Fibrin Stiffness Mediates Dormancy of Tumor-Repopulating ...Tumor Biology and Immunology Fibrin Stiffness Mediates Dormancy of Tumor-Repopulating Cells via a Cdc42-Driven Tet2 Epigenetic

Tumor Biology and Immunology

Fibrin Stiffness Mediates Dormancy ofTumor-Repopulating Cells via a Cdc42-DrivenTet2 Epigenetic ProgramYuying Liu1,2, Jiadi Lv1, Xiaoyu Liang1, Xiaonan Yin1, Le Zhang1, Degao Chen1, Xun Jin1,Roland Fiskesund1, Ke Tang3, Jingwei Ma3, Huafeng Zhang3,Wenqian Dong1, Siqi Mo1,Tianzhen Zhang1, Feiran Cheng1, Yabo Zhou1, Jing Xie1, Ning Wang4,5, and Bo Huang1,2,3

Abstract

Dormancy is recognized as a critical biological event for tumor-igenic cells surviving in an extremely harsh environment. Under-standing the molecular process of dormancy can unlock novelapproaches to tackle cancers. We recently reported that stem-liketumor-repopulating cells (TRC) sense mechanical signals andrapidly proliferate in a 90 Pa soft fibrin matrix. Here, we showthat a stiffmechanical environment induces TRCdormancy via anepigenetic program initiated by translocation of Cdc42, a cyto-solic regulator ofmechanotransduction, into the nucleus, where itpromotes transcription of hydroxymethylating enzyme Tet2. Tet2

epigenetically activated cell-cycle–inhibiting genes p21 and p27to induce dormancy, but also caused downregulation of integrinb3 to maintain dormancy. This stiffness-mediated dormancy wasrecapitulated in mouse models for both murine and primaryhuman melanoma TRCs. These data identify an epigenetic pro-gram directed by mechanics, which drives highly tumorigenicTRCs to enter dormancy in a stiff mechanical environment.

Significance:Amechanics-directed epigenetic programenablestumor-repopulating cells to enter dormancy in a stiff mechanicalenvironment. Cancer Res; 78(14); 3926–37. �2018 AACR.

IntroductionTumor dormancy is emerging as a key event for tumor long-

term survival from escaping intrinsic (immune surveillance) andextrinsic (toxic drugs) attacks (1–3). Despite the elusiveness ofhow it is initiated andmaintained, tumor dormancy appears to bedivided into tumor mass dormancy, a process reflecting an equi-librium between tumor growth and tumor killing, and tumor celldormancy defined at a cellular level as a process of induced cell-cycle arrest (1–3). Although quiescent tumorigenic cells may stayin a dormant status for years based on clinical observations (4–6),how this dormancy process is achieved and maintained inpatients with cancer remains an enigma. Cells apply contractile

forces to sense the physical stiffness of the surrounding micro-environment and respond accordingly (7, 8). Binding of extra-cellular matrix (ECM) proteins such as collagen and fibrin tointegrins leads to mechanotransduction along clustered integrinsto focal adhesions (8, 9). Thus, outside mechanical signals aresensed at focal adhesions and can be converted into biochemicalsignals inside the cells (10, 11). In line with this concept, celldormancy can be postulated to be regulated by the stiffness ofECM. To date, whether and how stiffness regulates the dormancyof tumorigenic cells that can repopulate tumors (tumor-repopu-lating cells or TRC) is unknown. Previously, we used a soft three-dimensional (3D) fibrin gel culture system to generate TRCs (12–16). This fibrin gel corresponds to 90 Pa in elastic stiffness and theTRCs were trapped individually in the gel for colony formation.Those TRCs grew into spheroid-like morphologic shapes resem-bling stem-like cells. Importantly, as few as 10 TRCs, after intra-venous injection via the tail vein, are able to grow tumors in thelungs of immunocompetent mice (12). These mechanicallyamplified TRCs are distinct from the cancer stem cells (CSC) thatare isolated using conventional cell surface markers and areunreliable (17, 18). The TRCs also appear to be different fromthe tumor-initiating cells (TIC) that have 3 distinct subtypes (19).Here, we have further used this unique 3D fibrin gel system toexamine the possibility of matrix stiffness-driven TRC dormancy.

Materials and MethodsAnimals and cell lines

Female C57BL/6 and NOD-SCID mice, 6 to 8 weeks old, werepurchased from the Center of Medical Experimental Animals oftheChineseAcademyofMedical Sciences (CAMS,Beijing,China).These animals were maintained in the Animal Facilities of theChinese Academy of Medical Science under pathogen-free

1National Key Laboratory of Medical Molecular Biology and Department ofImmunology, Institute of Basic Medical Sciences, Clinical Immunology Center,Chinese Academy of Medical Sciences, Beijing, China. 2Clinical ImmunologyCenter, Chinese Academy of Medical Sciences, Beijing, China. 3Department ofBiochemistry and Molecular Biology, Tongji Medical College, Huazhong Univer-sity of Science and Technology, Wuhan, China. 4Department of MechanicalScience andEngineering, College of Engineering, University of Illinois at Urbana–Champaign, Urbana, Illinois. 5Laboratory for Cellular Biomechanics and Regen-erative Medicine, College of Life Science and Technology, Huazhong Universityof Science and Technology, Wuhan, Hubei, China.

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

Y. Liu, J. Lv, and X. Liang contributed equally to this article.

Corresponding Author: Bo Huang, Chinese Academy of Medical Sciences, 5Dong Dan San Tiao, Beijing 100005, China. Phone: 86-10-69156447; Fax: 86-10-65229258; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-17-3719

�2018 American Association for Cancer Research.

CancerResearch

Cancer Res; 78(14) July 15, 20183926

on January 22, 2021. © 2018 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst May 15, 2018; DOI: 10.1158/0008-5472.CAN-17-3719

http://crossmark.crossref.org/dialog/?doi=10.1158/0008-5472.CAN-17-3719&domain=pdf&date_stamp=2018-6-28

http://cancerres.aacrjournals.org/

conditions. Mouse tumor cell lines B16 (melanoma), 4T1 (breastcancer) and H22 (liver cancer), and human tumor cell lines A375(melanoma), A549 (lung cancer) and HepG2 (liver cancer) werepurchased from the China Center for Type Culture Collection andcultured in DMEM (Thermo Fisher Scientific) with 10% fetalbovine serum (FBS; Gibco). Murine H22 hepatocellular carcino-ma cell line was grown in RPMI1640 medium (Gibco) with10% FBS. Cells were tested forMycoplasma detection, inter-speciescross contamination and authenticated by isoenzyme and short-tandem repeat (STR) analyses in Cell Resource Centre of PekingUnion Medical College before the study. Cell lines used in theexperiments were within 20 passages.

3D fibrin gel cell culture of tumor cellsTRC culture was conducted according to our previously

described method (12). Briefly, fibrinogen (Searun HoldingsCompany) was diluted into 2mg/mL or 16mg/mLwith T7 buffer(pH 7.4, 50 mmol/L Tris, 150 mmol/L NaCl). A 1:1 fibrinogenand cell solutionmixturewasmade bymixing the same volumeofthefibrinogen solution and the cell solution, resulting in 1mg/mL(90 Pa) or 8 mg/mL (1,050 Pa) fibrinogen and 5,000 cells/mL inthe mixture. A total of 250 or 50 mL cell/fibrinogen mixtures wereseeded into each well of 24- or 96-well plate and mixed well withpreadded 5 or 1.5 mL thrombin (0.1 U/mL, Searun HoldingsCompany). The cell culture plate was then moved into 37�C cellculture incubator for 30 minutes. Finally, DMEM medium con-taining 10% FBS and antibiotics was added. In order to harvestTRCs, 3D fibrin gels with TRCs were treated with dispase II(Roche) for 10 minutes at 37�C.

ReagentsMethotrexate, paclitaxel, ZCL278, ML141, wheat germ agglu-

tinin (WGA), and latrunculin were from Sigma-Aldrich. Salmonfibrinogen, thrombin, and dispase were from Reagent Proteins.Puromycin was from Invitrogen.

Animal experiments and treatment protocolAll studies involving mice were approved by the Animal Care

and Use Committee of the CAMS. C57BL/6 or NOD-SCID micewere subcutaneously embeddedwith 3�3mm90or 1,050Pa3Dfibrin gels with 5 B16 TRCs, GFP-SGRNA B16 TRCs, Cdc42-SGRNA B16 TRCs, Tet2-SGRNA, Cdkn1b-SGRNA1, Cdkn1b-SGRNA2, Cdkn1a-SGRNA1, Cdkn1a-SGRNA2, Cdkn1a/b-SGRNA,WT-Cdc42, NLS-Cdc42, or NES-Cdc42 B16 TRCs. The incidenceof tumor in mice, tumor size or tumor weight was recorded atindicated time point. For some experiments, C57BL/6 mice weresubcutaneously transplanted 5 B16 TRCs within 1,050 Pa fibringels pretreatedwith orwithout dispase (10mg/mL). In some cases,the above mice were treated with dispase (40 mg/mouse) on day30 once every 3 days for another 40 days. Tumor formation wasrecorded. All the animals were allocated randomly.

Isolating primary human melanoma cellsMelanoma tissues were taken frommice or patients. The study

was approved by the Clinical Trial Ethics Committee of PekingUnion Medical College. All patients have provided writteninformed consent to participate in the study. Then, these tissueswere cut into small pieces of 1 to 3 mm and minced in RPMI1640 medium supplemented with collagenase (Sigma-Aldrich,32mg/mL), hyaluronidase (Sigma-Aldrich, 500mg/mL) andDNA-ase I (Sigma-Aldrich, 5 mg/mL). These samples were incubated for

1 hour at 37�C under continuous rotation. After digestion, thesamples were filtered through Cell Strainer (BIOFIL). After cen-trifuged and lysed in RBC lysis buffer, the tumor cells wereresuspended in RPMI 1640 medium with 10% FBS, penicillinand streptomycin. The primary melanoma cells were confirmedby immunostaining with anti-S100 antibody.

ResultsElevation of matrix stiffness induces the entry of TRCs intodormancy

Physiologically, the stiffness of many tissues ranges fromapproximately 100 Pa to 3,000 Pa (20). Within the tissue, cellsexhibit amatched stiffness (15, 20).Given around100Pa stiffnessofmurine B16melanoma TRCs (12), we chose 90, 450, and 1,050Pa as soft, medium, and stiff fibrin gels, respectively, to study theinfluence of matrix stiffness on TRC growth. We found thatalthough TRCs grew rapidly and formed spheroids effectively in90 Pa gels, their growth was retarded and almost completelyinhibited in 450 and 1,050 Pa gels, respectively (Fig. 1A), con-sistent with previous report (15). In addition, further increase ofgel stiffness to 2,000 Pa also caused TRCs ceasing growth (Sup-plementary Fig. S1A); but in two-dimensional (2D) rigid plastic,these TRCs started to differentiate and proliferated rapidly like thegrowth of conventional tumor cells in plastic (Supplementary Fig.S1B). Moreover, when fibrin gels were crosslinked with factor XIIIto increase the stiffness, TRCs' growth was also remarkablydecreased (Supplementary Fig. S1C), suggesting that fibrin gelstiffness is capable of inhibiting TRC growth. Also, human mel-anoma A375 TRCs displayed a quiescent state in 1,050 Pa gels(Fig. 1B). In addition, TRCs from other tumor types (H22, A549,and HepG2) consistently grew rapidly in 90 Pa fibrin gels, whiletheir growth was inhibited in 1,050 Pa gels (Supplementary Fig.S1D—S1F). Such growth inhibition was not due to stiffness-caused TRC damage, because TRCs isolated from 1,050 Pa gelscould rapidly regain the growth in 90 Pa fibrin gels (Fig. 1A and B;Supplementary Fig. S1D—S1F). Thus, these observations suggestthat TRCsmight enter a dormant state in stiff fibrin gels. Recently,the orphan nuclear receptor COUP-TF1 or DEC2 has been iden-tified as tumor cell dormancymarker (21, 22). Both B16 andA375TRCs with Ki-67 negative staining were found to express COUP-TF1 orDEC2 in 1,050 Pa fibrin gels (Fig. 1C andD). In addition tothis molecular marker, cell dormancy can be functionally definedas G0–G1 arrest, being neither apoptotic nor senescent, consum-ing less glucose, decreasing responses to stimulation, andregrowth once the dormancy stimuli are removed. Based on thesecriteria, we found that B16 TRCs indeed underwent G0–G1 cell-cycle arrest, downregulated the expression of the PCNA gene (acell proliferation marker) and decreased glucose consumption(Fig. 1E–G). In addition, B16 TRCs did not undergo senescence asevaluated byb-galactosidase assay (Fig. 1H).Dormant tumor cellsmay decrease their response to xenobiotics, including chemother-apeutic drugs (2, 23, 24). Interestingly, our data showed that B16TRCs in 1,050 Pa fibrin gels were more resistant to methotrexateand paclitaxel than those in 90 Pa gels (Fig. 1I). The above-mentioned dormant properties were also observed in A375 TRCsin 1,050 Pa fibrin gels (Fig. 1E–I). To further confirm that TRCdormancy is induced by stiff fibrin gels, we added dispase enzymeto partially degrade 1,050 Pa fibrin gels and found that dormantTRCs resumed growth in a dispase dose-dependent manner(Fig. 1J), which could not be ascribed to the direct effect of dispase

Matrix Stiffness Induces TRC Dormancy

www.aacrjournals.org Cancer Res; 78(14) July 15, 2018 3927




on TRCs, because dispase had no effect on TRC growth (Supple-mentary Fig. S1G). Together, these data suggest that increasing 3Dmatrix stiffness alone is enough to induce TRCs into a dormantstate.

Matrix stiffness mediates the translocation of cytosolic Cdc42into the nucleus for TRC dormancy

Next, we investigated how the elastic stiffness-generatedmechanical signal was transmitted into a dormant signal. SmallG protein Cdc42, amember of the Rho family, plays a pivotal roleinmechanotransduction (25). Extracellular mechanical forces aresensed by integrins, transmitted to talin-based focal adhesions(FA) via integrins, propagated in the cytoplasm via F-actins (10),

and converted into biochemical signals at FAs (11) and othercytoplasmic sites (26). During this process, Cdc42 is recruited tothe focal adhesion and regulates the biogenesis of F-actin (27).Using a CRISPR-Cas9 technology, we found that Cdc42 knockouthad no effect on B16 cell proliferation in rigid plastic (Supple-mentary Fig. S2A) but led to dormant B16 TRC regrowth in1,050Pafibrin gels (Fig. 2A).Consistently, use ofCdc42 inhibitor,ZCL278 orML141, also resulted in the regrowth of B16 TRCs (Fig.2B), suggesting that Cdc42 regulates matrix stiffness-mediatedTRC dormancy. Cdc42 was located under the plasma membraneof differentiated B16 cells on rigid plastic (Fig 2C, top), whichhowever were differentially located in the cytosol of B16 TRCs in90 Pa fibrin gels (Fig. 2C, middle). Surprisingly, in 1,050 Pa stiff

Figure 1.

Three-dimensional stiff matrix induces TRC dormancy in vitro.A and B, B16 (A) or A375 (B) cells were seeded in 90, 450, or 1,050 Pa 3D fibrin gels for 5 days culture.The formed colony sizes at different time points are indicated (left, middle). In addition, the TRCs after 5 days culture in 90 and 1,050 Pa fibrin gels werereseeded into 90Pafibrin gels for colonygrowth as indicated (right). The colony size formed in90Paond1was set to 1. Bar, 50mm.C andD, Immunostaining ofCOUP-TF1 or hDEC2 and Ki-67 from B16 (C) or A375 (D) TRCs grown in 90 or 1,050 Pa fibrin gels for 3 days. Bar, 10 mm. E, Cell cycle was analyzed on day 3 in B16TRCs grown in 90and 1,050Pafibrin gels.F,B16orA375TRCswere cultured in 90or 1,050Pa 3Dfibrin gels for 3 days. ThemRNAexpression ofPCNAwasdeterminedby real time PCR. G, Glucose contents in the supernatants were measured with glucose detection kit. The glucose consumption decreased in 1,050 Pa morethan in 90Pafibrin gels.H,B16orA375TRCswere cultured in 90or 1,050Pafibrin gels for 3 daysand then treatedwith orwithout IFNg (100ng/mL) /TNFa (10 ng/mL)for 72 hours. SA-b gal staining was conducted. I, B16 or A375 TRCs cultured in 90 Pa fibrin gels or 1,050 Pa fibrin gels were treated with methotrexate (MTX,500ng/mL) or paclitaxel (PAX, 5mg/mL) for 24hours. Flowcytometry analysiswas performed todetermine cell apoptosis by stainingAnnexin V. J,B16orA375TRCscultured in 1,050 Pa fibrin gels were treated with different concentrations of dispase as indicated and the colony sizes were measured after 72 hours. Bar,50 mm.Data shown are representative of three independent experiments and error bars representmean� SEM; N.S., no significant difference; � , P <0.05; �� , P <0.01;�� , P < 0.001, by one-way ANOVA (A, B, H, and J) or Student t test (C–G and I).

Liu et al.

Cancer Res; 78(14) July 15, 2018 Cancer Research3928




fibrin gels, Cdc42 existed in the nucleus rather than the cytosol ofB16 TRCs (Fig. 2C, bottom). Similar nuclear location of Cdc42was also observed in 2,000 Pa gels (Supplementary Fig. S2B).Such distinct cytoplasmic and nuclear distribution of Cdc42was also observed in A375 TRCs (Supplementary Fig. S2C).This unusual observation was further supported by the additionof a nuclear pore inhibitor, WGA, which led to Cdc42 forming acircle around the nucleus (Fig. 2D). Intriguingly, blocking theentry of Cdc42 into the nucleus resulted in the regrowth of B16TRCs in 1,050 Pa gels (Fig. 2E). We thus constructed NLS- andNES-tagged Cdc42-expressing vectors to further clarify the roleof nuclear Cdc42 in TRC dormancy. We found that the growthof B16 or 4T1 TRCs in 90 Pa fibrin gels was completelyinhibited by importing Cdc42 into the nucleus (NLS-Cdc42),and exporting Cdc42 from the nucleus (NES-Cdc42) did notinhibit the growth (Fig. 2F; Supplementary Fig. S2D and S2E).In addition, although the above Cdc42 knockout led to dor-mant B16 TRC regrowth in 1,050 Pa fibrin gels (Fig. 2A), thetransduction of NLS-Cdc42 but not NES-Cdc42 was able toreset these proliferating TRCs into a quiescent state again(Fig. 2G; Supplementary Fig. S2F), suggesting that the entryof Cdc42 into the nucleus is a critical step for matrix stiffness-induced TRC dormancy. Microtubules play an important rolein transporting molecules from the cytosol to nucleus (28).As expected, the use of the microtubule inhibitor colchicinecould block the entry of Cdc42 into the nucleus in 1,050 Pafibrin gels (Fig. 2H). Together, these data suggest that Cdc42, asa cytosolic mechanotransducing molecule, is transported bymicrotubules to the proximity of nuclear membrane and fur-ther translocated into the nucleus via nuclear pores in order toregulate TRC dormancy.

Nuclear Cdc42 promotes the hydroxymethylation of Cdkn1aand Cdkn1b via upregulating Tet2

Next, we investigated the molecular mechanism throughwhich nuclear Cdc42 regulates TRC dormancy in stiff fibrin gels.Epigenetic modifications, including DNA methylation and his-tone posttranslational modifications, are known to governgenetic programs and profoundly affect many cellular pathways,including essential biological processes for managing cellularresponses to environmental stimuli (29–31). Currently, a criticalrole of ten–eleven-translocation methylcytosine dioxygenases(Tet) in epigenetic modification is highlighted (32, 33). Ourprevious studies have shown that integrin avb3-sensed mechan-ical forces regulate histone epigenetic modification in B16 TRCs(13, 15). Here, we further hypothesized that Tets were involvedin 1,050 Pa fibrin gel-mediated TRC dormancy through regula-tion of epigenetic modifications. Comparing 90 and 1,050 Pafibrin gels, we found that the increase of stiffness resulted in theupregulation of the mRNA expression of Tet2 and Tet3, but notTet1 (Supplementary Fig. S3A). However, fluorescent stainingshowed that a majority of Tet2 protein, but not Tet1 or Tet3, waslocalized in the nucleus of B16 TRCs in 1,050 Pa gels in a time-dependent manner (Supplementary Fig. S3B and S3C). There-fore, we focused on Tet2 and found that Tet2 knockout resultedin the regrowth of B16 TRCs in 1,050 Pa gels (SupplementaryFig. S3D), suggesting that Tet2 plays an important role instiffness-mediated TRC dormancy. Regarding the colocalizationof Cdc42 and Tet2 in the nucleus, we further investigated apossible link between these two molecules. We found that Tet2knockout did not affect the entry of Cdc42 into the nucleus(Fig. 3A), however, Cdc42 knockdown significantly downregu-lated the expression of Tet2 (Fig. 3B; Supplementary Fig. S3E).

Figure 2.

Three-dimensional matrix stiffness promotes Cdc42 translocation into the nucleus to induce TRC dormancy.A, B16 TRCs or Cdc42-SGRNAs-B16 TRCswere culturedin 1,050 Pa gels for different times as indicated. The colony size was presented graphically. B, B16 TRCs cultured in 1,050 Pa fibrin gels were treatedwith Cdc42 inhibitor ZCL278 (10 mmol/L) or ML141 (10 mmol/L) for 48 hours. The colony size was presented photographically and graphically. Bar, 50 mm. C, B16 cellsgrown on rigid 2D plastic and B16 TRC grown in 90 or 1,050 Pa 3D fibrin gels were immunostained with anti-Cdc42 (red) and DAPI (blue), and imaged by confocalmicroscopy. Bar, 10 mm. D, Nuclear pore inhibitor WGA (25 ng/mL) was induced into the nucleus of B16 TRCs by electroporation transformation, and thecells were seeded in 1,050 Pa 3D fibrin gels for 24 hours. These cells were fixed and immunostained with anti-Cdc42 (red) and DAPI (blue) and observed underconfocal microscope. Bar, 5 mm. E, B16 TRCs in 1,050 Pa fibrin gels were treatedwithWGA (25 ng/mL) for 2 hours and then cultured for another 72 hours. The colonysize was presented photographically and graphically. F, B16 cells stably expressing Cdc42-WT, Cdc42-NLS, or Cdc42-NES were seeded in 90 Pa fibrin gels. Theformed colony sizes are indicated at different time points. The colony size of vector-transduced B16 on day 1 was set as 1. G, Cdc42-SGRNA-B16 cells stablyexpressing Cdc42-WT, Cdc42-NLS, or Cdc42-NES were seeded in 1,050 Pa fibrin gels. The formed colony sizes are indicated at different time points. H, B16 TRCsin 1,050 Pa fibrin gels were treated with the microtubule inhibitor colchicine (1 mmol/L) for 30 minutes, and then colchicine was removed from the supernatant.After another 24-hour culturing, B16TRCswerefixedand immunostainedwith anti-Cdc42antibody (red) andDAPI (blue). Bar, 5mm.Data are from three independentexperiments and error bars represent mean � SEM. �� , P < 0.01; ��, P < 0.001, by one-way ANOVA (A–C, G, and F) or Student t test (D, E, and H).






Consistently, blocking the entry of Cdc42 into the nucleus byeither Cdc42 inhibitor or microtubule inhibitor also resulted inTet2 downregulation (Supplementary Fig. S3F). Moreover, thechromatin immunoprecipitation (ChIP)-qPCR result showed

that Cdc42 bound to the promoter of tet2 gene in B16 TRCs(Fig. 3C). Notably, we found that Cdc42 was present in thenucleus in a GTP-bound active form (Supplementary Fig. S3G),suggesting that the active form of Cdc42 in the nucleus probably

Figure 3.

Nuclear Cdc42 upregulates Tet2 expression that hydroxymethylates p21 and p27 genes for TRC dormancy. A, B16 TRCs or Tet2-SGRNAs-B16 TRCs were culturedin 1,050 Pa gels for 4 days. Then, these colonies were fixed and immunostained with anti-Cdc42 antibody (red) and DAPI (blue). Bar, 5 mm. B, B16 TRCsor Cdc42-SGRNAs-B16 TRCs were cultured in 1,050 Pa gels for 4 days. Then, these cells were isolated for immunostaining of Tet2 (red). Bar, 10 mm. C, ChIP-qPCRanalysis of Cdc42 binding to the promoter of Tet2 in B16 TRCs. ChIP-qPCR analysis was performed on B16 TRCs cultured in 90 Pa fibrin gels or in 1,050 Pafibrin gelswith anti-Cdc42 antibody andqPCRprimers specific for theTet2 promoter.D,Genome-widemappingof 5-mCand5-hmC inB16 TRCcultured in 90or 1,050Pa fibrin gels. Gene counts at the 5-hmC level were different in 90 Pa and 1,050 Pa B16 TRCs and the KEGG pathway analysis result for the associated genesis shown (left). RRHP analysis of B16 TRCs from 90 or 1,050 Pa fibrin gels was performed (n ¼ 2 for each group, right). E, The 5-mC and 5-hmC levels in Cdkn1aand Cdkn1b genes from B16 TRCs cultured in 90 Pa or 1,050 Pa fibrin gels are shown. F, Cdkn1a and Cdkn1b enrichment was analyzed by ChIP-qPCR assay. G,ThemRNAexpression (left) and protein expression (right) ofCdkn1a andCdkn1b fromB16 TRCs in 90Pa or 1,050 Pa fibrin gels.H and I, Tet2-SGRNAs-B16 TRCs (H) orCdc42-SGRNAs-B16 TRCs (I) and the control SGGFP-B16 TRCs were cultured in 1,050 Pa fibrin gels. The mRNA expression and protein expression of p27and p21 were analyzed. J, SGGFP-B16 TRCs, Cdkn1b-SGRNAs-B16 TRCs, or Cdkn1a-SGRNAs-B16 TRCs were cultured in the 3D 1,050 Pa fibrin gels for 3 to 5 days.The colony size was presented photographically and graphically. Bar, 50 mm. The data represent mean � SEM. N.S., no significant difference; ��, P < 0.01,by one-way ANOVA (A, B, and H–J) or Student t test (C, E, F, and G).

Liu et al.





acts as transcriptional activator that promotes Tet2 expressionfor TRC dormancy.

Next, we examined target genes by Tet2, which regulated TRCdormancy in 1,050 Pa fibrin gels. Because Tet2 exerts its functionby catalyzing 5-mC to 5-hmC,we thus analyzed 5-mCand5-hmCof B16 TRCs by reduced representation of hydroxymethylcytosineprofiling (RRHP) sequencing. Although the distribution patternof 5-mC or 5-hmC at the genome-wide levels revealed no differ-ence between 90 Pa and 1,050 Pa fibrin gels (Supplementary Fig.S3H) gene-specific changes of 5-hmC were observed under the

1,050 Pa condition (Fig. 3D). The KEGG pathway enrichmentanalysis indicated that those changed genes were closely associ-atedwith pathways related to variousmechanical forces includingECM-receptor interaction, adherens junction, focal adhesion,actin cytoskeleton, and cell-cycle regulation (Fig. 3D). Becausecell-cycle arrest is a typical feature of dormancy, various cell-cycle–related genes were then analyzed. We did not find the 5-mC and5-hmC alteration of cyclin-dependent kinases such as Cdks 2, 4,and 6 (Supplementary Fig. S3I), however genes encoding cell-cycle inhibitors p21 and p27 were highly hydroxymethylated

Figure 4.

TheCdc42–Tet2 signaling pathwaydownregulates integrinb3 expression for TRCdormancymaintenance.A,B16TRCsweregrown in90Paor 1,050Pafibrin gels andthe distribution of Cdc42 (red) was visualized by confocal microscopy at different time points during the culture. Bar, 10 mm. B, B16 TRCs were grown in1,050 Pa fibrin gels and the expression of F-actin (actin fibers, green) was visualized by confocal microscopy at day 1 (d1) or day 3 (d3). Bar, 10 mm. C, The mRNAexpression of Itgav from B16 TRCs cultured in 90 or 1,050 Pa fibrin gels. D, B16 or A375 TRCs were cultured in 1,050 Pa fibrin gels for 3 days. The mRNAand protein expression of integrin b3 was determined by real-time PCR (left) and Western blot (right). E, The same as D, but the expression of integrin b3 inthe membrane of B16 TRCs was determined by flow cytometry. F, B16 TRCs were cultured in 90 or 1,050 Pa fibrin gels for different time and the expression ofintegrin b3was determined by flow cytometry. G, ThemRNA expression (left) and protein expression (right) of integrin b3 in SGGFP-B16 TRCs or Tet2-SGRNAs-B16TRCs. H, B16 cells stably expressing Cdc42-Wt, Cdc42-NLS, or Cdc42-NES were seeded in 90 Pa fibrin gels for 4 days. The expression of integrin b3 wasdetermined byWestern blot. Error bars, mean� SEM; n¼ 3 independent experiments; N.S., no significant difference; �� , P < 0.01, by one-way ANOVA (A, G, and H)or Student t test (B–F).






under the 1,050 Pa condition (Fig. 3E). To further confirm thisresult, we conducted ChIP-qPCR using primers specific for therespective 5-hmC-containing region ofCdkn1a andCdkn1b genes.As expected, the analysis revealed an enrichment of 5-hmC inthese two genes (Fig. 3F). In line with the increased hydroxy-methylation of the Cdkn1a and Cdkn1b genes in 1,050 Pa gels,their expression was markedly upregulated (Fig. 3G). This induc-tionofCdkn1a andCdkn1b expressionwas readily inhibited by theknockout of Tet2 (Fig. 3H), suggesting that Tet2 indeed hydro-xymethylates Cdkn1a and Cdkn1b and promotes their expression.Consistently, the knockout ofCdc42 to downregulate Tet2 expres-sion also resulted in the downregulation of p21 and p27 in 1,050Pa gels (Fig. 3I). On the other hand, the overexpression of Cdc42resulted in theupregulationof p21andp27expression,whichwasfurther augmented by Cdc42-NLS overexpression; however,Cdc42-NES overexpression has no effect on p21 and p27 expres-sion (Supplementary Fig. S3J). To determine whether the resul-tant p21- and p27-mediated TRC dormancy in 1,050 Pa fibringels, we constructed B16 cell subline with stable knockout ofCdkn1a, Cdkn1b or both. We found that Cdkn1a, Cdkn1b or dualknockout led to the regrowth of B16 TRCs in 1,050 Pa stiff fibringels but without significant difference among the three groups(Fig. 3J; Supplementary Fig. S3K and S3L). In addition, from theunshown data, the knockout Tet2, Cdkn1a, or Cdkn1b did notaffect tumor cell proliferation in rigid plastic. Together, the datasuggest that Tet2 promotes the expression of cell-cycle inhibitorsp21 and p27 by hydroxymethylation of 5-mC to 5-hmC of thesetwo genes, leading to TRC dormancy.

Dormancy maintenance is explained by downregulatedintegrin b3

The above data showed that matrix stiffness induced theformation of TRC dormancy. We then further examined how thisdormancy was maintained in the stiff matrix. Given the criticalrole of cytosolic Cdc42 in force transduction and its involvementof TRC dormancy in the nucleus, we analyzed the expression/distribution kinetics ofCdc42 inB16TRCs thatwere seeded in stiff1,050 Pa fibrin gels. The immunostaining result showed thatunlike most Cdc42 existing in the cytoplasm in 90 Pa soft gels,Cdc42 was initially located in the cytosol and nucleus in 1,050 Pastiff gels at day 1, but a striking accumulation in the nucleus and amarked decrease in the cytoplasm were observed at day 3 (Fig.4A). Intriguingly, this distinctive distribution of Cdc42 did notshow difference between d3 and d5 (Fig. 4A), suggesting that TRCdormancy is initiated and formed during the first 3 days andmaintained after 3 days in the 1,050 Pa fibrin gels. In line withCdc42 regulating the biogenesis of F-actin (27), the overexpres-sion of Cdc42 by transfection of Cdc42-WT, -NES, or -NLS vectorinto B16 TRCs enhanced F-actin expression (Supplementary Fig.S4A). Significantly, the expression of F-actin decreased graduallyin B16 TRCs during the culture in 1,050 Pa (Fig. 4B), coupledwiththe simultaneous downregulation of Cdc42. The decreased F-actin and Cdc42 imply a decreasedmechanical force transductionin TRCs. Because the force transduction is initiated by integrinavb3 in fibrin gels (12, 34), we thus determined the expression ofintegrin avb3. Although the expression of Itgav was not altered(Fig. 4C; Supplementary Fig. S4B), b3 was markedly downregu-lated in B16 and A375 TRCs in 1,050 Pa stiff fibrin gels (Fig. 4Dand E). Further analyzing the time points, we found that theexpressionofb3hada stable level in90Pafibrin gels but graduallydecreased in 1,050 Pa gels within the initial 72 hours (Fig. 4F).

Intriguingly, knockout of Tet2 upregulated the expression ofintegrin b3 in B16 TRCs in 1,050 Pa gels (Fig. 4G), concomitantwith their rapid growth (Supplementary Fig. S3C), and transduc-tionofNLS-Cdc42but notNES-Cdc42 into B16TRCs in 90Pa gelsalso downregulated b3 expression (Fig. 4H). In addition,although Tet2 knockout resulted in promotion of TRC growthin 1,050Pafibrin gels, integrin b3blockade abrogated this process(Supplementary Fig. S4C). This requirement of b3 integrin forTRC growth was also observed in Itgb3-knockout B16 TRCs,whose growth was hindered in 90 Pa fibrin gels (SupplementaryFig. S4D). Together, these results suggest that the Cdc42–Tet2pathway is responsible for downregulating integrin b3 expressionso to decrease mechanotransduction in TRCs exposed to a stiffmechanical environment.

Stiff matrix induces TRC entering dormancy in vivoWe went on to investigate whether stiffness-induced TRC

dormancy could be modeled in vivo. To address this question,we adapted the soft 3D fibrin gel method to establish a modelusing 5 TRCs to form a tumor in wild-type immunocompetentmice (Fig. 5A).Weput 5 B16melanomaTRCswithin a small pieceof 90 Pa soft fibrin gel (around 50 mL volume) and inserted the gelinto a subcutaneous site of awild-typeC57BL/6mouse.We foundthat a melanoma was formed by the 5 TRCs in 19 out of 20 micewithin 3 weeks (Fig. 5B). However, if we put 5 TRCs into 90 Patype I collagen gel, no tumor was formed in mice (Fig. 5B),suggesting that the collagen gel is not suitable for TRC growth.Actually, around 80% B16 tumor cells underwent apoptosis aftera 24-hour growth in 90 Pa 3D collagen gel, compared with 20%apoptosis in the 90 Pa fibrin gel (Fig. 5C). In addition, when weseeded TRCs in 1,200 Pa 3D collagen gels, these TRCs also did notgrow and died on day 4 (Supplementary Fig. S5A). Also, we foundthat Cdc42 was not located in the nucleus but in the cytoplasm,when TRCs were cultured in 3D collagen gels with 90, 450, 1,050,or 1,200 Pa stiffness (Supplementary Fig. S5B).We then put 5 B16TRCs into 1,050 Pa stiff fibrin gels, and embedded this stiff gelunder the skin of mice (n ¼ 20). Surprisingly, no visible mela-noma was formed (Fig. 5D). While, even though nomacroscopicmelanoma was formed, microtumor nodules were stained bymelanoma-specific S100 antibody and confirmedbyhematoxylinand eosin (H&E) staining (Fig. 5E). Such microtumor nodulesshowed aCOUP-TF1þKi67� dormant phenotype (Fig. 5F), whichwas also evidenced by unchanged sizes on day 30, 50, or 70(Fig. 5E), suggesting that the stiffness of matrix induces TRCs intodormancy in vivo. In line with the long-term dormancy of themicronodules, the initially embedded 1,050 Pa fibrin persistentlyexisted, as evaluated on day 60 (Supplementary Fig. S5C). Thismight be due to an excellent histocompatibility of salmon fibrindue to its low immunogenicity and low virus infection propensity(35). Then, we used dispase to treat 5 TRC-contained 1,050 Pafibrin gels, followed by implanting the gel into the mice (n¼ 20).Surprisingly, 19 visible tumors were formed (Fig. 5D), suggestingthat fibrin-generated stiffness is crucial to induce TRCs into dor-mancy. Then, we tested whether breaking fibrin stiffness couldwake dormant microtumor nodules up. Mice, 30 days after beingembedded with 5 B16 TRCs-containing 1,050 Pa fibrin gels, weretreated locally with dispase. Forty days later, macroscopic tumorswere observed in 12 out of 20 mice (Fig. 5G), which was alsoconfirmed by anti-S100 andH&E staining (Fig. 5H). Consistently,the immunostaining of tumor section showed that dispase treat-ment remarkably decreased the percentage of COUP-TF1þKi67�

Liu et al.





cells (Fig. 5I). To verify that stiffness-induced dormancy is alsomediated through the above Cdc42–Tet2 pathway in vivo, weconducted immunohistologic analysis of the above microtumorsand found that the expression ofCdc42 andp27were upregulated(Supplementary Fig. S5D and S5E). Notably, Cdc42 was mainlylocated in the nucleus of the tumor cells (Supplementary Fig.S5D), suggesting that the Cdc42–Tet2 pathway is involved inregulating TRC dormancy in vivo. To further confirm this, we usedCdc42- and Tet2-knockedout TRCs to repeat the above in vivoexperiment, respectively. The result showed that once Cdc42 orTet2was knocked out, the above 5 B16 TRCs in 1,050 Pa fibrin gel

were able to grow a visible tumor in mice (Fig. 5J). Consistently,the knockout of Cdkn1a, Cdkn1b, or both also resulted in thegrowth of a visible tumor in mice in the above settings (Fig. 5K;Supplementary Fig. S5F and S5G). In addition, when we seeded 5TRCs overexpressing Cdc42-WT, -NES or NLSwith 1,050 Pa fibringel and inserted the gel into the subcutaneous site of mice, wefound that visible melanoma was formed in the Cdc42-NESgroup, while only microscopic tumor nodules with unchangedsizes on days 20 and 30 were formed in Cdc42-WT and Cdc-NLSgroups (Supplementary Fig. S5H). Meanwhile, these microscopicnodules showed a COUP-TF1þKi67� dormant phenotype

Figure 5.

Stiff matrix induces B16 TRC dormancy in vivo. A, Schematic diagram of transplanting B16 TRCs within 90 or 1,050 Pa gel into mice. B, Tumorigenicity of 5 B16 cellsembedded in 90 Pa 3D fibrin gels or 90 Pa 3D collagen I gels in C57BL/6 mice for 3 weeks. C, B16 cells were seeded in 3D 90 Pa fibrin or collagen I gelsfor 24 hours. Flow cytometry analysis was performed to determine cell apoptosis by staining Annexin V. D, Five B16 cells from rigid plastic or B16 TRCs within 90 or1,050 Pa 3D fibrin gel were implanted below the skin of C57BL/6 mice treated with or without dispase (40 mg/mouse) for 3 weeks. The tumor formation wasrecorded. E, Five B16 TRCs within 1,050 Pa gel were implanted into C57BL/6 mice. On days 30, 50, and 70, tissues at the implanted site were analyzed byimmunostaining against S100 or H&E staining. The tumor size is presented photographically and graphically. Bar, 30 mm. F, Tissues containing implanted tumorcell-gel were immunostained with anti-COUP-TF1, K-i67, and DAPI antibodies. Bar, 10 mm. G –I, the same as E, but mice were treated with PBS or 40-mgdispase onday 30of implantation for another 40days. Tumor formationwas recorded (G), and tissues fromeachgroupwere immunostainedwith anti-S100 andH&Estaining (H; bar, 50 mm) or anti-COUP-TF1/Ki-67 staining (I; bar, 10 mm). J, Five GFP-SGRNA-B16 TRCs, Cdc42-SGRNA-B16 TRCs, or Tet2-SGRNA-B16TRCswithin 1,050 Pa fibrin gel were implanted into C57BL/6mice. Twenty days after implantation, the tumor formationwas recorded. Also, the tumor wasweighed.Bar, 1 cm. K, Five GFP-SGRNA-B16 or Cdkn1b-SGRNAs-B16 TRCs were implanted into C57BL/6 mice. Thirty days after implantation, the tumor formation wasrecorded and weighed. Bar, 1 cm. The data represent mean � SEM. N.S., no significant difference; �� , P < 0.01; �� , P < 0.001, by Student t test (C, F, H, and I)or one-way ANOVA (E, J, and K).






(Supplementary Fig. S5I). Together, these data suggest that the invitro identified Cdc42–Tet2 pathway is also required for in vivoTRC dormancy.

Fibrin stiffness induces dormancy in primary humanmelanoma TRCs via the Cdc42–Tet2 pathway

Finally, in order to demonstrate the clinical importance of theCdc42–Tet2 pathway, we isolated primary human melanoma(PHM) cells and seeded them in 90 or 1,050 Pa 3D fibrin gelsfor 5 days' culture. Consistently, PHM TRCs were also inducedinto dormancy by the 1,050 Pa stiffness (Fig. 6A), which wasevidenced by cell-cycle arrest without senescence, lower glucoseconsumption and highly expressing COUP-TF1þKi67� dormantmarkers (Fig. 6B–D; Supplementary Fig. S6A). Reseeding suchdormant cells into 90 Pa fibrin gels caused their regrowth (Fig.6A).Or this dormant state could also be broken by the addition ofdispase to degrade fibrin in a dose-dependent manner (Supple-mentary Fig. S6B), suggesting that stiffness also induces primaryhuman TRCs into dormancy. Consistently, this stiffness-induceddormancy was also mediated by Cdc42. Treatment with Cdc42inhibitor ZCL278 or ML141 restored dormant TRCs of PHM togrow in stiff 1,050 Pa fibrin gels (Supplementary Fig. S6C).Moreover, Cdc42 was also translocated from the cytosol to thenucleus and its expression was elevated in line with Tet2 expres-sion in response to a stiff mechanical environment as shown byconfocal microscopy andWestern blot (Fig. 6E and F). Consistentwith the results fromB16 TRCs, stiffness also upregulated p21 andp27 expression in primary human melanoma TRCs in stiff 1,050Pa fibrin gels (Fig. 6G) and knockdown of CDKN1A or CDKN1Bwas able to restore the dormant TRC growth (Supplementary Fig.S6D and S6E). Significantly, transplantation of fibrin-embedded5 PHMTRCs into NOD-SCIDmice produced a high rate of tumorformation at 90 Pa (13/16) and no tumor formation in vivo at1,050 Pa (Fig. 6H), demonstrating the importance of environ-mental stiffness in regulating human TRC dormancy in vivo.Together, these data suggest that fibrin matrix stiffness induceshuman melanoma TRCs into dormancy through a Cdc42–Tet2–p27 pathway.

DiscussionElucidating exactly how the mechanical force signal is trans-

mitted to a dormant signal in highly tumorigenic TRCs is ofparamount significance. In the present study, we found that smallG proteinCdc42 in TRCswas changed based on its environmentalstiffness. In 90 Pa 3D fibrin gels, Cdc42 exists in its cytoplasmicform, but is translocated to the nucleus in 1,050 Pa stiff gels. Thistranslocation alters the role of Cdc42 from a mechanotransduc-tionmediator to a nuclear transcription regulator that upregulatesthe expression of Tet2, an epigeneticmodifier gene. Subsequently,5-hmC is generated in cell-cycle inhibitor genes via Tet2-catalyzedhydroxymethylation, leading to the upregulation of p21 and p27followed by cell-cycle arrest and cellular dormancy. Suchmechan-ical signals by matrix stiffness not only induce TRCs into dor-mancy but also maintain this dormant state through a feedback-regulated integrin b3 pathway (Fig. 6I). One of the key molecularevents during this dormancy process is the dissociation of Cdc42from focal adhesions. In another study, we found that TRCs in 90Pa fibrin gels used glycolysis as their metabolic feature, which,however, used fatty acid oxidation (FAO) as energy metabolismTRCs in 1,050 Pa gels. Because focal adhesions are formed in

proximity to cytoplasm membrane, Cdc42 might partition intomembrane rafts, subdomains of the plasma membrane for itsinteraction with focal adhesions. However, this process requirescovalent attachment of lipophilic moieties to Cdc42, such asattachment of farnesyl or geranylgeranyl moieties, so-called pro-tein prenylation (36). Farnesyl and geranylgeranyl moieties arederived from farnesyl pyrophosphate and geranylgeranyl pyro-phosphate, the metabolic intermediates of cholesterol biosyn-thesis. Notably, FAO (catabolism) antagonizes cholesterol bio-synthesis (anabolism), prompting us to hypothesize that B16TRCs in 1,050 Pa fibrin gels use FAO to antagonize cholesterolbiosynthesis, preventing Cdc42 prenylation and Cdc42 disasso-ciation from focal adhesions. This hypothesis is currently understudy. Although the detailed molecular mechanism underlyingCdc42 translocation and binding to target genes is not investi-gated in this study, using ChIP-qPCR technology, we found inanother study that nuclear Cdc42 also bound glucose metabo-lism-associated genes and regulated the expression of those genesexpression. Whether and how Cdc42-regulated glucose metabo-lism is involved in TRC dormancy is currently under study.

Fibrin is derived from soluble plasma protein fibrinogenthrough the catalysis of thrombin. Reports show that fibrin ispresent in the stromaofmalignant tumors (37, 38), thatfibrin andfibrinogen enhance the survival and metastatic potential of cir-culating tumor cells (39, 40), that fibrin–fibronectin complexespromote lung metastasis (41), and that fibrinogen depletiondownregulates pulmonary tumor formation in wild-type mice(42). Therefore, fibrin might be a fundamental ECM componentthat is profoundly involved in tumorigenesis; the 3D fibrin gelsmight not be simply considered as artificial biomaterials; and theimplantation offibrin gels tomice in this study is likely tomimic aphysiologic and/or pathologic process. In normal tissues, type Icollagen (collagen-1) actually is more widely distributed in ECM,compared with fibrin. However, soft 3D collagen-1 gels are not aseffective as fibrin gels in selecting and growing TRCs (12). More-over, TRCs in rigid 3D collagen-1 gel (1,050 Pa) do not enterdormancy but undergo apoptosis (Fig. 5C). Consistently, in thisstudy, 5 B16 TRCs in collagen-1 gels are not capable of forming atumor under the skin of a mouse (Fig. 5B). Therefore, fibrin gelsmight be a very unique natural biomaterial that is capable ofregulating either TRCgrowthor their dormancy, dependent on thegel's softness versus its stiffness. Such uniqueness might beexplained by the intrinsic trait of fibrin gels in transducing forceto cells. Even if the same force is given, different consequences aregenerated in regards to activating different transducing pathways.Fibrin uses integrin avb3 to transduce a weaker force to cells (43)but collagen-1 uses integrin a2b1, a1b1, or other b1 subtypes totransduce a stronger force to cells (44, 45). It is known thatdormant tumor cells can start to regrow by sprouting vessels ina zebrafish model (35). A recent report shows that stiff collagen-Imatrix can promote tumor angiogenesis (46). These findingstogether suggest that stiff collagen-I might negatively regulatetumor cell dormancy. However, it is not clear how it works ina complex matrix where fibrin is also there. Besides matrixstiffness, other factors may also affect tumor cell dormancy. It isrecently found that Her2-positive mammary cancer cells are ableto promote early dissemination and metastasis after dormancy(47). However, whether or how Her2 is influenced by matrixmechanics is not clear at this time.

In summary, our finding that a stiff 3D matrix can initiate andmaintain the dormancy of TRCs via a Cdc42–Tet2 epigenetic

Liu et al.





Figure 6.

TheCdc42–Tet2 pathway is involved in the dormancy of primary humanmelanoma cells.A,Primary tumor cells isolated fromhumanmelanomawere seeded in 90Pasoft 3D fibrin gels for TRC growth. TRCswere then isolated and seeded in 90 or 1,050 Pa 3D fibrin gels and cultured for 5 days. The colony sizeswere quantified on day3 (d3) and day 5 (d5; left). Or, the TRCs from 90 or 1,050 Pa gels were reseeded in 90 Pa fibrin gels and the colony sizes were quantified at differenttime points (right).B,Cell-cycle analysis of primary humanmelanoma TRCs cultured in 90 or 1,050 Pa 3D fibrin gels on day 3.C, Primary humanmelanomaTRCswerecultured in 90 or 1,050 Pa fibrin gels for 3 days and then treated with or without IFNg (100 ng/mL) /TNFa (10 ng/mL). SA-b gal staining was conducted.D, Primary humanmelanoma TRCs grown in 90 or 1,050 Pa 3D fibrin gels were immunostainedwith anti-COUP-TF1 (red), Ki-67 (green) and DAPI (blue), and imagedby confocal microscopy. E and F, the same as D, but TRCs were immunostained with anti-Cdc42 or anti-Tet2 (red) and DAPI (blue), and imaged by confocalmicroscopy (E). Or these TRCs were lysed for Cdc42 and Tet2 detection byWestern blot (F). Bar, 10 mm. G, The protein expression of p27 and p21 in primary humanmelanomaTRCsgrown in 90or 1,050Pa3Dfibrin gels.H,Fiveprimary humanmelanomaTRCswithin 90or 1,050Pa3Dfibrin gelwere implanted intoNOD-SCIDmice.The tumor formation was assessed after 10 days, and the tumor sizes were measured after 25 days. I, Schematic model for TRC dormancy by stiff matrix.Left, TRC growth in 90 Pa soft fibrin matrix. Middle, TRC dormancy formation in 1,050 Pa stiff fibrin matrix. The increased force results in the translocation ofCdc42 into the nucleus, where Tet2 expression is promoted, leading to hydroxymethylation of p21 and p27 genes and TRC dormancy. Right, TRC dormancymaintenance. The induced Tet2, in turn, downregulates integrin b3 in 1,050 Pa stiff fibrin matrix during the first 48 hours, leading to weakened mechanical forcetransduction and decreased F-actin filaments, thus forming a stable state of TRC dormancy. The data represent mean � SEM; n ¼ 3 separate experimentsfor A–G; N.S., no significant difference; ��, P < 0.01, by Student t test (A, B, and D–H) or one-way ANOVA (C).






program would stimulate more research in the area of tumordormancy and dormancy-related tumor recurrence andmetastasis.

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

Authors' ContributionsConception and design: Y. Liu, N. Wang, B. HuangDevelopment of methodology: Y. Liu, J. Lv, X. Yin, R. Fiskesund, S. Mo,F. ChengAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): Y. Liu, J. Lv, X. Liang, L. Zhang, D. Chen, R. Fiskesund,W. Dong, T. Zhang, Y. ZhouAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): Y. Liu, X. Liang, R. Fiskesund, K. TangWriting, review, and/or revision of the manuscript: Y. Liu, R. Fiskesund,N. Wang, B. Huang

Administrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): Y. Liu, J. Lv, X. Jin, J. Ma, H. Zhang, J. XieStudy supervision: Y. Liu, B. Huang

AcknowledgmentsThis work was supported by the National Natural Science Foundation

of China (81788101, 81661128007, 81530080, and 81773062) and theChinese Academy of Medical Sciences Initiative for Innovative Medicine(2016-I2M-1-007).

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 November 30, 2017; revised March 14, 2018; accepted May 11,2018; published first May 15, 2018.

References1. SosaMS, Bragado P, Aguirre-Ghiso JA.Mechanisms of disseminated cancer

cell dormancy: an awakening field. Nat Rev Cancer 2014;14:611–22.2. Goss PE, Chambers AF. Does tumour dormancy offer a therapeutic target?

Nat Rev Cancer 2010;10:871–7.3. Aguirre-Ghiso JA. Models, mechanisms and clinical evidence for cancer

dormancy. Nat Rev Cancer 2007;7:834–46.4. Meng S, Tripathy D, Frenkel EP, Shete S, Naftalis EZ, Huth JF, et al.

Circulating tumor cells in patients with breast cancer dormancy. ClinCancer Res 2004;10:8152–62.

5. Strauss DC, Thomas JM. Transmission of donor melanoma by organtransplantation. Lancet Oncol 2010;11:790–6.

6. Sagalowsky AI,Molberg K. Solitarymetastasis of renal cell carcinoma to thecontralateral adrenal gland 22 years after nephrectomy. Urology 1999;54:162.

7. Discher DE, Janmey P, Wang Y-l. Tissue cells feel and respond to thestiffness of their substrate. Science 2005;310:1139–43.

8. Irianto J, Pfeifer CR, Xia Y, Discher DE. SnapShot: Mechanosensing Matrix.Cell 2016;165:1820–.e1.

9. Seong J, Tajik A, Sun J, Guan J-L, Humphries MJ, Craig SE, et al. Distinctbiophysical mechanisms of focal adhesion kinase mechanoactivation bydifferent extracellular matrix proteins. Proc Natl Acad Sci U S A 2013;110:19372–7.

10. WangN, Tytell JD, IngberDE.Mechanotransduction at a distance:mechan-ically coupling the extracellular matrix with the nucleus. Nat Rev Mol CellBiol 2009;10:75–82.

11. Geiger B, Spatz JP, Bershadsky AD. Environmental sensing through focaladhesions. Nat Rev Mol Cell Biol 2009;10:21–33.

12. Liu J, Tan Y, Zhang H, Zhang Y, Xu P, Chen J, et al. Soft fibrin gelspromote selection and growth of tumourigenic cells. Nat Mater2012;11:734–41.

13. Li Y, Luo S, Ma R, Liu J, Xu P, Zhang H, et al. Upregulation of cytosolicphosphoenolpyruvate carboxykinase is a critical metabolic event in mel-anoma cells that repopulate tumors. Cancer Res 2015;75:1191–6.

14. Ma J, Zhang Y, Tang K, Zhang H, Yin X, Li Y, et al. Reversing drug resistanceof soft tumor-repopulating cells by tumor cell-derived chemotherapeuticmicroparticles. Cell Res 2016;26:713–27.

15. Tan Y, Tajik A, Chen J, Jia Q, Chowdhury F, Wang L, et al. Matrix softnessregulates plasticity of tumour-repopulating cells via H3K9 demethylationand Sox2 expression. Nat Commun 2014;5:4619–31.

16. Liu Y, Liang X, Dong W, Fang Y, Lv J, Zhang T, et al. Tumor repopulatingcells induce PD-1 expression in CD8þ T cells by transferring kynurenineand AhR activation. Cancer Cell 2018;33:480–94.

17. Quintana E, ShackletonM, SabelMS, FullenDR, Johnson TM,Morrison SJ.Efficient tumor formation by single human melanoma cells. Nature2008;456:593–8.

18. Persson AI, Petritsch C, Swartling FJ, Itsara M, Sim FJ, Auvergne R, et al.Non-stem cell origin for oligodendroglioma. Cancer Cell 2010;18:669–82.

19. Dieter SebastianM,Ball Claudia R,HoffmannChristopherM,Nowrouzi A,Herbst F, Zavidij O, et al. Distinct types of tumor-initiating cells form

human colon cancer tumors and metastases. Cell Stem Cell 2011;9:357–65.

20. Discher DE,MooneyDJ, Zandstra PW.Growth factors, matrices, and forcescombine and control stem cells. Science 2009;324:1673–7.

21. Fluegen G, Avivar-Valderas A,Wang Y, PadgenMR,Williams JK, Nobre AR,et al. Phenotypic heterogeneity of disseminated tumour cells is preset byprimary tumour hypoxic microenvironments. Nat Cell Biol 2017;19:120–32.

22. Sosa MS, Parikh F, Maia AG, Estrada Y, Bosch A, Bragado P, et al. NR2F1controls tumor cell dormancy via SOX9 and RARb driven quiescenceprograms. Nat Commun 2015;6:6170-.

23. Ghajar CM.Metastasis prevention by targeting the dormant niche. Nat RevCancer 2015;15:238–47.

24. Liu Y, Liang X, Yin X, Lv J, Tang K, Ma J, et al. Blockade of IDO-kynurenine-AhR metabolic circuitry abrogates IFN-g-induced immunologic dormancyof tumor-repopulating cells. Nat Commun 2017;8:15207.

25. Reymond N, Im JH, Garg R, Vega FM, Borda d'Agua B, Riou P, et al. Cdc42promotes transendothelial migration of cancer cells through b1 integrin.J Cell Biol 2012;199:653–68.

26. Na S, CollinO, Chowdhury F, Tay B,OuyangM,Wang Y, et al. Rapid signaltransduction in living cells is a unique feature of mechanotransduction.Proc Natl Acad Sci U S A 2008;105:6626–31.

27. Rohatgi R, Ma L, Miki H, Lopez M, Kirchhausen T, Takenawa T, et al. TheInteraction between N-WASP and the Arp2/3 complex links Cdc42-depen-dent signals to actin assembly. Cell 1999;97:221–31.

28. Rezaul K,GuptaD, Semenova I, Ikeda K, Kraikivski P, Yu J, et al. EngineeredTug-of-war between kinesin and dynein controls direction of microtubulebased transport in vivo. Traffic 2016;17:475–86.

29. Liu F, Wang L, Perna F, Nimer SD. Beyond transcription factors: howoncogenic signalling reshapes the epigenetic landscape. Nat Rev Cancer2016;16:359–72.

30. Brown R, Curry E, Magnani L, Wilhelm-Benartzi CS, Borley J. Poisedepigenetic states and acquired drug resistance in cancer. Nat Rev Cancer2014;14:747–53.

31. Jones C, Baker SJ. Unique genetic and epigenetic mechanisms drivingsignatures of paediatric diffuse high-grade glioma. Nat Rev Cancer2014;14:10.1038/nrc3811.

32. Shih AH, Abdel-Wahab O, Patel JP, Levine RL. The role of mutations inepigenetic regulators in myeloid malignancies. Nat Rev Cancer2012;12:599–612.

33. K B. Epigenetics: Enhancers under TET control. Nat Rev Mol Cell Biol2014;15:699.

34. Weisel JW. Fibrinogen andfibrin. Advances in protein chemistry: AcademicPress; 2005. p. 247–99.

35. Ju Y-E, Janmey PA, McCormick M, Sawyer ES, Flanagan LA. Enhancedneurite growth from mammalian neurons in three-dimensional salmonfibrin gels. Biomaterials 2007;28:2097–108.

36. Thurnher M, Gruenbacher G. T lymphocyte regulation by mevalonatemetabolism. Sci Signal 2015;8:re4.

Liu et al.





37. Dvorak HF, Senger DR, Dvorak AM. Fibrin as a component of the tumorstroma: origins and biological significance. Cancer Metastasis Rev 1983;2:41–73.

38. Costantini V, Zacharski LR. The role of fibrin in tumor metastasis. CancerMetastasis Rev 1992;11:283–90.

39. Palumbo JS, Degen JL. Fibrinogen and tumor cell metastasis. Haemostasis2001;31:11–5.

40. Palumbo JS, Kombrinck KW, Drew AF, Grimes TS, Kiser JH, Degen JL, et al.Fibrinogen is an important determinant of the metastatic potential ofcirculating tumor cells. Blood 2000;96:3302–9.

41. Malik G, Knowles LM, Dhir R, Xu S, Yang S, Ruoslahti E, et al. Plasmafibronectin promotes lung metastasis by contributions to fibrin clots andtumor cell invasion. Cancer Res 2010;70:4327–34.

42. van Sluis GL, Bruggemann LW, Esmon CT, Kamphuisen PW, Richel DJ,Buller HR, et al. Endogenous activated protein C is essential for immune-

mediated cancer cell elimination from the circulation. Cancer Lett2011;306:106–10.

43. Weisel JW. Fibrinogen and fibrin. Adv Protein Chem 2005;70:247–99.44. Hong H, Stegemann JP. 2D and 3D collagen and fibrin biopolymers

promote specific ECM and integrin gene expression by vascular smoothmuscle cells. J Biomater Sci Polym Ed 2008;19:1279–93.

45. Kapp TG, Rechenmacher F, Sobahi TR, Kessler H. Integrin modulators: apatent review. Expert Opin Ther Pat 2013;23:1273–95.

46. Szot CS, Buchanan CF, Freeman JW, Rylander MN. In vitro angiogenesisinduced by tumor-endothelial cell co-culture in bilayered, collagen Ihydrogel bioengineered tumors. Tissue Eng Part C Methods 2013;19:864–74.

47. Harper KL, Sosa MS, Entenberg D, Hosseini H, Cheung JF, Nobre R, et al.Mechanism of early dissemination and metastasis in Her2(þ) mammarycancer. Nature 2016;540:588–92.






2018;78:3926-3937. Published OnlineFirst May 15, 2018.Cancer Res Yuying Liu, Jiadi Lv, Xiaoyu Liang, et al. via a Cdc42-Driven Tet2 Epigenetic ProgramFibrin Stiffness Mediates Dormancy of Tumor-Repopulating Cells

Updated version

10.1158/0008-5472.CAN-17-3719doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2018/05/15/0008-5472.CAN-17-3719.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/78/14/3926.full#ref-list-1

This article cites 46 articles, 10 of which you can access for free 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/78/14/3926To 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-17-3719

http://cancerres.aacrjournals.org/content/suppl/2018/05/15/0008-5472.CAN-17-3719.DC1

http://cancerres.aacrjournals.org/content/78/14/3926.full#ref-list-1

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

mailto:[email protected]

http://cancerres.aacrjournals.org/content/78/14/3926


Fibrin Stiffness Mediates Dormancy of Tumor-Repopulating ...Tumor Biology and Immunology Fibrin Stiffness Mediates Dormancy of Tumor-Repopulating Cells via a Cdc42-Driven Tet2 Epigenetic

Documents