,
* Laboratory of Immunology and
Unit of Protein Biology, National Cancer Research Institute, Genoa, Italy; and
Laboratory of Tumor Immunology, Scientific Institute San Raffaele, Milan, Italy
Correspondence: Maria Raffaella Zocchi, Laboratory of Tumor Immunology, Scientific Institute San Raffaele, Via Olgettina 60 Milan, Italy. E-mail: zocchi.maria{at}hsr.it
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: AIDS • antigen-presenting cells • calcium • G proteins • phagocytosis
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
vß3 and vß5,
are crucial for the phagocytic process, mediating internalization of
apoptotic bodies and triggering of signal transduction pathways, which
precede antigen processing [2
3
4
]. Previously, we have
shown that the phagocytosis of apoptotic tumor cells, as other DC
functions [5
, 6
], is dependent on
extracellular calcium entry; in turn, apoptotic body engulfment
mediated by vß3 integrin elicits a calcium
influx in DC [2
]. Phagocytosis and calcium entry are
inhibited by HIV-1 Tat protein [7
], which is known to
contribute to immunosuppression in AIDS [8
], and this
inhibition is apparently related to the inactivation of L-type calcium
channels [9
]. With a similar mechanism, Tat impairs the
secretion by DC of interleukin-12 (IL-12), a cytokine that amplifies
the immune response [9
, 10
], thus
interfering with multiple functions of the APC. In the present work, we define a biochemical mechanism that follows extracellular calcium entry in DC, leading to the phagocytosis of apoptotic tumor cells. We show that apoptotic body engulfment is mainly dependent on the activation of calcium-calmodulin kinase II (CAMKII); the phosphoinositol-3-kinase (PI-3K) is partially involved; synthetic Tat blocks the activation of CAMKII elicited in DC by ß3 integrin; and pertussis toxin can prevent Tat-mediated inhibition, suggesting the involvement of an inhibitory guanosine monophosphate-binding (Gi) protein in DC-mediated phagocytosis.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
View this table: [in a new window] |
Table 1. Monocyte and Monocyte-Derived DC Phenotype by Flow Cytometry
|
v (CD51,
IgG1) from Serotec (Kidlingtone, Oxford, UK); anti-CD83 (IgG2a) from
Immunotech (Luminy Marseille, France); and anti-ß5
integrin rabbit antiserum obtained from Chemicon International
(Hofheim, Germany). On day 4, cells were checked for cytoplasmic
expression of CD83 (mostly surface-negative). Permeabilization was
performed with 0.01% Triton-X 100 after fixation with2%
paraforlmaldehyde. Cells (105 cells/sample) were stained
with the various antibodies for 30 min at 4°C, followed by
fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse (GAM) or
goat anti-rabbit (GAR) Ig. Samples were run on a FACSort (Becton
Dickinson), gated to exclude nonviable cells. At least 10,000 cells per
sample were analyzed, and results were plotted as percentage of
positive cells and mean fluorescence intensity [arbitrary units
(a.u.)].
Engulfment of apoptotic bodies by DC
Apoptosis was induced by sublethal irradiation (4000 rad) of the
human Jurkat lymphoma cell line (clone JA3), purchased from the
American Type Culture Collection (Manassas, VA). To check apoptosis,
JA3 cells were stained with 50 µg/ml propidium iodide
[2
]. Engulfment assay was performed as described
[2
]: Briefly, apoptotic bodies were labeled with
51Cr-Na (sodium chromate; NEN-DuPont, Cologno Monzese,
Italy) for 1 h at 37°C, washed, and cocultured with adherent DC
at the ratio of 3:1 at 37°C for 60 min. Noningested apoptotic cells
were removed by extensive washing, and DC-associated radioactivity was
measured in a 
-counter (Beckman Instruments, Irvine, CA) after cell
lysis. Phagocytosis of opsonized yeasts (not mediated by
vß3 integrin) or phagocytosis of apoptotic
cells at 4°C or in the presence of microtubule poisons was also
assayed (ref. [2
], and not shown in this paper). In
particular, DC could not phagocytose efficiently opsonized yeasts at
variance with activated monocytes [2
]. Results are
expressed as percentage of engulfment calculated as described
previously [2
]. In some experiments, DC were pretreated
for 15 min with 100 nM HIV-1 Tat or Tat peptides (see below; Technogen,
Piana di Monteverna, Italy). In other experiments, cells were exposed
to the vitronectin peptide (Vn-pep; 100 nM) DYMEQCKPQVTRGDF (PRIMM,
Milan, Italy), which binds to vß3
integrin, or the RGD-containing fibronectin peptide (Fn-pep; 100 nM;
Sigma Chemical Co., St. Louis, MO), which binds to
3ß1 or 5ß1
integrins or other fibronectin receptors [7
]. Some
samples were pretreated with
10-6–10-9 M cholera
toxin (CTX) or pertussis toxin (PTX), 10-8 M
adenylate-cyclase activator forskolin (FSK), and/or
10-8 M phosphodiesterase inhibitor
3-isobutyl-1-methylxantine (IBMX; Sigma Chemical Co.)
[11
]. In other experiments, DC were exposed to 10
µ–0.1 µM CAMKII inhibitors KN62 or KN93 or the inactive KN92
compund, the PI-3K inhibitors wortmannin (100 nM–1 nM) or LY294002 (10
µM–0.1 µM), the mitogen-activated protein kinase (MAPK) blocker
PD98059 (10 µM–0.1 µM), or 30–3 ng/ml rapamycin
(Calbiochem-MERCK, KgaA, Darmstadt, Germany).
CAMKII and Akt activation
CAMKII and Akt activation were measured in DC with the
respective assay kit, using the specific substrates and
32-adenosine 5'-triphosphate (ATP) after
immunoprecipitation with the specific anti-CAMKII or anti-Akt antibody
(Upstate Biotechnology, Lake Placid, NY). CAMKII and Akt activity were
tested 4 min after oligomerization of ß3 integrin with 5
µg/ml anti-CD61 mAb, 20 min at 4°C, followed by 10 µg/ml purified
F(ab')2 GAM Ig (Zymed Laboratories, San Francisco, CA) as
described [7
]. Control experiments using an
isotype-matched control antibody were also performed (ref.
[7
], and not shown in this paper). The timepoint of
maximal activation of either enzyme was assessed by experiments of
kinetics (not shown). As a control, cells were exposed to GAM
F(ab')2 alone and checked for kinase
activation. Some experiments were carried out in the presence of the
100 nM HIV-1 Tat or of the overlapping Tat-derived peptides: Tat20–39,
Tat24–51, Tat46–60, Tat56–70, and Tat65–80 (Technogen)
[12
]. In other experiments, cells were pretreated with
the Vn-pep or Fn-pep or with the specific kinase inhibitors KN62 or
LY294002, as above. Results are expressed as cpm x
10-3 and are the mean ± SD
of four independent experiments.
Intracellular cyclic adenosine monophosphate (cAMP) measurement
Aliquots of 2 x 106 DC were resuspended in
assay buffer and incubated with CTX or PTX
(10-6 M, 10-9 M) for
2 h at 37°C in shaking water bath; then, cells were lysed by
boiling for 5 min and centrifuged at 15,000 g for 10 min at
4°C. The cAMP content of each sample was determined using a
commercial kit (NEN-Dupont) [11
].
Results are expressed as pmoles/2 x 106 cells and are the mean ± SD of four independent experiments.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
vß3 integrin were up-regulated, and
CD83 began to be detectable in the cytoplasm on days 4–5 and at the
surface of 40–50% of the cells after 8 days (Table 1)
. Of interest,
these DC lacked the CD36 and CD68 antigens [2
,
7
], both markers of macrophages. At this day of culture,
DC were able to phagocytose apoptotic cells but not other meals such as
necrotic cells or opsonized yeasts [2
]. Phagocytosis of
apoptotic tumor cells by DC is dependent on extracellular calcium
entry, possibly through calcium channels [2
,
13
]; in particular, calcium interaction with calmodulin
induces conformational changes, which allow the binding of this protein
to CAMKII and its activation [14
]. Figure 1
shows that the CAMKII inhibitors KN62 and KN93, but not the
inactive compound KN92, reduced apoptotic body engulfment in a
dose-dependent way, from 90% at 10 µM concentration to 50% at 1
µM concentration ( Fig. 1A
). It is interesting that an inhibitory
effect was also observed, although to a lesser extent, using two
blockers of PI-3K [15
, 16
]: wortmannin
(20% inhibition at 100 nM concentration; Fig. 1B
) and LY294002 (40%
and 20% inhibition at 10 µM and 1 µM concentration, respectively;
Fig. 1C
). On the contrary, the MAPK blocker PD98059 did not affect
apoptotic body engulfment (Fig. 1C)
; likewise, rapamycin, which
inactivates the PI-3K substrate S6K but not Akt [15
,
17
], did not exhert any inhibition (Fig. 1D)
. These
results indicate that CAMKII is the major kinase involved in the
phagocytosis of apoptotic tumor cells by DC.
 View larger version (31K): [in a new window] |
Figure 1. Apoptotic body engulfment by DC is reduced by CAMKII and PI-3K
inhibitors. Apoptosis was induced by sublethal irradiation (4000 rad)
of the human Jurkat lymphoma cell line. 51Cr-labeled
apoptotic bodies were cocultured with adherent DC at the ratio of 3:1
at 37°C for 60 min. DC were pretreated (15 min) with the CAMKII
inhibitors KN62 and KN93 or the inactive compound KN92 (A), the PI-3K
inhibitor wortmannin (B), LY294002 or the MAPK inhibitor PD98059 (C),
or rapamycin (D) at the indicated concentrations. Noningested apoptotic
cells were removed, and DC-associated radioactivity was measured in a
-counter. Results are expressed as percentage of radioactivity
calculated as described previously [2
]. Mean ±
SD of six independent experiments. *, P <
0.05.
|
vß3 has
been described [19
, 20
]. Thus, we
investigated whether synthetic Tat or Tat peptides led to an impairment
in CAMKII or PI-3K activation upon ß3 engagement.
Figure 2
shows that the addition of synthetic Tat to DC blocks the
activation of CAMKII elicited via ß3 integrin, which is
also impaired by the use of the specific inhibitor KN62 but not by the
PI-3K blocker LY294002 (Fig. 2A)
. This would indicate that engagement
of ß3 integrin during phagocytosis triggers a CAMKII
pathway of activation. Along this line, synthetic Tat could only
slightly inhibit (<20%) Akt activation triggered via ß3
integrins (Fig. 2B)
, which in turn, is blocked by the PI-3K inhibitor
LY294002.
 View larger version (41K): [in a new window] |
Figure 2. ß3 Integrin-mediated activation of CAMKII and Akt is
regulated by HIV-1 Tat C-terminal domain. DC were incubated with
saturating amounts of ß3 integrin, followed by GAM; as a
control, cells were exposed to GAM alone. Some samples were pretreated
with synthetic Tat (100 nM), KN62 (10 µM), or LY294002 (10 µM) or
with Fn-pep, Vn-pep, or Tat peptides (100 nM; C), as indicated. Nil,
activation of CAMKII (A and C) or Akt (B) 4 min after ß3
integrin engagement in the absence of Tat. CAMKII and Akt activation
was measured in DC with the respective assay kit, using the specific
substrates and 32-ATP after immunoprecipitation with the
specific anti-CAMKII or anti-Akt antibody. Results are expressed as
cpm x 10-3 and are the mean ±
SD of four independent experiments. *, P <
0.05.
|
vß3-mediated signal
transduction.
The inhibition of apoptotic body engulfment by HIV-1 Tat involves a
pertussis toxin-sensitive G protein
Because CAMKII activation and calcium-channel function are
regulated by G proteins [13
, 14
], we
investigated the effect of PTX and CTX, two toxins able to
ADP-ribosylate distinct G-protein subunits [11
,
22
], on apoptotic body engulfment and on the inhibitory
effect of HIV-1 Tat. As shown in Figure 3 A
, treatment of DC with PTX, but not with CTX, enhanced the
phagocytosis of apoptotic cells by 20%. It is interesting that PTX, at
variance with CTX, could prevent the inhibition of apoptotic body
engulfment mediated by Tat (Fig. 3B)
or Tat C-terminal domain (Fig. 3C)
. It is also interesting that PTX was able to interfere with the
inhibition of engulfment exerted by the Vn-pep but not by the Fn-pep,
further supporting that the pathway inhibited by Tat is dependent on
vß3 integrin engagement. This effect is
unlikely because of a rise in intracellular cAMP concentrations:
Indeed, CTX and PTX can increase intracellular cAMP (Fig. 4 A
) [11
], and only PTX was able to prevent Tat
inhibition (Fig. 3B)
. Moreover, exposure of DC to FSK, which increases
cAMP levels through the activation of adenylate cyclase, or FSK plus
IBMX, an inhibitor of phosphodiesterase, did not affect apoptotic body
engulfment or the inhibitory effect of Tat (Fig. 4B)
. These data
suggest that DC-mediated phagocytosis is regulated by the recruitment
of a PTX-sensitive G protein.
 View larger version (76K): [in a new window] |
Figure 3. A putative PTX-sensitive G protein is involved in apoptotic body
engulfment by DC: interference by HIV-1 Tat. Phagocytosis of apoptotic
Jurkat cells by DC was evaluated as in Figure 1
. DC were pretreated (1
h) with the indicated concentrations of PTX or CTX alone (A) or before
exposure to synthetic Tat (100 nM; B) or to the indicated Tat peptides
or Vn-pep or Fn-pep (C). Noningested apoptotic cells were removed, and
DC-associated radioactivity was measured in a -counter. Nil,
percentage of engulfment by untreated DC. Results are expressed as
percentage of engulfment calculated as described previously
[2
] and are the mean ± SD of six
independent experiments. *, P < 0.05.
|
 View larger version (40K): [in a new window] |
Figure 4. Intracellular cAMP increase is not involved in apoptotic body
engulfment. (A) DC were incubated with CTX or PTX
(10-6 M), 10-8 M FSK,
10-8 M FSK, plus 10-8
M IBMX for 1 h at 37°C, lysed by boiling for 5 min, and
centrifuged at 15,000 g for 10 min at 4°C. The cAMP
content of each sample was determined using a commercial kit
[13
]. Nil, intracellular cAMP content in untreated DC.
Results are expressed as pmoles/2 x 106 cells and are
the mean ± SD of four independent experiments. (B)
Phagocytosis of apoptotic Jurkat cells by DC. 51Cr-labeled
Jurkat cells, induced to apoptosis by sublethal irradiation (4000 rad),
were cocultured with adherent DC at the ratio of 3:1 at 37°C for 60
min. DC were untreated (Nil) or pretreated (1 h) with
10-8 M FSK or 10-8 M
FSK plus 10-8 M IBMX in the presence or
absence of 100 nM Tat. Noningested apoptotic cells were removed, and
DC-associated radioactivity was measured in a -counter. Results are
expressed as percentage of engulfment calculated as described
previously [2
] and are the mean ± SD
of four independent experiments. *, P < 0.05.
|
 View larger version (37K): [in a new window] |
Figure 5. PTX prevents Tat-mediated inhibition of ß3
integrin-induced CAMKII activation. Engagement of ß3
integrin was performed as in Figure 2
. Some samples were pretreated (15
min) with PTX or CTX at the indicated concentrations before exposure to
100 nM synthetic Tat. Nil, activation of CAMKII 4 min after
ß3 integrin engagement in the absence of Tat. CAMKII
activation was measured using the specific substrate and
32-ATP after immunoprecipitation with the specific
anti-CAMKII. Results are expressed as cpm x
10-3 and are the mean ± SD
of four independent experiments. *, P < 0.05.
|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Phagocytosis of apoptotic cells by DC is now a well-known phenomenon
that contributes to the clearance of neoplastic or infected cells and
to antigen presentation to T lymphocytes [1
2
3
4
]. This is
an active process that needs the engagement of specific surface
receptors and an intracellular
[Ca2+]i rise
[2
3
4
]. At least two integrins contribute to apoptotic
body engulfment, vß3 and
vß5, mediating endocytosis and signal
transduction. In particular, the cytoplasmic tail of the
ß5 chain triggers Rac1 activation and phagosome formation
[3
], and ß3 elicits an intracellular
[Ca2+]i rise needed for the
phagocytic process [4
]. We found that phagocytosis of
apoptotic tumor cells by DC is dependent on the induction of CAMKII,
which is also triggered by ß3 engagement. Calcium
interaction with calmodulin elicits conformational changes, which allow
its binding to CAMKII and the activation of this enzyme
[13
]. In turn, CAMKII induces the opening of L-type
Ca2+ channels [23
], thus
creating a positive feedback, which possibly favors the phagocytic
process. This is in keeping with the observation that ß3
integrin-mediated signaling is strictly dependent on CAMKII activity
[18
] and with our previous findings that
vß3 complex mediated phagocytosis and
calcium mobilization in DC [2
]. So far, there is no
evidence that ß5 signaling leads to CAMKII activation.
Nevertheless, it has been shown that ß1,
ß3, and ß5 integrin-intracellular domains
share some common feature in signal transduction, raising the
possibility that CAMKII may be activated by any of these integrins,
which are able to elicit an extracellular calcium influx
[24
, 25
]. Along this line, it has been
shown recently that 5ß1 is regulated by
vß3 through the modulation of CAMKII
[26
]. Conversely, an endocytic route triggered via
vß;5 integrin, which in turn activates PI-3K, has been
described recently [27
], thus supporting the existence
of a cross-talk between the various integrin-dependent pathways of
kinase activation.
Previously, we have demonstrated that exogenous HIV-1 Tat inhibits the
engulfment of apoptotic bodies by interfering with extracellular
calcium entry elicited via ß3 integrin in DC
[7
]. We have also shown that Tat acts as a blocker of
L-type calcium channels that mediate calcium influx in DC and
lymphocytes [9
, 10
]. Herein, we provide
evidence that inhibition of apoptotic body engulfment is a result of
the block of CAMKII activation via ß3 integrin exerted by
the C-terminal, RGD-containing domain of Tat, which binds to ß
integrins [21
]. Because CAMKII contributes to the
up-regulation of L-type calcium-channel function [23
],
it is now clear why Tat inhibited the ß3
integrin-dependent calcium mobilization [7
].
PI-3K-dependent Akt activation via vß3 has
been described as well [19
, 20
]. Along this
line, we found that PI-3K is also involved in the phagocytosis of
apoptotic cells by DC, although to a lower extent than CAMKII. However,
synthetic Tat or Tat peptides led only to a slight inhibition of PI-3K
activation upon ß3 ligation. This is not surprising,
because this kinase is triggered by the engagement of the integrin
through a binding domain on the ß3 chain distinct from
the RGD recognition site [20
]. Moreover, at variance
with CAMKII, PI-3K activation is dependent on cytoplasmic calcium rises
as a result of not only extracellular calcium entry but also calcium
mobilization from intracellular stores, thus being less sensitive to a
selective impairment of extracellular calcium influx [15
,
17
].
In addition, we found that PTX could prevent the inhibition exerted by
HIV-1 Tat on apoptotic body engulfment by DC. This effect was not
simply because of a PTX-induced rise in intracellular cAMP, due to CTX
strongly increased cAMP levels without affecting phagocytosis or
Tat-mediated inhibition. Alternatively, a G protein might be involved.
This is of interest, because CAMKII activation and calcium-channel
function are regulated by different G proteins that maintain the
channels in the open (G) or closed (Gi)
state [13
, 14
]. One possible explanation
for our findings (see a model for this hypothesis in in Fig. 6
) is that PTX displaces a Gi protein involved in
calcium-channel inhibition [28
], thus allowing the
opening of these channels. The consequent rise in intracellular calcium
might activate CAMKII [18
], which in turn up-regulates
ß3 integrin function. Indeed, it has been demonstrated
that vß3-dependent cell migration,
inhibited by cyclic RGD peptides in vascular smooth muscle cells, can
be restored by maintaining elevated CAMKII activity, raising the
intracellular calcium levels with ionophores [18
]. A
role for Gi protein in down-regulating IL-12 production by
murine splenocytes has also been proposed: secretion of the cytokine
was restored by blocking Gi with PTX [29
].
It is interesting that we demonstrated that Tat inhibits IL-12
secretion by blocking L-type calcium channels in DC, and this
inhibition is not RGD-mediated. It is tempting to speculate that a
non-RGD domain located in the C-terminal portion of the protein
interacts with L-type calcium channels in the close state, e.g., when
they are blocked by a Gi protein. Another possibility is
that a PTX-sentive G subunit is engaged, allowing the
displacement of Gß complexes, which can contribute to
maintaining calcium channels in the open state. Indeed, calcium and
potassium channels are activated by the Gß subunits of
heterotrimeric G proteins [30
]. In particular,
engagement of G-protein-coupled receptors displaces Gß
from the G subunit [31
]. Thus,
Gß complexes can exert their biological effects,
including opening and regulation of calcium channels. In turn,
Gß dimers have been shown to stimulate vascular L-type
calcium channels via PI-3K [32
], thus providing an
additional explanation to our findings of a partial involvement of
PI-3K in apoptotic body engulfment by DC. Finally, we rule out a
potential activation of G-protein-coupled receptors by Tat in our
system, because the protein domains responsible for chemokine-receptor
binding did not exert any effect.
 View larger version (29K): [in a new window] |
Figure 6. Model of the interplay among Tat, G protein(s), and CAMKII.
Engagement of ß3 integrin during phagocytosis leads to
CAMKII activation, which in turn, contributes to the opening of calcium
channels and enhances the extracellular calcium entry, creating a
positive feedback. Occupancy of vß3 by
RGD-containing Vn-pep or Tat peptides would block this activation
pathway, which could be rescued by PTX as a result of the displacement
of a Gi protein involved in calcium-channel inhibition
[28
], thus allowing the opening of these channels. The
consequent rise in intracellular calcium might activate CAMKII, which
in turn, up-regulates ß3 integrin function
[18
]. Alternatively, a PTX-sensitive G
subunit is engaged, allowing the displacement of
Gß complexes, which can contribute to
maintaining calcium channels in the open state [30
],
possibly through the engagement of PI-3K [32
]. When the
vß3-integrin pathway is blocked (e.g., by
Tat), PI-3K might be activated alternatively via
vß5. Finally, the non-RGD domain, located
in the C-terminal portion of Tat, might interact with L-type calcium
channels in the close state, e.g., when they are blocked by a
Gi protein.
|
Received August 17, 2001; revised October 20, 2001; accepted October 25, 2001.
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
vß3 interin and requires intracellular and extracellular calcium Eur. J. Immunol. 27,1893-1900[Medline]This article has been cited by other articles:
 |
|
 |
|
  T. L. Herrmann, C. T. Morita, K. Lee, and D. J. Kusner Calmodulin kinase II regulates the maturation and antigen presentation of human dendritic cells J. Leukoc. Biol., December 1, 2005; 78(6): 1397 - 1407. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. R. Zocchi, P. Contini, M. Alfano, and A. Poggi Pertussis Toxin (PTX) B Subunit and the Nontoxic PTX Mutant PT9K/129G Inhibit Tat-Induced TGF-{beta} Production by NK Cells and TGF-{beta}-Mediated NK Cell Apoptosis J. Immunol., May 15, 2005; 174(10): 6054 - 6061. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Poggi, R. Carosio, D. Fenoglio, S. Brenci, G. Murdaca, M. Setti, F. Indiveri, S. Scabini, E. Ferrero, and M. R. Zocchi Migration of V{delta}1 and V{delta}2 T cells in response to CXCR3 and CXCR4 ligands in healthy donors and HIV-1-infected patients: competition by HIV-1 Tat Blood, March 15, 2004; 103(6): 2205 - 2213. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. L. Petersen, C. R. Morris, and J. C. Solheim Virus Evasion of MHC Class I Molecule Presentation J. Immunol., November 1, 2003; 171(9): 4473 - 4478. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
