e.g., silylated protection) but without success.Anyhow, the stereoselectivity of allylation was remarkably high: we could observe only one diastereoisomer, whose relative configuration was unambiguously established as transby NOE experiments.Despite the moderate yield of the allylation, 17 was easily prepared on > 10 g scale.
As mentioned before, the intermediate 17 was used for both synthetic routes.We will describe route A first.Transformation of 17 into an homologated alkyne required a regioselective hydroboration-oxidation, which was successfully achieved with 9-BBN (Scheme 4).The resulting primary alcohol was oxidized to the corresponding aldehyde and then subjected to the Corey-Fuchs protocol. 5,25The first step (formation of the vinylic dibromide) worked well with a 72% yield.More problematic was the subsequent elimination step with n-BuLi that, probably because of reactivity of the azetidinone, furnished in the best cases only a 30-40% yield of the desired alkyne 19.In order to by-pass this homologation we also tried to react directly the lithium enolate of 16 with 4-iodo-1-trimethylsilylbut-1-yne.However, only the elimination product (4-trimethylsilylbut-1-ene-3-yne) was detected.Desilylation of 19 and di-iodination afforded in 52% unoptimized yield the required substrate 20 for the final Danishefsky cyclization.To our disappointment, by using the conditions we had employed previously several times in the syntheses of Dynemicin analogs, 5,18 we were unable to detect even traces of the desired lactenediyne 21.We observed instead a rather unstable product whose structure was tentatively assigned as the bis-stannyl derivative 22.We thought that this failure might be because Danishefsky reaction is quite sensitive to conformational factors.
We hoped that route B, based on the more general Nozaki cyclization, might be more successful (Scheme 5).In this case, the double bond of 17 had to be oxidatively degraded.Previous work in our laboratory 15 has shown the possibility of ozonizing selectively a terminal double bond in the presence of a terminal (non-silylated) triple bond.Thus, 17 was desilylated to 23 and subjected to an ozonolysis-reduction to give the alcohol 24.This alcohol has a remarkable tendency to undergo intramolecular transacylation under basic catalysis, affording the corresponding lactone 26.Therefore, it is important to perform the NaBH 4 reduction at low temperature.
The alcohol 24 could, in principle, be subjected directly to Castro-Stephens-Sonogashira coupling with (Z)-1-chloro-4-trimethylsilylbut-1-ene-3-yne 26 in order to install the acyclic enediyne moiety.However, by using the usual method [Pd(PhCN) 2 Cl 2 , CuI, piperidine, THF, "sacrificial" alkyne], we observed extensive intramolecular transacylation to give the enediyne derivative of lactone 26. 10 Evidently, piperidine at room temperature is sufficient to promote this intramolecular process.Thus, we had to protect the primary hydroxyl before carrying out the coupling.In this way, enediyne 27 was obtained in excellent yields.Having installed all the needed carbon atoms, a series of high-yielding functional group transformations allowed conversion of 27 into the iodoaldehyde 29 in 77% yield for four steps.The final oxidation of 28 to 29 was not trivial.Swern oxidation or other attempted methods gave unsatisfactory yields.
Eventually the best results were achieved with IBX. 27The synthesis from 17 of this aldehyde, which is the substrate for the final Nozaki cyclization, turned out to be very efficient (49% for 8 steps).
We then subjected the iodoaldehyde 29 to the usual Nozaki conditions (CrCl 2 , NiCl 2 , THF).The starting material was completely consumed, and we could detect only three spots (A, B, and C) on TLC.They were isolated and accounted for an overall yield of only 27%.During work-up, however, we noticed a remarkable amount of poorly soluble gummy products, that were highly polar on TLC, and did not migrate, even with pure AcOEt.We think that these could be either decomposition products (deriving from opening of the β-lactam ring) or linear oligomers.Examination of the three spots A, B, and C in HPLC-MS indicated that none of the desired product 30 was present: they were all composed of cyclodimers, according to the molecular mass.While A and B were single products (from HPLC and NMR), C contained at least three isomeric products.It is worth noting that, since the starting aldehyde is racemic, and because of the formation of two new stereogenic centers, several stereoisomeric cyclodimers 31 are possible.In 31 there are six asymmetric centers, but two relative configurations are fixed (the βlactams are trans-).Therefore, according to the vant' Hoff rule, 16 stereoisomers (8 diastereoisomers) are expected.However, for symmetry reasons, some of them coincide and therefore there are only six possible diastereoisomers.In some of them, the two halves are identical by NMR (homotopic or enantiotopic).The major compound obtained (corresponding to spot A) accounts, alone, for 43% of the whole mixture.Thus, the cyclization displays a certain degree of stereoselectivity.A gives in the NMR a single set of signals for the two halves.Therefore, it should be a symmetric compound, with the same relative configuration between the stereogenic centers 20, 1, and 3 in the two halves.It was not possible, however, to establish unambiguously which is the relative configuration, nor whether the absolute configurations of the two halves are the same (meso-compound) or not (C 2 symmetry).A and B are white solids, poorly soluble in most solvents, but quite stable in the dry state.Thus the Nozaki reaction, which was so efficient for the synthesis of lactenediynes of type 1 (with yields ranging from 60 to 80% and with no formation of cyclodimers) in this case completely failed to afford the desired monomer.
The results obtained show that lactenediynes of general formula 5 are probably destined to remain a chimera.The feature that particularly attracted us, that is their expected steric strain, is most probably the very reason for the failure to obtain them through closure of the 10-membered ring.Maybe this time we expected too much, but we reasonably hoped that the steric strain would not completely prevent monomer formation.This study has demonstrated that the cost to be paid in terms of strain is, in this case, too high for cyclization to occur.However, apart from the last step-which can be optimized-the overall synthesis (route B) to 31 is efficient and these cyclodimers may find other interesting application (e.g., as host-guest systems).The partial stereoselectivity in the dimerization process is a positive feature toward this goal.

Experimental Section
General Procedures.NMR spectra were recorded in CDCl 3 at 200 or 300 MHz ( 1 H), and 50 or 75 MHz ( 13 C), using TMS as internal standard.Chemical shifts are reported in ppm (δ scale), coupling constants are reported in Hertz.Peak assignment in 1 H-NMR spectra was also made with the aid of double resonance and COSY experiments.In AB systems, the proton A is considered downfield and B upfield.Peak assignment in 13 C-spectra was made with the aid of DEPT experiments.
GC-MS were carried out on a HP-5971A instrument, using an HP-1 column (12 m long, 0.2 mm wide), electron impact at 70 eV, and a temperature of about 170°C.Only m/z > 33 were detected.All analyses were performed with a constant He flow of 0.9 ml/min, and (unless otherwise stated) with an initial temperature of 100°C, initial time 2 min, rate 20°C/min, final temp.260°C, final time 4 min, injection temperature 250°C, detector temp.280°C.Rt are in min HRMS was carried out on a hybrid q-TOF geometry tandem mass spectrometer (Q-STAR XL MS/MS system -Applied Biosystems MSD Sciex, Toronto, Canada) equipped with a MALDI ion source.2,5-Dihydroxybenzoic acid at a final conc. of 10 mg/ml in 70:30 0.1% TFA in H 2 O/ 0.05% TFA in CH 3 CN was used as matrix.All the measurements were carried out by mixing 500 fmoles of peptides obtained in-house (ALELFR, MW=747.4272,LFTGHPETLEK, MW= 1270.6550) with 500 fmoles of sample.Internal instrument calibration was performed using the main singly charged matrix fragment at m/z 137.0239 (from DHB) and singly charged ions at m/z 748.4352 and 1271.6630) of the above mentioned hexapeptides.The main peak obtained is that of the mono-sodium or potassium adduct.
IR spectra were measured with a Perkin-Elmer 881 instrument as CHCl 3 solutions.Melting points were measured on a Büchi 535 apparatus and are uncorrected.TLC analyses were carried out on silica gel plates and developed with U.V. or with molybdate reagent (21 g (NH 4 )MoO 4 •4H 2 O, 1 g Ce(SO 4 ) 2 , 469 ml H 2 O, 31 ml H 2 SO 4 ).RF values were measured after an elution of 7-9 cm.Chromatography was carried out on 220-400 mesh silica gel using the "flash" methodology.Petroleum ether (40-60°C) is abbreviated as PE.In extractive work-up, aqueous solutions were always re-extracted thrice with the appropriate organic solvent.Organic extracts were washed with brine, dried over Na 2 SO 4 and filtered before evaporation of the solvent under reduced pressure.Dry solvents were purchased from Fluka, with the exception of THF, which was freshly distilled from K/benzophenone.All reactions employing dry solvents were carried out under a nitrogen (or argon when specified) atmosphere.The purity of all compounds was established by TLC, 1 H-and 13 C-NMR, and by GC-MS or HPLC or HPLC-MS.
(4-Methoxybenzyl)oxyacetic acid (12).NaH, (60% in mineral oil; 10.99 g, 275 mmol) was placed under nitrogen in a flask equipped with internal thermometer and dropping funnel.Dry DMF (200 ml) and dry THF (100 ml) were added and the resulting suspension cooled in ice.A solution of 4-methoxybenzyl alcohol (15.82 g, 114.49mmol) in dry THF (20 ml) was added through the dropping funnel during ca. 10 min, keeping the internal temperature < 7-8°C.The ice bath was removed and the suspension stirred for further 15 min, cooled to below 4°C, and a solution of bromoacetic acid (17.50 g, 125.94 mmol) in dry THF (20 ml) added during 50 min, keeping the internal temperature < 15°C.After stirring at R.T. for 20 h, the reaction was quenched with H 2 O (250 ml), stirred for 20 min, diluted with H 2 O (250 ml) and washed with PE/Et 2 O 1:1 (150 ml).The aqueous phase was acidified with 4 M HCl (40 ml) to pH 1, and saturated with NaCl.Extraction with AcOEt (1 x 100 ml), AcOEt/EtOH 9:1 (1 x 100 ml) and AcOEt/EtOH 8:2 (4 x 80 ml), followed by evaporation at 20 mbar, and then at 0.1 mbar, gave an oil (25.80 g).This was taken up in Et 2 O and treated with NaOH (4.94 g, 122.5 mmol in H 2 O, 150 ml).The layers were separated and the aqueous one washed twice with AcOEt (70+50 ml).The combined organic layers were extracted again with satd aq.NaHCO 3 (50 ml).The united aqueous layers were acidified with conc.HCl (18 ml) to pH 1 and extracted with AcOEt (4 x 80 ml).The organic layer gave upon evaporation, and stripping at 0.1 mbar, a yellow-brown solid (18.458

(3R*,4S*)-4-(Ethyn-1-yl)-3-(2-hydroxyethyl)-3-(4-methoxybenzyloxy)-1-(4-methoxyphenyl) -2-azetidinone (24).
A solution of 23 (1.545 g, 4.093 mmol) in dry CH 2 Cl 2 (40 ml) was treated with dry MeOH (20 ml) and Sudan III (Solvent Red 23) dye (12 mg).The red solution was cooled to -78°C, and ozonized for about 10 min (at maximum power and at a flow of 90 l/h), until the red color started to fade.Ozone production was interrupted, the oxygen flow was substituted by a nitrogen flow, and the solution was treated with 0.5 ml of Me 2 S + 0.5 ml of cyclohexene in 2 ml of CH 2 Cl 2 .All these operations were carried out as quickly as possible.After 5 min, Et 3 N (500 µL), and NaBH 4 (774 mg, 20.47 mmol) were added and the apparatus put under a static nitrogen atmosphere.The temperature was allowed to rise slowly to -10°C in 3 h and 30 min, and the solution stirred at this temperature until conversion of the intermediate aldehyde to 24 was complete by tlc (30 min).In some instances it was necessary to add further NaBH 4 in order to reach completion.It is not recommended to allow the temperature to rise above -10°C in order to avoid formation of lactone 26 deriving from intramolecular lactam opening by the alcohol.The mixture was finally quenched by pouring into an Erlenmeyer flask containing 70 ml of 5% (NH 4 )H 2 PO 4 and 20 ml of 1 M HCl (caution: vigorous gas evolution!).Extraction with Et 2 O (3 times), washing with saturated NaCl, and evaporation afforded a crude product that was chromatographed at once through silica gel (75 g) eluting with PE/AcOEt 1:1 containing 1% of 96% EtOH.Pure 24 was obtained as a slightly yellow oil (1.343 g, 86%).
of this product at the dry state, it was conserved in solution and it was not possible to perform elemental analysis.The same applies for the intermediates for the synthesis of 28 from 27. R f 0.35 (PE/AcOEt 50:50);  27 (301 mg, 1.076 mmol) in DMSO (3 ml) was treated with a solution of alcohol 28 (454 mg, 814.5 mmol) in THF (7 ml).The solution was stirred for 21 h at 18-20°C.After a few hours a white precipitate formed.The suspension was diluted with Et 2 O (15 ml) and filtered, washing with Et 2 O.The filtrate was washed with H 2 O and saturated NaCl/H 2 O 1:1.After drying and evaporation, the residue was taken up with toluene and evaporated again.Chromatography on 54 g of silica (PE/AcOEt 60:40 to 55:45) gave a yellow oil, which tends to darken.It was azeotroped once with CH 2 Cl 2 /toluene and twice with benzene.The resulting dark foam was thoroughly dried at 0.05 mbar overnight.The final weight of the resulting aldehyde 29 was 431 mg (95%).
It was taken up in dry THF (10 ml) under Ar.Meanwhile, NiCl 2 (15.8 mg, 122 µmol) and good quality CrCl 2 (it should be off-white; green samples are not well suited for this reaction)(640 mg, 5.207 mmol) were suspended under Ar in dry THF (25 ml).The aldehyde solution was slowly added, with magnetic stirring, to the CrCl 2 suspension, during 1 h at R.T.After further stirring for 3 h, the reaction was judged complete by tlc.The mixture was poured into saturated aqueous NH 4 Cl (30 ml) and distilled water (50 ml).After dilution with Et 2 O (50 ml), the resulting biphasic system was stirred for 40 min at R.T. Between the dark green aqueous layer and the yellow organic layer, a colloidal brown precipitate was evident.It is not easily dissolved in Et 2 O, AcOEt, H 2 O or CH 2 Cl 2 .It is partially soluble in acetone, but it gives, at tlc, only a very polar U.V. detectable spot (that does not elute even with pure AcOEt).The phases were separated and the aqueous one re-extracted with AcOEt and CH 2 Cl 2 .The organic layer was washed with saturated NaHCO 3 (40 ml) + saturated NH 4 Cl (10 ml) (reextracting the aqueous layer with CH 2 Cl 2 ).
After drying and evaporation, the crude product (that showed three spots in tlc, called A (R f 0.61), B (R f 0.55) and C (R f 0.45) was chromatographed through 50 g of silica with CH 2 Cl 2 /AcOEt 92:8.This chromatography gave a fraction containing pure A (24.2 mg), a fraction containing A+B (1:1 ratio) (27.6 mg) and a fraction containing C (37.1 mg), contaminated by a little amount of A and B. The overall yield of A + B + C is therefore 26.7%.A and B are both white solids, that dissolves with some difficulty in CH 2 Cl 2 or CHCl 3 and are completely insoluble in Et 2 O or AcOEt.They seem to be stable at the dry state.On these fractions, HPLC-MS were carried out with an Agilent 1100 LC/MSD Trap SL instrument (electrospray ion trap analysis) with a C18 reverse phase Polarity column (Waters Corporation, MA, USA).In all cases, before introducing the eluent in the MS, a detection at 230-260 nm was performed using a diode array detector integrated in the system.The MS electrospray ion source parameters were set to maximize, from time to time, the interesting m/z ratios.The separation was performed in linear gradient from 100% H 2 O to 100% CH 3 CN in 60 min The column was maintained at 30°C and the flow was 350 µl/min A T-union was used for post-column infusion (30 µl/min) of aqueous NH 3 (1.3M) in order to optimize ionization.On HPLC-MS, A gives a single peak (48.14 min) with masses of (in order of increasing intensity) 859.A gives in the NMR a single set of signals, therefore compatible with a symmetric cyclodimer.On the other hand, B (the spectrum was deducted from the one of A/B mixture by subtracting the signals of A) gives two sets of signals (B1 and B2), in a 1:1 ratio.The ratio is the same in the head or tail fractions containing this spot.Since also in HPLC B gives a single peak, we believe that it is given by one of the unsymmetric stereoisomer of cyclodimer 31. A. 1