Surface nucleolin participates in both the binding and endocytosis
of lactoferrin in target cells
Dominique Legrand
1
, Keveen Vigie
´
1
, Elias A. Said
2
, Elisabeth Elass
1
, Maryse Masson
1
,
Marie-Christine Slomianny
1
, Mathieu Carpentier
1
, Jean-Paul Briand
3
, Joe¨ l Mazurier
1
and Ara G. Hovanessian
2
1
Unite
´
de Glycobiologie Structurale et Fonctionnelle et Unite
´
Mixte de Recherche n°8576 du CNRS, Institut Fe
´

surface-expressed nucleolin was then demonstrated through
competitive binding studies between Lf and the anti-human
immunodeficiency virus pseudopeptide, HB-19, which binds
specifically surface-expressed nucleolin independently of
proteoglycans. Interestingly, binding competition studies
between HB-19 and various Lf derivatives in proteoglycan-
deficient hamster cells suggested that the nucleolin-binding
site is located in both the N- and C-terminal lobes of Lf,
whereas the basic N-terminal region is dispensable. On intact
cells, Lf co-localizes with surface nucleolin and together they
become internalized through vesicles of the recycling/deg-
radation pathway by an active process. Morever, a small
proportion of Lf appears to translocate in the nucleus of
cells. Finally, the observations that endocytosis of Lf is
inhibited by the HB-19 pseudopeptide, and the lack of Lf
endocytosis in proteoglycan-deficient cells despite Lf bind-
ing, point out that both nucleolin and proteoglycans are
implicated in the mechanism of Lf endocytosis.
Keywords: lactoferrin; surface nucleolin; receptor binding;
HIV; cancer.
Lactoferrin (Lf) is an 80 kDa iron-binding glycoprotein
found in external secretions (mainly milk) and in the
secondary granules of leukocytes. It has important func-
tions, such as modulation of the inflammatory response and
inhibition of cancer cell proliferation [1,2]. Lf has also been
reported to have potent antiviral activity against human
immunodeficiency virus (HIV)-1 and human cytomegalo-
virus infection in in vitro cell cultures [3–5]. In the case of its
anti-HIV activity, Lf appears to inhibit virus binding and/or
entry into permissive cells [5]. Although most Lf-binding

´
des Sciences
et Technologies de Lille, 59655 Villeneuve d’Ascq cedex, France.
Fax: + 33 320436555, Tel.: + 33 320337238,
E-mail:
Abbreviations: AZT, azidothymidine; bLf, bovine Lf isolated from
milk; bLfc, bovine lactoferricin (residues 17–41 of bLf); CHO, Chinese
hamster ovary; FITC, fluorescein isothiocyanate; HB-19,
5[Kw(CH
2
N)PR]-TASP; HB-19-biotin, HB-19 labeled with biotin;
HB-19-fluo, HB-19 labeled with FITC; hLf, human Lf isolated from
milk; hLf-biotin, hLf labeled with biotin hydrazide; hTf, human
transferrin; Lf, lactoferrin; TRITC, tetrarhodamine isothiocyanate.
(Received 28 August 2003, revised 13 November 2003,
accepted 17 November 2003)
Eur. J. Biochem. 271, 303–317 (2004) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03929.x
involvement of higher affinity binding sites has been
hypothesized [7].
Nucleolin, a major ubiquitous 105 kDa nucleolar protein
of exponentially growing eukaryotic cells, has been des-
cribed as a cell surface receptor for several ligands, such as
matrix laminin-1, midkine, attachment factor J, apo-B and
apo-E lipoproteins [17–21]. This RNA-binding phospho-
protein was found primarily in the nucleus where it is
involved in the regulation of cell proliferation and growth,
cytokinesis, replication, embryogenesis and nucleogenesis
[22]. More recently, nucleolin has been described as a shuttle
between the cell surface and the nucleus [17,21,23] and it was
proposed as a mediator for the extracellular regulation of

MB-231 was grown in Eagle’s minimal essential medium, as
described previously [16]. Three CHO cell lines were used
and propagated in Ham’s F12 medium: wild-type cells
(CHO K1); mutant cells defective in heparan-sulfate
proteoglycan expression (CHO 677); or mutant cells
defective in heparan- and chondroitin-sulfate proteoglycan
expression (CHO 618) [27]. The human T lymphocyte cell
lines Jurkat and MT-4 were routinely grown in RPMI-1640,
and HeLa-CD4-LTR-lacZ cells (HeLa P4 cells) were
cultured in Dulbecco’s modified Eagle’s medium, as
described previously [8,19,25,26]. The HIV-1 LAI isolate
was propagated and purified as reported previously [25].
Proteins
Native human Lf (hLf) was purified from fresh human milk
(obtained from a single donor) by ion-exchange chroma-
tography, as described previously [28]. Bovine Lf (bLf) was
kindly provided by Biopole (Brussels, Belgium). Chicken egg
white lysozyme and human transferrin (hTf) were purchased
from Sigma. In order to avoid possible steric hindrance of
the interactions of the hLf polypeptide with nucleolin, hLf
used for microscopy studies was labeled with biotin
hydrazide (Pierce, Rockford, IL, USA) through its glycan
moiety after mild periodate oxidation of N-acetylneuraminic
acid residues, as described previously [9]. Radioiodination of
hLf was carried out as described previously [8]. The purity
of native Lf and Lf derivatives used in the experiments
was confirmed by the migration of single protein bands in
SDS/PAGE.
Antibodies
Antibodies to hLf and nonimmune rabbit polyclonal sera

was produced in E. coli BL21 cells transformed with the
expression plasmid and purified by passing the cell lysate
through a 1 mL glutathione Sepharose 4B column (Amer-
sham Pharmacia Biotech). After washing, the gel was
incubated with thrombin (Amersham Pharmacia Biotech)
(50 U in 1 mL of NaCl/P
i
) overnight at 20 °Cwithgentle
mixing. The 40 kDa protein released from the gel was
injected into rabbit.
Preparation of Lf derivatives
Mild enzymatic digestion of hLf gave the N-terminally
deleted proteins hLf
)2N
(residues 3–692), hLf
)3N
(residues
4–692) and hLf
)4N
(residues 5–692) [6], the 30 kDa hLf N-t
(residues 4–283), 50 kDa hLf C-t (residues 284–692) and
18 kDa hLf N2 (residues 91–255) fragments [30]. rhLf
EGS
,
304 D. Legrand et al. (Eur. J. Biochem. 271) Ó FEBS 2003
a recombinant hLf whose sequence 28RKVRGPP34 was
replaced with EGS (the 365–367 C-terminal counterpart of
sequence 28–34), was produced in a baculovirus expression
system, as reported previously [6]. The 30 kDa bLf N-t and
50 kDa bLf C-t fragments, which are homologous to their

labeling with fluorescein (HB-19-fluo) or biotin (HB-19-
biotin), were as described previously [32].
Affinity chromatography studies
Purified hLf was immobilized on an Ultralink hydrazide gel
(Pierce), according to the manufacturer’s instructions, and
used to study the binding of proteins from MDA-MB-231
cell lysates. Two milligrams of protein was bound per mL of
Ultralink hydrazide gel. A total of 50 · 10
6
MDA-MB-231
cells were washed twice with NaCl/P
i
and lysed in NaCl/P
i
,
1% Triton-X-100 (w/v) containing 1 m
M
of the protease
inhibitor Pefabloc [4-(2-aminoethyl)-benzenesulfonyl fluor-
ide] (Roche Diagnostics, Mannheim, Germany) for 1 h at
4 °C. After centrifugation at 10 000 g for 30 min, the
supernatant was recovered, diluted 10-fold with NaCl/P
i
containing 1 m
M
Pefabloc and incubated overnight at 4 °C
with 150 lL of hLf-Ultralink gel (250 lg of protein). The
hLf-Ultralink gel was collected by centrifugation at 600 g
for 5 min and washed with 10 mL of NaCl/P
i

Extranuclear nucleolin was prepared by lysis of 0.9–
1.2 · 10
9
NaCl/P
i
-washed Jurkat cells at 4 °Cfor1hin
25 mL of 20 m
M
Tris/HCl, pH 7.6, 150 m
M
NaCl, 5 m
M
MgCl
2
,5 m
M
b-mercaptoethanol, 0.5% (v/v) Triton X-100,
1m
M
Pefabloc and Complete (Roche Diagnostics), a
protease inhibitor cocktail. The nuclei were pelleted by
centrifugation at 1000 g for 5 min and the supernatant was
then centrifuged at 12 000 g for 10 min prior to storage at
)80 °C. A rapid two-step chromatography procedure was
used to purify nucleolin from nucleus-free extracts. All steps
were performed at 4 °C using ice-cold buffers and columns
in the presence of 1 m
M
Pefabloc and Complete protease
inhibitor cocktail. The cytoplasmic extract of Jurkat cells

eluted in 50 lL fractions with 2 mL of buffer A containing
0.6
M
ammonium sulfate. Five or six eluted fractions
contained a single 105 kDa protein band corresponding
to nucleolin, as confirmed by immunoblotting with anti-
nucleolin Ig. Nucleolin was pooled and dialyzed against
NaCl/P
i
containing 1 m
M
Pefabloc at 4 °Cfor2hbefore
storage at )80 °C. A further control on a 7.5% SDS
acrylamide gel, stained with Coomassie blue, confirmed the
presence of the purified nucleolin as a single 105 kDa
protein band. Two 70 and 50 kDa protein bands, corres-
ponding to partial degradation products of nucleolin
[23,32], were observed in amounts lower than 5% of the
total protein.
Analysis of Lf binding to nucleolin in a surface plasmon
resonance biosensor
All materials and chemicals were from BIAcore AB
(Uppsala, Sweden). Analyses were performed at 25 °Con
a BIAcore 3000 biosensor, and Hepes-buffered saline (HBS-
EP) was used as a running buffer and for the dilution of
ligands and analytes. Human nucleolin, purified from the
extranuclear fraction of Jurkat cells, was immobilized at a
concentration of 1.6 lgÆmL
)1
in 0.1

3.1 software).
Analysis of the inhibition of HB-19 binding
to CHO cells by Lf and Lf derivatives
The inhibition of HB-19 binding to CHO cells was
investigated by fluorescence flow cytometry on a FACScal-
ibur flow cytometer (Becton-Dickinson). Preconfluent cells,
propagated in six-well cell culture plates (Nalge Nunc,
Rochester, NY, USA), were removed from plastic using the
nonenzymatic cell dissociation solution (Sigma) and gentle
pipetting. Pooled cells were washed twice with NaCl/P
i
.
They were then resuspended in fresh RPMI containing 1%
heat-inactivated fetal bovine serum and distributed into
1.5 mL centrifuge tubes ( 500 000 cells per tube). The cells
were incubated at 15 °C for 45 min in 100 lL of cell culture
medium containing 1 l
M
HB-19-fluo and 0–8 l
M
hLf.
After seven washes with NaCl/P
i
, the cells were analyzed by
flow cytometry. Binding specificity and reversibility controls
were performed with 0–50 l
M
unlabeled HB-19. For studies
on the nucleolin-binding site of Lf, CHO 618 cells were
incubated at 15 °C for 45 min in 100 lL of cell culture

culture plates in Ham’s F12 containing 10% fetal bovine
serum. In some experiments, cells were incubated with
Ham’s F12 containing 1% fetal bovine serum for 12 h prior
to performing the binding assays. Cells were then incubated
for 1 h at 4 °C with 250 lL of 0–3 l
M
125
I-labeled hLf in
Ham’s F12 containing 1% fetal bovine serum. Non-specific
binding was measured in the presence of a 100-fold molar
excess of unlabeled hLf. Cells were washed seven times with
fresh Ham’s F12 medium containing 1% fetal bovine
serum, and then lysed with 0.1
M
NaOH. The cell lysates
were recovered for gamma counting. For the binding
competition assays between hLf and HB-19, CHO cells were
incubated at 15 °Cfor45minwith1l
M
125
I-labeled hLf
and 0–100 l
M
HB-19. Washes were performed five times
with NaCl/P
i
containing 1% BSA and twice with NaCl/P
i
containing 0.3
M

M
NaCl, the fluorescence intensity was measured by flow
cytometry.
Confocal microscopy
Indirect immunofluorescence staining and confocal micros-
copy were used to visualize the fate of hLf in MDA-MB-231
cells and its co-localization with nucleolin and endosome
markers. For these experiments, cells were grown on eight-
well glass slides (Laboratory-Tek Brand Products, Naper-
ville, IL, USA) coated with collagen. Cells in Eagle’s
medium containing 10% fetal bovine serum were incubated
at 15 or 37 °C for 1–14 h with 3 l
M
hLf-biotin, alone or in
the presence of either polyclonal anti-nucleolin rabbit Ig
(1 : 100) or mouse mAb to the hTf receptor (CD71)
(1 : 200). Thirty minutes before the end of incubation at
37 °C, the ligand-containing medium was replaced with
fresh 37 °C-warmed Eagle’s medium containing 10% fetal
bovine serum, to allow endocytosis of the cell bound ligand.
Cells were washed a further five times with NaCl/P
i
and
twice with NaCl/P
i
containing 0.3
M
NaCl, prior to fixation
with 4% paraformaldehyde in NaCl/P
i

with NaCl/P
i
containing 1% BSA, cells were examined
using an LSM 510 confocal microscopic system (Carl Zeiss,
Esslingen, Germany). Procedures used to evidence capping
of surface nucleolin on MT-4 cells and endocytosis of hLf
into CHO cells [19,26], are briefly described in the legends
of Figs 5 and 9, respectively.
Results
The purified hLf is functional as an inhibitor of cell
proliferation and virus infection
Lf from fresh human milk was purified as described
previously [28]. This purified preparation inhibited the
proliferation of breast cancer MDA-MB-231 cells in a
dose-dependent manner, as reported previously [16]. In
[
3
H]thymidine incorporation experiments, the 50% inhibi-
tion of cell proliferation was observed at 50 lgÆmL
)1
(0.62 l
M
) Lf (data not shown). To study its antiviral
activity, we investigated the action of hLf on infection of
HeLa-CD4-LTR-lacZ cells (HeLa P4 cells) by the HIV-1
LAI isolate. HIV entry and replication in HeLa P4 cells
resulted in activation of the HIV long terminal repeat
(LTR), leading to expression of the lacZ gene. Conse-
quently, the b-galactosidase activity could be measured in
cell extracts to monitor HIV entry into cells [25]. The value

of cancerous human mammary gland MDA-MB-231
cells, affinity chromatography was performed on immo-
bilized hLf. A complex pattern of protein bands was
retained on Ultralink-immobilized hLf and eluted by
increasing salt concentrations. One of the major proteins
that were preferentially and quantitatively retained on
immobilized hLf was a 105 kDa protein, which mostly
eluted at 0.5
M
NaCl (data not shown, see the Materials
and methods). This band was not observed in a control
experiment using Ultralink-immobilized hTf (data not
shown). Trypsin degradation of the 105 kDa protein
band generated peptides whose molecular ion masses
were used for identification by MALDI-TOF. As shown
in Table 1, the measured masses of seven out of nine
Fig. 1. Human lactoferrin (hLf) inhibition of HIV entry by blocking
virus particle attachment to cells. HIV entry (A) and attachment (B)
were assayed in HeLa P4 cells, as described previously, at 37 and
20 °C, respectively [25]. (A) Entry of the HIV-1 isolate, LAI, was
monitored in HeLa P4 cells by expression of the lacZ gene (corres-
ponding to b-galactosidase) under the control of the HIV-1 LTR. Cells
were infected in the presence of azidothymidine (AZT) (5 l
M
), HB-19
(1 l
M
), or hLf (0.25, 0.50, 1, 2 or 4 l
M
). The b-galactosidase activity

2312.7 2312.6
298
VEGTEPTTAFNLFVGNLNFNK
318
2501.8 2501.8
487
TLVLSNLSYSATEETLQEVFEK
508
Ó FEBS 2003 Nucleolin is a cell surface lactoferrin-binding site (Eur. J. Biochem. 271) 307
major peptides between 812.39 and 2501.79 Da matched
with the computed masses of peptides between residues
298 and 624 of human nucleolin (Swiss-Prot accession
number P19338). Finally, the identity of the 105 kDa
band as nucleolin was further confirmed by immunoblot-
ting using a mAb specific for human nucleolin (3G4B2)
(data not shown).
Lf binds to human nucleolin through medium–affinity
interactions
Mainly characterized as a nucleolar protein, nucleolin is
continuously expressed on the surface of different types of
cells along with its intracellular pool within the nucleus and
cytoplasm. Surface and cytoplasmic nucleolin are similar
and can be differentiated from nucleolar nucleolin by their
distinct isoelectric points, occurring, most probably, as a
consequence of post-translational modifications [23]. To
assess hLf binding to surface and/or cytosolic nucleolin, we
isolated the protein from the extranuclear pool and
investigated the binding parameters and kinetics in a surface
plasmon resonance biosensor. Jurkat cells were used as a
source for human nucleolin because they express substantial

, respectively, using Langmuir’s one-site model,
which gave the best fit at all concentrations used (v
2
<2).
The equilibrium dissociation constant (K
d
), calculated from
the ratio of the kinetic rate constants (k
off
/k
on
), was
238 ± 15 n
M
, a value very similar to that calculated from
the extent of binding observed near equilibrium using a
Scatchard plot (249 ± 45 n
M
) (Fig. 2B). R
max
estimated at
2005 ± 50 RU, correlates well with the maximal bind-
ing expected for hLf to sensorchip-immobilized nucleolin
(3200 RU). These results demonstrated that hLf binds with
fast kinetics and medium affinity to nucleolin. It should be
noted that under similar conditions, hTf did not bind
nucleolin (data not shown).
Evidence for hLf binding to proteoglycan-independent
sites on cells
CHO mutant cells [27], wild-type CHO K1 cells, and

)6
M
and n ¼ 555 000 ±
20 000) and CHO 677 (K
d
¼ 2.1 ± 0.3 · 10
)6
M
and
n ¼ 190 000 ± 10 000) cells are relevant to the presence
of proteoglycans.
Fig. 2. Surface plasmon resonance sensorgram of the binding of human
lactoferrin (hLf) to human extranuclear nucleolin. The raw data shown
are representative of a set of three experiments. Human nucleolin,
purified from nucleus-free extracts of Jurkat cells, was immobilized
onto a CM5 sensorchip. Human Lf, at different concentrations
(40–2560 n
M
), was incubated with immobilized nucleolin and analyzed
on a BIAcore 3000 apparatus. (A) Surface plasmon resonance sen-
sorgrams. (B) Binding curve and the Scatchard plot derived from these
data at equilibrium (insert). RU, response unit.
308 D. Legrand et al. (Eur. J. Biochem. 271) Ó FEBS 2003
Evidence that nucleolin is the major proteoglycan-
independent hLf-binding site on cells
The presence of proteoglycan-independent hLf-binding sites
on CHO cells led us to investigate the possible involvement
of surface nucleolin. Hence, competition experiments for
the binding to cell surface nucleolin were performed
between hLf and the nucleolin-specific pseudopeptide,

is probably a result of the fact that the cell-surface binding
of HB-19 is mostly due to nucleolin [24,26], whereas the
binding of Lf to the cell surface implicates several molecules,
including mainly proteoglycans and nucleolin. Taken
together, our results suggest that nucleolin is the major
proteoglycan-independent Lf-binding site on CHO cells. As
the C-terminal tail of nucleolin is the site for HB-19 binding
Fig. 3. The presence, on cells, of human lactoferrin (hLf)-binding site(s)
different from proteoglycans. Binding experiments were performed by
incubating wild-type Chinese hamster ovary (CHO) K1 and the
mutant cell lines CHO 677 (heparan sulfate-deficient proteoglycans)
and CHO 618 (heparan and chondroitin sulfate-deficient proteogly-
cans), with
125
I-labeled hLf at concentrations ranging from 0 to 3 l
M
.
(A) Specific binding of hLf to CHO K1 (d), CHO 677 (j)andCHO
618 (r) cells. (B) Scatchard analysis of the data showing two classes of
hLf-binding sites on CHO K1 and CHO 677 cells in contrast to a single
class on CHO 618 cells. Data shown represent mean values ± SEM of
three experiments conducted in duplicate.
Fig. 4. Competition between human lactoferrin (hLf) and HB-19 for
binding to Chinese hamster ovary (CHO) wild-type and mutant cell lines.
(A) Inhibition, by HB-19, of hLf binding to CHO cells. CHO cells were
incubated (45 min, 15 °C) with 1 l
M
125
I-labeled hLf and 0–100 molar
excesses of HB-19. Data are expressed as percentages ± SEM from

these experimental conditions, the nucleolin signal was
patched at one pole of the cell, which coincided with the hLf
signal (Fig. 5A). On the other hand, in control cells treated
similarly, but in the absence of hLf, the nucleolin signal was
evenly distributed in the plasma membrane in a diffused
state (Fig. 5B). Such a ligand-dependent capping of surface
nucleolin is a specific event because the distribution of
another surface protein, CD45, was not affected (data not
shown; see Fig. 1 in ref. [26]).
The two lobes of Lf, but not its basic N-terminal
region, bind to surface nucleolin
The basic sequences 2RRRR5 and 28RKVR31, located at
the N terminus of hLf, have been reported to contribute to
most of the ionic hLf interactions, particularly with
proteoglycans and nucleic acids [6,37,38]. The sequence
28RKVR31 was also proposed as a candidate for the
binding of hLf to its hypothetical receptor expressed on
lymphocytes [6]. In order to investigate the domain in Lf
implicated in its interaction with nucleolin, we investigated
the capacity of various Lf constructs and derivatives to
inhibit the binding of HB-19-fluo to CHO 618 cells
(Fig. 6A,B). Consistent with the proteoglycan-independent
binding of HB-19 to the cell-surface expressed nucleolin
[24,26], hLf
)2N
,hLf
)3N
,hLf
)4N
and rhLf

M
)thatrepresented 10% of the total
binding (2.6–3.2 · 10
5
sites per cell) [7]. We investigated
whether surface nucleolin is expressed on MDA-MB-231
cells and if it accounts for the proteoglycan-unrelated
binding of hLf to cells. To achieve this, flow cytometry cell-
binding experiments were performed using HB-19-fluo in
Fig. 5. Capping of surface nucleolin as a result of surface-bound human lactoferrin (hLf). (A) MT-4 cells were incubated in the presence (+ Lf) of
1 l
M
hLfat20 °C for 30 min before further incubation (20 °C for 60 min) in the presence of rabbit immune serum (1 : 50) raised against hLf. After
partial fixation in 0.25% paraformaldehyde, the co-aggregation of hLf with nucleolin was investigated using the murine mAb D3 against human
nucleolin [23]. (B) The same experiment as described in (A) but without hLf as a control. The rabbit antibodies were revealed by Texas Red dye (TR)
conjugated donkey anti-rabbit Ig, whereas the murine antibody was revealed by fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse
IgG. A cross-section for each staining is shown with the merge of the two colors and the respective phase contrast. Experimental conditions have
been described previously [19,26].
310 D. Legrand et al. (Eur. J. Biochem. 271) Ó FEBS 2003
experimental conditions that avoid binding to proteogly-
cans; i.e. washing with NaCl/P
i
that contains 0.3
M
NaCl. In
addition, the expression of surface nucleolin and its fate
following binding with Lf were investigated by confocal
microscopy using the biotin-labeled hLf (hLf-biotin)
and polyclonal antibodies against the C-terminal part of
nucleolin (residues 345–706).

human lactoferrin (hLf) derivatives used in binding competition studies with HB-19 on Chinese hamster ovary (CHO) 618 cells. Numbers at the
ends of the strips correspond to the first and last amino acid residues of polypeptides. The dotted lines show the approximate locations of the four
structural domains: N1 and N2 domains (N-t lobe) and C1 and C2 domains (C-t lobe) [52]. The black boxes in the strips show the location of basic
sequences 1-GRRRR-5 and 28-RKVRGPP-34. (B) CHO 618 cells were incubated (45 min, 15 °C) with 1 l
M
HB-19-fluo and 8 l
M
hLf derivatives:
intact hLf, the N-terminally deleted proteins hLf
)2N
,hLf
)3N
and hLf
)4N
, recombinant hLf mutated at residues 28–34 (rhLf
EGS
), the 30 kDa (hLf
N-t), 50 kDa (hLf C-t) and 18 kDa (hLf N2) hLf tryptic fragments, and a synthetic octadecapeptide corresponding to residues 20–37 of hLf (hLf
20–27); bLf polypeptides: intact bLf, the 30 kDa (bLf N-t) and 50 kDa (bLf C-t) tryptic fragments of bLf and bovine lactoferricin (bLfc); control
molecules: HB-19 (50 l
M
), hTf and chicken egg white lysozyme (8 l
M
). The intensity of green fluorescence associated with the cells was measured
by flow cytometry. Data are expressed as mean percentages ± SEM for three separate experiments, performed in duplicate, from the total HB-19
bound to CHO 618 cells without hLf.
Ó FEBS 2003 Nucleolin is a cell surface lactoferrin-binding site (Eur. J. Biochem. 271) 311
Previous studies showed a growth arrest effect on
several cancerous mammary gland cell lines incubated for
12–24 h with hLf [16,39], but its possible endocytosis was

protein constituent of caveolae implicated in endocytosis
via a clathrin-independent pathway [40]. Our results
demonstrate that hLf complexed with surface nucleolin
undergo active endocytosis into MDA-MB-231 cells via
the clathrin-dependent pathway.
Endocytosis of hLf requires both surface-expressed
nucleolin and proteoglycans
In a series of experiments using confocal immunofluores-
cence laser microscopy, we demonstrated that endocytosis
of hLf occurs in different types of cells (HeLa, MDA-MB-
231 and MT-4). Such endocytosis occurs at 37 °C, but not
at 20 °C, indicating that it uses an active internalization
process, consistent with other nucleolin-binding ligands
[19,23]. Endocytosis of hLf at 37 °Cwasalsotime
dependent, reaching saturation at 60–90 min (data not
shown). The results presented in Figs 3 and 4 suggest that
both proteoglycans and nucleolin are implicated in the
overall amount of hLf in CHO cell lines that bind to the cell
surface. In view of this, we investigated endocytosis of hLf
in CHO cell lines, the wild-type K1 cells and the proteo-
glycan-deficient cell lines CHO 677 and 618. Consistently,
we found that hLf becomes internalized at 37 °CintoCHO
K1 cells but not into CHO 677 cells deficient in heparan-
sulfate expression (Fig. 9A) or into CHO 618 cells deficient
in both heparan- and chondroitin-sulfate expression (data
not shown). Under similar experimental conditions, hLf was
found at the plasma membrane in both CHO K1 and CHO
677 cells (Fig. 9B), as expected from the results shown in
Figs 3 and 4. Therefore, despite efficient binding to the cell
surface, heparan-sulfate proteoglycan expression is required

M
HB-19-fluo alone (HB-19-fluo) or with 1 l
M
HB-19-fluo in
thepresenceof50l
M
HB-19 (HB-19-fluo + HB-19) or 8 l
M
hLf (HB-
19-fluo + hLf). (C) Binding of hLf to MDA-MB-231 cells and its
inhibition by HB-19. The figure displays a typical profile of cells
incubated with 1 l
M
hLf-biotin alone (hLf-biotin) or in the presence of
50 l
M
HB-19 (hLf-biotin + HB-19) or 8 l
M
hLf (hLf-biotin + hLf)
or anti-nucleolin polyclonal immunoglobulin (1 : 200 dilution) (hLf-
biotin + anti-Nucl pAb). Fluorescence staining was achieved with
streptavidin-FITC. Controls with cells incubated without proteins
(None) and with streptavidin-FITC only (Avidine-FITC control) are
shown.
312 D. Legrand et al. (Eur. J. Biochem. 271) Ó FEBS 2003
Discussion
Lf appears to participate in host defense mechanisms
against various infections and cancer. Although the pro-
posed mechanisms for the antiproliferation properties of Lf
in cancer are still controversial, its nuclear targeting has

protein is similar to that of nucleolin, they should represent
two distinct proteins because hLf residues 28–34 were
reported to be essential for binding to the 105 kDa receptor
[6]. Further investigations are needed to confirm that
nucleolin and the 105 kDa receptor are distinct proteins
and, if need be, to assess the relative importance of both
proteins in cells.
Although hLf binding to nucleolin occurs with average
affinity and dissociates with elevated salt concentrations,
our results are not in favor of simple ionic interactions
between basic hLf and the N-terminal anionic domain of
nucleolin [36]. In accordance with this, we show that the hLf
sequences 2RRRR5 and 28RKVR31, responsible for most
Fig. 8. Colocalization of human lactoferrin (hLf) with nucleolin, both on the surface of MDA-MB-231 cells and in vesicles of the recycling/degradation
pathway. A series of 10 optical sections at 0.4 lm was performed through cells. The figure shows cross-sections towards the middle of representative
pairs of cells for streptavidin-fluorescein isothiocyanate (FITC) bound to hLf-biotin (green), tetrarhodamine isothiocyanate (TRITC)- or Alexa
Fluor 546-stained antibodies bound to nucleolin or vesicle markers (red) with the merge of the two colors (yellow). Staining of cells with either
streptavidin-FITC or the red-labeled secondary antibodies alone was not significantly different from background. Bars, 5 lm. (A) Binding of hLf-
biotin and rabbit polyclonal anti-nucleolin Ig to MDA-MB-231 cells. Cells were incubated for 1 h at 15 °Cwith3l
M
biotin-labeled hLf (green) and
rabbit anti-nucleolin Ig (1 : 100 dilution) (red). (B) Endocytosis of hLf-biotin and rabbit polyclonal anti-nucleolin immunoglobulin into MDA-MB-
231 cells. Cells were incubated for 2 h at 37 °Cwith3 l
M
hLf-biotin (green) and rabbit antiserum to nucleolin (1 : 100 dilution) (red). White arrows
show clusters that were fluorescent through one to three successive planes in the middle of cells and inside the nucleus. (C) Co-localization of hLf-
biotin with EEA1, the early endosome antigen 1. Cells were incubated for 2 h at 37 °Cwith3l
M
biotin-labeled hLf. Fixed cells were then incubated
with mouse mAbs to EEA1 for 45 min at 37 °C.

hLf binding is about twofold higher in wild-type CHO K1
cells compared with CHO 677 cells deficient in the
expression of heparan-sulfate proteoglycans (Fig. 3),
although both of these cell types express similar levels of
surface nucleolin [26]. In contrast, the inhibition of hLf
binding to these CHO cell lines by low concentrations of the
nucleolin-specific HB-19 pseudopeptide is much more
efficient in CHO 677 cells, i.e. in those cells that do not
express heparan-sulfate proteoglycans (Fig. 4). A cooper-
ative mechanism between proteoglycans and nucleolin
could account for this latter effect. Similarly, a cooperative
mechanism appears also to be operational for the endo-
cytosis of hLf (Fig. 9). Indeed, endocytosis is prevented by
either the nucleolin-binding HB-19 pseudopeptide or the
absence of heparan-sulfate proteoglycans. It should be
noted that active internalization of the cytokine, midkine,
following binding to surface nucleolin does not require
proteoglycans, as midkine can be internalized at a similar
extent into CHO cell lines expressing or not expressing
heparan-sulfate or heparan/chondroitin-sulfate proteogly-
cans [19]. Therefore, the mechanism implicated in the
internalization of Lf should be different from that of
midkine, although both of these ligands require the presence
of accessible surface nucleolin. Interestingly, both Lf and
midkine are internalized at 37 °C, but not at 20 °C. As
fibroblast growth factor 2, which uses heparan-sulfate
proteoglycans as a low affinity receptor, is internalized,
even at 4 °C [19], endocytosis could occur either by an active
process through nucleolin or by a passive process through
heparan sulfate proteoglycans. The observation that endo-

was rabbit immune serum (1 : 50 dilution) raised against hLf, whereas
the secondary antibody was Texas Red-labeled donkey anti-rabbit IgG.
A scan corresponding to a cross-section towards the middle of the cell
monolayer is shown, together with the respective phase contrast.
314 D. Legrand et al. (Eur. J. Biochem. 271) Ó FEBS 2003
However, while loop GENQ of the bLf N2 domain is very
close to its hLf counterpart, GLDK is not. This feature
could explain the least inhibitory effect of HB-19 binding to
CHO 618 cells by bLf C-t. Further experiments, using
mutated Lfs, are required to assess the involvement, if any,
of these loops in Lf–nucleolin interactions.
Evidence for binding of Lf to surface nucleolin
may enlighten on the multifunctional properties of Lf.
MDA-MB-231 cells were previously used for studying the
antiproliferativeeffectofLfoncancerousmammarygland
cells [7,16]. Cell growth arrest was connected to both
inhibition of Cdk2 and Cdk4 activities and increase of Cdk
inhibitor p21 expression [7]. Whether some of these
responses are induced following interaction of Lf with
nucleolin remains to be elucidated. Our results show that
hLf co-localizes with nucleolin at the surface of MDA-MB-
231 cells and that both molecules undergo endocytosis by an
active process into vesicles of the classical endocytosis
pathway. In light of our results, it is probable that nucleolin
is responsible, at least in part, for the endocytosis of Lf and
its nuclear targeting. Internalization of specific antibodies
bound to surface nucleolin [23,29], which then gain access to
the nucleus, has been reported previously [29]. It has also
been reported that the pleiotropic activities of a number of
growth factors are mediated not only through receptor-

attachment of HIV particles to the cell surface occurs, on
the one hand, through the coordinated interactions with
heparan-sulfate proteoglycans and, on the other hand, with
the cell surface-expressed nucleolin [26]. Consequently,
targeting any one of these components could result in the
inhibition of HIV attachment. Indeed, HIV attachment
could be blocked either by the fibroblast growth factor 2
that binds heparan-sulfate proteoglycans, or by the anti-
HIV pseudopeptide HB-19 that binds nucleolin [24,26].
Therefore, the capacity to bind proteoglycans and surface
nucleolin makes Lf a potent inhibitor of HIV attachment
and thus infection. Previously, several laboratories had
reported the capacity of Lf to inhibit HIV infection with an
IC
50
value of 0.4–1.5 l
M
[3,4]. Here, we show that the
purified hLf inhibits HIV infection, in the experimental
model of HeLa P4 cells, with an estimated IC
50
value of
0.25 l
M
. Moreover, we show that the mechanism of this
inhibition is a result of the marked blockade of HIV
attachment to cells.
In conclusion, our results show that nucleolin, in addition
to proteoglycans, is a major Lf-binding site at the surface of
cells. As nucleolin is a ubiquitous multiligand protein

functions might rely on the capacity of Lf to enter cells, as
we have illustrated in cancerous mammary gland cells.
Other functions, such as the anti-HIV activity of Lf, might
be the consequence of competition for binding to surface
nucleolin and to proteoglycans. Further investigations are
necessary in order to elucidate the participation of Lf–
nucleolin and Lf–proteoglycan interactions in the overall
mechanism of action of Lf.
Acknowledgements
This work was supported, in part, by the Universite
´
des Sciences et
Technologies de Lille, the Centre National de la Recherche Scientifique
(UMR n°8576; Director: Dr J. -C. Michalski), and Agence Nationale
de la Recherche sur le SIDA (research grant to A. G. Hovanessian). We
thank S. Baveye for participating in preliminary studies. We thank
C. Slomianny and E. Perret for confocal microscopy and J P.
Decottignies, M. Benaı
¨
ssa and J. Svab for excellent technical assistance.
References
1. Ward, P.P., Uribe-Luna, S. & Conneely, O. (2002) Lactoferrin
and host defense. Biochem. Cell Biol. 80, 95–102.
2. Brock, J.H. (2002) The physiology of lactoferrin. Biochem. Cell
Biol. 80, 1–6.
3. Harmsen,M.C.,Swart,P.J.,deBethune,M.P.,Pauwels,R.,De
Clercq, E., The, T.H. & Meijer, D.K. (1995) Antiviral effects of
plasma and milk proteins: lactoferrin shows potent activity against
both human immunodeficiency virus and human cytomegalovirus
replication in vitro. J. Infect. Dis. 172, 380–388.

11. Huettinger, M., Retzek, H., Hermann, M. & Goldenberg, H.
(1992) Lactoferrin specifically inhibits endocytosis of chylomicron
remnants but not alpha-macroglobulin. J. Biol. Chem. 267, 18551–
18557.
12. Dhennin-Duthille, I., Masson, M., Damiens, E., Fillebeen, C.,
Spik, G. & Mazurier, J. (2000) Lactoferrin upregulates the
expression of CD4 antigen through the stimulation of the mito-
gen-activated protein kinase in the human lymphoblastic T Jurkat
cell line. J. Cell Biochem. 79, 583–593.
13. Oh, S.M., Hahm, D.H., Kim, I.H. & Choi, S.Y. (2001) Human
neutrophil lactoferrin trans-activates the matrix metalloproteinase
1 gene through stress-activated MAPK signaling modules. J. Biol.
Chem. 276, 42575–42579.
14. Garre, C., Bianchi-Scarra, G., Sirito, M., Musso, M. &
Ravazzolo, R. (1992) Lactoferrin binding sites and nuclear loca-
lization in K562 (S) cells. J. Cell. Physiol. 153, 477–482.
15. He, J. & Furmanski, P. (1995) Sequence specificity and tran-
scriptional activation in the binding of lactoferrin to DNA. Nature
373, 721–724.
16. Damiens,E.,ElYazidi,I.,Mazurier,J.,Duthille,I.,Spik,G.&
Boilly-Marer, Y. (1999) Lactoferrin inhibits G1 cyclin-dependent
kinases during growth arrest of human breast carcinoma cells.
J. Cell Biochem. 74, 486–498.
17. Kleinman, H.K., Weeks, B.S., Cannon, F.B., Sweeney, T.M.,
Sephel, G.C., Clement, B., Zain, M., Olson, M.O., Jucker, M. &
Burrous, B.A. (1991) Identification of a 110-kDa nonintegrin cell
surface laminin-binding protein which recognizes an A chain
neurite-promoting peptide. Arch. Biochem. Biophys. 290, 320–325.
18. Take, M., Tsutsui, J., Obama, H., Ozawa, M., Nakayama, T.,
Maruyama, I., Arima, T. & Muramatsu, T. (1994) Identification

nessian, A.G. (2000) The HB-19 pseudopeptide 5[Kpsi (CH2N)
PR]-TASP inhibits attachment of T lymphocyte- and macro-
phage-tropic HIV to permissive cells. AIDS Res. Hum. Retro-
viruses 16, 237–249.
26. Nisole, S., Said, E.A., Mische, C., Prevost, M.C., Krust, B.,
Bouvet, P., Bianco, A., Briand, J.P. & Hovanessian, A.G. (2002)
The anti-HIV pentameric pseudopeptide HB-19 binds the
C-terminal end of nucleolin and prevents anchorage of virus
particles in the plasma membrane of target cells. J. Biol. Chem.
277, 20877–20886.
27. Esko, J.D., Rostand, K.S. & Weinke, J.L. (1988) Tumor forma-
tion dependent on proteoglycan biosynthesis. Science 241, 1092–
1096.
316 D. Legrand et al. (Eur. J. Biochem. 271) Ó FEBS 2003
28. Spik, G., Strecker, G., Fournet, B., Bouquelet, S., Montreuil, J.,
Dorland, L., van Halbeek, H. & Vliegenthart, J.F. (1982) Primary
structure of the glycans from human lactotransferrin. Eur. J.
Biochem. 121, 413–419.
29. Deng, J.S., Ballou, B. & Hofmeister, J.K. (1996) Internalization of
anti-nucleolin antibody into viable HEp-2 cells. Mol. Biol. Report
23, 191–195.
30. Legrand, D., Mazurier, J., Aubert, J.P., Loucheux-Lefebvre,
M.H., Montreuil, J. & Spik, G. (1986) Evidence for interactions
between the 30 kDa N- and 50 kDa C-terminal tryptic fragments
of human lactotransferrin. Biochem. J. 236, 839–844.
31. Legrand, D., Mazurier, J., Colavizza, D., Montreuil, J. & Spik, G.
(1990) Properties of the iron-binding site of the N-terminal lobe of
human and bovine lactotransferrins. Importance of the glycan
moiety and of the non-covalent interactions between the N- and
C-terminal lobes in the stability of the iron-binding site. Biochem.

de Boer, H.A. & Nuijens, J.H. (1997) N-terminal stretch Arg2,
Arg3, Arg4 and Arg5 of human lactoferrin is essential for binding
to heparin, bacterial lipopolysaccharide, human lysozyme and
DNA. Biochem. J. 328, 145–151.
39. Hurley, W.L., Hegarty, H.M. & Metzler, J.T. (1994) In vitro
inhibition of mammary cell growth by lactoferrin: a comparative
study. Life Sci. 55, 1955–1963.
40. Smart, E.J., Graf, G.A., McNiven, M.A., Sessa, W.C., Engelman,
J.A., Scherer, P.E., Okamoto, T. & Lisanti, M.P. (1999) Caveolins,
liquid-ordered domains, and signal transduction. Mol. Cell. Biol.
19, 7289–7304.
41. Penco, S., Scarfi, S., Giovine, M., Damonte, G., Millo, E., Villaggio,
B., Passalacqua, M., Pozzolini, M., Garre, C. & Benatti, U. (2001)
Identification of an import signal for, and the nuclear localization
of, human lactoferrin. Biotechnol. Appl. Biochem. 34, 151–159.
42.Ghisolfi,L.,Joseph,G.,Amalric,F.&Erard,M.(1992)The
glycine-rich domain of nucleolin has an unusual supersecondary
structure responsible for its RNA-helix-destabilizing properties.
J. Biol. Chem. 267, 2955–2959.
43. Hanakahi, L.A., Sun, H. & Maizels, N. (1999) High affinity
interactions of nucleolin with G–G-paired rDNA. J. Biol. Chem.
274, 15908–15912.
44. Bouvet, P., Diaz, J J., Kindbeiter, K., Madjar, J J. & Amalric, F.
(1998) Nucleolin interacts with several ribosomal proteins through
its RGG domain. J. Biol. Chem. 273, 19025–19029.
45. Dougherty, D.A. (1996) Cation–pi interactions in chemistry and
biology: a new view of benzene, Phe, Tyr, and Trp. Science 271,
163–168.
46. Yu, R.H. & Schryvers, A.B. (1993) Regions located in both the
N-lobe and C-lobe of human lactoferrin participate in the binding