Key substrate recognition residues in the active site of a plantcytochrome P450, CYP73A1 Homology model guided site-directed mutagenesis Guillaume A.. A CYP73A1 homology model based on P
Trang 1Key substrate recognition residues in the active site of a plant
cytochrome P450, CYP73A1
Homology model guided site-directed mutagenesis
Guillaume A Schoch1, Roger Attias2, Monique Le Ret1and Danie`le Werck-Reichhart1
1 Department of Plant Stress Response, Institute of Plant Molecular Biology, Universite´ Louis Pasteur, Strasbourg, France;
2
Laboratoire de Chimie et Biochimie Pharmacologiques et Toxicologiques, Universite´ Paris V, 45 Paris, France
CYP73 enzymes are highly conserved cytochromes P450 in
plant species that catalyse the regiospecific 4-hydroxylation
ofcinnamic acid to form precursors oflignin and many other
phenolic compounds A CYP73A1 homology model based
on P450 experimentally solved structures was used to
iden-tify active site residues likely to govern substrate binding and
regio-specific catalysis The functional significance of these
residues was assessed using site-directed mutagenesis Active
site modelling predicted that N302 and I371 form a
hydro-gen bond and hydrophobic contacts with the anionic site or
aromatic ring ofthe substrate Modification ofthese residues
led to a drastic decrease in substrate binding and metabolism
without major perturbation ofprotein structure Changes to
residue K484, which is located too far in the active site model
to form a direct contact with cinnamic acid in the oxidized enzyme, did not influence initial substrate binding However, the K484M substitution led to a 50% loss in catalytic activity K484 may affect positioning ofthe substrate in the reduced enzyme during the catalytic cycle, or product release Catalytic analysis ofthe mutants with structural analogues ofcinnamic acid, in particular indole-2-carboxylic acid that can be hydroxylated with different regioselectivi-ties, supports the involvement ofN302, I371 and K484 in substrate docking and orientation
Keywords: active site; cinnamate 4-hydroxylase; homology modeling; plant cytochrome P450; site-directed muta-genesis
CYP73 designates a family of plant cytochromes P450 that
evolved with or before the evolution of vascular plants Up
to 20% ofthe woody plant biomass is processed by CYP73
enzymes to form lignin monomers, UV-shielding or insect
attracting pigments, and defensive compounds [1,2] CYP73
enzymes belong to the same subfamily, i.e share more than
55% amino acid identity, and catalyse the regiospecific
4-hydroxylation of trans-cinnamic acid into p-coumaric acid
[3–5] The importance ofthis reaction in plant biology seems
to have precluded further evolution and diversification of
the CYP73A P450 subfamily to the processing of other
endogenous metabolites CYP73A1 was one ofthe first
plant P450 genes isolated [6] Expression in yeast indicated
that the cinnamate 4-hydroxylase (C4H) activity proceeds
with a perfect coupling of oxygen consumption and
reducing equivalents to produce hydroxylated substrates
[3] CYP73A1 provides a good model for determining the
residues that control catalytic efficiency and optimal
substrate positioning in a typical plant P450 enzyme contributing to a high throughput anabolic pathway CYP73A1 is one ofthe most extensively studied plant P450 enzymes It has a quite high substrate specificity but can accommodate a diverse array ofcompounds, as far as they are structural analogues ofthe natural substrate Structural requirements for such analogues include a planar, aromatic structure, a small size ofabout two adjacent aromatic rings, and an anionic site opposite (i.e at about 8.5 A˚) to the position ofoxidative attack [7,8] A recent site-directed mutagenesis study that investigated the role of unusual residues in the most conserved regions involved in haem binding and oxygen activation [9], suggested that some are likely to contribute to the optimal coupling ofthe C4H reaction The protein residues that govern substrate recognition and orientation have not yet been identified
In order to obtain information on the orientation and positioning ofthe substrates in the active site, we have recently engineered a stable and water-soluble form of CYP73A1 that is suitable for 1H-NMR paramagnetic relaxation experiments [10] The results ofthe NMR analysis indicated that the average initial orientation of the substrates in the catalytic site ofthe resting Fe(III) protein is roughly parallel to the haem We decided to use a structure-based approach to site-directed mutagenesis in order to identify residues that affect substrate binding and turnover However, only one structure for a membrane-bound P450 protein was available [11] We thus had to rely
on a homology model based on soluble P450 structures
to predict residues that might participate in the substrate recognition and docking In this paper, we report the
Correspondence to D Werck-Reichhart, Department ofPlant Stress
Response, Institute ofPlant Molecular Biology, CNRS-UPR2357,
Universite´ Louis Pasteur, 28 rue Goethe, F-67000 Strasbourg, France.
Fax: +33 3 90 24 18 84, Tel.: + 33 3 90 24 18 54,
E-mail: daniele.werck@ibmp-ulp.u-strasbg.fr
Abbreviations: CA, trans-cinnamic acid; C4H, cinnamate
4-hydroxy-lase; IAA, indole-3-acetic acid; I2C, indole-2-carboxylic acid;
I3C, indole-3-carboxylic acid; 7MC, 7-methoxycoumarin;
NA, 2-naphthoic acid; SRS, substrate recognition site.
(Received 24 March 2003, revised 28 May 2003, accepted 2 July 2003)
Trang 2construction ofa CYP73A1 model, the identification of
residues likely to form contacts with the substrate, and the
confirmation by site-directed mutagenesis ofthe
involve-ment ofsome ofthese residues in the docking and catalysis
ofcinnamic acid The impact ofthe active site mutations on
the binding and metabolism ofcinnamic acid analogues is
also described
Experimental procedures
Chemicals
Trans-cinnamic acid (CA), trans-cinnamaldehyde,
3-acetic acid (IAA), 2-carboxylic acid (I2C),
indole-3-carboxylic acid (I3C), 7-methoxycoumarin (7MC),
2-naphthoic acid (NA), phenylpyruvic acid, NADPH and
umbelliferone were from Sigma-Aldrich (l’Isle d’Abeau
Chesnes, France) trans-Cinnamylic alcohol and
6-hydroxy-2-naphthoic acid were from Lancaster Synthesis
(Stras-bourg, France).L(–)-Phenylalanine and naphthalene-1-acetic
acid were from Merck (Schuchardt, Germany)
2-Amino-quinoline and 2-phenoxyacetamidine were from Maybridge
(Tintagel, UK), 5-hydroxy-2-indolecarboxylic acid was
from Acros Organics (Noisy-Le-Grand, France),
trans-[3-14C]cinnamate was from Isotopchim (Ganagobie,
France) 4-Propynyl-oxybenzoic acid was a gift from
W Alworth (Tulane University, New Orleans)
Mutagenesis
The modified CYP73A1 cDNAs were generated using
QuickChangeTM Site-Directed Mutagenesis (Stratagene)
using as a template the double-stranded wild-type
CYP73A1 cDNA from Helianthus tuberosus (GenBank
Z17369) subcloned as an EcoRI–BamHI fragment into the
shuttle vector pYeDP60 [12] and the primers listed in
Table 1 PCR mixtures (40 lL) contained 250 lMofeach
dNTP, 0.5 lM each primer, 30 ng template DNA, 2.5 U
Pfu DNA polymerase (Stratagene), 20 mM Tris/HCl
pH 8.75, 10 mM KCl, 6 mM (NH4)2SO4, 2 mM MgSO4,
0.1% Triton X-100 and 10 lgÆmL)1BSA The polymerase
was added after preheating for 2 min at 95C Thirteen
cycles ofamplification (90C, 1 min; specific annealing temperatures for each set of primers given in Table 1, 90 s and 72C, 22 min) followed by 10 min extension at 72 C Parental methylated DNA was selectively digested with DpnI before transformation of Escherichia coli The inserts ofthe selected neosynthetized vectors were fully sequenced
As neosynthetized DNA is not a template for the reaction, the amplification is linear, which is expected to keep the error frequency low in the final PCR product Two problems were, however, encountered in our experiments: additional mutations around the site ofmutagenesis and a large proportion ofwild-type vectors were frequently obtained As controls showed that the parental DNA was digested, this was attributed to poor primer synthesis or correcting properties ofthe polymerase
Yeast expression and microsome preparation The pYeDP60 vector [12] and the modified strain of Saccharomyces cerevisae W(R) over-expressing its own NADPH-P450 reductase were used for the expression of the constructs [13] Yeast transformation was performed as described in [14], growth and induction were based on the high density procedure described in [15] To achieve optimal expression, a yeast colony grown on an SGI plate was tooth-picked into 50 mL SGI and grown for 18 h at 30C
to a density of6· 107cellsÆmL)1 This preculture was diluted in YPGE to a density of2· 105cellsÆmL)1, and grown for 30–31 h until it reached a density of 8· 107 cellsÆmL)1 Protein expression was induced by addition of 10% aqueous solution ofgalactose at 200 gÆL)1 Final density after 17 h of induction at 28C was routinely around 2· 108 cellsÆmL)1 Microsomal membranes were isolated by ultracentrifugation after mechanical disruption ofthe yeast cells with glass beads [15] Microsomes from W(R) transf ormed with void pYeDP60 were used as a negative control
Spectrophotometric measurements and catalytic activity P450 content was calculated from CO-reduced vs reduced difference spectra [16] Low to high-spin conversion and
Table 1 PCR primers used for site-directed mutagenesis The DKR primer, meant to generate K248T/R249M double mutants, actually produced the D247E/K248T/R249M and K248T/R249M/I371K triple mutants.
N302F 5¢-CTTTACATTGTTGAATTCATCAATGTTGCAGC-3¢ 5¢-GCTGCAACATTGATGAATTCAACAATGTAAAG-3¢ 43
FEBS 2003 Key residues for substrate recognition in CYP73A1 (Eur J Biochem 270) 3685
Trang 3dissociation constants ofenzyme–ligand complexes were
evaluated from Type I ligand binding spectra using the
epeak-trough¼ 125ÆmM )1Æcm)1 [7] Integrity ofthe enzyme
was checked at the end ofeach titration experiment by
recording a difference spectrum of the CO-reduced protein
Cytochrome c reductase activity ofthe
NADPH-cyto-chrome P450 reductase was assayed as in [17] Trans-CA
hydroxylation was assayed using radiolabelled
trans-[3-14C]CA and TLC analysis ofthe metabolites [18] For
determination ofthe kinetic constants, data were fitted using
the nonlinear regression programDNRPEASYderived from
DNRP53 [19]
I2C and I3C hydroxylations were assayed in a total
volume of200 lL 100 mM sodium phosphate pH 7.4
containing 600 lM NADPH, 100 lM substrate and 70 lg
yeast microsomal protein Incubations, at 27C f or 20 min
for measurement of catalytic activity and for 90 min for
products identification, were stopped by the addition of
20 lL 4 N HCl Reaction products were extracted three
times with two vols ethyl acetate, the organic phases were
pooled and evaporated under argon The residue was
dissolved in acetonitrile, water, acetic acid (10 : 90 : 0.2,
v/v/v) and analysed by reverse-phase HPLC (LiChrosorb
RP-18 Merck, 4· 125 mm, 5 lm); flow rate 1 mLÆmin)1;
5 min isocratic, then 20 min linear gradient from 10 to 52%
acetonitrile Negative controls incubated with W(R) yeast
microsomes were used to test for CYP73-independent
reactions and to evaluate extraction yields Product elution
was monitored by photodiode array detection Retention
times ofI3C and its oxygenated product were 13 and
6.5 min, respectively Products ofI2C incubation were
collected, evaporated and submitted to MS analysis on a
BioQ triple quadrupole (Micromass)
Phenylalanine, which is insoluble at pH 7.4, was dissolved
in sodium borate 100 mM pH 8.3 Phenylalanine and
2-phenoxyacetamidine hydroxylations were assayed by
HPLC under similar conditions as I2C and I3C, excepted
for phenylalanine mobile phase (isocratic 5% acetonitrile,
7.5 mM(NH4)2PO4, 7.5 mMHCl) NA hydroxylation was
assayed by fluorometry [7] in 2 mL 100 mM sodium
phosphate pH 7.4 containing 0.2, 0.5 or 1 mg yeast
micro-somal protein, 600 lM NADPH, and 100 lM substrate
Product formation was monitored for 10 min at 30C
7MC hydroxylation was assayed as in Werck-Reichhart
et al.[20] with 1 mg microsomal protein in the assay
Modelling programs and calculations
Calculations were carried out on an Indy Silicon Graphics
computer Common structural blocks were determined
previously using the GOK interactive program [21] Side
chain atoms determination, distance and dihedral
con-straints calculation, rotamer selection, and data analysis
were performed by writing macros in BCLlanguage from
Accelrys (MSI), inAWKlanguage, inUNIXmacros, and by
using the functionalities of INSIGHTII and BIOPOLYMER
modules from Accelrys The programDYANAthat calculates
the initial minimized model, was designed for NMR
applications It was modified (mainly the array sizes) in
order to handle the large number ofconstraints generated
by this method (about 35 000 constraints were kept for the
present application) The input data to the modified
program are then no longer NMR constraints, but geometrical distances and torsions derived statistically from the templates.DYANAminimization includes Van de Waals’ interaction calculations, and proposes its best solution from
a starting conformation
Structures were analysed by using thePROCHECKpackage and Accelrys INSIGHTII Model minimization was further refined with the functionalities of Accelrys DISCOVER3 (version 97.0, Force Field CVFF and ESFF when including the haem iron atom) At this stage, electrostatic interactions are included in the minimization process At each ofthe modelling steps, models are selected on the basis ofquality scores supplied by the related program (ffactor inDYANA,
or PROCHECK G-factors scoring ideally above )0.5 for instance)
Construction of the models Homology models ofcytochrome P450 CYP73A1 were constructed using building blocks corresponding to com-mon P450 three-dimensional substructures (or comcom-mon structural blocks) ofthe four structures (P450BM3, P450CAM, P450TERP, and P450eryF) available from the Brookhaven PDB at the start ofthis work as entries 2HPD, 3CPP, 1CPT and 1OXA, respectively Common structural blocks were determined for the four structures by Jean et al [21] using the programGOKand the related strategy For specified tolerance parameters, this program performs a multiple structure comparison from internal coordinates (we used Alpha, Tau) Consensus sequence ofthe blocks were then independently located in CYP73A1 using a multiple alignment ofthe available CYP73 sequences For assigning three-dimensional coordinates to the common structural blocks in CYP73A1, we used a procedure implemented for modelling the CYP2Cs [22], and based
on the adaptation ofa technique designed for deriving structures from NMR data [23]
The atoms ofthe side chains showing identical spatial location when superimposing each set ofresidues were considered as conserved atoms They were identified and added to the list ofthe block backbone conserved atoms These side chain atoms also provide the resulting rotamer value for the related target residue Other rotamers, for residues with no conserved side chain atom, were attributed
by using a rotamer library [24]
From the three-dimensional coordinates ofthe common structural blocks, we derived a set ofgeometrical constraints (mean distances between two atoms, mean Phi and Psi values), and their standard deviations The distance cutoff between two atoms was set to 5 A˚, except for interblock CB atoms where no cutoff was given in order to reflect the more flexible relative location ofthe blocks These constraints constitute, within a tolerance interval, the spatial informa-tion that was used to build the model TheDYANAprogram was then used to calculate initial random coordinates ofthe target protein and performed minimization under this set of distance and dihedral constraints [25] The loops between the blocks were built with no constraints From each model, Phi and Psi additional constraints for nonconserved residues were derived in order to restrain them in an allowed region ofthe Ramachandran region.DYANA was then rerun and proposed a family of models Minimization refinements and
Trang 4docking were performed for a set of selected models The
PROCHECK program [26] was applied as a help to the
selection ofthe models
Results
Modelling the active site of CYP73A1
CYP73A1 does not show strong identity with any ofthe
P450 proteins that have been crystallized Also, it does not
reliably align with the sequences ofknown structures in
areas other than the most conserved regions common to all
P450 enzymes A CYP73A1 homology model was built
using the computational strategy, previously described by
Jean et al [21] and further improved by Minoletti [22], that
identifies substructures, or structurally conserved blocks, in
the crystal structures ofrelated proteins, and then locates
similar blocks in the target sequence Common structural
blocks ofthe four P450 structures (P450BM3, P450CAM,
P450TERP, and P450eryF) available from the Brookhaven
PDB at the start ofthis work were located in the CYP73A1
sequence as represented in Fig 1 Common structural
blocks were used to assign three-dimensional coordinates to
the corresponding blocks in CYP73A1 The resulting model
ofthe CYP73A1 core structure and active site region is
represented in Fig 2 The advantage ofthis approach is that
it merges structural information from several known
structures into the target protein rather than producing a
model that is based on a single structure All techniques are
limited by the prediction ofthe protein alignments, but
integration ofinformation from multiple structures has
some chances to be better when, as in our case, protein
identities are very low
The 6–8 A˚ distances between the substrate protons and
the haem iron were recently deduced from1H paramagnetic
relaxation experiments [10] indicate that CA initially binds
roughly parallel to the haem in the oxidized CYP73A1 The
carboxylic function, which can be replaced by other anionic
groups, was previously shown to be an essential determinant
ofsubstrate docking in the active site [7,8] An ionic or
hydrogen bond is likely to anchor CA to a cationic or
hydrophilic residue ofthe protein These data suggest that a
set ofresidues within 5–9 A˚ above haem iron could be
considered putative active site contacts and tested by
site-directed mutagenesis A search ofthe model for hydrophilic
residues likely to form a hydrogen bond with the substrate
pointed to N302 in the I helix as a good carboxylate binding
candidate as it is one turn away from the so-called oxygen
groove A set ofcationic residues that were predicted to
reside in the substrate-binding regions (substrate recognition
sites or SRS [27]), in particular SRS 1, 3 and 5, on the basis
ofa multiple alignments with bacterial and mammalian
enzymes, were also chosen for mutagenesis to circumvent
model-prediction inaccuracy Based on SRS predictions, the
modified cationic residues included R101, R103, K248,
R249, R366, R368 and K484 Only K484 was predicted in
the substrate pocket in our model However, its distance to
the haem seemed too large to allow direct interaction with
the substrate anionic site
Hydrophobic contacts with the aromatic ring ofthe
substrate were also investigated A306 modification was
previously shown to adversely affect the binding of
cinna-mate and the coupling ofthe hydroxylation reaction [9] This effect was probably due to a direct interaction of its side-chain with the aromatic ring ofthe substrate Our model supports this hypothesis The model predicts that I371 is another residue in close proximity to the substrate I371 aligns with F361 in the limonene 6-hydroxylase, a residue that was shown to control the regioselectivity of limonene hydroxylation by CYP71Ds [28] Finally, I303 is located close enough to the putative substrate pocket to form a hydrophobic contact However, such a contact would be precluded in the hypothesis ofa van der Waals’ interaction with I371 and a hydrogen bond to N302 Substitute residues were chosen to alter charge and hydrophilicity with minimal change alteration to side chain
Fig 1 Predicted location of the conserved structural blocks and SRSs
on the primary sequence of CYP73A1 Sequence alignments of CYP73A1 with the common structural blocks offour bacterial crystal structures (P450 BM3 , P450 CAM , P450 TERP , and P450 eryF ) predicted some ofthe substrate recognition sites regions SRS locations were corroborated on the basis ofa multiple alignment with the four bac-terial enzymes also including some members ofthe CYP2 and CYP73 families CYP73A1 putative SRSs determined on the basis of this alignment are underlined (numbered 1–6 from N to C terminal) and residues selected for directed mutagenesis are indicated by stars The region interblocks in CYP73A1 are displayed in grey For the bacterial sequences only the common structural blocks are represented, the identity between sequences is shaded in black, similarity is shaded in grey (threshold of70%).
FEBS 2003 Key residues for substrate recognition in CYP73A1 (Eur J Biochem 270) 3687
Trang 5bulk, except in the case ofhydrophobic contacts for which
the influence ofside-chain size was investigated The
consecutive residues K248 and R249 were modified
simul-taneously to avoid charge compensation As the desired
double mutations were not obtained, we analysed the triple
mutants D247E/K248T/R249M (DKR) and K248T/
R249M/I371K (KRI)
Impact of mutations on the structure and stability
of the protein
The impact ofamino acid substitutions on protein stability
was investigated using the initial CO-difference spectra of
the reduced enzyme to quantify amounts of properly folded
protein with correct incorporation ofhaem The time- and
temperature-dependent disappearance ofthe peak at
450 nm was monitored as well as any conversion ofP450
into P420 that would reflect disruption ofthe haem–thiolate
bond [29], to test the stability ofthe core structure High and
fast P450 disappearance usually correlated with decreased
yeast expression and indicated a link between improper
folding or stability loss and expression levels of the mutant
protein
Immunoblot quantification ofthe apoprotein content in
yeast microsomes using antibodies raised against purified
CYP73A1 [30] revealed decreases in polypeptide expression
ofthe mutants that did not exceed 40% compared to the
wild-type construct Carbon monoxide difference spectra
detected the presence ofhaem in all ofthe mutants,
although very low CO-binding was obtained with R366M,
R101M or for the triple mutants (Table 2) The
modifica-tions to residues I303, I371, R103 and K484 did not appear
to affect the production of haem protein
P450 disappearance followed pseudo-first order kinetics Under standard conditions, i.e when P450 spectra were recorded in the presence of0.5 mgÆmL)1sodium dithionite and 30% glycerol, the half-life (t1/2) ofthe wild-type CYP73A1 was around 3 h In the presence ofa higher sodium dithionite concentration (4.5 mgÆmL)1), the t1/2of CYP73A1 was 45 min when the buffer contained 3% glycerol, and 60 min with 30% glycerol Stability tests were performed at 4.5 mgÆmL)1dithionite, using different con-centrations ofglycerol depending on the stability ofeach mutant (Table 2) The results identified three classes of mutants The first group consisted ofN302D, I371F and I371K, that had a stability at least equal to that ofthe wild-type The second group included K484M, with a stability that was slightly decreased compared to the wild-type, and R103M, N302F and I371A that displayed a more pro-nounced decrease with a t1/2shift from 45 to approximately
15 min The third group, included all other mutants in particular R366M and R101M, which demonstrated a drastic loss in protein stability The R366M, R368F/K, R101M, R103M/E, DRK and KRI modifications resulted
in a very significant disruption ofthe tertiary structure of the CYP73A1 protein
Effect of mutations on cinnamic acid recognition and metabolism
The impact ofthe mutations on CA binding and metabo-lism was investigated (Table 2) The K484M mutation that
Fig 2 A preliminary model of the active site of CYP73A1 Construction ofthis first model was based on four bacterial crystallized structures Only part ofthe active site is shown Based on the1H-NMR data [10], the substrate is expected to be located between the spheres Generated by using SWISS - PDB viewer and rendered with POV - RAY
Trang 6Table 2 Impact of the mutations on protein stability and CA recognition and metabolism Expression levels were calculated from CO-difference spectra The initial proportion of P420 was estimated from these spectra Stability ofthe haem protein was assayed by monitoring the disappearance ofthe 450 nm peak from the CO difference spectra in recombinant yeast microsomes reduced with high sodium dithionite (4.5 mgÆmL)1) The half-lives (t1/2) calculated from the pseudo-first order kinetics of P450 decrease are reported Spectra were recorded every 0.5 or 1 min during 30 min (1) Microsomes were incubated at
30 C in sodium phosphate 100 m M pH 7.4 containing 3% glycerol (2) Low stability mutants were tested in buffer containing 30% glycerol to underline the differences between them C4H activity was measured using a concentration ofcinnamate (150 l M ) expected to be saturating for most ofthe mutants The binding constants were calculated from the amplitude ofthe type I difference spectra induced by increasing concentrations ofsubstrate, etype I being the molar absorption coefficient ofthe saturated P450-substrate complex (DAmax/P450 concentration) and Ks the dissociation constant Expression and activity values are relative to the wild-type (100%): P450 expression, 847 pmolÆmg)1microsomal protein; C4H activity, 287 pkatÆmg)1; cytochrome c reductase activity, 1520 pkatÆmg)1 Cytochrome c reductase activity is used as a control for protein induced expression and integrity Values ± SD are the mean of three or more experiments n.m not measurable.
Hydrophobic and hydrogen bonding residues Positively charged residues
Wild-type N302F N302D I303A I371F I371A I371K Wild-type R101M R103M R103E DKR KRI R366M R368K R368F K484M Yeast
expression
level (%)
100 ± 4.8 30 ± 0.3 71 ± 7.2 96 ± 5.1 95 ± 9.9 93 ± 2.8 105 ± 2.4 100 ± 4.8 11.2 ± 0.7 60 ± 3.5 78 ± 1.7 7.8 ± 0.2 11.6 ± 0.8 <5 81 ± 3.3 54 ± 4.3 89 ± 1.4
Initial P420
(%)
t1/2 (min) (1) 46 ± 7.5 13.8 ± 3.9 45.1 ± 5.7 5 ± 1 53.2 ± 7.1 12.7 ± 0.8 53.9 ± 7.6 46 ± 7.5 2 15 ± 4.7 2 <1 2 n.m 2 2 30.7 ± 0.6
C4H activity
(%)
100 ± 1.0 0.5 ± 0.2 10 ± 0.9 75 ± 4.6 0.09 ± 0.02 11.3 ± 1.5 1.1 ± 0.2 100 ± 1.0 0.2 ± 0.3 44 ± 4.5 36 ± 5.8 1.0 ± 0.1 0.1 ± 0.05 0.1 ± 0.1 60 ± 3.5 48 ± 4.1 55 ± 6.1 Cinnamate
binding
Ks (l M ) 7.1 ± 1.0 13.7 ± 2.6 45 ± 9.0 3.9 ± 0.3 >100 25 ± 3.0 11.1 ± 2.2 7.1 ± 1.0 no type I 16.7 ± 2.1 >50 5.2 ± 1.6 11.9 ± 0.7 >100 11 ± 0.5 11 ± 0.6 5.9 ± 0.2 etype I
(m M )1 Æcm)1)
128 ± 7.5 23 ± 4.7 7.5 ± 1.0 106 ± 3.0 1.0 ± 0.5 25 ± 1.5 15.6 ± 2.3 128 ± 7.5 – 120 ± 9.9 103 ± 2.2 35 ± 3.8 23 ± 9.2 n.m 133 ± 6.6 126 ± 5.7 112 ± 0.9 Cyt c reductase
activity (%)
100 ± 5.0 98 ± 6.9 166 ± 5.9 107 ± 4.9 101 ± 7.2 146 ± 13 106 ± 12 100 ± 5.0 91.1 ± 21 136 ± 14 119 ± 17 83 ± 1.7 100 ± 13 118 ± 6.0 102 ± 7.3 98 ± 6.4 125 ± 9.8
Trang 7did not significantly affect protein expression or stability
had no significant impact on the binding ofCA; however, it
did result in a 45% decrease in catalytic activity All other
modifications ofpositively charged amino acids adversely
affected expression and/or stability of the enzyme but had a
comparatively minor affect on substrate recognition and
metabolism Exceptions included R366M, R101M and the
triple mutations for which drastic decreases in stable haem
protein were paralleled by dramatic losses in activity
Despite the loss ofactivity and structural integrity, the
DKR mutation rather unexpectedly seemed to retain an
intact affinity for substrate binding
Modifications ofN302 and I371 resulted in limited or no
apparent perturbation ofprotein folding and stability but
led to dramatic decreases in CA binding and hydroxylation
N302 is likely to provide a hydrogen bonding side chain for
anchoring the carboxylate ofCA The conversion
ofaspa-ragine into negatively charged aspartic acid (N302D)
resulted in a drastic effect on substrate binding affinity
Whereas replacement with a bulky hydrophobic residue
(N302F) compromised overall protein structure and
cata-lysis
I371 is predicted to form a van der Waals’ contact with
the aromatic ring ofCA In the I371 mutants, I371A opens
more space in the active site and thus should allow for
increased substrate mobility Conversely, I371F and I371K
should create a steric hindrance to the binding ofthe
substrate above the haem iron As expected, the I371A
mutation substantially decreases CA affinity and the ability
to desolvate the active site Around 10% ofthe catalytic
activity is conserved, which would be in agreement with the
conservation ofthe carboxylate anchoring function ofthe
protein The I371F and I371K mutations lead to an almost
complete loss in C4H activity This activity loss is correlated
with impaired substrate binding A complete loss ofbinding
was also observed upon substitution ofI371 with the bulky
phenylalanine The insertion ofa positive charge in the 371
position does not completely prevent the binding ofthe
substrate, but almost totally hinders catalysis This probably
results from improper positioning of the substrate’s
aro-matic ring above the haem iron
Mutation ofI303, adjacent to N302, into alanine slightly
increased affinity but modified substrate positioning and
decreased catalytic activity This data is concordant with a
model where I303 is not a direct contact residue, but rather
contributes to optimal CA orientation in the active site
Binding of alternate ligands to CYP73A1 mutants
The mutants that showed strongly impaired CA binding
and metabolism, but that did not display a major structural
alteration in terms ofprotein stability and expression, were
further tested for their ability to recognize a set of structural
analogues ofCA This set included CA precursors, plus
other natural and synthetic compounds Some ofthese
compounds present a quite high intrinsic affinity for
wild-type CYP73A1, such as phenylpyruvic acid (Ks¼ 3.1 lM),
phenylalanine, indole-2-carboxylic acid or cinnamyl alcohol
(Ks¼ 12 lM), 2-aminoquinoline (Ks¼ 17 lM) and
indole-3-carboxylic acid (the natural auxin, Ks¼ 18 lM) These
compounds are ordered from gain to loss of binding to the
mutant proteins in Table 3
As shown in Table 3, the analogues investigated were better ligands for the mutants than the physiological substrate CA Relative to wild-type CYP73A1, the binding efficiency for CA decreases 10-fold in the mutant I371K, 50-fold in I371F and 100-fold in N302D In contrast, increases in binding efficiency are observed for a few ligands after modification of the protein The most notable increases are 15-fold for N302D with phenylalanine, 12-fold for I371K with 2-phenoxyacetamidine, and 10-fold for I371F with phenylalanine or cinnamylic alcohol
The I371F modification is likely to block access to the active centre for most of the potential substrates Only compounds with increased side chain flexibility or reduced bulkiness in the CA ring region are expected to have increased binding efficiencies compared to CA This is actually the case, with a gain in binding efficiency being observed only for phenylalanine, 2-phenoxyacetamidine, cinnamylic alcohol or 4-propynyl-oxybenzoic acid More relevant are the N320D and I371K mutations that could provide a new salt-bridge or hydrogen bonding opportu-nities in the active site region Increased affinity of several ligands indicates that new bonds are formed in the mutants and may reflect a reversed orientation ofthe ligands or occupation ofdifferent subpockets in the active site Noteworthy are the increased binding ofphenylalanine and 2-phenoxyacetamidine, which are highly polar mole-cules However, none ofthe compounds that displayed an increased affinity produced a large spin transition of the ferrous haem, which would be indicative of effective desolvation ofthe active site and appropriate positioning for an efficient oxidative attack Some of the analogues listed at the top ofTable 3 that showing better binding efficiencies than CA with the modified proteins, were analysed in catalytic assays
Metabolism of alternate substrates The metabolism ofCA analogues was assayed with the N302, I371 and K484 mutants (Table 4) Microsomes from yeast transformed with the empty expression plasmid, and also incubations without NADPH were used to control for CYP73-independent reactions No metabolism ofphenyl-alanine and 2-phenoxyacetamidine was detected with the wild-type or any ofthe mutants The sensitivity ofthe tests was low, due to high detection thresholds and the need to test phenylalanine metabolism at pH 8.3, which decreases C4H activity ofthe wild-type by 80%
NA was previously shown to be the best structural mimic and alternate substrate for wild-type CYP73A1 [7] NA was metabolized by all mutants with an efficiency very compar-able to that observed with CA This suggests that both compounds have a very similar positioning in the active site and validates use ofNA for fluorometric quantification of the enzyme activity [7] Metabolism ofI2C, I3C and 7MC does not parallel that ofCA in the different mutants For example the I371A and I371K mutations have less influence
on demethylation of7MC than on CA hydroxylation Also noteworthy is the opposite effect of several amino acid substitutions on I3C and I2C hydroxylations Most muta-tions have less impact on I2C than on I3C and CA metabolism, probably due to the symmetry axis ofI2C and
to the possible attack on two different carbon atoms
Trang 8Unexpectedly, the K484M substitution, which results in
close to 50% loss in C4H activity, does not affect I2C
hydroxylation As initial cinnamate binding is not
influ-enced by this mutation (Table 2) and binding kinetics are
first-order (indicating a single binding-site), this suggests
that K484 does not directly affect catalysis but might have a
selective role in substrate position adjustment during the
catalytic cycle
Modified regiospecificity of indole-2-carboxylic acid
hydroxylation
I2C metabolism by wild-type CYP73A1 was previously
shown to result in the formation of two products that were
not further characterized [7] On the basis of its HPLC
retention time, UV spectrum, and monoisotopic mass, the
most polar product P1 (RT 9.8 min) was unambiguously
identified as 5-hydroxy-I2C (Fig 3) P2 presents the same
mass as P1 and is thus a monohydroxylated product
Superimposition ofthe NA and CA structures, and oftheir
positions ofattack on those ofI2C, indicates that P2 is most
likely 6-hydroxy-I2C, although an authentic standard was not commercially available for verification In favour of the latter hypothesis, 5-hydroxy-I2C was tested as a substrate of CYP73A1 and was not further metabolized
As preliminary experiments indicated that the ratio between the two products varies upon metabolism by the different mutants, this ratio was used as a reporter of the influence ofthe mutations on substrate docking (Table 4)
In the wild-type CYP73A1, the formation of P2 is five times more frequent than that of P1
The K484M mutation does not significantly affect the P2/P1 ratio This is not surprising considering that it does not affect the global rate of I2C metabolism As the length ofthe I2C molecule and the distance between its carboxylate and the positions ofattack are slightly shorter than for CA
or NA, it is possible that the carboxylate ofI2C is beyond the area ofinfluence ofK484
The N302 mutations, in particular N302D, significantly increased the proportion ofP1 so that the P2/P1 ratio dropped closer to 1 This loss in regiospecificity in the mutant is concordant with the increased mobility ofthe
Table 3 Alternate ligand binding to the mutant protein 4-Propynyl-oxybenzoic acid and wild-type CYP73A1 is the only complex for which data fitting with the Michaelis–Menten equation indicated second order kinetics Binding efficiency is the e type I /K s ratio calculated for each complex The values listed are relative to the wild-type for each ligand Standard deviations (not shown) are less than 12% of these values.
FEBS 2003 Key residues for substrate recognition in CYP73A1 (Eur J Biochem 270) 3691
Trang 9molecule in the active site that is reflected by a low etype I
(Table 3) Taken together, the data support a role for N302
in controlling ofsubstrate orientation in the active site
The I371 mutations to K and A have opposite effects
The I371K mutation increases the preferential attack at the
putative 6-position, most likely by increasing steric
hin-drance near C5 ofthe indole ring In contrast, the I371A
mutation appears to remove the steric constraint existing in
the wild-type CYP73A1 and favours a P2/P1 ratio closer
to 1 The observed effects of both of these mutations
support the assumption ofa direct contact ofI371 with the
aromatic ring ofI2C or CA
Discussion
The computational homology modelling strategy
des-cribed by Jean et al [21] allows a reasonable prediction
ofthe most conserved P450 substructures, although
hypervariable regions cannot be predicted Our present
model was based on four crystallized bacterial enzymes
(Fig 2) and seems to correctly predict several residues forming contacts with CA
The model predicts that N302, which resides in the I helix and SRS 4, is likely to form a hydrogen bond with the carboxylate ofthe substrate Mutations ofthis residue lead to
a dramatic loss in CA binding efficiency (10-fold for the N302F and 100-fold for the N302D substitution) together with a very strong decrease in catalytic activity This confirms
a critical role for this residue in the initial binding and correct positioning ofCA during catalysis A role ofN302 in anchoring the side-chain carboxylate ofCA is further supported by the enhanced binding ofamine substituted ligands and also by the loss ofregiospecificity ofI2C hydroxylation when N302 is replaced by an aspartic acid Together with A306, I371 is predicted to form a hydrophobic pocket that positions the aromatic ring of the substrate in close proximity to the haem iron The adverse impact ofthe A306G substitution on substrate binding and metabolism as well as coupling ofthe reaction was described previously [9] Modifications ofI371, espe-cially I371F, produced a dramatic loss in binding and activity with CA and all other substrates The less detrimental effect of these substitutions on the binding of analogues, which are less rigid or bulky than CA, and differential impact on the regiospecificity of I2C ring-hydroxylation support the hypothesis that the side chain of I371 is an essential element ensuring correct positioning and orientation ofthe aromatic ring in the active site
Our model predicts that K484 is in the substrate pocket Its distance to the CA carboxylate in the oxidized enzyme model does not allow for any direct interaction and, as expected, modification ofK484 has no impact on the initial binding ofCA However, the K484M substitution leads to a 50% decrease in catalytic activity with both CA and NA A possible explanation is that K484 plays some role in the electron transfer from the P450 reductase to the haem iron However, this residue is located on the distal side ofthe haem, while interaction with the reductase and electron transfer should involve residues on the proximal side of the protein [31] The unchanged I2C hydroxylase activity in K484M when compared to that ofthe wild-type confirms that the mutant is not impaired in electron transfer Thus, K484 must exert some control on CA/NA positioning or product release during the catalytic cycle Although the K484 effect might be indirect and the interaction with the carboxylate ofCA might occur via a molecule ofsolvent, it can also be postulated that the reduction ofthe protein or binding ofoxygen results in a conformational change ofthe
Fig 3 Analysis of the products of I2C hydroxylation Upper panel:
HPLC analysis ofthe products ofthe metabolism of10 nmol I2C by
30 pmol recombinant CYP73A1 in 60 min and in a 100 lL assay.
Absorbance was monitored at 290 nm Lower panel: UV spectra
corresponding to the centre ofthe peaks P1 and P2 collected after
90 min incubation of120 nmol ofI2C were analysed by negative
ESI-MS Monoisotopic mass ofboth compounds was 176 Da P1 retention
time and UV spectrum was identical to that ofcommercial
5-hydroxy-2-indolecarboxylic acid.
Table 4 Metabolism of alternate substrates by mutant CYP73A1s Activities are expressed relative to wild-type CYP73A1 100% activity is
287 pkatÆmg)1microsomal protein for CA, 311 pkatÆmg)1for NA, 38.8 pkat mg)1for I2C, 20.4 pkat mg)1for I3C, 6.6 pkat mg)1for 7MC n.d not determined.
Trang 10protein, similar to that observed for P450BM3or P450CAM
substrate complexes [32–34] Such a change could bring
K484 much closer to CA In this case, ion pairing or
hydrogen bond between K484 and the CA carboxylate
could control the optimal positioning and orientation ofthe
substrate for catalysis NMR measurement of the distances
ofthe substrate protons to the haem iron indicate an initial
positioning ofCA approximately 6–8 A˚ from the iron in the
oxidized enzyme, which might not be optimal for catalysis
and would not particularly favour ring 4-hydroxylation If
these measurements are correct then a structural change
that brings CA closer to the iron and adjusts substrate
position, possibly tilting the substrate so as to favour attack
at the 4 position or on the 3–4 bond, would be needed for
efficient and regiospecific catalysis The K484M mutation
has no impact on I2C metabolism or the regioselectivity of
attack This observation is compatible with a role ofK484 in
CA reorientation as the slightly smaller size and different
shape ofI2C compared to that ofCA might prevent
interaction between its anionic site and K484
N302 and I371 align with residues that have been
shown to confer substrate specificity or regioselectivity to
many other plant or mammalian P450 enzymes Residues
corresponding to I371 govern the regiospecificity ofthe
hydroxylation of4S-limonene in CYP71D18 from
spear-mint and CYP71D15 from pepperspear-mint for the synthesis
ofcarvone and menthol, respectively [28] In the
mammalian CYP2B family, residues 294 and 363 are
equivalent as N302 and I371, respectively The CYP2B
mutations were shown to affect steroid regioselectivity
At position 363, a CYP2B1 mutant (V363L) exhibited a
twofold decrease in androgen activity [35], whereas in
CYP2B11 the reverse mutant shows a fivefold increase in
androgen activity [36] The same residue was identified as
a determinant ofsubstrate specificity in CYP2B2 [37],
CYP2B5 [38] and CYP2B6 [39] Likewise, residue 294
was shown to play a key role in androgen metabolism by
CYP2B1 [40] and CYP2B4 [38] A similar affect on
catalysis by these residue positions has been reported for
other mammalian enzymes For example in CYP2A5,
mutation ofM365, the equivalent ofI371, decreased the
metabolism ofaflatoxin B1 [41], while modification of
the corresponding residue (A370) in human CYP3A4
enhanced the hydroxylation ofsteroids [42,43]
A significant portion ofthe protein, which was not
reliably predicted in the model, is not shown in Fig 2 and
was not thoroughly investigated in our site-directed
experi-ments It is therefore likely that additional residues, such as
R or K that can form an ion pair with the carboxylate of
CA, may contribute to substrate recognition or docking
Mutation ofpositively charged residues found in the
putative SRSs (Fig 1), based on a multiple alignment did
not lead to the identification ofa residue that would be
critical for the recognition or positioning of CA Mutation
ofall arginines led to a significant loss in protein stability
suggesting that they are involved in protein fold structure
rather than binding ofthe substrate
The overall picture ofthe CYP73A1 active site provided
by our data is reminiscent ofP450BM3[44,45], as it involves
a hydrogen bond and possibly an ion pair for the anchoring
ofthe carboxylate on the substrate, and also a major
hydrophobic region for the docking of the aromatic ring As
in P450BM3, a substantial protein rearrangement must occur during the catalytic cycle [32,33], probably upon reduction,
to ensure an optimal positioning ofthe substrate relative to the ferryl-oxo intermediate for coupled, regiospecific attack ofthe ring at the 4 position While mutant analysis was in progress, the first X-ray structure was described for a membrane-bound mammalian P450, CYP2C5 [11] This new structure confirmed the conservation ofthe P450 spatial organization in eukaryotic microsomal enzymes The position ofSRS 4 that is located in the centre ofthe I helix, which includes N302 in CYP73A1, was highly conserved relative to the haem However, significant local changes were detected, particularly in all other SRSs For example, SRS 5, facing the I helix, shows a double bend due to two proline residues (P360 and P364) The resulting topology orients three leucine side chains toward the active site (L358, L359 and L363) In P450CAM[46] and P450TERP[47], SRS 5
is a b-strand partially involved in b-sheet formation with SRS 6 In P450BM3[44], the first bend found in CYP2C5 is present and the C-terminal part ofSRS 5 is a b-strand not involved in a b-sheet with SRS 6 The alignment ofSRS 5 ofCYP2C5 and the whole CYP2B family with those of CYP73A1 and CYP71Ds is not ambiguous The two prolines and the adjacent positive charge (H365) that bind the haem propionate in CYP2C5 are conserved This suggests that the double bend structure is present and confirms I371 as a central residue ofSRS 5 in CYP73A1 If the position ofthe SRS relative to the haem is conserved, the phenyl side chain in the I371F mutant should stack over the haem, which would explain the complete impairment of substrate binding and the increased stability ofthe mutant protein The orientation and size ofSRS 6 is quite variable between the different structures and reliable alignment of K484 with the crystallized sequences is not possible Consequently, the role ofK484 could not be correlated with the mammalian structure
CYP73A1 is more closely related to CYP2C5 (47% similarity) than to any ofthe bacterial proteins (36%, P450BM3; 28%, P450CAM; 27%; P450TERP; 30%, P450eryF), and the structure ofSRS 5 seems to be conserved between CYP2C5 and CYP73A1 In order to refine our understand-ing ofCYP73A1 and to gain structural information on SRS 5 topology, a new model was built based on the CYP2C5 structure exclusively (1DT6) CA was positioned
in the active site ofthis new model, taking into account the results ofthe previous NMR measurements [10] and information obtained from mutagenesis (Fig 4) In this new model, N302 easily forms a hydrogen bond with the carboxylate ofCA and I371 is well positioned for hydro-phobic contact with the substrate aromatic ring A306 was shown to be critical for substrate recognition [9] In this new model, its methyl group is 4.8 A˚ from the haem iron and 3.5 A˚ from the substrate K484 is still too far away to form a direct contact with the cinnamate
In conclusion, a combination ofhomology modelling and site-directed mutagenesis ofCYP73A1 has identified N302 and I371 as key determinants ofsubstrate binding and orientation for catalysis K484 is not involved in initial substrate binding, but seems to play a significant role in catalysis, possibly by contributing to substrate reorientation during the catalytic cycle Modification ofactive site residues improved affinity for substrate
FEBS 2003 Key residues for substrate recognition in CYP73A1 (Eur J Biochem 270) 3693