You are on page 1of 6

© 2000 Nature America Inc. • http://medicine.nature.

com
ARTICLES

Suppression of tumor growth through disruption of


hypoxia-inducible transcription

ANDREW L. KUNG, STREAM WANG, JEFFERY M. KLCO,


WILLIAM G. KAELIN, JR. & DAVID M. LIVINGSTON

The Dana-Farber Cancer Institute and Harvard Medical School, 44 Binney Street, Boston, Massachusetts 02115,USA
Correspondence should be addressed to D.M.L.; email: david_livingston@dfci.harvard.edu

Chronic hypoxia, a hallmark of many tumors, is associated with angiogenesis and tumor pro-
gression. Strategies to treat tumors have been developed in which tumor cells are targeted
with drugs or gene-therapy vectors specifically activated under hypoxic conditions. Here we
report a different approach, in which the normal transcriptional response to hypoxia is selec-
tively disrupted. Our data indicate that specific blockade of the interaction of hypoxia-in-
ducible factor with the CH1 domain of its p300 and CREB binding protein transcriptional
© 2000 Nature America Inc. • http://medicine.nature.com

coactivators leads to attenuation of hypoxia-inducible gene expression and diminution of


tumor growth. Thus, disrupting the normal co-activational response to hypoxia may be a new
and useful therapeutic strategy.

Cellular hypoxia triggers a multifaceted adaptive response transcriptional co-activators is a likely point of significant re-
that is primarily mediated by the heterodimeric transcription sponse amplification, we asked whether disruption of HIF-1α
factor hypoxia-inducible factor1 (HIF). Under normoxic con- interaction with p300/CBP could serve as a point of interven-
ditions, the α-subunit of the heterodimer (HIF-1α) is essen- tion in the hypoxia-response pathway.
tially undetectable2 due to its destruction by the
ubiquitin-proteosome system3,4. Ubiquitination of HIF-1α is Structural requirements for p300/CBP interaction with HIF-1α.
mediated by the von Hippel-Lindau tumor-suppressor pro- We mapped the detailed structural requirements for p300/CBP
tein5–7. Under hypoxic conditions, through as yet unclear binding to HIF-1α (Fig. 1a) by in vitro assays using a series of
mechanisms, HIF-1α stabilizes and accumulates2,8. HIF-1α het- glutathione S-transferase (GST)–fused truncation mutants cen-
erodimerizes with its constitutively expressed binding part- tered around the CH1 domain of p300, the primary interact-
ner, the aryl hydrocarbon receptor nuclear translocator9 ing unit of that protein10,12. We assayed these mutant fusion
(ARNT), a common binding partner of several bHLH-PAS do- proteins for binding to full-length, in vitro translated HIF-1α,
main proteins. The heterodimer interacts with p300/CBP(ref. which was found to require zinc in pilot experiments (Fig. 1b).
10–13) and SRC-1 family co-activators 13, binds to a cognate The minimal HIF-1α–binding domain encompassed 116
hypoxia-response element 9,14 (HRE) and thereby transacti- residues of p300 (aa 302–418), which included the CH1 do-
vates HRE-containing promoters and enhancers. Expression main and some flanking sequences (Fig. 1a and c). To map the
of HIF-1 target genes serves to maintain cellular homeostasis minimal p300-binding domain of HIF-1α, we generated a se-
at least in part by promoting anaerobic glycolysis, facilitating ries of Gal4-fused truncation mutants of the carboxyl-termi-
erythropoiesis and increasing blood delivery through vasodi- nus because this region had been shown previously to contain
latation and angiogenesis1. the p300/CBP interacting domain10,12. We translated these
As tumors evolve from a single malignant cell into a multi- Gal4-fused HIF-1α truncation mutants in vitro and probed
cellular mass, oxygen tension in the tumor microenvironment them for their ability to bind to GST–CH1 (aa 302–443). The
drops as the passive diffusional capacity of the existing blood C-terminal 41 residues of HIF-1α (aa 786–826), also known as
supply is surpassed15,16. Tumor hypoxia decreases the efficacy the C-terminal transactivation (TAD-C) domain25, were neces-
of many common therapies17, is a powerful stimulus for angio- sary and sufficient for binding to GST–CH1 (Fig. 1a and d).
genesis18 and is linked to tumor progression19,20 and metasta- To further study the requirements for HIF-1α/CH1 binding,
sis21. The importance of the HIF-1 response pathway in human we generated a series of GST–CH1 (aa 302–528) internal sub-
tumorigenesis is underscored by the finding that HIF-1α is stitution mutants. These serial, non-overlapping mutants
overexpressed in multiple human cancers19. Because tumor spanned residues 341–424 of human p300. In each mutant,
cells, unlike normal cells from the same tissue, are often we replaced five residues with the sequence Asn-Ala-Ala-Ile-
chronically hypoxic19, therapeutic strategies have been devel- Arg-Ser (NAAIRS), a flexible linker that can assume an α-heli-
oped that seek to exploit this difference by using drugs22 or cal or β-strand conformation26, and thus decreased the
gene-therapy vectors23,24 that are selectively activated under likelihood of generalized conformational disruption as a re-
hypoxic conditions. sult of the substitution. Almost all substitutions abolished
We investigated whether hypoxic tumor cells could be tar- CH1 binding to full-length HIF-1α (Fig. 1e), indicating that
geted by disrupting the normal homeostatic response to hy- the CH1 domain must assume a complex, zinc-coordinated
poxia. Specifically, because interaction of HIF-1 with conformation in order to interact with HIF-1α.

NATURE MEDICINE • VOLUME 6 • NUMBER 12 • DECEMBER 2000 1335


© 2000 Nature America Inc. • http://medicine.nature.com
ARTICLES

a a 786 826 d d Gal4-HIF


Gal4 766-826 786-826 796-826 801-826 786-822 786-818
HIF-1α

Bound

Bound

Bound

Bound

Bound

Bound

Bound
Input

Input

Input

Input

Input

Input

Input
bHLH PAS TAD-N TAD-C

302 418
p300
N CH1 KIX Bromo CH2 CH3 Q-rich

b b OP
c c
e e
GST-CH1 (302-528) NAAIRS substitution
GST-CH1

GST-CH1

1/5 Input
341-346

353-358

365-370

383-388

395-400

407-412

419-424
347-352

359-364

371-376
377-382

389-394

401-406

413-418
1/5 input
1/5 input

302-468

302-418

302-406

355-528
Zn 302-528

302-443

302-413

328-528

340-528
Native

GST
1 µM

GST
0

IVT IVT
HIF-1α HIF-1α

Fig. 1 Structural requirements for HIF-1α interaction with p300/CBP. mutants of GST–CH1, with amino acid boundaries as indicated, were as-
a, Schematic representations of HIF-1α and p300. The amino acid num- sayed for their ability to bind to full-length IVT HIF-1α. d, The indicated
bers indicated for each protein define the boundaries of the minimal in- C-terminal HIF-1α truncation mutants were fused to Gal4 and translated
teraction domains. b, Zinc is required for GST–CH1 interaction with in vitro. Binding assays were performed using GST–CH1 (aa 302–443).
HIF-1α. GST–CH1 (aa 302–528) was assayed for its ability to bind in 20% of the total IVT product used for each binding assay is indicated
vitro translated (IVT) full-length HIF-1α in its native state after purifica- (Input), along with the amount bound by GST–CH1 (Bound). e, Serial
tion from bacteria (Native) or after washing with 1,10-phenanthroline substitution mutants of GST–CH1 (aa 302–528) were generated by re-
(OP), a heavy metal chelator. The addition of 1µM zinc sulfate after placing the indicated amino acids with the sequence “NAAIRS,” and as-
© 2000 Nature America Inc. • http://medicine.nature.com

chelation fully restored binding activity. c, C- and N-terminal truncation sayed for their ability to bind to IVT full-length HIF-1α.

Suppression of hypoxia-inducible transcription attenuation by CH1-containing polypeptides (data not shown).


Overproduction of polypeptides corresponding to the HIF- Accordingly, subsequent experiments focused on the activity of
1α/TAD-C or p300/CH1 minimal binding domains led to attenu- a wild-type TAD-C polypeptide (aa 786–826 of HIF-1α) compared
ation of hypoxia-inducible reporter activity (Fig. 2a). In the case with that of an eight-residue truncated mutant (aa 786–818),
of TAD-C polypeptides, these were expressed as fusions to Gal4 which failed to bind to GST–CH1 in vitro (Fig. 1d) and failed to
which served to stabilize the polypeptides, as well as provide a exert a dominant-negative effect on hypoxia-inducible reporter
means of detection with antibodies against Gal4. Only those activity in vivo (Fig. 2a).
TAD-C and CH1 mutants that bound in vitro to p300 and HIF-1α, We assayed the interaction of Gal4-CH1 (aa 300–528) with
respectively (Fig. 1c and d), had a dominant-negative effect on VP16–HIF-1α (aa 723–826) in a mammalian two-hybrid assay to
hypoxia-inducible reporter activity in vivo (Fig. 2a). As an addi- assess the effect of TAD-C polypeptide overexpression on the in-
tional control, we observed that expression of an unrelated teraction of HIF-1α with the CH1 domain of p300 in vivo. Co-ex-
transactivational domain (VP16) had no effect on hypoxia-in- pression of TAD-C polypeptide recombined into an active
ducible reporter activity (Fig. 2a). In titration experiments, TAD- center of thioredoxin resulted in disruption of the HIF-1α inter-
C polypeptides were more potent suppressors of HIF function action with CH1 under normal as well as hypoxic conditions,
than CH1-containing dominant-negative polypeptides (data not whereas the truncated mutant polypeptide had no such effect
shown), perhaps because of the imputed conformational com- (Fig. 2b). Similar analysis of the interaction of p53 and SV40
plexity of CH1. At maximal effect, overexpression of TAD-C large T antigen, which served as an unrelated control, was not
polypeptides resulted in 80–90% attenuation of hypoxia-in- affected by either TAD-C or the mutant polypeptide (Fig. 2c).
ducible reporter activity compared with a maximum of 60–70% These data demonstrate that overexpression of TAD-C polypep-

Fig. 2 In vivo effects of CH1 or a 180 a b 250 c 250 b c


Epo-Luc Activity ( RLU )

TAD-C polypeptide overexpression. 160 Normoxia


Gal4-Luc Activity ( RLU )

Gal4-Luc Activity ( RLU)

200 Normoxia 200


140 Hypoxia
a, Cells were cotransfected with 100 120 150
Hypoxia
150
ng of the hypoxia-responsive re- 100
100
80
porter, Epo-Luciferase; 100 ng of an 100
60 50
internal control, pCMV-LacZ; and 40 50
0
300 ng of the expression plasmids 20 Gal4 +
0
0 VP16 +
indicated along the abscissa of the
Trx-MUT
Trx

Trx-TADC
pCMX-VP16

Gal4-CH1 + + + +
Vector

Vector
766-826
786-826

791-826
796-826

801-826

786-818

302-493

302-443

302-418

302-413

328-443
355-443

VP16-HIF + + + +
graph. Cells were subsequently ex- Trx + + +
Trx-TADC +
posed to either ambient (Normoxia, pCMX-Gal4-TADC pCMV-CH1
Trx-MUT +

) or a 1% (Hypoxia,) oxygen for


18 h before luciferase and β-gal as-
says. b, Cells were cotransfected with the indicated expression vectors. These plasmids encoded the Gal4 DNA-binding domain, alone (Gal4) or Gal4 fused to CH1
(aa 300–528, Gal4–CH1, 40 ng); the VP16 transactivation domain, alone (VP16), or VP16 fused to HIF-1α (aa 723–826, VP16-HIF, 40 ng); and thioredoxin alone
(Trx) or Trx fused to TAD-C (aa 786–826, Trx-TADC) or a mutant derivative (aa 786–818, Trx-MUT) TAD-C polypeptide (200 ng). Present in all co-transfection mix-
tures were a Gal4-Luciferase reporter (50 ng) and pCMV-LacZ (100 ng). Exposure to hypoxia or normoxia were as described above. c, Cells were co-transfected
with expression vectors encoding Gal4 DNA-binding domain fused to p53 (pM-53, 40 ng); the VP16 transactivation domain fused to SV40 large T antigen (pVP16-
T, 40 ng); and Gal4-Luciferase, pCMV-LacZ and the indicated thioredoxin expression vectors as described above. In all cases, results are expressed as corrected RLU
(mean of three independent transfections ± s.d.).

1336 NATURE MEDICINE • VOLUME 6 • NUMBER 12 • DECEMBER 2000


© 2000 Nature America Inc. • http://medicine.nature.com
ARTICLES

Fig. 3 Specificity of TAD-C polypeptide effects. a, Effect of TAD-C expres- a


sion on p300/CBP-dependent transactivation. Cells were co-transfected
a 80
Trx

Gal4-Luc Activity ( RLU )


70
with the indicated thioredoxin-constrained (Trx) expression plasmids (300 60
Trx-TADC
ng), the indicated Gal4 DNA-binding-domain–fused proteins (20 ng), Gal4- Trx-MUT
50 Hypoxia
luciferase reporter (100 ng), and pCMV-LacZ (50 ng). Where indicated, 40
cells were exposed to hypoxia for 18 h. Results are expressed as corrected 30
RLU (mean of three independent transfections ± s.d.). b, HepG2 cells were 20
infected at a MOI of 10 with retroviruses encoding Gal4 alone or Gal4 fused 10
to wild-type (Gal4-TADC) or mutant (Gal4-MUT) TAD-C polypeptide. Cells 0
were exposed to either ambient or 1% (Hypoxia) oxygen for 18 h prior to Gal4 Gal4- Gal4- Gal4- Gal4- Gal4- Gal4-
SRC1 SREBP RAR STAT2 HIF-1 α HIF-1 α
preparation of total RNA. VEGF and L32 (housekeeping control) mRNA
abundance was determined by ribonuclease protection assay. c, HepG2
cells were infected with retrovirus encoding the indicated Gal4 fusion pro- bb Hypoxia c c 30
teins. Cells were exposed to either ambient or 1% oxygen, then assayed for Normoxia *
25

Gal4-TADC

Gal4-TADC

Cell death (%)


Hypoxia

Gal4-MUT

Gal4-MUT
cell death by Annexin V staining and vital dye exclusion. Results are ex- 20
pressed as the mean of three independent experiments ± s.e.m. * indicates 15
statistical significance (P ≤ 0.05) in comparison with controls.

Gal4

Gal4
10
5
VEGF
0
tide disrupts the interaction of HIF-1α with the CH1 domain of L32 Gal4 Gal4- Gal4-
MUT TADC
p300 in vivo.
In comparison with its effects on HIF-1α–mediated tran-
© 2000 Nature America Inc. • http://medicine.nature.com

scriptional activity, TAD-C polypeptide overexpression had no pression of Gal4 and the Gal4-fused polypeptides in this and
effect on the transcription function of RAR, SREBP2 or SRC-1, related experiments (for example, see Fig. 4a inset).
which interact with the amino-terminal27, KIX28 and gluta- Concordant with the reporter assays noted above, VEGF
mine-rich29 domains of p300/CBP, respectively (Fig. 3a). mRNA abundance was reproducibly reduced in cells express-
However, TAD-C inhibited the transcriptional activity of ing wild-type but not mutant TAD-C polypeptide, under both
STAT2, which, like HIF-1α, uses CH1 to mediate its transcrip- normal and hypoxic conditions (Fig. 3b). The overall abun-
tional coactivation10,30. These results indicate that an intact dance of HIF-1α protein under hypoxic conditions did not
TAD-C polypeptide binds and blocks the access of multiple change with introduction of TAD-C polypeptide (data not
proteins to the CH1 domain without inducing a generalized shown).
suppressive effect on other p300/CBP functions. Expression of TAD-C polypeptide had no detrimental effect
We engineered VSV-G–pseudotyped retroviruses to express on cell survival under normal oxygen conditions, in compari-
unfused Gal4 carrier protein or Gal4 fused to either wild-type son with normoxic cells expressing Gal4 alone or Gal4 fused
or mutant TAD-C polypeptide to facilitate expression of the to mutant TAD-C polypeptide (Fig. 3c). Cell proliferation was
relevant TAD-C and mutant polypeptides. We infected also not affected under normoxic conditions because the dou-
human hepatoma HepG2 cells with these retroviruses at a bling time of cells expressing either TAD-C or mutant polypep-
multiplicity of infection (MOI) of 10, exposed them to either tide fused to Gal4 was identical to that of cells expressing Gal4
normal or hypoxic conditions and determined their vascular alone (data not shown). Under hypoxic conditions, however,
endothelial growth factor (VEGF) mRNA levels with ribonu- cell death increased in cells expressing wild-type TAD-C
clease protection assay. We regularly achieved equivalent ex- polypeptide in comparison with cells expressing Gal4 alone or

a c d
Gal4-TADC

3000
Gal4-TADC
Gal4-MUT

Mix (Input)
Gal4-MUT
Gal4-TADC

Tumor 1
Tumor 2
Tumor 3
Gal4-MUT

2500
Tumor Volume ( mm 3 )

Gal4
Gal4

2000

1500 Gal4-TADC
p53
Gal4-MUT
1000

500

0
0 1 2 3
Week Fig. 4 Anti-tumor effects of TAD-C polypeptide expression. HCT116, a, or MDA-
MB435S, b, cells were infected at a MOI of 5 with retroviruses encoding Gal4 alone
b 1400 (), Gal4 fused with mutant TAD-C polypeptide (Gal4-MUT, ) or with wild-type TAD-
Tumor Volume ( mm 3 )

1200 C polypeptide (Gal4-TADC, ). A total of 4 × 106 cells were implanted subcutaneously
1000 into nude mice. Results are expressed as the mean ± s.d. of 8–10 independent tumors.
800
The inset shows an immunoblot for Gal4 derived from infected cells, demonstrating
equivalent levels of expression. c, Immunoblot for p53 in HCT116 cells infected with
600
the indicated retroviruses at a MOI of 10. d, HCT116 cells were infected with Gal4-MUT
400
or Gal4-TADC at a MOI of 5. Cells were then mixed in equal proportion, and 6 × 106
200
cells implanted into nude mice. After three weeks, tumors were excised (Tumors 1–3),
0
and protein extracts were analyzed by immunoblotting for Gal4, along with extracts
0 1 2 3
Week prepared from cells before implantation (Mix).

NATURE MEDICINE • VOLUME 6 • NUMBER 12 • DECEMBER 2000 1337


© 2000 Nature America Inc. • http://medicine.nature.com
ARTICLES

a H&E αGal4 αCD31


b Gal4 Gal4-MUT Gal4-TADC

Gal4-MUT

Tum Tum Tum

c 3000

VP16 + Gal4-MUT
2500

Tumor Volume ( mm 3 )
VP16 + Gal4-TADC
2000 HIFVP+ Gal4-MUT
SQ SQ SQ HIFVP+ Gal4-TADC
1500
Gal4-TADC

1000
Tum Tum Tum
500

0
Nec Nec Nec 0 1 2 3
Weeks
© 2000 Nature America Inc. • http://medicine.nature.com

Fig. 5 The anti-tumor effect of TAD-C can be rescued by CH1-indepen- tumor (Tum) and necrotic areas (Nec) are indicated. c, HCT116 cells were
dent HIF-1α. a and b, Tumors arising from HCT116 cells infected with infected with either pBABE-puro-VP16 (VP16) or pBABE-puro-HIF-VP16
retrovirus encoding Gal4 fused with mutant (Gal4-MUT) or wild-type TAD- (HIFVP), and selected with puromycin. These cells were then infected with
C polypeptide (Gal4-TADC) at a MOI of 5 were excised after 3 weeks of retrovirus encoding either Gal4-MUT or Gal4-TADC at a MOI of 5. A total
growth. Paraffin sections were stained with hematoxylin and eosin, or im- of 6 × 106 doubly infected cells were implanted in nude mice. Results are
munostained for Gal4 or CD31/PECAM. The subcutaneous tissue (SQ), the mean of 8 independent tumors ± s.d.

Gal4 fused to mutant TAD-C polypeptide (Fig. 3c). Thus, ex- tumors arising from cells expressing Gal4 or Gal4 fused to mu-
pression of TAD-C polypeptide, in addition to blocking hy- tant (Gal4-MUT) polypeptide revealed small areas of necrosis,
poxia-inducible transcription, results in increased cell death but vast areas of viable tumor (Fig. 5a) that were grossly evident
under hypoxic tissue culture conditions. (Fig. 5b). Immunostaining for an endothelial cell marker
(CD31/PECAM) revealed abundant capillary formation within
Anti-tumor effect of disrupting HIF-mediated transcription control tumor segments compared with TAD-C–expressing tu-
To extend our findings, we assayed the growth of modified mors (Fig. 5a). It is unclear whether this was a primary or sec-
human tumor cells in a nude mouse xenograft model. We retro- ondary effect, that is, whether there were fewer tumor cells due
virally infected tumor cells at an equal MOI resulting in equiva- to diminished vessel density or vice versa.
lent expression of Gal4 or Gal4 fused to wild-type or mutant Because several known transcription factors interact with the
TAD-C polypeptide (Fig. 4a, inset). Tumor growth was signifi- CH1 domain of p300, and TAD-C polypeptide expression ap-
cantly and reproducibly attenuated in HCT116 colon carcinoma pears to generally disrupt CH1 function (Fig. 3a), we determined
(Fig. 4a) and MDA-MB435 breast carcinoma (Fig. 4b) cells ex- whether interference with HIF-1α binding and function is a crit-
pressing wild-type TAD-C polypeptide when compared with the ical component of the TAD-C–mediated anti-tumor effect. We
same cells expressing Gal4 alone or Gal4 fused to mutant engineered encoding a proteinHIF-1α in which the C-terminus,
polypeptide. This effect was independent of p53 status because which contains TAD-C and the residues required for oxygen-de-
HCT116 cells express wild-type p53 and MDA-MD435S cells ex- pendent degradation, was replaced with VP16 (HIF–VP16). This
press mutant p53. Immunoblot analysis for p53 showed no dif- created a stable, constitutively active, CH1-independent HIF-1α
ference in p53 abundance in HCT116 cells expressing these three molecule. We first infected HCT116 cells with either VP16 alone
ectopic Gal4 polypeptides (Fig. 4c). as a control, or with HIF–VP16. After selection, we infected these
We investigated whether the anti-tumor effect of TAD-C cells with Gal4 fused to wild-type or mutant TAD-C polypeptide.
polypeptide expression was caused by a paracrine effect, for ex- With co-expression of unfused VP16, tumor growth was attenu-
ample a cell producing TAD-C might elaborate a factor that sup- ated by TAD-C in comparison with the mutant polypeptide (Fig.
presses the growth of surrounding tumor cells. To this end, we 5c) in agreement with the results summarized above. With co-ex-
mixed tumor cells synthesizing either Gal4–TAD-C or Gal4-mu- pression of HIF–VP16, we saw identical tumor growth by cells co-
tant polypeptide in equal numbers before implantation. When expressing either TAD-C or mutant polypeptide (Fig. 5c). In
tumors were excised after three weeks of growth, we saw a strong immunoblot analysis, expression levels of VP16 and HIF–VP16
bias towards cells expressing the mutant polypeptide (Fig. 3d), were similar and did not alter co-expression of wild-type or mu-
indicating that the anti-tumor effect of TAD-C expression was tant TAD-C polypeptide (data not shown). We conclude that the
cell autonomous. anti-tumor effect of TAD-C expression is at least partly a result of
Necropsy and histological examination of the injection site of blocking access of endogenous HIF-1 to the CH1 domain of
tumor cells engineered to express TAD-C polypeptide revealed p300/CBP.
large areas of central necrosis with a small rim of viable tumor
cells adjacent to the dermal layer (Fig. 5a), even in the absence of Discussion
grossly measurable tumor at most of these sites (Fig. 5b). Control Though key data points to the importance of the HIF pathway

1338 NATURE MEDICINE • VOLUME 6 • NUMBER 12 • DECEMBER 2000


© 2000 Nature America Inc. • http://medicine.nature.com
ARTICLES

in tumor evolution, previous reports have differed on whether pCMV-CH1 mutants were constructed in pCMV/myc/nuc (Invitrogen,
interference with HIF operation is exploitable as an anti-tumor Carlsbad, California). A thioredoxin-constrained expression vector was cre-
strategy. Using HIF-1α–/– knock-out embryonic stem (ES) cells, ated by subcloning PCR fragments of Escherichia coli thioredoxin A into
the growth of subcutaneously implanted ES-cell tumors was ac- pcDNA3.1 (Invitrogen, Carlsbad, California), thereby creating new unique
restrictions sites within the active loop similar to prior reports44. The HIF-1α
celerated in one instance31 and decreased in another32. The de-
plasmid used for IVT, Epo-luciferase reporter, Gal4-luciferase reporter and
creased growth results are consistent with the finding that expression vectors encoding β-gal (pCMV-lacZ), VP16 (pCMX-VP16(N)),
growth of ARNT-deficient hepatomas33 and HIF-1α–deficient H- Gal4-SRC1, VP16-HIF-1α (aa 723–826), Gal4-CH1 (aa 300–528), Gal4-
ras–transformed cell lines34 was diminished compared with wild- SREBP2, Gal4-RAR, Gal4-STAT2 and Gal4-HIF-1α have been de-
type controls. These approaches demonstrate the consequences scribed10,29,30,45. Mammalian two-hybrid control vectors encoding Gal4-p53
of deleting key components of the HIF-1 pathway, but they do (pM-53) and VP16-T-antigen (pVP16-T) were obtained from Clontech (Palo
not indicate a tractable molecular mechanism for targeting the Alto, California). VSV-G-pseudotyped retroviruses were constructed in
intact pathway. pMMP, and packaged by cotransfection with pMD-MLV and pMD-G (gifts
from R. Mulligan, HHMI) in 293T cells. Titers were determined by infection
In principle, adaptation to hypoxia might be disrupted by in-
of the intended target cell line. HIF-VP16 and VP16 retroviruses were con-
terfering with one or more specific steps in the hypoxia-response structed by assembling PCR products corresponding to aa 1–529 of HIF-1α
pathway. The mechanisms that detect hypoxia and stabilize HIF- and/or VP16 in pBABE-puro, and packaged and titered, as described above.
1α represent attractive targets because HIF-1α is essentially unde- Details of plasmid construction are available on request.
tectable under normal conditions. However, the precise
mechanisms that result in accumulation of HIF-1α are not fully IVT and in vitro protein binding assays. We performed IVT in the presence
understood, although recent data indicate involvement of the of 35S-methionine (NEN) using T3 or T7 polymerase coupled with rabbit
PTEN/AKT pathway in some instances35,36. reticulocyte lysates (TNT, Promega) according to manufacturers instruc-
© 2000 Nature America Inc. • http://medicine.nature.com

Heterodimerization of HIF-1α with ARNT is crucial for HIF tions. GST-fusion proteins were produced in and purified from bacteria, and
binding of IVT proteins was performed as described45. Except where indi-
binding to an HRE and subsequent transactivation of hypoxia-
cated, binding and wash buffer consisted of 20 mM Tris-HCl (pH 8.0), 150
responsive genes25,37. ARNT is a common binding partner for sev- mM NaCl, 0.2% NP-40, 1 mM DTT, 10 µM ZnCl2, 2 µM EDTA, 50 µg/ml
eral bHLH-PAS proteins and is relatively abundant in cells in PMSF, 1 µg/ml pepstatin, 1 µg/ml leupeptin and 1.9 µg/ml aprotinin.
comparison with its HIF-1α partner2,8. In contrast, like HIF-1α, Where indicated, heavy metals were chelated by three changes of wash
CBP and p300 are limiting factors because of competition among buffer containing 10 mM 1,10-phenanthroline (Aldrich, St. Louis). For zinc
interacting proteins for binding to their overlapping interaction reconstitution after depletion of heavy metals, we used ultrapure zinc chlo-
domains38,39. Thus, disrupting the interaction of HIF-1α and ride (>99.999%, Aldrich, St. Louis). In all cases, standardized amounts of
purified GST fusion proteins, as determined by Coomassie stain, were used
p300/CBP might be more efficacious achieving interruption of
for in vitro binding assays.
complex formation than targeting HIF-1α heterodimerization
with an excess of ARNT. Our data indicate that p300/CBP are
Reporter assays. All reporter assays were performed in Hep3B cells.
crucial co-activators of hypoxia-inducible transcription and that Transfections were performed in 24-well plates using Fugene 6 (Roche) ac-
interfering with HIF-1α binding to p300/CBP represents a cording to the manufacturer’s instructions. Cells were subjected to nor-
tractable experimental means of abrogating hypoxia-induced moxia or hypoxia 24 h after transfection for the indicated periods of time.
transcription. Our data further show that interference with this Luciferase and β-gal assays were performed as described45. In all cases, raw
interaction has a potent anti-tumor effect, resulting at least luciferase activity was corrected for β-gal activity as an internal control for
partly from inadequate hypoxia-induced gene expression. transfection effciency and non-specific changes in transcriptional activity.
For each reporter experiment, at least two independent experiments were
Previous reports lend credence to the possibility of altering nu-
performed. Equivalent expression of mutants was verified by western-blot
clear protein structure/function with small molecule pharma-
analysis.
ceuticals40. A screen for small molecules capable of disrupting the
interaction of p300/CH1 with HIF-1α/TAD-C might provide a Hypoxia. Hypoxia was achieved by incubating cells in a humidified envi-
chemical means of reproducing the genetic alterations described ronment at 37 °C in a Cellstar three gas incubator (Harris, Asheville, North
here. In principle, systemic administration of such a small mole- Carolina) maintained at 10% CO2 and 1% O2.
cule may have a more potent anti-tumor effect than the synthe-
sis of TAD-C in tumor cells, because the functional-genetic Cell culture and retroviral infection. All cell lines were obtained from the
alterations described here were limited to the implanted tumor ATCC (Manassas, Virginia) and were cultivated in DMEM supplemented
cells with no molecular intervention specifically directed at with 10% FBS, penicillin, streptomycin and L-glutamine (Gibco-BRL,
tumor-stromal elements. Given the importance of the stroma in Rockville, Maryland). Cells in logarithmic growth phase were serially in-
fected, 12 h apart, with pMMP-Gal4 retroviruses to achieve an additive MOI
tumor evolution41 and in VEGF-mediated angiogenesis42, sys-
of 5 or 10, depending on the experiment. No selection was used, as im-
temic administration of drugs that disrupt the HIF-1 pathway munostaining routinely demonstrated ∼100% infection of cells at these
may lead to additional anti-tumor effects by additionally target- MOI. For pBABE-puro-VP16 or pBABE-puro-HIF-VP16, cells were infected at
ing tumor-infiltrating stromal cells, including tumor-associated a MOI of <1, and selected with puromycin (2 µg/ml for 4 d). Cells were al-
fibroblasts and endothelial cells. lowed to recover for 1 d and then infected at a MOI of 5 with pMMP-Gal4-
TAD-C or mutant TAD-C retroviruses.
Methods
Plasmids and retroviruses. Except where noted, all reagents were from Immunoblot, ribonuclease protection, and cell death assays.
Sigma (St. Louis). Plasmids were generated by subcloning PCR fragments Immunoblot analyses for Gal4 (RK5C1, Santa Cruz Biotechnology, Santa
using standard techniques43 and were verified by sequencing. All mam- Cruz, California), p53 (DO-1, Oncogene Research, Boston), and HIF-1α
malian expression vectors used cytomegalovirus (CMV) promoters. GST-fu- (OZ1510) were performed using standard techniques43. Total RNA was iso-
sion proteins were generated in pGEX-2T (Amersham). Gal4-fusion proteins lated (RNeasy, Qiagen, Valencia, California), and normalized amounts (10
were generated in pCMX-Gal4(N) (gift from R. Evans). Scanning NAAIRS µg) were subjected to ribonuclease protection assays using a multiprobe
mutagenesis of GST–CH1 (aa 302–528) was performed by ssDNA mutage- human angiogenesis kit (RiboQuant, Pharmingen, San Diego) according to
nesis (Muta-Gene, Bio-Rad) according to manufacturer’s instructions. the manufacturer’s instructions. Cell death was measured in HepG2 cells in-

NATURE MEDICINE • VOLUME 6 • NUMBER 12 • DECEMBER 2000 1339


© 2000 Nature America Inc. • http://medicine.nature.com
ARTICLES

fected with the indicated retroviruses at a MOI of 5, followed by exposure and intratumour heterogeneity. Acta Oncol. 33, 383–389 (1994).
to ambient oxygen or hypoxia for 36 h. Cell death was determined by stain- 16. Helmlinger, G., Yuan, F., Dellian, M. & Jain, R.K. Interstitial pH and pO2 gradi-
ents in solid tumors in vivo: high-resolution measurements reveal a lack of corre-
ing with Annexin V antibody (Pharmingen, San Diego) and 7-amino-actino-
lation. Nature Med. 3, 177–182 (1997).
mycin D (Pharmingen, San Diego), and analyzed by flow cytometry 17. Brown, J.M. Exploiting the hypoxic cancer cell: mechanisms and therapeutic
according to the manufacturer’s instructions. Statistical significance was de- strategies. Mol. Med. Today. 6, 157–162 (2000).
termined by Student’s t-test. 18. Shweiki, D., Itin, A., Soffer, D. & Keshet, E. Vascular endothelial growth factor in-
Xenograft tumor model. The indicated number of tumor cells was in- duced by hypoxia may mediate hypoxia-initiated angiogenesis. Nature 359,
843–845 (1992).
jected subcutaneously and bilaterally into the flanks of 6–8-wk nude mice 19. Zhong, H. et al. Overexpression of HIF-1α in common human cancers and their
(Taconic, Germantown, New York). Tumors were measured with calipers metastases. Cancer Res. 59, 5830–5835 (1999).
and tumor volume calculated as 0.5 × L × W2. For each cell line, at least two 20. Hockel, M. et al. Association between tumor hypoxia and malignant progression
independent experiments were performed. Where indicated, tumors were in advanced cancer of the uterine cervix. Cancer Res. 56, 4509–4515 (1996).
21. Brizel, D.M. et al. Tumor oxygenation predicts for the likelihood of distant
excised and fixed in 4% paraformaldehyde in PBS or minced and lysed in
metastases in human soft tissue sarcoma. Cancer Res. 56, 941–943 (1996).
RIPA/300 buffer composed of 50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 1% 22. Zeman, E. M., Brown, J. M., Lemmon, M. J., Hirst, V. K. & Lee, W. W. SR-4233: a
NP40, 0.5% DOC, 0.1% SDS, 1 mM DTT, protease inhibitors (Complete, new bioreductive agent with high selective toxicity for hypoxic mammalian
Roche). Immunostaining with anti-Gal4 (IKE, Santa Cruz Biotechnology, cells. Int. J. Radiat. Oncol. Biol. Phys. 12, 1239–1242 (1986).
Santa Cruz, California) or anti-CD31 (MEC13.3, Pharmingen, San Diego) 23. Binley, K., Iqball, S., Kingsman, A., Kingsman, S. & Naylor, S. An adenoviral vec-
tor regulated by hypoxia for the treatment of ischaemic disease and cancer.
was performed using the Vectastain ABC Elite system (Vector, Burlingame, Gene Ther. 6, 1721–1727 (1999).
California) according to manufacturer’s the instructions. 24. Shibata, T., Giaccia, A.J. & Brown, J.M. Development of a hypoxia-responsive
vector for tumor-specific gene therapy. Gene Ther. 7, 493–498 (2000).
ACKNOWLEDGEMENTS 25. Jiang, B.H., Zheng, J.Z., Leung, S.W., Roe, R. & Semenza, G.L. Transactivation
and inhibitory domains of HIF-1α. Modulation of transcriptional activity by oxy-
We thank R. Mulligan for retroviral vectors; and L. Sambucetti, S. Zabludoff, gen tension. J. Biol. Chem. 272, 19253–19260 (1997).
D. France and S. Bhattacharya for their advice. This work was supported by 26. Wilson, I. A. et al. Identical short peptide sequences in unrelated proteins can
© 2000 Nature America Inc. • http://medicine.nature.com

have different conformations: a testing ground for theories of immune recogni-


grants from the Novartis/DFCI Drug Discovery Program (to D.M.L.), and the tion. Proc. Natl. Acad. Sci. USA 82, 5255–5259 (1985).
Howard Hughes Medical Institute (to A.L.K. and W.G.K.). 27. Chakravarti, D. et al. Role of CBP/P300 in nuclear receptor signalling. Nature
383, 99–103 (1996).
28. Oliner, J.D., Andresen, J.M., Hansen, S.K., Zhou, S. & Tjian, R. SREBP transcrip-
RECEIVED 23 JUNE; ACCEPTED 11 OCTOBER 2000 tional activity is mediated through an interaction with the CREB-binding pro-
tein. Genes Dev. 10, 2903–2911 (1996).
1. Semenza, G.L. Regulation of mammalian O2 homeostasis by hypoxia-inducible 29. Yao, T.P., Ku, G., Zhou, N., Scully, R. & Livingston, D.M. The nuclear hormone
factor 1. Annu. Rev. Cell. Dev. Biol. 15, 551–578 (1999). receptor co-activator SRC-1 is a specific target of p300. Proc. Natl. Acad. Sci. USA
2. Wang, G.L., Jiang, B.H., Rue, E.A. & Semenza, G.L. Hypoxia-inducible factor 1 is a 93, 10626–10631 (1996).
basic-HLH-PAS heterodimer regulated by cellular O2 tension. Proc. Natl. Acad. 30. Bhattacharya, S. et al. Cooperation of Stat2 and p300/CBP in signalling induced
Sci. USA 92, 5510–5514 (1995). by interferon-α. Nature 383, 344–347 (1996).
3. Salceda, S. & Caro, J. HIF-1α protein is rapidly degraded by the ubiquitin-protea- 31. Carmeliet, P. et al. Role of HIF-1α in hypoxia-mediated apoptosis, cell prolifera-
some system under normoxic conditions. Its stabilization by hypoxia depends tion and tumour angiogenesis. Nature 394, 485–490 (1998).
on redox-induced changes. J. Biol. Chem. 272, 22642–22647 (1997). 32. Ryan, H.E., Lo, J. & Johnson, R.S. HIF-1α is required for solid tumor formation
4. Huang, L.E., Arany, Z., Livingston, D.M. & Bunn, H.F. Activation of HIF depends and embryonic vascularization. EMBO J 17, 3005–3015 (1998).
primarily upon redox-sensitive stabilization of its α subunit. J. Biol. Chem. 271, 33. Maxwell, P.H. et al. HIF-1 modulates gene expression in solid tumors and influ-
32253–32259 (1996). ences both angiogenesis and tumor growth. Proc. Natl. Acad. Sci. USA 94,
5. Maxwell, P.H. et al. The tumour suppressor protein VHL targets hypoxia-inducible 8104–8109 (1997).
factors for oxygen-dependent proteolysis. Nature 399, 271–275 (1999). 34. Ryan, H.E. et al. HIF-1α is a positive factor in solid tumor growth. Cancer Res. 60,
6. Cockman, M.E. et al. HIF-α binding and ubiquitylation by the von Hippel-Lindau 4010–015 (2000).
tumor suppressor protein. J. Biol. Chem. 275, 25733–25741 (2000). 35. Zhong, H. et al. Modulation of HIF-1α expression by the epidermal growth fac-
7. Ohh, M. et al. Ubiquitination of hypoxia-inducible factor requires direct binding tor/phosphatidylinositol 3-kinase/PTEN/AKT/FRAP pathway in human prostate
to the β-domain of the von Hippel-Lindau protein. Nature Cell Biol. 2, 423–427 cancer cells: implications for tumor angiogenesis and therapeutics. Cancer Res.
(2000). 60, 1541–1545 (2000).
8. Jiang, B.H., Semenza, G.L., Bauer, C. & Marti, H.H. HIF-1 levels vary exponentially 36. Zundel, W. et al. Loss of PTEN facilitates HIF-1-mediated gene expression. Genes
over a physiologically relevant range of O2 tension. Am. J. Physiol. 271, Dev. 14, 391–396 (2000).
C1172–1180 (1996). 37. Jiang, B.H., Rue, E., Wang, G.L., Roe, R. & Semenza, G.L. Dimerization, DNA
9. Jiang, B.H., Rue, E., Wang, G.L., Roe, R. & Semenza, G.L. Dimerization, DNA binding, and transactivation properties of HIF-1. J. Biol. Chem. 271,
binding, and transactivation properties of HIF-1. J. Biol. Chem. 271, 17771–17778 (1996).
17771–17778 (1996). 38. Kamei, Y. et al. A CBP integrator complex mediates transcriptional activation and
10. Arany, Z. et al. An essential role for p300/CBP in the cellular response to hy- AP-1 inhibition by nuclear receptors. Cell 85, 403–414 (1996).
poxia. Proc. Natl. Acad. Sci. USA 93, 12969–12973 (1996). 39. Horvai, A.E. et al. Nuclear integration of JAK/STAT and Ras/AP-1 signaling by
11. Ebert, B.L. & Bunn, H.F. Regulation of transcription by hypoxia requires a multi- CBP and p300. Proc. Natl. Acad. Sci. USA 94, 1074–1079 (1997).
protein complex that includes HIF-1, an adjacent transcription factor, and 40. Foster, B.A., Coffey, H.A., Morin, M.J. & Rastinejad, F. Pharmacological rescue of
p300/CREB binding protein. Mol. Cell. Biol. 18, 4089–4096 (1998). mutant p53 conformation and function. Science 286, 2507–2510 (1999).
12. Kallio, P.J. et al. Signal transduction in hypoxic cells: inducible nuclear transloca- 41. Olumi, A.F. et al. Carcinoma-associated fibroblasts direct tumor progression of
tion and recruitment of the CBP/p300 co-activator by the HIF-1α. EMBO J. 17, initiated human prostatic epithelium. Cancer Res. 59, 5002–5011 (1999).
6573–6586 (1998). 42. Fukumura, D. et al. Tumor induction of VEGF promoter activity in stromal cells.
13. Carrero, P. et al. Redox-regulated recruitment of the transcriptional co-activa- Cell 94, 715–725 (1998).
tors CREB-binding protein and SRC-1 to HIF-1α. Mol. Cell. Biol. 20, 402–415 43. Ausubel, F.M. et al. (eds.) Current Protocols in Molecular Biology (Wiley-
(2000). Interscience, New York, 1988).
14. Semenza, G.L. & Wang, G.L. A nuclear factor induced by hypoxia via de novo 44. LaVallie, E.R. et al. A thioredoxin gene fusion expression system that circumvents
protein synthesis binds to the human erythropoietin gene enhancer at a site re- inclusion body formation in the E. coli cytoplasm. Biotechnology (NY) 11,
quired for transcriptional activation. Mol. Cell. Biol. 12, 5447–5454 (1992). 187–193 (1993).
15. Nordsmark, M., Bentzen, S.M. & Overgaard, J. Measurement of human tumour 45. Bhattacharya, S. et al. Functional role of p35srj, a novel p300/CBP binding pro-
oxygenation status by a polarographic needle electrode. An analysis of inter- tein, during transactivation by HIF-1. Genes Dev. 13, 64–75 (1999).

1340 NATURE MEDICINE • VOLUME 6 • NUMBER 12 • DECEMBER 2000