(Received for publication, July 9, 1996, and in revised form, October 7, 1996)
From the Department of Organic Chemistry and
Biochemistry, Technical University of Munich, Lichtenbergstrasse 4, D-85747 Garching, Federal Republic of Germany and § F.
Hoffmann-La Roche Ltd., Pharma Preclinical Research & Development,
Department of Biotechnology, CH-4070 Basel, Switzerland
The biosynthesis of the pancreatic lipase
inhibitor lipstatin was investigated by fermentation experiments using
cultures of Streptomyces toxytricini, which were supplied
with soybean oil and a crude mixture of U-13C-lipids
obtained from algal biomass cultured with
13CO2. Lipstatin was analyzed by one- and
two-dimensional NMR spectroscopy. 13C total correlation
spectroscopy and INADEQUATE experiments show that two fatty acid
fragments containing 14 and 8 carbon atoms, respectively, are
incorporated en bloc into lipstatin. The 14-carbon fragment
is preferentially derived from the unsaturated fatty acid fraction, as
shown by an experiment with hydrogenated U-13C-lipid
mixture, which is conducive to labeling of the 8-carbon moiety but not
of the 14-carbon moiety. The data indicate that the lipstatin molecule
can be assembled by Claisen condensation of octanoyl-CoA with
3-hydroxy-5,8-tetradecanoyl-CoA obtained by
oxidation of linoleic acid. The formation of lipstatin from acetate
units by a polyketide-type pathway is ruled out conclusively by these
data. The data show that surprisingly clear labeling patterns can be
obtained in studies with crude, universally 13C-labeled
precursor mixtures that are proffered together with a large excess of
unlabeled material. One- and two-dimensional 13C total
correlation spectroscopy analyses are suggested as elegant methods for
the delineation of contiguously 13C-labeled biosynthetic
blocks.
Lipstatin (Fig. 1, 1) and its
tetrahydroderivative tetrahydrolipstatin (Fig. 1, 2) are
potent inhibitors of pancreatic lipase (1, 2, 3, 4). Tetrahydrolipstatin is
presently in clinical trials as an antiobesity agent.
Tetrahydrolipstatin has been shown to form a covalent adduct with a
serine moiety of human pancreatic lipase by transesterification
(5).
Lipstatin is produced by Streptomyces toxytricini (1). The
structure is characterized by a lacton ring carrying two aliphatic residues with chain lengths of 6 and 13 carbon atoms (Ref. 6 and Fig.
1). One of the side chains contains two isolated double bonds and a
hydroxy group esterified to N-formylleucine. Structurally, lipstatin is closely related to the esterase inhibitor esterastin, which contains a N-acetylasparagine side chain instead of
N-formylleucine (7). Tetrahydrolipstatin can be obtained by
catalytic hydrogenation of lipstatin (6). Total synthesis of
tetrahydrolipstatin has been reported (8).
Nothing is known about the biosynthesis of lipstatin. Initial biosynthetic studies with isotope-labeled acetate were not conducive to the incorporation of label into lipstatin.1
The long alkyl side chains of lipstatin could be biosynthesized from acetate via a polyketide pathway or from preformed fatty acids. S. toxytricini can be grown on medium containing large amounts of lipids, and it appeared likely that these were actively metabolized by the microorganism. In that case, 13C-labeled lipids could be incorporated into lipstatin by partial degradation of preformed fatty acids or by total degradation via acetyl-CoA.
Triglycerides with appropriate 13C labeling were not commercially available. We therefore decided to use a mixture of universally 13C-labeled lipids, which was obtained by acetone extraction of algal biomass grown with 13CO2. This complex mixture was extensively diluted with soybean oil (natural 13C abundance) for the preparation of culture media. In conjunction with advanced NMR techniques, this approach was conducive to the elucidation of the building blocks of lipstatin. This method appears to be generally useful for biosynthetic studies on compounds with a putative origin from fatty acid precursors.
The crude U-13C-lipid mixture used in this study was purchased from Dr. H. Oschkinat (European Molecular Biology Laboratory, Heidelberg, Germany). It was prepared by acetone extraction of algal biomass (Scenedesmus obliquus) grown on 13CO2. The acetone extract was evaporated to dryness under reduced pressure. The black, oily material was used as a supplement to the culture medium without purification.
Hydrogenation of U-13C-LipidA solution of 13C-lipid (crude algal extract, 277 mg) in 6 ml of ethanol was hydrogenated over 550 mg of Raney nickel (50 °C, 10-bar hydrogen, 3 h). The product still contained approximately 8% of C18:1 fatty acid according to GC2 analysis. The hydrogenation was therefore repeated using similar conditions, yielding 124 mg of fully hydrogenated material.
Fatty Acid Composition of 13C-Algal LipidAn
aliquot of the 13C-labeled lipid mixture (20 mg) and 1.25 mg of pentadecanoic acid were dissolved in 10 ml of 0.5 M
sodium methylate in methanol. The mixture was heated to 60 °C in a
screw cap glass for 30 min and was then acidified with 4 ml of 3.6% hydrochloric acid in methanol. An aliquot of 5 ml was diluted with 3 ml
of water, and the mixture was extracted with 5 ml of n-hexane. The resulting fatty acid methyl esters were
identified by GC/MS on a capillary column (DBWAX; 20 m, 0.3-µm
film; temperature gradient, 120-250 °C, 4 °C/min; splitless
injection). Quantitative GC analysis of the methyl esters was performed
on a capillary column (OV-225; 25 m, 0.25 µm; temperature
gradient, 140-220 °C, 3 °C/min). Pentadecanoic acid methylester
was used as internal standard for quantification, and
dihomo--linoleic acid ethyl ester (97.4% purity by weight) was used
as a reference substance. The 13C content of fatty acid
residues was determined by GC/MS analysis (electron impact ionization)
of the methyl esters.
For biosynthetic studies, the U-13C-labeled precursors had to be diluted extensively with unlabeled material. Per 35 ml of fermentation medium, 1.8 g of soybean oil and 0.1 g of the 13C-labeled lipid mixture or hydrogenated 13C-labeled lipid mixture were used. The lipids were emulsified in water using 0.5 g of soybean lecithin as an emulsifier. The medium also contained soybean flour (1.4 g) and glycerol (0.7 g). The pH was adjusted to 7.4 before sterilization (121 °C, 20 min). After seeding with S. toxytricini, the culture was incubated for 7 days at 27 °C and 70% relative humidity on a rotary shaker at 220 rpm and 5-cm throw.
Isolation of LipstatinFermentation broth (58 ml) was extracted with 150 ml of acetone and 100 ml of hexane. After separation of the organic layer, the aqueous layer was extracted three times with 100 ml of a 1:1 mixture of acetone and hexane. The combined organic extracts were dried with sodium sulfate and concentrated to yield a green oil (3.1 g) containing 173 mg of lipstatin. The material was dissolved in hexane (40 ml) and placed on a column of silica gel (Bond Elut, 10 g). The column was developed with hexane (40 ml) and hexane/ethyl acetate at dilutions of 20:1 (v/v, 300 ml), 10:1 (200 ml), 20:3 (200 ml), and 5:1 (200 ml), yielding 220 mg of crude material after evaporation of solvent. Second chromatography on silica gel using the same procedure afforded 163 mg of semipure material. Further purification was done by reversed phase chromatography. Lipstatin was dissolved in 50% aqueous isopropanol (25 ml) and placed on a column of Bond Elut C-18 (10 g) which was developed with 50% isopropanol (225 ml) and 60% isopropanol (150 ml). Fractions containing lipstatin were concentrated by evaporation under reduced pressure, and the residue was extracted with ethyl acetate. The extract was dried with sodium sulfate and was concentrated to dryness, yielding 112 mg of lipstatin with 88% purity (high performance liquid chromatography, percentage of area) containing 9% of lipstatin analogs.
NMR Spectroscopy1H and 13C NMR spectra were recorded at 360 and 90.6 MHz, respectively, with a Bruker AM 360 spectrometer equipped with fast power switching of the transmitter output, an external pulse amplifier (BFX5, Bruker), and a selective excitation unit (Bruker). Chloroform-d was used as a solvent.
1H NMR spectra were measured as follows: 50° pulse (4 µs); repetition time, 2.5 s; spectral width, 5.3 kHz; data set, 16 kilowords; temperature, 25 °C; and 0.1-Hz line broadening. 13C NMR spectra were measured as follows: 30° pulse (2 µs); repetition time, 2.8 s; spectral width, 25.0 kHz; data set, 32 kilowords zero filled to 64 kilowords; temperature, 25 °C; and 1H composite pulse decoupling.
Two-dimensional double quantum-filtered COSY, DEPT, and INADEQUATE experiments were performed according to standard Bruker software (DISR87). Phase-sensitive two-dimensional TOCSY spectroscopy was done according to the method of Bax and Davis (9). 1H-Detected multiple quantum 1H-13C chemical shift correlation experiments (HMQC and HMBC) were performed according to the methods of Bax and Subramanian (10) and Bax and Summers (11). Samples were not rotated during two-dimensional experiments.
Data acquisition and processing parameters for two-dimensional experiments were: COSY, 32 scans/t1 increment, 2.0-s relaxation delay, 480 × 2048 raw data matrix size, zero filled to 2048 words in t1 and processed with 2-Hz Gaussian in the f1 dimension, and 90°-shifted sine bell filtering in the f2 dimension; TOCSY, 32 scans/t1 increment, 2.0-s relaxation delay, 62-ms MLEV-17 mixing period preceded and followed by 2.5-ms trim pulses, 90° pulse width, 44 µs, 512 × 2048 raw data matrix size, zero filled to 2048 words in f1 and processed with 2-Hz Gaussian in the f1 dimension, and 90°-shifted sine bell filtering in the f2 dimension; HMQC, 256 scans/t1 increment, start of coherence experiment 159 ms after a bilinear rotation decoupling pulse, 3.5-ms delay period for evolution of 1JCH corresponding to a coupling of 143 Hz, 13C decoupling during acquisition by globally optimized alternating phase rectangular pulse sequence 1, 500 × 1024 raw data matrix size, zero filled to 1024 words in t1 and processed with 10-Hz Gaussian in f1, and 90°-shifted squared sine filtering in f2; HMBC, 64 scans/t1 increment, 1.8-s relaxation delay, 3.5-ms delay for suppression of 1JCH, 60-ms delay period for evolution of long-range couplings (2JCH and 3JCH) corresponding to a coupling of 8 Hz, and 350 × 2048 raw data matrix size, zero filled to 1024 words in t1 and processed with 20-Hz Gaussian in f1 and in f2; and INADEQUATE, 128 scans/t1 increment, 1.5-s relaxation delay, Ernst-type phase cycle, 5.0-ms delay for evolution of 1JCC, and 350 × 2048 raw data matrix size, zero filled to 2048 words in t1 and processed with 60°-shifted sine bell filtering in f1 and f2.
Phase-sensitive two-dimensional 13C TOCSY experiments were performed with a MLEV-17-based mixing period (9). The 13C excitation pulse was generated in the transmitter high power output level (90° pulse, 5 µs). 13C pulses for mixing were generated in the transmitter low power output level amplified with a BFX5 unit (90° pulse, 30 µs). The MLEV-17 mixing period was 45 ms and was preceded and followed by 2.5-ms trim pulses. The data were acquired in the phase-sensitive mode using time-proportional phase increments. Other data acquisition and processing parameters were: 48 scans per t1 increment, 2.0-s relaxation delay, and 400 × 2048 raw data matrix size, zero filled to 2048 in t1 and processed with 90°-shifted, squared sine bell functions in f1 and Gaussian broadening in f2.
One-dimensional 13C TOCSY experiments were performed with selective excitation using a Gaussian- or half-Gaussian-shaped pulse of 5 ms in length generated by the transmitter output of a selective excitation unit (Bruker). The transfer of magnetization between 13C atoms was achieved by a MLEV-17-based mixing period (45 ms) preceded and followed by trim pulses (2.5 ms). The pulses for mixing were generated in the low power output of the transmitter amplified with a BFX5 unit (90° pulse, 30 µs).
NMR Signal Assignment of Lipstatin
A 1H and 13C NMR analysis of lipstatin providing assignments for some of the carbon atoms has been reported (6). However, since the biosynthetic study depended crucially on unequivocal assignments for all 13C signals, a more detailed NMR analysis using double quantum-filtered COSY, 1H TOCSY, and inverse carbon-proton correlation experiments such as HMQC and HMBC was in order. Additional assignment information was afforded by 13C TOCSY and INADEQUATE analysis of multiple 13C-labeled lipstatin samples obtained in the labeling experiments described below. 1H and 13C NMR signal assignments of lipstatin are summarized in Table I.
|
Analysis of U-13C-Lipid Mixture
An acetone extract of totally 13C-labeled algal biomass obtained by growth of Scenedesmus obliquus on 99% enriched 13CO2 as a carbon source was obtained from Dr. H. Oschkinat. After evaporation of the solvent, the viscous oil appeared almost black. The dark color was in part due to the presence of chlorophylls in considerable amount.
For assessment of the fatty acid content, the crude algal lipid mixture was subjected to methanolysis, and the resulting fatty acid methyl esters were analyzed by coupled gas chromatography/mass spectrometry (G. Oesterhelt, Hoffmann-La Roche AG). The results are shown in Table II. The combined fatty acid residues account for about 40% (w/w) of the total material. Eleven fatty acids with chain lengths of 14-18 carbon atoms were determined in widely different abundance. In the saturated fraction, palmitic acid was the dominant component. The combined unsaturated fatty acids accounted for 32% (w/w) of the crude material. The dominant components were linolenic acid (about 21%) and linoleic acid (about 8%). The 13C abundance was approximately 97%.
|
An aliquot of the U-13C-lipid mixture from algae was subjected to two consecutive cycles of catalytic hydrogenation over Raney nickel. Gas chromatographic/mass spectrometric analysis of fatty acid methyl esters obtained by methanolysis confirmed that virtually all double bonds had been removed by hydrogenation (Table II). The dominant fatty acid in the hydrogenated mixture was stearic acid (about 25%, w/w).
Biological Studies
Fermentation with U-13C-LipidFermentation experiments were performed with S. toxytricini and the culture medium reported earlier (1). To detect the joint transfer of 13C-labeled atoms, it is important to dilute the U-13C-labeled precursors extensively with unlabeled material. We used the 13C-labeled lipid mixture and unlabeled soybean oil at a ratio of 1:17.4 (w/w) for the preparation of the fermentation medium. After cultivation for 7 days, lipstatin was isolated as described under "Materials and Methods."
The analysis of the 13C satellites in the 1H NMR spectrum of lipstatin gave 13C enrichments of about 4%. This indicated that the 13C-labeled algal lipid had been metabolized by the organism at a similar rate as soybean oil and had served as a precursor of the lipstatin molecule.
The 13C NMR spectrum of lipstatin revealed the presence of
extensive 13C-13C coupling in the biolabeled
molecule (Fig. 2). For example, the 13C
signal of C-3 was characterized by a central signal at 74.9 ppm and by
a doublet (1JCC, 39.2 Hz) resulting from
coupling to one adjacent 13C atom (C-2 or C-4). The
13C satellites of C-5 appeared at a distance of 78.2 Hz.
Coupling to either C-4 or C-6 should result in a coupling constant of
about 40 Hz due to the aliphatic nature of C-4 and C-6. Therefore, the observed coupling signature must result from simultaneous coupling to
two adjacent 13C atoms (i.e. C-4 and C-6)
yielding a pseudo-triplet (1JCC, 39.1 Hz) where
the central component overlaps with the uncoupled singlet (Fig. 2). It
follows that the carbon atoms C-4-C-6 were derived en bloc
from the uniformly 13C-labeled algal lipid. Similarly, the
signal of C-2 was characterized by simultaneous 13C
coupling to two 13C spins via one-bond coupling
(1JCC, 42.4 Hz and 38.0 Hz, respectively) and
to one 13C spin via two-bond coupling
(2JCC, 3.6 Hz) (Fig. 2).
The analysis of 13C-13C coupling in terms of multiplicities in the one-dimensional 13C NMR spectrum is summarized in Table III. These data show already that biosynthetic modules containing more than two carbon atoms were incorporated into the lipstatin molecule. However, the size of these respective building blocks can be addressed much better by two-dimensional NMR analysis as described below.
|
A section of a two-dimensional 13C INADEQUATE experiment is
shown in Fig. 3. The double quantum-filtering technique
monitors pairs of adjacent 13C atoms but not isolated
13C nuclei. Due to the low natural abundance of
13C, INADEQUATE is notorious for its low sensitivity.
However, pairs of 13C atoms that were contiguously
incorporated from multiple 13C-labeled precursors are
diagnosed with high sensitivity. The data in Fig. 3 indicate
13C-13C coupling between the carbon atoms 3 and
4, 4 and 5, and 5 and 6, suggesting the presence of molecules with
contiguous 13C labeling between C-3 and C-6. A more
detailed analysis (Table III) proves that
13C-13C coupling occurs between each individual
pair of adjacent carbon atoms in both alkyl side chains, thus
suggesting that these modules were incorporated en bloc from
their respective precursors. However, 13C-13C
coupling was not detected between C-3 and C-2, as shown by the absence
of signals at the positions marked by open circles in Fig.
3. It follows that the bond between C-2 and C-3 was formed in the
biosynthetic pathway and was not present in the totally labeled
precursor molecule. Bond formation involving one labeled precursor
molecule would most frequently involve an unlabeled reaction partner,
and the frequency of 13C-13C coupling along
this bond should be as low as about 4% (i.e. undetectable)
compared with about 97% for jointly transferred carbon pairs.
Contiguous 13C labeling throughout both alkyl side chains is also obvious by 13C TOCSY methods. The physical basis of the TOCSY experiment is the exchange of magnetization between directly coupled spins by a radiofrequency field (spin lock field). In the case of a two-dimensional experiment, the result of the mixing process is that a spin (1H or 13C) shows a correlation cross-peak to each of the nuclei in a contiguous spin system.
A transfer of magnetization under the influence of a spin lock field was introduced by Davis and Bax (12) in 1985. To improve the performance of the experiment, periodic phase alteration (MLEV-17) of the spin lock pulses was implemented (10). Traditionally, the TOCSY experiment is used to assign scalar coupled 1H spin systems. More recently, TOCSY pulse trains were also applied to transfer magnetization between 13C of labeled biopolymers (13). However, the 13C TOCSY experiment is used infrequently in the evaluation of biosynthetic pathways and therefore requires some technical comments.
The most important difference between 1H and
13C TOCSY methods is the larger spectral width of the
13C chemical shift range. Efficient transfer of
magnetization by the spin lock pulse requires a sufficiently high power
of spin lock field B1. Typically, B1 should
exceed the spectral width. However, the maximum strength of the
B1 field is limited by the thermal and electronic stability
of the probe head during the relatively long mixing period (typically
10-60 ms).
To optimize the experimental parameters, we performed a series of two-dimensional 13C TOCSY experiments with [U-13C]lysine. A spin lock field of 8 kHz applied for 45 ms enabled the transfer of magnetization over a relatively wide frequency range (15 kHz) between all of the six carbon atoms of lysine. However, it should be noted that the cross-peak intensity between the carboxylic carbon and the side chain carbon atoms decreased approximately by a factor 10 relative to the other cross-peaks, which had similar intensities.
To improve the limited digital resolution of this two-dimensional
experiment (e.g. for extraction of coupling constants), we
performed a selective excitation of a single 13C frequency
followed by a 13C TOCSY mixing process. The selective
excitation was achieved by a 5-ms Gauss or semi-Gauss pulse. In model
experiments with [U-13C]lysine, we were able to transfer
magnetization from the -carbon to C-3-C-6 with similar efficiency
and to C-1 with decreased intensity. Obviously, the spin lock field (8 kHz) was too weak to achieve efficient magnetization transfer between
C-1 and C-2 of lysine, which are separated by approximately 12 kHz from
each other. Since the intensities of C-3-C-6 in the one-dimensional
13C TOCSY spectrum were similar, the applied spin lock
field was optimal for a frequency range of 3-4 kHz.
These results served as a basis for the 13C TOCSY
experiments with lipstatin. Fig. 4 shows a part of a
phase-sensitive two-dimensional 13C TOCSY experiment with
lipstatin from the fermentation with U-13C-lipid
encompassing the aliphatic spectral region (80-10 ppm). The transfer
of magnetization is highlighted in Fig. 4 among C-3-C-6 and among C-2
and C-1-C-6
. Again, no transfer of magnetization was observed
between C-2 and C-3. Additionally, a series of selective one-dimensional 13C-TOCSY experiments was performed (Table
III). Fig. 5 shows spectra obtained by selective
excitation of C-3 and C-2, respectively, and subsequent isotropic
mixing. Due to the physical basis of the experiment, only signals were
observed that result from magnetization transfer from the excited
carbon. Consequently, the highly crowded 13C-coupled
one-dimensional 13C spectrum can be edited by this
spectroscopic technique (Fig. 5). It should be noted that the
13C spin lock field was not strong enough to achieve
magnetization transfer from the aliphatic C-6 to the alkene carbon
atoms. However, the signal of C-6 in the one-dimensional
13C TOCSY experiment is a pseudo-triplet (Fig.
5B), which clearly indicates contiguous 13C
coupling to the unsaturated C-7 of lipstatin. In combination with
signal multiplicities in the one-dimensional 13C NMR
spectrum (i.e. double doublets of the inner chain carbon atoms, indicating contiguous coupling) and the INADEQUATE results, this
proves the presence of isotopomers with contiguous 13C
labeling extending from C-16 to C-3, and from C-1 to C-6
(Fig. 6A). We conclude that the lipstatin molecule
is assembled from a 14-carbon (C-16-C-3) and an 8-carbon (C-1-C-6
)
moiety, which can both be derived en bloc from the
U-13C-lipid mixture supplied as a precursor.
Two pairs of labeled carbon atoms (i.e. C-1"/C-2", and C-4"/C-5") were incorporated into the leucine side chain from the 13C-labeled lipid mixtures. This signifies the diversion of multiple 13C-labeled components of the lipid mixture to the biosynthesis of the amino acid.
Fermentation with Hydrogenated U-13C-LipidThe presence of two isolated double bonds in the 14-carbon moiety suggested that it might be derived preferentially or exclusively from the unsaturated fatty acid fraction of the 13C-labeled precursor mixture (i.e. more specifically, from linoleic acid). To check this hypothesis, we used a totally hydrogenated U-13C-lipid mixture for an incorporation experiment.
The hydrogenated U-13C-lipid was mixed with soybean oil at a ratio of 1:17.4 (w/w), and the mixture was proffered to a growing culture of S. toxytricini. Lipstatin was isolated and analyzed by NMR spectroscopy as described above. The 13C NMR signals of the 8-carbon moiety were again characterized by the presence of 13C-coupled satellites, whereas the carbon atoms of the 14-carbon moiety showed no 13C-13C coupling in the one-dimensional NMR experiment (Table III). Analysis by two-dimensional INADEQUATE and 13C TOCSY spectroscopy is summarized in Table III and confirmed that the 8-carbon moiety but not the 14-carbon moiety was consecutively labeled (Fig. 6B).
It follows that the unsaturated 14-carbon moiety was not biolabeled in the experiment with the hydrogenated U-13C-lipid. Apparently, this part of lipstatin was exclusively derived from the unsaturated fraction of the unlabeled soybean oil supplement in this experiment. The labeling pattern of the leucine residue was the same as in the experiment described above.
We have demonstrated that the lipstatin backbone is assembled from two moieties consisting of 8 and 14 carbon atoms, respectively, which were both contiguously labeled with 13C from a U-13C-lipid mixture. Since the uniformly 13C-labeled lipids were proffered together with a large excess of unlabeled lipid material, this result shows that both building blocks were derived from precursor lipids by partial catabolism of fatty acid residues. The de novo synthesis of the building blocks from smaller units such as acetate could not possibly yield lipstatin molecules with uninterrupted 13C labeling of long alkyl chains, since labeled and unlabeled precursor molecules would be interspersed at random, thus conducing to noncontiguous 13C labeling, as shown in the formylleucine moiety of lipstatin. In this case, the TOCSY transfer of magnetization along the alkyl chain would be interrupted.
Catalytic hydrogenation destroys the potential of the
13C-labeled lipid mixture to serve as a precursor for the
14-carbon moiety. However, the hydrogenated lipid mixture is still used
efficiently as a precursor of the 8-carbon moiety and of leucine. It
follows that the 14-carbon moiety is specifically derived from an
unsaturated component of the U-13C-lipid mixture. Linoleic
acid has the same pattern of double bonds as the unsaturated side chain
of lipstatin and could be converted to an appropriate precursor by oxidation (Fig. 7). Alternative pathways for the partial
degradation of linoleic acid, such as the pathway described recently in
rat liver (14, 15), cannot be ruled out on basis of the available
data.
On the basis of the observed labeling patterns, we propose the
following pathway for the biosynthesis of lipstatin (Fig.
8). The 14-carbon moiety is derived from linoleic acid
by partial oxidation, resulting in the release of 2 acetate units.
The CoA ester of 3-hydroxy-
5,8-tetradecanoic acid (Fig.
8, 3) could serve directly as a component for Claisen
condensation. Similarly, the 8-carbon component could be obtained by
partial
oxidation of long chain fatty acids. The resulting
octanoyl-CoA (Fig. 8, 4) could then undergo the proposed
Claisen condensation with the CoA ester of
3-hydroxy-
5,8-tetradecanoic acid, resulting in the
formation of compound 5. This sequence of reactions explains the
formation of the carbon skeleton of lipstatin. The 14- and 8-carbon
modules can both be derived from U-13C-lipid or from
unlabeled lipid. The Claisen condensation will use labeled and
unlabeled fragments at random. Consequently,
13C-13C coupling is not observed between C-2
and C-3 of lipstatin.
A possible scenario for the subsequent reaction steps is immediately
obvious, but the details require further study. The hydroxy group of
the 14-carbon module could be aminoacylated prior to or after the
Claisen condensation. Reduction of the carbonyl group generated by the
Claisen condensation should yield a hydroxy group, and the -lactone
ring of compound 7 could be formed with the CoA moiety as a leaving
group.
It should be noted that this biosynthetic pathway was observed in a medium containing large amounts of saturated and unsaturated fatty acids. Under these conditions, the 14-carbon moiety is entirely derived from the unsaturated fatty acid pool, as shown by the experiments with the hydrogenated U-13C-lipid mixture. Apparently, desaturation of fatty acid does not play a significant role under these culture conditions.
The data also show that degradation products of the proffered lipid are
diverted to the biosynthesis of leucine. The incorporation of two
13C pairs from the precursor mixture is well in line with
the biosynthetic pathway of the amino acid, as summarized in Fig.
9. The biosynthesis of leucine involves the condensation
of pyruvate and acetyl-CoA molecules. A mixture of
13C3-pyruvate and
13C2-acetyl-CoA with the respective unlabeled
precursors should yield the observed labeling pattern, as shown in Fig.
9. Whereas the details have not been investigated,
13C3-pyruvate could be formed during the
fermentation from the glycerol part of the U-13C-lipid
mixture. This should be conducive to the observed isotope distribution.
The use of a crude mixture of 13C-labeled precursors for biosynthetic studies is an unusual approach. As shown with the present example, it can yield results of optimum clarity under appropriate conditions. The crucial part of the present experiments is the use of a mixture of totally 13C-labeled lipids with a large excess of unlabeled lipids. In conjunction with modern one- and two-dimensional NMR technology, this approach can define the length of the biosynthetic building blocks with a minimum of experimental effort. This strategy is not limited to mixtures of lipids. Indeed, we have shown independently that it can also be used successfully using crude mixtures of 13C-labeled carbohydrates or amino acids.3
We thank G. Oesterhelt for mass spectrometry and A. Kohnle for help with the preparation of the manuscript.