Steroid metabolite structure and stereochemistry by NMR
469
for comparison with literature data.5,6 Most 1H-detected
NMR spectra were acquired on a Bruker Avance DRX-600
spectrometer operating at a 1H frequency of 600.13 MHz,
using a 5-mm Bruker TXI triple-resonance 3-axis gradient
probe. 13C and DEPT spectra were acquired on a Bruker
amplitude of 15.7 kHz. The av180b 13C refocusing pulse20
was optimized using a 13C DPFGSE28 (double pulsed
field gradient spin-echo), setting the shaped pulse 90 ppm
(13.5 kHz) off-resonance and varying the pulse width with
the maximum power held constant. The best results were
1
°
obtained with a pulse width of 273 µs (19.1 times the 90 pulse
Avance DRX-500 spectrometer operating at a H frequency
of 499.93 MHz, using a 5-mm Bruker dual 13C-detect probe.
The modified HMBC experiment was optimized using a
width of the corresponding hard pulse, instead of the theoret-
ical ratio20 of 24.8) and a maximum B1 amplitude of 17.5 kHz.
Structure calculations were performed on a Silicon
Graphics Octane workstation using the Insight II software
package (Accelrys, Inc., San Diego, CA). Structures were built
using the Builder module and minimized using the Optimize
command. Energy minimization uses the consistent valence
force field (CVFF) with 1000 steps of steepest descent
followed by 1000 steps of conjugate gradient and finally 1000
steps using the Broyden-Fletcher-Goldfarb-Shanno (BFGS)
algorithm.
1
Varian Inova-600 spectrometer operating at a H frequency
of 599.7 MHz, with a 5-mm HCN Z-axis gradient probe.
°
Samples were analyzed in 5 mm NMR tubes at 25 C with
spinning for 1D experiments and without spinning for 2D
1
experiments. 1D H spectra were acquired with a spectral
width of 7507.5 Hz, acquiring 8192 complex data points.
Sample quantities were estimated by comparing absolute
integrals of resolved single-proton peaks with a calibration
curve derived from 10 different concentrations of sucrose in
D2O. 2D spectra were acquired in TPPI mode23 (gradient-
selected edited HSQC17), States mode24 (NOESY and Varian
gradient-selected DQF-COSY25) or echo-antiecho mode26
(HMBC and Bruker gradient-selected DQF-COSY25) with
1024 complex pairs in t2 and 750 FIDs. All NMR data were
processed using the Felix2000 software (Accelrys Inc., San
Diego, CA). 1H chemical shifts are referenced to residual
CHD2OD at 3.30 ppm and 13C chemical shifts are referenced
to solvent CD3OD at 49.15 ppm (1D spectra) or residual
CHD2OD at 49.43 ppm (2D spectra). J-coupling analysis of
1D 1H spectra was done using an unshifted sinebell window
and zero-filling from 8192 to 32 768 complex data points
before Fourier transformation, for a final digital resolution
of 0.23 Hz/point. 2D data were zero-filled to a final matrix
size of 2048 data points in F2 and 1024 data points in F1,
RESULTS
Steroid metabolism by CYP6A1 and isolation
of metabolites
The [14C] elution profile of the testosterone hydroxyla-
tion reaction is shown in Fig. 2. Incubation with recon-
stituted CYP6A1 resulted in the formation (Scheme 1) of
three major products: 15ˇ-hydroxytestosterone (1a), 2ˇ-
hydroxytestosterone (1b), and 12ˇ-hydroxytestosterone (1c),
as well as two minor dihydroxylated products: 2ˇ,15ˇ-
dihydroxytestosterone (1d) and 12ˇ,15ˇ-dihydroxytesto-
sterone (1e). The dihydroxy products were separated in a
subsequent HPLC purification step (data not shown).
Similarly, progesterone and androstenedione were
also oxidized by CYP6A1 (Scheme 1). Only two prod-
ucts were detected with androst-4-ene-3,17-dione as a
substrate: 15ˇ-hydroxyandrost-4-ene-3,17-dione (2a) and
2ˇ-hydroxyandrost-4-ene-3,17-dione (2b), while proges-
terone oxidation resulted in two major products, 15ˇ-
hydroxyprogesterone (3a) and 2ˇ-hydroxyprogesterone (3b),
and a minor product which was not isolated. The estimated
quantities in the NMR samples were 0.73, 0.19, 0.47, 0.47, 1.8,
0.33, 0.25, 0.26 and 0.15 mg for samples 1a, 1b, 1c, 1d, 1e, 2a,
2b, 3a, and 3b, respectively.
°
using (in both dimensions) a skewed, 45 -shifted sinebell
°
window function for HSQC, a 90 -shifted sinebell for HMBC
and NOESY and an unshifted sinebell for DQF-COSY.
The modified HMBC pulse sequence is shown in Fig. 1.
This sequence is based on the gradient-selected HMQC
sequence16 designed for phase-sensitive data presentation,
with a TANGO27 element and gradient at the beginning to
eliminate 13C-bound 1H coherence. The refocusing period
just prior to acquisition is eliminated to minimize signal
1
loss due to H T2 relaxation, and the 13C spin-echo period
is shortened to allow just enough time () for a gradient
and its recovery delay in each of the two delay periods. After
conversion of multiple quantum coherence (MQC) back to 1H
Structure and stereochemistry of metabolites
To determine regio-and stereospecificity of steroid oxidation,
the reaction products (0.15–1.8 mg) were analyzed by NMR
spectroscopy in CD3OD solution. Complete assignment of all
1H and 13C signals was accomplished for all 9 hydroxylated
steroids using the 2D HSQC, HMBC and COSY data
at 600 MHz (Tables 1 and 2). 1H–1H coupling constants,
derived from analysis of 1D 1H spectra, are given in Table 3.
1H chemical shifts of 1b in CDCl3 were identical to literature
values.5,6 Details of 1H multiplicities and 2D HMBC and
COSY correlations for each of the nine products are given in
the supplementary material, Tables S1–S9. Regiochemistry
of hydroxylation could be determined from 1H coupling
patterns of the downfield-shifted H–C–(OH) protons and
from the HMBC correlations of these protons and their
attached carbons. Each of the three hydroxylation positions
(2, 12 and 15) give unique multiplet patterns and chemical
1
single quantum coherence (SQC) a short ꢁ2ꢂ H spin-echo
is added, followed by another 13C spin-echo which refocuses
1H chemical shift evolution which occurred during the first
13C spin-echo. To achieve uniform excitation over the 13C
13
°
spectral window, all three C 180 pulses were converted
into broadband shaped pulses.
Shaped pulses were calibrated and optimized using a
doped 13CH3I in CDCl3 sample on a Varian Inova-600. The
pulse shapes were calculated in the pulse sequence using the
Pandora’s Box program with a 1.0 µs step size. The ad180
13C adiabatic inversion pulse20 was optimized using the
1
°
°
sequence: 180 (shaped) – 90 (hard) – FID( H-dec.). Excel-
lent inversion up to 120 ppm (18 kHz) off-resonance was
obtained with a pulse width of 477 µs and a maximum B1
Copyright 2006 John Wiley & Sons, Ltd.
Magn. Reson. Chem. 2006; 44: 467–474