Forward genetic screens in mice uncover mediatorsand suppressors of metastatic reactivationHua Gaoa,b,1, Goutam Chakrabortyb, Ai Ping Lee-Limb, Konstantinos J. Mavrakisc, Hans-Guido Wendelc,and Filippo G. Giancottib,1

aResearch Center for Translational Medicine, East Hospital, School of Life Sciences and Technology, Tongji University, Shanghai 200092, China; and bCellBiology Program and Metastasis Research Center and cCancer Biology and Genetics Program, Sloan Kettering Institute for Cancer Research, Memorial SloanKettering Cancer Center, New York, NY 10065

Edited by Erkki Ruoslahti, Sanford–Burnham Medical Research Institute, La Jolla, CA, and approved October 2, 2014 (received for review February 21, 2014)

We have developed a screening platform for the isolation of geneticentities involved in metastatic reactivation. Retroviral libraries ofcDNAs from fully metastatic breast-cancer cells or pooled microRNAswere transduced into breast-cancer cells that become dormant uponinfiltrating the lung. Upon inoculation in the tail vein of mice, thecells that had acquired the ability to undergo reactivation generatedmetastatic lesions. Integrated retroviral vectors were recoveredfrom these lesions, sequenced, and subjected to a second roundof validation. By using this strategy, we isolated canonical genesand microRNAs that mediate metastatic reactivation in the lung.To identify genes that oppose reactivation, we screened an expres-sion library encoding shRNAs, and we identified target genes thatencode potential enforcers of dormancy. Our screening strategyenables the identification and rapid biological validation of singlegenetic entities that are necessary to maintain dormancy or toinduce reactivation. This technology should facilitate the elucida-tion of the molecular underpinnings of these processes.

forward genetic screens | metastatic reactivation | cDNA library screen |shRNA library screen | microRNA library screen

The majority of cancer-related deaths are caused by metastaticrelapse (1). The process through which cancer cells acquire

metastatic capacity is complex. Unrestrained proliferation, resis-tance to proapoptotic insults, and invasion through tissue bound-aries are not sufficient for metastasis. To colonize distant organs,cancer cells must also adapt to the local microenvironment ofthe target organ and finally outgrow (2, 3). Mathematical mod-eling of clinical data and experiments in mouse models suggestthat cancer cells disseminating from prevalent cancers, such asthose of the breast and prostate, undergo an extended period ofdormancy at premetastatic sites (4). Insights into the mechanismsthat enable disseminated cancer cells to survive during dormancyand then outgrow into life-threatening lesions may lead to theidentification of novel therapeutic targets for the prevention ortreatment of metastatic disease.Advances in genomics and mouse modeling have fostered a

renaissance of studies on metastasis (2, 3). Current approachesto the study of metastasis can be divided in two categories. Inthe first, genetic methods are used to modify the function of acandidate gene in intact mice or in cells that are subsequentlytransplanted in mice (5, 6). In the second, genomic methods, suchas DNA microarray analysis or array comparative genomic hy-bridization (aCGH), are used to identify a restricted number ofcandidate genes, which are then tested in appropriate mousemodels (7, 8). Although these approaches have been extremelysuccessful, they are very laborious and do not necessarily yieldbiologically potent mediators of metastasis. Functional geneticscreens can lead to the rapid identification of strong mediatorsof a selectable phenotype (9, 10). In agreement with this notion,recent studies have revealed that RNAi screens in vivo canidentify mediators of skin carcinogenesis (11). However, formi-dable practical barriers have so far prevented successful appli-cation of gain-of-function screens to a process as complex asmetastasis. First, metastasis is a highly selective process. Only a

small percentage of cancer cells acquire the genetic alterationsnecessary to complete each of the sequential steps of metastasis,reducing the throughput of screens in which genetically modifiedtumor cells are injected at the primary site. Second, the com-pletion of each of the steps of metastasis requires the acquisitionof multiple capabilities, suggesting the possible involvement ofseveral genes. We have recently identified a mouse xenograftmodel that effectively mimics metastatic dormancy and reactiva-tion of breast cancer (12). Here, we describe a flexible and high-throughput functional genomic platform that enables the identifi-cation of single genes that enforce dormancy or mediate metastaticreactivation of breast cancer.

ResultsFeasibility Studies.To examine the feasibility of a functional genomicsapproach to the study of metastasis, we designed a cDNA screenthat uses the mouse as a filter to isolate prometastatic genes (Fig.S1A). In this screen, cDNA libraries derived from highly metastaticbreast-cancer cells are introduced into nonmetastatic cells, and therecipient cancer cells are injected in the mammary fat pad of mice.After a lag time, cancer cells that have acquired the capacity tometastasize to the lung will eventually form lesions in this organ.Integrated proviruses are rescued from the cells that constitute theselesions and sequenced. Finally, the provirus is reintroduced in lowmetastatic cells to confirm its prometastatic activity.To model metastasis as faithfully as possible, we elected to use

a progression series that comprises four mouse mammary car-cinoma cells seemingly arrested at defined steps of the metastaticcascade (Fig. 1A). Whereas the 67NR cells give rise to noninvasiveprimary tumors, the 168FARN cells are able to colonize loco-regional lymph nodes but do not gain access to the vasculature,and the 4TO7 cells enter into the vasculature and seemingly hometo the lung but do not produce visible metastatic lesions. In

Significance

Insights into the mechanisms that enable disseminated cancercells to survive during dormancy and then outgrow into life-threatening lesions may lead to the identification of noveltherapeutic targets for the prevention or treatment of meta-static disease. We have developed flexible and high-through-put functional genetic screens, which enable the identificationof single genetic entities that mediate metastatic reactivationof breast cancer in mice. These screens promise to facilitate theidentification of the core signaling pathways that govern meta-static dormancy and reactivation.

Author contributions: H.G. and F.G.G. designed research; H.G., G.C., and A.P.L.-L. per-formed research; K.J.M. and H.-G.W. contributed new reagents/analytic tools; H.G.,G.C., A.P.L.-L., and F.G.G. analyzed data; and H.G. and F.G.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence may be addressed. Email: f-giancotti@ski.mskcc.org or gaoh@tongji.edu.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1403234111/-/DCSupplemental.

16532–16537 | PNAS | November 18, 2014 | vol. 111 | no. 46 www.pnas.org/cgi/doi/10.1073/pnas.1403234111

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 3,

202

2

contrast, the 4T1 cells complete all these steps and producemacroscopic lung metastases (13).We constructed three high-complexity cDNA libraries from

size-fractionated mRNA isolated from 4T1 cells and clonedthem in the modified retroviral shuttle vector pEYK3.1, whichuses the MoLV-LTR promoter to drive expression of N-terminallyFlag-tagged proteins (total ∼2 × 106 independent clones) (Fig.S1B). FACS analysis indicated that a multiplicity of infection(MOI) of 3:1 leads to efficient transduction of more than 85% of67NR, 168FARN, and 4TO7 cells (Fig. S1 C and D). The cDNAlibraries from the 4T1 cells were thus transduced at an MOI of3:1 into each of the nonmetastatic cell types (Fig. S2A). Inter-estingly, upon transduction with 4T1 libraries and injection intothe mammary fat pad of syngeneic mice, the 67NR or 168FARNcells did not give rise to macroscopic lung metastases, suggestingthat the introduction of a single gene did not enable these cells topenetrate into the bloodstream and acquire all of the additionalcapabilities required for metastatic colonization. In contrast, the4TO7 cells infected with the 4T1 libraries produced a total ofnine lung nodules in multiple mice within 60 d (Fig. S2A) (12).Thirty mice injected with 4TO7 cells infected with empty vectordid not produce macroscopic lung metastases, indicating thatinsertional mutagenesis does not contribute to metastasis inthis system.Sequencing revealed that three of the cDNAs isolated were in the

reverse orientation and that one of them was out of frame (TableS1). The remaining three cDNAs were able to promote lung me-tastasis upon reintroduction in 4TO7 cells (12). One of thesecDNAs was recovered from two distinct lesions and found to en-code for an N-terminally truncated form of the secreted TGF-βligand inhibitor Coco (encoded by Dand5) (Fig. S2B). Subsequentstudies indicated that Coco promotes lung colonization by inhibitingparacrine bone morphogenetic proteins (BMPs) signaling and val-idated the relevance of Coco for human breast-cancer metastasis(12). The other two cDNAs encoded full-length mitochondrial ri-bosomal protein L3 (Mrpl3) and a 5′ fragment of the long non-coding RNA Malat1 (metastasis-associated lung adenocarcinomatranscript 1) (Fig. S2 C and D). Consistent with their ability tomediate a cancer-specific function, MRPL3 is overexpressed in non-small cell lung cancers (14) and MALAT1 in non-small cell lungcancers and hepatocellular carcinomas (15, 16). Although we

have not examined the mechanism by which these genes promotemetastasis, MRPL3 promotes the translation of mRNAs encodingmitochondrial enzymes (17) and the biogenesis of miRNA-likesmall RNAs (miRNAs) (18), and MALAT1 is involved in alter-native splicing (19) and transcriptional activation of growth-controlgenes (20). These results indicate that a gain-of-function cDNAscreen in a mouse model of mammary-tumor progression enablesthe identification of single genes implicated in lung colonization.Confocal imaging indicated that control 4TO7 cells infiltrate

the lung as efficiently as those expressing Coco after tail-vein in-jection. However, whereas the 4TO7 cells enter into proliferativequiescence immediately upon penetrating into the lung stromaand remain in this state for the entire observation period, thoseexpressing Coco undergo reactivation at around day 7 postin-jection (Fig. 1B) (12). Additional experiments demonstrated thatthe number of 4TO7 cells remain constant for at least 2 mo inthe lung stroma. These cells are viable, as indicated by their in-tact nuclear morphology and absence of reactivity with antibodiesto cleaved Caspase 3 (Fig. S3 A–D). Moreover, they are notsenescent because they do not exhibit senescence-associated het-erochromatic foci or phosphorylated histone γH2AX (Fig. S3 Eand F). These results demonstrated that the 4TO7 cells enter intobona fide solitary tumor dormancy upon infiltrating the lung.Phenotypic and functional analyses demonstrated that the

4TO7 and 4T1 cells exhibit traits associated with cancer stemcells, such as the ability to form oncospheres in vitro and toinitiate tumors in vivo (12). Moreover, immunofluorescent stain-ing indicated that the 4TO7 cells express Vimentin, but notE-cadherin, both in vitro and during solitary tumor dormancyin the lung, indicating that they have passed through an epithelial–mesenchymal transition (EMT) and retain this state during dor-mancy (Fig. S4). The fully metastatic 4T1 cells exhibited similarEMT traits in vitro and during the initial phase of latency in thelung (Figs. S4A and S5). After reactivation, a fraction of out-growing 4T1 cells exhibited reduced, but not absent, levels ofVimentin (Fig. S5 A and B). However, virtually none of the 4T1cells, within either incipient or frankly outgrowing lesions, dis-played detectable levels of E-cadherin (Fig. S5C). These resultssuggest that the 4T1 cells undergo reactivation without revertingto an epithelial phenotype.

Tail vein injection(3 x 105 library-transduced cells)

Growth in selectivemedium

Rescue of provirusand cDNA sequencing

Validation

cDNA libraryin pEYK3.1

Dormant cells (4TO7)

Metastatic cells (4T1)

BA

Loca

l inv

asio

n

Dis

sem

inat

ion

Rea

ctiv

atio

n

Prim

ary

tum

or

67NR

168FARN

4TO7

4T1

4TO7-TGL-vec 4TO7-TGL-Coco

PEC

AM

1 / G

FP /

DA

PI

C

Fig. 1. Screening platform for the isolation of genes that me-diate metastatic reactivation. (A) Schematic representation ofthe metastatic capabilities of the mouse mammary carcinomacells constituting the progression series used here. (B) 4TO7-TGL cells transduced with empty vector or Coco were inoculatedi.v. into syngeneic mice. Mice were killed after 21 d. Lung sec-tions were subjected to immunofluorescent staining of GFP(tumor cells) and PECAM-1 (endothelial cells), followed by con-focal analysis and 3D reconstruction. (C) Design of the tail-veincDNA screen. After transduction with the 4T1 libraries, the 4TO7cells are injected i.v. in syngeneic mice. Candidate mediators ofmetastasis are recovered from lung metastases.

Gao et al. PNAS | November 18, 2014 | vol. 111 | no. 46 | 16533

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 3,

202

2

A Gain-of-Function Platform for the Identification of Canonical GenesInvolved in Metastatic Reactivation. Building upon the identificationof a faithful model of metastatic dormancy and reactivation, wedesigned an improved strategy to screen for genes specifically in-volved in metastatic reactivation (Fig. 1C). We generated 4TO7cells expressing the triple-modality TGL reporter vector, whichencodes for thymidine kinase, GFP, and luciferase transducedthem with the libraries, and finally injected them directly into thetail vein of syngeneic mice. This approach bypasses the significantattrition that occurs during tumor initiation, primary tumor growth,and intravasation, thus increasing the throughput of the screen. Inaddition, bioluminescent imaging of luciferase activity enablesa quantitative assessment of colonization, eliminating the need todissect a large number of mice. Finally, expression of GFP facili-tates the identification of small metastatic lesions under a fluores-cent dissection microscope.Preliminary experiments indicated that the 4TO7-TGL cells do

not produce lung metastases when they are injected at 3 × 105 in thetail vein of syngeneic mice (n = 20). After transduction with thethree high-complexity cDNA libraries, the 4TO7-TGL cells werethen injected at this dose. To improve the screening output, therecipient cells were inoculated in different numbers of micedepending on the level of complexity of the library with which theywere transduced (Fig. 2A). Bioluminescent imaging and dissectionindicated that several mice developed one or more lung metas-tases (Fig. S6). Identical inserts encoding full-length TM4SF1,a divergent tetraspanin, were recovered from 44 lesions, andidentical inserts encoding ArhGEF2, a Rho-GEF, were re-covered from 5 lesions (Table 1). In contrast to control 4TO7cells, those stably transduced with TM4SF1 efficiently colonizedthe lung after tail-vein injection, suggesting that expression of

TM4SF1 is sufficient to promote reactivation in the lung (Fig.2B and Fig. S7A). In contrast, ArhGEF2 did not pass validation.Previous studies have indicated that TM4SF1 is associated withpoor prognosis in squamous lung carcinoma (21) and mesothe-lioma (22). In addition, TM4SF1 combines with synthenin-2, whichpromotes melanoma and breast-cancer metastasis by activatingprotein kinase C and focal adhesion kinase/Src family kinase sig-naling (23–25).

Gain-of-Function Screens Identify microRNAs That Promote MetastaticReactivation. Because of their ability to simultaneously regulatethe level of expression of multiple genes, microRNAs may beparticularly adept at coordinating gene-expression changes thatare required to complete several steps of the metastatic cascade(26). Although prior studies have indicated that expression ofmiR-335 and miR-126 can suppress micrometastatic outgrowthby blocking reinitiation and neo-angiogenesis, respectively (27,28), it is not currently known whether microRNAs can promoteexit from dormancy and reactivation.To identify microRNAs able to promote breast-cancer reac-

tivation in the lung, we screened a library of mouse microRNAscloned in the retroviral vector pMSCV (29). The original library wassubdivided in five pools, each containing ∼70 microRNAs, andsupplemented with an additional pool containing 27 microRNAs,which we had identified as up-regulated more than twofold in 4T1cells compared with 4TO7 cells by using microRNA microarrayanalysis (Fig. 3A). Because pMSCV drives expression of GFP froman internal ribosome entry site (IRES), we could directly determinethe titer of each pool and transduce the 4TO7 cells at an MOI ableto drive expression of an average of one microRNA per cell. Cellstransduced with each of the six pools of microRNAs were injectedin the tail vein of five syngeneic mice to ensure complete screeningof each collection of microRNAs (Fig. 3A). Five of the 30 micedeveloped multiple metastases in their lungs and 2 mice developedsingle lesions. Genomic sequencing of tumor cells isolated fromthese lesions revealed that they harbored nine integrated micro-RNA genes (Table 2). The microRNAs miR-340 and miR-346 andthe combination of miR-138 and miR-223 were recovered frommultiple lesions from different mice. To validate these hits, weconstructed retroviral miR-Vec vectors expressing each of the ninemicroRNAs from the CMV promoter and infected 4TO7-TGL cellsusing single vectors or combinations of vectors. Q-PCR (quantita-tive PCR) was used to verify that the microRNAs wereadequately expressed. Tail-vein injection followed by bio-luminescent imaging indicated that expression of miR-138 ormiR-346 efficiently induces lung colonization in the majority ofmice (3/5), indicating that these microRNAs promote exit fromdormancy in the lung (Fig. 3B and Fig. S7 B and C). In contrast,the other microRNAs did not pass validation. These results in-dicate that our screening platform permits the isolation ofmicroRNAs that mediate metastatic reactivation.

shRNA Screens Identify Genes That Enforce Tumor Dormancy. Wereasoned that screening a library of shRNAs might enable theidentification of genes that antagonize metastatic reactivation. Be-cause they need to be inactivated for successful outgrowth, thesegenes can be viewed as enforcers of the dormant state. To test this

ALibrary A(0-1 Kb)

Library B (1-3 Kb)

Library C(>3 Kb)

2.8 x 105

Independentclones (#)

Miceinoculated (#)

Cellsinjected (#)

8 mice3 x 105

TheoreticalRedundancy

(fold)2.4 x 106

Total cellsinjected (#)

8.6

1.6 x 106 12 mice3 x 105 3.6 x 106 2.25

4.6 x 104 8 mice3 x 105 2.4 x 106 53

B

1

10

102

0 1 2 3 4 5 6 7 Weeks

103

Nor

mal

ized

pho

ton

flux

4TO7-pEYK (n=5)4TO7-pEYK-Tm4sf1 (n=5)

30 min day 1 week 3 4 5 6

P=1.57E-4

4TO

7-pE

YK4T

O7-

pEYK

-Tm

4sf1

Fig. 2. Design of the gain-of-function cDNA library screen. (A) The flowchartdescribes the throughput and theoretical redundancy of the cDNA screen. (B)4TO7-TGL cells stably transduced with empty vector or TM4SF1 were in-oculated i.v. into syngeneic mice. Lung colonization was measured by bio-luminescent imaging. The panels show representative images (Left), and thegraph shows the normalized photon flux at the indicated times (Right). Notethat the scale for normalized photon flux is logarithmic.

Table 1. Results of the cDNA screen

Tm4sf1 encodes for the divergent tetraspanin TM4SF1 [also known as tumor-associated antigen L6 (TAL6)] and Arhgef2 for the Rho/Rac GEF 2 (also known as GEF-H1). Identical inserts encoding full-length TM4SF1 or Rho/Rac GEF 2 were recovered from 44 metastaticlesions. Two metastases, which presumably arose from the confluence of two distinct outgrowths, contained both inserts. Tm4sf1passed validation and is indicated in red. Numbers in parentheses indicate the lesion numbers, which contain inserts encoding TM4SF1or ArhGEF2. Mets, metastases; N/A, not available.

16534 | www.pnas.org/cgi/doi/10.1073/pnas.1403234111 Gao et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 3,

202

2

hypothesis, we screened a modified version of the Cancer 1,000mouse library (30, 31). The base library comprises ∼2,300 shRNAstargeting a list of ∼1,000 cancer-relevant genes (2–3 shRNAs pergene). To ensure potent knock down as single-copy integrants, theshRNAs were cloned in the context of an miR-30 backbone in anMSCV-based, GFP-expressing vector (32). Based on initial “re-construction” studies, the Cancer 1,000 library was subdivided in sixpools, each containing 288 shRNAs, and supplemented with a sev-enth pool comprising 260 additional shRNAs against cancer-relatedgenes (1–7 shRNAs per gene) (33). After transduction with theseven pools of shRNAs from the modified Cancer 1,000 library, the4TO7-TGL cells were injected into the tail vein of 49 recipient mice(Fig. 4A). A large fraction of mice (∼25%) developed metastaticlesions after 6 wk. Metastatic cells were isolated from 45 lunglesions, and the potentially prometastatic shRNAs that they har-bored were sequenced and identified. Primary hits were prioritized

based on their rate of recurrence. The target genes of shRNAsidentified in 2 or more independent lesions encoded a variety ofsignaling proteins and transcription factors (Table 3). To validatethese results, we conducted a secondary screen under similarconditions. Interestingly, shRNAs targeting three genes were re-covered in both rounds of screening, strongly suggesting that thesegenes must be inactivated for exit from dormancy. Intriguingly, thegenes targeted by these shRNAs constitute the Notch inhibitorNumb (34, 35), the histone methyl transferase Smyd5 (36), and theSmad E3 ubiquitin ligase Smurf2 (37). Because it is well establishedthat Smurf2 antagonizes TGF-β signaling, which promotes metas-tasis (2, 37), we focused on Numb and Smyd5. Customary specificitycontrols in RNAi screens involve testing a suite of shRNAs tar-geting the gene or rescuing the effect produced by the originalshRNA through expression of an shRNA-insensitive cDNA (9).Because we were interested in simplifying the overall phase ofvalidation, we decided to examine whether ectopic expression ofNumb or Smyd5 suppressed the ability of metastatic breast-cancer cells to colonize the lung. Vectors encoding these geneswere introduced in mammary-tumor cells from MMTV-Neu micebecause these cells provide a faithful model for HER2+ metastaticbreast cancer (12, 38) (Fig. S7 D and E). Interestingly, bio-luminescent imaging revealed that expression of Numb and Smyd5almost completely suppresses the ability of ErbB2-transformedcells to metastasize to the lung, consistent with the hypothesis thatNumb and Smyd5 mediate tumor dormancy (Fig. 4B). Althoughwe cannot at present exclude that some of the other shRNAsidentified silence genes that enforce dormancy, these resultsindicate that our screening strategy can lead to the rapid iden-tification of single genes that enforce tumor dormancy.

Bioinformatic Analysis of Existing Datasets Can Provide UsefulInformation on the Genes Identified. Oncoprint analysis of the pro-visional breast cancer TCGA dataset indicated that TM4SF1, andthe long noncoding RNAMALAT1 are amplified or overexpressedin 11% and 7% of the cases, respectively (Fig. S8A). Oncomine(www.oncomine.org/resource/login.html) analysis revealed thatTM4SF1 is expressed at higher levels in basal-like compared withluminal-like tumors in the Farmer dataset. In contrast, SMYD5 isup-regulated in apocrine tumors compared with other tumors (Fig.S8B). NUMB was down-regulated in both basal-like and luminal-like tumors in the Farmer dataset and in all tumors in the Curtisdataset (Fig. S8C). Finally, MALAT1 was up-regulated in breastcarcinoma compared with normal breast, and in lymph-node me-tastases compared with breast carcinoma in the Zhao dataset (Fig.S8D). No useful inference could be made regarding the other genesidentified. Comprehensive mechanistic studies and clinical vali-dation experiments will be necessary to establish the bi-ological and clinical relevance of each of the genes identified.

DiscussionWe have developed a screening strategy that uses the mouse as afilter to identify genetic entities that promote metastatic reac-tivation. Validation experiments have provided evidence that the

A

Pools 1-5 ~ 70 miR

Independentclones (#)

Miceinoculated (#)

Cellsinjected (#)

5 mice3 x 105

TheoreticalRedundancy

(fold)1.5 x 106

Total cellsinjected (#)

> 2 x 104

B

4TO7-miR-Vec (n=5)4TO7-miR-138 (n=3/5)

0.1 0 1 5 6 7 8 9

Weeks

Nor

mal

ized

pho

ton

flux

P=0.0110

102

103

104

1

p=6E-3

0.1

10

102

103

1

Nor

mal

ized

pho

ton

flux

4TO7-miR-Vec (n=5)4TO7-miR-346 (n=3/5)

0 1 2 3 4 6 7 5

*

Pool 6 5 mice3 x 105 1.5 x 10627 miR > 5 x 104

30 min week 7 8 30 min week 5 6

4TO

7-m

iR-V

ec4T

O7-

miR

-346

4TO

7-m

iR-V

ec4T

O7-

miR

-138

Weeks

Fig. 3. Gain-of-function microRNA library screen. (A) The flowchart describesthe throughput and theoretical redundancy of the microRNA screen. (B) 4TO7-TGL cells stably transduced with empty vector or the indicated microRNAs(miR-138, Left; miR-346, Right) were inoculated i.v. into syngeneic mice.Lung metastasis was measured by bioluminescent imaging. The panels showrepresentative images (Top), and the graphs show the normalized photon fluxat the indicated times (Bottom). Note that the end time point is 8 wk formiR-138 and relative controls and 6 wk for miR-346 and relative controls.

Table 2. Results of the microRNA screen

The microRNAs miR-25 and miR-27a were recovered from a single lesion and did not pass validation upon reexpression, either alone or in combination. ThemicroRNAs miR-138 and miR-223 were similarly recovered together from three lesions. miR-346 and miR-138 passed validation and are indicated in red.Numbers in parentheses indicate the lesion numbers, which contain inserts encoding corresponding miRNAs. BLU, bioluminescence units; N/A, not available.

Gao et al. PNAS | November 18, 2014 | vol. 111 | no. 46 | 16535

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 3,

202

2

canonical genes and microRNAs identified through this method arepotent mediators of reactivation whereas the targets of shRNAs,and potentially those of microRNAs, prevent reactivation and arethereby implicated in enforcing dormancy. Thus, forward genetic

screens in mice can identify genes that govern metastatic dormancyand reactivation.The screening platform we have developed has several notable

attributes. First, it enables a specific analysis of metastatic dor-mancy and reactivation. The recipient 4TO7 cells infiltrate thelung efficiently but immediately enter into solitary tumor-celldormancy, enabling genetic analysis of this process and thesubsequent reactivation step (12). In contrast, prior studies havenot distinguished between solitary tumor-cell dormancy andmicrometastatic dormancy, or they have focused on the latter(39–41). Moreover, the 4TO7 cells display features of cancerstem cells, such as the ability to form oncospheres in vitro and toinitiate tumors in mice (12). Because of their migratory and in-vasive ability, as well as their ability to enter into quiescence andto undergo reactivation in response to specific stimuli, the 4TO7cells provide a faithful model for migrating cancer stem cells(42). Our screening platform should therefore facilitate theidentification of the core signaling pathways that drive thereactivation of these metastasis-initiating cells.Second, our screens seem to be very stringent. The back-

ground is virtually negligible. The parental 4TO7 cells or thosetransduced with empty vector never give rise to macroscopicmetastases, indicating that stochastic genetic or epigeneticevents or insertional mutagenesis cannot convert these cellsinto fully metastatic cells at a detectable frequency. This con-clusion is consistent with the hypothesis that a limited numberof genes govern metastatic dormancy and reactivation. Fur-thermore, although we have injected the library-transduced4TO7 cells directly into the bloodstream, thus bypassing thesteps of metastasis leading to dissemination, the colonizationphase of metastasis is characterized by significant attrition. Onlya small fraction of the tumor cells, which are inoculated in-travenously, survive in the bloodstream, extravasate into thestroma of the target organ, and avoid apoptosis (43–45). In fact,only tumor cells that adhere to the abluminal endothelial base-ment membrane receive signals that promote their survival, butnot their proliferation, allowing entry into dormancy (46). Basedon prior results (12), we have estimated that only ∼8% ofinjected 4TO7 cells survive initial attrition and successfully enterinto dormancy in the lung. Moreover, only a small fraction ofthe tumor cells that have entered into dormancy eventuallyundergo reactivation, suggesting that formidable barriers oppose

ErbB2-vec (n=5)ErbB2-Numb (n=5)

0.1 0 1 2 3 4 5 6

Weeks

Nor

mal

ized

pho

ton

flux

P=0.016

10

102

103

105

1

30 min day 1 week 3 4 5

ErbB

2-ve

cEr

bB2-

Num

b

104

B

ErbB2-vec (n=5)ErbB2-Smyd5 (n=5)

0 1 2 3 4 5 6 Weeks

Nor

mal

ized

pho

ton

flux

P=0.015

10

102

103

105

1

104

30 min day 1 week 3 4 5

ErbB

2-ve

cEr

bB2-

Smyd

5

Pools 1-3 288 shRNAs

Independentclones (#)

Miceinoculated (#)

Cellsinjected (#)

6 mice3 x 105

TheoreticalRedundancy

(fold)1.8 x 106

Total cellsinjected (#)

> 6 x 103

Pools 4-6 7 mice3 x 105 2.1 x 106 > 7 x 103

A

Pool 7 10 mice3 x 105 3.0 x 106 > 1 x 104

288 shRNAs

260 shRNAs

Fig. 4. Loss-of-function shRNA library screen. (A) The flowchart describesthe throughput and theoretical redundancy of the shRNA screen. (B) ErbB2-TGL cells stably transduced with empty vector or the indicated genes (Numb,Left; Smyd5, Right) were inoculated i.v. into syngeneic mice. Lung metastasiswas measured by bioluminescent imaging. The panels show representativeimages (Top), and the graphs show the normalized photon flux at the in-dicated times (Bottom).

Table 3. Results of the shRNA screen

The primary screen yielded 45 hits. The table shows the targets of shRNAs recovered from more than twodistinct lesions. Hits are rank-ordered based on their cumulative frequency of occurrence. The genes targeted byshRNAs also recovered in a secondary screen are indicated in red.

16536 | www.pnas.org/cgi/doi/10.1073/pnas.1403234111 Gao et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 3,

202

2

metastatic reactivation (43–45). We suspect that the power ofour screens stems from both the low level of background and theselectivity of the step of metastasis examined.Third, our screening strategy is scalable. Currently, genome-

wide screens are limited by constraints in the generation of li-braries encoding or targeting the entire exome, in the rate oftransduction of recipient cells, and in proper regulation of geneexpression. Despite the suboptimal coverage of our libraries, wehave been able to identify several strong mediators of reactivationby performing a limited number of screens. Based on these results,we envision that repeated or larger-scale screening of existinglibraries or screening of newly generated genome-wide libraries,comprising full-length cDNAs or guide RNAs for CRISPR-Cas9genome editing (47, 48), will enable isolation of additional genesthat govern metastatic dormancy and reactivation.Finally, the platform we have described is flexible and can be

adapted to the study of metastatic reactivation in other organs andby other cancer types or to the study of other steps of metastasis.Unlike DNA microarray-based approaches for the identificationof candidate mediators of metastasis, our gain-of-function geneticscreening strategy affords the advantage of rapid biological valida-tion of single prometastatic entities. Because of its attributes, ourscreening platform should contribute to the identification of the

core signaling pathways that regulate metastatic dormancy andreactivation and possibly facilitate the development of noveltherapeutic strategies.

Materials and MethodsRetroviral cDNA, microRNA, and shRNA screens were performed in 6- to 12-wk-old BALB/c mice. Animals were imaged in an IVIS 100 chamber, and datawere recorded using Living Image software (Xenogen). All mouse studieswere conducted in accordance with protocols approved by the InstitutionalAnimal Care and Use Committee of the Memorial Sloan Kettering CancerCenter (MSKCC). Results are reported as mean ± SD or SE. Comparisonsbetween two groups were performed using an unpaired two-sided t test(P < 0.05 was considered significant).

ACKNOWLEDGMENTS. We thank R. Agami, G. Daley, G. Hannon, S. Lowe,and F. Miller for reagents; A. Barlas, N. Fan, and Z. Zhang for help withsome of the experiments; the staff of the Genomics Core and theMolecular Cytology Core facilities of Memorial Sloan Kettering CancerCenter for their services; and members of the F.G.G. laboratory fordiscussions. This work was supported by NIH Grants R01 CA175712 andP01 CA094060, Project 4 (to F.G.G.), the G. Beene Cancer Center (F.G.G.),and National Natural Science Foundation of China Grant 81372840 (toH.G.). H.G. was partially supported by the Program for Professor of SpecialAppointment (Eastern Scholar) at the Shanghai Institutions of HigherLearning (2013–2014).

1. Weinberg RA (2007) The Biology of Cancer (Garland Science, New York).2. Nguyen DX, Bos PD, Massagué J (2009) Metastasis: From dissemination to organ-

specific colonization. Nat Rev Cancer 9(4):274–284.3. Valastyan S, Weinberg RA (2011) Tumor metastasis: Molecular insights and evolving

paradigms. Cell 147(2):275–292.4. Giancotti FG (2013) Mechanisms governing metastatic dormancy and reactivation.

Cell 155(4):750–764.5. Van Dyke T, Jacks T (2002) Cancer modeling in the modern era: Progress and chal-

lenges. Cell 108(2):135–144.6. Heyer J, Kwong LN, Lowe SW, Chin L (2010) Non-germline genetically engineered

mouse models for translational cancer research. Nat Rev Cancer 10(7):470–480.7. Zender L, Lowe SW (2008) Integrative oncogenomic approaches for accelerated

cancer-gene discovery. Curr Opin Oncol 20(1):72–76.8. Bos PD, Nguyen DX, Massagué J (2010) Modeling metastasis in the mouse. Curr Opin

Pharmacol 10(5):571–577.9. Brummelkamp TR, Bernards R (2003) New tools for functional mammalian cancer

genetics. Nat Rev Cancer 3(10):781–789.10. Chang K, Elledge SJ, Hannon GJ (2006) Lessons from Nature: MicroRNA-based shRNA

libraries. Nat Methods 3(9):707–714.11. Beronja S, et al. (2013) RNAi screens in mice identify physiological regulators of on-

cogenic growth. Nature 501(7466):185–190.12. Gao H, et al. (2012) The BMP inhibitor Coco reactivates breast cancer cells at lung

metastatic sites. Cell 150(4):764–779.13. Aslakson CJ, Miller FR (1992) Selective events in the metastatic process defined by

analysis of the sequential dissemination of subpopulations of a mouse mammarytumor. Cancer Res 52(6):1399–1405.

14. Valles I, et al. (2012) Identification of novel deregulated RNA metabolism-relatedgenes in non-small cell lung cancer. PLoS ONE 7(8):e42086.

15. Ji P, et al. (2003) MALAT-1, a novel noncoding RNA, and thymosin beta4 predict metastasisand survival in early-stage non-small cell lung cancer. Oncogene 22(39):8031–8041.

16. Lin R, Maeda S, Liu C, Karin M, Edgington TS (2007) A large noncoding RNA is amarker for murine hepatocellular carcinomas and a spectrum of human carcinomas.Oncogene 26(6):851–858.

17. Galmiche L, et al. (2011) Exome sequencing identifies MRPL3 mutation in mito-chondrial cardiomyopathy. Hum Mutat 32(11):1225–1231.

18. Lee HC, et al. (2010) Diverse pathways generate microRNA-like RNAs and Dicer-independent small interfering RNAs in fungi. Mol Cell 38(6):803–814.

19. Tripathi V, et al. (2010) The nuclear-retained noncoding RNA MALAT1 regulates al-ternative splicing by modulating SR splicing factor phosphorylation. Mol Cell 39(6):925–938.

20. Yang L, et al. (2011) ncRNA- and Pc2 methylation-dependent gene relocation be-tween nuclear structures mediates gene activation programs. Cell 147(4):773–788.

21. Kao YR, et al. (2003) Tumor-associated antigen L6 and the invasion of human lungcancer cells. Clin Cancer Res 9(7):2807–2816.

22. Gordon GJ, et al. (2009) Four-gene expression ratio test for survival in patients un-dergoing surgery for mesothelioma. J Natl Cancer Inst 101(9):678–686.

23. Borrell-Pagès M, et al. (2000) The carboxy-terminal cysteine of the tetraspanin L6antigen is required for its interaction with SITAC, a novel PDZ protein. Mol Biol Cell11(12):4217–4225.

24. Boukerche H, et al. (2010) Src kinase activation is mandatory for MDA-9/syntenin-mediated activation of nuclear factor-kappaB. Oncogene 29(21):3054–3066.

25. Hwangbo C, Kim J, Lee JJ, Lee JH (2010) Activation of the integrin effector kinasefocal adhesion kinase in cancer cells is regulated by crosstalk between protein kinaseCalpha and the PDZ adapter protein mda-9/Syntenin. Cancer Res 70(4):1645–1655.

26. Ma L, Weinberg RA (2008) Micromanagers of malignancy: Role of microRNAs in

regulating metastasis. Trends Genet 24(9):448–456.27. Png KJ, et al. (2011) MicroRNA-335 inhibits tumor reinitiation and is silenced through

genetic and epigenetic mechanisms in human breast cancer. Genes Dev 25(3):226–231.28. Png KJ, Halberg N, Yoshida M, Tavazoie SF (2012) A microRNA regulon that mediates

endothelial recruitment and metastasis by cancer cells. Nature 481(7380):190–194.29. Mavrakis KJ, et al. (2010) Genome-wide RNA-mediated interference screen identifies

miR-19 targets in Notch-induced T-cell acute lymphoblastic leukaemia. Nat Cell Biol

12(4):372–379.30. Burgess DJ, et al. (2008) Topoisomerase levels determine chemotherapy response in

vitro and in vivo. Proc Natl Acad Sci USA 105(26):9053–9058.31. Meacham CE, Ho EE, Dubrovsky E, Gertler FB, Hemann MT (2009) In vivo RNAi

screening identifies regulators of actin dynamics as key determinants of lymphoma

progression. Nat Genet 41(10):1133–1137.32. Silva JM, et al. (2005) Second-generation shRNA libraries covering the mouse and

human genomes. Nat Genet 37(11):1281–1288.33. Oricchio E, et al. (2011) The Eph-receptor A7 is a soluble tumor suppressor for fol-

licular lymphoma. Cell 147(3):554–564.34. Gönczy P (2008) Mechanisms of asymmetric cell division: Flies and worms pave the

way. Nat Rev Mol Cell Biol 9(5):355–366.35. Pece S, Confalonieri S, R Romano P, Di Fiore PP (2011) NUMB-ing down cancer by

more than just a NOTCH. Biochim Biophys Acta 1815(1):26–43.36. Stender JD, et al. (2012) Control of proinflammatory gene programs by regulated

trimethylation and demethylation of histone H4K20. Mol Cell 48(1):28–38.37. Izzi L, Attisano L (2004) Regulation of the TGFbeta signalling pathway by ubiquitin-

mediated degradation. Oncogene 23(11):2071–2078.38. Guo W, et al. (2006) Beta 4 integrin amplifies ErbB2 signaling to promote mammary

tumorigenesis. Cell 126(3):489–502.39. Lu X, et al. (2011) VCAM-1 promotes osteolytic expansion of indolent bone micro-

metastasis of breast cancer by engaging α4β1-positive osteoclast progenitors. Cancer

Cell 20(6):701–714.40. Malanchi I, et al. (2012) Interactions between cancer stem cells and their niche govern

metastatic colonization. Nature 481(7379):85–89.41. Oskarsson T, et al. (2011) Breast cancer cells produce tenascin C as a metastatic niche

component to colonize the lungs. Nat Med 17(7):867–874.42. Brabletz T, Jung A, Spaderna S, Hlubek F, Kirchner T (2005) Opinion: Migrating cancer

stem cells—an integrated concept of malignant tumour progression. Nat Rev Cancer

5(9):744–749.43. Luzzi KJ, et al. (1998) Multistep nature of metastatic inefficiency: Dormancy of solitary

cells after successful extravasation and limited survival of early micrometastases. Am J

Pathol 153(3):865–873.44. Cameron MD, et al. (2000) Temporal progression of metastasis in lung: Cell survival,

dormancy, and location dependence ofmetastatic inefficiency. Cancer Res 60(9):2541–2546.45. Wong CW, et al. (2001) Apoptosis: An early event in metastatic inefficiency. Cancer

Res 61(1):333–338.46. Ghajar CM, et al. (2013) The perivascular niche regulates breast tumour dormancy.

Nat Cell Biol 15(7):807–817.47. Hsu PD, Lander ES, Zhang F (2014) Development and applications of CRISPR-Cas9 for

genome engineering. Cell 157(6):1262–1278.48. Shao DD, et al. (2014) KRAS and YAP1 converge to regulate EMT and tumor survival.

Cell 158(1):171–184.

Gao et al. PNAS | November 18, 2014 | vol. 111 | no. 46 | 16537

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 3,

202

2