ORGANIC MASS SPECTROMETRY, VOL. 29, 445-453 (1994)

Structural Analysis of Rapamycin and Related Compounds Using Liquid Secondary Ion Mass Spectrometry

+ Li]' Ions Generated by

J. P. Kiplinger* and M. A. Guadliana Pfizer Inc., Pfizer Central Research, Groton, Connecticut 06340, USA

Cationization of the macrocyclic immunosuppressant rapamycin with lithium ion upon liquid secondary ion mass spectrometric ionization yields a number of fragment ions, which are observable in the full-scan spectrum. These are clearly assigned using B/E linked scanning (fragment ion scanning), B2/E linked scanning (precursor ion scanning) and peak matching for accurate mass measurement. Many of the fragments are produced by processes that open the macrocyclic ring, and it is possible to observe several different pieces of the molecule as fragment ions. The diversity of fragments produced facilitates the elucidation of new raparnycin-like structures through mass spectrometry. Three structurally modified rapamycin analogues have been examined by this technique, and the modifications to the molecule may be located based on the nominal masses of their fragments.

INTRODUCTION

Rapamycin (1) is an antitumor, antifungal, and immu- nosuppressive agent isolated from cultures of Strep- tomyces hygroscopicus.' Its potent immunosuppressive activity and its structural homology with the well studied immunosuppressant FK-506' makes it an attractive lead structure for pharmaceutical companies. Since streptomyces cultures offer a potentially rich source of new compounds related to rapamycin, the ability to structurally characterize rapamycin analogues in small amounts is of potential value in a search for new drug candidates. It is therefore desirable to develop a mass spectrometric method for rapid screening of novel rapamycin-like molecules which gives useful structural information.

Unfortunately, rapamycin and its analogues present difficult mass spectrometric problems, producing little

OH

42

12

Raparnycin (1)

structural information by positive-ion chemical ioniza- tion (CI), electron impact (EI), fast atom bombardment (FAB) and ionspray method^.^ The FAB or ionspray mass spectra are often dominated by sodium-cationized molecules, even when efforts are made strictly to limit sodium salt contamination of the sample, solvents and liquid matrix. These data can offer molecular mass information, but sodiated ions usually produce few structurally useful fragment^.^ In our laboratory, attempts to ionize rapamycin using electrospray pro- duced only sodiated molecules, despite attempts to remove inorganic salts from the compound. With EI ionization, low-mass fragment ions dominate the spec- trum, and CI mass spectra are characterized by exten- sive fragmentation with the dominant fragments assigned to the loss of macrocyclic ring substituents such as OH and OMe. Negative-ion FAB produces a few structurally significant fragments in addition to molecular mass information, and is therefore one of the better choices for the analysis of rapamycin analogues. The negative-ion FAB mass spectrum of rapamycin was reported and interpreted previ~usly.~

As noted above, rapamycin produces little proto- nated molecule in liquid secondary ion mass spectrom- etry (LSIMS), but cationizes very well with sodium ions when present in the matrix. In our laboratory, alkali metal salts are often added to LSIMS matrices in order to cationize samples with alkali metal ions. Through such attempts to determine quickly molecular masses of rapamycin analogues, we have found that lithium ions also efficiently cationize rapamycin and related mol- ecules, producing abundant [M + Li]+ ions. Unlike sodium-cationized molecules, however, these [M + Li]+ ions produce abundant fragment ions, par- ticularly when they are activated collisionally. This is not an unexpected result; it has been reported pre- viously that lithiated ions fragment more extensively than their sodiated counterpart^.'-^ This phenomenon has been explained using bond energy argument^.^.^

CCC 0030-493X/94/080445-09 0 1994 by John Wiley & Sons, Ltd.

Received I 1 January I994 Revised 21 March I994

Accepted 21 March 1994

446 J. P. KIPLINGER AND M. A. GUADLIANA

2

&.-

Lithium cationization has been used previously to obtain useful structural information from fatty acids,' alcohols* and fatty acid esters," from disaccharides' ' and oligosaccharides'2 and from oligopeptides.' 3.14

More recently, the general utility of this method has been seen in its application to the structural analysis of a wider variety of compounds, including 1,2-diols,15 tri- terpenoids,' Diels-Alder adducts of benzoquinones,' acetogenins' (long-chain epoxyalkane natural products), various glycosides'' and nucleosides.20 Many of these studies suggest fragmentation mecha- nisms that proceed without the participation of the lithium ion, i.e. 'charge-rem~te' .~.~ ' Charge-remote frag- mentations are often touted as being more structurally informative than charge- or radical-site initiated reac- tions, which often involve extensive isomerizations or movement of the charge site. In this investigation of rapamycin and its analogues, it was hoped that the pro- pensity of lithium-cationized molecules to undergo charge-remote fragmentation would allow the observa- tion of significant ring cleavages, thus providing a struc- tural probe for analogues modified at different positions on the macrocyclic ring.

It is not our intention to present a complete analysis of the fragmentation of rapamycin. Rather, the goal of this work was to determine whether the observed frag- ments include all the sites on the molecule where struc- tural modifications are likely to be found. Rapamycin analogues derived from fermentation, such as 2-4, tend to differ from the original compound by small changes at only a few locations. In 2-4, these locations are the C(16) and C(27) ring carbons. If the C(16) and C(27) sites are included in different fragment ions, substitution of an OH for the OCH, group at one of these positions would shift the nominal mass of one of these fragments by 14 u, allowing easy structural assignment. If a great enough diversity exists in the fragmentation scheme, individual modifications in any rapamycin analogue may be quickly identified with a BJE scan of the [M + Li]+ ion.

EXPERIMENTAL

Rapamycin (1) and the rapamycin-like compounds 2, 3 and 4 were isolated from the rapamycin-producing

b-

3 4

culture NRRL 5941, also known as ATCC 29253. A 1 pI volume of a methanol solution of 10 mg cm-, of each of these was applied to the LSIMS probe tip atop 1 pl of matrix. 6LiCI was prepared from 6Li2C0, (Isotec, Miamisburg, OH, USA) by reaction with 1 M HCl, added dropwise to a strongly acidic pH. The resulting solution was lyophilized to yield solid 6LiCI, which was used without further purification. The liquid matrix chosen for most of the experiments was 'magic bullet,' a 5 : 1 mixture of dithiothreitol and dithioerythritol. 6LiCl and 7LiCl/magic bullet solutions were prepared to a concentration of 0.1 M by dissolving 0.1 mmol of LiCl and 1.0 cm3 of warm matrix.

LSIMS was carried out using a Kratos Concept-1S spectrometer (Kratos Analytical, Manchester, UK). The Cs' ion beam was accelerated to 1 1 k V above source potential and the desorbed ions were accelerated to 8 kV. Typical operating parameters included a resolving power of 1000 and a scan rate of 10 s per decade; high- resolution measurements were carried out at a resolving power of 10000-12000 for peak matching against poly(ethy1ene glycol) standards. Collisional activation was by collision with argon, admitted to a collision cell immediately after the source region at a pressure suffi- cient to reduce the precursor ion intensity by 50%. The linked-scan spectra represent an accumulation of 8-10 scans.

RESULTS

Figure 1 shows the LSI mass spectrum of rapamycin from a magic bullet matrix solution containing lithium chloride. The spectrum clearly shows a number of frag- ment ions, presumably produced from the rapamycin [M + Li]' ion at m/z 920. A linked-scan spectrum at constant BJE of m/z 920 (Fig. 2) contains a number of the fragment ions seen in the full-scan spectrum.

Owing to the amount of chemical noise inherent in LSI mass spectra, it is not possible to determine whether an ion seen in the full-scan LSI mass spectrum is a fragment of the [M + Li]+ ion without the confir- mation provided by the linked fragment ion scan. At present, the Kratos data system used to generate the scan function and to control the B and E fields allows the constant BJE ratio to be scanned down through the

STRUCTURAL ANALYSIS OF RAPAMYCIN AND RELATED COMPOUNDS 447

203

v

c.

c C - 0, 150 200 250 300 350 400 450 500 550 .-

920 1 586

A I , 1

81 8

800 900 1000 mT 0 - I &

600 700 rn/z

Figure 1. Full-scan LSI mass spectrum of rapamycin from a 0.1 M 'LiCI-magic bullet matrix.

fragment ion spectrum to about 50% of the precursor mlz value. The linked-scan spectra, therefore, may only be used to confirm precursor-product relationships for fragments in the upper half of the fragment ion spec- trum. Because of this, we do not feel entirely confident in assigning structures to most of the lower mass frag- ment ions.

Before attempting to assign structures to the major fragment ions observed in the full-scan and linked-scan spectra, it was considered important to exclude possible

contributions to the fragment ion spectrum from ions other than the desired precursor [M + Li]' ions. Such interferences result from the large precursor ion 'accep- tance window' inherent in the BIE scan. By varying the accelerating voltage V slightly, the range of this 'accep- tance window' may be observed on the Kratos spec- trometer, and it is nearly 9 u wide at 920 u (the mlz for the rapamycin [M + Li]' ion). It is possible, therefore, that fragments derived from both [M + H I + and [M + 2Li - HI' ions may be observed in the B / E scan

t-- x4

100

h

8 v

x v) != a, K a, > m a,

c .-

c .- .- - - a

586

604

818

I " " I " ' ' I " ' , I " " 1 ' ' ' ' I " ' ~

904

1 450 500 550 600 650 700 750 800 850 900

m/z Figure 2. B/E linked-scan spectrum (fragment ion scan) of them/z 920 [M + 'Li]+ ion of rapamycin.

448 J. P. KIPLINGER AND M. A. GUADLIANA

Table 1. Measured ma- of fragment ions present in the lithium-cationized rapamycin LSI mass spectrum (Fig. 1)

Fragment m/z Measured mass Formula A (ppm)

81 8 726 626 604 593 591 586 554 443 329

81 8.5047 726.4405 626.3480 604.3683 593.401 7 591.4249 586.3538 554.3252 443.301 1 329.2297

C46H69N011Li

C40H58N0 1 OLi2

C33H49N01 OLi

'32 H46

'35H540 ,L i

C3SH5606Li

C32H46N08Li2

'31 H42N07Li2

C25H4006Li

'1 9H3004Li

+1.99 +3.33 -5.83 +5.60 -2.1 2 +2.04 -0.94 -5.30 +5.88 -2.17

of the [M + Li]' ion. A straightforward labeling experiment using 6LiCI as the matrix additive was per- formed in order to differentiate these fragments, and several fragment ions were identified as dilithiated by their mass shift of 2 u. These ions are presumably derived from an [M + 2Li - H] + precursor ion, although this result was surprising in that such a pre- cursor is not observable (signal-to-background ratio z 2 : 1) in the full-scan spectrum. Labeling with 6Li+ has been used previously to aid in mechanistic studies of ion f r a g m e n t a t i ~ n , ' ~ * ~ ~ and we reported recently on its utility in quickly 'deconvolving' these linked-scan spectra of raparny~in.'~

N'

Q 0 0'

/ ' /

Q

- 512

& 0 o o ' OH I

OH Li'

0 0' 0

/ "

Li' "G1,

OH >fl: LI' \ 0

Scheme 1.

STRUCTURAL ANALYSIS OF RAPAMYCIN AND RELATED COMPOUNDS

Table 2. Masses of fragment ions produced by compounds 1-4 in B/E linked-scan experiments'

Aapamycin (1) C(27)-demethoxy (2) C ( 1 6 ) . C(27)-OH ( 3 ) C(Z7)-OH (4)

9 2 0 ( M + L i ] + 8 9 0 [ M + L i ] + 8 9 2 [ M + L i ] + 9 0 6 [ M + L i ] + 904 874 876 890 888 858 860 874 81 8 788 790 804 726 696 626 626 626 604 574 576 590 598 570 584 593 563 579 579 591 561 577 577 586 556 558 572 580 550 552 572 542 558 559 529 554 524 540 458 458 458 443 41 3 429 429 329 329 329

a Ions which are noticeably lower in abundance by comparison with analo- gous spectra are shown in italics.

449

Table 1 lists the more abundant fragment ions obser- vable in both the full-scan and B / E linked-scan spectra of lithiated rapamycin. The exception. is the ion at m/z 329, which was unobservable in the linked scan owing to the scanning restriction mentioned above, but which was verified as a fragment using a B 2 / E precursor ion scan. The ions listed in Table 1 were examined at high resolution in order to help determine their elemental compositions, and these are listed with deviations from the calculated exact masses. The dilithiated fragments were assumed to be derived from an [M - H + 2Li]+ precursor ion, but we should not similarly conclude that the monolithiated fragments are derived from a mono- lithiated precursor. The precursor ion for these was determined experimentally where necessary, by linked scanning at constant B2/E. No non-lithiated fragment ions were observed in this mass range, as determined by the 6Li' labeling experiment.

Since the appearance of dilithiated fragments in our product ion spectra is an artifact associated with the instrument and the technique, we have not made an effort to determine their structures. The fragmentation of the [M + Li]' ion of rapamycin via a fairly diverse series of processes is illustrated in Scheme 1. The frag- ment ion structures shown are rationalized below based on reasonable mechanistic assumptions, on precursor ion linked scanning and on the elemental compositions of the monolithiated ions.

Fragment ions observed in the full-scan and B / E linked-scan spectra of the rapamycin analogues 2 , 3 and 4 are listed in Table 2. The monolithiated fragments produced from these compounds are assumed to be formed through reactions equivalent to those shown in Scheme 1. Their nominal masses are listed alongside the nominal masses of the monolithiated rapamycin frag- ments for easy comparison. The observed fragmentation patterns of the analogues are in all cases consistent with predictions that may be made using Scheme 1.

DISCUSSION

The losses of 16 and 32 u seen in the B / E spectra of 1-4 have not been examined in detail, and are assumed to be due to facile losses from one or more of the ring functional groups. This region of the fragment ion spec- trum is complicated by a large amount of chemical noise, possibly due to the competitive losses of small neutral molecules. It does seem clear from the spectra of 1-4 that the losses of 16 and 32 u are important pro- cesses in all but 3, which is the only compound modified at the C(16)-methoxy group. in 3, the [M + Li' - 161' and [M + Li' - 321' ions are lower yield fragmenta- tions, resulting in peaks less than three times the back- ground level. This may indicate that the loss of methane or of methanol involving the C( 16)-methoxy group is especially facile owing to the resulting conjugation with the triene. Further testing of this hypothesis using other C( 16)-modified analogues is required before concluding that these fragments may be used as indicators of spe- cific structures.

Gross2' states that 'any means of tightly localizing a charge at one site will be sufficient to allow charge- remote fragmentations.' Examples of macrocyclic anti- biotics cationized both with proton and with sodium ion undergoing charge-remote fragmentations exist in the 1 i t e r a t~ r -e .~~ Although we have not investigated the role of the lithium cation in any of the proposed mecha- nisms, in most cases we invoke pathways (for example, the loss of methane discussed above) which are promi- nent in the literature of charge-remote fragmenta- t i o n ~ . ' ~ . ~ ~ The fragment ion structures are supported by the accurate mass data, by the labeling and linked- scanning experiments and by the observation or non- observation of analogous processes in compounds 2-4. The proposed mechanisms are offered as the simplest

450 J. P. KIPLINGER AND M. A. GUADLIANA

route from the precursor ion to the fragment. Generally, eliminations and simple rearrangements are used to account for the formation of the fragment ions. Since these do not require the participation of the lithium cation, they are shown as though they take place remote from the charge site. Consistent with these pro- cesses is the fact that no open-shell (odd-electron) ions are observed in the series we have examined. Open-shell ions have been observed in decompositions of closed- shell metal cationized molecules, but are typically formed through energetically favored allylic cleavages of single bonds near the charge site, and are thus usually low-mass fragment^.^-'^*'^ As stated previously, we have not attempted to assign structures to low-mass fragments observed in the LSI mass spectra of these compounds.

Two prominent odd-mass fragment ions are observed in the spectra of rapamycin, at m/z 593 and 591. The lack of open-shell fragments in this mass range should allow the prediction that all even-mass fragments retain the single nitrogen atom, while the odd-mass ions do not. Accurate mass measurements indicate that these fragments have, as expected, lost the nitrogen atom. By comparing the fragment ions produced by 1 with those produced by 2-4, it is readily apparent that functional- ity at the C(27) position is retained in these fragments. Compound 3, however, shows equivalent fragments that are only 14 u lower in mass than the rapamycin frag- ments, indicating that the methyl on the C(16) methoxy is lost. Formula assignments show that the production of the m/z 593 fragment involves the loss of C16Hz5N0,, while m/z 591 follows the loss of Cl5HZ3NO7. However, precursor ion scans show that while the ion at m/z 591 is derived from the [M + Li]' ion at m/z 920, m/z 593 appears to be derived from dili- thiated ions at m/z 604, 727 and 926. Admittedly, the m/z 593 peak contains a contribution from the isotope peaks of the abundant m/z 591 fragment, and this makes interpretation of the poorly resolved B 2 / E scan difficult. The proposed structure for the m/z 591 frag- ment ion has been included in Scheme 1, but no attempt has been made to settle on the structure for m/z 593.

In producing the m/z 591 ion, elimination across the C(34)-C(33) bond seems a reasonable ring-opening reaction of the ester linkage.19b It is more difficult to speculate on the pathways associated with cleavage at C(15) or C(16) without further study. The tetraene struc- ture shown fits the formula and the data from the ana- logues 2 4 . The precursor ion spectrum indicates that the fragment at m/z 888 may also produce an ion of m/z 591, so it seems likely that the complete mechanism involves an elimination of the substituent at C( 16) fol- lowed by an unknown hydride shift and rearrangement.

A third odd-mass fragment, less abundant than those at m/z 593 and 591, appears at m/z 443. This fragment has been determined to be a precursor of an abundant fragment ion at m/z 329 by precursor ion linked scans (constant Bz/E). Peak-matching experiments show that these fragments have lost neutral fragments of C,,H,,NO, and C32H,,N0, , respectively. The C(27) functionality is preserved in the fragment at m/z 443, but not in m/z 329, while that at C(16) is lost in both ions. An interesting result of the comDarison of frae-

present in 2 the C(27)-demethoxy analogue. We con- clude that, although the functionality at this position is not preserved in the fragment ion, an oxygen atom attached at C(27) is required in the mechanism produc- ing this ion. The accurate mass measurements, coupled with the requirement that the m/z 329 ion be produced by cleavage of the macrocyclic ring at the C(27FC(28) bond, lead to the mechanism proposed in Scheme 2 for production of both of these ions. We are unable to iden- tify clearly the hydride shift rearrangement responsible for the cleavage of the C(23FC(24) bond, but the struc- ture of the m/z 443 ion is supported by both precursor ion scanning and by accurate mass measurements. The 1,4-elimination reaction producing m/z 329 from m/z 443 has been invoked as a thermally allowed, charge- remote mechanism in the decomposition of fatty acids and alcohol^.'^ Although some controversy exists over the validity of this mechanism,'* a recent reviewz1 indi- cates that it is currently considered the most likely pathway in the fragmentation of long alkyl chains.

Li' H

1 ,? hydride shift

OH

6"- Li' 0+JH

ment ions from 1 2 is that the ion airnlz 329 is not Scheme 2.

STRUCTURAL ANALYSIS OF RAPAMYCIN AND RELATED COMPOUNDS 45 1

Likewise, the elimination at the C(33+C(34) ester linkage, common to the production of all the odd-mass ions, can be considered a reasonable charge-remote process. Also shown in Scheme 2 is the production of m/z 329 via an alternate pathway from a low-abundance fragment at m / z 458. Precursor ion linked scanning shows that this fragment is in turn derived from a monolithiated ion at m / z 572, but this fragment is present in the full-scan spectrum at too low abundance to obtain reliable accurate mass data. The same is true for the m/z 458 fragment, but is assumed to contain the nitrogen atom by analogy with the other derived struc- tures. The structure shown in Scheme 2 is thus a likely possibility, and production of m/z 329 takes place via the common elimination at the ester linkage.

An ion at m / z 626 is also produced from the rapamycin [M + Li]+ ion and by 3 and 4, but not by 2. In this fragment, as in the production of mlz 329, the C(16) function is lost and the analogous ion is not pro- duced by the C(27)-demethoxy compound. This leads to the conclusion that the macrocyclic ring is again cleaved at C(27)-C(28) bond with participation of the methoxy group, as proposed in Scheme 2. From this premise, and knowing the formula of the lost portion of the molecule (from Table l), it is only necessary to walk around the macrocyclic ring until the required portion of the molecule is removed. A reasonable mechanism is given in Scheme 3.

In the spectra of all four compounds, a distinct frag- ment ion formed by the loss of 102 u is present. Since the mass lost is unchanged with modifications at the C(27) and C(16) positions, these portions of the ecule must remain intact in the fragment ion, as

mol- must

the nitrogen atom in an even-mass fragment. The mea- sured mass of the m/z 818 fragment in the rapamycin spectrum indicates a loss of C,H,,02, which may be derived from the substituted cyclohexane ring through the two-step mechanism shown in Scheme 4. This elimi- nation and reverse Diels-Alder pathway is consistent with reported charge-remote mechanisms in metal- cationized molecule^.'^*^^ The precursor ion spectrum indicates that the fragment is derived directly from the [M + Li]+ ion at m/z 920, so the mechanism appar- ently takes place in a concerted fashion. Another possi- bility is that, following losses of water and methanol, the remainder of the lost mass is derived from either the piperidine ring or the hemiketallpyran ring. In either case, we find it difficult to write a reasonable mecha- nism which accounts for the fragments lost.

The fragment ions at m/z 586 and 604 were shown to be dilithiated by the 6Li labeling experiments and by the accurate mass data, so structures have not been pro- posed for them. Two monolithiated ions appear in the linked scan fragment ion spectrum, however, at mlz 580 and 598. These do not appear in the full-scan LSI mass spectrum at sufficient abundance to determine their masses accurately, but they were initially assumed to have the same basic structure as the dilithiated ions, replacing one of the lithium ions with a proton. Precur- sor ion scans show that the mlz 580 fragment is derived from the [M + 2Li - HI+ ion rather than the [M + Li]+, but the fragment at m/z 598 comes from the [M + Li]'. The equivalent ion appears in the spectra of 3 and 4, but not the demethoxy compound 2, which leads again to the conclusion that the C(27) oxygen- containing substituent must participate in the reaction.

9"

09 626

--

Scheme 3.

Scheme 4.

452 J. P. KIPLINGER AND M. A. GUADLIANA

Figure 3. Division of the rapamycin molecule into several regions, each of which is independently observable by exami- nation of the fragmentation pattern.

Walking around the macrocyclic ring to remove the required material leads to the structure shown in Scheme 1, which is again the result of an elimination at C(33+C(34). The mechanism for the ring cleavage at C(27+C(28) is unclear, although Table 2 shows that the C-27 substituent is retained in the product.

The variety of fragments whose structures may be assigned based on this work allows us to observe six separate regions of the rapamycin molecule indepen- dently. These are shown in Fig. 3. This ability to local- ize structural modifications in rapamycin analogues was the stated goal of this work. As a hypothetical example, consider a C(32) desoxorapamycin analogue. The [M + Li]' ion would appear at a nominal m/z of 906, indicating that the analogue differs from rapamycin by one less CH, group, by modification of an OCH, to an

OH or by complete reduction of a carbonyl to CH,. Accurate mass measurement of the [M + Li]' ion should clearly indicate that a carbonyl has been reduced and eliminate the other two choices. The mass differ- ence would be retained in all the observed fragments except m/z 598, which is only possible if the modifi- cation is between C(28) and C(37). Only one carbonyl is present in this region, at C(32). It should be generally possible to deduce the structure of any analogue that contains one or two such simple structural changes, although compounds with more modifications might present substantially greater challenges.

CONCLUSIONS

Since biosynthetic pathways producing rapamycin-like molecules cannot produce random, infinite variations in the compound, it is likely that most analogues will be simply modified in one or two places. Further, these modifications should be consistently found at certain locations within the molecule-C(l6) and C(27) in the examples presented here. In such cases, it should prove straightforward to identify many new structures based entirely on the fragment ion spectra generated from their [M + Li]' ions. For the natural products chemist accustomed to having to isolate and purify a relatively large amount of material for NMR structure determi- nation or crystallography, this method offers a highly desirable alternative source of structural information.

We have made confident structural assignments for only seven fragment ions, since the dilithiated fragments of the [M + 2Li - HI+ ions were excluded from con- sideration in this study. If the dilithiated fragments are considered, accurate mass information is available on ten product ions, and 13-16 distinct products are obser- vable in the spectra of 1-4. As work progresses with new analogues, it should become possible to locate structural modifications with greater precision, and better understanding of fragmentation mechanisms will be gained.

REFERENCES

1.

2. 3.

4. 5. 6. 7.

8.

9.

10.

11.

C. Vezina, A. Kudelski and S. N. Sehgal, J. Anfibiot. 28, 721 (1975). S. Schreiber Chem. Eng. News October 26, 22 (1992). J. Cantone, and K. Chan, in Proceedings of the 40th ASMS Conference on Mass Spectrometry and Allied Topics, Wash- ington, DC, M a y d l J u n e 5 1992, p. 1881. J. Adams, MassSpecfrom. Rev. 9, 141-186 (1990). K. Isa and Y. Terai, Nippon Kagaku Kaishi 4, 572 (1 988). R. P. Lattimer, J. Am. SOC. Mass Spectrom. 3, 225 (1 992). (a) J. Adams and M. L. Gross, Anal. Chem. 59, 1976 (1 987) ; (b) R. P. Grese and M. L. Gross, J. Am. Chem. SOC. 11 2, 5098 (1 990). J. Adams and M. L. Gross, J. Am. Chem. SOC. 108, 6915 (1986). J. Adams and M. L. Gross, Org. Mass Spectrom. 22, 307 (1987). M. J. Contado and J. Adams, Anal Chim. Acfa 246, 187 (1991). (a) G. E. Hofmeister, Z. Zhou and J. A. Leary, J. Am. Chem. SOC. 113, 5964 (1991); (b) A. Staempfli, Z. Zhou and J. A. Leary, J. Org. Chem. 57, 3590 (1 992).

12.

13.

14.

15.

16.

17.

18.

19.

Z. Zhou, S. Ogden and J. A. Leary, J. Org. Chem. 55, 5444 (1 990). (a) L. M . Teesch and J. Adams, J. Am. Chem. SOC. 113, 812 (1991); (b) R. P. Grese and M. L. Gross, J. Am. Chem. SOC. 112,5098 (1990). (a) J. A. Lean/, T. D. Williams and G. Bott, Rapid Commun. Mass Spectrom. 3,192 (1 989), (b) J. A. Leary, Z. Zhou, S. A. Ogden and T. D. Williams, J. Am. SOC. Mass Specrrom. 1,473 (1 990). J. A. Leary and S. F. Pedersen, J. Org. Chem. 54, 5650 (1 989). K. P. Madhusudanan and C. Singh, Org. Mass Spectrom. 32, 1329 (1 992). K. P. Madhusdanan, T. S. Dhami, S. N. Suryawanshi and D. S. Bhakuni Org. Mass Specfrom. 27, 837 (1 992). 0. Laprevote, C. Girard, B. C. Das, T. Laugel, F. Roblot, M . Leboeuf and A. Cave, Rapid Commun. Mass Spectrom. 6, 352 (1992). (a) M. Takayama, Org. Mass Spectrom. 28, 878 (1993); (b) K. P. Madhusudanan, A. Rani, B. Kumar and S. N. Suryama- nishi, Org. Mass Specfrom. 28, 892 (1 993).

STRUCTURAL ANALYSIS OF RAPAMYCIN AND RELATED COMPOUNDS 453

20. K. P. Madhusudanan, S. B. Katti and S. A. N. Hashmi, Org. Mass Spectrom. 28,970 (1 993).

21. M. L. Gross, Int. J. Mass Spectrom. Ion Processes 118, 137 (1 992).

22. 0. Burlet and S. J. Gaskell, J. Am. SOC. Mass Spectrom. 4. 461 (1993).

23. J. P. Kiplinger, Rapid Commun. Mass Spectrom. 7, 320 (1 993).

24. R. L. Foltz, A. F. Fentiman, Jr, L. A. Mitscher and H. D. D. Showalter, J. Chem. SOC., Chem. Commun. 872 (1973); ( b ) D. K. MacMillan, M. L. Gross, B. N. Pramanik and A. K.

Mallams, in Proceedings of the 39th ASMS Conference on Mass Spectrometry and Allied Topics, Nashville, TN, June 1991.

25. K. B. Tomer and M. L. Gross, Biomed. Environ. Mass Spec- trom. 15,89 (1 988).

26. 0. W. Hand and R. G. Cooks, Int. J. Mass Spectrom. Ion Pro- cesses 97, 35 (1 990).

27. J. Adams and M. L. Goss, J. Am. Chem. SOC. 111, 435 (1989).

28. V. H. Wysocki, M. M. Ross, S. R. Horning and R . G. Cooks, Rapid Commun. Mass Spectrom. 2, 214 (1 988).