Journal of Biomolecular NMR, 23: 57–61, 2002.
KLUWER/ESCOM
© 2002 Kluwer Academic Publishers. Printed in the Netherlands.
57
Refinement of the solution structure of the heparin-binding domain of
vascular endothelial growth factor using residual dipolar couplings
Melissa E. Stauffera,b,∗∗, Nicholas J. Skeltona & Wayne J. Fairbrothera,∗
aDepartment of Protein Engineering, Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080, U.S.A.;
bSchool of Molecular Biosciences, Washington State University, Pullman, WA 99164, U.S.A.
Received 21 December 2001; Accepted 22 February 2002
Key words: angiogenesis, protein structure, residual dipolar couplings, vascular endothelial growth factor
Abstract
Previous NMR structural studies of the heparin-binding domain of vascular endothelial growth factor (VEGF165)
revealed a novel fold comprising two subdomains, each containing two disulfide bridges and a short two-stranded
antiparallel β-sheet. The mutual orientation of the two subdomains was poorly defined by the NMR data. Het-
eronuclear relaxation data suggested that this disorder resulted from a relative lack of experimental restraints due
to the limited size of the interface, rather than inherent high-frequency flexibility. Refinement of the structure using
1HN-15N residual dipolar coupling restraints results in significantly improved definition of the relative subdomain
orientations.
Vascular endothelial growth factor (VEGF), an en-
dothelial cell (EC)-specific mitogen and motogen, is
a critical regulator of normal and pathological angio-
genesis (Ferrara, 2001). VEGF is a covalently-linked
homodimeric protein existing in at least six differ-
ent isoforms resulting from alternative exon splicing.
The different isoforms, containing 121, 145, 165,
183, 189, or 206 amino acids per monomer, share a
common 115-residue N-terminal domain that interacts
directly with the VEGF receptors Flt-1 (VEGFR-1)
and KDR (VEGFR-2). The most commonly expressed
isoform, VEGF165, and longer isoforms share the
same C-terminal heparin-binding domain. Removal
of this heparin-binding domain from VEGF165, ei-
ther by alternative exon splicing (i.e., VEGF121) or by
plasmin cleavage, is associated with a significant loss
(> 100-fold) in VEGF bioactivity (Keyt et al., 1996).
The enhanced bioactivity of the longer heparin-
binding isoforms of VEGF has been attributed recently
to the formation of ternary VEGF/KDR/neuropilin-1
complexes (Whitaker et al., 2001). Neuropilin-1 (NP-
∗To whom correspondence should be addressed. E-mail:
fairbro@gene.com
∗∗Present address: Department of Biochemistry, Vanderbilt
University, Nashville, TN 37232, U.S.A.
1) was identified previously as an isoform-specific
VEGF receptor that binds VEGF165 but not VEGF121;
the VEGF heparin-binding domain was identified as
the epitope for NP-1 binding (Soker et al., 1998).
When cotransfected into KDR-expressing cells, NP-
1 enhances the binding of VEGF165 to KDR and
increases the KDR-mediated mitogenic and chemo-
tactic activity of VEGF (Soker et al., 1998). More
recently, in vitro experiments have established that
the VEGF heparin-binding domain-mediated interac-
tion with NP-1 increases the affinity of VEGF165 for
KDR (Fuh et al., 2000). Furthermore, the affinity of
VEGF165 for the NP-1 extracellular domain is greatly
enhanced by the addition of heparin (Fuh et al., 2000).
Blocking VEGF165 binding to NP-1 with GST-fused
VEGF heparin-binding domain inhibits its binding to
KDR and its EC-mitogenic activity (Soker et al., 1997,
1998). Interestingly, NP-1 is also expressed by tumor
cells, where it acts as a positive modulator of angio-
genesis (Soker et al., 1998; Miao et al., 2000); over-
expression of a soluble variant of NP-1, that inhibits
VEGF165 binding to cell-bound NP-1, leads to tumors
with extensive hemorrhage, damaged blood vessels,
and apoptotic tumor cells (Gagnon et al., 2000). An-
tagonists of the VEGF heparin-binding domain/NP-1

58
interaction may therefore be useful for the treatment
of cancer.
We have reported previously the solution struc-
ture of the 55-residue C-terminal heparin-binding do-
main of VEGF165 (hereafter referred to as VEGF55)
(Fairbrother et al., 1998). The novel heparin-binding
domain fold comprises two sub-domains, each con-
taining a small two-stranded anti-parallel β-sheet and
two disulfide bonds; the C-terminal subdomain also
contains a short α-helix that packs against the β-sheet.
The orientation of the two subdomains with respect to
each other is poorly defined. 15N-relaxation data indi-
cate that the lack of definition results from a lack of
experimental restraints, due to the limited size of the
subdomain interface, rather than inherent flexibility on
the picosecond time scale. We report here refinement
of the structure of VEGF55 using 1HN-15N residual
dipolar coupling (RDC) data (Tjandra and Bax, 1997),
resulting in significantly improved definition of the
relative subdomain orientations.
Spectra were acquired at 27 ◦C on a Bruker DRX-
500 spectrometer equipped with a 5-mm inverse triple-
resonance probe with three-axis gradient coils, unless
stated otherwise. A broad-band inverse probehead was
used for acquiring 31P spectra. Spectra were processed
and analyzed using FELIX (Molecular Simulations,
Inc.).
The 15N-labeled sample of VEGF55 used for the
original structure determination (Fairbrother et al.,
1998) was used also in the present work. The
isotropic NMR sample contained 2.0 mM protein
in 25 mM sodium d3-acetate (pH 5.5), 50 mM
NaCl, 0.02% NaN3, 10% D2O. Partial alignment
was achieved by diluting the isotropic sample (3.5-
fold) into a liquid-crystalline bicelle medium com-
prising a mixture (4.0:1.0:0.2) of ditridecanoyl-
phosphatidylcholine (DTPC), dihexanoyl-phosphati-
dylcholine (DHPC), and cetyltrimethylammonium
bromide (CTAB) in the same buffer, to give 5% w/v to-
tal lipid (Ottiger and Bax, 1998; Losonczi and Preste-
gard, 1998). Based on 2H quadrupole splittings and
31P spectra the liquid-crystalline phase of this mixture
is stable between 25–35 ◦C.
1HN-15N splittings were measured under isotropic
and partially aligned conditions using 2D IPAP 1H-
15N HSQC experiments (Ottiger et al., 1998). Resid-
ual 1DNH dipolar couplings were extracted by sub-
tracting the 1JNH scalar coupling constant, measured
using the isotropic sample, from the 1JNH + 1DNH
values obtained using the liquid-crystalline bicelle
sample (Figure 1). Significant broadening observed
Figure 1. Superposition of selected sections from the 2D IPAP
1H-15N HSQC spectra of the (a) isotropic and (b) partially aligned
samples of VEGF55.
for some amide resonances suggests that the highly
basic VEGF55 interacts to some extent with the lipid
bicelles. Lack of significant chemical shift changes
indicates, however, that such interactions do not per-
turb the solution conformation of the protein. In cases
where the faster relaxing upfield component of the
IPAP 1H-15N HSQC was too broad for accurate mea-
surement, 1JNH +1 DNH was determined by compari-
son of the slower relaxing downfield component with
a regular decoupled 1H-15N HSQC spectrum acquired
using the same sample. Uncertainties in 1DNH were
estimated to be 1, 2 or 4 Hz depending on the de-
gree of line broadening. The residual 1DNH dipolar
couplings and uncertainties measured for VEGF55 are
summarized in Figure 2a.
The residual dipolar coupling between two nuclei
is given by:
D(θ,φ) = Da
{
(3 cos2 θ − 1)
+
3
2R(sin
2
θ cos 2φ)
}
, (1)
where R is the rhombicity defined as Dr/Da; Da
and Dr are the axial and rhombic components of the
alignment tensor given by 13 [Dzz − (Dxx + Dyy)/2]
and 13 [Dxx − Dyy], respectively; and θ and φ are the
cylindrical coordinates describing the orientation of
the internuclear vector in the principal axis system
of the molecular alignment tensor (Tjandra and Bax,
1997). The values of Da and R were estimated to
be 16.0 ± 0.5 Hz and 0.27 ± 0.04, respectively, by
fitting the experimental RDC values for 18 residues
in the well-defined C-terminal subdomain of VEGF55
to the original ensemble of 20 structures (PDB acces-
sion code 1VGH) using a simple Powell optimization