An evaluation of detergents for NMR structural studies of membrane proteins.

Abstract

Structural information on membrane proteins lags far behind that on soluble proteins, in large part due to difficulties producing homogeneous, stable, structurally relevant samples in a membrane-like environment. In this study 25 membrane mimetics were screened using 2D (1)H-(15)N heteronuclear single quantum correlation NMR experiments to establish sample homogeneity and predict fitness for structure determination. A single detergent, 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)] (LPPG), yielded high quality NMR spectra with sample lifetimes greater than one month for the five proteins tested - R. sphaeroides LH1 alpha and beta subunits, E. coli and B. pseudofirmus OF4 ATP synthase c subunits, and S. aureus small multidrug resistance transporter - with 1, 2, or 4 membrane spanning alpha-helices, respectively. Site-specific spin labeling established interhelical distances in the drug transporter and genetically fused dimers of c subunits in LPPG consistent with in vivo distances. Optical spectroscopy showed that LH1 beta subunits form native-like complexes with bacteriochlorophyll a in LPPG. All the protein/micelle complexes were estimated to exceed 100 kDaltons by translational diffusion measurements. However, analysis of (15)N transverse, longitudinal and (15)N[(1)H] nuclear Overhauser effect relaxation measurements yielded overall rotational correlation times of 8 to 12 nsec, similar to a 15-20 kDalton protein tumbling isotropically in solution, and consistent with the high quality NMR data observed.

Full text

Page 1

Journal of Biomolecular NMR 17: 43–57, 2004.
KLUWER/ESCOM
© 2004 Kluwer Academic Publishers. Printed in the Netherlands.
43
An evaluation of detergents for NMR structural studies of membrane
proteins
Ray D. Krueger-Koplina, Paul L. Sorgena, Suzanne T. Krueger-Koplina, Iv´ an O. Rivera-Torresa,
Sean M. Cahilla, David B. Hicksb, Leo Griniusc, Terry A. Krulwichb& Mark E. Girvina,∗
aBiochemistry Department, Albert Einstein College of Medicine, 1300 Morris Park Ave, Bronx, NY 10461, U.S.A.
bDepartment of Pharmacology and Biological Chemistry, Mt. Sinai School of Medicine, One Gustave L. Levy
Place, Box 1603, New York, NY 10029, U.S.A.
cCincinnati State Technical College, 3520 Central Parkway, Cincinnati, Ohio 45223, U.S.A.
Received 20 May 2003; Accepted 5 August 2003
Key words: detergent, lysolipids, membrane, protein, structure
Abstract
Structural information on membrane proteins lags far behind that on soluble proteins, in large part due to dif-
ficulties producing homogeneous, stable, structurally relevant samples in a membrane-like environment. In this
study 25 membrane mimetics were screened using 2D1H-15N heteronuclear single quantum correlation NMR
experiments to establish sample homogeneity and predict fitness for structure determination. A single detergent,
1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)](LPPG), yieldedhighqualityNMR spectra with
sample lifetimes greater than one month for the five proteins tested – R. sphaeroides LH1 α and β subunits, E. coli
and B. pseudofirmus OF4 ATP synthase c subunits, and S. aureus small multidrug resistance transporter – with
1, 2, or 4 membrane spanning α-helices, respectively. Site-specific spin labeling established interhelical distances
in the drug transporter and genetically fused dimers of c subunits in LPPG consistent with in vivo distances.
Optical spectroscopy showed that LH1 β subunits form native-like complexes with bacteriochlorphyll a in LPPG.
All the protein/micellecomplexeswere estimated to exceed 100 kDaltons by translationaldiffusionmeasurements.
However,analysisof15N transverse,longitudinaland15N{1H} nuclearOverhausereffectrelaxationmeasurements
yieldedoverallrotationalcorrelationtimesof8to12nsec, similartoa15–20kDaltonproteintumblingisotropically
in solution, and consistent with the high quality NMR data observed.
Abbreviations: APPC – 1-palmitoyl-2-acetoyl-sn-glycero-3-phosphocholine; βOG – n-octyl-β-D-glucopyranoside; CFTR cystic fibrosis
transmembrane conductance regulator; Chaps – 3-([3-cholamidopropyl] dimethylammonio)-1-propanesulfonate; Cholic acid – 3a,7a,12a-
trihydroxy-5-β-cholanic acid; DHPC – 1,2-dicaproyl-1-sn-glycero-3-phosphocholine; DHPG – 1,2-dicaproyl-1-sn-glycero-3-[phospho-RAC-
(1-glycerol)]; DLCP– dilysocardiolipin; DM–n-dodecyl-β-D-maltoside; DPC– dodecylphosphocholine; LDAO –N,N-dimethyldodecylamine
N-oxide; LH1 – light-harvesting complex I; LMPC – 1-myristoyl-2-hydroxy-sn-glycero-3-phosphocholine; LMPE – 1-myristoyl-2-hydroxy-
sn-glycero-3-phosphoethanolamine; LMPG – 1-myristoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)]; LOPC – 1-oleoyl-2-hydroxy-
sn-glycero-3- phosphocholine; LOPG – 1-oleoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)]; LPPC – 1-palmitoyl-2-hydroxy-sn-
glycero-3-phosphocholine; LPPE – 1-palmitoyl -2-hydroxy-sn-glycero-3-phosphoethanolamine; LPPG – 1-palmitoyl-2-hydroxy-sn-glycero-3-
[phospho-RAC-(1-glycerol)]; MLPA – sn-(3-myristoyl-2-hydroxy)-Glycerol-1-phospho-sn-3?-(1?-myristoyl-2?-hydroxy)-glycerol; pMAL-B –
[Poly (maleic anhydride-alt-1-tetradecene) substituted with 3-(Dimethylamino) propylamine]; 3-maleimido-PROXYL – 3-(maleimidomethyl)-
2,2,5,5-tetramethyl-1-pyrrolidinyloxy – free radical; RAC – racemic; SDS – Sodium-n-dodecylsulfate; Smr – small multi-drug resistance
protein from S. aureus; TMH – transmembrane helix; TRITON – polyoxyethylene(10) isooctylphenyl ether; TSP – [3-trimethylsilyl)propionic
acid]; Zwittergent 3-10 – Decyl-N-sulfobetaine; Zwittergent 3-12 – Dodecyl-N-sulfobetaine; Zwittergent 3-14 – Tetradecyl-N-sulfobetaine;
Zwittergent 3-16 – Hexadecyl-N-sulfobetaine.
Introduction
Cells use membranes to establish biologically dis-
tinct compartments, and rely on membrane-embedded
∗To
girvin@aecom.yu.edu
whomcorrespondence shouldbeaddressed;E-mail:
proteins to carry out transport, signaling, energytrans-
duction and other cellular functions across these barri-
ers. Roughly one-third of all structural genes code for
membraneproteins(Fraseret al., 1995).However,less
than one out of 500 high-resolution structures is from
this class of proteins. The main high-resolution struc-

Page 2

44
tural techniques, X-ray crystallography and solution
NMR, require removing the protein from the mem-
brane. Many of the difficulties in obtaining suitable
samples of membrane proteins for structural studies
stem from the need to satisfy interactions typically
provided by their multi-layered, quasi-planar envir-
onment. Structural studies require conditions where
the protein is soluble at high concentrations, mon-
odisperse, conformationally homogeneous, and stable
for days to weeks.
From a solution NMR perspective, a protein as-
sociated with a membrane mimetic tumbles as part
of a larger complex, which leads to signal broaden-
ing, poor sensitivity, and reduced spectral resolution.
This is especially problematic for membrane pro-
teins that often have low spectral dispersion due to
the preponderance of similar amino acid types and
helical secondary structures located in non-polar en-
vironments. However, recent advances, such as higher
magneticfieldsandTROSY methods(Pervushinet al.,
1997), have extended the size of systems accessible
to solution NMR towards 100 kDa (Wüthrich et al.,
1998). And, despite difficulties in over-expressionand
purification, many membrane proteins are available in
sufficient quantities for NMR studies. Still, few struc-
tures have actually been solved, in part due to the lack
of optimal membrane mimetics.
Detergent micelles are the smallest membrane
mimics that provide all of the interactions that in-
tegral membrane proteins require, prearranged in the
appropriate order. Mixed organic solvents and amphi-
pols likely organize in response to the characteristics
of the protein. However, detergents spontaneously
assemble intoahydrophobiccoreandaninterfacialre-
gion providing the membrane-like ability to induce or
accommodate structure (White et al., 2001). Bicelles
are an appealing membrane mimic (Sanders et al.,
1998), improving on micelles by providing a planar
interface between the hydrophobic and hydrophilic
domains. However,since bicelles are muchlarger than
micelles, detergents are the most promising candid-
ates for a general membrane mimetic for conventional
NMR solution structure determination.
Here we used NMR spectroscopy to evaluate de-
tergent micelle conditions that support concentrated,
stable, structurally homogeneous membrane protein
samples. The LH1 β subunit (5.5 kDa, 1 TMH)
of the R. sphaeroides photosynthetic bacterial light
harvesting 1 (LH1) complex had been tested previ-
ously with a subset of the detergents presented here,
and Zwittergent 3-12 yielded NMR spectra of suf-
ficient quality to determine the structure of the β
subunit. (Sorgen et al., 2002). However, the zwit-
tergents were poor mimetics for the LH1 α subunit
(6.8 kDa, 1 TMH), subunit c monomers of the E. coli
(8.3 kDa, 2 TMH) or B. pseudofirmus OF4 (7.0 kDa,
2 TMH) ATP synthase, and the small multidrug res-
istance (Smr) transporter from S. aureus (11.7 kDa,
4 TMH). In the work presented here, a number of
detergents were screened for their ability to sup-
port high quality NMR spectra of E. coli subunit c,
the β subunit, and Smr. A detergent derivative of
a standard lipid, 1-palmitoyl-2-hydroxy-sn-glycero-
3-[phospho-RAC-(1-glycerol)] (LPPG), stood out as
a superior membrane mimetic. Diffusion and15N
relaxation measurements revealed proteins tumbling
rapidly within the micelles, allowing higher spectral
quality than expected for these large complexes. Two
othermembraneproteins,LH1αsubunitandOF4sub-
unit c, made poor NMR samples in other membrane
mimetics but gave high quality 2D1H-15N HSQC
spectra in LPPG.
Materials and methods
Detergents
Lipid derivatives were purchased from Avanti Polar
Lipids (Alabaster, AL). Poorer protein reconstitution
was observed for about one-quarter of the LPPG
lot numbers used, but no significant lot variations
were observed for other detergents. pMAL-B was
purchased from Anatrace (Maumee, OH), and other
detergents from CalBiochem (San Diego, CA).
Subunit c of the E. coli F1FOATP synthase
Uniformly15N labeled wild type subunit c (uncE gene
product) and subunit c mutants were over-expressed
and purified as described previously (Girvin et al.,
1998), but using E. coli strain C43(DE3) (Miroux and
Walker, 1996) as the host. Typical yields from M63
minimal media were 4 mg/l. Subunit c and all other
purified proteins were characterized by SDS-PAGE,
NMR (2D1H-15N HSQC) and MALDI-TOF mass
spectrometry.
Dimers of E. coli subunit c
The expression vector for a fused subunit c dimer
was made by directionally cloning (NdeI, XhoI) a

Page 3

45
PCR product coding for uncE (G71W mutant) fol-
lowed by the linker sequence ANGSLNDG plus a
second uncE (G71C mutant) into the pET17b plas-
mid (Novagen). Subunit c dimer was over-expressed
from this plasmid in E. coli C43(DE3) in M63 min-
imal media with15NH4Cl as the sole nitrogen source.
Two hours after IPTG induction (0.7 mM), cells were
harvested by centrifugation, resuspended in 8 ml of
100 mM ammonium acetate per liter of culture, and
stored frozen at −20◦C. Subunit c dimer was extrac-
ted with 10 volumes of 1:1 CHCl3/CH3OH and the
cell debris removed by centrifugation. Adjusting the
extract to a ratio of 8:4:3 CHCl3/CH3OH/H2O caused
phase separation. The organic phase was removed and
concentrated to ∼4 ml, and the dimer was precipitated
with 5 volumes of diethylether at −20◦C. The precip-
itate was air dried then dissolved in a minimal volume
of 1:1 CHCl3/CH3OH and purified on a 2.5 × 100 cm
LH60 size exclusion column eluted isocratically with
1:1 CHCl3/CH3OH, 50 mM ammonium acetate. The
LH60 fractions were evaporated under a stream of ar-
gon to less than 10 ml and the subunit c dimer was
desalted and the solvent exchanged on a LH20 size
exclusion column (2.5 cm × 30 cm) eluted with 1:1
CHCl3/CH3OHandstoredat4◦C. Typicalyieldsfrom
minimal media were 3–5 mg/l.
Subunit c from B. pseudofirmus OF4
Uniformly
expressed in E. coli C43(DE3) (Miroux and Walker,
1996), and purified as described for subunit c dimer.
There was some overlap of the genomically exressed
E.colisubunitc(8.3kDa)andOF4subunitc(7.0kDa)
peaks eluted from the LH60 size exclusion column.
Calculated differences in optical absorbance (Pace
et al., 1995) of OF4 subunit c (0 Trp, 0 Tyr, and 3 Phe,
ε280∼ 0, ε257= 564 M−1cm−1), and E. coli subunit
c (0 Trp, 2 Tyr, and 4 Phe, ε280 = 2530 M−1cm−1,
ε257= 752M−1cm−1) were used to identifyunmixed
fractions. E. coli subunit c was confirmed to be be-
low the level of detection by 2D1H-15N HSQC NMR
experiments in mixed organic solvents before recon-
stituting the OF4 subunit c. Yields from M63 minimal
media were variable, occasionally reaching 3–4 mg/l.
15N labeled OF4 subunit c was over-
Small multidrug resistance protein (Smr) from
S. aureus
The S. aureus smr gene (Grinius et al., 1992) was dir-
ectionally cloned into pET17b. Over-expression and
purification was as described for subunit c dimer ex-
cept that the purified Smr was dried under a stream of
argon and stored at −20◦C. Typical yields from M63
minimal media were 3–5 mg/l.
LH1 α and β subunits from R. sphaeroides
For uniform isotopic labeling, the α and β subunits
were expressed in a genomic deletion strain of R.
sphaeroides SK102 as described previously (Sorgen
et al., 2002). Yields were typically 3–5 mg/l of each
subunit from R. sphaeroides in minimal media. Bac-
teriochlorophyll a (BChl a) purification and binding
assaysweredescribedpreviously(Sorgenetal., 2002).
Spin label-modified proteins
Single Cys and Trp substitutions were introduced us-
ing the Quick-Change mutagenesis protocol (Strata-
gene) or by PCR cassette mutagenesis. Unique Cys
residues were spin labeled after LH60 chromato-
graphy. The protein containing fractions were con-
centrated to near dryness in a rotovap at 37◦C then
diluted to approximately 0.1 mM in 4:4:1 chloro-
form/methanol/water, 20 mM TRIS pH 7.0 with a
10-fold molar excess of 3-maleimido-PROXYL, and
incubated for 60 min at 30◦C. Excess reagent and
buffer were removed, and the solvent exchanged on a
LH20 size exclusion column (2.5 cm × 30 cm) eluted
with 1:1 CHCl3/CH3OH.
Reconstitution in detergent micelles
Subunit c, subunit c dimer, OF4 subunit c, α subunit,
andSmrwereoptimallyreconstitutedbymixingdeter-
gent with the protein dissolved in 3:1 CHCl3/CH3OH
and drying to a thin, clear film (approximately 1 cm2
per mg of detergent present) under a stream of argon.
CHCl3and/or CH3OH were added a drop at a time
as necessary during drying to prevent precipitation.
The film was dried for 1–12 h more under argon to
remove residual CHCl3. Detergent/protein films were
dissolved in 1–50 mM aqueous potassium phosphate
buffer, pH 4.5–8 (as noted) containing 5% (v/v) D2O
and 1 mM TSP. When necessary, samples were heated
to 42◦C for up to 12 h and/or bath sonified for up
to 30 min to dissolve the film. Lyophilized β sub-
unit readily dissolved in buffered aqueous detergent
solutions.

Page 4

46
NMR spectroscopy
NMR data were acquired at 600 MHz. The sens-
itivity enhanced version of the 2D1H-15N HSQC
(Kay et al., 1992; Palmer et al., 1991) was used.1H
chemical shifts were referenced directly to TSP, and
15N was referenced indirectly (Wishart et al., 1995).
Translational diffusion coefficients were measured by
a bipolar pulse pair longitudinal-eddy-current delay
experiment (Wu et al., 1995). Protein/micelle sizes
were estimated from comparisons to protein stand-
ards, normalized for differences in viscosity using the
TSP standard. Spectra were processed and analyzed
using NMRPipe (Delaglio et al., 1995) and NMRView
(Johnson and Blevins, 1994).
15N relaxation experiments
The pulse sequences used for measuring15N T1and
T2relaxationtimesand15N{1H}NOEs(Skeltonetal.,
1993) included water suppression (Piotto et al., 1992;
Fushmanetal., 1997)andfortheNOE,waterflip-back
(Grzesiek and Bax, 1993).The T1and T2experiments
were collected with 4 K and 192 complex points in F2
(1H) and F1 (15N) respectively, with 32 scans per t1
point and a recycle delay of 1s. The NOE experiments
were collected with 2 K and 64 complex points in F2
(1H) and F1 (15N) respectively, with 96 scans per t1
point and a recycle delay of 6 s. All experiments used
a1H sweep width of 7800 Hz and a15N sweep width
of 2500 Hz with the1H and15N carriers set to 4.7 and
119 ppm, respectively. The T2relaxation delays were
8, 16(×2),24,32, 40,48, 64(×2),128,160,240(×2)
and 320 ms, with a 1 ms delay between15N pulses in
the Carr-Purcell-Meiboom-Gillsequence. The T1lon-
gitudinal relaxation delays were 8, 20 (×2), 40, 100
(×2), 160, 200, 300 (×2), 500, 800 ms, 2 and 4 s.
15N{1H}steady-state NOE values were obtained from
spectra with and without a 3 s GARP (Shaka et al.,
1985) proton decoupling period applied in the middle
of the amide spectral region during the recycle delay.
Relaxationdatawereanalyzedbythemodelfreeform-
alism (Lipari and Szabo, 1982a, b), using the program
Dasha (Orekhov et al., 1995).
Results
Detergent screen
Detergents were chosen to represent those used suc-
cessfully to extract and reconstitute membrane pro-
teins, those that have been used previously for NMR
and crystallographic studies, and those that are single
chain analogs of the most common phospholipids.
Three test proteins represented single TM helix (the
Rb. sphaeroides LH1 β subunit), two TM helix (E. coli
subunit c) and four TM helix (S. aureus Smr) classes.
For screening, the proteins were initially dissolved
at sub-millimolar concentrations (0.35 mM for β and
0.5 mM for c and Smr). Except where noted, the
samples remained clear and free of precipitation for
the duration of the studies. 2D1H-15N HSQC spec-
tra were used to evalute the sample properties in each
detergent and representative spectra are shown in Fig-
ures 1–3. The area under the amide proton region
in 1D1H spectra as a function of time was used as
an indicator of sample stability for the relatively suc-
cessful detergents. The percentage of expected cross
peaks initially observed, and the sample half-lives are
summarized in Table 1.
Figure 1 shows the 2D1H-15N HSQC spectra of
the three test proteins in the best detergent identified,
LPPG. Subunit c in a single conformation should give
rise to 84 cross peaks in this spectrum. As seen in Fig-
ure 1A, 96% of the expected cross-peaks for subunit c
were resolved in LPPG micelles, and the cross peaks
were uniform in line width and intensity. All of the
expected peaks were observed upon resolution into a
thirddimensionin13C-separatedexperiments.Asseen
in Figure 1B, LPPG also yielded excellent spectra for
the β subunit, where all 50 of the expectedcross-peaks
were observed. The Smr proton/drug antiporter from
S. aureus was perhaps the best test of LPPG as a de-
tergentforhigh-resolutionNMR studies. Smr contains
107 residues and four transmembrane helices (Grinius
and Goldberg,1994; Paulsen et al., 1995), and was the
largest monomer tested. 86% of the 113 cross-peaks
expected in a 2D1H-15N HSQC spectrum were ob-
served for Smr in LPPG (Figure 1C). Several of the
peaks had shapes and intensities consistent with two
or more overlapping signals.
As summarized in Table 1, and depicted for rep-
resentative detergents in Figure 2, many detergents
yielded poorer quality NMR data for these proteins.
The sparse cross peaks for subunit c and the β subunit
in APPC (Figures 2A,D) displayed a wide range of

Page 5

47
Table 1. 2D1H-15N HSQC spectral quality and lifetime for protein in detergent micelles
Detergent Subunit c
LH1 β subunit Smr
Peaks (%)a
T1/2(days)b
Peaks (%)T1/2(days)Peaks (%)T1/2(days)
APPC
βOG
Chaps
Cholic acid
DM
DPC
DHPC
DHPG
DLCP
LDAO
LMPC
LPPC
LOPC
LMPG
LPPG
LOPG
LPPE
LMPE
MLPA
PMAL-B
SDS, 37◦C
SDS, 42◦C
TRITON
Zwittergent 3–10
Zwittergent 3–12
Zwittergent 3–14
Zwittergent 3–16
30
33
ins.d
ins.
25
99
1
3
20
85
<1 n.a.c
18
ins.
ins.
1–2
<7
6
6
6
<1
<1
<1
<7
<1
9
9
2
93
96
n.a.
n.a.
10
100
76
20
100
100
94
n.a.
n.a.
n.a.
n.a.
126
n.a.
82
8
5
n.a.
<2
<0.1
<0.1
0
8
542
6
<1
<14
8
81
77
46
92
96
89
ins.
<1 50
28
18
29
86
86
ins.
n.a.
n.a.
n.a.
140
88
<0.1
<0.1
<0.1
<7
>30
>30
2.8
4.6
>40
>40
>40
<7
>30
>14
0
ins.
5
143
93
2
<1
<2
>0
>30
02
<1
<7
4
010
60
24
51
>7
>7
>7
>7
44
100
12
34
7–14
7–14
<1
18
14
4
<2
<2
n.a
aPercentage of expected 2D1H-15N HSQC cross-peaks actually observed. Subunit c and the β subunit were
measured at 37◦C, unless otherwise noted. Smr was measured at 42◦C. Subunit c and Smr were tested at a
minimum concentration of 0.5 mM and the β subunit at 0.3 mM. Detergent concentrations were 5% (w/v) for
subunit c, 8% (w/v) for the β subunit and 10% (w/v) for Smr. For the most successful detergents, 5% (w/v)
corresponds to 99 mM (LPPG), 94 mM (LOPG), 101 mM (LPPC), 142 mM (DPC) and 173 mM (SDS).
bNumber of days at which average signal intensity decreases to half its initial value.
cNot attempted.
dSolubility below the level of detection in the NMR experiments (see text).
intensities and line widths that were typical of spectra
where too few cross peaks were observed. Each pro-
tein had morethanthe expectednumberof cross peaks
in SDS at 37◦C (Table 1, and Figures2B,E).Although
the intensitieswerehigh, andsomelinewidthsnarrow,
the largenumberofamides showingtwo ormorecross
peaks suggest multiple conformationsfor each protein
in SDS at 37◦C. Increasing the temperature by 5◦C
gavespectrawithapproximatelytheexpectednumbers
of peaks for subunit c and Smr in SDS.
Some detergents gave protein spectra that had ap-
proximatelythecorrectnumberofcrosspeaks, buthad
other undesirableproperties. For all three proteins, the
number of cross peaks in LPPG and DPC were sim-
ilar. As seen in Figures 2C and F, the characteristics
of subunit c and the β subunit were less than ideal in
DPC (Smr was similar, data not shown). A wide range
of intensities, somewhat broader lines, and signific-
ant cross peak overlap were observed. Additionally,
DPC samples were not reproducible, especially for
subunit c. Often fewer than half of the expected cross
peaks were observed in contrast to the spectrum in
Figure 2C. This unpredictable behavior persisted with