contact information for elizabeth e. howell: phone, 865-974-4507
TRANSCRIPT
Role of K32 Residues in R67 Dihydrofolate Reductase
Probed by Asymmetric Mutations†
Stephanie N. Hicks, R. Derike Smiley, Lori G. Stinnett,
Kenneth H. Minor and Elizabeth E. Howell*
‡Department of Biochemistry, Cellular and Molecular Biology
University of Tennessee, Knoxville, TN 37996-0840
Running Title: Asymmetric K32M Mutations in R67 DHFR
Keywords: tetrahydrofolate dehydrogenase, trimethoprim, site-directed mutagenesis, binding
specificity, enzyme evolution, gene quadruplication, salt bridges, ionic strength dependence, salt
effects, ionic interactions, symmetry, non-productive binding.
† This research was supported by NSF grant MCB-0131394 (to E.E.H.).
Contact Information for Elizabeth E. Howell: phone, 865-974-4507; fax, 865-974-6306, email,
JBC Papers in Press. Published on August 27, 2004 as Manuscript M404484200
Copyright 2004 by The American Society for Biochemistry and Molecular Biology, Inc.
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
SUMMARY
R67 dihydrofolate reductase (DHFR) is a novel protein encoded by an R-plasmid that
confers resistance to the antibiotic, trimethoprim. This homotetrameric enzyme possesses 222
symmetry, which imposes numerous constraints on the single active site pore, including a “one-
site-fits-both” strategy for binding its ligands, dihydrofolate (DHF) and NADPH. Previous
studies uncovered salt effects on binding and catalysis [Hicks et al., (2003) Biochemistry 42,
10569-78], however the residue(s) that participates in ionic contacts with the negatively charged
tail of DHF as well as the phosphate groups in NADPH was not identified. Several studies
predict that K32 residues were involved, however mutations at this residue destabilize the R67
DHFR homotetramer. To study the role of K32 in binding and catalysis, asymmetric K32M
mutations have been utilized. To create asymmetry, individual mutations were added to a
tandem array of four in-frame gene copies. These studies show one K32M mutation is tolerated
quite well, while addition of two mutations has variable effects. Two double mutants,
K32M:1+2 and K32M:1+4, which place the mutations on opposite sides of the pore, reduce kcat.
However a third double mutant, K32M:1+3, that places 2 mutations on the same half pore,
enhances kcat four to five fold compared to the parent enzyme, albeit at the expense of weaker
binding of ligands. Since the kcat/Km values for this double mutant series are similar, these
mutations appear to have uncovered some degree of non-productive binding. This non-
productive binding mode likely arises from formation of an ionic interaction that must be broken
to allow access to the transition state. The K32M:1+3 mutant data suggest this interaction is an
ionic interaction between K32 and the charged tail of dihydrofolate. This unusual catalytic
scenario arises from the 222 symmetry imposed on the single active site pore.
2
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
INTRODUCTION
R67 dihydrofolate reductase (R67 DHFR) is an R-plasmid encoded enzyme that catalyzes
the NADPH dependent reduction of dihydrofolate (DHF) to tetrahydrofolate (THF). Its presence
in bacteria confers resistance to the antibiotic, trimethoprim. This enzyme is not similar in
sequence or structure to the chromosomally ended DHFRs.
R67 DHFR is a homotetramer and the pore that traverses the length of the molecule is the
active site. Surprisingly the structure possesses 222 symmetry (1), which imposes numerous
constraints on binding and catalysis. For example, the symmetry requires that for each binding
site, there must be three additional, symmetry related sites. However solution studies find only
two sites can be occupied simultaneously because of steric constraints. The possible binding
combinations are two NADPH molecules, or two folate/DHF molecules or one NADPH plus one
folate/DHF molecule (2). Only the latter is productive. Thus binding of neither ligand can be
optimized and a “one site fits both” approach is employed (3,4). Another constraint arising from
the symmetry is that addition of a mutation to the gene results in 4 mutations per single active
site pore. While addition of one mutation might help DHF binding in one half the pore, it will
not necessarily help NADPH binding in the other half of the pore (or vice versa). Breaking the
222 symmetry of R67 DHFR by introduction of asymmetric mutations should help unravel how
R67 DHFR functions.
To allow introduction of asymmetric mutations, a tandem array of four, in-frame R67
DHFR gene copies has been constructed (5,6). The resulting protein product (named Quad3)
possesses a mass four times that of the R67 DHFR monomer. Each domain in Quad3
3
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
corresponds to a monomer in R67 DHFR. The linker sequence connecting the gene copies is the
natural N-terminus. The quadruplicated gene product is almost fully active with only a slight
decrease in kcat (1.6 fold) and a slight increase in Km values (1.5 fold). All other studies suggest
the quadruplicated gene product mimics the homotetramer very closely.
Which mutations should be introduced into Quad3 to allow analysis of R67 DHFR
catalysis? Previous studies have identified K32, Q67, I68 and Y69 as the most critical residues
in binding and catalysis2 (4,7-9). Therefore these residues were targeted for introduction of
asymmetric mutations. The first asymmetric mutation studied was Q67H (6). This study focuses
on the role of K32 residues and a third study considers the effect of Y69F mutations (Stinnett et
al., see companion manuscript).
What role is proposed for K32? Since both substrates are negatively charged (DHF
possesses two carboxylate groups in its p-aminobenzoyl–glutamic acid (pABA-glu) tail and
NADPH possesses a 2’ phosphate on the AMP ribose as well as the pyrophosphate bridge), the
role of counter charges was considered. The only positively charged residue in the active site is
K32 and the symmetry related positions of this residue are shown in Figure 1. To examine the
role of K32 in R67 DHFR function, mutagenesis was employed. Unfortunately, all mutants
destabilized the homotetramer, therefore salt effects were used to indirectly probe the role of
K32 (7,10). Increasing ionic strength was found to dramatically increase the Km values for both
ligands. A smaller, but significant rise in kcat was also noted, indicating ionic interactions are
involved in binding and catalysis (11,12).
4
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Further support for the importance of K32 in binding and catalysis comes from NMR
experiments which find a chemical shift in the resonances associated with K32 upon NADPH
binding (13). Also, docking of NADPH into the R67 DHFR•folate complex predicts that K32
likely forms an ionic interaction with the 2’phosphate of NADPH, while a symmetry related K32
likely interacts with the γ-carboxyl group on the glutamate tail of DHF (3). Our simple model
for binding and catalysis suggests K32 guides the negatively charged ligands to the active site
pore by establishing a positive electrostatic potential (3). Once the ligands near the active site,
symmetry related K32 residues form direct ionic interactions with both NADPH and DHF. After
formation of the ground state complex, at least one ionic interaction breaks, leading to hydride
transfer.
To evaluate the role of K32 in R67 DHFR, asymmetric K32M mutations were generated
using the tandem gene array that encodes Quad3. From studies of these asymmetric mutants, a
revised model for K32 in binding and catalysis is proposed.
EXPERIMENTAL PROCEDURES
Construction of Asymmetric Mutants Construction of asymmetric mutations in our tandem
array is straightforward as the gene copies are separated by unique restriction enzyme sites. We
mutagenize each gene copy separately and then reconstruct the tandem array. This process has
been previously described for production of asymmetric Q67H mutations (6). All K32M
mutants were verified by DNA sequencing.
5
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Growth and Purification E.coli STBLII cells are able to maintain the tandem gene array without
recombination (14). Cells containing the desired clone were grown in TB media at 30°C for 60
hours in the presence of only 200µg/ml ampicillin, since introduction of trimethoprim into the
growth media results in random mutations within the gene sequences. DNA from each of the 12
liters for the K32M:1+3 and K32M:1+4 mutants was isolated and sequenced to verify that each
culture still possessed the appropriate mutations. Cells were lysed and protein purified as
described previously (6). During the purification process, 0.1g/L polyethylene glycol 3350
(PEG) was added to all solutions since this addition minimizes aggregation (15) and leads to
higher protein yields. Mass spectrometry (intact mass by electrospray-ion trap (4)) was utilized
to verify the K32M:1+2 mutant protein had the correct molecular mass.
pH Titrations Wild type R67 DHFR is a homotetramer that undergoes a pH dependent
dissociation to 2 dimers (16). This process can be monitored by fluorescence as symmetry
related tryptophan 38 residues occur at the two symmetry related dimer-dimer interfaces (17).
At pH 8, W38 is buried (tetramer), while at pH 5 it is exposed to solvent (dimer). The pH
dependence of dissociation arises from protonation of symmetry related H62 residues that also
occur at the dimer-dimer interfaces. While Quad3 cannot dissociate due to the linker sequences
tethering each gene copy product, it can undergo a transition from a "closed" form (active
conformation, pH 8) to an "open" form (inactive, pH 4) (5,6). The effects of pH on Quad3 were
assessed in varying salt conditions at room temperature. The intensity averaged emission
wavelength, <λ>, for each emission spectrum was calculated according to Royer et al. (18). The
pH profiles were fit to a simple ionization equation and normalized according to Bradrick et al.
(5).
6
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Circular Dichroism Circular dichroism (CD) spectra were recorded at 22°C in 10 mM potassium
phosphate buffer (pH 8) with 10 µM Quad3, K32M:1+2, K32M:1+3, and K32M:1+4 mutant
R67 DHFRs using an Aviv Model 202 series circular dichroism spectrometer as previously
described (4).
Steady-State Kinetics with Mutants The steady-state kinetic behavior of each mutant was
monitored as previously described (19). Experiments were performed at 30°C in either 50 mM
MES + 100 mM Tris, + 50 mM acetic acid (MTH) polybuffer (20) or TE buffer (10 mM Tris, 1
mM EDTA, pH 7) in the presence of various concentrations of NaCl to adjust the ionic strength.
At least five subsaturating concentrations of NADPH and DHF were used to measure activity.
Global fitting of the data to an equation describing the bi-substrate kinetic reaction of R67 DHFR
used a non-linear subroutine of SAS (version 8.2, (6)). The NLINEK macro for use in SAS is
available on the internet at http://animalscience.ag.utk.edu/faculty/saxton/software.htm. The
extinction coefficient for the R67 DHFR reaction at 340 nm is 12,300 L mol-1cm-1 (21). Ligand
concentrations were calculated using an extinction coefficient of 28,000 L mol-1 cm-1 at 282 nm
for DHF (22) and 6220 L mol-1cm-1 at 340 nm for NADPH (23). Kinetic data for the K32M:1+3
mutant were obtained at 360 nm to more accurately assess the Km values. The extinction
coefficient for the DHFR reaction is 5020 L mol-1cm-1 at this wavelength. Extinction coefficients
for NADPH and DHF are 4020 L mol-1cm-1 and 2630 L mol-1cm-1 respectively at 360 nm.
Fluorescence Quenching Binding of NADPH or DHF to 2 µM protein at 4oC was monitored in
TE buffer containing 0.03M NaCl (µ=0.05) using tryptophan fluorescence (24). Data were fit to:
7
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Fl = Fo – 0.5 Fo [Ptot + Kd + Ltot – [(Ptot + Kd + Ltot)2 – 4 Ptot Ltot)1/2] where Fl is the observed
fluorescence, Ltot is the total ligand concentration, and Ptot, Kd and Fo are variables describing the
number of enzyme binding sites, dissociation constant and fluorescence yield per unit
concentration of enzyme, respectively (25).
Sedimentation velocity Sedimentation velocity experiments were conducted in a Beckman
Optima XL-I ultracentrifuge equipped with both interference and absorbance optics. Protein
samples were dialyzed into TE buffer, pH 7 with 0.03M NaCl added (µ= 0.05) and the dialysate
was used as an optical reference. Varying concentrations from 0.6 µM to 5 µM of each mutated
protein were loaded (400 µl loading volume) into double-sector charcoal-filled epon centerpieces
and sedimentation velocity was carried out at 50,000 rpm at 4°C using an An50 Ti eight-hole
rotor. Sedimentation velocity analysis was performed by direct boundary modeling by solutions
of the Lamm equation using the program Sedfit ((26), see
http://www.analyticalultracentrifugation.com). Partial specific volume, buffer density and
viscosity were determined using the software SEDNTERP (John Philo at AMGEN Corp, see
http://www.jphilo.mailway.com/download.htm).
RESULTS
Nomenclature The following system will be used to name each of the asymmetric mutants. The
residue, residue number, and mutation will be listed first followed by a colon. The asymmetric
location of the particular mutations will be indicated numerically where 1 refers to gene copy 1,
2 refers to gene copy 2, etc. For example, a single K32M:1 mutant was constructed3 where the
mutation was placed in gene copy 1. While three other single mutants could have been
8
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
constructed, the 222 symmetry of wt R67 DHFR predicts they will be equivalent. Three double
mutants were also constructed. The K32M:1+3 mutant contains two mutations: lysine 32 in gene
copy 1 has been mutated to methionine, as has lysine 32 located in gene copy 3. The K32M:1+2
and K32M:1+4 double mutants also have two mutations in their respective gene copies. The
non-equivalency of these constructs can be illustrated using the crystal structure for
homotetrameric R67 DHFR (see Figure 1) where each monomer would correspond to a domain
in Quad3 (2,27). The four domains are labeled in a counter-clockwise fashion where domain 1 is
located at the top, left position. Panels B-D in Figure 1 compare the different topologies of the
three double mutants. For the K32M:1+3 mutant, one side of the active site pore contains two
wild-type K32 residues while the opposite side of the pore contains two mutant K32M residues
(panel C). In contrast, both the K32M:1+2 and K32M:1+4 mutants contain one K32 and one
K32M residue on each side of the pore (panels B and D), although with different geometries
between the mutations. Figure 1E depicts the arrangement of the mutations with respect to a
reverse image of the active site pore and shows our simple model for bound NADPH and folate
derived from docking NADPH into R67 DHFR•FolI (3). Since the K32M mutation destabilizes
homotetrameric R67 DHFR, triple and quadruple mutants were not constructed.
Physical Studies of the Apo Proteins To examine whether the gene quadruplication event
altered the salt sensitivity of the protein, pH titrations of Quad3 were performed in MTH buffer
with various additions of NaCl (Figure 2) (7). The data were fit and the corresponding pKa
values are given in the figure legend. It is apparent that addition of salt slightly stabilizes the
protein. However, there is not a significant difference in the titrations near pH 7, where the
majority of experiments are performed.
9
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
pH titrations of the double mutants were attempted, however asymmetric addition of the
K32M mutation leads to increased aggregation. Visible turbidity usually appeared during
dialysis or during strong mixing/shaking conditions, particularly for concentrated solutions. To
minimize aggregation, the proteins were purified and maintained in buffer solutions containing
0.1g/L PEG 3350 (15). While low concentrations of protein are stable at ≤ 4oC, we were unable
to perform pH titrations of 2µM protein at room temperature.
Since aggregation was a problem with these mutants, we evaluated whether the
oligomerization state of the protein was altered using sedimentation velocity experiments.
Sedimentation velocity profiles for each double mutant at 4oC and at varying concentrations
(≤5µM) were fit using a continuous c(S) distribution model (Sedfit; (26)). The results clearly
demonstrate the presence of a single species for all samples, with S values of 2.4, 2.4, and 3.4 for
the K32M:1+2, K32M:1+3, and K32M:1+4 proteins, respectively. For all fits, conversion to a
continuous c(M) distribution model demonstrated that all three mutants are of the same
molecular mass (~30kDa). (From the gene sequence, the expected molecular mass is ~33.5 kDa,
thus these mutants are monomeric under these conditions.) It is likely that the observed
difference in the S value for K32M:1+4 reflects an alteration in its molecular shape compared to
the K32M:1+2 and K32M:1+3 mutants. Since the K32M mutations in this asymmetric mutant
both occur at what would correspond to one dimer-dimer interface in homotetrameric R67
DHFR, this “interface” appears perturbed in the apo K32M:1+4 mutant (see Figure 1D).
10
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
CD spectra for each of the K32M double mutants at pH 8 are given as supplemental data.
The signals for the K32M:1+2 and K32M:1+3 mutants show some changes with respect to
Quad3, while the spectrum for the K32M:1+4 mutant displays greater differences. These
differences could describe some level of conformational change in the proteins and/or alterations
in the contribution of aromatic residues to the signal. Aromatic residues contribute significantly
to the CD signal in R67 DHFR (17), in accord with general observations by Woody (28). Since
K32 is near W38 (closest approach ~ 4 angstroms), mutation to K32M could potentially affect
the CD signal.
All the above physical studies suggest addition of a K32M mutation(s) results in some
level of structural destabilization. Ocam’s razor suggests the simplest interpretation is alteration
of the structure at what would correspond to the dimer-dimer interface(s) in homotetrameric R67
DHFR. While only limited conformational changes cannot be proven, the steady state kinetic
data below suggest addition of ligand(s) provides sufficient contacts to restore the active site
structure.
Binding of NADPH Monitored By Fluorescence Quenching To monitor how the mutations
affect NADPH binding, a fluorescence quenching approach was used (at 4oC). This technique
only monitors binding at the first tight site, yielding Kd1 (2). The titrations are shown in Figure
3A, and the fitted values are given in Table 1. The tightest Kd1 values are associated with Quad3
and the single mutant, K32M:1, followed by K32M:1+3, and then by the K32M:1+2 and
K32M:1+4 double mutants, which are quite similar in their behavior.
11
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Another difference between the titrations is the percent quench observed. Similar
quenches are observed for Quad3, K32M:1 as well as the K32M:1+3 mutant, suggesting related
binding modes. Smaller quenches are observed for the K32M:1+2 and K32M:1+4 mutants,
suggesting either NADPH binds differently, or the environment of W38 (which monitors the
titration (17)) is altered by a proximal K32M mutation.
Binding of DHF Monitored By Fluorescence Quenching Binding of DHF to the various
mutants was also performed at 4oC. Particularly for weaker binding mutants, the background
quench associated with addition of DHF becomes substantial, precluding accurate fitting. Thus
the DHF binding data monitored by fluorescence quenching are presented qualitatively in Figure
3B. Binding of DHF to both Quad3 and the single mutant (K32M:1) proteins yields titrations
that are mostly complete upon addition of ~7 µM ligand. An intermediate titration with a
reduced quench is observed for DHF binding to the K32M:1+4 mutant. Finally, the K32M:1+2
and 1+3 double mutants both show minimal quenching as well as much weaker binding.
Alterations in the percent quench could describe either changes in the number of DHF molecules
bound and/or how DHF is juxtaposed near W38. Weaker binding most likely describes lost
interactions between DHF and K32 and/or effects on the positive cooperativity that occurs
during binding of 2 DHF molecules.
A Comparison of Steady-State Kinetic Parameters Since PEG was present in the assays, a test
of its effects on the steady state behavior of Quad3 was performed. As shown in Table 2, no
significant changes were noted. The steady state kinetic parameters for the various mutants are
also listed in Table 2. The behavior of the single mutant protein, K32M:1, is quite similar to that
12
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
of Quad3. More significant changes are observed for the double mutants. The K32M:1+2
mutant displays a 4-fold increase in Km (NADPH), a 2-fold increase in Km (DHF), and a 4-fold
decrease in kcat as compared to Quad3. The K32M:1+4 mutant possesses a similar Km (NADPH) to
Quad3 with a 1.6 fold decrease in the affinity for DHF. However, kcat is decreased 8-fold by this
mutation. Finally, the K32M:1+3 mutant displays the most obvious changes as its Km (NADPH) is
≥160 µΜ, Km (DHF) is ≥ 330 µM and kcat is ≥ 3.6 s-1. These kinetic values are presented as lower
limits due to the large elevation in Km values for both substrate and cofactor coupled with the
absorbance limit of the spectrophotometer. This mutant (1+3, where half the pore contains wt
K32 residues and the other half pore contains K32M mutations) clearly displays at least a 4.5
fold enhancement of the kcat value compared to Quad3.
While the apo structure of the K32M:1+4 mutant (see Figure 1D) appears perturbed by
CD and sedimentation velocity measurements, addition of ligands appears to provide a
reasonable level of compensation since this mutant is as active as the K32M:1+2 mutant. Most
likely binding of NADPH and DHF provides sufficient binding contacts to restore the necessary
structure for activity.
Salt Effects on Kinetic Behavior To evaluate the ionic strength dependence of these steady state
kinetic parameters, salt effects were analyzed for the double mutant series. Data are shown in
Figure 4 and the slopes of the plots are listed in Table 3. For both the K32M:1+2 and K32M:1+4
mutants, kcat is only minimally affected by salt concentration while Km (NADPH) and Km (DHF)
increase with increasing ionic strength. The kinetic behavior for these two double mutants (with
13
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
single K32M mutations in each half of the pore) is similar, suggesting R67 DHFR can
accommodate several topologies of bound ligands.
Finally, we attempted to determine if kinetic values could be obtained for the K32M:1+3
mutant in TE buffer plus 0.03M NaCl, as smaller Km values are usually observed in lower salt
conditions. For this buffer, it appeared that Km (NADPH) could be bracketed, but Km (DHF) could
not. Since the resulting values are linked, they provide only a lower limit and are: Km (DHF) ≥ 490
µM, Km (NADPH) ≥ 60 µM, and kcat ≥ 4.9 s-1.
DISCUSSION
The simplest model of catalysis in R67 DHFR proposes that DHF occupies half the pore
and cofactor the other half. The pteridine ring of DHF and the nicotinamide ring of NADPH
encounter each other at the center of the pore (29) where the reaction occurs. From a comparison
of Kd1 values, a strong preference exists for NADPH to bind first (2). It would presumably
interact with the tightest binding site available. When DHF then binds to the enzyme•NADPH
complex, it is forced into the other half of the pore, which could lack one to two K32 contacts in
this asymmetric mutant series. Binding models for the various mutants are schematized in
Figure 5 and will be discussed below.
Does K32 play a direct role in binding NADPH? Docking of NADPH into R67 DHFR•folate
predicts an ionic interaction between K32 and the 2’phosphate of NADPH (see Figure 1E) (3).
While the pyrophosphate bridge is docked further away from a symmetry related K32 residue,
14
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
sidechain movement could allow formation of a second ionic interaction. Experimental evidence
supporting a second ionic interaction comes from our previous salt effect data (7) where binding
of NADPH to wild type (homotetrameric) R67 DHFR was found to be salt sensitive and a log-
log plot of ionic strength vs. Kd1 yielded a slope of 2. The slopes of these types of plots have
previously been taken to describe Z, the number of ionic interactions involved in binding
(11,12,30).
To evaluate the above model where 1-2 ion pairs are predicted to occur between NADPH
and symmetry related K32 residues, asymmetric mutants were used. A single K32M mutation
did not alter the Kd1 associated with NADPH binding. In the double mutant series, the
K32M:1+3 mutant shows the least effect, with an ~ 5 fold increase in Kd1 and a similar percent
quench as Quad3. The K32M:1+2 and K32M:1+4 double mutants show ~10 fold weaker
binding and reduced quenching of protein fluorescence, suggesting more perturbations are
associated with NADPH binding. Effects on binding are likely to include loss of an interaction
between K32 and NADPH as well as any alterations in the electrostatic potential, and/or in κ, the
orientational effect (refer to Figure 5 for models). (Effects on κ, the orientational effect in the
modified Smoluchowski equation describing diffusion, would be expected to decrease kon (31).)
When NADPH binds to the K32M:1+3 double mutant, it may bind to either the half pore
possessing wildtype K32 residues or the half pore containing mutant K32M residues (see Figure
5). Presumably, binding to the wildtype half pore would involve fewer perturbations and result
in minimal effects on Kd1. This expectation is consistent with the observed dissociation constant.
However as mutations are added to each half pore (as in the K32M:1+2 and K32M:1+4 double
15
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
mutants), larger effects would be expected if an ionic interaction is lost upon NADPH binding.
Surprisingly, the observed effect on Kd1 for the K32M:1+2 or K32M:1+4 mutants is small (~10
fold when compared to Quad3 and <2 fold when compared to K32M:1+3), suggesting that either
R67 DHFR utilizes only one ionic interaction in binding NADPH or that if a second ionic
interaction occurs, it is quite weak.
When the Kd1 (NADPH) values are compared with the Km (NADPH) values (TE with 0.03M
NaCl pH 7.0), it is clear that some Km values must include kinetic terms. For example, while the
Kd1 (NADPH) for the K32M:1+3 double mutant is 5.2 µM, the corresponding Km (NADPH) value is ≥
60 µM.
When kcat/Km (NADPH) values are compared, the double mutants become equivalent (~0.01-
0.02 s-1µM-1) and are approximately 10-20 fold less efficient than Quad3. The observation of
similar kcat/Km (NADPH) values for all double mutants over a ≥ 36 fold kcat range indicates some
degree of non-productive binding likely occurs. Non-productive binding dictates that kcat/Km
values remain constant while additional binding modes lead to apparently tighter binding.
However since only a fraction of these binding modes is productive, kcat is also decreased in
parallel (32). Non-productive binding in R67 DHFR would not be surprising due to its 222
symmetry. Since the highest kcat and Km values occur for the K32M:1+3 double mutant, this
protein likely possesses the most productive binding mode (in the double mutant series). In other
words, loss of ionic contacts in one half of the pore in this mutant (see Figure 5) leads to more
productive binding. Conversely, non-productive binding likely describes ionic contacts in one
half of the pore between K32 and DHF or NADPH.
16
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
What is the role of K32 in binding DHF? From binary complex binding studies (Figure 3B),
K32 clearly plays a role in binding DHF. Loss of one K32 residue is well tolerated, but loss of
two K32 residues affects both the percent quench as well as binding affinity. Changes in
affinity must remain qualitative as the curves were not fit.
If our cartoon describing the ternary complex in Figure 5 (also see Figure 1E)
approximates productive ground state binding, then minimal effects on kcat and Km would be
expected for the asymmetric mutant that matches the (predicted) productive ternary complex
topology (K32M:1+2), with larger effects on the non-preferred topology (K32M:1+4). However
we find that the 1+2 and 1+4 double mutants have similar kinetic behavior (<3 fold differences
in both kcat and Km values). This is a highly unusual result and we conclude that both topologies
allow comparable interactions between K32 and the glu tail of DHF. This behavior correlates
with the disorder observed for the glu tail of bound folate in the crystal structure (1) as well as
NMR studies that find the glu tail is mobile (29).
These data present a conundrum, how can the DHF tail be mobile, yet nearby residues
affect its’ binding? Perhaps the tail could maintain solvent separated ion pairs (SSIP) with both
symmetry related K32 residues. SSIP are a common phenomenon in small molecules (33-35) &
may play a role in enzymes (36,37). Solvent separation could minimize the desolvation penalty
associated with ion pair formation (38). Also some host-guest studies indicate tight binding can
be associated with high ligand mobility in the cavity (39-41).
17
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
From our binding model (2), NADPH binds first and interacts with the wt half of the
pore, while DHF is forced into the mutant half of the pore lacking K32 contacts (see Figure 5).
This model is supported by the asymmetric mutant data garnered in this study as Km (DHF) is
weakened >40 fold in the K32M:1+3 mutant when compared to Quad3. DHF is unlikely to bind
first as our fluorescence quenching studies do not indicate appreciable formation of a
K32M:1+3•DHF complex (up to 27µM DHF), while NADPH binds reasonably well to this
mutant under identical conditions (Figure 3). Since interligand cooperativity is important to R67
DHFR function (2,9), weaker binding of DHF affects NADPH Km values as well.
What is the role of K32 in catalysis? In wt R67 DHFR, an increase in kcat is associated with
increasing salt conditions (7). An increase in kcat also occurs in the K32M:1+3 mutant with a ≥
4.5 fold increase over Quad3 and a ≥ 18-fold increase over the other double mutants. Thus, this
mutant, which appears to have lost its potential for an ionic contact(s) in one half the pore due to
the substitution of both K32 residues, to some degree mimics the effects of salt on kcat for wt R67
DHFR. These data indicate that charge neutralization facilitates formation of the transition state.
What is the salt sensitive interaction monitored by kcat? Since salt sensitivity for kcat
remains when NADH is used as the alternate cofactor in wt R67 DHFR, the 2’phosphate moiety
is not involved (7). Either the pyrophosphate bridge of NADPH or the pABA-glu tail of DHF
remain as possible candidates. Our results for the K32M:1+3 double mutant suggest the salt
sensitive interaction may be with the pABA-glu tail of DHF. A weak, salt sensitive interaction
can be envisioned as beneficial as it could be more readily broken. Also the ability of the
18
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
K32M:1+2 and 1+4 double mutants to tolerate various positions of the pABA-glu tail could
correlate with the requirement to break a salt bridge with K32 to reach the transition state.
Another obvious effect of the asymmetric K32M mutations is on the slopes of log-log
plots of ionic strength vs. kcat. The slope for wt R67 DHFR is 0.9 while the slope for the
K32M:1+2 and 1+4 mutants decreases to 0.2. Clearly an interaction with K32 is involved in loss
of some level of salt sensitivity. Since the change in slope is not unitary and some salt sensitivity
remains, does this correspond to loss of a single ionic interaction? This scenario seems unlikely
as some level of non-productive binding appears to have been unmasked in the double mutants.
Since the slopes for kcat/Km remain within error of each other, it seems likely that the decrease in
slope for kcat corresponds to different salt sensitivities associated with the productive and non-
productive binding modes, with the non-productive mode being less susceptible in this ionic
strength range.
Conclusions Three critical observations from our asymmetric K32M mutations that provide
further detail on R67 DHFR function are: 1) a non-productive binding mode has been uncovered
in R67 DHFR as all three double mutants show similar kcat/Km values. The nonproductive
binding mode appears to correlate with interactions between K32 and the pABA-glu tail of DHF.
2) Similar kcat and Km values for the K32M:1+2 and K32M:1+4 double mutants are observed and
indicate alternate pABA-glu tail positions of DHF are tolerated by the enzyme. 3) The
K32M:1+3 double mutant shows an enhanced kcat coupled with weaker binding of DHF. This
result indicates charge neutralization enhances transition state formation, albeit at the expense of
19
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
ground state binding. These opposing trends arise from the 222 symmetry of the enzyme, which
dictates that each residue must serve numerous roles, none of which are likely to be optimized.
REFERENCES
1. Narayana, N., Matthews, D. A., Howell, E. E., and Xuong, N. H. (1995) Nat. Struct. Biol.
2, 1018-1025.
2. Bradrick, T. D., Beechem, J. M., and Howell, E. E. (1996) Biochemistry 35, 11414-
11424.
3. Howell, E. E., Shukla, U., Hicks, S. N., Smiley, R. D., Kuhn, L. A., and Zavodszky, M. I.
(2001) J. Comput. Aided Mol. Des. 15, 1035-1052.
4. Strader, M. B., Smiley, R. D., Stinnett, L. G., VerBerkmoes, N. C., and Howell, E. E.
(2001) Biochemistry 40, 11344-11352.
5. Bradrick, T. D., Shattuck, C., Strader, M. B., Wicker, C., Eisenstein, E., and Howell, E.
E. (1996) J. Biol. Chem. 271, 28031-28037.
6. Smiley, R. D., Stinnett, L. G., Saxton, A. M., and Howell, E. E. (2002) Biochemistry 41,
15664-15675.
7. Hicks, S. N., Smiley, R. D., Hamilton, J. B., and Howell, E. E. (2003) Biochemistry 42,
10569-10578.
8. Park, H., Bradrick, T. D., and Howell, E. E. (1997) Protein Eng. 10, 1415-1424.
9. Strader, M. B., Chopra, S., Jackson, M., Smiley, R. D., Stinnett, L. G., Wu, J., and
Howell, E. E. (2004) Biochemistry 43, 7403 - 7412
10. Hamilton, J. B. (1997) MS thesis in Biotechnology, University of TN, Knoxville
11. Park, C., and Raines, R. T. (2001) J. Am. Chem. Soc. 123, 11472-11479.
20
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
12. Record Jr., M. T., Lohman, T. M., and De Haseth, P. (1976) J. Mol. Biol. 107, 145-158.
13. Pitcher, W. H., 3rd, DeRose, E. F., Mueller, G. A., Howell, E. E., and London, R. E.
(2003) Biochemistry 42, 11150-11160
14. Strader, M. B., and Howell, E. E. (1997) Gibco-BRL Focus 19, 24-25.
15. de Bernandez Clark, E., Schwartz, E., and Rudolph, R. (1999) Methods Enzymol. 309,
217-235.
16. Nichols, R., Weaver, C. D., Eisenstein, E., Blakley, R. L., Appleman, J., Huang, T. H.,
Huang, F. Y., and Howell, E. E. (1993) Biochemistry 32, 1695-1706
17. West, F. W., Seo, H. S., Bradrick, T. D., and Howell, E. E. (2000) Biochemistry 39,
3678-3689.
18. Royer, C. A., Mann, C. J., and Matthews, C. R. (1993) Protein Sci. 2, 1844-1852
19. Howell, E. E., Warren, M. S., Booth, C. L., Villafranca, J. E., and Kraut, J. (1987)
Biochemistry 26, 8591-8598.
20. Ellis, K. J., and Morrison, J. F. (1982) Methods Enzymol. 87, 405-426
21. Baccanari, D., Phillips, A., Smith, S., Sinski, D., and Burchall, J. (1975) Biochemistry 14,
5267-5273
22. Blakley, R. L. (1960) Nature 188, 231-232.
23. Horecker, B. L., and Kornberg, A. (1948) J. Biol. Chem. 175, 385-390.
24. Zhuang, P., Yin, M., Holland, J. C., Peterson, C. B., and Howell, E. E. (1993) J. Biol.
Chem. 268, 22672-22679
25. Dunn, S. M., Lanigan, T. M., and Howell, E. E. (1990) Biochemistry 29, 8569-8576.
26. Schuck, P. (2000) Biophys. J. 78, 1606-1619
21
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
27. Reece, L. J., Nichols, R., Ogden, R. C., and Howell, E. E. (1991) Biochemistry 30,
10895-10904
28. Woody, R. W. (1995) Methods Enzymol. 246, 34-71
29. Li, D., Levy, L. A., Gabel, S. A., Lebetkin, M. S., DeRose, E. F., Wall, M. J., Howell, E.
E., and London, R. E. (2001) Biochemistry 40, 4242-4252.
30. Jensen, D. E., and von Hippel, P. H. (1976) J. Biol. Chem. 251, 7198-7214.
31. von Hippel, P. H., and Berg, O. G. (1989) J. Biol. Chem. 264, 675-678
32. Fersht, A. (1999) Structure and Mechanism in Protein Science, W.H. Freeman and Co.,
New York
33. Dougherty, R. C., and Howard, L. N. (2003) Biophys. Chem. 105, 269-278
34. Bagnol, L., Horner, J. H., and Newcomb, M. (2003) Org. Lett. 5, 5055-5058
35. Harder, S., Feil, F., and Repo, T. (2002) Chemistry 8, 1991-1999
36. Roca, M., Marti, S., Andres, J., Moliner, V., Tunon, I., Bertran, J., and Williams, I. H.
(2003) J. Am. Chem. Soc. 125, 7726-7737
37. Umezurike, G. M. (1987) Biochem. J. 241, 455-462
38. Chong, L. T., Dempster, S. E., Hendsch, L. P., and Tidor, B. (1998) Protein Sci. 7, 206-
210
39. Garel, L., Lozach, B., Dutasta, J., and Collet, A. (1993) J Am Chem Soc 115, 11652-
11653
40. Schneider, H., Blatter, T., and Simova, S. (1991) J Am Chem Soc 113, 1996-2000
41. Houk, K. N., Leach, A. G., Kim, S. P., and Zhang, X. (2003) Angew Chem Int Ed Engl
42, 4872-4897
22
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
42. Kuntz, I. D., Blaney, J. M., Oatley, S. J., Langridge, R., and Ferrin, T. E. (1982) J. Mol.
Biol. 161, 269-288.
43. Shoichet, B. K., and Kuntz, I. D. (1993) Protein Eng. 6, 723-732
44. Brito, R. M., Reddick, R., Bennett, G. N., Rudolph, F. B., and Rosevear, P. R. (1990)
Biochemistry 29, 9825-9831
45. Brito, R. M., Rudolph, F. B., and Rosevear, P. R. (1991) Biochemistry 30, 1461-1469
FOOTNOTES
1. Abbreviations: DHFR, dihydrofolate reductase; wt, wild type; TMP, trimethoprim; DHF,
dihydrofolate, NADP(+/H), nicotinamide adenine dinucleotide phosphate (oxidized/reduced);
NMNH, reduced nicotinamide mononucleotide; MTH buffer, 50 mM MES + 100 mM Tris, + 50
mM acetic acid polybuffer; µ, ionic strength; TE, 10 mM Tris + 1 mM EDTA buffer; ITC,
isothermal titration calorimetry; CD, circular dichroism; pABA-glu, the para-aminobenzoic acid-
glutamate tail of dihydrofolate; and Quad3, the protein product of the quadruplicated R67 DHFR
gene. Mutant enzymes containing amino acid substitutions are described by the wild type
residue and numbered position in the sequence, followed by the amino acid substitution. For
example, K32M R67 DHFR describes the lys32 → met mutant.
2. The amino acids in the first monomer (A) are labeled 1-78; those in the second monomer (B),
101-178; those in the third monomer (C), 201-278; and those in the fourth monomer (D), 301-
378. For brevity, when a single residue is mentioned, all four symmetry related residues are
implied. The monomer arrangement going clockwise in the crystal structure (1VIE.pdb) is
23
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
ABDC. To minimize confusion in the quadruplicated gene construct, we have re-labeled the
monomers 1234 (= ABCD) going clockwise.
3. For simplicity as well as to provide consistent nomenclature with other asymmetric mutants
(e.g. (6) and Stinnett et al., companion paper), we use the K32M:1, K32M:1+2 or 1+3 or 1+4
nomenclature. However the single mutation was actually placed in the second gene copy.
Because of the 222 symmetry, the K32M:2 construct should be equivalent to a K32M:1
construct. The double mutants actually constructed were K32M:3+4, K32M:2+4, and
K32M:2+3 (which should be equivalent to the K32M:1+2, K32M:1+3 and K32M:1+4 constructs
from the 222 symmetry). These constructs were made as we were concerned that two K32M
mutations at what corresponds to a dimer-dimer interface in wildtype (wt) R67 DHFR could be
destabilizing. Thus we placed the two K32M mutations into gene copies 2 and 3. Our reasoning
for this placement was to allow the wild type (wt) sequences in copies 1 and 4 to interact at the
N- and C-termini and provide contacts to hold the domains together.
ACKNOWLEDGEMENTS
We thank Cynthia Peterson for her assistance with the sedimentation velocity experiments and
for critical reading of the manuscript. Special thanks to Nathan C. VerBerkmoes and Robert L.
Hettich at the Organic and Biological Mass Spectrometry group at Oak Ridge National
Laboratory for intact mass spectrometry analysis of the K32M:1+2 mutant.
24
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
FIGURE LEGENDS
Figure 1. The structure of R67 DHFR (1VIE in the protein data bank). Each monomer in wt
homotetrameric R67 DHFR corresponds to a domain in Quad3. Monomer 1 is colored red,
monomer 2 is yellow, monomer 3 is magenta, and monomer 4 is green. The K32 positions are
shown using a CPK representation. Panel A shows a ribbon drawing of the homotetramer
looking end on with the active site pore in the center. The N-termini point to the left and right
sides of the image while the C-termini point up and down. The protein in this crystal form was
obtained by chymotrypsin treatment such that 16 amino acids were cleaved off the N-terminus
(1,27). Panels B-D, the ribbon structures are related to panel A by a 90° rotation along the y-
axis. The expected positions of the mutations are shown in white whereas wild-type K32
residues are indicated in color. Panel B illustrates the K32M:1+2 double mutant, its mutations
lie across the pore from one another in a diagonal orientation (i.e. across the monomer-monomer
interface (red and yellow). Panel C depicts the K32M:1+3 double mutant; both mutations lie in
the same half- pore. Panel D describes the K32M:1+4 mutant; it possesses mutations on
opposite sides of the pore, arranged consecutively (i.e. across along the dimer-dimer interface
(red and green). Panel E shows a reverse image of the active site pore generated by DOCK
(3,42,43). The orientation of this sphere cluster (with respect to the homotetramer) corresponds
to panels B-D. Each white point corresponds to a potential atom position used by the docking
program. The position of a docked NADPH molecule that meets the NMR constraints
(29,44,45) is shown in cyan and the position of the highest scoring docked folate molecule is
shown in orange. The positions of the 4 symmetry related K32 residues in the surrounding
protein are colored and numbered as described above. The rest of the protein is not shown for
25
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
simplicity. Only 1 out of 4 symmetry related binding modes is shown. The other binding modes
can be generated by 180o rotations around the x, y and z axes.
Figure 2. pH titration curves describing the behavior of Quad3 in the presence of various
concentrations of NaCl. This experiment examines salt effects on the transition from a “closed”
(pH 6-8) to an “open” (pH 4-5) form. Titration experiments were performed in the presence of
MTH buffer ( , ), MTH buffer containing 0.2M NaCl ( , ), MTH containing 0.5M
NaCl ( , ), and MTH containing 0.75M NaCl ( , ). Best fits for these curves yield
pKa values of 5.9 ± 0.01, 6.0 ± 0.01, 5.7 ± 0.01 and 5.6 ± 0.02, respectively.
Figure 3. Fluorescence quenching of Quad3 variants by NADPH or DHF. Panels A and B
show the NADPH and DHF titrations respectively. The data were normalized to allow ready
comparison. The titrations generated for Quad3 (hexagons, fit by solid line), K32M:1 (∇ points
and dot-dash-dash line), K32M:1+2 ( points and a dot-dash line), K32M:1+3 ( points and a
dashed line), and K32M:1+4 mutants ( points and a dotted line) are shown. Best-fit values for
the NADPH titrations are given in Table 1. The DHF titrations were not fit as the baseline
quench for DHF varied and became substantial for the weaker binding mutants.
Figure 4. Log-log plots of each kinetic parameter vs. ionic strength generated for the K32M:1+2
( points) and K32M:1+4 ( points) double mutants. Steady-state kinetics were performed in
TE buffer pH 7 (30°C) with various salt concentrations added to adjust the ionic strength. There
is a linear correlation between each parameter and ionic strength. Slopes of the various plots are
given in Table 3.
26
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Figure 5. Cartoons describing how NADPH and DHF may bind to the various mutants. The
dumbbell shaped outline represents the active site pore. Bound NADPH is shown by a light gray
patch on the right and bound DHF is depicted by a darker gray area on the left. The 2’phosphate
and pyrophosphate groups of NADPH are represented by a circle and square and the glu tail of
DHF by a diamond. For Quad3, only one of four possible symmetry related binding modes is
shown. The positions of the four symmetry related K32 residues in the surrounding protein are
also shown in a cartoon fashion. They are numbered 32, 132, 232 and 332 and are colored black,
dark gray, light gray and dotted respectively. The rest of the protein is not shown for simplicity.
Potential ionic interactions are indicated by the juxtaposition of the K32 residues with the circle,
square or diamond. As residues are mutated to methionine, the K32 patches are removed,
indicating the loss of any potential contact. As mutations are added, the occupancy of binding
sites containing K32M substitutions is expected to decrease. For example, as NADPH has a
clear preference to bind before DHF (2), we have placed it in the least perturbed half pore to
maintain the maximal number of contacts, see K32M:1 and K32M:1+3. When K32M mutations
are added to each half pore, the preferred binding configuration is expected to be a function of
the strongest interaction(s). For example, if a preference exists for an ionic interaction between
the 2’phosphate of NADPH and K32, then that site will be occupied (see K32M:1+2 as well as
the K32M:1+4 model where the ligands are flipped and occupy a symmetry related site).
27
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Table 1. Fluorescence quenching of Quad3 and the K32M mutants by NADPH. These
experiments were performed at 4°C in TE buffer, pH 7 in the presence of 0.03M NaCl (µ=0.05)
and 0.1g/L PEG.
Species Kd1 (µM)
Quad3 1.0 ± 0.1
K32M:1 1.2 ± 0.04
K32M:1+2 9.7 ± 0.7
K32M:1+3 5.2 ± 0.5
K32M:1+4 8.4 ± 0.4
28
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Table 2. Steady-state kinetics were performed with the K32M asymmetric single and double
mutants in MTH buffer, pH 7 at 30°C. Data were fit using a non-linear global option in SAS to
an equation describing bi-substrate kinetics.
Protein Km (NADPH) (µM)
Km (DHF) (µM)
kcat (s-1)
kcat / Km (NADPH) (s-1 µM-1)
kcat / Km (DHF) (s-1 µM-1)
Wild Type R67 DHFRa
3.0 ± 0.06 5.8 ± 0.02 1.3 ± 0.1 0.43 ± 0.03 0.22 ± 0.01
Quad 3b
4.4 ± 0.4 6.7 ± 0.4 0.80 ± 0.02 0.18 ± 0.02 0.12 ± 0.01
Quad3 + PEG
3.9 ± 0.2 7.6 ± 0.4 0.85 ± 0.03 0.22 ± 0.02 0.11 ± 0.01
K32M:1
3.6 ± 0.2 6.0 ± 0.3 0.56 ± 0.01 0.16 ± 0.01 0.093 ± 0.006
K32M:1+2
17.1 ± 0.1 14.4 ± 0.1 0.20 ± 0.01 0.012 ± 0.001 0.014 ± 0.001
K32M:1+3 ≥ 145 ≥ 400 ≥ 3.7
0.026c 0.009c
K32M:1+4 5.4 ± 0.2 10.5 ± 0.3 0.10 ± 0.01 0.019 ± 0.003 0.010 ± 0.001
a from (27).
b from (6). c presented without errors as saturation with neither substrate could be obtained.
29
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Table 3. Slopes of log-log plots of various kinetic parameters vs. ionic strength for the
K32M:1+2 and K32M:1+4 mutants compared to wt R67 DHFR.
DHFR
Species Slope kcat/Km
(NADPH)
Slope kcat/Km (DHF)
Slope kcat Slope Km (NADPH)
Slope Km (DHF)
wt R67 DHFRa
-0.6 ± 0.09 -0.9 ± 0.03 0.9 ± 0.08 1.5 ± 0.1 1.8 ± 0.2
K32M:1+2 -1.1 ± 0.13 -0.72 ± 0.16 0.2 ± 0.05 1.3 ± 0.02 0.9 ± 0.02
K32M:1+4 -1.2 ± 0.1 -1.2 ± 0.2 0.2 ± 0.05 1.4 ± 0.04 1.4 ± 0.09
a from (7).
30
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
Circular dichroism spectra for Quad3 and each of the K32M double mutants at pH 8. The spectrum for Quad3 is given by the solid line; that for K32M:1+2 by a dashed line; K32M:1+3 by a dotted line; and K32M:1+4 by a dash-dot-dot line.
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from
HowellStephanie N. Hicks, R. Derike Smiley, Lori G. Stinnett, Kenneth H. Minor and Elizabeth E.Role of K32 residues in R67 dihydrofolate reductase probed by asymmetric mutations
published online August 27, 2004J. Biol. Chem.
10.1074/jbc.M404484200Access the most updated version of this article at doi:
Alerts:
When a correction for this article is posted•
When this article is cited•
to choose from all of JBC's e-mail alertsClick here
Supplemental material:
http://www.jbc.org/content/suppl/2004/09/10/M404484200.DC1
by guest on April 4, 2018
http://ww
w.jbc.org/
Dow
nloaded from