Proc. Nati. Acad. Sci. USA
Vol. 91, pp. 4082-4085, April 1994
Thalidomide is an inhibitor of angiogenesis
(fibroblast growth factor/rabbit cornea)
ROBERT J. D'AMATO*, MICHAEL S. LOUGHNAN, EVELYN FLYNN, AND JUDAH FOLKMAN
Department of Surgery, Children's Hospital, Harvard Medical School, Boston, MA 02115
Communicated by John A. Glomset, January 3, 1994
dysmelia (stunted limb growth) in humans. We have demon-
strated that orally administered thalidomide is an inhibitor of
aniogenesis induced by basic fibroblast growth factor in a
rabbit cornea micropocket assay. Experiments including the
analysis of thalidomide analogs revealed that the antanio-
genic activity correlated withtheteratogenicity butnotwith the
sedative or the mild immunosuppressive properties of tha-
lidomide. Electron microscopic examination of the corneal
neovascularization ofthafidomide-treated rabbits revealed spe-
cific ultrastructural changes similar to those seen in the de-
formed limb bud vasculature of thalidomide-treated embryos.
These experiments shed light on the mechanism of tha-
lidomide'steratogenicity and hold promise for the potential use
ofthalidomide asan orally adminsered drug for the treatment
of many diverse dease dependent on angiogenesis.
Thalidomide is a potent teratogen causing
Thalidomide is a potent teratogen. It was developed by
Chemie Grunenthal in the 1950s as a sedative that appeared
so nontoxic in rodent models that a LD50 could not be
established. In 1961, McBride (1) and Lenz (2) described the
association between limb defects in babies and maternal
thalidomide usage. Although humans are exquisitely sensi-
tive to the teratogenic effects ofthalidomide, experiments in
rodents failed to reveal similar effects (3, 4). Teratogenic
effects could be experimentally reproduced by the adminis-
tration of thalidomide to pregnant rabbits at an oral dose of
100-300 mg per kg per day (5, 6). Over the past 30 years, the
mechanism of thalidomide's teratogenicity has been exten-
sively studied but has remained unsolved (7).
We now postulate that the limb defects seen with thali-
domide were secondary to an inhibition of blood vessel
growth in the developing fetal limb bud. The limb bud is
unique in requiring a complex interaction of both angiogen-
esis and vasculogenesis during development (8). Vasculo-
genesis is the formation of a capillary bed from endothelial
cells that have differentiated from mesenchymal precursors.
Angiogenesis is the formation of new blood vessels from
sprouts ofpreexisting vessels. Therefore, the limbbudwould
be a particularly vulnerable target to a teratogen that inhib-
ited endothelial cell function. We chose to examine the effect
of thalidomide on growing vasculature in the chicken chori-
oallantoic membrane and in the rabbit cornea.
MATERIALS AND METHODS
Chicken chorioallantoic membrane (CAM) assays were per-
formed as described (9, 10) and the effects on the developing
vasculature were recorded at 48 h after implantation of the
0.5% carboxymethylcellulose pellet containing various
drugs. Corneal neovascularization was induced by an im-
planted pellet of poly(hydroxyethyl methacrylate) (Hydron;
Interferon Sciences, New Brunswick, NJ) containing 650 ng
(bFGF) bound to sucralfate (sucrose aluminum sulfate; Bukh
Meditec, Copenhagen) (11). The addition of sucralfate to the
pellet protects the bFGF from degradation (12) and provides
for its slow release, thus producing consistent aggressive
angiogenesis that is more pronounced than that induced by
bFGF/Hydron alone. Release ofbFGF from pellets contain-
ing sucralfate/Hydron could be detected in vitro for up to 4
days after the pellets were formed compared tojust 1 day for
pellets with Hydron alone (11). Pellets were made by mixing
110 pI of saline containing 12 Mgof recombinant bFGF
(Takeda, Osaka) with 40 mg of sucralfate; this suspension
was added to 80 pI of 12% (wt/vol) Hydron in ethanol.
Aliquots (10 Al) of this mixture were then pipetted onto
Teflon pegs and allowed to dry producing approximately 17
pellets. A pellet was implanted into corneal micropockets of
each eye of an anesthetized female New Zealand White
rabbit, 2 mm from the limbus, followed by a single topical
application of erythromycin ointment on the surface of the
cornea. Histologic examination on consecutive days demon-
strated progressive blood vessel growth into the cornea
toward the pellet with only rare inflammatory cells seen. This
angiogenic response was not altered by severe immune
suppression with total body irradiation, and pellets with
sucralfate alone did not induce angiogenesis (data not
shown). Unlike other models of corneal angiogenesis that
utilize inflammation to stimulate neovascularization, the new
vessels are primarily inducedbythebFGF. The animals were
fed dailyfrom 2 days afterimplantationby gastric lavage with
either drug suspended in 0.5% carboxymethylcellulose or
vehicle alone. Thalidomide was purchased from Andrulus
Pharmaceutical (Beltsville, MD) and EM-12 and Supidimide
were kindly provided by Grunenthal (Stolberg, F.R.G.).
Immunosuppressed animals received total body radiation of
6 Gy for 6 min immediately prior to implantation of the
pellets. This dose ofradiation resulted in a marked leukocy-
topenia with >80%o reduction in the leukocyte count by day
2 and >90o reduction by day 3, results that are consistent
with previous reports (13, 14).
The animals were examined with a slit lamp every other
day in a masked manner by the same corneal specialist
(M.S.L.). The area ofcorneal neovascularization was deter-
mined by measuring with a reticule the vessel length (L) from
the limbus and the number of clock hours (C) of limbus
involved. A formula was used to determine the area of a
circular band segment: C/12 x 3.1416 [r2 - (r - L)2], where
r = 6 mm, the measured radius ofthe rabbit cornea. We have
utilized various mathematical models to determine the
amount of vascularized cornea and have found that this
formulaprovides the most accurate approximation ofthe area
of the band of neovascularization that grows toward the
pellet. Only the uniform contiguous band of neovasculariza-
Abbreviations: bFGF, basic fibroblast growth factor; CAM, chicken
chorioallantoic membrane; PGA, phthaloylglutamic anhydride; PG
acid, phthaloylglutamic acid; TNF-a, tumor necrosis factor a.
*To whom reprint requests should be addressed.
The publication costs ofthis article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement"
in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Proc. Natl. Acad. Sci. USA 91 (1994)
tion adjacent to the pellet was measured. The noncontiguous
neovascularization, which can be seen superiorly, was not
quantified due to its irregular shape. These vessels that often
grow concurrently toward the pellet from the superior limbus
arise from vessels ofthe superior rectus supplying the limbus,
are directly induced by the bFGF/sucralfate pellet, and are
histologically identical to the inferior limbal vessels. How-
ever, it should be noted that this superior neovascularization
was commonly seen in control animals and was never seen in
thalidomide-treated animals. Thus, the total inhibition of
neovascularization is conservatively underestimated.
Our initial investigations were performed on the CAM.
Neither thalidomide nor EM-12, a related teratogenic analog
(15), exhibited any inhibitory activity onblood vesselgrowth.
This result was expected as it has been proposed that
thalidomide must be metabolized by the liver to form an
epoxide that may be the active teratogenic metabolite (16).
Other thalidomide analogs that have been shown to be
teratogenic in rodents (17), including phthaloylglutamic an-
hydride (PGA) and phthaloylglutamic acid (PG acid), were
also analyzed (Fig. 1). Interestingly, weak antiangiogenic
activity ofthe developing vasculature was seen with both PG
acid and PGA when 100jigof either compound was placed
on the CAM in a pellet of 0.5% carboxymethylcellulose.
Despite frequent mild scarring, avascular zones were pro-
duced in 15% of the CAMs with PGA compared to control
0.5% carboxymethylcellulose pellets in which no avascular
zones were seen (data not shown).
Based on these initial findings, we decided to test tha-
lidomide's effect on angiogenesis induced by bFGF in the
rabbit corneal micropocket model. Treatment with a terato-
Structure of thalidomide and related analogs.
genic dose (200 mg/kg) of thalidomide resulted in an inhibi-
tion ofthe area ofvascularized cornea that ranged from 30 to
51% in three experiments with a median inhibition of 36%
(Figs. 2A and 3) (n = 30 eyes; P = 0.0001, two-way ANOVA
with ranked data). The inhibition of angiogenesis by tha-
lidomide was seen after only two doses (Fig. 2B). The rabbits
did not demonstrate obvious sedation and there were no signs
of toxicity or weight loss. The teratogenic analog EM-12,
which shares the other properties of thalidomide, was also
inhibitory, with a median inhibition of 42% (n = 10 eyes; P
= 0.002, one-way ANOVA with ranked data). Supidimide, a
nonteratogenic analog that retains the sedative properties of
thalidomide, exhibited no activity (area 107% ofcontrol; n =
10 eyes; not statistically different from control). Other ana-
logs, PGA and PG acid, displayed weaker inhibitory effects
Animal treatment groups
tion by thalidomide and related analogs expressed as percent of
median control on day 8. Pellets containing bFGF and sucralfate
were implanted into micropockets of both corneas of rabbits (18).
Vessel ingrowth into the clear corneafrom the limbus was first noted
on day 2 and treatments (200 mg/kg orally) were begun on this day.
The area of corneal neovascularization was measured from day 4
through day 12. Day 8 measurements were used for comparison
between groups. No regression ofvessels and near maximal neovas-
cularization was seen at this time point. Statistical analysis was
performed with ANOVA with ranked data to account for interex-
perimental variation and to guard against anonnormal distribution of
data (i.e., outliers) by utilizing a nonparametric method. (b) Time
course of inhibition of neovascularization with thalidomide. Mean
areas of corneal neovascularization with standard error bars are
presented from one experiment that is representative of the three
experiments performed with thalidomide on nonirradiated animals.
Data presented from the first time point after administration of the
drug through the completion ofthe study are statistically different (n
= 10 eyes; P < 0.005 for all time points, one-way ANOVA with
(A) Inhibition of bFGF-induced corneal neovasculariza-
Medical Sciences: D'Amato et al.
Medical Sciences: D'Amato et al.
than thalidomide (data not shown). The density of vessel
ingrowth in thalidomide-treated animals was also markedly
reduced. Due to the lack ofan objective grading scale, these
results are not presented.
Thalidomide has immunosuppressive properties that might
have indirectly affected angiogenesis. Recently, thalidomide
has been used for its immunosuppressive properties in the
treatment ofleproma reactions (19) and chronic graft versus
host disease (18, 20-23). However, in humans its immuno-
suppressive properties are weak and delayed with little effect
in acute graft versus host disease (24). Because the effect of
thalidomide on the immune system is similar butweakerthan
that of cyclosporin A (25), we tested cyclosporin A at the
highest tolerated dose of 25 mg/kg. No statistically signifi-
cant effect was observed compared to control. To investigate
Proc. Natl. Acad. Sci. USA 91 (1994)
further the immune interactions, we pretreated the rabbits
with the maximally tolerated immunosuppressive dose of
total body irradiation. Immunosuppressed animals re-
sponded equally well tothalidomide, withamedian inhibition
of neovascularization of52% (n= 12; P = 0.0001, one-way
ANOVA with ranked data) as compared to irradiated place-
bo-treated controls (Fig. 2A).
Electron microscopic examination of corneas from tha-
lidomide-treated and control animals revealed ultrastructural
differences. Vessels in the thalidomide-treated group dem-
bFGF pellets from control (A) and thalidomide-treated (B) rabbits.
There is prominent corneal neovascularization (arrows) in the con-
trol with associated corneal clouding, which was demonstrated
histologically to be stromal edema without inflammation. The tha-
lidomide-treated animal has markedly less neovascularization with
minimal corneal edema.
Representative corneas at 8 days after implantation of
served in a thalidomide-treated rabbit 10 days after implntation of
a pellet containing bFGF. (A) High-magnification (x40,000) view of
typical fenestrations (arrow) in an endothelial cell from corneal
neovascularization in thalidomide-treated rabbit. (B) High-
magnification (x60,000) view of an area of cell thinning (asterisk)
adjacent to a celljunction in thalidomide-treated corneal neovascu-
larization. These changes were not seen in control day 10 corneal
neovascularization. (Bars = 0.1 amn.)
Electron micrographs of corneal neovascularization ob-
f..... ;7 .1
Proc. Natl. Acad. Sci. USA 91 (1994) Download full-text
onstrated fenestrations not seen in control animals (Fig. 4A).
Fenestrations have been previously reported to be specific to
regressing corneal blood vessels after removal of the angio-
genic stimulus (26). However, in that model, endothelial cell
regression was associated with platelet plugging and cellular
hypoxic changes such as swollen mitochondria, which were
not seen in the thalidomide-treated animals. Interestingly,
histologic changes previously described in the vasculature of
the limb budsfrom chicken embryos treated with thalidomide
(27) were also seen in the corneal neovascularization of our
thalidomide-treated rabbits including vesicular projections
into the lumen and extreme thinning of cell processes (Fig.
4B). In general, the corneal neovascularization from thali-
domide-treated rabbits appeared more immature than that
observed in control animals with poorly formed cell junc-
tions, incomplete basement membrane, and fewer associated
pericytes. These findings support the hypothesis that tha-
lidomide has a direct effect on the growing vasculature.
Orally administered thalidomide is an inhibitor of angiogen-
esis induced by bFGF in the rabbit cornea micropocket
assay. The mechanism by which thalidomide inhibits angio-
genesis is unknown. Thalidomide has shown no effect on
bFGF-induced proliferation of endothelial cells in culture
(data not shown). Current studies are focused on the identi-
fication of the most active thalidomide metabolite. The
formation ofan active metabolite by the liver in vivo provides
an explanation of the observation that the effect of tha-
lidomide on growing vessels is seen only when given sys-
Thalidomide has been shown to suppress tumor necrosis
factor a (TNF-a) production from macrophages (28). How-
ever, macrophages were rarely seen in histologic examina-
tions of our model of corneal neovascularization. Further-
more, studies examining the role ofTNF-a in corneal angio-
genesis have failed to detect TNF-a production in a model of
inflammatory corneal angiogenesis in which macrophages
were prominent (29). TNF-a is only weakly angiogenic in
vivo. It acts by inducing secondary inflammation in contrast
to bFGF, which stimulates angiogenesis without inflamma-
tion (30). Thus, the ability of thalidomide to inhibit angio-
genesis induced by pharmacologic doses of bFGF supports
the hypothesis that thalidomide directly inhibits an essential
component of angiogenesis and does not operate through
effects on TNF-a production.
In conclusion, thalidomide is a potent angiogenesis inhib-
itor in vivo. In this model of corneal angiogenesis, we have
tested many putative angiogenesis inhibitors (including an-
timitotic agents, cis-retinoic acid, tamoxifen, and others).
Thalidomide was the only agent capable of inhibiting angio-
genesis after oral administration. Evaluation of thalidomide
analogs demonstrated a correlation between teratogenicity
and antiangiogenic potential. The weak and delayed immu-
nosuppressive action ofthalidomide when used clinically, its
inhibition of angiogenesis in radiation-immunosuppressed
animals, and the lack of effect of the functionally related
immunosuppressive agent cyclosporin A argue for a direct
effect of thalidomide on angiogenesis. Further support for
this hypothesis is derived from the ultrastructural changes
seen in thalidomide-treated animals. There are clear impli-
cations for the use of this drug in the treatment ofpathologic
angiogenesis that occurs in diabetic retinopathy, macular
degeneration, and solid tumors. Because antiangiogenic ther-
apy is likely to be long-term, there is a need for an orally
Special thanks to Klio Chatzistefanou, Geri Jackson, Evelyn
Gonzalez, Pat D'Amore, Helene Sage, Michael O'Reilly, Michael
Kaplan, Tony Adamis, Elizabeth N. Allred, Ramzi Cotran, and
Dianna Ausprunk for their assistance and advice. We also thank E.
Frankus and K. Zwingenberger ofGrunenthal GMBH for providing
technical information, EM-12, and supidimide. R.J.D. is a Howard
Hughes Medical Institute physician research fellow. M.S.L. is partly
supported by the Ruth Rae Davidson cornealfellowship endowment.
Animal studies were reviewed and approved by the animal care and
use committee ofChildren's Hospital and are in accordance with the
guidelines of the Department of Health and Human Services.
McBride, W. G. (1961) Lancet u1, 1358.
Lenz, W. (1962) Lancet 1, 45.
Cahen, R., Sautai, M., Montagne, J. & Pessonier, J. (1964)
Med. Exp. 10, 201-224.
Christie, G. A. (1962) Lancet ii, 249.
Helm, F. (1966) Arzneim. Forsch. 16, 1232-1237.
Stertz, H., Nothduft, H., Lexa, P. & Ockenfels, H. (1987)
Arch. Toxicol. 60, 376-381.
Stepens, T. D. (1988) Teratology 38, 229-239.
Seifert, R., Zhao, B., Christ, B. (1992) Anat. Embryol. 186,
Auerbach, R., Kubai, L., Knighton, D. & Folkman, J. (1974)
Dev. Biol. 41, 391-394.
Crum, R., Szabo, S. & Folkman, J. (1985) Science 230, 1375-
Gonzalez, E., Adamis, T. & Folkman, J. (1994) Invest. Oph-
thalmol. Vis. Sci. 33 Suppl., 424.
Folkman, J., Szabo, S., Stovroff, M., McNeil, P., Li, W. &
Shing, Y. (1991) Ann. Surg. 214, 414-427.
Bogman, M., Cornelissen, I., Berden, J., De Jong, J. & Koene,
R. (1984) J. Immunol. Methods 70, 31-38.
Bigelow, C., Adler, L., Appelbaum, F. (1987) Transplantation
Helm, F., Frankus, E., Friderichs, E., Graudums, I. & Flohe,
L. (1981) Arzneim. Forsch. Drug Res. 31, 941-949.
Blake, D. A. & Balasubramanian, V. (1981) Proc. Natl. Acad.
Sci. USA 78, 2545-2548.
Kocher-Becker, U., Kocher, W. & Ockenfels, H. (1984) Z.
Naturforsch. 36, 904-909.
Vogelsang, G. B., Taylor, S., Gordon, G. & Hess, A. D. (1986)
Transp. Proc. 18, 904-906.
Sheskin, J. (1965) Clin. Pharm. Ther. 6, 303-306.
Field, E. O., Gibbs, J. E., Tucker, D. F. & Hellman, K. (1966)
Nature (London) 211, 1308-1313.
Lim, S. H., McWannell, A., Vora, A. J. & Boughton, B. J.
(1988) Lancet 1, 117.
McCarthy, D. M., Kanfer, E. J. & Barrett, A. J. (1987)
Biomed. Pharmacother. 43, 693-697.
Saurat, J. H., Camenzind, M., Helg, C. & Chapuis, B. (1988)
Lancet 1, 359.
McCarthy, D. M., Kanfer, E. J. & Barrett, A. J. (1989)
Biomed. Pharmacother. 43, 693-697.
Keenan, R., Eiras, G., Burckart, G., Stuart, R., Hardesty, R.,
Vogelsang, G., Griffith, B. & Zeevi, A. (1991) Transplantation
Ausprunk, D., Falterman, K. & Folkman, J. (1978)Lab. Invest.
Jurand, A. (1966) J. Embryol. Exp. Morphol. 16, 289-300.
Sampaio, E., Sarno, E., Galilly, R., Cohn, Z. & Kaplan, G.
(1991) J. Exp. Med. 173, 699-703.
Sunderkotter, C., Roth, J. & Sorg, C. (1990)Am. J. Pathol. 137,
Frater-Schroder, M., Risau, W., Hallman, R., Gautschi, P. &
Bohlen, P. (1987) Proc. Natl. Acad. Sci USA 84, 5277-5281.
Medical Sciences: D'Amato et al.