a combinatory strategy for detection of live...

Models and Technologies

A Combinatory Strategy for Detection of LiveCTCsUsingMicrofiltration and aNewTelomerase-Selective AdenovirusYanchun Ma1,2, Sijie Hao3, Shuwen Wang2,4, Yuanjun Zhao2, Bora Lim5, Ming Lei1,David J. Spector6,Wafik S. El-Deiry5, Si-yang Zheng3, and Jiyue Zhu2,4

Abstract

Circulating tumor cells (CTC) have become an importantbiomarker for early cancer diagnosis, prognosis, and treatmentmonitoring. Recently, a replication-competent recombinantadenovirus driven by a human telomerase gene (hTERT) pro-moter was shown to detect live CTCs in blood samples ofpatients with cancer. Here, we report a new class of adeno-viruses containing regulatory elements that repress the hTERTgene in normal cells. Compared with the virus with only thehTERT core promoter, the new viruses showed better selectivityfor replication in cancer cells than in normal cells. In particular,Ad5GTSe, containing three extra copies of a repressor element,

displayed a superior tropism for cancer cells among leukocytesand was thus selected for CTC detection in blood samples. Tofurther improve the efficiency and specificity of CTC identifi-cation, we tested a combinatory strategy of microfiltrationenrichment using flexible micro spring arrays and Ad5GTSeimaging. Our experiments showed that this method efficientlydetected both cancer cells spiked into healthy blood andpotential CTCs in blood samples of patients with breast andpancreatic cancer, demonstrating its potential as a highly sen-sitive and reliable system for detection and capture of CTCs ofdifferent tumor types. Mol Cancer Ther; 14(3); 1–9. �2015 AACR.

IntroductionCancers shed their tumor cells into the peripheral blood,

generating metastases that confer lethality. Circulating tumorcells (CTC) have been detected in patients with many types ofcancers, even thosewith no signs of clinically overtmetastases (1).Because of their easy accessibility, CTCs are clinically relevant tocancer detection and prognosis. Repeated sampling of CTCs maybe used for real-time assessment of cancer progression and treat-ment response (2).

CTC detection and capture are technically demanding. One ofthe challenges is their extremely low abundance in the peripheralblood: the clinically relevant numbers of CTCs are as low as one

CTC per 105 to 107 nucleated blood cells. CTCs are also highlyvariable and their genetic and epigenetic characteristics differ notonly among tumors of different tissue origins, but also within thesame cancer type, or even within a patient. Most current detectionstrategies are based on the epithelial markers like epithelial celladhesion molecule (EpCAM) and cytokeratins (CK). However,evidence is accumulating that, in certain tumor types, theseepithelial markers may be downregulated during tumor celldissemination, potentially hampering the detection of CTCs(3). To date, theCellSearch assay for patientswith breast, prostate,and colon cancer, based on EpCAM expression, is the only CTCdetectionmethod cleared by the FDA. We and others have shownthat, in at least some cases, CellSearch detection of CTC is lessefficient than detection after enrichment by microfiltration (4, 5).Thus, simple and fail-safe techniques for efficient CTC enumer-ation and capture are still needed for clinical applications.

Most adult cancers are of epithelial origin and CTCs in thesepatients are larger than blood leukocytes. CTC enrichment bymicrofiltration, basedon intrinsic physical parameters such as sizeand deformability, is cost-effective (6–8). To further improve theefficiency of microfiltration and the viability of enriched cells, wepreviously described aflexiblemicro spring array (FMSA)device, anew design of microfilters that can enrich CTCs from 7.5 mL ofblood samples with 104-fold enrichment, 90% capture efficiency,and over 80% viability within 10 minutes (9). Thus, filtrationby FMSA is less disruptive to cell integrity than other microfiltra-tion devices, providing a platform capable of enrichment andanalysis of viable CTCs from clinically relevant volumes ofblood samples (4).

Telomerase expression is a hallmark of cancer (10) and isrequired for limitless proliferation of tumor cells. The catalyticsubunit of human telomerase, hTERT, is silenced in normalhuman somatic cells but activated in the majority of cancers.

1College of Life Science, Northwest A&F University, Taicheng Road,Yangling, Shaanxi, China. 2Department of C&MPhysiology, Penn StateCollege of Medicine, Hershey, Pennsylvania. 3Micro & Nano IntegratedBiosystem Laboratory, Department of Biomedical Engineering andMaterials Research Institute, Pennsylvania State University, UniversityPark, Pennsylvania. 4Department of Pharmaceutical Sciences,Collegeof Pharmacy, Washington State University, Spokane, Washington.5Division of Hematology-Oncology, Penn State Hershey Cancer Insti-tute, Hershey, Pennsylvania. 6Department of Microbiology and Immu-nology, Penn State College of Medicine, Hershey, Pennsylvania.

Note: Supplementary data for this article are available at Molecular CancerTherapeutics Online (http://mct.aacrjournals.org/).

Y. Ma and S. Hao contributed equally to this article.

Current address for W.S. El-Deiry: Fox Chase Cancer Center, 333 CottmanAvenue, Room P2035, Philadelphia, PA 19111.

Corresponding Author: Jiyue Zhu, Washington State University College ofPharmacy, PBS 323, PO Box 1495, Spokane, WA 99210. Phone: 509-368-6565; Fax: 509-368-6561; E-mail: [email protected]

doi: 10.1158/1535-7163.MCT-14-0693

�2015 American Association for Cancer Research.

MolecularCancerTherapeutics

www.aacrjournals.org OF1

on April 17, 2018. © 2015 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst January 14, 2015; DOI: 10.1158/1535-7163.MCT-14-0693

http://mct.aacrjournals.org/

This common trait of cancer cells was used in the developmentof the GFP-containing recombinant adenovirus OBP-401, inwhich an hTERT promoter drives the viral E1 "master regulator"gene. This virus infects CTCs ex vivowithout prior enrichment ofblood samples, providing a new logic for CTC detection (11).However, evidence about the specificity of this virus for cancercells is limited (12, 13). Although the strategy of CTC detectionand enumeration by adenovirus infection remains promising,questions about its cancer cell specificity persist and the recom-binant virus is not readily available for most academic scien-tists. Here, we report the development of a class of recombinantadenoviruses with DNA elements that repress hTERT promoterin normal cells (14). Our results show that multiple copies ofthe repressive element rendered the virus more selective forreplication in cancer cells than in normal cells, especiallyleukocytes, as determined by GFP imaging. Furthermore, wedeveloped a combinatory strategy for CTC detection and enu-meration, by taking advantage of both efficient FMSA enrich-ment technique and the specificity and sensitivity conferred byimaging cells infected with the improved adenovirus. Thiscombinatory approach generated a simple and fail-safe, yethighly efficient, system for CTC analysis in the clinical settingwithout labor-intensive cytologic staining procedures.

Materials and MethodsRecombinant adenoviruses

Adenoviral constructs were derived from pAdZ5, a bacterialartificial chromosome (BAC) containing a wild-type adenovirustype V genome (15). The E3 gene in pAdZ5 was first replacedby a GFP cassette from pEGFP-N1 (Clontech), resulting inpAd5G. The hTERT promoter-containing adenoviral constructswere created by replacing the 179-bp E1 promoter with threeversions of the hTERT promoter, using the hTERT ATG codon asthe E1a initiation codon (Fig. 1). BAC modifications wereperformed using a two-step recombineering strategy (16).Viruses were generated by transfecting linearized BACs into293 packaging cells and purified by CsCl2 gradient ultracentri-fugation. The structures of viral DNAs, prepared by Hirt'smethod, were verified by restriction enzyme digestion anddirect sequencing of the junctions. Viral titers were determinedby plaque-forming assays using Hela cells.

Cell culture and telomerase expressionNormal human foreskin fibroblasts (NHF)were obtained from

Dr. Thea Tlsty's laboratory at University of California, San Fran-cisco (San Francisco, CA) in 1995 and the fifth passage of thisstock was used. Normal human breast epithelial cells (HBEC)were obtained in 2011 from Dr. Andrea Manni's laboratory atHershey Medical Center (Hershey, PA) and cultured in MEGMmammary epithelial cell growth medium (Lonza). All cancer celllines were obtained in 2011 fromDr. El-Deiry's laboratory wherethey were regularly authenticated by growth, morphologic obser-vation, and protein expression that was monitored by Westernblotting. The cells were used within two passages in the recom-mended culture conditions without further authentication afterthey were transferred. Human colon cancer cell lines HCT-116and HT-29 were cultured in McCoy's 5A medium with 10% heat-inactivated FBS. Breast cancer cell lines MDA-MB-231 and T-47Dwere cultured in DMEM/F12 medium with 5% FBS and RPMI-1640 medium with 10% FBS, respectively. MCF-7 and 293 cells

were cultured in DMEM with 10% FBS. MCF-10A cells werecultured in DMEM/F12 containing 10 mL/mL insulin, 30 ng/mLEGF, 100 ng/mL cholera toxin, and 0.5 mg/mL hydrocortisone.Humanfibroblast lines 167bandNHFwere cultured inMEMwith10% FBS. HX-98 cells were immortalized from HBECs usingretrovirus pBABE-hTERT (17). mCherry-expressing cells wereobtained by infecting cells with pQCXIP-mCherry or pBABE-neo-mCherry (for HX-98 cells) retroviruses, followed by selectionwith puromycin or neomycin and further enriched by flow-cytometry (BD FACSAria cell sorter). Telomerase activities andhTERTmRNA expression in human cell lines were determined byTRAP assay and real-time RT-PCR analyses, as described previ-ously (18, 19).

White blood cell preparation and adenoviral infectionWhole bloodwas centrifuged at 2,000� rpm for 5minutes and

the plasma phase was removed. Erythrocytes were lysed byincubating the blood cells twice with red blood cell lysis buffercontaining ammonium chloride at 37�C for 5 to 10 minutes,followed by centrifugation and suspension in PBS. mCherry-marked cancer cells, mixed with or without white blood cells(WBC), were infected with adenoviruses and plated at 37�C innontissue culture 96-well plates.

Fluorescence signal detection and immunostainingCells in 96-wellflat clear bottomblack plateswere infectedwith

adenoviruses for 72 hours. GFP and mCherry fluorescence were

hTERT promoter Repressive elements

pAd5GTS

pAd5GTL

pAd5GTSe

(450 bp)

(1.4 kb)

ITR ITRΨ E1A

E1B

∆179bp∆E3

CMV-eGFP

∆1.7kb

pA

pAd5G

A

AP-1 MZF-2 ETSE-box X-box

ATG

–500–1.4k TSS+1

Long promoter – pAd5GTL

Short promoter – pAd5GTS

–77

Core promoter

Repressive element

B

Figure 1.Recombinant adenoviruses. A, a schematic illustration of recombinantadenoviruses. The upper diagram shows the genome of adenovirus type 5(�35 kb) and the positions of the E1 and E3 genes. In Ad5G, the E3 gene isreplaced by a CMV-eGFP expression cassette. In Ad5GTS and Ad5GTL, the179-bp E1 promoter is replaced by a 450-bp and a 1.4-kb hTERT promoterfragment, respectively, with the hTERT ATG codon serving as the E1Ainitiation codon. Ad5GTSe contains the short hTERT promoter fragment andthree extra copies of a repressor element. ITR, inverted terminal repeat; y,packaging signal; pA, polyA signal. B, the hTERT promoter. The hTERT corepromoter, depicted as an oval, contains five GC-boxes and an initiatorelement. The long and short promoter fragments are within the brackets andthe position of a repressor element is indicated by a hatched bar below. ATG,initiation codon; TSS, transcriptional start site.

Ma et al.

Mol Cancer Ther; 14(3) March 2015 Molecular Cancer TherapeuticsOF2




quantitated every 12hours using amicroplate reader (SpectraMaxM5, Molecular Devices) with excitation/emission at 485 nm/528nm and 585 nm/610 nm, respectively. Fluorescent images werecaptured with an EVOS FL microscope. To identify leukocytes,mixtures of WBCs and cancer cells were fixed in BD fixation/permeabilization solution (BDCytofx/CytopermKit), transferredonto glass slides by cytospin, and stained sequentially with aCD45 antibody (Cell Signaling Technology) and Alexa Fluor 594-conjugated antibody (Invitrogen).

FMSA filtration7.5 mL of blood samples were passed through a FMSA

filtration device under gravity flow, followed by washing with7.5 mL of PBS. During filtration, the FMSA device was retainedbetween two sealing O-rings (polydimethylsiloxane, DOWCorning) and clamped within a housing consisting of twoplastic jigs (4). A syringe barrel served as a sample-loading

chamber at the top of the filtration device and the bottom wasconnected to a waste trap. Immediately after filtration, themicrofilters with the attached O-rings were transferred to awell of a 6-well plate containing 1.5 mL of medium. Ad5GTSevirus was added to the top chamber formed between the O-ringand FMSA, followed by incubation at 37�C for 24 hours. Forimmunostaining, cells on the FMSA devices were fixed with 4%paraformaldehyde and permeabilized by addition of 0.3%Triton X-100. After blocking in 5% goat serum, cells wereincubated sequentially with anti-cytokeratin 8/18/19 (clone2A4, Abcam) and goat anti-mouse IgG conjugated to DyLight550 (Thermo Scientific). Cells on the devices were blockedagain with 5% goat serum, incubated with an anti-CD45antibody conjugated to Alexa Fluor 647 (clone 35-Z6, SantaCruz Biotechnology), and nuclei were stained with DAPI. Eachfluorophore was imaged separately using a monochromaticcamera (QImaging, Retiga EXi Blue) and composite images

Figure 2.hTERT expression and replication ofrecombinant adenoviruses. A, hTERTexpression in cancer cell lines andnormal cells. hTERT mRNA levels weredetermined by qRT-PCR. Values werenormalized to the levels of 18Sribosomal RNA. � , normal cells withlimited life spans. B, replication ofrecombinant adenoviruses. Twelvehours before infection, 1.5 � 104 cellswere plated into individual wells of 96-well plates and infected withrecombinant adenoviruses at MOI 10.GFP and mCherry fluorescent signalswere recorded every 12 hours afterinfection using a microplate reader.Viral replication was reported as theratio of GFP signals to those ofmCherry. C, relationship of hTERTmRNA levels and adenoviralreplication. The data in A were plottedagainst the levels of fluorescent signalsat 24 hours in B. All datawere collectedas averages of triplicates.

CTC Detection by Microfiltration and Adenoviral Imaging

www.aacrjournals.org Mol Cancer Ther; 14(3) March 2015 OF3




were created with pseudocolor added using QCapture Pro 7software (QImaging). Each mCherryþ and/or GFPþ cells wereconfirmed to be DAPIþ (data not shown).

For comparison, 7.5mL of blood samples was also subjected tothe CellSearch test at the Penn State Cancer Center (Hershey, PA).The CellSearch Analyzer II software preselects events that arepotential CTCs based on EpCAM and CK positivity in closeproximity to a DAPI signal. The standard CellSearch procedurewas performed and candidate CTCs were scored by a trainedoperator.

Statistical analysisStatistical analyses were performed using GraphPad Prism 6,

and results are reported as mean � SD.

ResultsConstruction of recombinant adenoviruses

Recombinant adenoviruses were generated from the BACconstruct pAdZ5 (15). First, the E3 gene, dispensable for virusreplication in cell culture, was replaced by a GFP cassette togenerate pAd5G. Three telomerase-specific adenoviral vectorswere engineered from pAd5G (Fig. 1). pAd5GTS contains a450-bp hTERT promoter, which is active in many cell types(20, 21). This DNA fragment, which contains the hTERT ini-tiation codon, replaced a 179-bp sequence 50 of and includingthe E1a initiation codon. This configuration retains the viralpackaging signal. In pAd5GTL, a 1.4-kb hTERT promoter substi-tuted for the E1a promoter. The additional sequence in this

long promoter contained four MZF-2 sites and an AP-1 site,which are important for repression of hTERT expression innormal cells (22, 23). pAd5GTSe is similar to pAd5GTS exceptthat it consists of three extra copies of a 20-bp negative controlelement (CGCACGTGGGAAGCCCTGG), containing overlap-ping E-box, Ets, and X-box sites (14, 24, 25). The replication ofthese recombinant viruses was examined in MCF7 cells (Sup-plementary Fig. S1). At a multiplicity of infection (MOI) of 10,GFP-positive cells were detected as early as 12 hours afterinfection and the numbers of infected cells continued toincrease for at least 2 days after infection.

Adenoviral replication and telomerase expressionTo determine the relationship between viral replication

and hTERT expression, we used a collection of normal andcancer-derived cells of lung, breast, and colon origins. As shownin Fig. 2A and Supplementary Fig. S2, all the cancer cell lines testedexpressed high levels of hTERT mRNA and telomerase activity.Fibroblast cell line 3C167b also contained abundant hTERTmRNA and telomerase activity. Furthermore, immortalized butuntransformed breast epithelial MCF10A cells expressed detect-able hTERTmRNA and telomerase activity. In contrast, NHFs andHBECs had little hTERT mRNA and undetectable telomerase.

An important objective of this work was to evaluate whethercancer cells mixed with normal cells would be preferentiallylabeled by virus infection of the mixed population (see below).Accordingly, we needed to establish a means of tracking thecancer cells in the mixtures. This strategy would also permit usto control for any differences among the cells lines in their

Figure 3.Adenoviral infection of different celltypes. A, efficiency of fluorescentlabeling of normal and cancer cells byinfection with recombinantadenoviruses. Approximately 200 cellswere mixed with 100 mL of adenovirusstock and added to individual wells of a96-well plate with no tissue culturecoating. GFP-expressing cells werecounted 24 hours after infection. Totalnumbers of cells were determined bycountingmCherryþ cells. B, infection ofWBCsby adenoviruses.WBCs (103, 104,or 105) were infected with 100mL ofadenovirus stock for 24 hours.

Ma et al.





ability to express GFP. To do so, cells were transduced withretroviruses containing a mCherry protein and sorted by FACS,yielding over 98% mCherry-positive (mCherryþ) cells. Thesecells were infected with adenoviruses and GFP/mCherry signalratios were determined (Fig. 2B). All of the recombinant ade-noviruses replicated more efficiently in telomerase-positivecancer cells than in normal cells, although cellular hTERTmRNA levels were not strictly correlated to viral GFP signals(Fig. 2C). The parental Ad5G virus replicated better than hTERTpromoter-driven viruses in all of the cell lines. Ad5GTL andAd5GTSe replicated even slower than Ad5GTS, especially inNHFs and HBECs, probably because the repressive elementsreduced the rate of accumulation of E1a gene products andconsequently delayed the viral lytic cycle.

Labeling cancer cells and white blood cells by adenovirusesTo determine the abilities of the recombinant adenoviruses

to label cancer cells, about 200 cells in a single well of 96-wellplates were infected with an individual adenovirus at differentdoses and GFPþ cells were quantified 24 hours later. As shownin Fig. 3A, over 80% of all immortal and cancer cells displayedGFP signals when infected with 108 pfu/mL of any of theviruses. With 107 pfu/mL of virus, the labeling efficiencies werestill above 80% for most cell lines, except for 3C167b and T-47D lines. For normal and telomerase-nonexpressing NHFsand HBECs, the labeling efficiencies were lower than for cancercell lines in most cases. Among different viruses, Ad5G infectedboth cancer cell lines and normal cells most efficiently, whereasthe infection efficiencies by Ad5GTSe were lower than those byother viruses, requiring at least 107 pfu/mL to label 70% to 80%of the cancer cells. Nevertheless, our data showed that 107 pfu/

mL Ad5GTSe was best suited for further investigations ofdifferentially labeling cancer cells, although none of the viruseswas strictly specific for cancer cells.

CTC preparations contain large amounts of blood cells evenafter enrichment, so we examined the ability of the adenovirusesto infect WBCs. WBCs were infected with adenoviruses and GFP-labeled cells were examined 24hours after infection. Less than 5%of the WBCs were labeled at the highest Ad5G dose tested (108

pfu/mL; Fig. 3B), much less than Ad5G-infected epithelial orfibroblast cells. The hTERT promoter-driven viruses infectedWBCs at efficiencies much lower than Ad5G did. Ad5GTSe wasparticularly ineffective in labeling WBCs at 107 pfu/mL dose. Thedata indicate that Ad5GTSe is suitable for distinguishing cancercells from WBCs.

Preferential infection of cancer cells in mixtures with WBCsTo establish the feasibility of CTC detection using adenovirus-

mediated imaging, mCherryþ HCT-116 colon cancer cells weremixed withWBCs and infected with 107 pfu/mL of an adenovirusfor 24 hours (Supplementary Fig. S3). When 100 or 1,000 HCT-116 cells were mixed with 103 to 105 WBCs, similar numbers ofmCherryþ and GFPþ cells were observed. When 10 cancer cellswere added to the WBCs, there were more GFPþ cells thanmCherryþ cells especially after infection with Ad5G, indicatingthat Ad5G must have infected significant numbers of WBCs.Ad5GTSe-infected cellswere a notable exception: similar numbersofGFPþ andmCherryþ cellswere always observed, suggesting thatvery few WBCs if any were labeled by this virus. To illustrate thespecificities of infection of HCT-116 cells by adenoviruses, GFPþ

only, mCherryþ only, and double positive cells in the mixturescontaining 100 HCT-116 cells were also counted (Fig. 4, left

Figure 4.Adenovirus-mediated labeling ofmixtures of cancer and normal cellswith WBCs. 100 mCherryþ HCT-116(left charts), MCF7 (left charts), orHBECs (right charts) were mixed with103, 104, or 105 WBCs, and infectedwith 107 pfu/mL adenoviruses in 100mL. GFP, mCherry, and doubly-labeledcells were scored after 24 hourincubation.






column). Most of mCherryþ cells were labeled by GFP, indicatingthat HCT-116 cells were efficiently infected by all four viruses.Conversely, upon infection by hTERT promoter-driven viruses,fewer GFPþmCherry� cells were detected than those in Ad5G-infected mixtures. In the case of Ad5GTSe, about five GFPþ

mCherry� cells were found even when 105 WBCs were used,indicating a strong tropism of this virus for HCT-116 cells. Tocheck the infection of WBCs in the mixtures directly, one of theAd5GTSe-infected cell mixtures (1000 HCT-116 cells and 105

WBCs) was stained with an antibody against CD45, a commonleukocyte antigen. In this experiment, unlabeled HCT-116 cellswere used so that the mCherry fluorescence would not interferewith the visualization of CD45 staining in the red channel.Supplementary Fig. S4 shows that no GFPþ/CD45þ cells wereobserved. We conclude that Ad5GTSe is highly selective forlabeling of HCT-116 cells in mixtures with WBCs.

To further evaluate cancer cell imaging using the recombinantadenoviruses, similar experiments were performed to comparethe imaging efficiencies of cancer and normal breast epithelialcells, MCF7 and HBEC, respectively, in mixtures with WBCs.Hundred cells were mixed with varying numbers of WBCs,infected with 107 pfu/mL adenoviruses, and imaged 24 hoursafter infection. GFP�/mCherryþ cells were uninfected epithelialcells andGFPþ/mCherry� cellsweremost likely infectedWBCs. Asshown in the middle column of Fig. 4, there were significantnumbers of GFPþ/mCherry� MCF7 cells in the Ad5G-infectedmixtures in all conditions tested (grey bars), indicating that thisvirus produced significant false-positive signals due to its infec-tion ofWBCs. Ad5GTS andAd5GTL labeled somemCherry� cells,especially when 105 WBCs were present. Notably, Ad5GTSelabeled over 70% MCF7 cells in all conditions, but infected nomore than 5 mCherry� cells even when 105 WBCs were used. Asabove, we did not detect any Ad5GTSe-labeled cells that expressedCD45 in a mixture of MCF7 cells and 105 WBCs (SupplementaryFig. S5).

The data in Figs. 2 and 3 showed that the recombinantadenoviruses also infect normal human cells. Indeed, theseviruses infected and labeled many HBECs in the mixtures (Fig.4, right column). Fewer cells were labeled by infection withAd5GTL or Ad5GTSe, likely due to the negative regulatoryelements in the hTERT promoter in these viruses. The resultssupport the conclusion that, because it displayed the bestselectivity in labeling cancer cells over both HMECs and WBCs,Ad5GTSe is a viable and attractive candidate for improved CTCdetection.

Enrichment and imagingof cancer cells using FMSAmicrofiltersBecause none of the viruses was absolutely specific for cancer

cells, enrichment of cancer cells before adenoviral imagingmay reduce false-positive signals. To explore the possibility ofcombiningmicrofiltration enrichment and adenoviral imaging inCTC detection, approximately 300 mCherry-labeled coloncancer HCT-116 or breast cancer MCF7 cells were added to 7.5mL of blood from healthy donors and the mixtures were passedthrough a FMSA filtration device (Figs. 5A and B). Themicrofilterswere then transferred into single wells of a 6-well plate andinfected with 107 pfu/mL Ad5GTSe for 24 hours. As shownin Table 1, the labeling efficiencies of cancer cells ranged from75% to 93% in four experiments. However, there were three tofive GFPþ/mCherry� cells, either rare mCherry� cancer cells orinfectedWBCs, in each experiment. This result demonstrated that

Ad5GTSe could detect live cancer cells following their enrichmenton FMSA.

Identification of CTCs in patient blood samplesNext, we applied this combinatory strategy in the analyses of

blood samples from patients with breast and pancreatic can-cers. Two 7.5 mL blood samples were collected from eachpatient. One sample was passed through a FMSA device andcells retained on the microfilter were infected with 107 pfu/mLAd5GTSe. To distinguish cells of epithelial origin from those ofhematopoietic origin, the cells on FMSA were then stained with

B mCherry GFP

10 μm10 μm

A

Infection with 107 pfu

adenoviruses

Filtration7.5 mLblood

Microfilter

PhaseCK/DAPICD45CKGFP

P1

P2

C1

C

50 μm 50 μm 50 μm

Figure 5.A combinatory strategy to detect cancer cells in blood samples. A, schematicof the procedure. 7.5mLof bloodwas passed through aFMSAfiltration deviceand the microfilter was transferred to a single well of a 6-well plate. Cellscaptured on the microfilters were infected with Ad5GTSe. B, infection ofHCT116 and MCF7 cells by Ad5GTSe on a FSMA. 7.5 mL of whole blood fromhealthy donors mixed with approximately 300 mCherryþ cancer cells wasenriched on FMSA and infected with 107 pfu Ad6GTSe virus for 24 hours at37�C. Cancer cells were detected in fluorescent images, but WBCs could onlybe visualized by phase microscopy. Images of HCT 116 cells on FSMA areshown. C, expression of surface markers by Ad5TSe-infected CTCs in cancerpatient samples. 7.5 mL of patient blood was passed through a FSMA deviceand the captured cells were infected with 107 pfu of Ad5GTSe. Cells on themicrofilters were stained with antibodies against CK8/18/19 (DyLight 550)and CD45 (Alexa Fluor 647), and visualized in their respective wavelengths.Nuclei were stained with DAPI. P1, P2, and C1 refer to Patients #1, #2, andcontrol #1 (see Table 2).

Ma et al.





antibodies against CK, an epithelial marker, and CD45 (Fig.5C). As shown in Table 2, GFPþ cells were identified in allpatient samples, ranging from 19 in patient #4 to 32 in patient#3. Most of these GFPþ cells were CKþ and CD45�, consistentwith an epithelial origin characteristic of most breast andpancreatic cancer cells. Interestingly, most CKþ/CD45� cellswere also infected by Ad5GTSe, as indicated by GFP expressionin these cells, indicating that most of the CTCs in the clinicalsamples were actually viable.

Although a few GFPþ cells were detected in blood samplesfrom two healthy donors (controls #1 and #2), none of themwere CKþ/CD45�, suggesting that the GFPþ/CKþ/CD45� cellsin the samples from patients with cancer were tumor cells. Theother 7.5 mL of blood sample from each patient was subjectedin parallel to the CellSearch CTC test. Eight and one positivecells were detected in 7.5 mL of blood from patients #1 and #3,respectively, but none was found in other two samples. Thesedata suggest that Ad5GTSe selectively detected CTCs in patientswith breast and pancreatic cancer with potentially higher sen-sitivity than the Veridex CellSearch system. The results dem-onstrate proof of principle that the combinatory strategy ofmicrofiltration and adenoviral imaging using Ad5GTSe is likelyan extremely sensitive and specific method for CTC detectionand enumeration for patients with cancer.

DiscussionMajor challenges in CTC detection are to process large

amounts of blood cells in a short period of time and to usesensitive and specific methods to identify rare and often-var-iable tumor cells. Most reported CTC methods are based onantibodies against cell surface antigens. Recently, the recom-binant adenovirus OBP-401 was used to image viable CTCs inpatient blood samples (11). Although it was reported that thisvirus was specific for telomerase-expressing cancer cells, datasupporting this conclusion have thus far been limited (12, 13).On the basis of extensive studies of hTERT gene regulation, the

promoter fragment used in OBP-401 should provide onlylimited cancer cell specificity (20, 21). Here, we generated anew class of replication-competent adenoviruses, Ad5GTS,Ad5GTL, and Ad5GTSe, with variants of the hTERT promoter.Our data demonstrated that Ad5GTL and Ad5GTSe, with extrarepressive elements, were more selective for replication incancer cells versus normal cells than Ad5GTS, which containsan hTERT promoter fragment comparable with OBP-401.Among the new viruses, Ad5GTSe was best suited for CTCdetection because this virus labeled the fewest number ofWBCs.

Expressionof the E1 genes of Ad5GTS andOBP-401 is drivenby450 or 455-bp of the hTERT promoter upstream of the ATGcodon, respectively. On the basis of the similarity of their geneticorganization and sequences, these two viruses are expected tohave comparable replication efficiencies and cancer cell specifi-cities. The region controlling E1 expression in the Ad5GTLgenome has an additional 1 kb of sequence upstream of the corehTERT promoter, including several DNA elements involved inhTERT repression in normal cells (Fig. 1B; refs. 22, 23). Instead ofthis region, the Ad5GTSe genome has three extra copies of acomposite element known to repress the hTERT promoter innormal but not cancer cells (14). The E-box site in this elementmediates hTERT repression in differentiated cells by the Mycantagonist Mad1 (25). A putative X-box–binding proteinNFX1-91 represses hTERT expression in epithelial cells. As partof its oncogenic activity, the HPV E6 protein activates hTERTexpression by targeting NFX1-91 for ubiquitin-mediated degra-dation (24). As predicted, replication of the two viruses contain-ing additional repressive elements, especially Ad5GTSe, was lessefficient innormal cells than inmost cancer cell lines. Twonotableexceptionswere 3C167b fibroblasts and T-47Dbreast cancer cells.The hTERT gene is translocated and amplified in some immortalcell lines and cancer cells (26) and is rearranged in 3C167bcells (27). Such alterations of the hTERT locus might contributeto the atypical response of these cell lines to viral infectionand also help explain why we observed poor correlation between

Table 1. Labeling efficiencies of HCT116 and MCF7 cells retained on microfilters

Cell linesTotal mCherryþ

cellsGFPþmCherryþ

cellsInfection efficiency(GFPþ/mCherryþ)

GFPþmCherry�

cells

HCT-116 239 221 93% 5277 252 91% 3

MCF-7 236 178 75% 3267 220 82% 5

NOTE: Experiments were performed as described in Fig. 5B.

Table 2. Detection of potential CTCs in blood samples of patients with cancer

Patient 1 Patient 2 Patient 3 Patient 4 Control 1 Control 2Patients GFPþ GFP� GFPþ GFP� GFPþ GFP� GFPþ GFP� GFPþ GFP� GFPþ GFP�

CKþ/CD45� 18 4 17 2 30 3 13 4 0 0 0 0CKþ/CD45þ 0 ND 0 ND 2 ND 4 ND 0 ND 0 NDCK�/CD45þ 0 tmtc 1 tmtc 0 tmtc 2 tmtc 1 tmtc 6 tmtcCK�/CD45� 3 ND 5 ND 0 ND 0 ND 5 ND 1 NDTotal GFPþ cells 21 23 32 19 6 7Veridex counting 8 0 1 0 ND NDCancer types Breast

(triple-negative)

Breast(ERþ)

Pancreatic(stage IV)

Pancreatic(stage IIB)

No No

Metastasis Bone Bone Liver No No No

NOTE: Numbers indicate the quantities of cells detected on each microfilter.Abbreviations: ND, not determined; tmtc, too many to count.






viral replication, measured by GFP expression, and endogenoushTERT mRNA levels (Fig. 2). In addition, many cancer cellscontain mutations at the hTERT locus, such as promoter muta-tions in familial and sporadic melanoma patients (28, 29). Thesemutations of the endogenous hTERT gene could also alter hTERTtranscription and contribute to the miscorrelation between cel-lular hTERT mRNA levels and viral GFP signals in our study.

Another new observation in this study is that the hTERTpromoter-driven adenoviruses replicate much less efficiently inWBCs than the wild-type Ad5G (Fig. 3B). The data suggest thatthe hTERT promoter has a low activity in normal blood cells.This low activity results in the poor replication of the hTERTpromoter-containing viruses in WBCs and is likely the molecularbasis for selective CTC detection in patient blood samples bythese viruses (11).

We found that Ad5GTSe was best suited for discriminatingbetween CTCs and WBCs in blood samples; 107 infectious par-ticles of this virus labeled about five WBCs per 105 cells in allexperiments. Thus, the lower limit of CTC detection by Ad5GTSeis about one in 104 WBCs. Although this was the best among allthe recombinant adenoviruses, such an efficiency is unlikely toachieve accurate CTC detection in patients with cancer due to theoverwhelmingly large numbers ofWBCs in blood samples. There-fore, enrichment of cancer cells was required to reduce false-positive signals. To this end, we designed a novel strategy com-bining microfiltration enrichment using FMSA devices and ade-noviral infection (4). In this combinatory approach, FMSA filtra-tion was used to enrich the CTCs against the WBCs by a factor ofapproximately 104. In a sample of 7.5 mL undiluted blood, thetotal number of WBCs on the filter was thus reduced to less than104–5, falling within the range in which an optimal detectionsensitivity and specificity of CTCs could be achieved by usingAd5GTSe.

Unlike microfilter devices used previously, the FMSA devicehas flexible micro springs that deform in response to pressure,allowing a higher flow-through rate under low differentialpressure, minimizing cellular mechanical stress, and thus pre-serving viability (9). With this combined strategy, we identifiedmore potential CTCs than the CellSearch assay in our experi-ments using a limited number of cancer patient samples (Table1). Previously, Lin and colleagues also reported that immu-nostaining of patient samples enriched by a different type ofmicrofilter identified potential CTCs in more patients withcancer as well as more CTCs in most patients (5). Thus, thesuperior recovery rate of CTCs by microfiltration in combina-tion with adenoviral infection of live cells may lead to afaster and more efficient strategy of viable CTC detection andenumeration.

In summary, our current combinatory strategy using imaging ofadenovirus-infected cells has the potential to be an excellentmethod for detecting CTCs in clinical settings. First, adenoviralinfection and imaging are simple procedures, requiring no sophis-ticated equipment or highly technicalmanipulations. Second, the

method labels only live cells, because infection and replication inliving cancer cells are required to obtain GFP expression. Third,the hTERT promoter-dependent viruses should infect many typesof cancer cells, regardless of their origins and development stages.This broad host range is an important feature, because cells inmost tumors are highly heterogeneous. Susceptibility to adeno-virus infection is unlikely to be affected by changes in the molec-ular features of tumor cells, such as EMT, a transition fromepithelial tumor cells to an invasive and mesenchymal cell-likephenotype, and so virus infection should provide a reliable CTCevaluation. Finally, though Kojima and colleagues reported thedetection of CTCs by OBP-401 without enrichment (11), ourstudy revealed that even the improved virus Ad5GTSe labelednormal leukocytes in patient samples. However, although thecombinatory strategy promises to result in a more sensitivedetection with fewer false positives than straight adenoviralinfection, additional studies will be needed to test its efficacy invarious clinical settings.

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

Authors' ContributionsConception and design: Y. Ma, S. Hao, S. Wang, W.S. El-Deiry, J. ZhuDevelopment of methodology: Y. Ma, S. Hao, S. Wang, Y. Zhao, M. Lei,D.J. Spector, W.S. El-Deiry, S.-Y. Zheng, J. ZhuAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): S. Hao, S. Wang, B. Lim, D.J. Spector, S.-Y. ZhengAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): Y. Ma, S. Hao, S. Wang, B. Lim, W.S. El-Deiry,S.-Y. Zheng, J. ZhuWriting, review, and/or revision of the manuscript: S. Hao, S. Wang, M. Lei,D.J. Spector, W.S. El-Deiry, S.-Y. Zheng, J. ZhuAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): S.-Y. Zheng, J. ZhuStudy supervision: S. Wang, W.S. El-Deiry, S.-Y. Zheng, J. ZhuOther (helped with proper analysis of data and summarized clinical aspectof patients from whom we obtained samples): B. Lim

AcknowledgmentsThe authors thank Dr. Richard Stanton of Cardiff University for adenoviral

BAC constructs. D.J. Spector thanks Richard Courtney and members of theDepartment of Microbiology and Immunology for their support.

Grant SupportThis work was supported by the NIH grants R03CA165182 and

R01GM071725 (to J. Zhu) and DP2CA174508 (to S.-Y. Zheng). J. Zhu wasalso supported by the W.W. Smith Charitable Trust.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received August 19, 2014; revised December 15, 2014; accepted December29, 2014; published OnlineFirst January 14, 2015.

References1. Kang Y, Pantel K. Tumor cell dissemination: emerging biological insights

from animal models and cancer patients. Cancer Cell 2013;23:573–81.2. Lin H, Balic M, Zheng S, Datar R, Cote RJ. Disseminated and circulating

tumor cells: Role in effective cancer management. Crit Rev Oncol Hematol2010;77:1–11.

3. Balic M, Williams A, Lin H, Datar R, Cote RJ. Circulating tumor cells: frombench to bedside. Annu Rev Med 2013;64:31–44.

4. HarouakaRA, ZhouMD,YehYT, KhanWJ,Das A, Liu X, et al. Flexiblemicrospring array device for high-throughput enrichment of viable circulatingtumor cells. Clin Chem 2014;60:323–33.


Ma et al.




5. Lin HK, Zheng S, Williams AJ, Balic M, Groshen S, Scher HI, et al. Portablefilter-based microdevice for detection and characterization of circulatingtumor cells. Clin Cancer Res 2010;16:5011–8.

6. Vona G, Sabile A, Louha M, Sitruk V, Romana S, Schutze K, et al. Isolationby size of epithelial tumor cells: a new method for the immunomorpho-logical and molecular characterization of circulating tumor cells. Am JPathol 2000;156:57–63.

7. Zheng S, Lin H, Liu JQ, Balic M, Datar R, Cote RJ, et al. Membranemicrofilter device for selective capture, electrolysis and genomic analysisof human circulating tumor cells. J Chromatogr A 2007;1162:154–61.

8. Zheng S, Lin HK, Lu B, Williams A, Datar R, Cote RJ, et al. 3D microfilterdevice for viable circulating tumor cell (CTC) enrichment from blood.Biomed Microdevices 2011;13:203–13.

9. Harouaka RA, Zhou MD, Yeh TY, Khan WJ, Allerton J, Zheng SY. Viablecirculating tumor cell enrichment by flexible micro spring array. Conf ProcIEEE Eng Med Biol Soc 2012;2012:6269–72.

10. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell2011;144:646–74.

11. Kojima T, Hashimoto Y, Watanabe Y, Kagawa S, Uno F, Kuroda S, et al. Asimple biological imaging system for detecting viable human circulatingtumor cells. J Clin Invest 2009;119:3172–81.

12. Kishimoto H, Kojima T, Watanabe Y, Kagawa S, Fujiwara T, Uno F, et al. Invivo imaging of lymph node metastasis with telomerase-specific replica-tion-selective adenovirus. Nat Med 2006;12:1213–9.

13. Maida Y, Kyo S, Sakaguchi J, Mizumoto Y, Hashimoto M, Mori N, et al.Diagnostic potential and limitation of imaging cancer cells in cytologicalsamples using telomerase-specific replicative adenovirus. Int J Oncol2009;34:1549–56.

14. Horikawa I, Cable PL, Mazur SJ, Appella E, Afshari CA, Barrett JC.Downstream E-Box-mediated regulation of the human telomerasereverse transcriptase (hTERT) gene transcription: evidence for an endog-enous mechanism of transcriptional repression. Mol Biol Cell2002;13:2585–97.

15. Stanton RJ, McSharry BP, Armstrong M, Tomasec P, Wilkinson GW. Re-engineering adenovirus vector systems to enable high-throughput analysesof gene function. Biotechniques 2008;45:659–62, 664–8.

16. Wang S, Zhao Y, Leiby M, Zhu J. A new positive/negative selection schemefor precise BAC recombineering. Mol Biotechnol 2009;42:110–6.

17. Zhu J, Wang H, Bishop JM, Blackburn EH. Telomerase extends the lifespanof virus-transformed human cells without net telomere lengthening. ProcNatl Acad Sci USA 1999;96:3723–8.

18. Jia W, Wang S, Horner JW, Wang N, Wang H, Gunther EJ, et al. A BACtransgenic reporter recapitulates in vivo regulation of human telomerasereverse transcriptase in development and tumorigenesis. Faseb J 2011;25:979–89.

19. Mathew R, Jia W, Sharma A, Zhao Y, Clarke LE, Cheng X, et al. Robustactivation of the human but not mouse telomerase gene during theinduction of pluripotency. Faseb J 2010;24:2702–2715.

20. Ducrest AL, Amacker M, Mathieu YD, Cuthbert AP, Trott DA, NewboldRF, et al. Regulation of human telomerase activity: repression bynormal chromosome 3 abolishes nuclear telomerase reverse transcrip-tase transcripts but does not affect c-Myc activity. Cancer Res 2001;61:7594–602.

21. Wang S, Zhu J. Evidence for a relief of repressionmechanism for activationof the human telomerase reverse transcriptase promoter. J Biol Chem2003;278:18842–50.

22. Fujimoto K, Kyo S, Takakura M, Kanaya T, Kitagawa Y, Itoh H, et al.Identification and characterization of negative regulatory elements of thehuman telomerase catalytic subunit (hTERT) gene promoter: possible roleof MZF-2 in transcriptional repression of hTERT. Nuc Acid Res2000;28:2557–62.

23. Takakura M, Kyo S, Inoue M, Wright WE, Shay JW. Function of AP-1 intranscription of the telomerase reverse transcriptase gene (TERT) in humanand mouse cells. Mol Cell Biol 2005;25:8037–43.

24. Gewin L, Myers H, Kiyono T, Galloway DA. Identification of a noveltelomerase repressor that interacts with the human papillomavirus type-16 E6/E6-AP complex. Genes Dev 2004;18:2269–82.

25. XuD, PopovN,HouM,WangQ, BjorkholmM,Gruber A, et al. Switch fromMyc/Max to Mad1/Max binding and decrease in histone acetylation at thetelomerase reverse transcriptase promoter during differentiation of HL60cells. Proc Natl Acad Sci USA 2001;98:3826–31.

26. Zhu J, Zhao Y, Wang S. Chromatin and epigenetic regulation of thetelomerase reverse transcriptase gene. Protein & Cell 2010;1:22–32.

27. Zhao Y, Wang S, Popova EY, Grigoryev SA, Zhu J. Rearrangement ofupstream sequences of the hTERT gene during cellular immortalization.Genes Chromosomes Cancer 2009;48:963–74.

28. Horn S, Figl A, Rachakonda PS, Fischer C, Sucker A, Gast A, et al. TERTpromoter mutations in familial and sporadic melanoma. Science 2013;339:959–61.

29. Huang FW, Hodis E, Xu MJ, Kryukov GV, Chin L, Garraway LA. Highlyrecurrent TERT promoter mutations in human melanoma. Science2013;339:957–9.






Published OnlineFirst January 14, 2015.Mol Cancer Ther Yanchun Ma, Sijie Hao, Shuwen Wang, et al. Microfiltration and a New Telomerase-Selective AdenovirusA Combinatory Strategy for Detection of Live CTCs Using

Updated version

10.1158/1535-7163.MCT-14-0693doi:

Access the most recent version of this article at:

Material

Supplementary

http://mct.aacrjournals.org/content/suppl/2015/01/14/1535-7163.MCT-14-0693.DC1

Access the most recent supplemental material at:

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

.http://mct.aacrjournals.org/content/early/2015/02/19/1535-7163.MCT-14-0693To request permission to re-use all or part of this article, use this link



http://mct.aacrjournals.org/lookup/doi/10.1158/1535-7163.MCT-14-0693

http://mct.aacrjournals.org/content/suppl/2015/01/14/1535-7163.MCT-14-0693.DC1

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

mailto:[email protected]

http://mct.aacrjournals.org/content/early/2015/02/19/1535-7163.MCT-14-0693


a combinatory strategy for detection of live...

Documents