Page 1
1
Combined HDAC and Bromodomain Protein Inhibition Reprograms Tumor Cell Metabolism and
elicits Synthetic Lethality in Glioblastoma
Yiru Zhang1, Chiaki Tsuge Ishida1, Wataru Ishida2, Sheng-Fu L. Lo2, Junfei Zhao3, Chang Shu1,
Elena Bianchetti1, Giulio Kleiner4, Maria J. Sanchez-Quintero4, Catarina M. Quinzii4, Mike-
Andrew Westhoff5, Georg Karpel-Massler6, Peter Canoll1 and Markus D. Siegelin1
1 Department of Pathology & Cell Biology, Columbia University Medical Center, New York, NY,
U.S.A., 2 Department of Neurosurgery, The Johns Hopkins University School of Medicine,
Baltimore, Maryland, U.S.A., 3 Department of Biomedical Informatics, Columbia University, New
York, NY, U.S.A., 4 Department of Neurology, H. Houston Merritt Neuromuscular Research
Center, Columbia University Medical Center, New York, NY, U.S.A., 5 Department of Pediatrics
and Adolescent Medicine, Ulm University Medical Center, Ulm, Germany, 6 Department of
Neurosurgery, Ulm University Medical Center, Ulm, Germany
Running title: Targeting the Epigenome reprograms glioblastoma metabolism.
Correspondence to:
Markus D. Siegelin, MD, Dr. med., Department of Pathology & Cell Biology, Columbia University
Medical Center, 630 West 168th Street, P&S Rm. 15-415, New York, NY 10032, Phone: 212
305 1993, [email protected] , [email protected]
Financial Support:
M.D. Siegelin: NIH NINDS R01NS095848, K08NS083732, Louis V. Gerstner, Jr. Scholars
Program (2017-2020) and American Brain Tumor Association Discovery Grant 2017
(DG1700013)
Conflict of interest statement: The authors have declared that no conflict of interest exists.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 2
2
Abstract
Purpose: Glioblastoma remain a challenge in oncology in part due to tumor heterogeneity.
Experimental Design: Patient-derived xenograft and stem-like glioblastoma cells were used as
the primary model systems. Results: Based on a transcriptome and subsequent gene set
enrichment analysis (GSEA), we show by using clinically validated compounds that combined
histone deacetylase (HDAC) inhibition and Bromodomain protein (BRD) inhibition results in
pronounced synergistic reduction in cellular viability in patient-derived xenograft and stem-like
glioblastoma cells. Transcriptome based GSEA analysis suggests that metabolic
reprogramming is involved with synergistic reduction of oxidative and glycolytic pathways in the
combination treatment. Extracellular flux analysis confirms that combined HDAC inhibition and
BRD inhibition blunts oxidative and glycolytic metabolism of cancer cells, leading to a depletion
of intracellular ATP production and total ATP levels. In turn, energy deprivation drives an
integrated stress response, originating from the endoplasmic reticulum. This results in an
increase in pro-apoptotic Noxa. Aside from Noxa, we encounter a compensatory increase of
anti-apoptotic Mcl-1 protein. Pharmacological, utilizing the FDA-approved drug sorafenib, and
genetic inhibition of Mcl-1 enhanced the effects of the combination therapy. Finally, we show in
orthotopic patient-derived xenografts of GBM, that the combination treatment reduces tumor
growth, and that the triple therapy, involving clinically validated compounds, panobinostat,
OTX015 and sorafenib further enhances these effects, culminating in a significant regression of
tumors in vivo. Conclusion: Overall, these results warrant clinical testing of this novel,
efficacious combination therapy.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 3
3
Introduction
Over the last decade, there were tremendous new insights in the characteristics of solid
cancers, such as melanoma and glioblastoma. Notable findings are for instance the discovery of
BRAF mutations in melanoma and IDH1 mutations in gliomas (1-6). Regarding gliomas,
comparable efforts have been taken, albeit these approaches have not matured to the extent as
in melanoma. Nevertheless, given the lack of durability new or amended strategies are required
to address the problem of solid tumors.
In conjunction with the discovery of driver mutations, there are a number of small molecule
inhibitors available that target deregulated pathways in cancers. Similarly, transcription is
altered in cancer by various mechanisms, e.g. DNA methylation or histone acetylation. Very
recently, mutations in the histone 3.3 protein (H3) at codon 27 were identified and it was shown
in preclinical model systems that malignancies with this aberration benefit from treatment with a
histone-deacetylase inhibitor. In this regard, broad HDAC-inhibitors, such as panobinostat or
vorinostat, have reached FDA-approval and are in clinical testing for solid malignancies as well.
However, as with other monotherapies, tumors are either primarily or become secondarily
resistant to these agents. Therefore, combination therapies that target multiple pathways may
address this problem. One of these signaling cascades is c-myc, which is highly upregulated in
many cancers and which became druggable through the discovery of bromodomain protein
inhibitors (BRDs) (7,8), such as the prototype JQ1 and clinically more amenable chemical
derivatives, such as OTX015.
Here, we provide a mechanism-based drug combination therapy that synergistically reprograms
tumor cell transcription, leading to metabolic reprogramming, energy starvation and activation of
the integrated stress response followed by significant intrinsic apoptosis. Moreover, the
combination therapy causes a partial cyto-protective effect by elevating Mcl-1 and interference
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 4
4
with Mcl-1 through an FDA-approved drug leads to significant tumor regression and survival
extension in two glioblastoma xenograft models in vivo.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 5
5
Materials and Methods
Reagents
OTX015 (OTX), vorinostat (Vr), panobinostat (Pb) and sorafenib (Sf) were purchased from
Selleckchem.
Cell cultures and growth conditions
U87MG, LN229, U87-EGFRvIII and T98G human glioblastoma cell lines and A375 malignant
melanoma cells were obtained from the American Type Culture Collection (Manassas, VA) or
the Coriell Institute for Medical Research, respectively. NCH644 and NCH421K stem cell-like
glioma cells were obtained from Cell Line Services (CLS, Heidelberg, Germany). The GBM14
PDX cells have been described elsewhere (9-12). The respective cell line depository
authenticated the cells. All cell lines were cultured as previously described (13-16).
Cell viability assays
Viability assays were performed as previously described (13-18).
Measurement of apoptosis and mitochondrial membrane potential
For Annexin V/propidium iodide staining the Annexin V Apoptosis Detection Kit (BD
Pharmingen) was used as previously described (16,19). Tetramethylrhodamine ethyl ester
(TMRE) staining was performed according to the manufacturer’s instructions (Mitochondrial
Membrane Potential kit, Cell Signaling Technology, Danvers, MA). The data were analyzed with
the FlowJo software (version 8.7.1; Tree Star, Ashland, OR).
Extracellular flux analysis
Extracellular flux analysis was performed on the Seahorse XFe24 analyzer. The mitochondrial
stress assay was utilized in accordance with the instructions by the manufacturer and as
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 6
6
described earlier (14). The glycolytic stress test was used in accordance with the instructions by
the manufacturer. Briefly, 40,000 cells were seeded and starved. During the assay, cells were
exposed to glucose, followed by oligomycin and completed by 2-DG. The extracellular
acidification rate (ECAR) was measured.
Transfections of siRNAs or transduction of shRNAs
Briefly, cells were incubated for 6h with the formed complexes of Oligofectamine® 2000
(Invitrogen, Carlsbad, CA) and the respective siRNA (12-well condition) in DMEM without FBS
and antibiotics. After 6h, FBS was added to a total concentration of 1.5%. Transduction of
lentiviral shRNAs is performed as described earlier (14).
Western blot analysis, protein capillary electrophoresis and immunoprecipitation
analysis
Specific protein expression in cell lines was determined by Western blot analysis as described
before (18). Co-immunoprecipitations were performed as described earlier in (13).
Subcutaneous xenograft model
1 x 106 A375 cells, 1 x 106 U87-EGFRvIII or GBM12 PDX xenograft tumors were implanted
subcutaneously into the flanks of 6-8 week-old SCID SHO mice as described before (16).
Measurements were performed with a caliper and tumor sizes were calculated as (length x
width2)/2. Treatments were performed intraperitoneally as described in the respective figure
legends of each individual experiment. Columbia IACUC has approved these studies.
Orthotopic glioblastoma PDX and stem-cell like xenograft model
For the GBM12 orthotopic GBM model (34), 300,000 cells and for the stem-like GBM model,
20,000 NCH644 cells were injected in a manner as described earlier. The drugs for the
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 7
7
indicated treatments were dissolved as described earlier (10). Columbia IACUC has approved
these studies.
Microarray and gene set enrichment analysis
Transcriptome and gene set enrichment analysis (GSEA) was performed as previously
described (14). For the present experiments, two biological replicates per condition were
submitted. The experiment was deposited online with GEO with the following ID: GSE108958.
Statistical analysis
Statistical significance was assessed by Student’s t-test using Prism version 7.00 (GraphPad,
La Jolla, CA). Data was considered statistically significant at * p<0.05, ** p<0.01, *** p<0.001, or
**** p<0.0001 level. n.s. indicates not significant. The CompuSyn software (ComboSyn, Inc.,
Paramus, NJ) was used to detect synergistic, additive or antagonistic effects as previously
described.
Study approval
All procedures were in accordance with Animal Welfare Regulations and approved by the
Institutional Animal Care and Use Committee at the Columbia University Medical Center.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 8
8
Results
HDAC and BRD inhibition leads to synergistic reduction in tumor cell proliferation
To ensure that the utilized compounds act on-target, we validated as to whether or not HDAC
inhibition enhances acetylation of known targets. To this end, patient-derived xenograft cells,
GBM14, were treated with panobinostat in the presence or absence of OTX015. Protein
expression analysis for total histone H3 and acetylated histone H3 validated on-target activity
(Supplementary Figure 1A). To demonstrate that combined HDAC and BRD inhibition is a
potential efficacious strategy to inhibit growth of glioblastoma cells in a synergistic manner, we
tested two clinically validated drug compounds that interfere with these pathways, panobinostat
(HDAC inhibitor) and OTX015 (BRD inhibitor), in established (U87, U87-EGFRvIII, T98G,
LN229), patient-derived xenograft (GBM14) and stem-like glioma cells (NCH644). In all model
systems tested, the drug combination over a range of concentrations elicited a synergistic
interaction as evaluated by combination index value analysis (CI values) (Figure 1A and 1B).
Notably, the most significant synergistic interaction was identified in patient-derived xenograft
cells, GBM14. Similar results were obtained when FDA-approved, vorinostat, was administered
in lieu of panobinostat (Supplementary Figure 1B). These results suggest that combined
targeting of HDACs and BRD is efficacious.
HDAC and BRD inhibition causes enhanced apoptotic cell death
The combination of panobinostat and OTX015 resulted in morphological signs of apoptosis. In
support of this finding, GSEA showed that the combination treatment activated a transcriptional
pro-apoptotic state (Figure 2C). Therefore, we determined as to whether or not features of
apoptotic cell death can be confirmed biochemically. To this purpose, LN229, T98G and U87
GBM cells or stem-cell like GBM cells, NCH421k and NCH644 were treated with panobinostat
(or vorinostat), OTX015 or the combination of both and stained with Annexin V/propidium iodide
and analyzed by multi-parametric flow cytometric analysis. Consistently, we found that the
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 9
9
combination treatment of OTX015 and Panobinostat led to more apoptotic cells than control or
single treatments. Similar findings were made when vorinostat was used in lieu of panobinostat
(Figure 2A and 2B, Supplementary Figure 2A-C). Intrinsic apoptosis is accompanied by loss of
mitochondrial membrane potential. Consistently, the combination treatments
(panobinostat+OTX015; vorinostat+OTX015) reduced the amount of TMRE positive cells
stronger than single treatments or control in LN229, T98G and U87 cells (Figure 2D and 2E and
Supplementary Figure 2D). To assess whether or not caspases are involved in the death, we
treated GBM cells in the presence or absence of pan-caspase inhibitor, zVAD-fmk. We found
that zVAD-fmk partially protected the cells from DNA fragmentation induced by the combination
treatment, suggesting that caspases are involved in the death (Supplementary Figure 2E). This
finding was also supported by enhanced cleavage of PARP by the combination treatment
(Figure 2F and Supplementary Figure 2H).
Combined HDAC and BRD inhibition modulates the expression of pro- and anti-apoptotic
Bcl-2 family members
Given the activation of apoptosis, we were wondering about the regulation of this form of cell
death induced by the combination treatment. To this purpose, we analyzed protein expression
of both anti-apoptotic Bcl-2 family members, Mcl-1, Bcl-xL and Bcl-2, and pro-apoptotic Bcl-2
family members, Noxa and BIM, in established, stem-like and patient derived xenograft cells of
GBM. Our findings demonstrate that combined treatment with OTX015 and HDAC-inhibitors
results in an increase in BIM and Noxa protein levels (Figure 2G and 2H and Supplementary
Figure 2F and 2G). Aside from these pro-apoptotic molecular changes, we found a
compensatory, transitory up-regulation of Mcl-1 and its deubiquitinase Usp9X induced by the
combination treatment (Figure 2G and 2H and Supplementary Figure 2F and 2G). Next, we
assessed the transcriptional changes by real-time PCR analysis and found that mRNA for
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 10
10
Usp9x, Mcl-1, Noxa and BIM were all increased with the exception of BIM in NCH644 cells
(Figure 2I and 2J).
Next, we evaluated the functional impact of the observed protein changes on the apoptotic
efficacy of the combination treatment. Given the increase in Noxa and Mcl-1, we mainly focused
on the MCl-1/Noxa/BAK cascade in which high levels of Noxa bind to Mcl-1 and mediate the
BAK/Mcl-1 dissociation. In turn, BAK engages in mitochondrial outer membrane
permeabilization. To this purpose, we silenced the expression of BAK with two siRNAs.
Silencing of BAK was confirmed by standard western blotting (Figure 3A). Knockdown of BAK
provided a partial protection from panobinostat+OTX015 mediated apoptosis (Figure 3B). Next,
we tested the impact of Noxa on the combination treatment and silencing of Noxa was
confirmed by capillary electrophoresis (Figure 3A). We found that silencing of Noxa provided a
partial protection from panobinostat+OTX015 mediated apoptosis (Figure 3C and
Supplementary Figure 3E), in keeping with the results on BAK. To gain a further understanding
about the molecular interactions implicated in the combination treatment, we conducted co-
immunoprecipitation analysis by pulling down Mcl-1 in the presence or absence of the
combination treatment (Figure 3D). While the IgG control did not pull down any of the analyzed
proteins, anti-Mcl-1 prominently precipitated Mcl-1 and its associated described binding
partners, Usp9X, BIM, Noxa, BAK, but not GAPDH (negative control), confirming high specificity
of our experiment (Figure 3D). Regarding the effects of the combination treatment on the
binding partners of Mcl-1, we observed that the panobinostat+OTX015 treatment resulted in an
increased binding of Noxa to Mcl-1 (Figure 3D). Enhanced binding of Noxa to Mcl-1 is
accompanied by a dissociation of two pro-apoptotic, Bcl-2 family members, BIM and BAK, from
Mcl-1 (20,21). Moreover, a reduced interaction between Usp9x and Mcl-1 was observed (Figure
3D).
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 11
11
Given that we observed an increased expression of Mcl-1 and Usp9x at early time points upon
treatment with the combination treatment, we hypothesized that these effects mediate a
transitory “pro-survival” effect. Indeed, later time points (>24h after treatment) show a decline in
Mcl-1, coinciding with enhanced death (Figure 3G and 3H). To test this hypothesis, we utilized
two Mcl-1 specific siRNAs and one Mcl-1 pool siRNA, which potently suppressed Mcl-1 protein
levels (Figure 3E and Supplementary Figure 3A-D). Indeed, silencing of Mcl-1 significantly
enhanced apoptosis in LN229 GBM cells (a cell line that responds slower to the combination
treatment) and T98G GBM cells (Figure 3E and 3F and Supplementary Figure 3A-D). In like
manner, we evaluated the effects of USP9X silencing on the combination treatment and found
that Usp9X silencing enhanced apoptosis induced by the combination treatment further (Figure
3F and Supplementary Figure 3C).
These results intimate that inhibition of Mcl-1 might further enhance the efficacy of the drug
combination therapy. To further evaluate this point, we tested whether the kinase inhibitor,
sorafenib, a known modulator of Mcl-1 levels and FDA-approved drug (22,23), is capable of
counteracting the combination treatment mediated increase of Mcl-1 protein levels. In keeping
with this expectation, sorafenib blunted Mcl-1 up-regulation mediated by panobinostat and
OTX015 (Figure 3G and 3H). Remarkably, sorafenib also curtailed Usp9X expression,
suggesting that sorafenib abrogated two main pro-survival effects elicited by the combination
treatment (Figure 3G and 3H). To evaluate whether sorafenib also enhances apoptosis
induction by the combination treatment, LN229 or GBM14 cells were treated with the
combination treatment in the presence or absence of sorafenib. While in the absence of
sorafenib the combination treatment displayed some apoptosis induction, this effect was
significantly enhanced in the presence of the multi-kinase inhibitor (Figure 3J and
Supplementary Figure 3F). Similar results were obtained in T98G GBM cells (DNA –
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 12
12
fragmentation) (Supplementary Figure 3G). These findings establish a novel, rational triple drug
therapy, called OTX+Pb+Sf (OTX015 + panobinostat + sorafenib).
Combined HDAC and BRD inhibition suppresses oxidative phosphorylation and
glycolysis in a synergistic manner
Given the pronounced synergistic reduction of cellular proliferation over a broad range of
different cell lines, we wondered about the underlying molecular mechanisms, orchestrating and
driving these changes. To this end, GSEA with subsequent interaction analysis assisted us to
identify pathways that are synergistically down regulated by panobinostat and OTX015. This
analysis was proved powerful since it permits to elucidate which molecular processes require
both compounds for most potent pathway inhibition. It turned out that the key energy producing
pathways in tumor cells, glycolysis and oxidative phosphorylation, were drastically suppressed
(Figure 4A), presumably leading to an intracellular energy crisis. Consistently, the combination
treatment suppressed protein levels of transporters and enzymes, involved in glycolysis
(GLUT1, LDHA, GAPDH) (Figure 4B).
Next, we determined the underlying molecular mechanism, governing these significant
metabolic changes mediated by HDAC and BRD inhibition. We hypothesized that these
metabolic aberrations might likely be a result of interference with the expression of the master-
regulator, c-Myc. First, c-Myc is pivotal for the regulation of tumor cell metabolism (24). Second,
c-Myc is a master-regulator in glioma stem cells (7). Third, both HDAC- and BRD- inhibitors
have been reported to suppress c-Myc levels (25). Based on this reasoning, we found that
combined inhibition of BRD and HDAC results in silencing of MYC transcript as shown by real-
time PCR and GSEA analysis in glioblastoma stem-like cells, NCH644 (Figure 4C, 4D and
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 13
13
Supplementary Figure 4A). Along with the reduction in c-Myc, the combination treatment also
affected markers of stemness, such as SOX2 and Nanog. Although in these model systems
CD133 and Nestin cannot be considered as bona-fide stem-cell markers, their expression was
suppressed as well (Supplementary Figure 4B). Most relevantly, c-Myc was down-regulated on
protein level, showing the strongest c-Myc suppression in the combination treatment of OTX015
and panobinostat (Figure 4B).
The above findings support the hypothesis that glioblastoma cells treated with the combination
of OTX015 and panobinostat have low energy levels. Indeed, the combination treatment
depletes GBM14 patient-derived xenograft cells of ATP more potently than single treatments or
vehicle (Figure 4E). This is also in keeping with the related gene set enrichment analysis,
showing a transcriptional signature of “starvation” (GO_RESPONSE_TO_STARVATION)
(Figure 5A).
Given the impact on the transcripts and proteins related to energy metabolism, we conducted
extracellular flux analysis to validate whether or not the combination treatment affects oxidative
phosphorylation and glycolysis in glioblastoma cell cultures (Figure 4F-M)). To this end, patient-
derived xenograft cells, GBM14, were treated with OTX015, panobinostat, the combination
treatment in the presence or absence of sorafenib and were subjected to analysis for oxygen
consumption rate (OCR) in the context of a mitochondrial stress test. We found a significant
reduction in basal OCR (mitochondrial oxygen consumption rate) in the combination treatment,
which was mirrored also by the amount of OXPHOS related ATP production (Figure 4F, 4J and
4K). Similarly, we found that the maximal respiration and spare respiratory capacity was
prominently reduced by the combination treatment, which was further enhanced by the
presence of sorafenib, suggesting that sorafenib further inhibits oxidative phosphorylation, in
keeping with its suppression on Mcl-1 protein levels (Figure 4L-M). Next, we assessed the
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 14
14
impact of the combination treatment on glycolysis by determining ECAR (extracellular
acidification rate) in the context of a glycolysis stress test. Akin to OXPHOS, we found that the
combination treatment reduced key indicators of glycolysis more potently than each compound
alone (Figure 4G-I). In the presence of sorafenib, the combination treatment completely
abrogated the glycolytic reserve, suggesting that sorafenib further enhances the inhibitory
effects of the combination treatment on energy metabolism (Figure 4I). To address the
applicability of these findings to other model systems, we performed the same experiments in
another glioblastoma cell culture, LN229. The findings in LN229 are overall similar to GBM14
except that the triple therapy (OTX+Pb+Sf) had a more drastic impact on OXPHOS related ATP
production and that the combination treatment was in some instances less efficacious
(Supplementary Figure 5A-D). Finally, we evaluated the impact of c-Myc levels on glycolysis in
patient-derived xenograft GBM cells, GBM14, and found that silencing of c-Myc levels
suppressed glycolysis (Supplementary Figure 5E). These results are in keeping with our
findings related to the combination treatment of OTX015 and panobinostat, showing enhanced
suppression of glycolytic key indicators and confirm that c-Myc suppression results in an
inhibition of glycolysis. All in all, our findings suggest that the combination treatment interferes
with energy metabolism, and these effects are further enhanced in the presence of sorafenib.
The combined inhibition of BRD and HDAC leads to endoplasmic reticulum stress
followed by enhanced expression of pro-apoptotic Noxa
Based on the findings related to energy metabolism, we hypothesized that the state of energy
deprivation elicited by the combination treatment should affect the homeostasis of the
endoplasmic reticulum (ER) since ATP is necessary for protein folding and loss of ATP will
result in the accumulation of unfolded proteins. GSEA confirms this notion, which was also
validated by RT-PCR analysis for common ER-stress related markers, e.g. GRP78, ATF3, ATF4
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 15
15
and CHOP in both NCH644 and LN229 GBM cells (Figure 5A-C). To confirm that the
transcriptional changes are also found at protein level, we conducted western blot analysis as
well as capillary electrophoresis for common ER-stress related markers. The combination
treatment of OTX015 and panobinostat displayed the most significant ER-stress related
signature with up-regulation of GRP78, phosphorylated eIF2, total eIF2, ATF3, ATF4 and
CHOP (Figure 5D and Supplementary Figure 2H). Moreover, the induction of ER-stress by the
indicated treatments correlated with the cleavage of PARP, suggesting that ER-stress is
important for the induction of apoptosis by the combination treatment (Supplementary Figure
2H). Overall, these results confirm the notion that the combination treatment elicits a potent ER-
stress response likely related to energy starvation.
To determine the down-stream consequences of ER-stress, we hypothesized that the ER-stress
signaling is responsible for the enhanced induction of Noxa by the combination treatment, which
is supported by earlier findings by others and our group (26,27). In this context, we first focused
on the PERK-ATF4-Noxa cascade and hypothesized that the combination treatment utilizes this
stress pathway to upregulate Noxa expression. To evaluate this claim, we silenced the
expression of PERK by siRNA in LN229 GBM cells. Silencing of PERK decreased ATF4 levels
and suppressed OTX015, panobinostat and the combination treatment mediated increase of
Noxa protein levels (Figure 5E). Another transcription factor related to ER-stress signaling and
to Noxa regulation is ATF3 (26,27). Given that ATF3 was potently increased by the combination
treatment, we tested the hypothesis that silencing of ATF3 will attenuate Noxa up-regulation by
the combination treatment. To this purpose, we silenced ATF3 in U87 and LN229 GBM cells
and subsequently treated the cells with OTX015 and panobinostat. In agreement with our
hypothesis, we found that silencing of ATF3 suppressed Noxa increase in response to treatment
with the drug combination (Figure 5F).
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 16
16
Combined inhibition of BRD and HDAC facilitates the stability of Mcl-1 in a GSK3
dependent manner
Aside from the pro-apoptotic changes induced by the combination treatment, we detected an
increase of Mcl-1 and its associated deubiquitinase Usp9X at early time points, which based on
our mechanistic analysis exerts pro-survival features since interference with both targets
enhances the apoptotic effects of the combination treatment. Therefore, we sought to establish
a deeper understanding by which mechanisms Mcl-1 is up-regulated. Given the increase of
Usp9X, it appears obvious that the combination treatment modulates the stability of Mcl-1 aside
from its effects on mRNA levels. To this purpose, we conducted a cycloheximide (CHX) block
experiment in the presence or absence of OTX015 and panobinostat combination in LN229 and
T98G cells. While control treated cells displayed a rapid decline in Mcl-1 protein levels, the
combination treatment enhanced the stability of Mcl-1 in both cell lines (Figure 5G and 5H),
indicating that OTX015 and panobinostat combination affects Mcl-1 half-life.
Concerning protein stability, it is well known that Mcl-1 possesses a short-half life and is prone
to rapid degradation by proteasomes. Proteasomal degradation is preceded by phosphorylation
of Mcl-1 through several kinases. Amongst those regulators, GSK3is known to phosphorylate
Mcl-1 at serine 159 (28), preceding ubiquitin conjugation and subsequent energy dependent
degradation by proteasomes. Given these implications, we hypothesized that the combination
treatment inhibits the activity of GSK3 (phosphorylated form at serine 9) and that in turn Mcl-1
becomes dephosphorylated at serine 159, rendering Mcl-1 more stable (29). To this purpose,
GBM14 and LN229 cells were treated with OTX015, panobinostat or the combination. Protein
expression analysis revealed that both cell lines demonstrate an increase in Mcl-1, which was
most pronounced in the combination treatment, which was accompanied by a reduction in Mcl-1
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 17
17
phosphorylation at serine 159 (Figure 5I). In agreement with the reduction of Mcl-1
phosphorylation, the combination treatment led to an enhanced phosphorylation of GSK3 at
serine 9 (Figure 5I), rendering the kinase less active. These findings reinforce the notion that
Mcl-1 is regulated at the posttranslational level by the combination treatment.
To address the related up-stream mechanisms that govern the Mcl-1 increase mediated by the
combination treatment, we reasoned that it may be related to the effects of the combination
treatment on the transcription factor, c-Myc, since suppression of c-Myc has been linked to
inhibit PP2A (30), which binds to and de-phosphorylates GSK3. To assess whether this
previously described relationship of c-Myc and GSk3 exists in the setting of glioblastoma
models as well, we created two stable cell lines derived from the patient-derived xenograft cell
culture, GBM14, either infected with a lentiviral shRNA targeting c-Myc, or a lentiviral driving
ectopic over-expression of c-Myc. The activity of the corresponding constructs was verified by
protein capillary electrophoresis, confirming the silencing or over-expression of c-Myc in the
respective cell cultures (Figure 5J). In agreement with our earlier hypothesis, we found that
while silencing of c-Myc results in an enhancement of phosphorylation of GSK3, coupled with
an increase in total Mcl-1 and reduced phosphorylation of Mcl-1 at serine 159, over-expression
of c-Myc results in a reciprocal expression phenotype with reduced phosphorylation of GSK3
and lower protein levels of Mcl-1 (Figure 5J). Similar results were observed in LN229 GBM cells
in which stable knockdown of c-Myc results in increased levels of Mcl-1 and over expression
leads to down-regulation of Mcl-1. To exclude that c-Myc modulation affects Mcl-1 mRNA levels,
we conducted real-time PCR analysis in c-Myc over-expressing and silenced cells. In both
situation, we found that c-Myc modulation only marginally affected Mcl-1 mRNA, suggesting that
most likely c-Myc regulates Mcl-1 in a posttranslational manner (Figure 5K). These results
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 18
18
confirm that c-Myc levels affect Mcl-1 protein levels likely through enhanced stability mediated
by reduced GSK3 activity.
The triple therapy of OTX015, panobinostat and sorafenib leads to tumor regression and
significant host survival extension
To determine the efficacy of the combination treatment of OTX015 and panobinostat, we utilized
several xenograft model systems, involving glioblastoma and malignant melanoma. To this
purpose, we implanted U87-EGFRvIII cells in the subcutis of immunocompromised mice. After
establishment of tumors, treatment groups were formed, consisting of vehicle, OTX015,
panobinostat, OTX015+panobinostat, sorafenib and the triple combination of
OTX015+panobinostat+sorafenib. Our findings show that the combination treatment of OTX015
and panobinostat reduced tumor growth much more potently than each compound alone (Figure
6A, 6B and Supplementary Figure 7A-B). We found similar results in a BRAF V600E mutated
melanoma xenograft model (A375 cells), albeit less efficient compared to the U87-EGFRvIII
model (Supplementary Figure 7E-F). Given the limited efficacy of the combination treatment in
vivo, we addressed whether or not sorafenib would further enhance the combination treatment.
As anticipated with our in vitro findings, the triple combination therapy had the strongest impact
on tumor growth, leading to a regression of tumors (Figure 6A, 6B and Supplementary Figure
7A-B). To determine as to whether or not single, combination and triple therapies display similar
efficacy in a patient-derived xenograft GBM model, we tested the GBM12 PDX model and found
that in agreement with the findings obtained from the U87-EGFRvIII model system, the triple
therapy was most effective and leads to a significant suppression of tumor growth in this model
system as well (Figure 6C-D and Supplementary Figure 7D). Next, we evaluated the impact of
the various treatment on the histopathological level. We found that the triple therapy results in a
significantly reduced cellularity and mitotic rate, while at the same time an increase in
apoptosis/necrosis was encountered in the U87-EGFRvIII and GBM12 PDX model system
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 19
19
(Figure 6E-H, and Supplementary Figure 6A-C and 7A). Consistently, we found a reduced
amount of Ki67 staining in tumors that received the triple therapy, consistent with a reduced
mitotic rate (Figure 6F and 6H and Supplementary Figure 6B and 7A). To confirm cell death, we
conducted TUNEL-staining and found the highest amount of TUNEL positive cells in the triple
therapy treated tumors (Figure 6G-H and Supplementary Figure 6C and 7A). In addition, we
confirmed that the triple combination therapy suppresses c-Myc protein levels in vivo
(Supplementary Figure 7C).
Finally, we determine as to whether or not this treatment approach is active in orthotopic models
of glioblastoma, which bear a closer resemblance to the clinical situation and are more
challenging to treat. First, since the combination treatment regulated c-Myc and several other
transcription factors that orchestrate the stem-cell like phenotype in NCH644 GBM cells
(Supplementary Figure 4B and Figure 4B-C), we used NCH644 cells and implanted them in into
the right striatum of immunocompromised mice (Figure 6I). Treatment groups were formed and
animals were treated three times a week over the course of three weeks. While single
treatments (OTX or Pb) and the combination treatment (OTX+Pb) did not result in a significant
life extension for the host animals (compared to vehicle treated animals), the triple therapy
(OTX+Pb+Sf) resulted in a statistical significant survival benefit compared to the control and
combination treatment (Figure 6I). Next, we evaluated our drug treatments in the current gold-
standard for preclinical drug therapy, an orthotopic patient-derived xenograft of glioblastoma. To
this end, we utilized the GBM12 PDX model (9-12) and implanted cells into the right striatum of
nude mice (Figure 6J). Thereafter, animals were assigned to groups and treated according to
the same schedule of the NCH644 orthotopic model system. The triple therapy resulted in a
significant extension of host survival akin to the results obtained in the heterotopic xenograft
model (Figure 6J). These findings suggest that our treatment strategy is active in orthotopic
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 20
20
xenograft models, suggesting that these agents are capable of crossing the blood-brain barrier
and exert efficacy in the brain microenvironment.
Finally, we provide a summarizing scheme of the involved mechanisms of the combination and
triple therapy (Figure 6K).
Discussion
Although multi-combination therapies are a logical step ahead due to the inherent heterogeneity
of most solid tumors (31,32), one of the concerns remains toxicity and the challenging clinical
burden of proof that even a triple therapy is more efficacious and safe than single treatments or
combination therapies, consisting of only two drugs. However, a lesson again might be learnt
from HIV therapy. While in the early 1980’s HIV constituted a death sentence, this changed
dramatically of the course of several decades when HIV was started to be treated with “drug
combination cocktails”. In our study, we have taken a similar approach by combining two drug
compound classes, involving HDAC inhibitors and BRD inhibitors, and in the course of
mechanism studies we identified a rational triple therapy, involving HDAC inhibitors
(panobinostat, vorinostat), BRD inhibitor (OTX015) and multikinase inhibitors (sorafenib).
Our findings reveal that combined inhibition of HDAC by panobinosat or vorinostat, and BRD
proteins by OTX015 results in a significant synergistic reduction of cell growth that is
predominantly mediated through enhanced cell death in the form of apoptosis, which was
supported by gene set enrichment analysis showing up-regulation of multiple apoptosis
promoting genes. Cell death was protected by the pan-caspase inhibitor, zVAD-fmk, which
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 21
21
demonstrating that caspases are involved in the death pathway elicited by the combination
treatment. These findings are in agreement with earlier studies, involving single treatments with
HDAC inhibitors or BRD inhibitors (25,33-35), that demonstrated activation of intrinsic
apoptosis. We hypothesized that the enhanced apoptotic cell death by the combination
treatment is driven by an alteration of pro- and anti-apoptotic factors and identified Noxa as one
of the key players to mediate apoptosis. Noxa is a pro-apoptotic Bcl-2 family member that is
known for its inhibitory activity on anti-apoptotic Mcl-1. Induced by several stress responses,
Noxa is increased in the context of various stimuli and known to be down-stream of the
integrated stress response when ATF4 is up regulated. In keeping with this, the drug
combination treatment resulted in an activation of a profound ER-stress response and increased
Noxa via at least two mechanisms, involving either the PERK-ATF4 pathway or ATF3
(13,27,36,37).
Based on our transcriptome analysis, we found evidence for energy deprivation induced by the
combination treatment, which was further supported by the fact that gene sets related to energy
metabolisms were synergistically down-regulated. We validated this observation by determining
total ATP levels and confirmed these findings by extracellular flux analysis, which supported the
notion that the combination treatment suppressed glycolysis and oxidative energy metabolism.
To the best of our knowledge, our findings are the first to show that simultaneous inhibition of
BRD and HDAC proteins results in an energy crisis in glioblastoma cells, which then in turn
elicits a profound stress response with activation of ER-stress.
To provide an explanation for the energy suppression mediated by the combination treatment,
we hypothesized that c-Myc is involved in this context since c-Myc is known to modulate energy
metabolism in cancer cells (24,38-40). Indeed, our findings unanimously showed that the
combination treatment reduced c-Myc mRNA and protein levels in a synergistic manner and
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 22
22
along with c-Myc suppression we found a decline in enzymes and transporters related to
glycolysis, which are known transcriptional targets of c-Myc. Indeed, selective stable knockdown
of c-Myc suppressed glycolysis in patient-derived xenograft cells. Although c-Myc seems to be a
major player in the regulation of our observed metabolic changes, we cannot exclude that
others factors regulate energy metabolism in the combination treatment.
Another major aspect of our findings relates to the observation that the combination treatment
not only increased the levels of pro-apoptotic molecules, but also anti-apoptotic factors, such as
Mcl-1 and its deubiquitinase Usp9X, which were detected in essentially every cell culture tested,
suggesting that this response appears to be robust and broadly relevant. Given this implication,
it was tempting to speculate as to whether or not the combination treatment can be further
enhanced through selective genetic or pharmacological inhibition. Our findings confirmed the
anti-apoptotic role of Mcl-1, and sorafenib (23) potently enhanced the effects of the combination
treatment, giving rise to a novel, rational triple therapy. Interestingly, the combination treatment
affected Mcl-1 levels both at the transcriptional as well as the posttranslational level. At the post-
translational level, the combination treatment increased Mcl-1 through a feed-forward
mechanism, involving the suppression of c-Myc with a subsequent inactivation of GSK3 and
stabilization of Mcl-1. Being an FDA-approved drug, sorafenib (23,41,42) appeared to be the
ideal compound to counteract the Mcl-1 increase mediated by the combination treatment
because sorafenib has been shown to suppress Mcl-1 levels through multiple mechanisms,
involving inhibition of GSK3, but also suppression of MCL1 transcripts (22,43). In keeping with
this, sorafenib suppressed Mcl-1 up-regulation by the combination treatment in our model
systems and in turn facilitated apoptosis induced by the combination treatment. Most relevantly,
our in vivo studies showed that adding sorafenib to the combination treatment resulted in a
potent enhancement of efficacy in the highly aggressive U87-EGFRvIII model as well as in a
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 23
23
patient-derived xenograft system. While initially we examined only heterotopic xenografts,
similar impressive results were obtained with the triple combination therapy in orthotopic
models, significantly extending overall survival. These findings intimate that the triple therapy is
feasible, well-tolerated and efficacious. In the broader picture, this research suggests that akin
to infectious disease the consideration of multi-combination therapies might be a viable
approach to enhance treatment efficacy of recalcitrant malignancies.
References
1. Fu X, Chin RM, Vergnes L, Hwang H, Deng G, Xing Y, et al. 2-Hydroxyglutarate Inhibits
ATP Synthase and mTOR Signaling. Cell Metab 2015;22(3):508-15 doi
10.1016/j.cmet.2015.06.009.
2. Sasaki M, Knobbe CB, Itsumi M, Elia AJ, Harris IS, Chio, II, et al. D-2-hydroxyglutarate
produced by mutant IDH1 perturbs collagen maturation and basement membrane function.
Genes Dev 2012;26(18):2038-49 doi 10.1101/gad.198200.112.
3. Gross S, Cairns RA, Minden MD, Driggers EM, Bittinger MA, Jang HG, et al. Cancer-
associated metabolite 2-hydroxyglutarate accumulates in acute myelogenous leukemia with
isocitrate dehydrogenase 1 and 2 mutations. J Exp Med 2010;207(2):339-44 doi
10.1084/jem.20092506.
4. Yan H, Parsons DW, Jin G, McLendon R, Rasheed BA, Yuan W, et al. IDH1 and IDH2
mutations in gliomas. N Engl J Med 2009;360(8):765-73 doi 10.1056/NEJMoa0808710.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 24
24
5. Parsons DW, Jones S, Zhang X, Lin JC, Leary RJ, Angenendt P, et al. An integrated
genomic analysis of human glioblastoma multiforme. Science 2008;321(5897):1807-12 doi
10.1126/science.1164382.
6. Balss J, Meyer J, Mueller W, Korshunov A, Hartmann C, von Deimling A. Analysis of the
IDH1 codon 132 mutation in brain tumors. Acta Neuropathol 2008;116(6):597-602 doi
10.1007/s00401-008-0455-2.
7. Fang X, Zhou W, Wu Q, Huang Z, Shi Y, Yang K, et al. Deubiquitinase USP13 maintains
glioblastoma stem cells by antagonizing FBXL14-mediated Myc ubiquitination. J Exp Med
2017;214(1):245-67 doi 10.1084/jem.20151673.
8. Wang X, Cunningham M, Zhang X, Tokarz S, Laraway B, Troxell M, et al.
Phosphorylation regulates c-Myc's oncogenic activity in the mammary gland. Cancer research
2011;71(3):925-36 doi 10.1158/0008-5472.CAN-10-1032.
9. Parrish KE, Cen L, Murray J, Calligaris D, Kizilbash S, Mittapalli RK, et al. Efficacy of
PARP Inhibitor Rucaparib in Orthotopic Glioblastoma Xenografts Is Limited by Ineffective Drug
Penetration into the Central Nervous System. Molecular cancer therapeutics 2015;14(12):2735-
43 doi 10.1158/1535-7163.MCT-15-0553.
10. Gupta SK, Kizilbash SH, Carlson BL, Mladek AC, Boakye-Agyeman F, Bakken KK, et al.
Delineation of MGMT Hypermethylation as a Biomarker for Veliparib-Mediated Temozolomide-
Sensitizing Therapy of Glioblastoma. J Natl Cancer Inst 2015;108(5) doi 10.1093/jnci/djv369.
11. Allen C, Opyrchal M, Aderca I, Schroeder MA, Sarkaria JN, Domingo E, et al. Oncolytic
measles virus strains have significant antitumor activity against glioma stem cells. Gene Ther
2013;20(4):444-9 doi 10.1038/gt.2012.62.
12. Sarkaria JN, Carlson BL, Schroeder MA, Grogan P, Brown PD, Giannini C, et al. Use of
an orthotopic xenograft model for assessing the effect of epidermal growth factor receptor
amplification on glioblastoma radiation response. Clinical cancer research : an official journal of
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 25
25
the American Association for Cancer Research 2006;12(7 Pt 1):2264-71 doi 10.1158/1078-
0432.CCR-05-2510.
13. Karpel-Massler G, Ishida CT, Bianchetti E, Shu C, Perez-Lorenzo R, Horst B, et al.
Inhibition of Mitochondrial Matrix Chaperones and Antiapoptotic Bcl-2 Family Proteins Empower
Antitumor Therapeutic Responses. Cancer research 2017;77(13):3513-26 doi 10.1158/0008-
5472.CAN-16-3424.
14. Karpel-Massler G, Ishida CT, Bianchetti E, Zhang Y, Shu C, Tsujiuchi T, et al. Induction
of synthetic lethality in IDH1-mutated gliomas through inhibition of Bcl-xL. Nat Commun
2017;8(1):1067 doi 10.1038/s41467-017-00984-9.
15. Karpel-Massler G, Ramani D, Shu C, Halatsch ME, Westhoff MA, Bruce JN, et al.
Metabolic reprogramming of glioblastoma cells by L-asparaginase sensitizes for apoptosis in
vitro and in vivo. Oncotarget 2016 doi 10.18632/oncotarget.9257.
16. Karpel-Massler G, Horst BA, Shu C, Chau L, Tsujiuchi T, Bruce JN, et al. A Synthetic
Cell-Penetrating Dominant-Negative ATF5 Peptide Exerts Anticancer Activity against a Broad
Spectrum of Treatment-Resistant Cancers. Clinical cancer research : an official journal of the
American Association for Cancer Research 2016;22(18):4698-711 doi 10.1158/1078-
0432.CCR-15-2827.
17. Karpel-Massler G, Ba M, Shu C, Halatsch ME, Westhoff MA, Bruce JN, et al.
TIC10/ONC201 synergizes with Bcl-2/Bcl-xL inhibition in glioblastoma by suppression of Mcl-1
and its binding partners in vitro and in vivo. Oncotarget 2015;6(34):36456-71 doi
10.18632/oncotarget.5505.
18. Pareja F, Macleod D, Shu C, Crary JF, Canoll PD, Ross AH, et al. PI3K and Bcl-2
inhibition primes glioblastoma cells to apoptosis through downregulation of Mcl-1 and Phospho-
BAD. Molecular cancer research : MCR 2014;12(7):987-1001 doi 10.1158/1541-7786.MCR-13-
0650.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 26
26
19. Karpel-Massler G, Ramani D, Shu C, Halatsch ME, Westhoff MA, Bruce JN, et al.
Metabolic reprogramming of glioblastoma cells by L-asparaginase sensitizes for apoptosis in
vitro and in vivo. Oncotarget 2016;7(23):33512-28 doi 10.18632/oncotarget.9257.
20. Elgendy M, Abdel-Aziz AK, Renne SL, Bornaghi V, Procopio G, Colecchia M, et al. Dual
modulation of MCL-1 and mTOR determines the response to sunitinib. J Clin Invest 2016 doi
10.1172/JCI84386.
21. Sheridan C, Brumatti G, Elgendy M, Brunet M, Martin SJ. An ERK-dependent pathway
to Noxa expression regulates apoptosis by platinum-based chemotherapeutic drugs. Oncogene
2010;29(49):6428-41 doi 10.1038/onc.2010.380.
22. Wang R, Xia L, Gabrilove J, Waxman S, Jing Y. Sorafenib Inhibition of Mcl-1
Accelerates ATRA-Induced Apoptosis in Differentiation-Responsive AML Cells. Clinical cancer
research : an official journal of the American Association for Cancer Research 2016;22(5):1211-
21 doi 10.1158/1078-0432.CCR-15-0663.
23. Kiprianova I, Remy J, Milosch N, Mohrenz IV, Seifert V, Aigner A, et al. Sorafenib
Sensitizes Glioma Cells to the BH3 Mimetic ABT-737 by Targeting MCL1 in a STAT3-
Dependent Manner. Neoplasia 2015;17(7):564-73 doi 10.1016/j.neo.2015.07.003.
24. Xiao ZD, Han L, Lee H, Zhuang L, Zhang Y, Baddour J, et al. Energy stress-induced
lncRNA FILNC1 represses c-Myc-mediated energy metabolism and inhibits renal tumor
development. Nat Commun 2017;8(1):783 doi 10.1038/s41467-017-00902-z.
25. Nagaraja S, Vitanza NA, Woo PJ, Taylor KR, Liu F, Zhang L, et al. Transcriptional
Dependencies in Diffuse Intrinsic Pontine Glioma. Cancer cell 2017;31(5):635-52 e6 doi
10.1016/j.ccell.2017.03.011.
26. Dey S, Sayers CM, Verginadis, II, Lehman SL, Cheng Y, Cerniglia GJ, et al. ATF4-
dependent induction of heme oxygenase 1 prevents anoikis and promotes metastasis. J Clin
Invest 2015;125(7):2592-608 doi 10.1172/JCI78031.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 27
27
27. Fels DR, Koumenis C. The PERK/eIF2alpha/ATF4 module of the UPR in hypoxia
resistance and tumor growth. Cancer biology & therapy 2006;5(7):723-8.
28. Pan R, Ruvolo V, Mu H, Leverson JD, Nichols G, Reed JC, et al. Synthetic Lethality of
Combined Bcl-2 Inhibition and p53 Activation in AML: Mechanisms and Superior Antileukemic
Efficacy. Cancer cell 2017;32(6):748-60 e6 doi 10.1016/j.ccell.2017.11.003.
29. Nifoussi SK, Ratcliffe NR, Ornstein DL, Kasof G, Strack S, Craig RW. Inhibition of
protein phosphatase 2A (PP2A) prevents Mcl-1 protein dephosphorylation at the Thr-163/Ser-
159 phosphodegron, dramatically reducing expression in Mcl-1-amplified lymphoma cells. The
Journal of biological chemistry 2014;289(32):21950-9 doi 10.1074/jbc.M114.587873.
30. Liu L, Eisenman RN. Regulation of c-Myc Protein Abundance by a Protein Phosphatase
2A-Glycogen Synthase Kinase 3beta-Negative Feedback Pathway. Genes Cancer
2012;3(1):23-36 doi 10.1177/1947601912448067.
31. Wei W, Shin YS, Xue M, Matsutani T, Masui K, Yang H, et al. Single-Cell
Phosphoproteomics Resolves Adaptive Signaling Dynamics and Informs Targeted Combination
Therapy in Glioblastoma. Cancer cell 2016;29(4):563-73 doi 10.1016/j.ccell.2016.03.012.
32. Prados MD, Byron SA, Tran NL, Phillips JJ, Molinaro AM, Ligon KL, et al. Toward
precision medicine in glioblastoma: the promise and the challenges. Neuro-oncology 2015 doi
10.1093/neuonc/nov031.
33. Lee DH, Ryu HW, Won HR, Kwon SH. Advances in epigenetic glioblastoma therapy.
Oncotarget 2017;8(11):18577-89 doi 10.18632/oncotarget.14612.
34. Chen S, Zhang Y, Zhou L, Leng Y, Lin H, Kmieciak M, et al. A Bim-targeting strategy
overcomes adaptive bortezomib resistance in myeloma through a novel link between autophagy
and apoptosis. Blood 2014;124(17):2687-97 doi 10.1182/blood-2014-03-564534.
35. Gammoh N, Lam D, Puente C, Ganley I, Marks PA, Jiang X. Role of autophagy in
histone deacetylase inhibitor-induced apoptotic and nonapoptotic cell death. Proceedings of the
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 28
28
National Academy of Sciences of the United States of America 2012;109(17):6561-5 doi
10.1073/pnas.1204429109.
36. Ishida CT, Shu C, Halatsch ME, Westhoff MA, Altieri DC, Karpel-Massler G, et al.
Mitochondrial matrix chaperone and c-myc inhibition causes enhanced lethality in glioblastoma.
Oncotarget 2017;8(23):37140-53 doi 10.18632/oncotarget.16202.
37. Ishida CT, Bianchetti E, Shu C, Halatsch ME, Westhoff MA, Karpel-Massler G, et al.
BH3-mimetics and BET-inhibitors elicit enhanced lethality in malignant glioma. Oncotarget
2017;8(18):29558-73 doi 10.18632/oncotarget.16365.
38. Satoh K, Yachida S, Sugimoto M, Oshima M, Nakagawa T, Akamoto S, et al. Global
metabolic reprogramming of colorectal cancer occurs at adenoma stage and is induced by
MYC. Proceedings of the National Academy of Sciences of the United States of America
2017;114(37):E7697-E706 doi 10.1073/pnas.1710366114.
39. Mertz JA, Conery AR, Bryant BM, Sandy P, Balasubramanian S, Mele DA, et al.
Targeting MYC dependence in cancer by inhibiting BET bromodomains. Proceedings of the
National Academy of Sciences of the United States of America 2011;108(40):16669-74 doi
10.1073/pnas.1108190108.
40. Wang J, Wang H, Li Z, Wu Q, Lathia JD, McLendon RE, et al. c-Myc is required for
maintenance of glioma cancer stem cells. PloS one 2008;3(11):e3769 doi
10.1371/journal.pone.0003769.
41. Allen JE, Prabhu VV, Talekar M, van den Heuvel AP, Lim B, Dicker DT, et al. Genetic
and Pharmacological Screens Converge in Identifying FLIP, BCL2, and IAP Proteins as Key
Regulators of Sensitivity to the TRAIL-Inducing Anticancer Agent ONC201/TIC10. Cancer
research 2015;75(8):1668-74 doi 10.1158/0008-5472.CAN-14-2356.
42. Sheng Z, Li L, Zhu LJ, Smith TW, Demers A, Ross AH, et al. A genome-wide RNA
interference screen reveals an essential CREB3L2-ATF5-MCL1 survival pathway in malignant
glioma with therapeutic implications. Nature medicine 2010;16(6):671-7 doi 10.1038/nm.2158.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 29
29
43. Tong J, Tan S, Zou F, Yu J, Zhang L. FBW7 mutations mediate resistance of colorectal
cancer to targeted therapies by blocking Mcl-1 degradation. Oncogene 2017;36(6):787-96 doi
10.1038/onc.2016.247.
Figures and figure legends:
Figure 1: Dual inhibition of HDAC (Pb) and BRD (OTX) causes synergistic reduction in
cellular proliferation of a broad range of glioblastoma model systems in vitro.
A, Established glioblastoma cells, U87, U87-EGFRvIII, T98G and LN229, patient-derived
xenograft cells, GBM14 and stem-like GBM cells, NCH644, were treated with OTX, Pb or the
combination over a broad range of concentrations for 72h. Thereafter, cells were analyzed for
cellular viability: OTX (blue), Pb (red) and the combination OTX+Pb (green). Shown are means
and SD. n=3 biological replicates. All concentrations are in M. #: Combination vs OTX, p<0.05;
+: Combination vs Pb, p<0.05; -: Combination vs OTX or Pb, p>0.05. B, Combination index (CI)
is plotted for the cells treated as in A. A CI value of less than 1.0 indicates synergy, whereas a
CI value larger than 1.0 shows antagonism. A CI value of 1.0 defines additivity. The average CI
value of all data points is provided in the upper portion of each diagram.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 30
30
Figure 2: Dual inhibition of HDAC (Pb) and BRD (OTX) causes synergistic activation of
apoptosis in glioblastoma cells
A, B, LN229 GBM cells were treated with indicated drugs for 24h, stained with Annexin V/PI and
analyzed by flow cytometry. n=3 biological replicates. Shown are means and SD. C, Microarray
was performed with NCH644 stem-like GBM cells treated with the combination of OTX+Pb and
vehicle DMSO control. Shown is GSEA enrichment plot with FDR q-values (false discovery
rate), NES (normalized enrichment score) and p-values. n=2 biological replicates for each
condition. D, E, LN229 GBM cells were treated with indicated drugs and stained with
tetramethylrhodamine, ethyl ester (TMRE) and analyzed by flow cytometry for the change in
mitochondrial membrane potential. n=3 biological replicates. Shown are means and SD. F,
Western blotting analysis of LN229 GBM cells treated with indicated drugs. TF: Total form, CF:
cleaved form. All concentrations are in M. G, H, LN229, U87, NCH644 or GBM14 cells were
treated with indicated drugs and analyzed for the levels of the indicated proteins by conventional
western blotting or capillary electrophoresis. All concentrations are in M. Blots or capillary
electrophoresis were quantified for the levels of Mcl-1 and Noxa normalized with its related
loading control. I, J, LN229 or NCH644 GBM cells were treated with indicated drugs and
analyzed by real-time PCR for the indicated makers. Shown are means and SD (n=3), and
statistical analysis was performed.
Figure 3: Molecular requirements of apoptosis induction by combined inhibition of HDAC
and BRD.
A, LN229 GBM cells were transfected with non-targeting (siNT), BAK, Noxa siRNA. Indicated
protein levels were shown. B, C, The same transfected LN229 GBM cells from A were subjected
to the combination treatment, OTX+Pb, stained with propidium iodide for flow cytometric
analysis. D, T98G GBM cells were treated with vehicle or the OTX+Pb combination for 16h.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 31
31
Thereafter, protein lysates were immunoprecipitated with a species control IgG or Mcl-1 specific
antibody and detected by conventional western blotting. The left portion shows the
immunoprecipitated lysates, whereas on the right side 1% input lysates were loaded. All
concentrations are in M. E, LN229 GBM cells were transfected with indicated siRNA for 3
days, then treated with indicated drugs for 24h and analyzed for the expression of the indicated
proteins by either standard western blotting or capillary electrophoresis. Arrows indicate total
and cleavage forms of PARP. All concentrations are in M. F, LN229 cells treated as in E were
stained with propidium iodide and analyzed by flow cytometry for DNA – fragmentation. n=3
biological replicates. Shown are means and SD. G, H, LN229 or NCH644 GBM cells were
treated with indicated drugs for 24h and 48h and analyzed for the indicated proteins by capillary
electrophoresis. All concentrations are in M. Quantifications are provided for Mcl-1 normalized
with vinculin. J, LN229 GBM cells were treated with indicated drugs and stained with Annexin
V/PI and analyzed by flow cytometry. Displayed are representative flow plots and quantification
of the results. n=3 biological replicates. Shown are means and SD.
Figure 4: Metabolic reprogramming of energy metabolism elicited by combined inhibition
of HDAC and BRD.
A, Microarray was performed with NCH644 stem-like GBM cells treated with the combination of
OTX+Pb and DMSO control. Shown are GSEA enrichment plots with FDR q-values (false
discovery rate), NES (normalized enrichment score) and p-values. n=2 biological replicates for
each condition. B, GBM14, LN229 and NCH644 GBM cells were treated as indicated and
analyzed for the expression of the indicated proteins by capillary electrophoresis. * indicates 14-
3-3 protein. C, NCH644 stem-like GBM cells were treated with indicated drugs and analyzed by
real-time PCR for the levels of c-Myc. n=3 biological replicates. Shown are means and SD. D,
GSEA enrichment plot for MYC targets from the same cells in A. FDR q-values (false discovery
rate), NES (normalized enrichment score) and p-values. E, GBM14 GBM cells were treated as
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 32
32
indicated for 24h and ATP levels were determined. n=3 biological replicates. Shown are means
and SD. F, GBM14 GBM cells were treated with indicated drugs and subjected to extracellular
flux analysis for oxygen consumption rate (OCR) in the context of a mitochondrial stress assay.
O: Oligomycin, F: FCCP, A/R: Antimycin/rotenone. Shown are means and SD, n=3 biological
replicates. G, GBM14 cells were treated as in F and subjected to extracellular flux analysis in
the setting of glycolysis stress test. G: Glucose, O: Oligomycin, 2-DG: 2-Deoxyglucose. Shown
are means and SD, n=3 biological replicates. H, I, The glycolytic capacity and glycolytic reserve
were calculated based on the experiment, shown in G. Shown are means and SD. n=3
biological replicates. J-M, Functional OXPHOS related parameters were calculated based on
the experiment shown in F, ATP production, basal OCR, maximal respiration and spare
respiratory capacity. Shown are means and SD. n=3 biological replicates.
Figure 5: Dual inhibition of HDAC (Pb) and BRD (OTX) causes endoplasmic reticulum
stress and mediates enhanced stability of Mcl-1 in a GSK3 dependent manner.
A, B NCH644 stem-like GBM cells were treated with the combination of OTX and Pb and
subjected to microarray analysis. Shown are GSEA enrichment plots with with FDR q-values
(false discovery rate), NES (normalized enrichment score) and p-values. n=2 biological
replicates for each condition. C, NCH644 and LN229 GBM cells were treated with indicated
drugs, and Real-time PCR analysis was performed for the expression of the ER-stress related
transcripts. Shown are means and SD. n=3 biological replicates. D, U87 and LN229 cells were
treated with indicated drugs for 7h and analyzed for the protein expression of ER-stress related
markers (ATF3 and Vinculin were run on capillary electrophoresis). All concentrations are in
M. E, LN229 GBM cells were transfected with non-targeting or PERK specific siRNA for 3 days
following with the treatment of indicated drugs for 7h, and analyzed for expression of the
indicated makers by conventional western blotting. All concentrations are in M. F, U87 and
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 33
33
LN229 GBM cells were transfected with non-targeting or ATF3 specific siRNA for 3 days
following with the treatment of indicated drugs for 7h, and analyzed by capillary electrophoresis
for the indicated markers. All concentrations are in M. G, H, LN229 or T98G GBM cells were
pretreated with combined OTX and Pb or Ctrl, and then cycloheximide (CHX) was added for
indicated time (minutes). Lysates were analyzed for the expression of the indicated markers by
capillary electrophoresis. I, GBM14 and LN229 GBM cells were treated with indicated drugs and
analyzed for the expression of the indicated markers. J, GBM14 and LN229 GBM cells were
transduced with indicated lentivirus coding non-targeting or c-myc shRNA (sh: shRNA), c-myc
overexpression (O/E: overexpression). Whole cell protein lysates were analyzed for the
indicated markers. K, GBM14 or LN229 GBM cells were transduced as indicated in J and
analyzed for the mRNA levels of c-Myc and Mcl-1.
Figure 6: Dual inhibition of HDAC (Pb) and BRD (OTX) causes synergistic reduction in
tumor growth and extends host survival of animals bearing orthotopic patient derived-
glioblastoma in the presence of the FDA-approved multi-kinase inhibitor, sorafenib.
A, B, U87-EGFRvIII GBM cells were implanted in the subcutis of immunocompromised mice
(Nu/Nu). After establishment of tumors, groups were formed as indicated. Tumor volumes were
plotted. n=9-15 biological replicates. C, D, GBM12 patient-derived xenograft cells were
implanted in the subcutis of immunocompromised mice (Nu/Nu). After establishment of tumors,
groups were formed as indicated. Tumor volumes were plotted. n=6-12 biological replicates. E,
F, G Shown are H&E, Ki67, and Tunel stained representative sections from a vehicle or
OTX015+panobinostat+sorafenib treated GBM12 tumor (from D). H, Quantifications of mitosis
and apoptotic/necrotic cells from the GBM12 tumor sections collected at day 20 post tumor
transplantation. 5 fields of each section were counted. Shown are means and SD. I, Stem-like
GBM cells, NCH644, were implanted in the right striatum of nude mice. Thereafter, mice were
randomly assigned to indicated groups. Treatments were performed three times a week for 3
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 34
34
weeks. Kaplan-meier survival curves were plotted. Median survival days of each group were:
DMSO: 24.5, OTX: 26, Pb: 23, OTX+Pb: 25 and OTX+Pb+Sf: 34. J, Kaplan-meier survival
curves of mice xenograft with GBM12 patient derived GBM cells. Treatments were performed
three times a week for 5 weeks until animals became moribund. Median survival days of each
group were: DMSO: 23, OTX: 31, Pb: 31, OTX+Pb: 35 and OTX+Pb+Sf: 50. K, Graphical
summary of the proposed mechanisms of action by the OTX, Pb and Sf drug combinations.
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 35
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 36
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 37
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 38
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 39
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 40
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260
Page 41
Published OnlineFirst May 15, 2018.Clin Cancer Res Yiru Zhang, Chiaki Tsuge Ishida, Wataru Ishida, et al. Lethality in GlioblastomaReprograms Tumor Cell Metabolism and elicits Synthetic Combined HDAC and Bromodomain Protein Inhibition
Updated version
10.1158/1078-0432.CCR-18-0260doi:
Access the most recent version of this article at:
Material
Supplementary
http://clincancerres.aacrjournals.org/content/suppl/2018/05/15/1078-0432.CCR-18-0260.DC1
Access the most recent supplemental material at:
Manuscript
Authoredited. Author manuscripts have been peer reviewed and accepted for publication but have not yet been
E-mail alerts related to this article or journal.Sign up to receive free email-alerts
Subscriptions
Reprints and
[email protected] at
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)
.http://clincancerres.aacrjournals.org/content/early/2018/05/15/1078-0432.CCR-18-0260To request permission to re-use all or part of this article, use this link
Research. on September 8, 2020. © 2018 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 15, 2018; DOI: 10.1158/1078-0432.CCR-18-0260