A metal-chelating lipid for 2D protein crystallization via coordination of surface histidines

Article abstract of DOI:10.1021/ja964099e

Two-dimensional protein crystallization on lipid monolayers is becoming a powerful technique for structure determination as well as materials applications. However, progress has been hindered by the requirement of a unique affinity lipid for each new protein of interest. Metal ion coordination by surface-accessible histidine side chains provides a convenient and general method for targeting of proteins to surfaces. Here we present the synthesis and characterization of a metal-chelating lipid which has been designed to target proteins to Langmuir monolayers and promote their two-dimensional crystallization based on histidine coordination. The lipid utilizes the metal chelator iminodiacetate (IDA) as the hydrophilic headgroup and contains unsaturated, oleyl tails to provide the fluidity necessary for two-dimensional protein crystallization. The lipid is shown to bind copper from the subphase strongly when incorporated in Langmuir monolayers. In addition, it is possible to form copper-containing monolayers by spreading the premetalated lipid on the subphase in the absence of copper. Fluorescence microscopy reveals the binding and crystallization of the protein streptavidin, promoted by the simultaneous coordination of two surface-accessible histidine side chains to the IDA-Cu lipid.

Full text of DOI:10.1021/ja964099e

J. Am. Chem. Soc. 1997, 119, 2479-2487
2479
A Metal-Chelating Lipid for 2D Protein Crystallization Via
Coordination of Surface Histidines
Daniel W. Pack, Guohua Chen, Kevin M. Maloney, Chao-Tsen Chen,† and
Frances H. Arnold*
Contribution from the DiVision of Chemistry and Chemical Engineering, 210-41,
California Institute of Technology, Pasadena, California 91125
ReceiVed NoVember 27, 1996X
Abstract: Two-dimensional protein crystallization on lipid monolayers is becoming a powerful technique for structure
determination as well as materials applications. However, progress has been hindered by the requirement of a unique
affinity lipid for each new protein of interest. Metal ion coordination by surface-accessible histidine side chains
provides a convenient and general method for targeting of proteins to surfaces. Here we present the synthesis and
characterization of a metal-chelating lipid which has been designed to target proteins to Langmuir monolayers and
promote their two-dimensional crystallization based on histidine coordination. The lipid utilizes the metal chelator
iminodiacetate (IDA) as the hydrophilic headgroup and contains unsaturated, oleyl tails to provide the fluidity necessary
for two-dimensional protein crystallization. The lipid is shown to bind copper from the subphase strongly when
incorporated in Langmuir monolayers. In addition, it is possible to form copper-containing monolayers by spreading
the premetalated lipid on the subphase in the absence of copper. Fluorescence microscopy reveals the binding and
crystallization of the protein streptavidin, promoted by the simultaneous coordination of two surface-accessible histidine
side chains to the IDA-Cu lipid.
Introduction
lowed near-atomic resolution structures to be obtained from 2D
streptavidin crystals grown on biotinylated lipid monolayers.12
Furthermore, 2D protein crystals may be useful for seeding the
epitaxial growth of three-dimensional protein crystals13,14 and
may allow the crystallization of proteins or multiprotein
complexes which are difficult to crystallize by other methods.
2D protein crystallization requires the immobilization and
orientation of protein at the monolayer surface in high concen-
tration, without loss of mobility.15 Lateral and rotational
mobilities of the bound protein are required to allow its self-
organization, formation of the optimum intermolecular contacts,
and efficient packing within the crystal. This mobility is
supplied largely by the lipid monolayer. Crystallization is
generally performed using fluid-phase monolayers, and lipids
with unsaturated hydrocarbon tails are employed to provide this
fluidity.16-19
To date, a significant number of proteins and peptides have
been crystallized on Langmuir monolayers.15 Various classes
of interactions have been employed for targeting proteins to
the interface. In most cases, specific ligands have been used
to promote crystallization of antibodies, enzymes, polypeptide
toxins, and other proteins (e.g., streptavidin, annexin VI). In
each instance, a unique lipid containing the appropriate affinity
ligand was required, and in many cases a novel synthesis of
Living cells are capable of utilizing their inherently ordered
cellular membranes as a template to direct the growth of
organized materials. One example is the growth of two-
dimensional protein or glycoprotein crystals known as S-layers
which form the outermost layer of a wide variety of Gram-
positive and Gram-negative bacteria.1-3 In their natural envi-
ronments, S-layers function as structure-maintaining elements,
protective coats, molecular sieves, and recognition elements.
These naturally ordered structures have been employed as
molecular filters with a steep exclusion cutoff,4 patterned
immobilization and affinity matrices,5 biosensors,6 and pattern-
ing masks for nanolithography.7
2D protein crystals can also be grown ex ViVo, on Langmuir
monolayers of lipids.8,9 Crystalline protein arrays have ap-
plications in such diverse fields as structural biology and
materials science. Perhaps the most obvious application of 2D
protein crystals is for structure determination by electron
diffraction.10,11 Recent technological developments have al-
* To whom correspondence should be addressed. Fax: (818) 568-8743.
E-mail: frances@cheme.caltech.edu.
† Current address: Box 768 Havemeyer Hall, Columbia University, New
York, NY 10027.
X Abstract published in AdVance ACS Abstracts, February 15, 1997.
(1) Messner, P.; Sleytr, U. B. AdV. Microb. Physiol. 1992, 33, 213-
275.
(11) Jap, B. K.; Zulauf, M.; Scheybani, T.; Hefti, A.; Baumeister, W.;
Aebi, U.; Engel, A. Ultramicroscopy 1992, 46, 45-84.
(12) Avila-Sakar, A. J.; Chiu, W. Biophys. J. 1996, 70, 57-68.
(13) Edwards, A. M.; Darst, S. A.; Hemming, S. A.; Li, Y.; Kornberg,
R. D. Struct. Biol. 1994, 1, 195-197.
(2) Beveridge, T. J. Curr. Opin. Struct. Biol. 1994, 4, 204-212.
(3) Taylor, K. A.; Deatherage, J. F.; Amos, L. A. Nature 1982, 299,
840-842.
(4) Sa´ra, M.; Sleytr, U. B. J. Membr. Sci. 1987, 33, 27-49.
(5) Sa´ra, M.; Ku¨pcu¨, S.; Weiner, C.; Weigert, S.; Sleytr, U. B. In
AdVances in Bacterial Paracrystalline Surface Layers; Beveridge, T. J.,
Koval, S. F., Eds.; Plenum Press: New York, 1993; pp 195-204.
(6) Pum, D.; Sa´ra, M.; Sleytr, U. B. In Immobilized Macromolecules:
Application Potential; Sleytr, U. B., Messner, P., Sa´ra, M., Pum, D., Eds.;
Springer-Verlag: London, 1992; pp 141-160.
(14) Darst, S. A.; Edwards, A. M. Curr. Opin. Struct. Biol. 1995, 5,
640-644.
(15) Kornberg, R. D.; Darst, S. A. Curr. Opin. Struct. Biol. 1991, 1,
642-646.
(16) Ribi, H. O.; Reichard, P.; Kornberg, R. D. Biochemistry 1987, 26,
7974-7979.
(17) Lebeau, L.; Regnier, E.; Schultz, P.; Wang, J. C.; Mioskowski, C.;
Oudet, P. FEBS Lett. 1990, 267, 38-42.
(7) Douglas, K. In Biomimetic Materials Chemistry; Mann, S., Ed.; VCH
Publishers, Inc.: New York, 1996; pp 117-142.
(18) Lebeau, L.; Oudet, P.; Mioskowski, C. HelV. Chim. Acta 1991, 74,
1697-1706.
(8) Fromherz, P. Nature 1971, 231, 267-268.
(9) Uzgiris, E. E.; Kornberg, R. D. Nature 1983, 301, 125-129.
(10) Newman, R. H. Electron Microsc. ReV. 1991, 4, 197-203.
(19) Kubalek, E. W.; Le Grice, S. F. J.; Brown, P. O. J. Struct. Biol.
1994, 113, 117-123.
S0002-7863(96)04099-1 CCC: $14.00 © 1997 American Chemical Society

2480 J. Am. Chem. Soc., Vol. 119, No. 10, 1997
Pack et al.
this lipid was necessary. For many potentially interesting
proteins, however, no convenient affinity ligand is available.
Nonspecific electrostatic interactions may be sufficient to induce
crystallization when the surface charge on the protein is
asymmetric, promoting adsorption with a specific orientation,
but this approach is not generally applicable. The difficulty of
identifying and obtaining appropriate lipids for targeting and
crystallization is certainly a serious limitation to progress in
this field.
It has recently been shown that proteins can be targeted to
lipid surfaces through coordination of amino acid side chains
to lipid-bound metal ions.20-23 At neutral pH, histidine side
chains are primarily responsible for the affinity of metal ions
for protein surfaces (e.g., Ka ≈ 103.5 M-1 for a single histidine
interaction with iminodiacetate-chelated Cu2+); at slightly basic
pH, other groups such as lysine and the N-terminus can also
contribute to metal binding.24 This interaction should be useful
for 2D crystallization studies of proteins displaying surface
histidines. In fact, nitrilotriacetate-nickel(II) (NTA-Ni) lipids
have been utilized by Kubalek et al. for the 2D crystallization
of a polyhistidine-tagged HIV-1 reverse transcriptase.19 Imi-
nodiacetate-copper (IDA-Cu) lipids provide a higher affinity
for histidine than NTA-Ni and can therefore effectively target
proteins Via coordination of naturally-occurring surface his-
tidines which lipids containing the NTA-Ni complex cannot.20
Thus, IDA-Cu lipids can potentially extend this approach to a
wide variety of proteins with solvent-accessible histidines, either
natural or genetically engineered.
Here we report the synthesis and monolayer characterization
of the metal-chelating lipid 1,1′-[[9-[2,3-bis[(Z)-octadec-9-
enyloxy]propyl]-3,6,9-trioxanonyl]imino]diacetic acid (DOIDA)
which was specifically designed for 2D crystallization. Unsat-
urated oleyl tails are utilized to provide the required monolayer
fluidity. The metal chelator IDA, which strongly binds divalent
transition metal ions such as copper (Ka ≈ 1010.6 M-1),25 is used
as the hydrophilic headgroup. The IDA is attached to the
glycerol backbone Via a triethylene glycol spacer (∼11.5 Å long)
which ensures the IDA-Cu complex is displayed in the aqueous
subphase and is available for ligand binding. The utility of this
lipid for 2D protein crystallization is demonstrated using the
tetrameric protein streptavidin, known for its ability to very
tightly bind 4 equiv of the small molecule biotin (Ka ≈ 1015
M-1).26 Streptavidin has been crystallized previously in two
dimensions on Langmuir monolayers of biotin-containing lip-
ids27 and on monolayers of poly(1-benzyl-L-histidine) Via
nonspecific, electrostatic interactions.28 Our initial observations
of streptavidin crystallization on DOIDA-Cu monolayers using
Brewster angle microscopy have been reported previously.29 In
the present work, fluorescence microscopy is used to confirm
and further characterize the streptavidin crystallization process.
Figure 1. (a) MsCl, CH2Cl2, Et3N, 12 h. (b) Trityl chloride, DMAP,
Et3N, CH2Cl2, 12 h, 42%. (c) Trityl chloride, DMAP, Et3N, THF, 12
h, 76%. (d) MsCl, CH2Cl2, Et3N, 12 h, 86%. (e) Powdered KOH,
DMSO, 80 °C, 36 h, 33%. (f) TsOH, THF/CH3OH, room temperature,
12 h, 67%. (g) NaH, 1-(methylsulfonyl)-9-(triphenylcarbinyl)-3,6,9-
trioxynonane (2), THF, reflux, 12 h. (h) TsOH, THF/CH3OH, room
temperature, 12 h, 90%. (i) CBr4, PPh3, THF, 0 °C to room temperature,
12 h, 75%. (j) Diethyl iminodiacetate, Et3N, THF, 72 h, 21%. (k) NaOH,
THF/MeOH/H2O, reflux, 1 h, 41%.
Results
Lipid Synthesis. Dioleyl lipid 12 (DOIDA) was synthesized
as outlined in Figure 1. DOIDA features a glycerol backbone
connecting two ether-linked oleyl tails at the 2 and 3 positions
and a hydrophilic spacer to the iminodiacetate headgroup on
the 1 position. Trityl-protected glycerol 4 and oleyl mesylate
6 were coupled to produce 7. Deprotection of the trityl group
and esterification with 1-(methylsulfonyl)-10-(triphenylcarbinyl)-
1,4,7,10-tetraoxadecane (2), prepared from triethylene glycol,
introduced the triethylene glycol spacer group to the main lipid
body 9. The hydroxyl group of 9 was then substituted with
bromine, which was subsequently substituted by diethyl imi-
nodiacetate. Hydrolytic cleavage of the ethyl ester groups of
11 led to the final product DOIDA (12).
Langmuir Monolayer Characterization of DOIDA. Sur-
face pressure-molecular area (π-A) isotherms are sensitive to
changes in molecular organization of amphiphiles at the air-
buffer interface and can reflect the presence of molecules in
the aqueous subphase which are capable of interacting with the
monolayer components.30 Changes in the π-A isotherms of
the metal-chelating DOIDA have been used to observe Cu2+
binding by the IDA headgroups. In order to check for oxidation
(20) Shnek, D. R.; Pack, D. W.; Sasaki, D. Y.; Arnold, F. H. Langmuir
1994, 10, 2382-2388.
(21) Ng, K.; Pack, D. W.; Sasaki, D. Y.; Arnold, F. H. Langmuir 1995,
11, 4048-4055.
(22) Dietrich, C.; Schmitt, L.; Tampe´, R. Proc. Natl. Acad. Sci. U.S.A.
1995, 92, 9014-9018.
(23) Dietrich, C.; Boscheinen, O.; Scharf, K.-D.; Schmitt, L.; Tampe´,
R. Biochemistry 1996, 35, 1100-1105.
(24) Johnson, R. D.; Todd, R. J.; Arnold, F. H. J. Chromatogr. 1996,
725, 225-235.
(25) Martell, A. E.; Smith, P. M.Critical Stability Constants; Plenum
Press: New York, 1974; Vol. 6.
(26) Green, N. M. AdV. Protein Chem. 1975, 29, 85-133.
(27) Darst, S. A.; Ahlers, M.; Meller, P. H.; Kubalek, E. W.; Blankenburg,
R.; Ribi, H. O.; Ringsdorf, H.; Kornberg, R. D. Biophys. J. 1991, 59, 387-
396.
(28) Furuno, T.; Sasabe, H. Biophys. J. 1993, 65, 1714-1717.
(29) Frey, W.; Schief Jr., W. R.; Pack, D. W.; Chen, C.-T.; Chilkoti, A.;
Stayton, P.; Vogel, V.; Arnold, F. H. Proc. Natl. Acad. Sci. U.S.A. 1996,
93, 4937-4941.
(30) Gaines, G. L. Insoluble Monolayers at Liquid-Gas Interfaces; Wiley
(Interscience): New York, 1966.

A Metal-Chelating Lipid for 2D Protein Crystallization
J. Am. Chem. Soc., Vol. 119, No. 10, 1997 2481
grow, and the monolayer became very stiff as evidenced by its
resistance to convection. Protein aggregates of all sizes tend
to associate with one another, as groups of aggregates are more
common than individual “tiles”. Furthermore, the tiles were
observed to interlock primarily at the corners or on the flat sides;
contacts between corners and flat sides were rare. However,
the sharp edges and square corners were maintained even in
very dense groupings of large tiles. (It should be noted that
the relatively bright background in these images is due to free,
out-of-focus protein in solution through which the excitation
and emission light must pass, due to the inverted microscope
configuration.)
Figure 2. Surface pressure-molecular area isotherms of pure DOIDA
lipid in the presence of varying amounts of CuCl2: (a) no divalent
metal; (b) 0.20 µM CuCl2; (c) 0.35 µM CuCl2; (d) 0.50 µM CuCl2; (e)
1.0 µM CuCl2; (f) 10 µM CuCl2 on a buffered subphase (20 mM MOPS,
100 mM NaCl, pH 7.5, 25 °C).
Two control experiments indicate that streptavidin binding
by the IDA lipid requires the presence of copper ions. Injection
of the soluble metal chelator ethylenediaminetetraacetate (EDTA)
into the subphase beneath mature streptavidin crystals caused
the protein crystals to disappear within 15-30 min. EDTA can
remove Cu2+ from the IDA lipid (log K ) 18.7 compared to
log K ) 10.6 for IDA25). In addition, no protein was bound to
the interface after injection of the same amount of streptavidin
of the unsaturated tails, π-A isotherms were measured on
monolayers under both Ar and ambient atmospheres. No
difference in the π-A isotherms was observed, and the
experiments described herein were performed under ambient
atmosphere.
Isotherms of Langmuir monolayers of DOIDA with varying
amounts of CuCl2 (0-10 µM) in the buffered subphase are
shown in Figure 2. Owing to their unsaturated tails, the DOIDA
lipids remain in an expanded phase throughout the compression;
no phase transition is observed in the isotherms at any surface
pressure or any copper concentration. However, we do observe
a decrease in area with increasing copper concentration in the
subphase. Adding 10 µM CuCl2 to the subphase results in a
decrease in the take-off area from 145 to 120 Å2/molecule and
a similar, yet smaller, decrease in area at all pressures (e.g.,
from 75 to 66 Å2/molecule at π ) 20 mN/m).
For many applications, including 2D crystallization of
streptavidin, it is necessary to form a metal-containing mono-
layer in the absence of free Cu2+ in the subphase. Thus,
monolayers were spread from a premetalated lipid solution; the
lipid headgroups were loaded with copper by adding a slight
molar excess (less than 10 mol %) of CuCl2 in MeOH to the
lipid solution in CHCl3. Isotherms of the premetalated lipids
on a metal-free subphase are identical to those of the metal-
free lipid spread on a subphase containing 10 µM CuCl2 (data
not shown).
2D Protein Crystallization. Because Cu2+ in solution was
found to inhibit binding and crystallization of streptavidin even
under a biotinylated lipid monolayer,29 the chelating lipids were
loaded with Cu2+ prior to spreading the monolayers. In each
experiment, streptavidin was injected beneath the DOIDA-Cu
monolayer at a surface pressure of 3-4 mN/m to final protein
concentrations in the aqueous subphase of 20-120 nM.
Although the presence of Ar had no effect on π-A isotherms
of DOIDA, all crystallization experiments with DOIDA-Cu
were performed under an inert Ar atmosphere to prevent
oxidation of the unsaturated tails during the often long time
period (∼4 h) required for crystallization.
underneath a DOIDA monolayer in the absence of Cu2+
.
Crystallization experiments were carried out under both
constant surface pressure and constant area conditions. Domain
formation and growth proceeded in a similar fashion in both
cases. Protein binding could be observed through changes in
the surface pressure-molecular area behavior of the monolayer.
Typical plots of the changes in surface pressure and molecular
area as functions of time are shown in Figure 4. In constant
area experiments, surface pressure began to increase im-
mediately after protein injection (Figure 4a). The first ag-
gregates appeared after a pressure increase of 1.8 mN/m. As
the tiles grew, the surface pressure continued to increase and
reached saturation (∆π ) 5 mN/m) at about the same time the
tiles stopped growing and the monolayer became stiff. Similar
results were obtained in constant surface pressure experiments
(Figure 4b). The first aggregates appeared after an area increase
(∆A/A) of 0.20. The area continued to increase as the tiles grew,
and the area change reached saturation (∆A/A ) 0.45) as the
streptavidin tiles reached their maximum size (∼75 × 75 µm2).
Polarized fluorescence microscopy was used to investigate
the degree of ordering and visualize the relative orientations of
the streptavidin domains. In this experiment, two polarizing
filters are placed on a slider in the excitation light path with
their directions of polarization perpendicular to one another.
Fluorophores whose absorption dipoles are aligned with the
polarization of the excitation light are selectively excited and
are, therefore, observed. Fluorophores whose absorption dipoles
are aligned perpendicular to the excitation light do not absorb
and do not emit fluorescence. An amorphous phase, in which
fluorophores are randomly aligned, will emit fluorescence of
equal intensity regardless of the orientation of the polarizing
filters. In contrast, a crystalline phase is expected to display a
single orientation of fluorophores. Thus, a crystal which appears
bright when viewed through a particular polarizer will appear
darker when viewed through a perpendicularly oriented polar-
izer.
Fluorescence micrographs demonstrating the crystallization
of rhodamine-labeled streptavidin on a DOIDA-Cu monolayer
are shown in Figure 3. In this experiment, the first aggregates
began to appear nearly 2 h after protein was injected. The
earliest observed aggregates are irregular in shape. At longer
times, the aggregates took on a characteristic rectangular shape.
The aggregates continued to grow in size over approximately 4
h, maintaining the rectangular shape at all times. Finally, the
aggregates became quite large (∼50 × 50 µm2), covering most
of the monolayer area. At this point, the aggregates ceased to
Pairs of fluorescence micrographs showing streptavidin
domains viewed under perpendicular orientations of the polar-
izers are shown in Figure 5. These tiles, while appearing to be
largely aggregated with one another, retain the characteristic
sharp edges and square corners of crystals grown on DOIDA-
Cu monolayers. The inverted contrast in each pair of images
is readily apparent, suggesting the protein domains are indeed
crystalline.31 Surprisingly, the tiles appear to be oriented with
one another within individual aggregates, indicating the tiles

2482 J. Am. Chem. Soc., Vol. 119, No. 10, 1997
Pack et al.
Figure 3. Time sequence of fluorescence micrographs of 2D streptavidin self-assembly beneath a DOIDA-Cu monolayer (3 mN/m) on a buffered
subphase (20 mM MOPS, 250 mM NaCl, pH 7.8) at room temperature. The streptavidin concentration is ∼80 nM. Images were taken (a) 115, (b)
125, (c) 150, (d) 190, (e) 200, and (f) 250 min after protein injection. Scale bar ) 100 µm.
either arose from a single nucleation site or, more likely, became
aligned as they aggregated.
In addition, biotin-loaded streptavidin failed to bind DOIDA-
Cu monolayers or an IDA-Cu-derivatized chromatography
matrix.29 These results are rather surprising as biotin is not
expected to affect the IDA-Cu-His complex directly, and its
binding by streptavidin causes only small perturbations in the
protein structure.32
A variety of biotinylated molecules spanning a wide range
of molecular sizes is capable of disrupting the protein domains.
Addition of biotinylated fluorescein (see the Experimental
Section), at a 4-fold molar excess over the streptavidin
concentration, resulted in disappearance of protein from the
interface within 5 min. Injection of a biotinylated lysozyme
caused the disappearance of streptavidin from the interface
Biotin Disrupts 2D Streptavidin Domains. Mature strepta-
vidin domains grown on DOIDA-Cu monolayers rapidly
disappeared upon injection of biotin into the subphase at 4-40-
fold molar excess over the total protein concentration. Even
when stoichiometric quantities of biotin (4-fold excess) were
added, the protein disappeared from the interface within 5 min.
(31) Polarization fluorescence microscopy of the streptavidin domains
requires the protein to be labeled at a unique site. The succinimidyl ester
chemistry preferentially labels lysine residues and is therefore expected to
label a number of residues on the protein surface. However, the polarized
fluorescence results indicate the streptavidin is, in fact, labeled at a unique
site. Similar results were obtained previously with fluorescein isothiocyanate-
labeled streptavidin (Blankenburg, R.; Meller, P.; Ringsdorf, H; Salesse,
C. Biochemistry 1989, 28, 8214-8221.).
(32) Weber, P. C.; Ohlendorf, D. H.; Wendoloski, J. J.; Salemme, F. R.
Science 1989, 243, 85-88.

A Metal-Chelating Lipid for 2D Protein Crystallization
J. Am. Chem. Soc., Vol. 119, No. 10, 1997 2483
indicates the lipid is in a fluid state at all surface pressures. In
fact, the shape of the π-A curve (Figure 2) is very similar to
that of fluid dioleoylphosphocholine monolayers.34 Chelation
of Cu2+ by the IDA headgroups results in a condensation of
the DOIDA monolayer. However, the lipid remains in the fluid
state at all copper concentrations. The fluid state is required
for growth of 2D protein crystals.10,15 The condensation of the
DOIDA monolayers is very similar to that observed for
analogous IDA lipids with differing hydrophobic tails.35
The decrease in monolayer area upon metal binding is most
likely due to neutralization of the headgroup charge.35 Changes
in the π-A isotherm cease as the CuCl2 concentration is
increased above 0.50 µM; π-A isotherms obtained for mono-
layers on 0.50, 1.0, and 10 µM CuCl2 are identical. Thus, it is
believed the IDA headgroups are saturated with chelated Cu2+
at and above 0.50 µM CuCl2. Given a saturation point between
0.35 and 0.50 µM CuCl2, one can estimate a lower limit on the
apparent affinity of Cu2+ for the IDA headgroup of (1.3-1.8)
× 108 M-1 at pH 7.5. This association constant is smaller than
that of IDA for Cu2+ in bulk solution (4 × 1010 M-1).25 The
discrepancy in binding affinity is likely a result of steric or
electrostatic effects due to the very high local concentration of
IDA at the interface.
Lipids which have been premetalated in solution prior to
spreading of the monolayer result in π-A isotherms identical
to the unmetalated lipid spread on subphase containing greater
than 0.50 µM CuCl2. Thus, premetalated DOIDA-Cu mono-
layers are believed to be fully loaded with copper. Kinetic
studies of the π-A behavior of the monolayers have determined
that premetalated IDA lipids retain ∼75% of the Cu2+ in the
headgroup even after 10 h at the air-buffer interface.36
Streptavidin Crystallization via Metal Coordination.
Streptavidin forms two-dimensional domains on Langmuir
monolayers of the metal-chelating lipid DOIDA. Several lines
of evidence indicate that these domains are crystalline. Firstly,
their sharp edges and consistent shape, both during growth in a
single experiment and from experiment to experiment, strongly
suggest the domains are crystalline. Secondly, the polarized
fluorescence microscopy results (Figure 5) support their ordered
nature. Recently, Fourier analysis of an electron micrograph
of DOIDA-Cu-bound streptavidin aggregates indicates the
protein is ordered and has shown lattice parameters identical to
those of biotin lipid-bound streptavidin crystals.37 In addition,
a filtered image in negative stain produced a low-resolution
projection map of the electron density37 very similar to the
known 2D projection of biotin-bound streptavidin crystals at a
comparable resolution.27,38 Streptavidin is known to bind to
IDA-Cu lipids Via His 87, which is located on the same face
of the protein as the entrance to the biotin binding pockets.29
This fact, in combination with the apparent identity of the
DOIDA-Cu- and biotin lipid-bound crystal parameters, suggests
the protein is oriented similarly in the two cases, with two biotin
binding pockets adjacent to the monolayer and two pockets
facing away from the monolayer, exposed to the aqueous
subphase.
Figure 4. Surface pressure (a) and relative monolayer area change
(b) of a DOIDA-Cu monolayer on a buffered subphase (20 mM
MOPS, 250 mM NaCl, pH 7.8) during 2D streptavidin crystallization.
The time of protein injection is indicated by the gray arrows.
Fluorescence micrographs at 62, 184, and 278 min in (a) show the
crystal growth in relation to the pressure change. Fluorescence
micrographs at 150, 220, and 270 min in (b) show the crystal growth
in relation to the area change.
within 15 min.33 Finally, a bisbiotin linker was prepared and
loaded into a solution of streptavidin. Each of the biotin binding
pockets contained one bisbiotin molecule, and the resulting
protein displayed four biotin moieties capable of further binding
to apostreptavidin. Addition of bisbiotin streptavidin beneath
mature tiles of apostreptavidin disrupted the tiles, and the protein
completely disappeared from the interface in approximately 45
min.
Discussion
Crystals could be produced from subphase streptavidin
concentrations as low as 20 nM. This concentration appears
low, given that only two simultaneous IDA-Cu-His 87
interactions are possible, which could indicate that protein-
protein interactions play a significant role in stabilization of
Metal Chelation by DOIDA at the Air-Buffer Interface.
The relatively large compressibility of the DOIDA monolayers
(33) Native, unbiotinylated lysozyme in the same concentration also
disrupted streptavidin tiles. However, without biotin the process of crystal
disruption was accompanied by a large increase in the surface pressure to
>20 mN/m and required more than 45 min. Thus, disruption of streptavidin
domains by native lysozyme appears to occur by a different mechanism.
Most likely, lysozyme binds the IDA-Cu headgroups outside the strepta-
vidin domains, increasing the surface pressure; lysozyme is known to have
significant affinity for immobilized metal ions. The surface pressure increase
could be responsible for the disappearance of protein from the interface
(Vogel, V.; Schief Jr., W. R.; Frey, W. Supramol. Sci., in press).
(34) Bohorquez, M.; Patterson, L. K. Langmuir 1990, 6, 1739-1742.
(35) Pack, D. W.; Arnold, F. H. Chem. Phys. Lipids, in press.
(36) Pack, D. W. Ph.D. Thesis, California Institute of Technology, 1996.
(37) Chang, W.-H.; Kornberg, R. D. Personal communication.
(38) Ku, A. C.; Darst, S. A.; Robertson, C. R.; Gast, A. P.; Kornberg,
R. D. J. Phys. Chem. 1993, 97, 3013-3016.

2484 J. Am. Chem. Soc., Vol. 119, No. 10, 1997
Pack et al.
Figure 5. Fluorescence micrographs of streptavidin aggregates grown on DOIDA-Cu monolayer on a buffered subphase (20 mM MOPS, 250
mM NaCl, pH 7.8) imaged through polarizing filters. Images on the right and left were taken under perpendicular orientations of the polarizers.
White arrowheads indicate equivalent points on the monolayer in each of the pairs of images. Scale bar ) 100 µm.
streptavidin binding to the DOIDA-Cu monolayer. In addition,
at pH 7.8, other basic groups on the protein such as lysine and
the N-terminus can participate in metal coordination and may
be contributing to the apparent high affinity of streptavidin for
the IDA-Cu lipids.
monolayer area observed during crystallization (Figure 4). We
might expect crystallization on IDA-Cu lipid monolayers to
be more sensitive to surface pressure than that on biotin lipid
monolayers due to the vastly larger affinity of streptavidin for
the biotin lipids. This larger binding energy will allow the
protein to remain attached to the lipids at higher surface
pressures where steric interactions with the monolayer make
adsorption less favorable. In fact, streptavidin crystals grown
on biotin lipid may be compressed to surface pressures as high
as 32 mN/m with no effect on the crystal morphology.39
Biotin Binding Affects Metal Binding to Streptavidin. The
ability to decorate the crystal surface by binding additional
biotinylated molecules to the exposed pockets is desirable for
a variety of applications.40-42 However, addition of biotin, or
biotinylated molecules spanning a wide range of sizes, to the
subphase below mature streptavidin crystals caused the protein
to disappear from the monolayer. The fluorescein-biotin
compound disrupted DOIDA-Cu-bound streptavidin crystals
as quickly as did free biotin (within 5 min). Lysozyme (MW
14 300) is small enough (∼45 × 30 × 30 Å3) to access both
biotin binding pockets simultaneously (given ∼3600 Å2 per
streptavidin27) and disrupted the mature streptavidin crystals in
15 min. Similarly, bisbiotin-containing streptavidin disrupted
mature apostreptavidin crystals within 45 min of its injection
into the subphase. Loading streptavidin with biotin also
Streptavidin crystallization on DOIDA-Cu was performed
at surface pressures near 3 mN/m. Higher pressures greatly
reduced the formation of crystals, and in fact, compression of
the monolayer to pressures of 15-20 mN/m destroys the mature
crystals.39 Such low surface pressures in the IDA-Cu lipid
monolayers may be required in order to provide sufficient
mobility of the lipid and bound protein. However, this seems
unlikely as DOIDA-Cu monolayers appear to be in the liquid
state at all pressures and are thus expected to be very mobile.
Streptavidin crystallization on biotinylated lipid monolayers
is known to depend on the length of the spacer attaching the
biotin to the lipid backbone. Very long spacers allow crystal-
lization over a large range of surface pressures, shorter spacers
allow crystallization only in the fluid state, and, with no spacer,
crystallization occurs only in noncompressed monolayers.27 The
longer spacer is believed to decrease penetration of the protein
into the monolayer. Due to differences in (1) the IDA-Cu and
biotin lipid structures and (2) the binding site on the protein, it
is unclear in which class of spacer length the IDA-Cu lipids
should be placed. Nevertheless, the inhibition of crystallization
with increasing surface pressure is likely a result of decreasing
ability of streptavidin to penetrate the DOIDA-Cu monolayer.
A requirement for partial penetration of the protein into the lipid
bilayer also explains the increase in surface pressure or
(40) Ahlers, M.; Blankenburg, R.; Grainger, D. W.; Meller, P.; Ringsdorf,
H.; Salesse, C. Thin Solid Films 1989, 180, 93-99.
(41) Herron, J. N.; Mu¨ller, W.; Paudler, M.; Riegler, H.; Ringsdorf, H.;
Suci, P. A. Langmuir 1992, 8, 1413-1416.
(42) Samuelson, L. A.; Miller, P.; Galotti, D. M.; Marx, K. A.; Kumar,
J.; Tripathy, S. K.; Kaplan, D. L. Langmuir 1992, 8, 604-608.
(39) Vogel, V.; Schief Jr., W. R.; Frey, W. Supramol. Sci., in press.

A Metal-Chelating Lipid for 2D Protein Crystallization
J. Am. Chem. Soc., Vol. 119, No. 10, 1997 2485
129 mmol) was added, and the resultant mixture was vigorously stirred
at room temperature overnight. CH2Cl2 (70 mL) and water (100 mL)
were then added to the flask. The reaction mixture was transferred to
a separatory funnel, the solution was shaken, the layers were separated,
and the organic layer was collected. The aqueous layer was then further
extracted with CH2Cl2 (100 mL). The combined organic layers were
washed sequentially with 10% NaHCO3 (70 mL × 2) and brine (100
mL), and dried over MgSO4. Solvent was then removed on a rotovap
to yield a yellow viscous oil, which was used directly in the next step
without further purification.
eliminates the strong affinity exhibited by apostreptavidin for
copper-IDA in metal affinity chromatography.29
These effects may reflect biotin-induced structural changes
in streptavidin.32 Alternatively, biotin binding may disrupt the
IDA-Cu-grown crystals by disturbing protein-protein contacts.
In fact, binding of biotin to two streptavidin binding pockets
on biotin-lipid monolayers affects the protein-protein contacts,
resulting in asymmetric crystal growth.38 However, streptavidin
crystallizes on biotin-lipid monolayers, and the two empty
biotin sites of these crystals can be loaded with biotin or biotin
derivatives in limited quantities without affecting those 2D
crystals.40 Thus, biotin-induced structural changes do not affect
the ability of the protein to crystallize in 2D. The effects of
biotin on metal ion binding are clearly not fully understood at
this time.
The oil obtained from the previous step and Et3N (7.4 mL, 53.1
mmol) were dissolved in CH2Cl2 (60 mL) and were cooled to 0 °C. To
this solution was slowly added MsCl (4.1 mL, 53.1 mmol). The
reaction mixture was then allowed to warm to room temperature and
stirred vigorously overnight. Water (50 mL) and CH2Cl2 (30 mL) were
added to the solution, and the mixture was transferred to a separatory
funnel. The mixture was shaken, the layers were separated, and the
organic layer was collected. The aqueous layer was further extracted
with CH2Cl2 (50 mL × 2). The combined organic layers were then
washed with 10% NaHCO3 (50 mL) and brine (50 mL), and dried over
MgSO4. Solvent was then removed on a rotovap to yield a yellow oil.
The oily residue was chromatographed, eluting with EtOAc/hexanes
(50:50, Rf ) 0.30) to give 13.2 g of a white solid (42% yield): mp
Conclusions
Metal ion coordination by surface-accessible histidines, either
genetically engineered or naturally occurring, can provide a
general method for targeting of proteins to metal-chelating lipid
membranes. The ability to immobilize protein on a mobile lipid
monolayer is the first step toward 2D protein crystallization. In
the present study, coordination of two histidines to DOIDA-
Cu monolayers leads to 2D crystallization of streptavidin. The
metal-chelating DOIDA-Cu lipid provides the necessary target-
ing and mobility to promote 2D protein crystallization. Thus,
this lipid has the potential to greatly accelerate the investigation
of proteins for 2D crystallization and should facilitate the
exploration of organized protein arrays as functional materials.
1
89.0-90.5 °C; H NMR (CDCl3) δ 2.94 (s, 3H), 3.23 (t, J ) 4.4 Hz,
2H), 3.66 (m, 6H), 3.78 (t, J ) 4.4 Hz, 2H), 4.36 (t, J ) 4.4 Hz, 2H),
7.27 (m, 9H), 7.45 (m, 6H); 13C{1H} NMR (CDCl3) δ 37.59, 63.26,
69.07, 69.25, 70.69, 70.73, 86.54, 126.97, 127.77, 128.66, 143.99; IR
(neat) υ 3015.2, 2932.4, 2870.0, 1595.9, 1489.6, 1448.0, 1351.8, 1205.3,
1174.5, 1148.2, 1079.5, 1012.4, 970.5, 917.9, 798.3, 775.2, 763.1, 748.4,
707.5, 650.1, 632.3, 528.4 cm-1. Anal. Calcd for C26H30O6S: C, 66.36;
H, 6.43. Found: C, 66.13; H, 6.22.
1-(Triphenylcarbinyl)glycerol (4). Glycerol (20.0 g, 217 mmol),
trityl chloride (15.0 g, 53.8 mmol), and DMAP (150 mg, 1.23 mmol)
were placed in a 250 mL round-bottom flask containing 40 mL of dry
THF. To this heterogeneous mixture was added Et3N (9 mL, 64.63
mmol), and the resultant mixture was vigorously stirred at room
temperature overnight. EtOAc (70 mL) and water (50 mL) were then
added to the flask. The reaction mixture was transferred to a separatory
funnel, the solution was shaken, the layers were separated, and the
organic layer was collected. The aqueous layer was then further
extracted with EtOAc (50 mL × 2). The combined organic layers were
washed sequentially with 10% NaHCO3 and brine, and dried over Na2-
SO4. Solvent was then removed on a rotovap to yield a yellow oil
which was recrystallized with a benzene/hexanes mixture to give 13.6
g of a white solid (76% yield): mp 103-107 °C; 1H NMR (CDCl3) δ
1.97 (t, J ) 3.7 Hz, 1H), 2.49 (d, J ) 3.9 Hz, 1H), 3.24 (m, 2Η), 3.63
(m, 2Η), 3.87 (m, 1Η), 7.0-7.5 (m, 15Η); 13C{1H} NMR (CDCl3) δ
64.45, 70.18, 71.04, 86.59, 126.98, 127.78, 128.64, 143.80; IR (neat)
υ 3020.2, 2933.0, 1498.8, 1477.8, 1203.2, 1148.8, 764.5, 706.0, 632.0,
494.7 cm-1. Anal. Calcd for C22H22O3: C, 79.00; H, 6.64. Found:
C, 78.88; H, 6.52.
Experimental Section
General Methods. All reactions were performed in oven-dried (160
°C) glassware under positive Ar atmosphere. 1H NMR spectra were
recorded on a QE 300 spectrometer and referenced to tetramethylsilane
(TMS) at 0.00 ppm in chloroform-d (CDCl3). 13C NMR spectra were
recorded on a QE 300 spectrometer operating at 75 MHz and referenced
to 77.0 ppm in chloroform-d. Infrared spectra were recorded on a
Perkin-Elmer 1600 Series FTIR spectrometer. Mass spectra were
obtained from the Center of Biomedical and Bio-organic Mass
Spectrometry of Washington University, and from the University of
California at Riverside mass spectrometry facility. Chemical analyses
were performed by Desert Analytics (Tucson, AZ). Melting points
were determined on a Laboratory Devices Mel-Temp II. All flash
column chromatography was performed with Merck grade 60 silica
gel. Reactions were monitored as a function of time by TLC using J.
T. Baker plastic silica gel plates (JT7024-2) with a 40 µm particle size
and adsorbent thickness of 250 µm, purchased from VWR Scientific
(Chicago).
1-[(Methylsulfonyl)oxy]-(Z)-octadec-9-ene (6). Oleyl alcohol 5
(85% purity, 111 g, 0.351 mol) and Et3N (73.5 mL, 0.527 mol) were
dissolved in CH2Cl2 (1.2 L) and cooled to 0 °C. To this solution was
slowly added MsCl (32.6 mL, 0.422 mol). A white precipitate was
observed during the addition. The reaction mixture was then slowly
allowed to warm to room temperature and stirred vigorously overnight.
Water (500 mL) was added to the solution, and the mixture was
transferred to a separatory funnel. The mixture was shaken, the layers
were separated, and the organic layer was collected. The aqueous layer
was further extracted with CH2Cl2 (500 mL × 2). The combined
organic layers were then washed with 1 N HCl (500 mL), 10% NaHCO3
(500 mL) and brine, and dried over Na2SO4. The solution was
concentrated in Vacuo to give a brown oil, which was dissolved in 1 L
of mixed solvent of 80% ethyl acetate and 20% hexane. The solution
was stored at -5 °C overnight. The solid formed was filtered off.
Concentration of the filtrate yielded 105 g (86%) of 1-[(methylsulfonyl)-
oxy]-(Z)-9-octadecene as a colorless oil: 1H NMR (CDCl3) δ 0.88 (t,
J ) 6 Hz, 3H), 1.27 (br s, 12H), 1.30 (br s, 10H), 1.75 (d, J ) 7 Hz,
2H), 2.01 (d, J ) 3 Hz, 4H), 3.01 (s, 3H), 4.22 (t, J ) 6.6 Hz, 2H),
5.33-5.37 (m, 2H); 13C {1H} NMR (CDCl3) δ 13.96, 22.53, 25.26,
27.04, 28.87, 28.97, 29.18, 29.38, 29.87, 31.76, 37.074, 70.08, 129.53,
Materials. Chemical reagents and starting materials were purchased
from Aldrich Chemical Co. unless otherwise noted. Tetrahydrofuran
was distilled from sodium-benzophenone ketyl. Dichloromethane was
distilled from CaH2. DMSO was dried over 3 Å molecular sieves.
Methanesulfonyl chloride, triethylamine (Et3N), and diethyliminodi-
acetate (Kodak) were dried and distilled prior to use. Fluorescein biotin
[5-[N-[5-[N-[6-(biotinoylamino)hexanoyl]amino]pentyl]thioureidyl]-
fluorescein], 5-(and 6)-carboxytetramethylrhodamine succinimidyl ester
(RSE), and succinimidyl 6-biotinamidohexanoate (biotin-SE) were
obtained from Molecular Probes, Inc. (Eugene, OR). Native strepta-
vidin was obtained from Boehringer-Mannheim as a lyophilized powder.
Chloroform, methanol, and ethanol (all ACS HPLC grade), hexane
(99+%), and CuCl2 (99.999%) were obtained from Aldrich (Milwaukee,
WI). 3-(N-Morpholino)propanesulfonic acid (MOPS) was obtained
from Sigma (St. Louis, MO). All other reagents were used as received
without further purification.
1-(Methylsulfonyl)-10-(triphenylcarbinyl)-1,4,7,10-tetraoxade-
cane (2). Triethylene glycol (1) (10.0 g, 66.6 mmol), trityl chloride
(14.8 g, 53.1 mmol), and DMAP (320 mg, 2.62 mmol) were dissolved
in CH2Cl2 (100 mL) in a 250 mL round-bottom flask. Et3N (18.0 mL,

2486 J. Am. Chem. Soc., Vol. 119, No. 10, 1997
Pack et al.
129.80; IR (neat) υ 3000, 2922, 2853, 1465, 1356, 1177, 833, 724
cm-1; LRFAB (EI+) m/z (relative intensity) 364 [(M + NH4)+, 100],
317 (40), 243 (73); HRFAB calcd for C19H42NO3S [(M + NH4)+] m/z
364.2885, found 364.2877.
cm-1; HRMS (FAB) calcd for C45H89O6 [(M + H)+] m/z 725.6659,
found 725.6641.
9-[2,3-Bis[(Z)-octadec-9-enyloxy]propyl]-1-bromo-3,6,9-triox-
anonane (10). Compound 9 (1.64 g, 2.17 mmol) was dissolved in
THF (15 mL) and cooled to 0 °C. To the solution were added CBr4
(1.44 g, 4.33 mmol) and PPh3 (1.14 g, 4.33 mmol) as solids. The
resultant mixture was stirred at 5 °C for 1 h and then stirred at room
temperature overnight. The solution turned yellow, and a white
precipitate was also observed within 1/2 h after the complete addition
of the reagents. The solid was removed by filtration and washed with
cold CH2Cl2. The filtrate was then concentrated in Vacuo to yield a
bright yellow oil. Flash column chromatography of the oil eluting with
EtOAc/hexanes (25:75, Rf ) 0.26) furnished 1.34 g (75% yield) of the
desired compound as a colorless oil: 1H NMR (CDCl3) δ 0.88 (t, J )
6 Hz, 6H), 1.28 (br s, 44H), 1.50-1.65 (m, 4H), 1.85-2.15 (m, 8H),
3.35-3.70 (m, 19 H), 3.81 (t, J ) 6.3 Hz, 2H), 5.25-5.45 (m, 4H);
13C{1H} NMR (CDCl3) δ 14.10, 22.66, 26.07, 27.18, 29.24, 29.29,
29.48, 29.62, 29.74, 30.25, 31.88, 32.58, 70.04, 70.55, 70.59, 70.69,
71.19, 71.53, 129.82, 129.89; IR (neat) υ 2922, 2853, 1738, 1412, 1352,
1201, 1134, 737, 532 cm-1; HRMS (FAB) calcd for C45H88O5Br [(M
+ H)+] m/z 787.5815, found 787.5796.
Diethyl 1,1′-[N-[9-[2,3-bis[(Z)-octadec-9-enyloxy]propyl]-3,6,9-
trioxanonyl]imino]diacetate (11). Compound 10 (1g, 1.22 mmol),
diethyl iminodiacetate (DEIDA) (0.916 g, 9.64 mmol) and Et3N (0.675
mL, 9.64 mmol), were dissolved in dry CH3CN (16 mL)/THF (3 mL),
and the reaction mixture was brought to reflux and maintained at this
temperature for three days. The reaction was then cooled to room
temperature. Excess solvent was removed on a rotovap to yield an
oil, which was dissolved in EtOAc (100 mL). The solution was then
washed with 1 N HCl (50 mL), H2O (50 mL), and brine (50 mL) and
dried over Na2SO4. The solvent was removed on a rotovap. The yellow
oil residue was subjected to flash column chromatography eluting with
Et2O/hexane (80:20, Rf ) 0.30) to give 225 mg of the desired product
(21% yield): 1H NMR (CDCl3) δ 0.88 (t, J ) 6 Hz, 6H), 1.27 (br s,
50H), 1.45-1.60 (m, 4H), 1.95-2.10 (m, 8H), 2.97 (t, J ) 7.2 Hz,
2H), 3.35-3.75 (m, 23 H), 4.16 (q, J ) 7.2 Hz, 4H), 5.25-5.45 (m,
4H); 13C{1H} NMR (CDCl3) δ 14.04, 14.17, 22.61, 26.02, 27.13, 29.20,
29.24, 29.43, 29.56, 29.69, 31.83, 53.60, 55.64, 60.45, 69.99, 70.03,
70.28, 70.44, 70.56, 71.47, 129.75, 129.83, 170.98; IR (neat) υ 3849,
3745, 2923, 2853, 1738, 1208, 1153, 501 cm-1; HRMS (FAB) calcd
for C53H102NO9 [(M + H)+] m/z 896.7554, found 896.7556.
1,1′-[[9-[2,3-Bis[(Z)-octadec-9-enyloxy]propyl]-3,6,9-trioxanonyl]-
imino]diacetic Acid (12, DOIDA). Compound 11 (452 mg, 0.488
mmol) and powdered NaOH (195 mg, 4.88 mmol) were placed in a
round-bottom flask charged with a reflux condenser. THF (7 mL),
MeOH (7 mL), and H2O (2 mL) were added to the flask, and the
mixture was then brought to reflux for 1 h. The reaction was cooled
to room temperature and acidified to pH ∼1 with concentrated HCl.
The solvent was removed in Vacuo. Flash column chromatography of
the yellow oil residue eluting with CHCl3/MeOH/H2O (70:25:5, Rf )
0.3) gave 168 mg (0.2 mmol, 41% yield) of the desired product: 1H
NMR (CDCl3) δ 0.88 (t, J ) 6 Hz, 6H), 1.28 (br s, 44H), 1.50-1.65
(m, 4H), 1.90-2.10 (m, 8H), 3.25-4.10 (m, 25 H), 5.25-5.45 (m,
4H), 9.48 (br s, 2H); 13C{1H} NMR (CDCl3) δ 14.16, 22.72, 26.19,
27.27, 29.09, 29.37, 29.58, 29.81, 30.03, 30.20, 31.95, 32.67, 53.74,
70.42, 70.64, 70.78, 70.91, 71.01, 71.08, 71.21, 71.45, 71.55, 71.72,
76.83, 77.26, 77.30, 77.81, 77.93, 129.86, 129.83, 130.30, 130.38,
177.80; IR (neat) υ 2970, 2361, 2335, 1738, 1366, 1216, 1207, 1150,
772, 639, 510 cm-1; LRFAB (EI+) m/z (relative intensity) 878.5 [(M
+ K)+, 15], 448.2 (10.0), 386.2 (100); HRMS (FAB) calcd for C49H94-
NO9 [(M + H)+] m/z 840.6928, found 840.6950.
1-[(Triphenylcarbinyl)oxy]-2,3-bis[(Z)-octadec-9-enyloxy)pro-
pane (7). Tritylglycerol 4 (8.00 g, 23.09 mmol), dissolved in THF
(40 mL), was added to a suspension of powdered KOH (3.3 g, 58.93
mmol) in dry DMSO (100 mL). The mixture was heated to 80 °C and
maintained at this temperature for 4 h. Oleyl mesylate 6 (19.2 g, 55.42
mmol) dissolved in 30 mL of THF was added to the reaction mixture
at 80 °C Via a syringe. The mixture became gel-like at first and was
heated at 80 °C for 36 h. The reaction was cooled to room temperature,
and then EtOAc (150 mL) and H2O (150 mL) were added to the
mixture, the mixture was shaken, and the layers were separated. The
aqueous layer was extracted with EtOAc (100 mL × 2). The combined
organics were washed with water (150 mL) and saturated NaCl (150
mL), and dried over Na2SO4. The solution was concentrated in Vacuo,
and the oily residue was chromatographed, eluting with Et2O/hexanes
(5:95, Rf ) 0.30) to obtain 6.6 g of a colorless oil (33% yield): 1H
NMR (CDCl3) δ 0.88 (t, J ) 6 Hz, 6H), 1.27 (br s, 44H), 1.50-1.65
(m, 4H), 1.90-2.05 (m, 8H), 3.17 (br s, 2H), 3.40 (t, J ) 6.6 Hz, 2H),
3.52-3.60 (m, 5H), 5.33-5.50 (m, 4H), 7.18-7.80 (m, 15H), 7.46 (d,
J ) 7.2 Hz); 13C{1H} NMR (CDCl3) δ 14.10, 21.01, 22.57, 22.68,
25.90, 27.20, 28.59, 28.84, 28.99, 29.21, 29.31, 29.40, 29.51, 29.73,
29.75, 31.90, 32.59, 64.65, 76.58, 129.77, 129.96, 130.20; IR (neat) υ
2926, 2854, 2361, 2249, 1723, 1448, 1377, 1219, 1091, 908, 734, 706,
649, 632 cm-1; LRFAB (EI+) m/z (relative intensity) 857.7 [(M + Na)+,
40], 641.5 (25.0), 621.6 (100), 591.5 (5.0); HRFAB calcd for C58H90O3-
Na [(M + Na)+] m/z 857.6787, found 857.6769.
2,3-Bis[(Z)-octadec-9-enyloxy]propanol (8). Compound 7 (6.50
g, 7.50 mmol) and TsOH‚H2O (220 mg, 2.4 mmol) were placed in a
mixture of THF (40 mL)/MeOH (40 mL) and stirred overnight at room
temperature. Et3N (0.3 mL, 2.3 mmol) was added to consume excess
TsOH‚H2O, and the reaction solution was concentrated on a rotovap
to yield an oil which was purified by flash chromatography, eluting
with 75% CH2Cl2/hexane (75:25, Rf ) 0.30) to give 2.96 g of the desired
product (67% yield): 1H NMR (CDCl3) δ 0.86 (t, J ) 6 Hz, 6H), 1.25
(br s, 24H), 1.27 (br s, 20H), 1.42-1.75 (m, 4H), 1.85-2.25 (m, 9H),
3.40-3.75 (m, 9H), 5.20-5.40 (m, 4H); 13C{1H} NMR (CDCl3) δ
14.11, 22.68, 26.09, 27.20, 29.26, 29.32, 29.51, 29.60, 29.76, 30.06,
31.90, 63.05, 70.37, 70.86, 71.82, 78.23, 129.80, 129.93; IR (neat) υ
3300, 3030, 2920, 2854, 1735, 1460, 1380, 1205, 1160, 1070, 720,
450 cm-1; HRFAB calcd for C39H77O3 [(M + H)+] m/z 593.5872, found
593.5871.
9-[2,3-Bis[(Z)-octadecen-9-yloxy]propyl]-3,6,9-trioxanonanol (9).
To a suspension of NaH (60% mineral oil dispersion, 1.68 g, 42.1
mmol) in 60 mL of THF was added compound 8 (5.00 g, 8.44 mmol)
dissolved in 60 mL of THF. The resultant suspension was stirred for
2 h at room temperature. 1-(Methylsulfonyl)-10-(triphenylcarbinyl)-
1,4,7,10-tetraoxadecane (2) (5.95 g, 12.65 mmol) was added to the
suspension as a solid, and the reaction mixture was brought to reflux
overnight. The reaction mixture was then cooled to room temperature,
and water was added. The heterogeneous solution was stirred for a
few minutes before being transferred to a separatory funnel. EtOAc
(200 mL) was added, the mixture was shaken, the layers were separated,
and the organic layer was collected. The aqueous layer was further
extracted with EtOAc (200 mL × 2). The combined organic layers
were washed with brine and dried over Na2SO4. The filtrate was
concentrated on a rotovap to yield a blond residue, which was taken
up in THF (50 mL)/MeOH (50 mL). TsOH‚H2O (400 mg, 43.5 mmol)
was then added to the reaction mixture, which was stirred overnight at
room temperature. Et3N (0.40 mL) was added to consume excess
TsOH‚H2O. The solution was then concentrated in Vacuo to yield a
yellow residue. Flash column chromatography of the residue eluting
with EtOAc/hexanes (75:25, Rf ) 0.28) furnished 5.50 g (90% yield)
of the desired compound as a colorless oil: 1H NMR (CDCl3) δ 0.88
(t, J ) 6 Hz, 6H), 1.28 (br s, 44H), 1.50 (br s, 4H), 1.90-2.20 (m,
9H), 3.46-3.75 (m, 21 H), 5.30-5.40 (m, 4H); 13C {1H} NMR (CDCl3)
δ 14.03, 22.60, 25.99, 27.12, 29.18, 29.24, 29.43, 29.50, 29.68, 31.83,
32.52, 61.60, 69.94, 70.26, 70.52, 71.48, 72.51, 129.73, 129.81; IR
(neat) υ 3400, 2924, 2854, 1737, 1412, 1349, 1196, 1127, 740, 501
Streptavidin Crystallization. Lipid solutions were prepared in
chloroform. The IDA lipids were premetalated by adding an aliquot
of concentrated CuCl2 in methanol to the lipid solutions. Excess Cu2+
in the lipid solution is no more than 10%. The monolayer subphase
was buffered with 20 mM MOPS and 250 mM NaCl, pH 7.8. The
buffer, prepared with Milli-Q water (Milli-Q UV Plus, Millipore,
Bedford, MA), was shaken with chloroform to remove organic
impurities and, subsequently, shaken with hexane to remove traces of
chloroform. Streptavidin concentrations were determined by UV-vis
absorption spectroscopy at 280 nm using an extinction coefficient of
136 000 M-1 cm-1
.

A Metal-Chelating Lipid for 2D Protein Crystallization
J. Am. Chem. Soc., Vol. 119, No. 10, 1997 2487
Crystallization was performed under two sets of conditions: constant
pressure and constant area. Constant pressure experiments were
performed in a computer-controlled Langmuir trough (KSV Instruments,
Helsinki) with a surface area of 75 × 230 mm2. The entire film balance
was placed inside a Plexiglas box, and the atmosphere was purged with
Ar. The surface pressure was measured by the Wilhelmy plate method
using plates cut from filter paper (Whatman No. 1). The plate was
rinsed thoroughly with chloroform and calibrated with stearic acid prior
to use. Monolayers were compressed at a constant rate of 3 Å2/
(molecule‚min). A 15 min waiting period prior to compression allowed
for solvent evaporation. Once the desired surface pressure (3 mN/m)
was attained, the instrument was switched to constant pressure mode,
and the monolayer was allowed to equilibrate for 15-30 min. The
desired amount of protein, dissolved in the above buffer, was injected
into the subphase either behind the barrier or through the monolayer
and allowed to mix by diffusion. Changes in surface area were recorded
as a function of time. After each experiment, the trough was thoroughly
cleaned with detergent (RBS 35), Milli-Q purified water, and chloro-
form. All measurements were performed at room temperature, 20-
27 °C.
Constant area experiments were performed in a Petri dish hydro-
phobized with octadecyltrichlorosilane. Holes cut in the lid of the Petri
dish allowed access of the Wilhelmy plate and entrance of Ar. Surface
pressure was again measured using Wilhelmy plates cut from filter
paper. Monolayers were spread dropwise until the desired surface
pressure (3 mN/m) was attained. Time was allowed between drops
for stabilization of the surface pressure. Thirty to forty-five minutes
after spreading, a concentrated streptavidin solution was injected into
the 15 mL subphase through the monolayer. Surface pressure was
measured as a function of time.
dry DMF, was added to the reaction mixture at a biotin/lysozyme ratio
of 8:1. The ability of the biotin-tagged lysozyme to bind streptavidin
was confirmed qualitatively. An aliquot of iminobiotin-derivatized
agarose (Immunopure Immobilized Iminobiotin, Pierce, Rockford, IL)
was incubated with fluorescein-labeled streptavidin, and the gel was
pelleted by centrifugation. The yellow color of the fluorescein was
localized in the gel, indicating the protein had been bound by the
iminobiotin moieties. The supernatant was drawn off, and a solution
of rhodamine/biotin-labeled lysozyme was added to the gel. The protein
and gel were mixed briefly, and the gel was pelleted by centrifugation.
Again, most of the red color of the rhodamine was localized in the
gel, indicating the lysozyme was bound to the free biotin pockets of
the previously bound streptavidin.
Fluorescence Microscopy. Fluorescence microscopy was performed
essentially as described.21 For polarized fluorescence studies, plastic
polarizing sheets (Edmund Scientific C43,781) were placed in the
excitation path just before the excitation filter. The excitation polarizers,
oriented 90° with respect to each other, were placed in a slider such
that the polarizers could be interchanged rapidly. The images were
obtained Via a variable-gain image intensifier (Hamamatsu C2400-68)
with a solid-state CCD camera (Hamamatsu C2400-60) and were
recorded on a video cassette recorder or digitized directly using a frame
grabber (AG-5, Scion Corp., Frederick, MD) on a Macintosh compatible
computer (Power 100, Power Computing, Austin, TX). (The AG-5
frame grabber allows multiple frames to be averaged, improving image
quality of low-contrast, low-light samples. The images presented are
most commonly averages of five consecutive frames.) Images were
manipulated using the program NIH Image (developed at the U.S.
National Institutes of Health and available on the Internet at http://
rsb.info.nih.gov/nih-image/).
Labeling of Proteins. Streptavidin was labeled with 5-(and 6)-
carboxytetramethylrhodamine succinimidyl ester (RSE) following the
standard procedure43 with a slight modification as described.36 Briefly,
10 mM RSE in dry DMF was added dropwise, with stirring, to a
solution of 20-100 µM streptavidin (in 0.1 M NaHCO3 and 0.1 M
NaCl, pH 8.3) to yield a dye/protein ratio of 2:1. The reaction mixture
was incubated at room temperature for 1 h. Labeled protein was
separated from unreacted dye by gel filtration chromatography on
prepacked Sephadex G-25 columns (PD-10, Pharmacia, Uppsala,
Sweden) preequilibrated with 20 mM MOPS and 250 mM NaCl, pH
7.8.
Acknowledgment. We thank Dr. Viola Vogel, Dr. Wolfgang
Frey, and William R. Schief, Jr., University of Washington,
Seattle, for helpful discussions. In addition, we thank Wei-Hau
Chang and Dr. Roger D. Kornberg, at Stanford University, for
performing the electron diffraction analysis of the 2D strepta-
vidin crystals. This work was supported by the Army Research
Office. D.W.P. acknowledges a Landau fellowship and a
training fellowship from the National Institute of General
Medical Sciences, NRSA Award 1 T32 GM 08346-01. G.C.
and K.M.M. are supported by postdoctoral fellowships from
the National Institute of General Medical Sciences, NRSA
Awards 5 F32 GM17448-02 and 5 F32 GM16953-02,
respectively.
Lysozyme was simultaneously labeled with rhodamine and tagged
with biotin. The procedure was essentially the same as that described
above for labeling of streptavidin with rhodamine succinimidyl ester.
However, in addition to the RSE, a biotin-SE derivative, dissolved in
(43) Brinkley, M. Bioconjugate Chem. 1992, 3, 2-13.
JA964099E