Journal of the American Chemical Society
COMMUNICATION
products were identical to those of the authentic standards for
3a/b and 4a/b (Figures 3 and S1). However, the catalytic
efficiency of S-LovA for the synthesis of 4a/b was at least
110-fold higher than that of HS-LovA, taking into consideration
that HS-LovA showed 11.9-fold higher expression than S-LovA.
LC-MS analyses of the four compounds confirmed that the
masses of these compounds were consistent with those for 3a/b
and 4a/b. In the negative-ion (ꢀ)LC-MS, 3a and 4a showed
[M ꢀ H]ꢀ ions of m/z 321 and 337, respectively. In the positive
ion (þ)LC-MS, the [M þ H]þ ions of 3a and 4a were labile and
underwent dehydration (ꢀ18 Da) to form [M þ H ꢀ H2O]þ
ions whose m/z values corresponded to the predicted values of
positive ions for 3a (m/z = 305) and 4a (m/z = 321). In
addition, [M þ H]þ ions of the lactone compounds (3b and
4b) were consistent with the predicted masses (m/z = 305 for
3b and m/z = 321 for 4b).
It is potentially feasible that both free acid and lactone forms
(1a and 1b) could be used as LovA substrates. In order to test if
1b can be used as LovA substrate, 100 μM 1b was fed to the yeast
expressing S-lovA and cpr, and the culture was incubated for 8 h.
However, no conversion of 1b was detected by HPLC-DAD
(Figure 3). When the same sample was analyzed by a highly
sensitive LC-MS, the four compounds (3a/b and 4a/b) could be
detected, but their abundance was about 250-fold lower than
those converted from 1a. Therefore, it appears that 3a and 4a
were synthesized from 1a and then converted to the correspond-
ing lactones in an acidic yeast culture medium. In order to verify
this, the same feeding experiments were performed in extended
incubation times (24 h) with varying final pH (3.0ꢀ6.8) in the
medium using different buffer strengths. In acidic conditions
(pH 3), almost all of the monacolin L and monacolin J were
present as their lactone forms (3b and 4b), whereas their free
acid forms (3a and 4a) were dominant in the medium with final
pH 6.8 (Figure S2). This result together with the data from the
1b-feeding assay suggested that 1a is the LovA substrate. The 3b
and 4b apparently resulted from non-enzymatic lactonization in
acidic yeast medium.
Using the pH-optimized conditions, yeast feeding assays were
scaled up (1 L), and 3a and 4a were purified by HPLC. The
structure of the final product 4a was verified by spectral
comparison to authentic standard, and standard NMR analyses
were used to confirm the structure of 3a (Supporting In-
formation). By using FT-ICR-MS, the exact m/z of the
[M ꢀ H]ꢀ for 3a was determined to be 321.20700 and for 4a
to be 337.20212. These values were less than 0.4 ppm deviations
from the theoretical masses. To ensure the reactions were
catalyzed by LovA, in vitro enzyme assays were done using
microsomes prepared from yeast expressing S-lovA and cpr.
When 1a was incubated with the microsomes, 3a and 4a were
produced as shown by LC-MS analysis (Figure 4A). Two
additional [M ꢀ H]ꢀ ions displaying m/z 339 were detected.
One of these compounds is likely to be 3R-hydroxy-3,5-dihy-
dromonacolin L acid (2a), a reported intermediate in the
lovastatin biosynthesis, and we propose that the other compound
is its isomer, 4aR-hydroxy-4a,5-dihydromonacolin L acid (Figure
S3, Supporting Information). No conversion was detected when
1b was incubated with the microsomes, consistent with the
in vivo feeding experiment. As many P450 enzymes catalyze
epoxidations, it has been proposed that 3,4-epoxy-dihydromo-
nacolin L could be a LovA reaction intermediate.7 To examine
this possibility, the pure R and β isomers of 3,4-epoxy-dihy-
dromonacolin L (open forms a, m/z 339) were chemically
Figure 4. In vitro LovA enzyme assays. Total negative ion scans were
performed by LC-MS. Selective ions of m/z 323, 339, 321, and 337 were
used to detect the metabolites shown in Figure 1B. Beside the peaks, m/z
values of [M ꢀ H]ꢀ ions are given. (A) 1a, 4a,5-dihydromonacolin L
acid (substrate), was incubated with the microsomes from yeast expres-
sing either cpr only or cpr and S-lovA. Asterisks indicated compounds
displaying negative ions of m/z 339, proposed to be 2a and its isomer
4aR-hydroxy-4a,5-dihydromonacolin L acid (Supporting Information).
(B) 3a, monacolin L acid, was used as substrate in the same experimental
conditions as described for panel A.
synthesized from 1b and also independently incubated with
the microsomes, but these were not transformed further (data
not shown). In addition, MS/MS analysis suggested that the two
m/z 339 compounds from the in vitro assays are not 3,4-epoxy-
dihydromonacolin L (Figure S4, Supporting Information). On
the basis of these results, we propose that 2a is synthesized by a
hydrogen (1a C-4a hydrogen) abstraction and subsequent oxy-
gen re-bound (i.e., C-3 hydroxyl group) onto the allylic radical
(Figure S3). Using scaled-up yeast cultures (1 L), we attempted
to purify the two compounds with m/z 339 after feeding 1a, but
the low abundance of these two compounds did not allow us to
acquire sufficient amounts for NMR analyses.
In the assays described thus far using 1a, it cannot be excluded
that the second reaction (the conversion of 3a to 4a) is catalyzed
by an unknown yeast enzyme. Also, 3a could, in principle, be a
reaction shunt product that is released from LovA but cannot be
re-introduced into the LovA biosynthetic pathway. To address
these questions, the purified 3a was used as a substrate for in vitro
LovA assays. In these assays, clear conversion of 3a to 4a was
observed, with no trace of catalytic conversion in the control
microsomes (Figure 4B). The Km value of LovA for 3a was
determined to be 6.2 ( 1.1 μM, and the microsomes showed
Vmax = 9.1 ( 0.5 pmol minꢀ1 mgꢀ1. The sufficiently low Km
value supported the physiological relevance of LovA activity in
A. terreus. These results demonstrate that 3a is a true intermedi-
ate in the lovastatin biosynthetic pathway.
8080
dx.doi.org/10.1021/ja201138v |J. Am. Chem. Soc. 2011, 133, 8078–8081