The National Cancer Institute ALMANAC: A Comprehensive ... · To date, over 100 small-molecule oncology drugs have been approved by the FDA. Because of the inherent heterogeneity

Therapeutics, Targets, and Chemical Biology

The National Cancer Institute ALMANAC:A Comprehensive Screening Resource for theDetection of Anticancer Drug Pairs withEnhanced Therapeutic ActivitySusan L. Holbeck1, Richard Camalier1, James A. Crowell1, Jeevan Prasaad Govindharajulu2,Melinda Hollingshead1, Lawrence W. Anderson1, Eric Polley1, Larry Rubinstein1,Apurva Srivastava2, Deborah Wilsker2, Jerry M. Collins1, and James H. Doroshow1,3

Abstract

To date, over 100 small-molecule oncology drugs have beenapproved by the FDA. Because of the inherent heterogeneity oftumors, these small molecules are often administered in combi-nation to prevent emergence of resistant cell subpopulations.Therefore, new combination strategies to overcome drug resis-tance in patients with advanced cancer are needed. In this study,weperformed a systematic evaluationof the therapeutic activity ofover 5,000 pairs of FDA-approved cancer drugs against a panel of60 well-characterized human tumor cell lines (NCI-60) to uncov-er combinations with greater than additive growth-inhibitoryactivity. Screening results were compiled into a database, termedthe NCI-ALMANAC (A Large Matrix of Anti-Neoplastic Agent

Combinations), publicly available at https://dtp.cancer.gov/ncialmanac. Subsequent in vivo experiments in mouse xenograft mod-els of human cancer confirmed combinations with greater thansingle-agent efficacy. Concomitant detection of mechanistic bio-markers for these combinations in vivo supported the initiation oftwo phase I clinical trials at the NCI to evaluate clofarabine withbortezomib and nilotinib with paclitaxel in patients withadvanced cancer. Consequently, the hypothesis-generatingNCI-ALMANAC web-based resource has demonstrated value inidentifying promising combinations of approved drugs withpotent anticancer activity for further mechanistic study and trans-lation to clinical trials. Cancer Res; 77(13); 3564–76. �2017 AACR.

IntroductionDespite recent advances in translational oncology (1), new

therapies for cancer are urgently needed, as are new approaches tofacilitate the rapid translation of promising preclinical data toclinical evaluation. Recognizing that combining anticancer drugs,whether cytotoxic or molecularly targeted, with distinct mechan-isms of action is the approach most likely to overcome single-agent resistance and produce sustained clinical remissions (2, 3),we determined that it would be of value to systematically screenpairwise combinations of 104 FDA-approved oncology drugs inthe NCI-60 panel of human tumor cell lines. Our objective wasto create a database, the NCI-ALMANAC (A Large Matrix of

Anti-Neoplastic Agent Combinations), that could be searched toidentify combinations with greater antitumor activity than eitheragent alone. Promising combinations could then be selected forfurther evaluation in vitro or in human tumor xenografts models.By screening only approved drugs with proven activity andestablished safety profiles, combinations identified from theNCI-ALMANACdatabase offer potential for rapid translation intoInvestigational New Drug–exempt clinical trials.

The NCI-60 panel is one of many drug developmentresources that are, and will continue to be, supported by theNCI; each of the cell lines has been extensively characterized atthe molecular level; exome sequence, mRNA expression,miRNA expression, protein quantification, protein modifica-tion, DNA methylation, enzyme activity, and metabolomicsdata are all publicly available (4, 5). These datasets enablemolecular data mining in the context of combination drugstudies to assess the engagement of multiple potential targetsand/or downstream markers of drug action, both in animalmodel systems and in patients during early-phase clinical trials(6). They also support research into the multiple mechanismsdriving resistance to cancer therapeutic agents, including muta-tion or amplification of the gene encoding the drug target,enhanced drug efflux or metabolism, activation of compensa-tory signaling networks that bypass the effects of target engage-ment, changes in DNA damage response or epigenetic path-ways, or alterations in the tumor microenvironment (7).

Although a variety of active clinical trials are testing drugcombinations based on accepted precepts of tumor cell biology,

1Division of Cancer Treatment and Diagnosis, National Cancer Institute, NIH,Bethesda, Maryland. 2Clinical Pharmacodynamics Program, Applied/Develop-mental ResearchDirectorate, LeidosBiomedical Research Inc., FrederickNation-al Laboratory for Cancer Research, Frederick, Maryland. 3Center for CancerResearch, National Cancer Institute, NIH, Bethesda, Maryland.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Corresponding Author: James H. Doroshow, Division of Cancer Treatment andDiagnosis, National Cancer Institute, NIH, 31 Center Drive, Bldg. 31 Room 3A-44,Bethesda, MD 20892. Phone: 301-496-4291; Fax: 301-496-0826; E-mail:[email protected]

doi: 10.1158/0008-5472.CAN-17-0489

�2017 American Association for Cancer Research.

CancerResearch

Cancer Res; 77(13) July 1, 20173564

on June 1, 2020. © 2017 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst April 26, 2017; DOI: 10.1158/0008-5472.CAN-17-0489

http://crossmark.crossref.org/dialog/?doi=10.1158/0008-5472.CAN-17-0489&domain=pdf&date_stamp=2017-6-17

https://dtp.cancer.gov/ncialmanac


http://cancerres.aacrjournals.org/

the preclinical hypotheses supporting many of these studies maybe difficult to validate in patientswithout clearly definedmechan-isms of action, resistance, and toxicity for each component of thecombination, as well as for the combination itself (8). Indeed, theheterogeneity observed across individual tumors hints at thepossibility of distinct cooccurring resistance mechanisms withina patient, such that different regions of the samemalignancy mayvary in their sensitivity (or resistance) to an individual agent at thetime that treatment is initiated (9–11). Hence, developing newcombination strategies to overcome drug resistance in patientswith advanced cancer is a therapeutic imperative (8, 12). Prom-ising drug combinations may be discovered with synthetic lethalscreens (13) utilizing siRNA or shRNA libraries, or CRISPR-Cas9genome editing (14), to identify targets for commonly usedanticancer drugs (15, 16); however, in this study, we utilized anunbiased, "hypothesis-free" phenotypic screening approach (8).The results of our comprehensive combination screening effortare available in the NCI-ALMANAC database, a web-basedresource intended for hypothesis-generating assessments ofoncology drugs combinations. On the basis of the outcome ofthis screening program, several promising drug combinationswere examined in human tumor xenografts developed fromNCI-60 cell lines. From our in vitro-to-in vivo translation effort,two promising drug combinations never before tested in humanstudies, clofarabine with bortezomib and paclitaxel with niloti-nib, were chosen for clinical evaluation.

Materials and MethodsCell lines and drug screening

NCI-60 cell lines were obtained from the NCI DevelopmentalTherapeutics Program Tumor Repository in the mid-1990s toearly 2000s and distributed for drug screening and xenograftexperiments in 2010–2012. For each lot of cells, the Repositoryperformed Applied Biosystems AmpFLSTR Identifiler testing withPCR amplification to confirm consistency with the publishedIdentifiler STR profile (17) for the given cell line. Each cell linewas tested for mycoplasma when it was accepted into the repos-itory; routinemycoplasma testing of lotswas not performed. Cellswere kept in continuous culture at NCI for no more than 20passages.

Oncology drugs were obtained through commercial sources.Experiments were performed at three locations (NCI's FrederickNational Laboratory for Cancer Research, SRI International, andUniversity of Pittsburgh). At NCI, drug treatment was performedaccording to the standard NCI-60 testing protocol (https://dtp.cancer.gov/discovery_development/nci-60/default.htm), using96-well plates and 2-day drug exposures; the endpoint is deter-mination of Sulphorhodamine B levels in duplicate wells. Foreach drug pair, both single agents were tested at 5 concentrations.For drug combinations, one agent was tested at 5 concentrationstogether with one of 3 different concentrations of the secondagent, generating 5 � 3 combination matrices. At SRI Interna-tional and University of Pittsburgh, this protocol was modified:cells were plated in 384-well plates, FBS was increased to 10%,CellTiter-Glo was used as an endpoint, and drugs were tested at 3concentrations as single agents and in 3 � 3 concentrationmatrices for combinations, in individual wells. Drug concentra-tions were chosen based on clinical relevance; where possible,concentrations were selected to be below the human peak plasmaconcentration (Cmax) at the FDA-approved clinical dose and also

to yield measurable activity in our assay systems. In a few cases,one or more concentrations above the Cmax were selected toachieve some growth inhibition in vitro. Growth percent wascalculated as described in the standard NCI-60 testing protocol.Inclusion of a time zero measurement in all experiments allowedfor determination of cell killing and growth inhibition. Fifty-nineof the NCI-60 cell lines passed predefined quality control para-meters for >90% of the drug pairs. The SK-MEL-2 melanoma celllinewas an outlier; data for only 65%of the drug pairsmet qualitycontrol standards as the cells grew poorly in 384-well plates atboth testing locations.

Development and use of the NCI ComboScoreDetermination of combination benefit ("ComboScore") uti-

lized a modification of Bliss independence (18), as follows: LetYi

ApBq be the growth fraction for the ith cell line exposed to the pth

concentration of drug A and the qth concentration of drug B,defined as:

YApBq

i ¼ 100� TApBq

1 � T0T01 � T0

where T0 is the time zero endpoint measurement, TApBq1 is the

endpoint measurement after 2 days, and T01 is the endpoint

measurement after 2 days for the control well. Define YiAp; Yi

Bq

as the growth fractions when only exposed to drug A or drug B,respectively. The expected growth fraction for the combination is:

ZApBq

i ¼min YAp

i ; YBq

i

� �if YAp

i � 0 or YBq

i � 01

100~YAp

i � ~YBq

i

� �otherwise

8<:

where ~Yi ¼ minðYi; 100Þ truncates the growth fraction at 100.The final combination score for the cell line and the drug com-bination is the sum of the differences in expected versus observedgrowth fractions:

YABi ¼

Xp; q

YApBq

i � ZApBq

i

These calculations set a ceiling of 100% growth, that is, theexpected combination growth percentage cannot be greaterthan the control. One screening site had a high incidence of"reversals," where the growth percent increased as the concen-tration of one or both drugs was increased, that is, apparentenhancement of growth by drug treatment. While the phenom-enon of hormesis can be real, the frequency with which this wasobserved, at only one of the screening locations, argued againstthis being actual hormesis. All data for a drug pair/cell line wereremoved from the dataset if the growth percent increased by50% or greater between any 2 adjacent doses of either drug.ComboScores are available for download at https://dtp.cancer.gov/ncialmanac.

Evaluation of combination therapies registered withclinicaltrials.gov

For initial analysis of the clinical status of all NCI-ALMANACcombinations, a drug pair was designated "clinically tested" if thetwo names were provided as an intervention in the same triallisted in the clinicaltrials.gov database (see Supplementary Meth-ods for further detail).

NCI ALMANAC of Approved Cancer Drug Combinations

www.aacrjournals.org Cancer Res; 77(13) July 1, 2017 3565



https://dtp.cancer.gov/discovery_development/nci-60/default.htm

https://dtp.cancer.gov/discovery_development/nci-60/default.htm




HeatmapJMP 11 (SAS institute) software was utilized for preparation of

heatmaps and statistical analyses. Two-way hierarchical clusteringutilized the Ward method with data standardization and withoutimputation of missing values.

Human tumor xenograftsAnimal experiments were performed at both SRI International

and Frederick National Laboratory for Cancer Research. Both SRIInternational and Frederick National Laboratory are accredited byAAALAC International and follow the PublicHealth Service Policyfor the Care and Use of Laboratory Animals. Animal care wasprovided in accordance with the procedures outlined in the"Guide for Care and Use of Laboratory Animals" (NationalResearch Council; 2011; National Academies Press; Washington,D.C.).

Formouse inoculation, tumor cells from theNCI-60panelwereused at the 4th to 6th in vitro passage of cryopreserved cell stocks.Cells (1 � 107 cells/0.1 mL/injection) were subcutaneously inoc-ulated bilaterally into female athymic nu/nu NCr mice, andtherapeutic studies were initiated upon reaching a target volumeof 100mm3. Sample sizeswere n¼ 10–20mice for vehicle-treatedgroups and n ¼ 5–10 mice for single-agent- or combination-treated groups. Mice were treated with drug doses that had beendemonstrated to be the MTD for the single agents and thecombination in prior experiments by the NCI DevelopmentalTherapeutics Program (see Supplementary Data for dosage regi-mens); in some studies, treatments at lower doses were alsoperformed. For each mouse, the time from treatment initiationto tumor volume doubling was estimated using exponentialgrowth interpolation when the event was between two timepoints. Mice for which tumor volume had not doubled by theend of studywere considered censored at the last observation. Thedistribution of time-to-tumor-volume-doubling was estimatedfor each treatment cohort using the Kaplan–Meier method, andthe log-rank test was used to determine significant differencesbetween combination and single-agent treatment groups.

Biomarker analyses of human tumor xenograftsAnalyses of apoptosis (19) and DNA damage response bio-

markers (20, 21) were performed as described previously (seeSupplementary Methods for details).

ResultsThe NCI-ALMANAC dataset and web tools

Pairwise combinations of 104 FDA-approved anticancer drugs(Supplementary Table S1)were screened in eachof the cell lines ofthe NCI-60 panel to discover those with enhanced growth inhi-bition or cytotoxicity profiles compared with either single agent(Fig. 1A). Combination activity was reported as a "ComboScore"(22); positiveComboScores indicated greater than additive in vitroactivity and negative scores indicated less-than-additive activity(see the Materials and Methods). A total of 5,232 drug pairs wereevaluated in each of the cell lines; 304,549 experiments wereperformed to test each drug at either 9 or 15 combination dosepoints, for a total of 2,809,671 dose combinations. The NCI-ALMANAC is publicly available at https://dtp.cancer.gov/ncialmanac; the data can be visualized as a heatmap summarizing theentire dataset, as a bar graph of the ComboScores for a particulardrug pair in all cell lines, or as dose–response curves for one drug

combination in a given cell line (Supplementary Fig. S1). Finally,users can obtain an overview of which drugs or mechanisticcategories of drugs perform best in combination with a givendrug of interest.

We first researched what fraction of the NCI-ALMANAC drugpairs appeared on clinicaltrials.gov for an estimation of clinicalexperience (see Materials and Methods for details). At the start ofthis study in June 2011, nearly 75% of the > 5,000 pairs of FDA-approved drugs examined had not been reported to clinicaltrials.gov as being part of an ongoing or completed study (Fig. 1B).Next, to identify patterns in cellular sensitivity and combinationactivity, we conducted a two-dimensional hierarchical clusteringanalysis. The resulting heatmap of ComboScores across cell linesand drug combinations revealed that most drug combinationshave selective activity for a subset of cell lines, as indicated by theblue–red heterogeneity throughout most of the heatmap,although some combinations did exhibit patterns of predomi-nantly greater than additive (highly red) or predominantly less-than-additive (highly blue) activity acrossmost of the NCI-60 celllines (Fig. 2A and B). In addition, except for leukemia cell lines,sensitivity to drug combinations appeared to be independent ofcellular histology, as cell lines from the same source tissue did notcluster together.

Examination of promising anticancer drug combinations fromthe NCI-ALMANAC in human tumor xenografts

Because it was not feasible to examine all drug combinations infollow-up animal model experiments, a subset of combinationswith greater than additive activity was selected for in vivo analysisbased on theComboScore, the ability of relevant cell lines to growas xenograft implants, and clinical investigator input regardingclinical utility. An initial study was performed using doses at ornear the MTD of each single agent to eliminate those combina-tions with excessive toxicity (requiring dose reduction >50%). Forthe remaining combinations, the more clinically pertinent crite-rion of "greater-than-single-agent" activity was applied to thefollow-up xenograft analyses to identify combinations that couldpotentially confer additional efficacy in patients. We also differ-entiated between "novel" combinations (those that had not beenpreviously tested together exclusively) and "non-novel" combi-nations (those tested together exclusively in a clinical trial arm),according to clinicaltrials.gov and the literature.

Various methods have been used to assess drug efficacy inxenograft studies, although to our knowledge, no method hasbeen validated for comparing results of a high-throughput xeno-graft screen. For our purposes, we chose Kaplan–Meier and log-rank analyses of time-to-tumor-volume-doubling. Out of 13"non-novel" combinations that were selected for xenograft anal-ysis based on high in vitro ComboScores (Supplementary TableS2), 5 yielded greater-than-single-agent efficacy in at least one ofthe xenograft models tested. Furthermore, each of these 5 com-binations has demonstrated appreciable efficacy (partial and/orcomplete responses) in clinical trials (Supplementary Table S3).Despite the small number of combinations examined, this anal-ysis suggested that the NCI-ALMANAC results may be used as astarting point to select novel combinations for further study andpotential clinical use.

Given these promising results, 20 novel combinations wereevaluated for greater-than-single-agent efficacy in one or morexenograft models derived from NCI-60 cell lines, yielding a totalof 44 unique "combination þ model" experiments (in several

Holbeck et al.

Cancer Res; 77(13) July 1, 2017 Cancer Research3566



https://dtp.cancer.gov/ncialmanac;

https://dtp.cancer.gov/ncialmanac;


cases, multiple schedules and/or doses were examined for each"combinationþmodel" set; in these cases, only the best outcomewas scored). The analysis revealed that 14 of the 44 experimentsyielded significantly improved antitumor efficacy for the combi-nation compared with both single-agents (P < 0.05; Fig. 3;Supplementary Table S4). In addition, apanel of xenograft expertsdesignated experiments for which the tumor regression appearedsuperior for the combination compared with the respective singleagents, despite lack of statistical significance in log-rank analysisof time-to-tumor-doubling; 7 such experiments were identified(Fig. 3; Supplementary Table S4). For the remaining experiments,the combination did not yield significantly improved efficacycompared with either single agent (Supplementary Table S4; Fig.3); in 2 experiments, this was due to substantial single-agentefficacy (i.e., � 20% of mice treated with the single agent expe-rienced a tumor doubling). Following this analysis, two of thecombinations that exhibited significantly improved antitumoractivity compared to their respective single agents were selectedfor further efficacy and mechanistic analysis.

Clofarabine and bortezomibThe purine nucleoside analogue clofarabine and the protea-

some inhibitor bortezomib play important roles in the treatmentof hematopoietic malignancies; however, neither agent had pre-viously been demonstrated to exhibit appreciable activity against

human solid tumors (23, 24).On thebasis of greater than additiveactivity in vitro (Fig. 4A), we tested this combination in five solidtumor xenograft models and determined that the combinationcaused significant tumor regressions in two colon cancer models,HCT-116 (P < 0.01) and HT29 (P < 0.001), and in the non–smallcell lung cancer model NCI-H522 (P < 0.01; Fig. 4B; Supplemen-tary Table S4). In HCT-116 xenografts treated with bortezomib(0.75 mg/kg), clofarabine (60 mg/kg), or both agents in combi-nation on an every two day (Q2D) � 5 dosing schedule for twotreatment cycles, the combination had significantly greater effi-cacy than either single agent (P < 0.01) as determined by log-rankanalysis of the time-to-tumor-doubling (Supplementary TableS4); in a second experiment, a single treatment cycle at thesedoses yielded slightly greater-than-single-agent efficacy (P <0.1; Fig. 4B; Supplementary Table S4). This combination waswell tolerated, with no observed median body weight loss > 30%(data not shown). In contrast, although the clofarabine andbortezomib combination had greater than additive efficacy inthe ovarian cancer cell lineOVCAR-3 and themelanoma lineM14in vitro (Fig. 4A), neither single-agent nor combination treatmentelicited antitumor effects in the corresponding xenograft models(Fig. 4C; Supplementary Table S4).

Because both bortezomib and clofarabine have been demon-strated to induce apoptosis in a variety of human tumor cell lines(25, 26), we examined whether the activity of the combination in

Figure 1.

NCI-ALMANAC screen strategy and summary results. A, Workflow diagram for the NCI-ALMANAC screen and follow-up preclinical and clinical studies.Double-headed arrows indicate the iterative flow of information between different stages of the project. B, NCI-ALMANAC reveals a large number ofpotentially active, clinically novel drug combinations (according to clinicaltrials.gov analysis). Data were downloaded from clinicaltrials.gov and processed toidentify the cancer trials that had utilized the drug combinations examined in this study. Each drug is shown as both the abscissa and ordinate, alphabetically,with green indicating drug pairs that were identified as having been reported as tested in combination in a clinical trial and blue indicating pairs that have not.






the treatment-responsive HCT-116 xenograft models might beexplained by an enhanced modulation of apoptotic cascadeproteins. To this end, tumor samples from one responsive(HCT-116) and one nonresponsive (M14) model were collectedat various time points after treatment with clofarabine, bortezo-mib, or the combination, and tumor extracts were surveyed forchanges in the apoptotic pathway using a 15-biomarker apoptosismultiplex panel (ref. 19 and Supplementary Methods). On thebasis of results from a pilot time-course study (data not shown),we selected time points at 6 and 24 hours postdose-5 for subse-quent experiments.

In the responsive HCT-116 model, bortezomib alone hadvirtually no effect compared with vehicle on apoptosis markersat 6 hours, while clofarabine alone significantly altered 3 ofthe 5 examined apoptosis markers (cleaved caspase-3 and

mitochondrial/nuclear lamin-B and survivin) relative to vehicle(Fig. 5A). In contrast, compared with both vehicle and bortezo-mib alone, the bortezomib–clofarabine combination significantaltered levels of all 5 of themeasured apoptosismarkers: cytosolicsurvivin, cleaved caspase-3, and mitochondrial/nuclear lamin-B,BAX, and survivin (Fig. 5A). These trends persisted at the 24-hourposttreatment time point (Supplementary Fig. S2A). In addition,Western blot analysis showed that treatment of HCT-116 xeno-graft models with clofarabine alone or with the combinationyielded high levels of p53 upregulation/stabilization relative tovehicle control (Supplementary Fig. S2B). These results in thesensitive HCT-116 model contrasted with those observed for thenonresponsive M14 model, in which bortezomib–clofarabinecombination treatment did not produce this apoptosis-associatedbiomarker signature (Fig. 5B; Supplementary Fig. S2C). Overall,

●C

CR

F-C

EM

●M

OLT

-4●

SR

●H

L-60

(TB

)●

HO

P-6

2●

SF-

268

●S

F-29

5●

SK

-ME

L-5

●U

AC

C-6

2●

K-5

62●

TK-1

0●

HS

-578

T●

MD

A-M

B-4

68●

EK

VX

●O

VC

AR

-4●

NC

I-H32

2M●

NC

I-H22

6●

SK

-ME

L-28

●H

CT-

15●

UA

CC

-257

●N

CI-H

522

●B

T-54

9●

OV

CA

R-5

●O

VC

AR

-8●

M14

●M

DA

-MB

-435

●H

CC

-299

8●

SW

-620

●N

CI/A

DR

-RE

S●

HO

P-9

2●

NC

I-H23

●A

CH

N●

CA

KI-1

●S

K-O

V-3

●78

6-0

●R

PM

I 822

6●

HT2

9●

MC

F7●

SN

B-1

9●

SN

12C

●D

U-1

45●

RX

F 39

3●

MD

A-M

B-2

31●

A54

9/A

TCC

●N

CI-H

460

●H

CT-

116

●C

OLO

205

●M

ALM

E-3

M●

T-47

D●

IGR

OV

1●

PC

-3●

UO

-31

●K

M12

●S

F-53

9●

A49

8●

U25

1●

OV

CA

R-3

Dru

g co

mbi

natio

nsA

B

1.5

14.018.6

49.4

13.5

3.00

10

20

30

40

50

60

0 1–5 6–10 11–30 31–50 >50

Perc

enta

ge o

f dru

g pa

irs

Number of cell lines in which ComboScore > 0

Figure 2.

In vitro activity of drug combinationsin the NCI-60 panel. A, Two-wayhierarchical clustering was used tocluster ComboScores for 57 cell lines(the SK-MEL-2, SNB-75, and LOX-IMVIlines hadmanymissing values andwereexcluded) and 4,629 drugcombinations. Red, positiveComboScores (i.e., better-than-additive effects; blue, negative values(i.e., less-than-additive effects). Tumortissue derivations for the NCI-60 celllines are indicated by colored circles:red, leukemia; green, colon; blue,non–small cell lung; gray, CNS; orange,melanoma; purple, ovarian; yellow,renal; turquoise, prostate; and pink,breast.B, Themajority of drug pairs hadin vitro activity in 11–30 cell lines of theNCI-60 panel, indicating that most drugpairs only have greater than additiveactivity in a subset of the NCI-60cell lines.

Holbeck et al.





in the sensitive HCT-116 xenograft model, the intrinsic apoptoticpathway was activated primarily by clofarabine, and the additionof bortezomib enhanced this activation.

Clofarabine is known to induce DNA damage response signal-ing (27); therefore, we evaluated the effects of these treatments onmarkers of DNA damage by immunofluorescence analysis oftumor segments using a validated assay (20, 21). The combina-tion yielded a significantly greater induction of gH2AX comparedwith both bortezomib alone and clofarabine alone at the 6-hourtime point (Fig. 5C and F), the latter indicating a role forbortezomib in potentiating clofarabine-induced DNA damage.Such substantial changes in gH2AX levelswere not observed in thenonresponsive M14 model (Fig. 5C, right). The DNA damagemarker pNbs1 was significantly increased in the clofarabine- andcombination-treated groups relative to the vehicle-treated groupas early as 2 hours (data not shown), at 6 hours, and at 24 hours in

HCT-116, but not M14, models (Fig. 5D and F). Finally, themitoticmarker histoneH3–pSer10 (pHH3)was lower in both theclofarabine-only and combination groups relative to vehicle at 6and 24 hours in the HCT-116 model (Fig. 5E and F), but was notconsistently lower in M14 xenografts (Fig. 5E), suggesting thatsustained cell-cycle arrest was specific to the responsive model. Asexpected, treatment with bortezomib alone did not induce mar-kers of DNA damage or cell-cycle arrest (Fig. 5C–F). Together,these data provide mechanistic evidence that DNA damage, cell-cycle arrest, and apoptosis contribute to the efficacy of the borte-zomib–clofarabine combination in the responsive HCT-116xenograft model.

Paclitaxel and nilotinibThe combination of the microtubule-stabilizing agent pacli-

taxel and the BCR-Abl kinase inhibitor nilotinib also hadnot beenexplored clinically. All 6 hematopoietic tumor cell lines tested hadpositive ComboScores for the paclitaxel–nilotinib combination,and this drug combination gave positive ComboScores in threeof the triple-negative breast cancer NCI-60 cell lines (Hs578T,MDA-MB-468, and BT-549), but negative scores in the two estro-gen receptor–positive breast cancer lines (MCF7 and T-47D; Fig.6A). On the basis of these data, the effects of the nilotinib–paclitaxel combination on breast cancer xenografts derived fromtriple-negative cell lines were examined. Nilotinib–paclitaxel waswell tolerated and highly efficacious in the MDA-MB-468 xeno-graftmodel; the combination (at either of the examined paclitaxeldoses) was superior to both single agents (Fig. 6B; SupplementaryTable S4). Strikingly, the combination treatment resulted incomplete tumor regression, with no tumor regrowth observedfor over 80 days following the end of therapy (day 134; Fig. 6B).Single-agent nilotinib had no significant effects on tumor growthcompared with vehicle, and single-agent paclitaxel yielded onlypartial tumor control, with tumor regrowth commencing aroundday 65–75 (Fig. 6B). Hs578T and BT-549 grew poorly as xeno-grafts and were therefore not assessable in vivo, so we insteadexamined two additional triple-negative models derived fromnon�NCI-60 cell lines, SUM-52 PE and SUM-149PT. In both ofthese models, the nilotinib–paclitaxel combination yieldedgreater-than-single-agent efficacy (Supplementary Table S4).An additional triple-negative NCI-60 cell line, MDA-MB-231,did not yield valid in vitro screen results for the nilotinib–paclitaxel combination due to a "reversal" profile (see Meth-ods) and also did not yield greater-than-single-agent efficacywhen tested as a xenograft model (Supplementary Table S4).Beyond breast cancer models, the nilotinib–paclitaxel combi-nation also had greater-than-single-agent efficacy in the RXF393 renal carcinoma model, in line with its greater thanadditive activity in the corresponding cell line in vitro (Supple-mentary Table S4). In contrast, the U251 cell line, where thecombination was less efficacious in vitro (ComboScore of 21),yielded no greater-than-single-agent efficacy when tested as axenograft model (Supplementary Table S4).

To identify potential biomarkers of response to this combina-tion, we first queried the NCI-60 molecular markers database tolook for associations between in vitro ComboScore and drug-relatedmolecular signatures.We foundno significant correlationsbetween nilotinib-paclitaxel ComboScores and either: (i) proteinand RNA expression levels for the non-BCR-Abl kinases known tobe inhibited by nilotinib (28), or (ii) activity and RNA expressionlevels of the ABC family efflux pumps that have been reported to

(13; 54%)

14(32%)

7(16%)

2(4%)

21(48%)

Combo has > single-agent activity (time-to doubling)

Combo has > single-agent activity (expert panel)

Highly efficacious single agent

Combo no better than single agents

Figure 3.

Antitumor efficacy of NCI-ALMANAC–derived drug combinations in the testedNCI-60 xenograft models. A total of 44 unique xenograft efficacy experiments,involving 20 different drug combinations that had not previously undergoneclinical testing (according to clinicaltrials.gov), were performed using sets ofdrug combinations and NCI-60–derived models that demonstrated positivein vitro ComboScores. Experiments in which combination treatment yieldedimproved antitumor efficacy comparedwith both single-agent treatments wereidentified by Kaplan–Meier and log-rank analysis of time-to-tumor-volume-doubling (P < 0.05; dark pink) or by a panel of xenograft experts (light pink).Experiments with no greater-than-single-agent activity are shown in darkblue, while those for which one single agent was highly efficacious are shown inlight blue. See Supplementary Table S4 for a complete list of xenograftexperiments and results.






–200

–100

0

100

200

300

HL-

60(T

B)

K-5

62M

OLT

-4R

PM

I-822

6S

RA

549/

ATC

CE

KV

XH

OP

-62

HO

P-9

2N

CI-H

226

NC

I-H23

NC

I-H32

2MN

CI-H

460

NC

I-H52

2C

OLO

205

HC

C-2

998

HC

T-11

6H

CT-

15H

T29

KM

12S

W-6

20S

F-26

8S

F-29

5S

F-53

9S

NB

-75

SN

B-1

9U

251

LOX

IMV

IM

ALM

E-3

MM

14M

DA

-MB

-435

SK

-ME

L-2

SK

-ME

L-28

SK

-ME

L-5

UA

CC

-257

UA

CC

-62

IGR

OV

1O

VC

AR

-3O

VC

AR

-4O

VC

AR

-5O

VC

AR

-8N

CI/A

DR

-RE

SS

K-O

V-3

786-

0A

498

AC

HN

CA

KI-1

RX

F 39

3S

N12

CTK

-10

UO

-31

PC

-3D

U-1

45M

CF7

MD

A-M

B-2

31/A

TCC

HS

578

TB

T-54

9T-

47D

MD

A-M

B-4

68

Com

boS

core

CC

RF-

CE

M

0

200

400

600

800

1,000

1,200

15 20 25 30 35 40 45

Med

ian

tum

or v

olum

e (m

m3 )

Days post implantation

0

200

400

600

800

1,000

1,200

5 10 15 20 25 30 35 40 45 50 55

Med

ian

tum

or v

olum

e (m

m3 )


A

BHCT-116 M14

VehicleBortezomib (0.35)Bortezomib (0.75)Bortezomib (1)Clofarabine (60)Clofarabine (100)Clofarabine (60) + bortezomib (0.35)Clofarabine (60) + bortezomib (0.75)Clofarabine (100) + bortezomib (0.35)Clofarabine (100) + bortezomib (0.75)

C

D


Pro

babi

lity

of e

vent

-free

sur

viva

l

0 10 20 30 40 500

0.2

0.4

0.6

0.8

1.0

#

BreastCNSColonLeukemia NSC Lung Melanoma Ovarian ProstateRenal

Figure 4.

Therapeutic activity of the NCI-ALMANAC–derived combination of bortezomib and clofarabine. A, The combination of bortezomib and clofarabine in vitroyielded positive ComboScores across several cell types. Tumor tissue derivations for the NCI-60 cell lines are indicated by bar color: red, leukemia; green, colon; blue,non-small cell lung; gray, CNS; orange, melanoma; purple, ovarian; yellow, renal; turquoise, prostate; and pink, breast. A leukemia cell line for which no ComboScorewas available for this combination is indicated by gray text. Models for which xenograft experiments revealed the presence or absence of in vivo greater-than-single-agent activity are indicated by filled triangles or empty triangles, respectively. B and C, Clofarabine–bortezomib combination treatment exhibits enhancedefficacy relative to the respective single-agent treatments in the human colon cancer HCT-116 xenograft model (B) but not in the M14 melanoma xenograftmodel (C). Median tumor volumes are shown for mice treated with vehicle, single-agent bortezomib [5 � every two days (Q2D), intraperitoneal injection],single-agent clofarabine [5 � every two days (Q2D), oral administration], or the combination (schedule and administration for each agent as indicated for therespective single-agent treatments); treatment commenced onday6 for theHCT-116 experiment (B) and onday 17 for theM14 experiment (C; dosing period indicatedby gray-shaded area in both graphs). D, Kaplan–Meier curves for time-to-tumor doubling analysis of HCT-116 xenograft models treated with the clofarabine–bortezomib combination. The y-axis indicates probability of event-free survival, where an event is defined as one tumor doubling or drug-related death(the latter occurred for only one animal in the 60mg/kg clofarabineþ0.75mg/kg bortezomib-treated group). #, the combination-treated group exhibiting greater-than-single-agent time-to-tumor-doubling compared with both bortezomib-alone and clofarabine-alone groups at the corresponding doses (P < 0.1), asdetermined by log-rank tests. Doses (mg/kg) for each treatment group are indicated in the legend (n¼ 16mice for the vehicle group; n¼ 8mice for all other groups).Error bars, SE of the median.

Holbeck et al.





A B

γH2A

X

(% n

ucle

ar a

rea

posi

tive)

0

20

80

60

40

6 h 24 h 6 h 24 h

HCT-116 M14

D

pNbs

1(%

nuc

lear

are

a po

sitiv

e)

0

20

80

60

40

6 h 24 h 6 h 24 h

HCT-116 M14

E

pHH

3(%

pos

itive

cel

ls)

6 h 24 h 6 h 24 h

HCT-116 M14

0

1

4

3

2

5

VehicleBortezomibClofarabineBortezomib + clofarabine

Veh

icle

Bor

tezo

mib

Clo

fara

bine

Bor

tezo

mib

+ c

lofa

rabi

ne

γH2AX pNbs1 pHH3

CF

Per

cent

cha

nge

rela

tive

to v

ehic

le

0

50

200

150

100

300

250

350

Cytosolic Survivin

Cleaved Caspase-3

Mito.+Nuc. Lamin-B

Mito.+Nuc. BAX

Mito.+Nuc. Survivin

Per

cent

cha

nge

rela

tive

to v

ehic

le

0

50

200

150

100

250

Cytosolic Survivin

Cleaved Caspase-3

Mito.+Nuc. Lamin-B

Mito.+Nuc. BAX

Mito.+Nuc. Survivin

M14HCT-116VehicleBortezomibClofarabineBortezomib + clofarabine

Figure 5.

Bortezomib–clofarabine combination treatment modulates markers of apoptosis and DNA damage in responsive, but not unresponsive, xenograft models. Micebearing HCT-116 or M14 xenografts were treated with either vehicle, 0.75 mg/kg bortezomib [5 � every two days (Q2D), intraperitoneally], 60 mg/kgclofarabine [5 � every two days (Q2D), oral administration], or the combination of bortezomib and clofarabine with the same dosage regimens. Tissues wereharvested at 6 and 24 hours postdose 5 and processed into cell lysates (for apoptosis marker assays, A and B) or were formalin-fixed and paraffin-embedded forquantitative immunofluorescence analysis of DNA damage response and cell-cycle markers (C–F). Apoptosis markers were measured in the therapeuticallyresponsive HCT-116 (A) or nonresponsive M14 (B) xenograft models at 6 hours following the day 5 dose. Error bars, SEM; n ¼ 6 mice per treatment group.Statistically significant differences between the combination and single-agent treatment groups are indicated in red, while those compared with vehicle areindicated in black (� , P < 0.05; ��, P < 0.01; P values derived from unpaired, nonparametric Mann–Whitney tests). One outlier data point for caspase-3 in theclofarabine HCT-116 groupwas removed. Effects of drug treatment on nuclear levels of gH2AX (C), pNbs1 (D), or pHH3-Ser10 (E) were examined in HCT-116 (left) andM14 (right) models at 6 and 24 hours following the day 5 dose. Representative images from the HCT-116 6-hour group are shown (F). Error bars (C–E), SD (n¼ 6miceper treatment group, except for the HCT-116 bortezomib 24-hour group [n¼ 5 due to unsuitable tissue quality and theM14 bortezomib 6-hour group (n¼ 4)], and anaverage of 5,000 cells per mouse tumor were analyzed. Statistically significant differences between the combination and single-agent treatment groups areindicated (� , P < 0.05; �� , P < 0.01); the sole statistically significant greater-than-single-agent difference between the HCT-116 clofarabine group and thecombination group is bolded for emphasis (C). In the HCT-116 experiments (C–E, left), biomarker levels for both the clofarabine group and combination groupwere significantly different than vehicle at 6- and 24-hour time points (P < 0.01 vs. vehicle for all; P values derived from nonparametric Mann–Whitney tests).






–500

–300

–100

100

300C

CR

F-C

EM

HL-

60(T

B)

K-5

62M

OLT

-4R

PM

I-822

6S

RA

549/

ATC

CE

KV

XH

OP

-62

HO

P-9

2N

CI-H

226

NC

I-H23

NC

I-H32

2MN

CI-H

460

NC

I-H52

2C

OLO

205

HC

C-2

998

HC

T-11

6H

CT-

15H

T29

KM

12S

W-6

20S

F-26

8S

F-29

5S

F-53

9S

NB

-75

SN

B-1

9U

251

LOX

IMV

IM

ALM

E-3

MM

14M

DA

-MB

-435

SK

-ME

L-2

SK

-ME

L-28

SK

-ME

L-5

UA

CC

-257

UA

CC

-62

IGR

OV

1O

VC

AR

-3O

VC

AR

-4O

VC

AR

-5O

VC

AR

-8N

CI/A

DR

-RE

SS

K-O

V-3

786-

0A

498

AC

HN

CA

KI-1

RX

F 39

3S

N12

CTK

-10

UO

-31

PC

-3D

U-1

45M

CF7

MD

A-M

B-2

31/A

TCC

HS

578

TB

T-54

9T-

47D

MD

A-M

B-4

68

A

BreastCNSColonLeukemia NSC Lung Melanoma Ovarian ProstateRenal

Com

boS

core

HO

P-6

2H

OP

-92

SK

-ME

L-2

OV

CA

R-5

MD

A-M

B-2

31/A

TCC

% C

hang

e re

lativ

e to

veh

icle

0200400

Cyto. Cleaved Caspase-3

Mito.+Nuc. Mcl-1

Mito.+Nuc. BAX

BAX–Bcl-2Complex

600800

1,0001,2001,4001,6001,800 Vehicle

Nilotinib PaclitaxelNilotinib + paclitaxel

0

500

1,000

1,500

2,000

2,500

20 40 60 80 100 120

Med

ian

tum

or v

olum

e (m

m3 )


VehicleNilotinib (75)Paclitaxel (15)Paclitaxel (10)Nilotinib (75) + paclitaxel (15)Nilotinib (75) + paclitaxel (10)

MDA-MB-468 CB


Pro

babi

lity

of e

vent

-free

sur

viva

l

0 20 40 60 80 1000

0.2

0.4

0.6

0.8

1.0

120

**

D

Figure 6.

In vitro activity, in vivo efficacy, and mechanistic analysis of the NCI-ALMANAC–derived nilotinib–paclitaxel combination. A, The combination of nilotinib andpaclitaxel in vitro yielded positive ComboScores across several cell types. Tumor tissue derivations for the NCI-60 cell lines are indicated by bar color: red,leukemia; green, colon; blue, non–small cell lung; gray, CNS; orange,melanoma; purple, ovarian; yellow, renal; turquoise, prostate; and pink, breast. Cell lines forwhichno ComboScore data are available for this combination are indicated by gray text. Models for which xenograft experiments revealed the presence or absenceof in vivo greater-than-single agent activity are indicated by filled triangles or empty triangles, respectively. (Continued on the following page.)

Holbeck et al.





be inhibited by nilotinib (29–31) (Supplementary Fig. S3A). Tofurther examine the possibility of nilotinib-mediated reduction inpaclitaxel efflux via ABC transporter inhibition, we analyzedtumor paclitaxel levels in MDA-MB-468 xenograft models andfound comparable tumor paclitaxel concentrations in animalstreated with the nilotinib–paclitaxel combination and thosetreated with paclitaxel alone (Supplementary Fig. S3B), providingfurther evidence that the mechanism of action for this combina-tion does not involve modulation of paclitaxel efflux.

Because apoptosis, among other forms of cell death, has beenpostulated to play a role in tumor cell killing by paclitaxel and/ornilotinib (32, 33), we examined biomarkers of apoptosis inresponse to single-agent and combination treatment in theMDA-MB-468 breast cancer xenograft model. No increases in theapoptosis marker cleaved caspase-3 were observed in either sin-gle-agent or combination treatment groups compared with thevehicle control (Fig. 6D), supporting a nonapoptotic mechanismof antitumor activity for this drug pair. Corroborating thisfinding,levels of mitochondrial Mcl-1, associated with antiapoptoticactivity (34), were higher in the combination and single-agenttreatment groups comparedwith vehicle (Fig. 6D). Although totallevels of the canonically antiapoptotic BAX�Bcl-2 complex werehigher following nilotinib–paclitaxel combination treatment rel-ative to the vehicle and single-agent treatments (Fig. 6D), sub-cellular fractionation experiments revealed that this was primarilydue to an increase in ER-associated BAX�Bcl-2 (data not shown),which is involved in the perturbation of calciumhomeostasis thatoccurs in both apoptosis and necroptosis (35). Together, theseresults suggest that the nilotinib–paclitaxel combination likelysuppresses tumor growth through a caspase-3–independentmechanism of cell death.

DiscussionMore than 100 small-molecule drugs have been approved for

cancer therapy by the FDA (36), and are often used in combina-tion with other agents to address themultiple molecular pathwayaberrations, and consequent drug resistance, associated withmany malignancies. However, a broad range of potential FDA-approved drug combinations has never been evaluated clinically.We determined that it would be valuable to systematically screenall possible two-drug combinations in vitro to identify novel pairsof FDA-approved drugs with greater in vitro growth inhibitionthan would be expected from the sum of their single-agentactivities. A number of combinations identified in theNCI-ALMANAC screen were further evaluated in animal models

for greater-than-single-agent efficacy. Although combinationexperiments have been conducted before (16, 37–39), to ourknowledge, the NCI-ALMANAC is the most extensive anticancerdrug combination screening effort for which the data and searchtools are publicly available.

Our comprehensive in vitro screen identified 1,898 drug pairswith positive ComboScores in at least 1 cell line. Consequently,we performed in vivo testing of a small subset of drug pairs thathave not, to our knowledge, been clinically tested (Supplemen-tary Table S4). Agents were tested at or near their MTD to assesswhether combination therapy could achieve efficacy superior tothat of the single agents. Forty-eight percent of these novel, NCI-ALMANAC–derived combinations exhibited greater-than-single-agent activity in at least one model (Fig. 3), suggesting that theNCI-ALMANAC may be helpful in identifying additional combi-nations with clinical efficacy in the future.

The bortezomib–clofarabine combination conferred an anti-tumor activity advantage over single-agent clofarabine or borte-zomib in the HCT-116 xenograft model (Fig. 4B). In view of theknown induction of apoptosis in response to single-agent borte-zomib (40) or clofarabine (27), we investigated the effects of thiscombination on apoptotic pathway markers. We found that theantitumor activity of the clofarabine–bortezomib combinationwas accompanied by several significant apoptosis-associatedmolecular changes relative to the single-agent- and/or vehicle-treated arms, including decreases in cytosolic andmitochondrial/nuclear survivin and mitochondrial/nuclear-associated lamin-B,as well as increases in caspase-3 activation and mitochondrial/nuclear-associated BAX (Fig. 5A). In contrast, for the nonrespon-sive M14 xenograft model, in which no antitumor efficacy advan-tagewas observed for the combination,weobserved no consistentalterations in levels of these apoptotic biomarkers, suggesting thatthe apoptotic pathway activation observed in the responsiveHCT-116 model is associated with the antitumor efficacy of the clofar-abine–bortezomib combination. The mechanism whereby clo-farabine–bortezomib treatment reduces survivin levels in HCT-116 xenografts could be either p53-dependent (41), as indicatedby our Western blot analysis (Supplementary Fig. S1E), or p53-independent (42). Regardless, the greater-than-single-agent activ-ity of this combination has some precedent; previous in vitroexperiments have demonstrated that bortezomib-induced cyto-toxicity and proapoptotic signaling are enhanced by purinenucleoside analogues (43).

Because clofarabine is both a DNA-damaging agent and a DNAsynthesis inhibitor (27), we evaluated the role of DNA damage in

(Continued.) B, Nilotinib/paclitaxel combination treatment exhibits enhanced efficacy relative to the respective single-agent treatments in the MDA-MB-468triple-negative breast cancer xenograft model. Median tumor volumes are shown for mice treated with vehicle, single-agent nilotinib [75 mg/kg every day (QD)�5, oral administration for four cycles beginning on days 25, 32, 39, and 46], single-agent paclitaxel [on one of two dose schedules for two cycles beginning on days 25and 39: (i) 15mg/kg, every 7 days (Q7D)�2, i.v.; or (ii) 10mg/kg, every day (QD)�3, i.v.; or (iii) the respective combinations (schedule and administration as indicatedfor the respective single-agent treatments; dosing period indicated by gray-shaded area]. n¼ 15 mice for vehicle group, and n¼ 7–8 mice for all other groups. Errorbars, SE of the median. C, Kaplan–Meier curves for time-to-tumor doubling analysis of MDA-MB-468 xenograft models treated with the nilotinib–paclitaxelcombination. The y-axis indicates probability of event-free survival, where an event is defined as one tumor doubling. Asterisks denote combination-treated groupsexhibiting significantly greater-than-single-agent time-to-tumor-doubling compared with both nilotinib-alone and paclitaxel-alone groups at thecorresponding doses (� , P < 0.001), as determined by log-rank tests (n¼ 16mice for the vehicle group; n¼ 8mice for all other groups). Treatment groups, and doses(mg/kg) for each, are indicated in the legend in B. D, Nilotinib–paclitaxel combination treatment does not induce apoptotic pathway biomarkers in therapeuticallyresponsive MDA-MB-468 xenograft models. Mice were treated with either vehicle, single-agent nilotinib [75 mg/kg every day (QD)�19, orally], single-agentpaclitaxel [15 mg/kg, every 7 days (Q7D)�3, i.v.], or the combination (schedule and administration as indicated for the respective single-agent treatments).Tissues were harvested on day 15 following the start of treatment and processed into cell lysates, which were then subjected to subcellular fractionation formeasurement of the indicated apoptosis markers. Error bars, SEM; n ¼ 6 mice per treatment group. Statistically significant differences between thecombination and single-agent treatment groups are indicated in purple, while those compared with vehicle are indicated in black (�, P < 0.05; �� , P < 0.01);P values derived from nonparametric Mann–Whitney tests.






the activity of the bortezomib-clofarabine combination. The com-bination and clofarabine alone both significantly increased levelsof the DNA damage response markers gH2AX and pNbs1 anddecreased levels of the cell proliferationmarker pHH3 inHCT-116xenografts compared with vehicle and single-agent bortezomibtreatment (Fig. 5C–F); gH2AX levels were also significantly higherin the combination arm relative to the clofarabine-only arm at the6-hour time point (Fig. 5C). Hence, the ability of clofarabine toinduce DNA damage (which could then trigger p53-mediatedapoptosis) and/or directly alter apoptotic pathway components(27) might lower the apoptosis threshold in HCT-116 xenografts,rendering these tumorsmore susceptible to the cytotoxic effects ofbortezomib-induced proteasome inhibition.

Although the pharmacodynamic effects of clofarabine andbortezomib converge on apoptosis, potential upstream interac-tions remain to be explored; for example, it is possible thatbortezomib-mediated proteasome inhibition stabilizes a proteinrequired for the uptake, recycling/stability, mechanism of action,or activating phosphorylation of clofarabine. A whole-proteomeanalysis in Jurkat cells revealed that bortezomib treatment yieldeda reproducible 12- to 13-fold upregulation of nucleoside diphos-phate kinase, which catalyzes the final step of clofarabine activa-tion, as well as more modest upregulation of the clofarabine-activating enzyme deoxycytidine kinase and the clofarabine tar-gets DNA polymerase-a and e (27, 44). Another possibility is thatproteasome inhibition by bortezomib stabilizes DNA damageresponse proteins, therebypotentiating this response following itsinduction by clofarabine. Although these mechanistic detailsremain to be elucidated, the in vivo efficacy of the clofarabine–bortezomib combination, along with a prior NCI review of therelationship between efficacy in xenograft models and activity inhuman phase II trials (45) and the development of a validatedmultiplex assay for apoptotic pathway components (19), provid-ed support for initiating a phase I trial of this drug pair in patientswith advanced, refractory cancers (clinicaltrials.gov identifier:NCT02211755).

The novel NCI-ALMANAC–derived combination of nilotiniband paclitaxel exhibited greater-than-single-agent efficacy in ourexperiments with the MDA-MB-468 triple-negative breast cancermodel and others. Previous studies have found that nilotinibenhances intracellular accumulation of paclitaxel and other agentsin ABC transporter�overexpressing cell lines via inhibition of ABCtransporter�mediated drug efflux (46, 47). Furthermore, niloti-nib–paclitaxel combination treatment has been shown to yieldgreater-than-single-agent efficacy in xenograft models overexpres-singABC transportersABCB1orABCC10(30).However, the lackofassociation between in vitro ABC transporter activity or expressionand nilotinib-paclitaxel ComboScore (Supplementary Fig. S3A)suggests that ABC transporter inhibition may not account for thegreater than additive activity of this combination that we observed.The comparable tumor paclitaxel concentrations in MDA-MB-468xenograft models treated with paclitaxel alone versus those treatedwith the nilotinib–paclitaxel combination (Supplementary Fig.S3B) cast further doubt on this hypothesis for the enhanced activityof this combination. Another potentialmechanismof actionmightbe the known inhibition by nilotinib of the cytochrome P450s(including CYP2C8) involved in themetabolismof paclitaxel (48),which would increase systemic paclitaxel exposure; however,we also found comparable plasma paclitaxel concentrationsin MDA-MB-468 xenograft models treated with paclitaxelalone versus the nilotinib-paclitaxel combination (data not

shown), rendering this hypothesized mechanism of action alsounlikely.

Recent studies have demonstrated that both paclitaxel andnilotinib induce a variety of overlapping apoptotic and necrop-totic cell deathmechanisms (32, 33), that might contribute to theefficacy of this drug pair. Our analysis of apoptotic pharmacody-namic biomarkers revealed the absence of apoptosis-associatedchanges in cleaved caspase-3 or Mcl-1 levels following nilotinib–paclitaxel combination treatment in MDA-MB-468 xenografts,suggesting that an alternative mechanism of cell death may beresponsible for the antitumor activity of this combination. Pac-litaxel has previously been shown to induce both apoptosis andnecrosis in a cell-cycle stage–specific manner (49), and a numberof kinase inhibitors are known to induce necrosis (50), suggestinga potential role for nilotinib in shifting the apoptotic–necroptoticbalance in paclitaxel-induced cell death. Regardless, the impres-sive in vitro activity and in vivo efficacy of this combination havemotivated phase 1 evaluation of this drug pair (clinicaltrials.govidentifier: NCT02379416).

Despite the limitations inherent in the translation of in vitroactivity into in vivo efficacy (due in part to the size of the screen,which precluded comprehensive investigation of the effect ofdifferent drug administration doses and schedules), and the factthat large-scale animal studies could not easily be conducted tovalidate all of our screening results, theNCI-ALMANAC generatedinteresting hypotheses that were further addressed with mecha-nistic biomarker profiling and xenograft efficacy studies. Toexplore drug combinations of interest identified in the NCI-ALMANAC screen, we enriched our xenograft testing for pairsthat had not been studied previously in a clinical trial (as definedby searching clinicaltrials.gov). Because only FDA-approved smallmolecules were used in the NCI-ALMANAC screen, our resultshave the potential for rapid clinical translation. Thus, theNCI-ALMANAC should prove to be a valuable resource that willinform the development and selection of novel combinations ofFDA-approved anticancer agents for future clinical trials.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

DisclaimerThe content of this publication does not necessarily reflect the views or

policies of the Department of Health andHuman Services, nor doesmention oftrade names, commercial products, or organizations imply endorsement by theU.S. Government.

Authors' ContributionsConception and design: S.L. Holbeck, J.A. Crowell, L. Rubinstein, J.M. Collins,J.H. DoroshowDevelopment of methodology:M.G. Hollingshead, E.C. Polley, L. Rubinstein,A.K. Srivastava, D.F. Wilsker, J.M. Collins, J.H. DoroshowAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): J.P. Govindharajulu, M.G. Hollingshead, L.W. Ander-son, A.K. Srivastava, J.H. DoroshowAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): S.L. Holbeck, R. Camalier, J.P. Govindharajulu, L.W.Anderson, E.C. Polley, L. Rubinstein, A.K. Srivastava, D.F. Wilsker, J.M. Collins,J.H. DoroshowWriting, review, and/or revision of the manuscript: S.L. Holbeck, J.P. Govind-harajulu, L.W. Anderson, E.C. Polley, L. Rubinstein, A.K. Srivastava, D.F.Wilsker, J.M. Collins, J.H. DoroshowAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): S.L. Holbeck, R. Camalier, J.P. Govindharajulu,A.K. Srivastava, J.M. Collins, J.H. Doroshow


Holbeck et al.




Study supervision: J.A. Crowell, M.G. Hollingshead, A.K. Srivastava, J.M.Collins

AcknowledgmentsThe authors wish to thank the University of Pittsburgh and SRI International

for acquisition of in vitro data, SRI for performing follow-up xenograft experi-ments, Dr. Karen Schweikart for excellent project management, Dr. RobertKinders for helpful discussion regarding the biomarker analyses, Dr. DanielZaharevitz for construction of the NCI-ALMANAC website, and Drs. MariamKonat�e and Sarah Miller, Kelly Services, for editorial assistance in the prepara-tion of this manuscript.

Grant SupportThis project has been funded in whole or in part with federal funds from the

National Cancer Institute, NIH, under contract no.HHSN261200800001E. Thisresearch was also supported in part by the Intramural Research Program of theNIH, National Cancer Institute, Center for Cancer Research and in part byAmerican Recovery and Reinvestment Act funds.

The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received February 15, 2017; revised April 13, 2017; accepted April 24, 2017;published OnlineFirst April 26, 2017.

References1. Doroshow JH, Kummar S. Translational research in oncology-10 years

of progress and future prospects. Nat Rev Clin Oncol 2014;11:649–62.2. Devita VT, Schein PS. Drugs in combination for treatment of cancer -

rationale and results. N Engl J Med 1973;288:998–1006.3. Chen S-h, Lahav G. Two is better than one; toward a rational design of

combinatorial therapy. Curr Opin Struct Biol 2016;41:145–50.4. Holbeck SL, Collins JM, Doroshow JH. Analysis of Food and Drug Admin-

istration-approved anticancer agents in the NCI60 panel of human tumorcell lines. Mol Cancer Ther 2010;9:1451–60.

5. National Cancer Institute.NCI-60 human tumor cell lines screen. Availablefrom: https://dtp.cancer.gov/discovery_development/nci-60.

6. Doroshow JH, Parchment RE. Oncologic phase 0 trials incorporatingclinical pharmacodynamics: from concept to patient. Clin Cancer Res2008;14:3658–63.

7. Rebucci M, Michiels C. Molecular aspects of cancer cell resistance tochemotherapy. Biochem Pharmacol 2013;85:1219–26.

8. Carragher NO, Brunton VG, Frame MC. Combining imaging and pathwayprofiling: an alternative approach to cancer drug discovery. Drug DiscovToday 2012;17:203–14.

9. Alifrangis CC, McDermott U. Reading between the lines: understandingdrug response in the post genomic era. Mol Oncol 2014;8:1112–19.

10. Gerlinger M, Horswell S, Larkin J, Rowan AJ, Salm MP, Varela I, et al.Genomic architecture and evolution of clear cell renal cell carcinomasdefined by multiregion sequencing. Nat Genet 2014;46:225–33.

11. Jamal-Hanjani M, Quezada SA, Larkin J, Swanton C. Translationalimplications of tumor heterogeneity. Clin Cancer Res 2015;21:1258–66.

12. Wang LH, Pfister TD, Parchment RE, Kummar S, Rubinstein L, Evrard YA,et al. Monitoring drug-induced gammaH2AX as a pharmacodynamicbiomarker in individual circulating tumor cells. Clin Cancer Res 2010;16:1073–84.

13. You J, Li Y, Fang N, Liu B, Zu L, Chang R, et al. MiR-132 suppresses themigration and invasion of lung cancer cells via targeting the EMT regulatorZEB2. PLoS One 2014;9:e91827.

14. Schmitt SM, Gull M, Brandli AW. Engineering Xenopus embryos forphenotypic drug discovery screening. Adv Drug Del Rev 2014;69–70:225–46.

15. Astsaturov I, Ratushny V, Sukhanova A, EinarsonMB, Bagnyukova T, ZhouY, et al. Synthetic lethal screen of an EGFR-centered network to improvetargeted therapies. Sci Signal 2010;3:ra67.

16. Mathews Griner LA, Guha R, Shinn P, Young RM, Keller JM, Liu D, et al.High-throughput combinatorial screening identifies drugs that cooperatewith ibrutinib to kill activated B-cell-like diffuse large B-cell lymphomacells. Proc Natl Acad Sci U S A 2014;111:2349–54.

17. Lorenzi PL, Reinhold WC, Varma S, Hutchinson AA, Pommier Y, ChanockSJ, et al. DNA fingerprinting of the NCI-60 cell line panel. Mol Cancer Ther2009;8:713–24.

18. Bliss CI. The toxicity of poisons applied jointly. Ann Appl Biol 1939;26:585–615.

19. Srivastava AK, Jaganathan S, Stephen L, Hollingshead MG, Layhee A,Damour E, et al. Effect of a Smac mimetic (TL32711, Birinapant) on theapoptotic program and apoptosis biomarkers examined with validatedmultiplex immunoassays fit for clinical use. Clin Cancer Res 2016;22:1000–10.

20. Do K, Wilsker D, Ji J, Zlott J, Freshwater T, Kinders RJ, et al. Phase I study ofsingle-agent AZD1775 (MK-1775), a Wee1 kinase inhibitor, in patientswith refractory solid tumors. J Clin Oncol 2015;33:3409–15.

21. LoRusso PM, Li J, Burger A, Heilbrun LK, Sausville EA, Boerner SA, et al.Phase I safety, pharmacokinetic, and pharmacodynamic study of the poly(ADP-ribose) polymerase (PARP) inhibitor veliparib (ABT-888) in com-bination with irinotecan in patients with advanced solid tumors. ClinCancer Res 2016;22:3227–37.

22. Drewinko B, Loo TL, Brown B, Gottlieb JA, Freireich EJ. Combinationchemotherapy in vitro with adriamycin. Observations of additive, antag-onistic, and synergistic effects when used in two-drug combinations oncultured human lymphoma cells. Cancer Biochem Biophys 1976;1:187–95.

23. KantarjianHM,Gandhi V, Kozuch P, Faderl S, Giles F, Cortes J, et al. Phase Iclinical and pharmacology study of clofarabine in patients with solid andhematologic cancers. J Clin Oncol 2003;21:1167–73.

24. Huang Z, Wu Y, Zhou X, Xu J, Zhu W, Shu Y, et al. Efficacy of therapy withbortezomib in solid tumors: a review based on 32 clinical trials. FutureOncol 2014;10:1795–807.

25. Sayers TJ, Brooks AD, Koh CY, Ma WH, Seki N, Raziuddin A, et al. Theproteasome inhibitor PS-341 sensitizes neoplastic cells to TRAIL-mediatedapoptosis by reducing levels of c-FLIP. Blood 2003;102:303–10.

26. Genini D, Adachi S, Chao Q, Rose DW, Carrera CJ, Cottam HB, et al.Deoxyadenosine analogs induce programmed cell death in chronic lym-phocytic leukemia cells by damaging the DNA and by directly affecting themitochondria. Blood 2000;96:3537–43.

27. Zhenchuk A, Lotfi K, Juliusson G, Albertioni F. Mechanisms of anti-canceraction and pharmacology of clofarabine. Biochem Pharmacol 2009;78:1351–59.

28. Manley PW,Drueckes P, Fendrich G, Furet P, Liebetanz J,Martiny-BaronG,et al. Extended kinase profile and properties of the protein kinase inhibitornilotinib. Biochim Biophys Acta 2010;1804:445–53.

29. Eadie LN, Hughes TP, White DL. Interaction of the efflux transportersABCB1 and ABCG2 with imatinib, nilotinib, and dasatinib. Clin Pharma-col Ther 2014;95:294–306.

30. Tiwari AK, Sodani K, Dai CL, Abuznait AH, Singh S, Xiao ZJ, et al. Nilotinibpotentiates anticancer drug sensitivity in murine ABCB1-, ABCG2-, andABCC10-multidrug resistance xenograft models. Cancer Lett 2013;328:307–17.

31. Villar VH, Vogler O, Martinez-Serra J, Ramos R, Calabuig-Farinas S,Gutierrez A, et al. Nilotinib counteracts P-glycoprotein-mediated multi-drug resistance and synergizes the antitumoral effect of doxorubicin in softtissue sarcomas. PLoS One 2012;7:e37735.

32. Eum KH, Lee M. Crosstalk between autophagy and apoptosis in theregulation of paclitaxel-induced cell death in v-Ha-ras-transformed fibro-blasts. Mol Cell Biochem 2011;348:61–68.

33. Shaker ME, Ghani A, Shiha GE, Ibrahim TM, Mehal WZ. Nilotinib inducesapoptosis and autophagic cell death of activated hepatic stellate cells viainhibition of histone deacetylases. Biochim Biophys Acta 2013;1833:1992–2003.

34. Clohessy JG, Zhuang J, de Boer J, Gil-Gomez G, Brady HJ. Mcl-1interacts with truncated Bid and inhibits its induction of cytochromec release and its role in receptor-mediated apoptosis. J Biol Chem2006;281:5750–59.





https://dtp.cancer.gov/discovery_development/nci-60


35. Zhivotovsky B, Orrenius S. Calcium and cell death mechanisms: A per-spective from the cell death community. Cell Calcium 2011;50:211–21.

36. CenterWatch. Available from: https://www.centerwatch.com/drug-information/fda-approved-drugs/therapeutic-area/12/oncology.

37. MillerML,Molinelli EJ, Nair JS, Sheikh T, SamyR, Jing X, et al. Drug synergyscreen and network modeling in dedifferentiated liposarcoma identifiesCDK4 and IGF1R as synergistic drug targets. Sci Signal 2013;6:ra85.

38. Schmidt L, Kling T, Monsefi N, Olsson M, Hansson C, Baskaran S, et al.Comparative drug pair screening across multiple glioblastoma cell linesreveals novel drug-drug interactions. Neuro Oncol 2013;15:1469–78.

39. Roller DG, Axelrod M, Capaldo BJ, Jensen K, Mackey A, Weber MJ, et al.Synthetic lethal screening with small-molecule inhibitors provides a path-way to rational combination therapies for melanoma. Mol Cancer Ther2012;11:2505–15.

40. Milano A, Perri F, Caponigro F. The ubiquitin-proteasome system as amolecular target in solid tumors: an update on bortezomib. OncoTargetsTherapy 2009;2:171–8.

41. Mirza A, McGuirk M, Hockenberry TN, Wu Q, Ashar H, Black S, et al.Human survivin is negatively regulated by wild-type p53 and participatesin p53-dependent apoptotic pathway. Oncogene 2002;21:2613–22.

42. BhatUG,Gartel AL.Differential sensitivity of human colon cancer cell linesto the nucleoside analogs ARC and DRB. Int J Cancer 2008;122:1426–29.

43. DuechlerM, Linke A,Cebula B, ShehataM, Schwarzmeier JD, Robak T, et al.In vitro cytotoxic effect of proteasome inhibitor bortezomib in combina-

tion with purine nucleoside analogues on chronic lymphocytic leukaemiacells. Eur J Haematol 2005;74:407–17.

44. Mertins P, Qiao JW, Patel J, Udeshi ND, Clauser KR, Mani DR, et al.Integrated proteomic analysis of post-translational modifications by serialenrichment. Nat Methods 2013;10:634–7.

45. Johnson JI, Decker S, Zaharevitz D, Rubinstein LV, Venditti J, Schepartz S,et al. Relationships between drug activity in NCI preclinical in vitro and invivo models and early clinical trials. Br J Cancer 2001;84:1424–31.

46. Tiwari AK, Sodani K, Wang SR, Kuang YH, Ashby CRJr., Chen X, et al.Nilotinib (AMN107, Tasigna) reverses multidrug resistance by inhibitingthe activity of the ABCB1/Pgp and ABCG2/BCRP/MXR transporters. Bio-chem Pharmacol 2009;78:153–61.

47. Shen T, Kuang YH, Ashby CR, Lei Y, Chen A, Zhou Y, et al. Imatinib andnilotinib reversemultidrug resistance in cancer cells by inhibiting the effluxactivity of the MRP7 (ABCC10). PLoS One 2009;4:e7520.

48. Kim MJ, Lee JW, Oh KS, Choi CS, Kim KH, Han WS, et al. The tyrosinekinase inhibitor nilotinib selectively inhibits CYP2C8 activitiesin human liver microsomes. Drug Metab Pharmacokinet 2013;28:462–67.

49. Liao PC, Lieu CH. Cell cycle specific induction of apoptosis and necrosis bypaclitaxel in the leukemic U937 cells. Life Sci 2005;76:1623–39.

50. MengMB,WangHH,Cui YL,WuZQ, Shi YY, ZaorskyNG, et al.Necroptosisin tumorigenesis, activation of anti-tumor immunity, and cancer therapy.Oncotarget 2016;7:57391–413.


Holbeck et al.



https://www.centerwatch.com/drug-information/fda-approved-drugs/therapeutic-area/12/oncology

https://www.centerwatch.com/drug-information/fda-approved-drugs/therapeutic-area/12/oncology


2017;77:3564-3576. Published OnlineFirst April 26, 2017.Cancer Res Susan L. Holbeck, Richard Camalier, James A. Crowell, et al. Enhanced Therapeutic Activity

withScreening Resource for the Detection of Anticancer Drug Pairs The National Cancer Institute ALMANAC: A Comprehensive

Updated version

10.1158/0008-5472.CAN-17-0489doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2017/04/25/0008-5472.CAN-17-0489.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/77/13/3564.full#ref-list-1

This article cites 48 articles, 16 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/77/13/3564.full#related-urls

This article has been cited by 20 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/77/13/3564To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-17-0489

http://cancerres.aacrjournals.org/content/suppl/2017/04/25/0008-5472.CAN-17-0489.DC1

http://cancerres.aacrjournals.org/content/77/13/3564.full#ref-list-1

http://cancerres.aacrjournals.org/content/77/13/3564.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/77/13/3564


Documents

The National Cancer Institute ALMANAC: A Comprehensive ... · To date, over 100 small-molecule oncology drugs have been approved by the FDA. Because of the inherent heterogeneity