Solid-state NMR Observation of Cysteine and Lysine Michael Adducts of Inactivated Estradiol Dehydrogenase*

The inactivation of estradiol dehydrogenase by en- zyme-generated 3-hydroxy-14,15-secoestra-1,3,5(10)-trien- 15-yn-17-one is accompanied by the formation of a lysine enaminone. The experiments leading to this conclusion involved degradation of the inactivated enzyme with Pronase and subsequent analysis by solu- tion-state "C NMR. The present paper reports solid-state "C NMR experiments on lyophilized intact inactivated enzyme which are free from problems due to Pronase digestion. These experiments combine conventional cross-polarization and magic-angle spinning with selective irradiation of resonances arising from a "C double label in the steroid. Magnetization transfer between neighboring "C nuclei is used to simplify the spectra and to identify peaks due to label. The forma- tion of cysteine and lysine Michael adducts of the enzyme is established by comparisons with chemical shifts of solid model adducts. We recently described (1) the conversion of 14,15-secoestra-1,3,5(10)-trien-15-yne-3,17/3-diol (1) to ketone Z1 by human term placental 17/3,20a-hydroxysteroid dehydrogenase (estra-diol dehydrogenase). Following conversion, ketone 2 inacti-vates varied between

correlation times that characterized the mobility of the large dimer in solution. Comparison of spectra of sodium dodecyl sulfate-solubilized peptide(s) from inactivated enzyme with spectra of model amino acid-acetylenic ketone adducts suggested that a lysine residue reacted with the steroid. Poor mobility of the 13C labels in the solubilized digested sample raised concerns that Michael adducts other than lysine enaminones were present but remained undetected by solutionstate NMR because of their excessive line widths. In general, proteolytic digestion of the inactivated enzyme prior to NMR analysis is undesirable since rearrangement or loss of labile adducts could arise during sample preparation.
In this paper, we show that the complications with the solution-state NMR experiments described above are overcome by performing CPMAS 13C NMR on lyophilized intact enzyme inactivated by enzyme-generated [l3Cz]ketone 2. Differences of CPMAS spectra obtained with and without selective irradiation show only the irradiated peaks and peaks arising from magnetization transfer by spin diffusion. These difference spectra are compared to the corresponding difference spectra of model compounds to provide evidence for the formation of l3Cz-labeled cysteine and lysine Michael adducts. Red 120-agarose (type 3000-CL) were obtained from Sigma.

Purification of Estradiol Dehydrogenase
The procedure of Murdock et al. (6) was modified after the heat treatment. The protein was precipitated with ammonium sulfate (0.313 g/ml supernatant), then resuspended and dialyzed against potassium phosphate buffer (50 mM, pH 7.0, 20% glycerol). Reactive Portions of this paper (including part of "Experimental Procedures," Figs. SlGS5, Table SI, and Structures 4 and 5 ) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.
Red 120-agarose was added, and the filtered resin was batch-washed in phosphate buffer to remove unbound proteins. The washed resin was poured into a column and eluted with 0.1 mM NADP+ and 0.1 mM estrone in phosphate buffer (10 mM, pH 7.0, 20% glycerol). The eluted enzyme was precipitated with ammonium sulfate (0.313 g/ml), then resuspended in and dialyzed against phosphate buffer. Each placenta afforded 5-10 mg of homogeneous estradiol dehydrogenase with a specific activity of 2.5 0.5 units/mg protein using the enzyme and protein assays previously reported (1).

Enzyme Inactivations for CPMAS NMR Experiments
Estradiol dehydrogenase (100 mg, 250 units) was incubated with 100 p~ NAD' and 40 p~ alcohol 1 (I3Cz-enriched or natural abundance) in 0.96 liter of 0.1 M sodium carbonate/bicarbonate buffer (pH 9.2,20% glycerol, 0.5% ethanol) at 25 "C for 15-16 h ( 4 0 % of initial activity remained). The incubation mixture was concentrated by ultrafiltration to about 50 ml and dialyzed first against 1 liter of 10 mM potassium phosphate buffer (pH 7.0, 1 g/liter bovine serum albumin) for 16-17 h with one change of buffer, then against the same buffer without bovine serum albumin (1 liter) for 19-20 h with one change (all at 4°C). The contents of the dialysis bag were filtered through Whatman paper and concentrated by suction ultrafiltration to 1-2 ml. The thick solution was filtered through cotton, diluted with 2 ml of water, and lyophilized. The dried samples weighed 81-85 mg.

CYSTEINE STEROIDAL MODEL
13C CPMAS NMR 13C NMR spectra were obtained at 50.3 MHz using matched spinlock cross-polarization transfers at 38 kHz with 2-ms contact times. The samples were spun at either 3205 or 2398 Hz in a double-bearing rotor. Detailed descriptions of cross-polarization and magic-angle spinning are given elsewhere (7)(8)(9).
Selective irradiation of 13C in solids was obtained with a CPMAS-DANTE (10) pulse sequence (Fig. S1, bottom, Miniprint Supplement). A DANTE pulse train of 65 0.7-ps pulses spaced 75 ps apart was applied at the resonance frequency of the 13C peak to be selectively irradiated. Spectra were also obtained with the CPMAS-delay pulse sequence (Fig. S1, top, Miniprint Supplement) which omits the DANTE pulses. When a CPMAS-DANTE spectrum is subtracted from the corresponding CPMAS-delay spectrum, the resulting DANTE difference spectrum shows only the effects of selective irradiation and subsequent magnetization transfer (see Miniprint Supplement).
The stoichiometry of inhibitor to enzyme was calculated from the integrated intensity S of 13C CPMAS spectra (S = kCin;CJ;, where k is a constant, ni is the number of molecules of type i, and C; is the number of carbon atoms/molecule with the fractional 13C abundance fJ. One enzyme subunit and one ketone 2 molecule contain 1502 and 18 carbon atoms, respectively. The integrated spectral intensities for where nk and ne are the respective number of ketone 2 molecules and enzyme subunits. Substituting the integrated spectral intensities into the above equations gives the number of molecules of ketone 2/ enzyme subunit.  Solid-state NMR of Dehydrogenase Inuctiuation quence and then subtracting the resulting spectrum from one obtained without DANTE irradiation. Resonances in this difference spectrum result either from selective irradiation or from magnetization transfer to the selectively irradiated nucleus. The labeled inactivated enzyme was selectively irradiated at 148.6 ppm with the CPMAS-DANTE pulse sequence (see Miniprint Supplement) while spinning at 3205 Hz. The resulting spectrum (Fig. 2, bottom left), which was obtained with the variable delay r set to 30 ps, appears similar to the normal CPMAS spectrum (Fig. 1, bottom right) except for reduced intensity near 150 ppm. The difference obtained by subtracting this spectrum from the corresponding spectrum obtained without DANTE irradiation consists of a 150-ppm center band and its side bands (Fig. 2, middle left). When the experiment was repeated with T = 10 ms, the appearance of a new 117-ppm peak and the reduction of the 150-ppm peak (Fig. 2, top left) result from magnetization transfer from the carbon associated with the 117-ppm resonance to the carbon associated with the 150-ppm resonance. This rapid magnetization transfer indicates that the 117-and 150-ppm peaks arise from directly bonded 13C nuclei (12).

Solid-state
This experiment was repeated with DANTE irradiation at 156.5 ppm and with magic-angle spinning at 2398 Hz. The spinning speed was changed to prevent overlap between a side band resonance of the irradiated nucleus and the center band resonance of the neighboring 13C nucleus. (See the discussion in the Miniprint Section associated with Fig. 53). The resulting difference spectra (Fig. 2, middle and top right) indicate that magnetization was transferred from the carbon associated with the 86-ppm resonance to the carbon associated with the 157-ppm resonance, The minor 86-ppm peak in the 7 = 30 ps difference spectrum results from magnetization transfer during the 4-ms period of DANTE irradiation (see Miniprint Supplement). Chemical shifts and shift anisotropies associated with '%-labeled pairs of covalently bound carbon atoms of the inactivated enzyme are presented in Table I. 13C NMR of Model Compounds-To mimic possible adducts formed with ketone 2, a series of model adducts was generated by Michael reaction of a nonsteroidal ketone with nucleophilic side chain moieties representative of those found in amino acids (see Miniprint Supplement). In particular, the acetylenic ketone, 2,2-dimethyl-4-(l-oxo-2-propynyl)-l,3-dioxolane (3), was prepared with 99% 13C at both acetylenic carbon atoms. Chemical shifts from solution-state spectra of olefinic carbons of various nonsteroidal model adducts are presented in Table 11. The olefinic-carbon chemical shifts of the cysteine and lysine model adducts match, respectively, the shifts of the 117-and 150-ppm pair and the 86-and 15'i"ppm pair of coupled resonances in the solid-state spectra of the 13C2labeled enzyme. In addition, both the isotropic chemical shifts and the relative spinning side band intensities obtained from solid-state spectra of the 13C2-labeled steroidal model adducts ( Fig. 3) also match those of the inactivated enzyme ( Table I).   exhibited magnetization transfer consistent with the presence of a lysine adduct.

DISCUSSION
Identification of Enzyme Adducts-The solid-state 13C NMR spectra of labeled intact (undigested) enzyme, which contain resonances arising from over 1500 different carbon nuclei, are sufficiently simplified by DANTE irradiation and by differencing techniques that chemical shifts and relative side band intensities obtained from the difference spectra can be used to identify directly the %enriched pairs of covalently bound carbon atoms (Fig. 2). We believe that comparisons of these results with those obtained for model compounds (Tables I and 11) are only consistent with the formation of cysteine and lysine Michael adducts.
Examination of undigested inactivated estradiol dehydrogenase by CPMAS 13C NMR shows the presence of an enzyme-steroid cysteine adduct that was not detected in solution-state NMR experiments on the digested inactivated enzyme (2). Chemical experiments were performed subsequently with the nonsteroidal model compounds in an attempt to understand the apparent loss of the cysteine adduct during proteolytic degradation (see last paragraph under "Results"). The cysteine model adduct was shown by solution-state 13C NMR to be stable under proteolytic conditions. However, it was quantitatively converted to the lysine model adduct via a Michael addition-elimination reaction when the lysine mimic was present (data not shown). Proteolytic degradation of the inactivated enzyme creates a free amino group at each cleavage site and may also expose inaccessible lysine side chains. Accessible lysine e-amino groups may then react with a cysteine adduct and possibly transfer the steroid to lysine residues via a Michael addition-elimination reaction.
Mechanism of Inactivation-Chemical experiments with nonsteroidal model compounds show that j3-thioenones are converted to B-enaminones in the presence of an amine. These results suggest that there are two ways for lysine adducts to form during enzyme inactivation. First, lysine adduct formation may involve the initial bonding of ketone 2 to a cysteine residue followed by the transfer of the ketone 2 to the Cnitrogen of a lysine residue via a Michael addition-elimination reaction. The cysteine residue would then be free to bond to a different ketone 2 molecule. The other possibility is that cysteine and lysine residues are independently modified by ketone 2. The experiments reported here make no distinctions between these two paths, each of which might reduce the enzyme's catalytic activity.
Stoichiometry-In the present study, a stoichiometry of 4.2 f 0.3 eq of steroid/subunit was calculated from the NMR experiments. Previous experiments using tritiated alcohol 1 gave a stoichiometry of one to two Michael adducts/enzyme subunit (1). Since solid-state 13C NMR measurements on proteins are, in general, quantitatively reliable (see Miniprint Supplement), systematic errors in the NMR stoichiometric measurement are restricted to chemical impurities in the enzyme sample. However, we were unable to find significant concentrations of impurities in our samples. The ultrafiltrate arising from a preparation of inactivated enzyme was examined by solution state 13C NMR and found to contain only excess ['3C2]alcohol 1 (& = 83 and 84 ppm) and glycerol adducts with [l3C2]ketone 2 ( 6~ = 41 and 101 ppm). No significant intensities with these chemical shifts are present in the CPMAS difference spectrum of inactivated enzyme (Fig. 1, upper right). We conclude that no more than about 10% of the 13C signal from label can arise from impurities.
Stoichiometric determinations by radiometric methods usually involve separate protein and radioactivity measurements. Since only microgram guantities of enzyme were used in our radiolabeling experiments, significant errors in stoichiometry may have arisen either from sample loss due to nonspecific binding of the protein (possibly to glass surfaces) or from incomplete solubilization during scintillation counting. We plan to resolve uncertainties regarding stoichiometry by performing solid-state 13C NMR and radiolabeling experiments on a single sample of enzyme inactivated with 13C-and 3Hlabeled ketone 2.

Solid State NMR Obsewation of Cysteine and Lysine nichael Adducts of Inactivated Estradiol Dehydrogenase
Richard J. AUChus, Douglas F. Covey, Vincent Bork. and Jacob Schaefer I" spinnL;y bykids provides a method t o probe the chemical environment of speclfic C-C linkages in complex materials. The 13C-13C 1inNaqe is unambiguously identified by the magnetization transfer observed after selective irradiation of one Of the 13C labels. The chemical shifts Of the 13c-13c oair reflect their bondina t o Other atoms. In the ~S e 5 e n t Selective Irrediation ~n Spuming Solids ----Selective ztradlaitkon study, the dbuble-labeled material is-a SBCosterOid enzyme inactlvator and the s m p l e -~ molecule with the unknown binding 51te is the enzyme, estradml dehydrogenase. selective lrradiation in solids is obtained by cornbinlng DANTE pulse sequence5 with conventional CPXAS technlgues. The DANTE-CPMAS pulse sequence (Figure 51 bottom1 begins with a lH-llC  respectively) and t i e Coupling constant Jcc were obtained (Table SI) from the A0 or AX patterns observed for the ad)acent 13c nuclei. The Stereochemistry about the C=C bond in each adduct was determined from spectra prepared with natural abundance ketone 1 1 6 ) . -4 X"