Download (Hsp90) inhibitors

COMMUNICATION
Redefining the phenotype of Heat shock protein 90 (Hsp90)
inhibitors
Yao Wang,b Yen Chin Koaya, and Shelli R. McAlpine*a
The phenotypes produced when cells are treated with the Hsp90
inhibitors AUY922 or 17-AAG (classical inhibitors) are different to
those produced when cells are knocked down with Hsp90. Pulldown assays using classical inhibitors suggest that these molecules
bind to multiple targets other than Hsp90. Classical inhibitors also
induce similar protein markers as other anti-cancer therapies
Cisplatin and bortezomib that do not target Hsp90. Together these
data suggest that AUY922 and 17-AAG acts on multiple targets and
likely kills cells through multiple mechanisms. Comparing these
classical inhibitors to the effects seen when treating cells with Cterminal Hsp90 modulators reveals that C-terminal modulators
effectively bind to Hsp90, and induce phenotypic markers consistent
with the Hsp90 CRISPR knockdown data. Our findings challenge
the current interpretation of Hsp90 inhibitors and suggest that a large
body of literature that describes the Hsp90 phenotype and inhibitors
is re-examined in this context.
Heat shock protein 90 (Hsp90) is a well-investigated
chemotherapeutic target, with 15 unique clinical structures
reported to act by controlling this protein [1]. Hsp90 plays a
major role in controlling proteins that regulate many
oncogenic pathways[2], and it was anticipated that Hsp90
inhibitors would be highly successful at blocking drug
resistance and metastasis [3]. However, these clinically tested
drugs have produced disappointing results. Mechanistically,
all 15 clinical trial drugs (termed “classical Hsp90 inhibitors”)
are similar: they are all reported to bind inside the ATPbinding pocket on Hsp90’s N-terminus, blocking ATP-Hsp90
interactions, thereby suppressing Hsp90’s functions. All
classical inhibitors produce a similar cellular phenotype that
includes the up-regulation of heat shock factor 1 (HSF-1),
heat shock protein 70 (Hsp70) and heat shock protein 27
(Hsp27). Indeed, the most common approach for identifying
Hsp90 inhibitors is to screen for molecules that induce an
increase in Hsp70 protein, with the assumption that they will
be Hsp90 inhibitors. It is also well established that these
three heat shock proteins initiate anti-apoptotic cascades and
promote drug resistance,[4] which is a highly problematic
property in a cancer drug.
Another explanation for these classical inhibitors’ clinical
failure is that act via multiple mechanisms in a cellular
environment. Binding to multiple proteins perhaps in addition
to Hsp90 and producing off-target effects would explain the
perplexing data reported for these classical inhibitors.
Specifically,
several
N-terminal
inhibitors,
including
Geldanamycin
(GM)
and
17-(Allylamino)-17-
[a]
[b]
Y.C.Koay, Assoc. Prof. S.R.McAlpine
School of Chemistry, University of New South Wales,
Sydney, NSW 2052, Australia.
E-mail: [email protected]
Dr. Y..Wang
Department of Medicine
University of New South Wales
E-mail: [email protected]
demethoxygeldanamycin (17-AAG), have a binding affinity for
isolated or purified Hsp90 protein in the low micromolar
range[5], which is at odds with these inhibitors’ nanomolar GI50
values for inhibiting cancer cell growth. Surprisingly, studies
using siRNA or CRISPR technology to knockdown or
knockout Hsp90 have not been reported and compared to the
impact of treating cells with these extensively utilized
molecules. Although pull-down assays using tagged
molecules of classical Hsp90 inhibitors have been published,
close analysis of these data suggest that the molecules do
not target Hsp90 alone but rather have non-specific binding
affinities for multiple proteins[5c, 6]. In the past 10 years over
10,000 papers have been published reporting studies on
Hsp90 (Scopus), of which over 5,000 use classical inhibitors
to investigate Hsp90 pathways. Thus, it is important to
understand how these classical inhibitors behave in a cellular
environment.
Herein we show that the phenotype of Hsp90
knockdown cells using CRISPR technology is significantly
different to that generated when cells are treated with
classical inhibitors. We reveal that drugs acting via
mechanisms unrelated to Hsp90 induce a similar phenotype
to that produced by classical Hsp90 inhibitors. We also show
that in contrast to an earlier report, [5c] classical molecules are
not highly effective at binding to Hsp90 in cancer cells despite
their toxicity. We compare the growth inhibitory effect of these
classical versus C-terminal inhibitors (SM inhibitors) to their
binding affinity for Hsp90. Despite the C-terminal modulators
lower cellular potency than the classical inhibitors, we show
that the C-terminal modulators outperform classical inhibitors
when pulling down Hsp90 from cancer cell lysates. The SM
inhibitors also produce a phenotype that is identical to that of
Hsp90 knockdown cells, whereas the classical inhibitors
produce a very different phenotype. Taken together our data
indicate that SM molecules likely function primarily via
inhibition of Hsp90. In contrast, classical inhibitors likely act
via multiple mechanisms (perhaps including Hsp90), and
produce a stress response that is unrelated to Hsp90
inhibiton within the cell.
Early work has shown that the behavior of classical[7]
and C-terminal inhibitors[8] is cell line independent. It has
been well documented that both classical inhibitors and Cterminal inhibitors behave similarly in all cancer cells and this
is logical given that Hsp90 is an important chaperone in all
cancer cells. Our Hsp90 knockdown studies were conducted
in HeLa cells using the CRISPR/Cas9 knockout system,
however, the data is translatable to other cell lines. When the
protein level of Hsp90α, which is the inducible form of Hsp90
and encoded by HSP90AA1, was decreased by ~ 45% of the
control level in HeLa cells, there was no observable increase
in the protein expression of HSF-1, Hsp70 and Hsp27 (Fig
1a). Thus a decrease of Hsp90 protein levels does not
produce cellular stress, at least as measured by increases in
Hsp70 or Hsp27. In contrast, cells treated with AUY922
COMMUNICATION
IS
PR
R
C
Sc
ra
m
C
on
t
ro
l
a)
HeLa
bl
ed
(which according to the literature targets Hsp90) showed a
strong up-regulation of these heat shock proteins (HSPs) (Fig
1a). The protein expression level of Hsp90α isoform was
significantly increased by AUY922 treatment (Fig 1a ~ 3 fold).
These data indicate that the increase in HSPs is likely
unrelated to Hsp90 inhibition in cells. These data also
suggest that the cellular stress caused by AUY922 even at
low levels of AUY922 treatment, is induced via a mechanism
that does not involve Hsp90. Comparing cells treated with Cterminal modulators SM258 (or supplemental data SM253) to
the knockdown showed that both modulators produced a
similar impact on HSP production to that seen in Hsp90α
knockdown cells (Fig. 1b). Thus, in contrast to the classical
inhibitors, SM compounds produce a similar phenotype to
that produced when knocking down Hsp90 α.
AUY922
Hsp90 inhibition, we investigated the impact of Bortezomib
(PS-341 or PS) and Cisplatin (Cis), two well-established
chemotherapeutic drugs, on the protein expression of HSF-1
and Hsp70 in HCT116 human colorectal carcinoma cells. PS
acts as a proteasome inhibitor that blocks protein degradation.
Cis crosslinks DNA and leads to DNA damage in cells. At
concentrations that were 5 and 10 fold over the GI 50 of each
drug, PS, Cis and AUY922 all increased the protein levels of
HSF-1 and Hsp70 in HCT116 cells after 24 h-treatments (Fig.
2). Treatments with PS resulted in much higher levels of
Hsp70 protein than caused by AUY922 treatments.
Specifically, the Hsp70 expression level in PS-treated
HCT116 cells was ~ 9 fold the control level, whereas Hsp70
increased by ~ 5 fold over the control in AUY922-treated cells
(Fig. 2).
HCT116
AUY-922 (nM)
a)
DMSO 25 nM 50 nM 100 nM
Hsp90a
HSF-1
HCT116
AUY-922 (nM)
Hsp70
Hsp27
Actin
R
IS
PR
C
am
50
HCT116
PS-341 (nM)
100
HSF-1
HSF-1
Hsp70
Hsp70
Actin
Actin
0
30
60
HCT116
Cisplatin (µM)
HSF-1
HSF-1
HSF-1
Hsp70
Hsp70
Hsp70
Actin
Actin
Actin
IV
V
VI
VII
b)
AUY922
SM253 SM258
50 nM 100 nM 30 mM
25 mM
Hsp90a
Hsp70
Hsp27
Actin
12
0
30
0
30
60
HSP Expression in HCT116 Cells
(24 h-treatments)
HSF-1
Hsp70
10
8
6
4
2
0
I
II
III
IV
V
VI
VII
Figure 1. Hsp90α knockdown study in HeLa cells. a) The impact of
Hsp90α knockdown and AUY922 treatments on the protein expression of
Hsp90α, HSF-1, Hsp70 and Hsp27. b) Comparison of the impact of
Hsp90α knockdown, AUY922 (50 nM and 100 nM), SM253 (30 µM) and
SM258 (25 µM) on the production of Hsp90α, total Hsp90, HSF-1, Hsp70
and Hsp27. Actin was used as the protein loading control. All
experiments were performed at least three times, and representative
images are shown.
The knockdown data provides evidence that the upregulation of HSPs caused by AUY922 treatment is unlikely
to be the result of suppressing Hsp90’s function. However, it
may be the outcome of a cellular stress response caused by
off-target toxicity. In order to assess whether similar stress
responses can be triggered by mechanisms unrelated to
A
U
A Y9 D
U 2 M
Y9 2
S
22 (50 O
( 1 nM
0
PS 0 )
( nM
P S 30 )
n
C ( 60 M )
is n
(
C 30 M)
is µ
(6 M
0 )
µM
A
)
U
Y
A 9 D
U 2 M
Y9 2
S
22 (50 O
( 1 nM
0
PS 0 )
( nM
P S 30 )
n
C ( 60 M )
is n
(
C 30 M)
is µ
(6 M
0 )
µM
)
Lane:
HCT116
PS-341 (nM)
100
Fold Change Relative to
Control Treatment
III
bl
ed
II
Sc
r
HeLa
or
m
al
b)
I
N
Lane:
0
50
0
Figure 2. The protein expression of HSF-1 and Hsp70 in HCT116 cells
treated with PS and Cis at indicated concentrations for 24 hrs. a) The
representative Immunoblot images are shown. Actin was used as the
protein loading control. b) Immunoblot results are plotted as fold change
relative to the untreated control, which is set as 1. All values are average ±
s.e.m. from three independent experiments.
Although not as effective as PS or AUY922, Cis also
induced up-regulation of Hsp70 by 2 fold over of the control
(Fig. 2). These data indicate that toxic molecules can produce
high levels of HSPs regardless of mechanism, and show that
inducing high levels of Hsp70 and HSF-1 are a phenotype of
general toxicity. Taken together the knockdown data and the
induction of Hsp70 and HSF-1 when treated with three
60
COMMUNICATION
different types of drugs are evidence that classical inhibitors
produce stress effects that are unrelated to Hsp90.
Verification of a molecule’s affinity for a protein in cell
lysate is typically accomplished using pull-down assays.
Pulldowns have been performed using biotin-tagged
Geldanamycin (GM-tag)[5c], where competitive binding assays
were used to demonstrate that 17-AAG bound effectively to
Hsp90. Through competitive binding assays using GM-Tag, we
tested two classes of inhibitors: classical inhibitors 17-AAG, and
AUY922 (structures in Supplementary Fig. 1a and b), and Cterminal modulators SM253 and SM258 (structures in
Supplementary Fig. 1c and d).
Pulling down Hsp90 from cell lysates required tagged
versions of these inhibitors, thus PEG-Biotin-tagged SM253 and
SM258 (SM253-Tag and SM258-Tag, Supplementary Fig. 1e
and f) were synthesized, and PEG-Biotin-tagged GM (GM-Tag,
Supplementary Fig. 1g) was purchased. 17-AAG binds to Hsp90
with an IC50 of 700 nM despite killing cells at concentrations mid
to low nM (GI50 =10-50 nM). Thus, it is likely acting via multiple
mechanisms given its low affinity for Hsp90. Similarly, AUY922,
which kills cells at low nM (GI50 ~10 nM), only has an affinity for
Hsp90 of 230 nM. (Fig. 3a). These two data are inconsistent for
a compound that acts primarily via Hsp90. AUY922 should be
more effective at binding to Hsp90 than killing cells because
there is no cellular membrane involved in the pulldown studies.
Calculating the amount of Hsp90 present in the cell lysate
confirms the hypothesis that AUY922 and 17-AAG are likely not
acting via as Hsp90 their primary mechanism. Hsp90 is ~ 3% of
all protein in HCT116 cancer cells. Using 50 µg of HCT116 cell
lysate and GM-Tag, there is ~110 nM of Hsp90, which is 55 nM
of the active Hsp90 dimer. Yet 700 nM of 17-AAG or 230 nM
AUY922 are required in order to bind effectively to the Hsp90
dimer, while the concentrations of 17-AAG and AUY922 required
to inhibit 50% of HCT116 cell growth is 50 nM and 10 nM
respectively. These data show that it is unlikely classical
inhibitors kill cells via an Hsp90 mechanism because the
concentration required to bind to Hsp90 is up to 23 fold greater
than the amount required to kill cells. The significantly higher
concentrations required to bind to Hsp90 in the presence of cell
lysate versus the much lower doses for suppressing cell growth
indicate that these classical inhibitors must be killing cells via
alternative mechanisms.
In contrast, the SM molecules are relatively effective at
binding to Hsp90 with comparable binding affinities to their GI 50
values for killing cells. Specifically, they have IC50 binding
affinities for Hsp90 of ~20 µM,[9] which is similar to the GI50 (~5
µM).[8] Thus the SM compounds likely reach the necessary
concentration to bind effectively to Hsp90, and kill cells via this
mechanism. Comparison of tagged SM compounds to GM-Tag
showed that GM-Tag pulled down similar levels of Hsp90 as the
negative controls. In contrast, SM253-Tag and SM258-Tag
pulled down Hsp90 by ~ 6 and 14 fold more than the GM-Tag,
respectively (Fig. 3b).
b)
Figure 3. Binding affinity of Hsp90 inhibitors for Hsp90 in HCT116 cells. a)
Competitive pull-down assays using GM-Tag as the probe were performed to
evaluate the binding affinity of 17-AAG and AUY922 for Hsp90 in HCT116 cell
lysate. Treatment with DMSO in the absence of GM-Tag was used as the
negative control. b) A protein pull-down assay comparing the binding affinity of
SM253-Tag (i.e. SM253-T), SM258-Tag (i.e. SM253-T) and GM-Tag (i.e. GMT) for Hsp90 in HCT116 cell lysate. DMSO and PEG-Biotin tag (i.e. Tag) were
tested as negative controls. All experiments were performed in triplicate, and
representative immunoblot images are shown (the images of full length
Coomasie stained gels are in the Supplementary Fig. 3).
In summary, classic inhibitors, GM, 17-AAG, and AUY922,
all exhibit poor binding affinity for Hsp90 in cell lysate, and they
induce a stress response in cells, which includes the
overproduction of multiple HSPs at protein expression levels.
This cellular phenotype is different from that seen in the
Hsp90
slightly decreased. The large increase in Hsp70 produced when
cells are treated with classical inhibitors is consistent with other
drugs that induce cellular stress. Our combined CRISPR
phenotype and pulldown data show that classical Hsp90
inhibitors likely suppress cell growth via multiple mechanisms,
which may or may not include the inhibition of Hsp90.
Treatment with SM compounds against cancer cells
suppresses HSP production, which is similar to the phenotype
observed in Hsp90 knockdown cells. C-terminal modulators of
Hsp90, SM253 and SM258, selectively pull down Hsp90 in cell
lysate, indicating their capability for targeting Hsp90 in the cell
system. Indeed, their binding affinity for Hsp90 in a cellular
environment correlates to drug concentrations needed to kill
cells. These data indicate that the SM molecules likely act on
Hsp90 as the primary target. Taken together, these data
reinforce the fact that phenotypes identified using small
molecules should be compared to knockdown phenotypes of
their proposed targets before a target is validated for these
small molecules.
Acknowledgements
This work was supported by the US National Institute of Health
grant NCI R01CA137873 and the Australian National Health and
Medical Research Council (NHMRC) grant GNT1043561. We
thank UNSW for support of Y.W. and S.R.M., and the NHMRC
COMMUNICATION
for support of Y.W. We also thank the Endeavour program for
supporting Y.C.K.
[6]
Keywords: Heat shock protein 90 • phenotype • CRISPR •
knockdown • AUY922
[1]
[2]
[3]
[4]
[5]
See <http://www.clinicaltrials.gov>
D. Hanahan, R. A. Weinberg, Cell 2011, 144, 646-674.
a) U. Banerji, Proc. Am. Assoc. Cancer Res. 2003, 44, 677; b) E. A.
Sausville, Curr. Cancer Drug Targets 2003, 3, 377-383; c) L. H. Pearl,
C. Prodromou, Curr. Opin. Struct. Bio. 2000, 10, 46-51; d) A. Dehner, J.
Furrer, K. Richter, I. Schuster, J. Buchner, H. Kessler, ChemBioChem
2003, 4, 870-877; e) F. X. Hagn, K. Richter, J. Buchner, H. Kessler,
Proc. experimental nuclear magnetic resonance conference 2005; f) J.
M. Jez, J. C. Chen, G. Rastelli, R. M. Stroud, D. V. Santi, Chem. Biol.
2003, 10, 361-368.
J. R. Mcconnell, S. R. Mcalpine, Bioorg. Med. Chem. Lett. 2013, 23,
1923-1928.
a) B. Panaretou, C. Prodromou, S. M. Roe, R. O'brien, J. E. Ladbury, P.
W. Piper, L. H. Pearl, EMBO J 1998, 17, 4829-4836; b) S. M. Roe, C.
Prodromou, O. B. R., J. E. Ladbury, P. W. Piper, L. H. Pearl, J. Med.
[7]
[8]
[9]
Chem. 1999, 42, 260-266; c) A. Kamal, L. Thao, J. Sensintaffar, L.
Zhang, M. F. Boehm, L. C. Fritz, F. J. Burrows, Nature. 2003, 425, 407.
a) G. Chiosis, H. Jhuezo, N. Rosen, E. Mimgaugh, L. Whitesell, L.
Neckers, Mol. Cancer Ther. 2003, 2, 123-129; b) K. Moulick, J. Ahn, H.
Zong, A. Rodina, L. Cerchietti, D. E. Gomes, E. Caldas-Lopes, K.
Beebe, F. Perna, K. Hatzi, L. Vu, X. Zhao, D. Zatorska, T. Taldone, P.
Smith-Jones, M. Alpaugh, S. S. Gross, N. Pillarsetty, T. Ku, J. S. Lewis,
S. M. Larson, R. Levine, H. Erdjument-Bromage, M. L. Guzman, S. D.
Nimer, A. Melnick, L. Neckers, G. Chiosis, Nat. Chem. Biol. 2011, 25,
818-826.
a) U. Banerji, M. Walton, F. Raynaud, R. Grimshaw, L. Kelland, M.
Valenti, I. Judson, P. Workman, Clin. Cancer Res. 2005, 11, 70237032; b) L. E. Xu, J. L.; Egorin, M. J.; D’argenio, D. Z.
, J.
Pharmacokinet Pharmacodyn. 2003, 30, 185-219.
Y. C. Koay, J. R. Mcconnell, Y. Wang, S. R. Mcalpine, RSC Advances
2015, 5, 59003-59013.
R. P. Sellers, L. D. Alexander, V. A. Johnson, C.-C. Lin, J. Savage, R.
Corral, J. Moss, T. S. Slugocki, E. K. Singh, M. R. Davis, S. Ravula, J.
E. Spicer, J. L. Oelrich, A. Thornquist, C.-M. Pan, S. R. Mcalpine,
Bioorg. Med. Chem. 2010, 18, 6822-6856.
COMMUNICATION
Entry for the Table of Contents (Please choose one layout)
Layout 1:
COMMUNICATION
Text for Table of Contents
Yao Wang, Yen Chin Koay, Shelli R
McAlpine*
Hsp90
Hsp90
Page No. – Page No.
KnockDown
C-terminal
Hsp90
modulator
Redefining the phenotype of Hsp90
inhibitors
Phenotype A
Classical
Inhibitor
AUY922
Phenotype B
hihere))