Neutron resonance scattering shows specific binding of plutonium to the calcium-binding sites of the protein calmodulin and yields precise distance information

Article abstract of DOI:10.1021/ja9633124

We have successfully substituted ²⁴⁰Pu³⁺ for Ca²⁺ in the calcium- binding protein calmodulin and used neutron resonance scattering from the bound ²⁴⁰Pu to demonstrate that the Pu binds specifically to the Ca²⁺ sites and also to measure the distance between the ion binding sites within individual domains of the protein. ²⁴⁰Pu has a strong nuclear resonance at 0.278 ?, and at this wavelength the coherent scattering from ²⁴⁰Pu is > 1000 times that of any other nucleus present in a protein. The ionic radius of Pu³⁺ is very similar to that of Ca²⁺, and hence we chose this species to substitute for Ca²⁺ in the protein. We identified solution conditions that stablize Pu³⁺ in solution at near neutral pH for 6-7 h in order to form the Pu/calmodulin complex under conditions favorable for both complex formation and maintaining the structural integrity of the protein. We collected small-angle neutron scattering data from solutions of 4(²⁴⁰Pu³⁺)·calmodulin, which contain periodic terms that are directly related to the distances between the Ca²⁺-binding sites. The shorter Pu-Pu distance, i.e., the average distance between the two sites within each globular domain of calmodulin, is found to be 11.8 ± 0.4 ?, in excellent agreement with the value of 11.7 ? from crystallographic determinations. This is the first use of neutron resonance scattering as a structural probe in a protein.

Full text of DOI:10.1021/ja9633124

5118
J. Am. Chem. Soc. 1997, 119, 5118-5125
Neutron Resonance Scattering Shows Specific Binding of
Plutonium to the Calcium-Binding Sites of the Protein
Calmodulin and Yields Precise Distance Information
†
‡
P. A. Seeger, S. E. Rokop, P. D. Palmer, S. J. Henderson, D. E. Hobart, and
J. Trewhella*
Contribution from the Chemical Science and Technology DiVision and
Manuel Lujan Jr. Neutron Scattering Center, Mail Stop G758,
Los Alamos National Laboratory, Los Alamos, New Mexico 87545
X
ReceiVed September 20, 1996. ReVised Manuscript ReceiVed April 11, 1997
Abstract: We have successfully substituted 40Pu3+ for Ca2+ in the calcium-binding protein calmodulin and used
2
neutron resonance scattering from the bound 2 Pu to demonstrate that the Pu binds specifically to the Ca sites and
40
2+
2
40
also to measure the distance between the ion binding sites within individual domains of the protein.
Pu has a
2
40
strong nuclear resonance at 0.278 Å, and at this wavelength the coherent scattering from Pu is >1000 times that
3
+
2+
of any other nucleus present in a protein. The ionic radius of Pu is very similar to that of Ca , and hence we
chose this species to substitute for Ca2 in the protein. We identified solution conditions that stablize Pu in solution
at near neutral pH for 6-7 h in order to form the Pu/calmodulin complex under conditions favorable for both complex
formation and maintaining the structural integrity of the protein. We collected small-angle neutron scattering data
+
3+
240
3+
from solutions of 4( Pu )‚calmodulin, which contain periodic terms that are directly related to the distances between
2+
the Ca -binding sites. The shorter Pu-Pu distance, i.e., the average distance between the two sites within each
globular domain of calmodulin, is found to be 11.8 ( 0.4 Å, in excellent agreement with the value of 11.7 Å from
crystallographic determinations. This is the first use of neutron resonance scattering as a structural probe in a protein.
Introduction
and exposure of hydrophobic amino acid side chains that play
a key role in target enzyme binding. Earlier small-angle
scattering studies by us were interpreted as indicating that the
globular domains of CaM were on average closer together in
solution than in the crystal structure, probably through flexibility
Ca2 is widely used in nature as a messenger to transmit
signals by reversibly binding to specific receptor proteins and
triggering conformational transitions that in turn result in
activation of specific biochemical pathways. Calmodulin (CaM)
is the major intracellular receptor for Ca2 in eukaryotic cells
and is responsible for the regulation of a diverse array of target
enzymes (reviewed in refs 1 and 2). CaM is a relatively small
protein (16700 Da, 148 residues), and its structure shows an
unusual dumbbell shape with two globular domains connected
by a solvent-exposed R-helix of 7-8 turns.3 Each of the
+
6
in the interconnecting helix region. While initially a point of
+
disagreement, this conclusion is now widely accepted in the
7
,8
9,10
1,12
light of more recent data from NMR and crystallography,
as well as neutron small-angle solution scattering studies1
on CaM complexed with peptides corresponding to its binding
domains in various target enzymes. These studies show that
CaM collapses around the helix-forming peptides via the
conformational flexibility available in the central helix, which
would also facilitate CaM binding to the different stereochemical
surfaces presented by different target enzymes.
Neutron-scattering techniques are extremely useful for study-
ing structure and dynamics in condensed matter, including
biological systems. Neutron scattering is a quantum mechanical
process that depends upon the properties of the compound
nucleus that forms between an incoming neutron and a nucleus
2+
globular domains contains two structurally similar Ca -binding
sites having a helix-loop-helix structural motif, with the loop
regions each presenting seven oxygen atoms (six from the
4
protein and one from a water molecule) in a pentagonal-
byramidal geometry to bind a single Ca2+ with a mean oxygen-
calcium distance of approximately 2.4 Å (reviewed in ref 5).
2+
Ca binding in these loops results in an “opening” of the
globular domains, via a hinge defined within the loop regions,
*
To whom correspondance should be addressed.
†
Present address: Chemical Engineering and Materials Sciences, Un-
versity of Oklahoma, Norman, OK 73019.
(6) Heidorn, D. B.; Trewhella, J. Biochemistry 1988, 27, 909-915.
(7) Ikura, M.; Clore, G. M.; Gronenborn, A. M.; Zhu, G.; Klee, C. B.;
Bax, A. Science 1992, 256, 632-638.
‡
Present address: RE/Spec Inc., Albuquerque, NM, and Glenn T. Seaborg
Institute of Transactinide Science, Lawrence Livermore Laboratory, Liv-
ermore, CA 94551.
(8) Barbato, G.; Ikura, M.; Kay, L. E.; Pastor, R. W.; Bax, A.
Biochemistry 1992, 31, 5269-5278.
X
Abstract published in AdVance ACS Abstracts, May 15, 1997.
(9) Meador, W. E.; Means, A. R.; Quiocho, F. A. Science 1992, 257,
1251-1255.
(1) Means, A. R.; VanBerkum, M. F. A.; Bagchi, I.; Lu, K. P.;
Rasmussen, C. D. Pharm. Ther. 1991, 50, 255-270.
(10) Meador, W. E.; Means, A. R.; Quiocho, F. A. Science 1993, 262,
1718-1721.
(
4.
2) O’Neil, K. T.; Degrado, W. F. Trends Biochem. Sci. 1990, 15, 59-
6
(11) Heidorn, D. B.; Seeger, P. A.; Rokop, S. E.; Blumenthal, D. K.;
Means, A.R.; Crespi, H.; Trewhella, J. Biochemistry 1989, 28, 6757-6764.
(12) Trewhella, J.; Blumenthal, D. K.; Rokop, S. E.; Seeger, P. A.
Biochemistry 1990, 29, 9316-9324.
(3) Babu, Y. S.; Sack, J. S.; Greenhough, T. G.; Bugg, C. E.; Means, A.
R.; Cook, W. J. Nature 1985, 315, 37-40.
(4) Kretsinger, R. H.; Moncrief, N. D.; Goodman, M.; Czelusniak, J.
The Calcium Channel: Structure, Function and Implications; Morad,
Naylor, Kazda, and Schramm, Eds.; Springer-Verlag: Berlin, 1988; pp 16-
(13) Mughabghab, S. F. Neutron Scattering Cross Sections: Neutron
Resonance Parameters and Thermal Cross Sections; Academic Press:
London, 1984; Vol. 1, Parts A and B.
3
5.
(
5) Strynadka, N. C. J.; James, M. N. G. Annu. ReV. Biochem. 1989, 58,
(14) Lynn, J. E.; Seeger, P. A. At. Data Nucl. Data Tables 1990, 44,
191-207.
9
51-958.
S0002-7863(96)03312-4 CCC: $14.00 © 1997 American Chemical Society

Specific Binding of Pu to Ca-Binding Sites of Protein
J. Am. Chem. Soc., Vol. 119, No. 22, 1997 5119
in a sample. As a result, neutron-scattering amplitudes vary
irregularly with atomic number (in contrast to X-ray scattering
factors that increase linearly with the number of electrons),
isotopes of the same element can have very different neutron-
scattering properties, and there are strong wavelength-dependent
resonance effects. The large differences in neutron-scattering
1
2
properties between the isotopes of hydrogen ( H and H) have
been taken advantage of for studies of biological molecules and
complexes over the past several decades. Modern accelerator-
based spallation neutron sources have opened up the potential
for applications of neutron resonance scattering, since they
provide pulses of neutrons containing a wide spectrum of
wavelengths and time-of-flight methods can be readily used to
select the neutrons with the desired wavelength. To date,
however, it has been difficult to use neutron resonance scattering
as a structural probe due largely to the low neutron fluxes at
the wavelengths of interest. There are many nuclei with
resonances below ∼2 eV, but if we include only those whose
coherent scattering is at least 1% of the total cross section, then
1
13
115
the list is limited to 11 isoptoes: soft metals Cd and In,
rare earths Sm, Eu, Gd, Dy, Yb, and 177Hf, noble
149
151
157
163
168
metal Ir, and actinides Th and 240Pu.
1
93
230
13,14
A number of
Figure 1. (A) Coherent neutron-scattering length (complex value) near
the 1.057-eV resonance of 2 Pu. (B) Neutron transmission at the
resonance, for various area densities.
the resonant nuclei have properties that are similar to those of
40
biologically relevant metals. In particular, 2 Pu is a positively
40
3+
2+
charged ion with a similar ionic radius as Ca , and a good
candidate for binding with high affinity to Ca2 sites in proteins.
+
this allows a precise computation of the wavelength dependence
of both the complex coherent scattering amplitude (which leads
to small-angle scattering) and the total cross section (which is
essential in correcting for the transmission of the sample in the
neighborhood of the resonance). The wavelength at the peak
of the resonance is 0.278 Å. Although this is much shorter
than the wavelengths usually used for small-angle scattering,
and hence will limit the ability to measure small momentum
transfer, neutrons of this wavelength are available and well
resolved at a pulsed spallation source such as the Manuel Lujan
Jr. Neutron Scattering Center (MLNSC) at the Los Alamos
Neutron Science Center (LANSCE). Small-angle neutron
scattering is determined by coherent interference of all pairs of
nuclei in the sample. If a neutron of wavelength λ (velocity V
The differences in charge-to-radius ratio between Ca2 and Pu
+
3+
would, however, be expected to result in differences in the
dynamics of the reversible ion binding, and possibly in the ion-
dependent structural transitions that are critical to the normal
function of proteins like CaM. Disruption of biochemical
regulation by metals via such mechanisms has been dis-
cussed;1
5,16
however, in the case of Pu, there have been no
structural studies to date demonstrating binding to specific sites
in a protein.
The neutron resonance scattering experiments we report here
were motivated by the desire to see if the technique could work
as a structural tool, in particular for proteins in solution, and to
2
+
learn something of the specificity of binding of Pu to Ca
regulatory proteins. In proposing an experiment involving
)
h/Mnλ, where h is Plank’s constant and Mn is the neutron
240
resonant scattering from
Pu substituted for calcium in a
mass) scatters at an angle of 2θ, then the magnitude of its
momentum change (in units of h/2π) is
biological molecule, it is necessary to address questions of
physics (suitability of the resonant scatterer), chemistry (dis-
solving the atoms in the needed ionization state in an appropriate
medium), and biochemistry (binding the substituted atoms at
the relevant sites).
Q ) 4π(sin θ)/λ
(1)
and the direction of the Q vector is at an angle (π/2 + θ) from
the initial velocity. The scattering amplitudes must be summed
with a phase factor depending on wavelength and angle, so that
Experimental Methods and Results
1
8
the total amplitude of the scattered wave is
Physics of Resonant Scattering. Resonant neutron scattering
is analogous to anomalous X-ray scattering in that it is sensitive
to rapid variations of the interaction probability of a specific
atomic species as a function of wavelength. To be useful in
resonant scattering, the nuclear resonance must be at a low
enough energy (long enough wavelength) to sample appropriate
distance scales, must be relatively narrow so that variations are
rapid, and must have a strong component of coherent scattering.
Of all available nuclei, the neutron resonance at 1.057 eV in
A(Q) ) b exp(-iQ‚r )
(2)
∑
k
k
where rk is the vector from an arbitrary origin to atom k which
has (complex) scattering length bk. The observable scattering
intensity is the square of this amplitude:
I(Q) ) A*(Q) A(Q) )
b* b exp(-iQ‚r )
k j jk
∑∑
∑∑
2
40
Pu meets these physics criteria best. Parameters of this
)
b* b [cos(Q‚r ) -
k j jk
resonance have been fitted to existing nuclear data by using a
generalized single-level formalism.17 As shown in Figure 1,
i sin(Q‚r )] (3)
jk
(
15) Cox, J. L.; Harrison, S. D. Biochem. Biophys. Res. Commun. 1983,
15, 106-111.
16) Pounds, J. G.; Rosen, J. F. Toxicol. Appl. Pharmacol. 1988, 94,
1
where rjk is the vector from atom j to atom k. In solution,
assuming nonaligned particles, all orientations of rjk are equally
(
3
31-341.
(
17) Lynn, J. E. personal communication of the fitted parameters of the
(18) Guinier, A.; Fournet, G. Small-Angle Scattering of X-Rays; John
Wiley: New York, 1955; pp 5-8.
2
40
Pu resonance in the formalism of ref 14.

5120 J. Am. Chem. Soc., Vol. 119, No. 22, 1997
Seeger et al.
probable. In particular the sine term in eq 3 is odd and will
vanish when averaged. Also, in the double sum over j and k
the odd (imaginary) terms in the coefficient of the even cosine
will cancel. The result of spherical averaging is
I(Q) )
Re(b* b ) sin(Qr )/Qr
jk
(4)
∑
∑
k
j
jk
This is essentially the X-ray scattering equation of Debye,19
modified to include complex neutron scattering lengths.
For small-angle scattering, the scattering-length densities are
generally averaged over larger volumes of space (unlike high-
resolution crystal diffraction). In the simple case of a solution
of molecules of uniform composition and density in a uniform
solvent, the scattering depends on the difference between the
average scattering-length density within the molecule and that
of the solvent. In the case of resonant scattering, however, the
scattering length for the resonant species is so much greater
than that of any other atom in the system that we use eq 4
explicitly for all atoms of the resonant species and treat
everything else with an average (real) scattering-length density
F. Using subscript “r” to represent resonant scattering and “0”
for nonresonant, we can express I(Q,λ) in terms of nonresonant,
interference, and resonant form factors:
4
+
Figure 2. Plutonium solution absorbance spectra. (A) Pu (dash line)
and Pu3 (solid line) in acid solutions (1 M HClO
+
); (B) Pu in 1 mM
3+
2
2
4
I(Q,λ) ) F F (Q) + Re{b(λ)}FF (Q) + |b(λ)| F (Q) (5)
0
0
0r
rr
tris(hydroxylmethyl)aminomethane, pH 4.8 (the same spectra were
obtained for pH 4.8-6.5); and (C) 4Pu3 ‚CaM in 1 mM tris-
(hydroxymethyl)aminomethane (500 mM KCl, pH ∼ 6.0).
+
The known and distinctive wavelength dependencies (Figure
1
4
) will allow the form factor Frr(Q) to be extracted. From eq
, including only terms jk for the resonant species,
2
1
containing the plasmid pCam23PL. Purification was carried
11
out as previously described. (The first sample used for neutron
measurements also included about 1/3 bovine CaM purchased
from Calbiochem and used without further purification.) CaM
was prepared for complexing with Pu by first dialyzing against
0 mM tris(hydroxymethyl)aminomethane, 0.1 M KCl, and 2
mM EGTA (ethylene glycol bis(â-aminoethyl ether)-N,N,N′,N′-
tetraacetic acid) at pH 7.0 to remove all the calcium, as well as
other metal ions. Next it was dialyzed against 1 mM tris-
Frr(Q) )
sin(Qr )/Qr
jk
(6)
∑∑
jk
3+
and information about the interatomic distances rjk of the
5
resonant species can be obtained.
Preparation of Pu . Preparation of Pu3+ began with
isotopically enriched 2 PuO2 which was dissolved by heating
in concentrated HClO4 with a few drops of 5% hydrofluoric
acid. The Pu in solution was oxidized to PuO22 with fuming
HClO4. Hydrazine dihydrochloride was added to reduce the
3+
40
(
hydroxymethyl) aminomethane, 500 mM KCl, pH 6.0. Neutron-
+
scattering samples were prepared in D2O rather than H2O in
order to minimize the large incoherent neutron scattering from
H. For the first neutron sample D2O:H2O exchange was
achieved by repeated concentrations (4 times 40 mL to 10 mL)
in an Amicon ultrafiltration device (YM5 filter). For the second
neutron-scattering sample, dialysis was used instead of ultra-
filtration to avoid protein loss through absorption on the
membrane.
2
+
3+
3+
PuO2 to Pu , and NaNO2 was added to oxidize the Pu to
4
+
4+
Pu . The Pu solution was then purified with an anion
exchange column. Pu4 was eluted with 0.5 M HCl and stored
as a stock solution. An aliquot of this stock solution was used
for addition to CaM solutions. The acid content was reduced
further by again loading the Pu4 on an anion exchange column
+
+
4+
and eluting with 0.25 M HCl. The Pu in 0.2 M HCl was
Preparation of the 4(240Pu3+)‚CaM Complex. Preparation
of a Pu/CaM complex presents a number of challenges. Pu
has an ionic radius of 1.00 Å, which is essentially the same
as that of Ca . Pu would therefore be expected to bind
readily to the Ca -binding sites in CaM. Further, the greater
ratio of charge-to-radius for Pu is likely to give rise to tighter
binding than is seen for Ca . In order to make a Pu /CaM
complex, however, it is necessary to form a solution of Pu in
the pH range 5.5-6.0 since the protein will aggregate or
precipitate at lower pH. This is a challenge because at pH higher
than 4.0, Pu oxidizes rapidly and forms an insoluble Pu
colloid. We approached this problem by trying to identify
3
+
reduced to Pu in a standard three-electrode electrochemical
cell; the working electrode of the cell was a Pt screen, the
counter electrode was a Pt wire separated from the experimental
solution with a Vicor frit, and the reference electrode was a
saturated calomel. After all the Pu was reduced to Pu , a
strong reducing potential was applied in order to drive off H
3+
22
2+
3+
2+
2
0
3+
3+
+
2+
3+
at the Pt screen electrode. This further reduced the acid content
of the solution. Solid Na2CO3 was then added to bring the pH
to ∼4.0. The solution was centrifuged to remove any solid Na2-
3+
CO3 or Pu4 colloid. The purity of the Pu in the supernatant
+
3+
3+
4+
was assessed by spectrophotometry. Figure 2 shows the
4+
3+
characteristic absorption spectra for Pu and Pu . Concentra-
tions of Pu3 were determined from the absorbance at 600 nm.
Protein Preparation. Protein samples were prepared with
bacterially synthesized CaM from E. coli N5151 or AR68
3+
agents that would chelate Pu , holding it in solution and
+
inhibiting oxidation long enough to facilitate binding to CaM.
The chelator also needed to be mild enough to release the Pu3
to the CaM when needed. A variety of solutions were surveyed
+
(
19) Debye, P. Ann. Phys. 1915, 46, 809-823.
(21) Putkey, J. A.; Draetta, G. F.; Slaughter, G. R.; Klee, C. B.; Stull, J.
T.; Means, A. R. J. Biol. Chem. 1986, 261, 9896-9903.
(22) Shannon, R. D. Acta Crystallogr. 1976, A32, 751-767.
(20) Newton, T. W.; Hobart, D. E.; Palmer, P. D. Los Alamos National
Laboratory Report, LA-UR-86-967, 1986.

Specific Binding of Pu to Ca-Binding Sites of Protein
J. Am. Chem. Soc., Vol. 119, No. 22, 1997 5121
Table 1. Solution Conditions Tested for Stabilizing Pu3 Near Neutral pH
+
solvent
pH
7.0
Pu3 solubility
+
sodium Phosphate buffer
blue precipitate
(
50 mM NaPO4, 100 mM KCl)
MOPS buffer
50 mM 3[N-morpholino]propanesulfonic acid,
00 mM KCl)
sodium citrate buffer
50 mM HOC(COONa)(CH
00 mM KCl)
formed immediately
yellow precipitate
formed immediately
7.0
6.0
6.5
(
1
complete oxidation
to Pu4 in 20 min
+
(
2 2
COONa) ,
1
a
3+
cacodylate buffer
50 mM(CH
ascorbate buffer
20% of Pu oxidized
to Pu4 in 1 h
+
(
3
)
2
AsO‚OH, 100 mM KCl)
Pu3 stayed in solutionb
Pu precipitated overnight
Pu precipitated overnight
Pu precipitated overnight
+
(50 mM C
(50 mM C
(50 mM C
(10 mM C
5
5
5
H
H
H
cH
8
8
8
8
O
6
O
6
O
6
O
6
)
)
)
4.0
4.5
5.0
6.0
5
, 100 mM KCl)
Tris solutions
Pu3 stable 2-4 h
Pu3 stable 2-4 h
Pu3 stable 2-4 h
+
(10 mM tris(hydroxymethyl)aminomethane,
00 mM KCl)
(10 mM tris(hydroxymethyl)aminomethane,
00 mM KCl)
(1 mM tris(hydroxymethyl)aminomethane,
00 mM KCl)
6.5
6.0
6.0
1
+
5
+
5
a
CaM binding experiments with Pu in cacodylate buffers resulted in no Pu binding to the protein, possibly because the cacodylate chelated the
Pu too tightly. Low pH causes CaM to precipitate. c The Tris solutions were not buffers, and the pH drifted down to between 5 and 6 for these
b
experiments.
for their ability to stabilize Pu3+ at pH 5.5-6.0, and the results
are given in Table 1. KCl was used to keep the ionic strength
of the solution in a range known to be favorable to keeping the
protein in solution, free of aggregation and structurally stable.
The Tris solutions used had very little buffering capacity, but
were the most effective at keeping the Pu3 in solution for long
enough times to make the complexes with CaM. Thus to
+
prepare Pu-complexed CaM, the 2 Pu stock was diluted with
40
3+
1
mM tris(hydroxymethyl)aminomethane (500 mM KCl, pH 6.0)
in D2O. A flat-bottomed glass tube that had been soaked in
EDTA (ethylenediaminetetraacetic acid, to remove metal ions)
and thoroughly rinsed was used to contain the solutions in the
binding process. A small Teflon-coated stirbar provided rapid
3
+
stirring to lessen local acid shock to the protein. The Pu was
added in 100-mL aliquots by placing the pipet tip near the stirbar
3
+
3+
and releasing the Pu slowly. Sufficient Pu was added to
3+
achieve a 4:1 Pu :CaM stoichiometry, with no excess. Small
amounts of CaM came out of solution but quickly redissolved.
The final product was a clear blue solution characteristic of Pu .
Absorbance spectra were used to characterize Pu and Pu/
3+
Figure 3. Absorbance intensities from the Pu3+/CaM complex as a
function of mole equivalents of addition of Pu3 . Several of the features
+
3+
3+
CaM complexes, as shown in Figure 2. The comparison of Pu
show a linear increase between 0 and 4 mol equiv of Pu , followed
and Pu4 in strong acid (Figure 2A) shows the characteristic
differences between these species, and Figure 2B demonstrates
the purity of Pu3 in 1 mM tris(hydroxymethyl)aminomethane.
+
3+
by a flattening. This indicates that there are four binding sites for Pu
in CaM.
+
3+
Neutron-scattering data were collected from samples of
4(240Pu )‚CaM made in two independent preparations. The
calculated protein concentrations for the two samples were 50
and 44 mg/mL, prepared for the first and second neutron
experiments, respectively. The freshly prepared samples for
neutron scattering were transferred into a quartz cell 27 mm
wide by 28 mm high by 4 mm thick, for a total volume of 3
cm . This cell was in turn encapsulated in a secondary contain-
ment vessel with quartz windows and a boron nitride (BN) mask
to define the exposed area. The calculated CaM molecular area
On forming the complex with CaM, the Pu spectrum (Figure
3+
C) remains Pu3 -like, with some perturbations. The broad
+
2
double peak around 560 nm becomes a narrower single peak.
The single peak at 600 nm splits into two features at 597 and
6
03 nm, while the sharper peak at 662 nm is relatively
3+
unchanged. The Pu ‚CaM spectrum also shows a rising
background with increasing intensity toward shorter wave-
lengths, due to the protein absorbing strongly in the UV. A
titration experiment was done adding mole equivalents of Pu
sequentially to CaM, and the intensity of the Pu peaks in the
spectrum was plotted (Figure 3). The spectral peaks at 502,
3
3
+
3+
1
8
18
-2
densities were 0.72 × 10 and 0.63 × 10 cm . Over the
5
97, and 603 nm show a distinctive linear increase in intensity
period of the neutron-scattering experiments (several days), some
of the initial characteristic Pu blue color faded indicating some
oxidation of the Pu . Since there was no accompanying
evidence of Pu colloid formation and precipitation, we assume
3+
3+
with 1 through 4 mol equiv of Pu added, and then a flat
plateau. The peak at 662 nm which was unperturbed by
complexing Pu with CaM increases linearly for 1 through 6
mol equiv added. These results indicate that there are indeed
3
+
3
+
4+
the oxidized Pu remained bound to the protein. This conclusion
is supported by the fact that the neutron resonance scattering
3
+
four specific Pu -binding sites in CaM.

5122 J. Am. Chem. Soc., Vol. 119, No. 22, 1997
Seeger et al.
yielded the characteristic distance separation expected for the
bound Pu.
Neutron-Scattering Data Acquisition. Two independent
sets of neutron measurements were made on the Low-Q
Diffractometer (LQD) at the MLNSC on the two independent
sample preparations. To optimize for short-wavelength neu-
trons, three changes were made to the usual operational
procedures of LQD.23 First, the single-crystal MgO filter
normally used to remove short-wavelength neutrons was rotated
out of the beam. The time-of-flight range in the software
was adjusted to begin at an earlier time (corresponding to
0
0
.243-Å neutrons) and the time resolution was set to δt/t )
.5%. Finally, the multi-aperture collimation system with
24
five separate beams was used; this improves angular resolu-
tion by using smaller apertures, but simultaneously increases
count rate by using larger areas of the moderator and sample.
All of these changes required a recalibration of the instru-
ment. It had previously been determined that one of the
five aperture sets could not be aligned to the same point on
the detector, and hence had to be blocked with a mask of
Figure 4. Measured 240Pu transmission from the second neutron-
scattering experiment. The histogram is a theoretical calculation with
two free parameters.
Since the LQD is on a pulsed source, neutron time of flight
is used to measure the velocity (and hence wavelength) of each
neutron. The flight time is measured from the instant at which
protons strike the spallation target, producing the neutron pulse,
and is about 900 µs at the resonant wavelength. All data were
6
LiF. Short runs were also made with the “normal” LQD
configuration to measure the radius of gyration, Rg, of the
3
+
4
Pu ‚CaM.
In the first neutron experiment, the 4(240Pu )‚CaM sample
3+
collected as 3-dimensional (X, Y, t) histograms of counts per
was run for a total of 15.7 beam hours with an average proton
beam current on the neutron-producing target of 60 µA. An
identical cell containing only the Tris/KCl/D2O solvent was
measured for 7.4 h at an average current 64 µA. For each case
the sample transmission was measured by attenuating the beam
with 6.3 mm of BN and removing the beamstop from in front
of the detector for about 30 min. All runs were normalized to
integrated counts of the beam-current signal provided by the
facility. A short “blocked beam” run was also made, using the
same BN attenuator and with the beamstop in place. In
retrospect it was found that this signal was not useful in
determining non-beam-related background because the transmis-
sion of the BN near 0.278 Å was too large; lack of this
information meant that interpretation of the transmission data
was difficult, and only qualitative analysis of the data was
possible following the first experiment.
2
3
2
.4 × 3.4 mm detector pixel and per 0.5% time slice. Only
4 of the time slices (wavelengths of 0.261-0.295 Å) were used
directly for resonance analysis.
Neutron-Scattering Data Reduction and Analysis. The
process of data reduction is to use the signal and transmission
histograms, normalized and corrected for efficiencies and various
forms of backgrounds, to obtain I(Q) and eventually Frr(Q), the
2
40
form factor for the separations of the Pu atoms. The first
stage was to analyze the long-wavelength data to see if the
3
+
overall solution conformation of 4Pu ‚CaM is the same as
2+
4Ca ‚CaM. The probable frequency distribution of interatomic
vectors (P(r), calculated as the inverse Fourier transform of I(Q))
for the Pu-complexed CaM showed the same general features
2+
6
as that previously published for 4Ca ‚CaM , i.e. a peak at ∼18
Å, with a shoulder at ∼45 Å, and a maximum linear dimension
of ∼65 Å, indicating the overall dumbbell shape of the protein
is preserved. The radius of gyration derived from the second
moment of P(r) is Rg ) 22.3 ( 0.2 Å, which is ∼5% larger
For the second neutron experiment the liquid hydrogen
moderator normally serving LQD was replaced with an ambient
temperature water moderator which provided a spectrum of
neutrons that should have better favored the 2 Pu resonance
experiment (providing greater flux at 0.278 Å). In addition, a
beam-monitoring detector was installed in the beamstop, al-
lowing an independent direct measurement of the transmitted
40
2+
than the previously measured value for 4Ca ‚CaM in D2O
6
solution, 21.3 ( 0.2 Å . This small difference could reflect
3+
2+
differences in the Pu vs Ca induced conformational changes
within the individual globular domains of the protein. It is
known that metal binding to CaM results in a “hinge” opening
within each globular domain. Indeed such hinge opening is
flux during the experiment.25 The second 4( Pu )‚CaM
sample was run for 28.3 h at 68 µA, and the solvent was run
for 19.9 h at 74 µA. These runs were normalized by the
respective transmitted beam monitor counts. The blocked beam
run was made by using a thickness of 15.4 mm of BN with a
240
3+
2
+
26-28
observed upon Ca binding,
resulting in a 9% increase
in Rg while preserving the overall dumbbell conformation of
6
2+
the protein. The different charge-to-radius ratios for the Ca
and Pu
-
4
3+/4+
transmission less than 10 at the resonant wavelength for 6.8
h at 77 µA. The background measurement and the transmission
measurements from the second run made it possible to extract
quantitative information from the first data set. Both sets of
neutron data showed evidence for the same periodic term in
the resonance-scattering term. Shown here is the analysis of
the first data run, which had a flatter non-resonant-scattering
term, possibly due to the lower background scattering from the
liquid hydrogen moderator.
ions might be expected to result in different degrees
of hinge opening.
The sample transmission as measured by a 6LiF glass
2
5
scintillation detector mounted in the beamstop is illustrated
for the second neutron-scattering measurement in Figure 4. The
blocked-beam signal (which has been subtracted from both data
sets) is not a significant term for this detector. The solvent
signal was normalized at 80-100 µs on either side of the
(
26) Zhang, M.; Tanaka, T.; Ikura, M. Nature Struct. Biol. 1995, 2, 758-
767.
(27) Kuboniwa, H.; Tjandra, N.; Grzesiek, S.; Ren, H.; Klee, C. B.; Bax,
(
78.
23) Seeger, P. A.; Hjelm, R. J. P. J. Appl. Crystallogr. 1991, 24, 467-
4
(
(
24) Seeger, P. A. Physica (Utrecht) 1986, 136B, 106-109.
25) Seeger, P. A.; Hjelm, R. J. P; Nutter, M. J. Mol. Cryst. Liq. Cryst.
A. Nature Struct. Biol. 1995, 2, 768-776.
(28) Finn, B. E.; Even a¨ s, J.; Drakenberg, T.; Waltho, J. p.; Thulin, E.;
Fors e´ n, S. Nature Struct. Biol. 1995, 2, 777-783.
1
990, 180A, 101-117.

Specific Binding of Pu to Ca-Binding Sites of Protein
J. Am. Chem. Soc., Vol. 119, No. 22, 1997 5123
resonance of the 4(240Pu )‚CaM signal (a correction of 2.5%
from the beam current monitor signal) and then fitted with a
straight line in the vicinity of the resonance. The theoretical
3+
240
Pu cross section must be convoluted with Doppler broadening
and with the instrument time-of-flight resolution, which has
terms due to the width of the recorded time slice (4.47 µs), the
proton pulse width (0.27 µs), and the pulse shape distribution
of the moderator. This last function has been calculated by a
detailed Monte Carlo simulation of the MLNSC target/moderator
2
9,30
system
by an Ikeda-Carpenter function, (t/λ) e
of the neutrons emitted directly and 30% convoluted with an
and is very well fitted near the resonance wavelength
31
2 -(0.257t/λ)
, with 70%
-
(t/6.8)
exponential decay, e
(for t in µs and λ in Å). Doppler
Figure 5. Weighted averages of resonant scattering length, including
effects of the asymmetric neutron pulse shape, resonant absorption,
and Doppler broadening for the first neutron-scattering experiment.
3
2
240
broadening results from thermal motion of the Pu nuclei.
The velocities of the CaM molecules are small compared to
the neutron velocity, but the 2 Pu atoms are vibrating within
their CaM binding sites. If we neglect Doppler broadening,
the resulting computed resolution is too narrow, and if we use
the value of the mean-square velocity appropriate for metallic
Pu as in ref 32, it is too broad. We have therefore taken the
thermal width to be a free parameter and have chosen 12 meV
as the best fit. Since the resolution is asymmetric and the
transmission is nonlinear, Monte Carlo is the best technique
for applying both the Doppler and the resolution corrections.
A program using library MCLIB33 determined the number of
counts per time slice, averaged over all wavelengths which may
40
blocked-beam measurement, and suppose the relative normal-
ization factors of the three measurements are N , N , and N .
S
B
K
Then we can subtract the blocked-beam component of the
signals before normalizing the transmitted beams as
(S) - (K)N /NK (B) - (K)N /NK
S
B
I )
-
(
T )
(T )
S
B
(
S)
(B) (K) N
S
N
B
)
-
-
-
(7)
2
40
(TS)
(T )
NK
[
(TS)
(T )
]
contribute to the slice, as a function of the density of Pu
atoms. The histogram line in Figure 4 is thus a theoretical curve,
with density and Doppler width as the two free parameters. The
B
B
The second form is preferred because it allows propagation of
errors without correlations, and also shows (K) as a correction
to the standard form. Since (TS)/NS and (TB)/NB are exactly the
functions which were fitted in the construction of Figure 4, the
coefficient of (K) can be computed precisely. To minimize the
statistical uncertainties in the transmission measurements the
measured (TB) was fitted over the relevant time range with a
second-order curve and (TS) was computed by multiplying by
the transmission curve corresponding to Figure 5. The blocked-
beam correction was a major fraction of the signal and
contributed to the statistical uncertainty of the data points; it
did not, however, affect the final result.
240
area density of Pu atoms was found by iteration, giving 2.37
1
8
2
×
10 atoms/cm , which is 6% lower than 4 times the estimated
CaM density; this is partly because the sample did not
completely cover the exposed area of the cell, but may also
represent some small deterioration of the sample.
The histogram in Figure 4 shows that the wavelength
calibration is excellent and the shape of the resonance is very
well understood. We can therefore use this measurement,
corrected for the different 240Pu density, in the reduction of the
first neutron data set. Also, the Monte Carlo procedure provides
independent averages of λ(t) and its standard deviation,
Re{b(t)}, Im{b(t)}, and |b(t)| appropriately weighted by resolu-
tion and transmission probability. These averages depend on
For each of 24 time slices around the resonance, the detector
(X, Y, t) histogram was binned in a uniform Q-scale. For each
2
40
value of Q the three orthogonal form factor terms of eq 5 were
then extracted by least squares. Correlations of the errors with
the small F00 and F0r terms prevent assignment of simple error
bars to Frr obtained in this extraction. Therefore simple
expressions for F00 and F0r as a function of Q were subtracted
from the data and Frr was extracted again. The error bars could
then be propagated from the counting statistics (see Figure 6).
In this process all of the errors are assigned to Frr. Systematic
bias may occur in the processes of dividing by the computed
transmission and normalizing the background and blocked-beam
subtractions. The dashed line on Figure 6 is a linear correction
computed in the model-fitting process. The resolution in Q is
dominated by the size and shape of the neutron beam spot, which
is a triangular distribution with a measured full width half
maximum of 28 mm. The resulting Q resolution, with full width
the Pu density; e.g., the peak value of |b| is reduced because
fewer neutrons are transmitted at the peak. The terms needed
for the decomposition of eq 5 are shown for the first neutron
sample density in Figure 5, which may be compared to Figure
1
A. The wavelength resolution (rms), including the width of
the recorded slices, is 0.63%.
The usual normalization and background subtraction for LQD
data is to divide the recorded histogram for the signal, (S), by
the transmitted beam, (TS), and subtract the buffer, (B), divided
by its transmitted beam, (TB). If, however, there is a significant
difference in the two transmission functions, and if there is also
a significant background that is sample independent, then an
additional correction term is necessary. Let (K) represent a
(29) Pitcher, E. J.; Russell, G. J.; Seeger, P. A.; Fergusson, P. D. ICANS-
XIII; Paul Scherrer Institut: Villigen, Switzerland, Oct 11-14, 1995; PSI
-
1
half maximum ) 0.15 Å , is illustrated in Figure 6.
Proceedings 95-02, pp 323-329.
(30) P. D. Fergusson provided Monte-Carlo calculations of the perfor-
If we assume the CaM molecules are non-interacting, then
the range of the double sum of eq 6 is just the four binding
sites within the molecule, a total of sixteen terms. We further
simplify by assuming the distances between the two sites within
each globular domain are equal, d, based on the knowledge that
the domains are structurally homologous (the crystal structure
mance of the as-built MLNSC target-moderator system.
(
44.
31) Ikeda, S.; Carpenter, J. M. Nucl. Instrum. Methods 1985, A239, 536-
5
(
32) Seeger, P. A.; Taylor, A. D.; Brugger, R. M. Nucl. Instrum. Methods
1
985, A240, 98-114.
(
33) Seeger, P. A. ICANS-XIII, Paul Scherrer Institut: Villigen, Swit-
zerland, Oct 11-14, 1995; PSI Proceedings 95-02, 323-329. The Monte
Carlo program referenced here and its description are available by
anonymous ftp from azoth.lansce.lanl.goV/pub/mclib.
2
+
2+
of 4Ca ‚CaM shows the separations of Ca ions within each
globular domain to be 11.85 and 11.49 Å which are equal within

5124 J. Am. Chem. Soc., Vol. 119, No. 22, 1997
Seeger et al.
by less than 1%, and if the blocked-beam correction is reduced
by 50%, d becomes 11.6 Å.
Discussion and Conclusions
In the experiments reported here we were successful in
3
+
identifying conditions whereby Pu is stable in near neutral
pH solutions for 6-7 h. Tris(hydroxymethyl)aminomethane
3
+
was shown to chelate Pu sufficiently strongly to keep it from
oxidizing and precipitating over this time period, but the
chelation was also weak enough so that addition of CaM to the
3
+
3+
Pu /Tris solution resulted in the release of the Pu from the
tris(hydroxythmethyl)aminomethane and its binding to CaM.
The Pu/CaM titration data combined with the distance measure-
ment from the neutron resonance scattering demonstrate that
2+
the Pu binds specifically to the four Ca -binding sites in CaM.
This is the first demonstration of binding of Pu to specific sites
in a protein. We have also demonstrated that interatomic
distances of specific sites in macromolecules in solution may
be determined with a precision of 3% by small-angle neutron
scattering, by substitution of nuclei with strong nuclear reso-
nances, especially Pu. The value of 11.8 ( 0.4 Å obtained
for the average distance between the two Ca -binding sites in
the same globular domain of CaM agrees with the crystal-
Figure 6. Resonant scattering form factor, Frr, calculated for the
scattering data from the first neutron experiment. Error bars are
propagated from counting statistics; data-reduction procedures were
chosen to minimize correlations between points. Systematic bias
estimated from the least-squares model fit is indicated by the dashed
straight line. The triangular distribution is the instrumental (angular)
resolution. The resulting smeared model fit is the solid curve.
2
40
2
+
3
4
lographic average of 11.7 Å. Thus the measurement is both
precise and accurate. The precision of our measurement based
on data of relatively poor statistical quality (Figure 6) results
from using the detailed shape of the nuclear resonance to select
neutron wavelengths and fitting the model parameter over a
relatively wide Q range. The dynamic range of the experiment
was limited by the angular resolution of the small-angle
instrument collimation to distances of about 7-30 Å. Future
spallation neutron sources, now in design stages, will have the
much higher intensity required to support instruments with
higher angular resolution which will extend the available range
to both higher and lower values. Such sources will also make
it possible to do experiments with smaller samples and lower
concentrations, and/or with higher precision. Further, they will
facilitate use of other resonance scatterers (see the list in the
Introduction) that may have less “ideal” resonant properties
the precision of our measurements). We also assume the four
distances between sites in opposite domains may be represented
by an average, D. The resulting model is
Frr(Q) ) 4 + 4 sin(Qd)/Qd + 8 sin(QD)/QD
(8)
More generally, D may be replaced by a probability density
and the final term by an integral over the range of D.
Reasonable values for the range of D, from the crystal structures
3
4,9
of extended and of collapsed configurations of CaM, would
be from about 31-48 Å, with the average of the four distinct
distances being 45.8 and 36.1 Å for the two configurations
respectively.
Before fitting to the experimental data, the model must be
240
compared with Pu, but are easier to work with. Collimation
“smeared” by the instrumental resolution; this was accomplished
systems and detector shielding of future instruments must be
designed to minimize the “blocked-beam” contribution to the
background, which made a major contribution to the statistics
of the data in this experiment.
by a (triangular) weighted sum of eq 8 evaluated at 11 points
-1
in the range Q ( 0.15 Å . Weighted nonlinear regression was
performed in SigmaPlot,35 which uses the Marquardt-Leven-
berg algorithm. Although a precise and robust value of d was
found, the program was unable to solve for D. Constraining D
to fixed values in the range 20-60 Å, the fit improves very
slightly up to D ) 40 Å and is identical for all higher values.
This is because whenever the full width of the resolution
function covers more than two oscillations of the function, i.e.,
for D > 2 π/0.15 ) 43 Å, all information has been completely
lost. Since we begin to lose sensitivity at half this value, D >
The specific binding of Pu3 to the Ca -binding sites of a
key regulatory protein leads one to speculate on possible
mechanisms for health effects from Pu. Biochemical regulation
is critical to healthy function, and breakdown of regulation leads
to uncontrolled cell growth and proliferation and, if not
corrected, cell death. Nature regulates biochemical activity
through the use of a number of “messengers”, among which
the divalent calcium ion is perhaps the simplest and most
+
2+
2
2 Å, the experiment did not provide any direct information on
2
+
widespread. Ca -dependent regulation generally involves
equilibrium binding to a receptor protein. The kinetics of the
D in the range of interest. A value must be included for D,
however, to give the rise of the data at low Q. The curve
through the data in Figure 6 is the fitted result:
2+
“
on” and “off” rate constants for the Ca binding are delicately
balanced so that when there is a particular extracellular stimulus,
2+
an increase in local Ca concentrations within a cell is effected,
d ) 11.8 ( 0.4 Å
2+
which in turn results in Ca binding to specific proteins that
eventually cause specific pathways to be switched “on”. A
pathway is switched “off” again when the local Ca concentra-
with D set to ∼40 Å. This value for d is robust in the sense
that it does not depend strongly on the resolution function. For
instance, if Doppler broadening is neglected the value d changes
2+
2+
2+
tion drops. Many Ca -binding regulatory proteins have Ca -
binding sites that strongly resemble those in CaM. The fact
that the 4Pu‚CaM samples were stable for several days suggests
(
34) Babu, Y. S.; Bugg, C. E.; Cook, W. J. J. Mol. Biol. 1988, 204,
91-204.
35) Jandel Corporation. SigmaPlot for Windows, Version 3.0, 1995.
3
+
the Pu binding is very tight, consistent with our expectation
based on its higher charge-to-radius ratio compared with Ca .
1
2+
(

Specific Binding of Pu to Ca-Binding Sites of Protein
J. Am. Chem. Soc., Vol. 119, No. 22, 1997 5125
4
+
Other ionic species of Pu, including the more stable Pu , might
Acknowledgment. This work was performed under the
auspices of the United States Department of Energy, under
contract No. W-7405-ENG-36 to the University of California.
The neutron work was done at the Manuel Lujan Jr. Neutron
Scattering Center, a national user facility funded by DOE Office
of Basic Energy Sciences. The project was also supported by
DOE Office of Health and Environmental Research project KP-
04-01-00-0. Additional support was provided by National
Institutes of Health Grant No. GM40528.
bind even more tightly; thus even at low Pu concentrations there
could be permanent binding to Ca2 regulatory sites resulting
in complete disruption of the normal kinetics. One is tempted,
therefore, to speculate that Pu binding to regulatory proteins
could be a factor in breakdown of cellular regulation and
subsequent expression of toxic effects, or even neoplastic
transformation of a cell. Such mechanisms of metal toxicity
+
have been discussed.1
5,16
Binding to Ca regulatory sites could
2+
also provide a means by which Pu is transported directly to
DNA-containing sites in cells, thus facilitating radioactive
damage to DNA.
JA9633124