Content uploaded by Kristin White
Author content
All content in this area was uploaded by Kristin White
Content may be subject to copyright.
ganic peroxides, nitro compounds, azo compounds, or
inorganic sulfur, was detected.
26. J. I. Kroschwitz Ed., Encyclopedia of Polymer Science
and Engineering (Wiley, New York, 1985).
27. This behavior is consistent with our observation that
the addition of polar organic solvents such as acetone
or ethanol to the latex emulsion induces latex coag-
ulation, but addition of water alone does not produce
this effect (26).
28. We thank Mexico’s Consejo de Arqueologia at the
Instituto Nacional de Antropologia e Historia (INAH)
and J. Garcia-Ba´rcena, C. Rodriguez, and P. Ortiz of
INAH for permissions to perform this study; in Chia-
pas, J. Gasco, F. Guillen, L. Guillen, and A. Castan˜eda;
in the United States, C. Coggins; at Harvard Univer-
sity, the staff at the Peabody Museum and Botany
libraries; and V. Williams, M. Rubner, C. Scott, W.
Williams, H. Lechtman, and the Undergraduate Re-
search Opportunities Program at the Massachusetts
Institute of Technology.
23 December 1998; accepted 30 April 1999
Requirement for Croquemort in
Phagocytosis of Apoptotic Cells
in Drosophila
Nathalie C. Franc,
1
Pascal Heitzler,
4
R. Alan B. Ezekowitz,
2,3
Kristin White
1
*
Macrophages in the Drosophila embryo are responsible for the phagocytosis of
apoptotic cells and are competent to engulf bacteria. Croquemort (CRQ) is a
CD36-related receptor expressed exclusively on these macrophages. Genetic
evidence showed that crq was essential for efficient phagocytosis of apoptotic
corpses but was not required for the engulfment of bacteria. The expression of
CRQ was regulated by the amount of apoptosis. These data define distinct
pathways for the phagocytosis of corpses and bacteria in Drosophila.
Phagocytosis is the terminal event of the
apoptotic process (1,2) and is also critical for
the engulfment of microorganisms (3). It has
been proposed that the recognition of both
nonself (microorganisms) and effete self
(corpses) may share common receptors (4).
Blocking experiments have implicated a
number of receptors as important for target
recognition (2– 4 ). Genetic studies indicate
that some of these receptors participate in
phagocytosis of pathogens in vivo (5,6).
However, the multiplicity and redundancy of
recognition mechanisms in mammalian sys-
tems have made it difficult to evaluate the
relative roles of these receptors in the phago-
cytosis of corpses. Although several genes of
Caenorhabditis elegans are involved in the
phagocytosis of corpses (7–9), none of these
molecules seem to act directly as a receptor in
the recognition of the corpse.
In Drosophila embryos, like in mammals
and in contrast to worms, the clearance of apo-
ptotic cells is primarily mediated by macro-
phages, hemocytes that become phagocytic at
the initiation of developmentally regulated apo-
ptosis (10). Croquemort (CRQ), a Drosophila
CD36-related receptor, is specifically expressed
on all embryonic macrophages (11). Human
CD36 acts as a scavenger receptor (12–14) and
also binds apoptotic cells in combination with
the macrophage vitronectin receptor and throm-
bospondin (15,16). CD36 has the ability to
confer phagocytic activity on nonphagocytic
cells on transfection (17,18). CRQ expression
in nonphagocytic Cos7 cells allows these cells
to recognize and engulf apoptotic thymocytes
(11). Thus, CRQ may participate in the removal
of apoptotic cells during Drosophila embryo-
genesis. We genetically evaluated the relative
Fig. 1. Macrophages in crq-deficient embryos have very poor phagocytic
activity for apoptotic cells. (Ato F) In confocal micrographs, peroxidasin-
stained hemocytes appear green, CRQ staining appears blue, 7-AAD–
stained apoptotic corpses appear as bright red round particles, and the
nuclei of viable cells appear as large red diffused components. All images
are the sum of eight focal planes. (A) to (C) show a ⫻40 magnified lateral
view of the head region of (A) a In(2L)Cy homozygous embryo, (B) a
Df(2L)al homozygous embryo, and (C) a W88 homozygous embryo. (D)
to (F) show high-magnification views (⫻400) of their respective macro-
phages. As compared with the wild-type distribution (A) and phagocytic
activity (D) of macrophages within In(2L)Cy homozygous embryos, mac-
rophages in Df(2L)al (B) and W88 (C) homozygous embryos accumulate
in the head and around the amnioserosa and show very poor phagocytic
activity despite their recruitment at sites of abundant apoptosis (E and F).
Asterisks indicate the nucleus of each macrophage seen in these fields. (G)
A chart summarizes the efficiency of phagocytosis of apoptotic corpses
observed within each genotyped embryos assayed. Results shown are the
mean P.I. ⫾SE; nis the total number of macrophages scored for each
genotype. Dark blue, w; In(2L)Cy/In(2L)Cy; red, w; In(2L)Cy/Df(2L)al; yellow,
w; Df(2L)al/Df(2L)al; and light blue, w; W88/W88.
REPORTS
www.sciencemag.org SCIENCE VOL 284 18 JUNE 1999 1991
role of this receptor in phagocytosis of apoptotic
cells and in other macrophage functions in vivo.
To look at the crq-null phenotype, we used
two overlapping deletions of the 21C region,
Df(2L)al (19) and Df(2L)TE99(Z)XW88 (W88)
(20). The Drosophila genome project sequence
indicates that crq is at position 21C4 between
expanded (ex)(21) and u-shaped (ush)(22,23).
Df(2L)al removes about 180 kb from the ari-
staless gene (al )(19)toush, whereas W88
uncovers about 100 kb from ex to ush (20).
Homozygous embryos for either of these defi-
ciencies can be distinguished by morphological
defects (19,22). Both polymerase chain reac-
tion (PCR) on single embryos (24) and CRQ
immunostaining (11) confirmed that these ho-
mozygous embryos are crq null.
We assayed phagocytosis of apoptotic
corpses in Df(2L)al and W88 homozygous
embryos with a double fluorescent immuno-
labeling for CRQ and peroxidasin, a hemo-
cyte marker (10), and a nuclear dye, 7-amino
actinomycin D (7-AAD) (25). Macrophages
in wild-type embryos and embryos homozy-
gous for the balancer chromosome were
phagocytic for apoptotic corpses (Fig. 1, A
and D) with a mean phagocytic index (P.I.) of
3.96 corpses per macrophage (Fig. 1G) (26).
Although they accumulated at the site of cell
death, macrophages within Df(2L)al and W88
homozygous embryos remained very small
and round (Fig. 1, B, C, E, and F), with P.I.s
of 0.26 and 0.21, respectively (Fig. 1G).
In the absence of crq single mutants, we
could not definitively conclude that the phe-
notype observed in Df(2L)al or W88 homozy-
gous embryos resulted solely from the dele-
tion of crq. Therefore, we generated a UAS-
crq transgene (27) and tested the ability of
ubiquitously expressed CRQ to rescue the
engulfment defect in Df(2L)al homozygous
embryos, using a hsGal4 transgene to drive
expression. In heat-shocked mutant embryos
that carried both the hsGal4 and UAS-crq
transgenes, macrophages showed substantial
CRQ expression and phagocytic activity for
apoptotic cells, with a P.I. of 2.20 (Fig. 2, C,
F, and G). This indicates that crq is sufficient
to rescue the phagocytosis defect in Df(2L)al
homozygous embryos.
In serial sections of embryos that ubiqui-
tously expressed CRQ, we observed that apo-
ptotic corpses were not engulfed by cells other
than macrophages (28,29). Thus, ectopic ex-
pression of CRQ is not sufficient to confer
phagocytic ability on other cells in the embryo.
This finding is in contrast with our previous
observation that CRQ expression was sufficient
to confer phagocytic activity on Cos7 cells (13).
However, in UAS-crq; hsGal4 embryos, CRQ
was found at only low levels in cells other than
macrophages, suggesting that CRQ might be
unstable in other cells.
A human macrophage receptor, CD14, par-
ticipates in both recognition and engulfment of
pathogens as well as of apoptotic cells (30,31).
We tested whether crq participates in phagocy-
tosis of pathogens by Drosophila embryonic
macrophages. We injected fluorescently labeled
bacteria into living stage 11 wild-type and W88
embryos and monitored their fate by confocal
microscopy (32). Wild-type embryonic macro-
phages engulfed both Gram-negative (Esche-
richia coli) (Fig. 3, A to C) and Gram-positive
bacteria (Staphylococcus aureus)(28). In W88
homozygous embryos identified by their
u-shaped phenotype (Fig. 3D), macrophages
also engulfed bacteria (Fig. 3, D and E). Al-
though these assays are not quantitative, we
conclude that crq is specifically required for the
phagocytosis of apoptotic corpses and is not
essential for the engulfment of bacteria. crq is
also not necessary for endocytosis of acetylated
low density lipoproteins (LDLs) (33) (Fig. 3F)
or for the production of the extracellular matrix
components peroxidasin and MDP-1 (Fig. 1, B
and C) (28).
The onset of CRQ expression corresponds
to the time at which developmentally regu-
lated apoptosis begins. We tested whether the
presence of apoptotic cells might regulate
CRQ expression by examining CRQ protein
levels in embryos with altered amounts of
apoptosis. In Df(3L)H99 (H99) homozygous
embryos, apoptosis does not occur, as a result
of the deletion of the cell death regulators
reaper (rpr), grim, and head involution de-
fective (hid)(34 –36 ). However, macro-
phages in H99 homozygotes engulf corpses
when apoptosis is induced by high levels of
x-ray irradiation (34). When quantified by
1
Cutaneous Biology Research Center,
2
Laboratory of
Developmental Immunology, and
3
Department of Pe-
diatrics, Massachusetts General Hospital, Harvard
Medical School. Charlestown, MA 02129, USA.
4
Insti-
tut de Ge´ne´tique et de Biologie Mole´culaire et Cellu-
laire (IGBMC), Centre National de la Recherche Sci-
entifique/Institut National de la Sante´etdelaRe-
cherche Me´dicale/Universite´ Louis Pasteur de Stras-
bourg (CNRS/INSERM/ULP), 67404 Illkirch Cedex,
France.
*To whom correspondence should be addressed. E-
mail: kristin.white@CBRC2.MGH.Harvard.edu
Fig. 2. Expression of a croquemort transgene reinstates the
ability of macrophages in Df(2L)al homozygotes to recognize
and engulf apoptotic corpses. A UAS-crq transgene was ex-
pressed under the control of a hsGal4 driver in all cells of
Df(2L)al embryos, and phagocytosis of apoptotic corpses was
assessed as in Fig. 1 (25). (Ato C) Single focal planes. (Dto F)
Images are the sum of eight focal planes. (A) to (C) show a ⫻100
view of macrophages within the head of embryos with the
following genotypes: (A) control: w; CyO,S/CyO,S; hsGal4/⫹; (B)
homozygous mutant, as recognized by aberrant morphology: w;
Df(2L)al/Df(2L)al; hsGal4/hsGal4; and (C) transgene rescue of
homozygous mutant: w, UAS-crq; Df(2L)al/Df(2L)al; hsGal4/⫹.
All embryos were heat-shocked for 1 hour at 39°C and assayed
2 hours later. (D) to (F) show high-magnification views (⫻400)
of macrophages within the embryos shown in (A) to (C), respectively. Asterisks indicate the nucleus
of each macrophage seen in these fields. Arrows indicate free apoptotic corpses. (G) A chart
summarizes the efficiency of phagocytosis of apoptotic corpses observed within each category of
embryos assayed. Results shown are the mean P.I. ⫾SE; nis the total number of macrophages
scored for each genotype. Blue, w; CyO,S/CyO,S; hsGal4/hsGal4; red, w; Df(2L)al/Df(2L)al; hsGal4/
hsGal4; and yellow, w, UAS-crq; Df(2L)al/Df(2L)al; hsGal4/⫹.
REPORTS
18 JUNE 1999 VOL 284 SCIENCE www.sciencemag.org1992
confocal microscopy, CRQ expression was
decreased by 74% in H99 embryos (Fig. 4, B
and E) as compared with wild-type embryos
(Fig. 4, A and D) (37). Some hemocytes in
these embryos do not express detectable lev-
els of CRQ (Fig. 4H). However, after x-ray
irradiation, apoptosis is induced in H99 em-
bryos (34), and CRQ expression increases
(28). This suggests that rpr,grim, and hid
themselves do not regulate CRQ expression
but that the absence of apoptotic corpses
results in CRQ down-regulation. MDP-1 ex-
pression is also down-regulated in H99 em-
bryos (38), suggesting that multiple macro-
phage functions might be activated in the
presence of apoptotic cells.
We tested whether increased apoptosis re-
sulted in increased CRQ expression by sub-
jecting wild-type embryos to x-ray irradiation
(34). In such embryos, giant macrophages
were seen that had engulfed many apoptotic
corpses. In these embryos, macrophages
showed a 3.3-fold increase in CRQ expres-
sion as compared with wild-type embryos
(Fig. 3, C and F) (37). CRQ expression was
similarly up-regulated after treatment of
l(2)mbn cells with ecdysone (39), which in-
duces increased apoptosis and increases the
phagocytic activity in these cells (40). Thus,
signals generated by dying cells cause in-
creased expression of CRQ, which could fa-
cilitate the clearance of the cell corpses. The
expression of the related protein CD36 in
human monocytes is also increased by bind-
ing to one of its ligands, oxidized LDL (41).
This work characterizes a phagocytosis mu-
tant in Drosophila and indicates that the CRQ
protein is necessary, but probably not sufficient,
for efficient phagocytosis of apoptotic cells in
the embryo. Blocking studies on mammalian
macrophages predicted a role for CD36 in the
engulfment of apoptotic cells, and our in vivo
data support this model. Because phagocytosis
of apoptotic cells was not completely abolished
in crq-deficient embryos (Fig. 1F), other recep-
tors are probably involved in this process. Two
other Drosophila macrophage receptors, the
Scavenger Receptor dSR-C1 and Malvolio (42,
43), may share overlapping functions with
CRQ in the engulfment of apoptotic corpses.
However, the rather low efficiency of the resid-
ual phagocytic activity in crq-deficient embryos
implicates the CRQ pathway as a major partic-
ipant in the phagocytosis of apoptotic cells.
CRQ is not required for the phagocytosis of
bacteria by embryonic macrophages, but be-
cause dSR-C1 and Malvolio are similar to mol-
ecules involved in mammalian immune re-
sponses (42,43), they may be specific pattern-
recognition receptors for pathogens.
A genetic dissection of phagocytosis in
Drosophila should further elucidate the
phagocytic pathways for apoptotic corpses
during development and for the engulfment
of pathogens during an immune response. A
greater understanding of the molecular mech-
anisms of both these processes in the fly, as
well as of the macrophage responses they
trigger, is likely to provide insights relevant
to mammalian systems.
References and Notes
1. J. Savill, Br. Med. Bull. 53, 491 (1997).
2. N. Platt, R. Da Silvia, S. Gordon, Trends Cell Biol. 8,
365 (1998).
3. D. T. Fearon and R. M. Locksley, Science 272,50
(1996).
4. N. Franc, K. White, R. Ezekowitz, Curr. Opin. Immunol.
11, 47 (1999).
5. H. Suzuki et al.,Nature 386, 292 (1997).
6. R. Jack et al.,ibid. 389, 742 (1997).
7. Q. Liu and M. Hengartner, Cell 93, 961 (1998).
8. Y. Wu and H. Horvitz, Nature 392, 501 (1998).
9. 㛬㛬㛬㛬 , Cell 93, 951-960 (1998).
10. U. Tepass, L. I. Fessler, A. Aziz, V. Hartenstein, Devel-
opment 120, 1829 (1994).
11. N. C. Franc, J. L. Dimarcq, M. Lagueux, J. Hoffmann,
R. A. Ezekowitz, Immunity 4, 431 (1996).
Fig. 3. Wild-type and
crq-deficient Drosophi-
la embryonic macro-
phages are competent
to recognize and engulf
bacteria and perform
endocytosis. Stage 11
yw and W88 homozy-
gous embryos were in-
jected with TRITC-la-
beled bacteria (32). The
fate of the bacteria was
monitored by confocal
microscopy. (Ato C)
High-magnification im-
ages (⫻100) of a mac-
rophage in a yw inject-
ed embryo taken at
three successive focal
planes. The overlay of
Nomarski and fluores-
cent images shows a
macrophage that has
engulfed several la-
beled bacteria. (D)A
Nomarski image of a magnified view (⫻40) of the head region of a W88 embryo injected with
TRITC-labeled bacteria. In this panel, a very large cell can be seen that has the typical morphology of
a macrophage (arrowhead). (E) As seen at high magnification (⫻400), macrophages in this mutant can
engulf bacteria. (F) A high-magnification image (⫻100) of a macrophage (arrowhead) in a W88 embryo
that had been injected with dioctadecyl tetramethyl indocarbocyanine–labeled acetylated LDL (DiI
AcLDL) (33). Nomarski and fluorescent images of a single focal plane were merged so that the
morphology of a macrophage that had taken up DiI AcLDL (red stain) can be seen.
Fig. 4. Croquemort expression is
regulated by the amount of ap-
optosis. Projected confocal im-
ages of a CRQ immunostaining
in the head of a wild-type em-
bryo (A,D, and G), an H99 ho-
mozygous embryo (B,E, and H),
and an irradiated wild-type em-
bryo (34)(C,F, and I). (A) to (C)
are at ⫻40; (D) to (I) are isolated
macrophages at ⫻400. Images
shown in (A) to (F) were taken
with constant excitation and de-
tection settings to show relative
levels of CRQ staining. (G) to (I)
show overlays of (D) to (F) with
corresponding Nomarski images.
CRQ expression is considerably
down-regulated in H99 homozy-
gous embryos (B and E). In x-
ray–irradiated wild-type embry-
os, the amount of apoptosis is
considerably increased. Macro-
phages become greatly enlarged
as they engulf numerous apoptotic corpses, and the level of CRQ expression in each macrophage
is remarkably up-regulated (C and F). (E) and (H) show three small macrophages side by side, one
of which does not appear to express CRQ (arrowhead). (D), (G), (F), and (I) show single
macrophages.
REPORTS
www.sciencemag.org SCIENCE VOL 284 18 JUNE 1999 1993
12. A. Nicholson, S. Frieda, A. Pearce, R. Silverstein, Ar-
terioscler. Thromb. Vasc. Biol. 15, 269 (1995).
13. G. Endemann et al.,J. Biol. Chem. 268, 11811 (1993).
14. S. Nozaki et al.,J. Clin. Invest. 96, 1859 (1995).
15. J. Savill, I. Dransfield, N. Hogg, C. Haslett, Nature 343,
170 (1990).
16. J. Savill, N. Hogg, Y. Ren, C. Haslett, J. Clin. Invest. 90,
1513 (1992).
17. Y. Ren, R. Silverstein, J. Allen, J. Savill, J. Exp. Med.
181, 1857 (1995).
18. S. W. Ryeom, J. R. Sparrow, R. L. Silverstein, J. Cell Sci.
109, 387 (1996).
19. K. Schneitz, P. Spielmann, M. Noll, Genes Dev. 7, 114
(1993).
20. P. Heitzler, unpublished observations.
21. M. Boedigheimer and A. Laughon, Development 118,
1291 (1993).
22. L. Frank and C. Rushlow, ibid. 122, 1343 (1996).
23. Berkeley Drosophila Genome Project, unpublished
data.
24. Single embryos were squished in 10 lof10mM
tris-HCl (pH8.0), 1 mM EDTA, and 25 mM NaCl
containing Proteinase K (200 g/ml; Boehringer
Mannheim) and incubated at 37°C for 30 min, fol-
lowed by 2 min at 95°C. The PCR was performed with
2.5 U of Taq polymerase and 100 ng of each primer.
Two pairs of primers were designed to amplify 634
base pairs (bp) of the crq genomic sequence and 200
bp of the genomic region of the doom gene as an
internal control [A. J. Harvey, A. P. Bidwaldi, L. K.
Miller, Mol. Cell. Biol. 17, 2835 (1997)]. Crq-specific
primers were 5⬘-TGCCACCGATGCTTGCAGAT-3⬘and
5⬘-AGCCGAATATGATTCCGTACTG-3⬘.Doom-specif-
ic primers were 5⬘-AGGGTAAACGGCCACAGAATGT-
3⬘and 5⬘-GATATCGTTGTAGTTGGCCCG-3⬘. The PCR
cycles were 94°C for 1 min, 65°C for 1 min, and 72°C
for 1 min for 30 cycles. In embryos from the W88
stock, 20 of 79 were missing the crq-specific band.
25. CRQ immunostaining was used to genotype each em-
bryo. Peroxidasin immunostaining detected all hemo-
cytes [R. E. Nelson et al.,EMBO J. 13, 3438 (1994)]. The
nuclear dye 7-AAD labeled all DNA and allowed for the
identification of apoptotic corpses. Unless otherwise
specified, stage 11 to 16 embryos were fixed with
standard procedures (44). Fixed devitellinized embryos
were incubated in phosphate-buffered saline (PBS),
0.0125% saponin, 1% bovine serum albumin, and 4%
normal goat serum (PSN) for 1 hour at room temper-
ature and then incubated with the primary antibodies at
a 1:1000 dilution in PSN overnight at 4°C. After several
washes in PBS, the embryos were incubated for 1 hour
at room temperature with the following secondary
antibodies: fluorescein isothiocyanate–conjugated goat
antibody to mouse and Cy5-conjugated goat antibody
to rabbit ( Jackson Immunoresearch) used at a 1: 1000
dilution in PSN. Finally, embryos were washed three
times in PBS for 20 min and subsequently incubated
with 7-AAD (5 g/ml) in PBS for 30 min. Embryos were
quickly washed twice in PBS, mounted in Vectashield
(Vector), and viewed by confocal microscopy (Leica TCS
NT 4D).
26. The efficiency of engulfment was quantified by
counting the number of engulfed corpses per macro-
phage in at least five fields of four embryos of each
genotype. A P.I., that is, the mean number of engulfed
corpses per macrophage, was calculated for each
embryo, and the mean P.I. was derived for each
genotype.
27. C. Phelps and A. Brand, Methods 14, 367 (1998).
28. N. Franc and K. White, data not shown.
29. Stage 12 to 14 w, UAS-crq;hs-Gal4/⫹and control w;
hs-Gal4 embryos were heat-shocked for 1 hour at
39°C, aged for 2 hours at 25°C, and fixed and em-
bedded in Spurr’s resin (34). Serial 1-m sections of
two embryos of each genotype were stained with a
solution of methylene, toluidine blue, and borax (44)
and viewed by standard microscopy.
30. J. Pugin et al.,Immunity 1, 509 (1994).
31. A. Devitt et al.,Nature 392, 505 (1998).
32. Stage 11 embryos were microinjected with a solution
of PBS and 2 mM NaN
3
containing about 6 ⫻10
8
tetramethyl rhodamine isothiocyanate ( TRITC)–la-
beled fluorescent E. coli (K-12 strain) or S. aureus
(Wood strain) bioparticles (heat-killed bacteria; Mo-
lecular Probes) with standard microinjection proce-
dures (44). After injection, embryos were kept in the
dark at 18°C for 14 to 16 hours, incubated for 1 hour
at 4°C, mounted, and viewed under Nomarski and
fluorescence with a confocal microscope.
33. J. Abrams, A. Lux, H. Steller, M. Krieger, Proc. Natl.
Acad. Sci. U.S.A. 89, 10375 (1992).
34. K. White et al.,Science 264, 677 (1994).
35. M. E. Grether, J. M. Abrams, J. Agapite, K. White, H.
Steller, Genes Dev. 9, 1694 (1995).
36. P. Chen, W. Nordstrom, B. Gish, J. M. Abrams, ibid. 10,
1773 (1996).
37. After CRQ immunostaining, the amount of fluores-
cence seen in five isolated macrophages of each
genotype was quantified with a confocal microscope.
For each macrophage, the fluorescence of serial sec-
tions of 0.5 m was quantified from top to bottom of
the cell. After subtracting the background fluores-
cence, the total amount of fluorescence within each
macrophage was calculated.
38. M. Hortsch et al.,Int. J. Dev. Biol. 42, 33 (1998).
39. J.-L. Dimarcq et al.,Insect Biochem. Mol. Biol. 27, 877
(1997).
40. E. Gateff et al.,inInvertebrate Systems In Vitro,E.
Kurstak, K. Maramorosch, A. Dubendorfer, Eds. (North-
Holland, Elsevier, Amsterdam, 1980), pp. 517–533.
41. L. Nagy, P. Tontonoz, J. Alverez, H. Chen, R. Evans,
Cell 93, 229 (1998).
42. A. Pearson, A. Lux, M. Krieger, Proc. Natl. Acad. Sci.
U.S.A. 92, 4056 (1995).
43. V. Rodrigues, P. Cheah, K. Ray, W. Chia, EMBO J. 14,
3007 (1995).
44. M. Ashburner, Drosophila: A Laboratory Manual (Cold
Spring Harbor Laboratory Press, Cold Spring Harbor,
NY, 1989).
45. We thank J. Fessler and M. Hortsch for providing
antibodies; E. Noll, N. Perrimon, the Bloomington
Stock Center, and L. Dobens for fly stocks; M. Krieger
for the DiI AcLDLs; Y. Ge and W. Fowle for assistance
with confocal microscopy and histology; the lab of T.
Orr-Weaver for advice on nuclear dyes; I. Ando for
suggesting the bacteria assay experiment; and J.
Settleman and the members of the Ezekowitz and
White laboratories for helpful comments on this
work. This work was supported by grants from the
Shiseido Company of Japan to Massachusetts Gener-
al Hospital/Harvard Medical School (N.C.F. and K.W.),
from NIH (K.W. and A.E.), and from the Human
Frontiers in Science Program (A.E. and N.C.F.).
5 February 1999; accepted 12 May 1999
Vessel Cooption, Regression,
and Growth in Tumors
Mediated by Angiopoietins and
VEGF
J. Holash,
1
P. C. Maisonpierre,
1
D. Compton,
1
P. Boland,
1
C. R. Alexander,
1
D. Zagzag,
2
G. D. Yancopoulos,
1
*
S. J. Wiegand
1
*
In contrast with the prevailing view that most tumors and metastases begin as
avascular masses, evidence is presented here that a subset of tumors instead
initially grows by coopting existing host vessels. This coopted host vasculature
does not immediately undergo angiogenesis to support the tumor but instead
regresses, leading to a secondarily avascular tumor and massive tumor cell loss.
Ultimately, however, the remaining tumor is rescued by robust angiogenesis at
the tumor margin. The expression patterns of the angiogenic antagonist an-
giopoietin-2 and of pro-angiogenic vascular endothelial growth factor (VEGF)
suggest that these proteins may be critical regulators of this balance between
vascular regression and growth.
It is widely accepted that most tumors and
metastases originate as small avascular mass-
es that belatedly induce the development of
new blood vessels once they grow to a few
millimeters in size (1–3). Initial avascular
growth would be predicted for tumors that
arise in epithelial structures that are separated
from the underlying vasculature by a base-
ment membrane and for experimental tumors
that are implanted into avascular settings
(such as the cornea pocket) or into a virtual
space (such as the subcutaneum) (2,3). How-
ever, there is also evidence to suggest that
tumors in more natural settings do not always
originate avascularly, particularly when they
arise within or metastasize to vascularized
tissue (4). In such settings, tumor cells may
coopt existing blood vessels (4). The inter-
play between this coopting of existing vessels
and subsequent tumor-induced angiogenesis
has not been extensively examined nor has
the role of angiogenic factors in this process.
The pro-angiogenic vascular endothelial
growth factors (VEGFs) and the angiopoietins
are the only known growth factor families that
are specific for the vascular endothelium be-
cause expression of their receptors is restricted
to these cells (5,6). The angiopoietins include
both receptor activators [angiopoietin-1 (Ang-
1)] and receptor antagonists [angiopoietin-2
1
Regeneron Pharmaceuticals, 777 Old Saw Mill River
Road, Tarrytown, NY 10591, USA.
2
Microvascular and
Molecular Neuro-Oncology Laboratory, Department
of Pathology, Kaplan Cancer Center, New York Uni-
versity Medical Center, New York, NY 10016, USA.
*To whom correspondence should be addressed. E-
mail: gdy@regpha.com (G.D.Y.); stan.wiegand@regpha.
com (S.J.W.)
REPORTS
18 JUNE 1999 VOL 284 SCIENCE www.sciencemag.org1994