Dehydroamino acid derivatives from D-arabinose and L-serine

Robert W. Armstrong, Andrew P. Combs, Paul A. Tempest, S. David Brown, and Thomas A. Keating. Accounts of Chemical Research 1996 29 (3), 123-131...
0 downloads 0 Views 2MB Size
J. Org. Chem. 1993,58,7848-7859

7848

Dehydroamino Acid Derivatives from D-Arabinose and L-Serine: Synthesis of Models for the Azinomycin Antitumor Antibiotics' Edmund J. Moran, John E. Tellew, Zuchun Zhao, and Robert W. Armstrong' Department of Chemistry and Biochemistry, University of California at Los Angeles, Los Angeles, California 90024

Received November 19, 1992 (Revised Manuscript Received September 14, 1993.)

Synthesis of aldehydes 17 from D-arabinose and 31 from L-serine provided key precursors for the generation of highly functionalized dehydroamino acid derivatives upon condensation with glycyl phosphonates. Subsequent bromination and intramolecular additiodelimination afforded the azabicyclo[3.1.0lhex-2-ylidenering system postulated to exist in the azinomycinantitumor antibiotics. Introduction The antitumor antibiotics azinomycin A (1) and B (2) were recently isolated from the fermentation broth of Streptomyces griseofuscus.2 An in vitro assay has established that azinomycin B induces interstrand crosslinks in duplexed DNA between G and purine residues two base-pairs r e m ~ v e d .Azinomycin ~ B also monoalkylates single-stranded DNA and duplex DNA exclusively at G residues. Piperidine treatment of monoalkylation intermediates results in depurination and strand scission as has been observed for adducb of DNA with synthetic aziridine reagents.4 We have postulated that the bicyclic and vinylic aziridine proposed to exist in the natural products is responsible for the monoalkylation of G residues. In the case of duplex DNA, a second alkylation of a purine residue two bases removed in the 5' direction results in cross-link formation. In the context of our work on the mechanism of action of the azinomycins, we have been engaged in the total synthesis of these compounds and of selected analogs. We recently reported in communication form the syntheses of l-azabicyclo[3.1.01 hex2-ylidene dehydroamino acid derivatives 362, 412, and 41E.6*6These are the first synthetic compounds to contain the intact vinylic bicyclic aziridine moiety proposed for the azinomycin antibiotics. These compounds were obtained via Horner-Emmons condensation of a glycine phosphonate with aldehyde 17.6 Described herein are two syntheses of aldehydes of the type represented by structure 5,' starting from D-arabinose and L-serine. Full details on the conversion of the dehydroamino acid derivatives to (2)and (E)-azabicyclo[3.1.0lhex-2-ylidenes are also provided.

* Abstract published in Advance ACS Abstracts, November 1,1993.

(1) Taken in part from the Ph.D. thesis of E. J. Moran. (2) Nagaoka, K.; Mataumoto, M.; Oono, J.; Yokoi, K.; Ishizaki, S.; Nakashima, T. J. Antibiot. 1986, 39, 1527. Yokoi, K.; Nagaoka, K.; Nakashima, T. Chem.Pharm. Bull. 1986,34,4554. I s h d , S.; Ohtauka, M.; Irinoda, K.; Kukita, K.; Nagaoka, K.; Nakashima, T. J. Antibiot. 1987,40,60. Hata, T.; Koga, F.; Sano, Y.; Kanamori,K.; Mataumae,A.; Sugawara,R.; Shima, T.; Ito, S.;Tomizawa,S.J. Antibiot., Ser. A (Tokyo) 1954, 7, 107. (3) Armstrong,R. W.; Salvati, M. E.; Nguyen, M. J. Am. Chem. SOC. 1992,114,3144. (4) Mattes, W. B.; Hartley, J. A.; Kohn, K. W. Nucl. Acids Res. 1986, 14, 2971. (5) Armstrong,R. W.; Moran, E. J. J . Am. Chem. SOC.1992,114,371. (6) Armstrong, R. W.; Tellew, J. E.; Moran, E. J. J. Org. Chem. 1992, 57, 2208. (7) For other syntheses of aldehydes or related precursor8represented by structure 5, see: Miller, 5.C. Ph.D. Thesis, University of Rochester, 1991. Coleman, R. S.; Dong, Y.; Carpenter,A. J. J. Org. Chern. 1992,57, 3732.

H

O

RY H

-

Results and Discussion Our retrosynthetic strategy for the azinomycin skeleton is outlined in eq 1. Disconnectionof the labile l-azabicyclo[3.1.0] hex-2-ylidene ring system and separation of the upper and lower halves generates key intermediate target compounds 4 and 5. Aldehyde 5 is linear, bears heteroatom substitution on each of its five carbons, and contains three contiguous stereocenters. These features suggested that it could be synthesized through transformation of an appropriate sugar. The relative configuration of D-arabinose (6)matched that assigned to the natural product with the exception of (2-4, which could be generated via a Walden inversion involving a nitrogen nucleophile (eq 2). 0

0

D-Arabinose was converted to ita l,l-bis(ethylthio)-4,5isopropylidene derivative 7 as previously described8 (Scheme I). Treatment of 7 with p-methoxybenzyl chloride and NaH in DMF resulted in formation of bisprotected product 8. The discovery that a mesyl or tosyl (8) Scarlato, G. R. Thesis, University of California,Los Angeles, 1990.

0022-3263/9311958-7848$04.00/0 0 1993 American Chemical Society

MpM

J. Org. Chem., Vol. 58, No.27,1993 7849

Models for the Azinomycin Antitumor Antibiotics

Scheme I1

Scheme I SCH2CH3

I

14

mrnTrCI, pyr. EtsN, -78'-OO C 6*%

I

n A F , w Fc rl

NmmTr

MPMO

82%

I

RO

2. NaBH4, EtOH

0

MPMO

74%

CICOCOCI, DMSO EtaN, CH2C12. -78' C MPMO

PPTS, MeOH,

MoCI. pyr. CH2C12

c

50'C

66%

NaN3, DMF, 50'C

MPM I

6% from 11

OH (Ph)3P,toluene c

I

4@C, 76%

group in the C-4 or C-5 position was incompatible with a C-1 thioacetal derivative convinced us that, before activation of either of these oxygens for nucleophilic displacement, an alternative protecting group at c-1would have to be introduced. Thus, NBS deprotectiong of 8 followed immediately by reduction of the crude aldehyde with NaBH4 in EtOH led to alcohol 9. Silylation under standard conditions afforded 10 followed by removal of the isopropylidene protecting group with mild acid to generate diol 11. Azide 13 was obtained via selective mesylation at C-5 and subsequent treatment with sodium azide. Transformation to the aziridine 14 was effected by concomitant reduction and cyclization via the aza-ylide intermediate under Staudinger conditions."J Consideration of the synthetic strategy at this point required a judicious choice of a protecting group for the aziridinein 14. Initial studies with carboethoxy and FMOC protecting groups were discouraging, since the resulting carbamate aziridines spontaneously underwent intramolecular aziridineopening to form tetrahydrofuran products upon fluoride deprotection of the primary alcohol. A report by Nakajimall on the stability of an aziridine-2carboxylate to trifluoroacetic acid (TFA)-promoted Ntrityl deprotection prompted us to prepare the monomethoxytrityl protected aziridino aldehyde 17 from 14 (Scheme 11). In contrast to the carbamate derivatives, alcohol 16 proved to be sufficientlystable when generated through deprotection of 15 with TBAF in THF followed, by chromatography. Prompt oxidation to aldehyde 17 prevented any undesirable cyclizationproducts. Although this aldehyde could not be chromatographed on silica gel, lH NMR analysis of the crude reaction mixture indicated that the product was of high purity. Subsequent experimenta using 17 convinced us that the monomethoxytrityl-protected aziridine was remarkably stable to a variety of conditions (vide infra). However, we sought a synthesis of aldehydes of type 5 which would (9) Corey, E. J.; Erickson, B. W. J . Org. Chem. 1971,96, 3663.

(10) Staudurgsr, H.; Meyer, M. Helu. Chem. Acta 1919,2,619. Ittah,

Y.;Sasson, Y.; hahak, I.; Twoon, S.; Blum, J. J. Org. Chem. 1978,43,

4271. (11) N+ajima, K.; Takai, F.;Tanaka, T.; Okawa, K. Bull. Chem. SOC. Jpn. 1978,61, 1677.

MPMO

provide greater latitude in the selection of blocking groups for the C-2 and C-3hydroxyls and which specificallywould allow for silyl protecting groups in these positions. These factors led us to the synthesis of 31 from L-serine (Scheme 111). Serine methyl ester (18)was converted to alcohol 21 without purification of intermediates in excellent overall yield. This material proved to be stable to freezer storage (>6 months) without racemization. Oxidation and homologation provided ester 23 which was subsequently reduced and blocked with dimethoxytrityl chloride, affording 25. Sharpless asymmetric hydroxylation12provided an inseparable mixture of diastereomers 26 and 27 in a 4.1:l ratio, respectively.l3 Subsequent protection with TESCl (or TBSOTF4) afforded 28 containing a trace of the minor diastereomer. Acid treatment followed by selective reprotection afforded alcohol 30 as a single diastereomer which was oxidized to 31 using Swern conditions. Like 17, aldehyde 31 was unstable to chromatography, but could be used as crude reaction mixture in subsequent condensation reactions. Our first attempts at condensation of 17 with a glycine anion equivalent were via the acidicErlenmeyer reaction.l6 Before attempting this reaction,we confiied the stability of the (monomethoxytrity1)aziridineto the reaction conditions by recovering 15 in high yield following treatment with Pb(0Ac)dAczO in THF at 60 OC for approximately 12 h.16 However, condensation of 17 with hippuric acid using the Erlenmeyer conditions followed by addition of primary amines provided only low yields of dehydroamino acid (DHAA) products. Instead of pursuing variations on the Erlenmeyer synthesis, we turned instead to the more versatile glycine phosphonate approach for a DHAA synthesis. Both a-phosphono esters and phosphono amides have demonstrated usefulness in natural product synthesis." Attempted condensation of ethyl N-benzoyl-ddiethy1phosphono)glycinate (32)18with aldehyde 17 (Scheme IV) using t-BuOK in CHzCl2 at low temperature led to extensive decomposition of 17.19 Use of LDA in THF, however,provided good yields of the desired dehydroamino ester products 33E and 332. Ester 332 could be suc(12) Wai, J. S. M.; Marko, I.; Svendeen, J. S.;Finn,M. 0.;Jacobwn, E. N.; Sharpleas, K. B. J. Am. Chem. Soe. 1989,111,1123-1126.

(13) In the absence of the Sharplean chiral auxiliary, the wmylation yielded a 1:l mixture of diaetereomem. (14) This material WBB carried to the bis-TBS analogof compound 81. (16) Moran,E. J . ; h t r o n g , R .W. TetrahedronLett. 1991,32,9807. (16) Cativiela, C.; Melendez, E. Synthesis 1978,832. (17) Schmidt, U.;Lieberknecht, A,;Kazmaier, U.; Grieawr, H.; Jung, G.; Metzger, J. Synthesis 1991,49. For areviewshowcasingmme natural products synthesizedfrom thia method see: Schmidt, U.; Lieberknecht, A.; Wild, J. Synthesis 1988, 169. (18) Kober, R.; Stieglich, W. Leibigs Ann. Chem. 1983,699. (19) High 2 stereoselectivity is observed in the reaction of glycyl phoaphonate estera with aldehydes when potassium tert-butoxideis uaed aa a base in methylene chloride at low temperatures.

Mora et al.

7860 J. Org. Chem., Vol. 58, No,27,1993

U

(Et0)2POCH$02Me mmTr

W mmTr

-

LDA, THF, -1OOC

O

D

M

T

OsO*, O°C, NMO

r

N*ODMTr

,

DHODpClBz

25

/N

mmTr

mmTr

86%

?TES 26/1,

TESCI, imidazole CHG12

..

87%

A

1 "

mmTr

O

D

28

M

T

r

CF3-H.

I

64%

OTES

1

OH

30 R-mmTr

EtSH

CH2C12, -78OC to -2OOC

OTES

1

. ...

28

( M 7 ,4.4:l)

OTES

RM

OH

+ d*rre-r

0

II

CIcocOcI, -78OC DMS0,EtjN

~TES

~TES mmTr

31

mmTrCI, -78OC CHzCh, EtjN

cessfully hydrolyzed to the carboxylic acid, coupled with 1-amino-2-propanolvia DCC/HOBt activation, and oxidized to the 2-ketopropanamido derivative 342 without purification of intermediates. This was necessary because neither the acid nor the hydroxy amide intermediates withstood exposure to silica gel. Unfortunately, the E-carboxylic acid decomposed without conversion to its corresponding amide under the same conditions successfully used for the 2 isomer. Dehydroamino amide 342 was brominated using Brz in CH2C12 at -78 OC, followed by DABCO. A single @-bromo dehydroamino amide isomer (352) was isolated in 71% yield. However, attempts to observe an NOE between either amide proton and the allylic methylene of 352 were unsuccessful, leaving the olefin geometry undetermined. TCA deprotection of the mmTr group in CDCls led to a set of signals ('H NMR) d o d i e l d from the aziridine resonances in 352 and consistent with those of the azabicyclo[3.1.0] hex-2-ylidenering system of the natural product. FAB-MS analysis of the crude mixture, however, indicated that only a very small percentage of product gave rise to a mass corresponding to 586 [MHl+,expected for 362. Instead, two major signals corresponding to masses of 666 (with a monobromo isotope pattern) and 828 (with a monobromo-trichloro isotope pattern) were reported, in accord with structures 37 and 38. Therefore, the shift of the aziridine resonances in the lH NMR was not due to the vinylogous urea character of 362 but rather was due to protonation of 37 by TCA. Furthermore, TCA was adding to 37 resulting in 38. Indeed, treatment of 15 with TCA in CDC13 led to nucleophilic, opening of the

aziridine followed by acyl transfer to provide amide 39 (eq 3). To prevent acid hydrolysis, we repeated the experi-

L

MPMQ,,..,

TCA ,RCDCi3

CI3CCO"

MPMO

1s

s..,

Nmmlr

)

HO

*OMPM

(3)

39

menta with 362 but added triethylamine to the NMR tube following TCA deprotection. The aziridine resonances immediately shifted back upfield after triethylamine addition indicating deprotonation by the stronger base.20 The deprotected aziridine product was chromatographed on silica gel, redissolved in CDCla, and, after its lH NMR spectrum was recorded, cyclized to the azabicyclo[3.1.01 compound 362 in the presence of triethylamine at 50 "C over an 8-h period. Full characterization of 362 was facilitated by its stability to purification on a silica gel column. An lH NMR NOE differenceexperiment conclusivelyestablished the 2 stereochemistry of the product. Not only was an enhancement of the allylic methine H-1321observed upon irradiation of "-16, but irradiation of "-5 produced an enhancement of the exocyclic aziridine (H-10 exo). Consistent with the data on the natural product azine (20) Deyrup, J. A. In Heterocyclic Compounde; Hasener,A., Ed.;John Wiley and Sone, Inc.: New York, 1983, Vol. 42, Part 1, p 9. (21) The bromination of N-carbarnateblocked dehydronmino acid derivatives gives different ratioe of vinyl bromidw. Coleman, R. S.; Carpenter,A. J. J. Org. Chem. 1993,68,4452-4481.

J. Org. Chem., Vol. 68, No.27,1993 7881

Models for the Azinomycin Antitumor Antibiotics Scheme IV

n

-780-oo c 3.7:l UE

+ 17

MPW+

NmmTr

26'" nn:

332

0

2.2-propanol 1. LOH HOBt, ,THF, CHpCIz, amine, 50' C rt DCC ,

33 E

3. oxalyl chloride, DMSO, NEt3, CHpCIz, -78' C

0

Br2, -78O C

DABCO , CHpClz

NmmTr

71%

MPMO 342

5Q%

0

0 Br

OMPM

OMPM

Br

=/

37

38

HO

Table I. Comparison of 'H NMR Coupling Constants and Chemical Shifts for Azinomicyn B (2) and Synthetic l-Azabicyclo[3.l.0]hex-2-ylidenes 362,412, and 41E 20

H 13 12 11

1 0 , 1oendo

J (Hz) 4.0,O.F 4.0,5.6 5.6,5.4,4.0 5.4,o.g 4.0

6 5.50 4.65 3.39 2.75 2.29

36Zb

6 5.12 4.42 3.03 2.40 2.18

J (Hz) 1,lC 1,4.9 4.9,5.3,3.6 5.3,l 3.6

412b

4lEb

6

J (Hz)

6

5.05 4.41 3.00 2.39 2.14

1.5,1.SC 1.5,4.2 4.2,5.5,3.7 5.5,l.B 3.7

4.57 4.53 3.00 2.41 2.14

J (Hz) 4.9,lC 4.9,4.9 3.7,4.9,4.9 5.0,l 3.7

*

400-MHz. CDCh. 360 MHz, CDCln. The second value corresponds to long-range coupling between 13-H and 10-H, ae observed for azinomycin B and ail three synthetic bicyclic analogs.

mycin B, we also observed an NOE between the H-lOendo aziridine hydrogen and the allylic methine H-13. The contrast between the coupling constants in the 13.1.01 ring system in 362 and the azinomycins was apparent, however (Table I). Although there is excellent agreement between the data for proton resonances H-lO,,, H-IOendo, and H-11, those for H-12 and H-13 differ considerably. The substituents on these latter two positions are not the same and obviously could affect the conformation of the fiveatom ring. In an effort to evaluate the substrate specificity of brominations, we reacted dehydroamino ester 332 with either NBS or Br2. Treatment of 332 with Br2 in the presence of 2,6-lutidine in CH2C12 at -78 "C over 1.5 h followed by DABCO treatment led to an inseparable mixture of vinyl bromides 402and 40E(1.51, respectively).

The mixture was then subjected to TCA in CDCla. Deprotection of one isomer occurred much more rapidly than the other such that,after the addition of triethylamine and warming to 50 OC overnight, a single vinyl bromide 40E and a single azabicyclo[3.1.01 compound 412 were recovered. Assignment of the stereochemistry of 412was based on the close correlation of chemical shifts and coupling constants of the 13.1.01 ring protons with those of 362 (Table I). Compelling evidence for the stereochemistry of 40Ewas to follow. In contrast to brominations with Br2, the reaction of 332 with N-bromosuccinimidein CHzClz produced vinyl bromide 40E exclusively. When 40E was deprotected by TCA (in CDsCN) and treated with EtsN, it cyclized to a single [3.1.01 product 41E.The coupling constants for the ring protons of 41E were in contrast with those for 412 and 362. However,

7882 J. Org. Chem., Vol. 58,No.27, 1993

Moran et al.

Scheme V 0

33 z

Brp, 2.6-lutidine -7s0 c-00 c 77% 1:l.S En

OCHpCH:,

0 40E+40Z

nr

NBS, CHCl3, rt 76%, E only

0 1. TCA , CD3CN, rt

40Z+40E

0 OMPM

2. EtaN, SO0 C

I.TCA , CD3CN c

Ph ) (0 NHtOCH&H3

40E

/

MPMO

41E

they correlate well with those for the natural products 1 and 2, in particular for H-12 and H-13. The stereochemistry of 41E was firmly established by an observed NOE relationship between the amide hydrogen H-5 and the allylic methine H-13 in the [3.1.01 ring. Both 41E and 412 show the expected cross ring enhancements between the allylic methine and the endocyclic aziridine methylene confirming their bicyclic structure. In parallel with the natural products, they also exhibit a long-range five-bond coupling (J = 1-2 Hz) between H-13 and H-lo,,. All cyclizations of the vinyl bromides were completely stereospecific. The lack of NOE or chemical shift correlations for vinyl bromides 362,40E, and 402 leads to a reliance on the stereochemical outcome of their cyclizations for this assignment. The additiodelimination of the aziridine as a multistep process results in a carbanion intermediate upon aziridine addition followed by elimination of the @-halide.22If each isomer executes the same mode of facial addition, then E and Z give rise to intermediates which differ only by rotation about the CaC@ bond axis. The intermediate Michael adduct must make either a 60" rotation (retention) or a 120" rotation (inversion) about the a-@ bond before expulsion of the @-bromidecan occur (Scheme VI). There is a considerable barrier to the 120° rotation in 8-halo-substituted ethyl anion systems due to the loss in hyperconjugative stabilization of the a-carbanion by the halide upon rotation. The 60° rotation, in contrast, serves to maximize the stabilization, placing the interacting orbitals parallel for elimination. If the two isomers execute opposite modes of facial addition, then they generate diastereomeric intermediate carbanions, both of which must perform a 120" rotation before eliminating to give complete inversion of stereochemistry. Regardless of the mode of facial addition, the unfavorable 120° a+ bond rotation makes retention of starting olefin geometry highly likely. In the course of investigating the condensation of phosphonate 32 with the aldehyde 31, we inadvertently (22) A vinylic substitution occurring aa a single-step procees would produce Z and E [3.1.01 produds from Z and E vinyl bromides, respectively. See, Rappaport, Z.Acc. Chem. Res. 1981,14,7.

deprotected the aziridines during preparative thin-layer chromatography (silica gel) purification of a mixture of 42E and 422 (Scheme VII). This resulted in isolation of asingle diastereomer from each of the elution bands: 42E, the higher eluting band, produced the saturated l-azabicyclo[3.1.01 43, while 422, the lower eluting band, produced diastereomer 44. The stereochemical assignments for these bicyclic structures were established by NOE experiments. We had expected the NOE data to indicate that the products shared identical ring stereochemistry and we had assumed that protonation a to the carbonyl had occurred on opposite faces of the enolate intermediate. Surprisingly, the olefin geometry controlled the stereochemistry at the @-carbon. The observation of NOE enhancements between the @-hydrogenand the endo aziridine methylene in both 43 and 44 suggest that they exist in boat-like conformations. The diaxial relationship of the triethylsilyloxy substituents likely destabilizes the chair conformations. This is especially true in the case of 43 which has all substituents in anti relationships in the chair. If the transition states assume boat-like conformations, then the aziridine in 422 adds exclusively to the pseudoaxial rotamer, and in 42E exclusively to the pseudoequatorial rotamer. An intramolecular hydrogen bond is possible between the amide NH and the allylic silyl ether oxygen in 422 and not in 42E for the boat conformations,potentially explaining the opposite rotamer selectivity. In both cases, it appears that the a-anion is protonated stereospecifically. This result is difficult to justify by invoking steric interactions with an external acid source, but could arise from an intramolecular synprotonation by the aziridine following Michael addition. The stereochemistry of the a-amino acid center has yet to be established. These saturated [3.1.01 derivatives of glycine bear a resemblance to the natural product ficel10mycin.~~ This method of preparation offers a synthetic approach to the bicyclic ring in this antibiotic. Conclusion We have described the first synthesis of the (E)- and (Z)-l-azabicycl0[3.1.01hexd-ylidene derivatives of N-benzoyl dehydroamino esters and amides. The oxidation of the @-carbonof dehydroamino esters bearing acid-sensitive functionality was successfully implemented using both bromine and NBS, resulting in moderate to high selectivity information of the vinyl bromides. However, this reaction appears to be highly substrate specific. The resulting vinyl bromides undergo conversion to the l-azabicyclo[3.1.01hex-2-ylidenederivatives with retention of stereochemistry with respect to the starting olefin geometry. The dehydroamino ester and amide precursors were synthesized via condensation of glycyl phosphonatas with highly functionalizedaldehydes of type 6. These aldehydes were synthesized via two independent routes, proceeding from D-arabinose and L-serine. Experimental Section General Information. 1H and 1% NMR spectra were recorded at the field strength specified in megahertz. Chemical shifta are reported in ppm with CHCls, acetone, or DMSO as internal standards. E t absorption frequencies are reported in cm-'.Tetrahydrofuran, diethyl ether, and toluene solventswere distilled from sodium benzophenone ketyl under argon. Meth(23)Kuo, M.S.;Yurek, D.A.; M U ,S.A. J. Antibiot. 1989,42,357.

J. Org. Chem., Vol. 58, No. 27, 1993 7853

Models for the Azinomycin Antitumor Antibiotics Scheme VI sYn

anti

0

0

vinyl bromides

412

[3.1.O] produds

0

0

-

41E

Scheme VI1

&

=so,,., TESO

LDA, THF 42 E

NmmTr ' 3 2

31

OTES

+

L

J

i

1

1

422

'OTES

&TES

0 II

PhOOCNH'

ylene chloride was distilled from P205. Dimethylformamide and dimethyl sulfoxide were distilled from barium oxide and CaH2, respectively,under N2 and stored over 4-A molecular sieves. E t N and 2,6-lutidine were passed through a column of basic alumina immediately prior to use. All other reagents were used aa supplied or synthesized according to literature procedures. All reactions were carried out under an inert N2 atmosphere. Unless otherwise noted, flash chromatography was performed on Merck silica gel 60 (230-400 mesh) using various gradients of hexanes/ethyl acetate as eluants. Acid-sensitive compounds were chromatographed using EbN-deactivated silica. Small-scale separations (