Capping Protein Regulator and Myosin 1 Linker 3 Is Required for Tumor Metastasis

Published OnlineFirst November 6, 2019; DOI: 10.1158/1541-7786.MCR-19-0722

MOLECULAR CANCER RESEARCH | CANCER GENES AND NETWORKS

Capping Protein Regulator and Myosin 1 Linker 3 Is Required for Tumor Metastasis Huan Wang1, Chao Wang2, Guang Peng2, Doudou Yu1, Xin-Gang Cui3,4, Ying-Hao Sun2, and Xiaojing Ma1,5

ABSTRACT ◥ Metastasis accounts for 90% of deaths caused by solid tumors, but protein, that are involved in actin cytoskeletal organization, which the multitude of mechanisms underlying tumor metastasis remains is required for cell polarization and focal adhesion formation. poorly understood. CARMIL1 and 2 proteins are capping protein Moreover, molecular pathway enrichment analysis reveals that lack (CP) interactants and multidomain regulators of actin-based mobil- of CARMIL3 leads to loss of cell adhesions and low CARMIL3 ity. However, CARMIL30s function has not been explored. Through expression in breast cancer patient specimens is implicated in bioinformatic metadata analysis, we find that high CARMIL3 epithelial–mesenchymal transition. We also find that CARMIL3 expression correlates with poor survival of patients with breast and sustains adherens junction between tumor cells. This is accom- prostate cancer. Functional studies in murine and xenograft tumor plished by CARMIL3 maintaining E-cadherin transcription down- models by targeted diminution of CARMIL3 expression or forced stream of HDACs through inhibiting ZEB2 protein level, also via expression demonstrate that CARMIL3 is vitally important for protecting b-catenin from ubiquitination-mediated degradation tumor metastasis, especially for metastatic colonization. Consistent initiated by the destruction complex. with a predominantly cell-intrinsic mode of action, CARMIL3 is also crucial for tumor cell migration and invasion in vitro. Coim- Implications: This study uncovers CARMIL3 as a novel and critical munoprecipitation coupled with mass spectrometric analyses iden- regulator of metastatic progression of cancers and suggests thera- tifies a group of CARMIL3-interacting proteins, including capping peutic potentials to target CARMIL3.

Introduction primary tumor and is switched on again in MET during distant colonization (11). EMT and MET take place sequentially with Metastasis is the leading cause of cancer-related death (1). morphologic changes conferring on cancer cells the metastatic During metastasis, the phenotypic transformation of tumor cells plasticity (12, 13). Thus, actin cytoskeleton remodeling, cell adhe- is closely related with epithelial-to-mesenchymal transition (EMT) sion,EMT,andMETarefinely orchestrated during tumor metas- and mesenchymal-to-epithelial transition (MET), which facilitates tasis. Despite extensive studies to elucidate the mechanisms, the tumor cells to gain the plasticity to undergo the multiple processes molecular pathways that govern these processes are still not well including local invasion, intravasation, survival in the circulation understood. system and arrest, extravasation, seeding, and colonization (2–6). There are three members in the capping protein regulator and Many of these processes require cell motility to overcome tissue myosin 1 linker (CARMIL) family in mammals. CARMIL proteins barriers, which is driven by actin cytoskeleton remodeling and is are capping protein (CP) interactants and multidomain regulators regulated by cell adhesion (7–9). E-cadherin, a major component of of actin-based mobility (14–18). Several human diseases are adherens junction (AJ), is anchored to the actin cytoskeleton via its reported to associate with CARMILs. A common CARMIL1 (also cytoplasmic tail binding with b-catenin and a-catenin (10), and it is known as LRRC16A) variant is found to be associated with downregulated in EMT during local invasion of cancer cells to leave increased susceptibility to gout (19). Mutations in CARMIL2 (also known as RLTPR) cause abrogated CD28 costimulation of T cells 1 State Key Laboratory of Microbial Metabolism, Sheng Yushou Center of Cell and decreased activation of NF-kB signaling (20, 21). In T cells Biology and Immunology, School of Life Science and Biotechnology, Shanghai fi Jiao Tong University, Shanghai, China. 2Department of Urology, Changhai isolated from patients de cient of CARMIL2, F-actin levels were Hospital, Second Military Medical University, Shanghai, China. 3Department of decreased at the leading edge and the microtubule network was Urinary Surgery, Gongli Hospital, Shanghai, China. 4Department of Urinary disorganized (22). CARMIL3 (also known as LRRC16B) was first Surgery, The Third Affiliated Hospital (Eastern Hepatobiliary Surgery Hospital), reported by Hsu and colleagues, using public information of 5 Second Military Medical University, Shanghai, China. Department of Microbi- expressed sequence tags in a systematic analysis of gene expression ology and Immunology, Weill Cornell Medicine, New York, New York. patterns, as an oncofetal protein with transforming capability and Note: Supplementary data for this article are available at Molecular Cancer reexpression in ovarian and colorectal cancer tissues (23). Our Research Online (http://mcr.aacrjournals.org/). group discovered it independently in a study of phagocytes ingest- Corresponding Authors: Xiaojing Ma, Weill Cornell Medicine, 1300 York Ave- ing apoptotic cells in which LRRC16B was identified, via DNA nue, New York, NY 10021. Phone: 212-746-4404; Fax: 212-746-4423; E-mail: affinity purification coupled with mass spectrometric analysis, to be [email protected]; Xin-Gang Cui, Gongli Hospital, Second Military induced to bind to the IL12p40 promoter region promoting its Medical University, 219 Miaopu Road, Shanghai 200135, China. Phone: 8602- transcription (unpublished data). Recently, CARMIL3 was reported 1588-58730; Fax: 8602-1338-2163; E-mail: [email protected]; and Ying- Hao Sun, [email protected] to play a role in spinogenesis by localizing CP to developing synapses (24), but the involvement of CARMIL3 during cancer Mol Cancer Res 2019;XX:XX–XX progression remains virtually unknown. Here we demonstrate for doi: 10.1158/1541-7786.MCR-19-0722 the first time a critical role of CARMIL3 in tumor metastasis and 2019 American Association for Cancer Research. exploit the underlying molecular mechanism.

AACRJournals.org | OF1

Wang et al.

Materials and Methods subtype of TCGA BRCA cohort was based on TCIA (ref. 25; https:// Bioinformatic analysis of clinical cancer patients tcia.at/home) and the clinical survival data of TCGA was retrieved The CARMIL3 CNV data of The Cancer Genome Atlas (TCGA) was according to the pipeline of Liu and colleagues (26). Survival analysis derived from Genomic Data Commons (GDC, https://gdc-portal.nci. in Fig. 1I was based on Kaplan–Meier plotter database (http://www. nih.gov/legacy-archive/) and the CARMIL3, ZEB2, CDH1 mRNA KMplot.com) and the clinical data used in Fig. 1J and Supplementary (RNA-seq), and b-catenin protein (RPPA) expression data of cancer Table S1B was retrieved from GSE42568 dataset. All the graphs were patients’ specimens was retrieved from Firehose (http://gdac.broad plotted with R version 3.4.4 (https://www.r-project.org/about.html). institute.org/), GEO (https://www.ncbi.nlm.nih.gov/geo/), or cBio- Portal (http://www.cbioportal.org/) accordingly. Kaplan–Meier anal- Cell lines ysis of cancer patients’ percent survival was grouped by median The mouse mammary tumor cell line 4T1 was a gift from Prof. CARMIL3 mRNA expression unless stated. The PAM50 luminal A Xuanming Yang of Shanghai Jiao Tong University (SJTU, Shanghai,

Figure 1. High CARMIL3 expression positively correlates with human breast cancer progression. A–G, Analysis of CARMIL3 mRNA expression in breast cancer samples, in TCGA-BRCA cohort grouped by sample types (A) and tumor stages (B); in patients with TNBC of dataset GSE61723 (C and D); in dataset GSE4922 grouped by breast cancer disease grade (E); in liver metastases of dataset GSE56493 (F); and in TCGA BRCA cohort, normal tissue versus primary tumor of PAM50 luminal A subtype (G). H–J, Kaplan–Meier analysis of breast cancer patient survival, in the cohort of TCGA BRCA PAM50 luminal A subtype (H); in the lymph node positive cohort of patients with breast cancer based on Kaplan–Meier plotter dataset (I); in the cohort of GSE42568 grouped by mean CARMIL3 mRNA expression (J).

OF2 Mol Cancer Res; 2019 MOLECULAR CANCER RESEARCH

CARMIL3 Regulates Tumor Metastasis

China). The 4TO7 cell line was a gift from Prof. Guohong Hu of two times independently, verifying that the targeting Carmil3 Shanghai Institute of Nutrition and Health (Shanghai, China). The genomic locus was correctly edited by one base pair loss in the mouse prostate tumor cell line TRAMP C-1 was purchased from exon, resulting in a frame shift. The identified control and Carmil3 ATCC. HEK293T and MDA-MB-231 cell lines were stored in Ma lab KO clones were used for further experiments. of SJTU (Shanghai, China). These cells were maintained in DMEM (Gibco) supplemented with 10% FBS (Gibco) and 100 mg/mL peni- Mouse tumor models cillin/streptomycin (Gibco). The PC-3 and PC-3M cells were stored in All mouse studies were conducted in accordance with protocols Sun lab of Second Military Medical University (Shanghai, China) and approved by the Institutional Animal Care and Use Committee of cultured with RPMI1640. All the cells were tested Mycoplasma-free by SJTU. For 4T1 tumor model, 2.5 105 cells were injected to the forth PCR and were used for experiments within 8 passages after thawing, mammary fat pad of 8-week BALB/c mice. Tumor growth was which were incubated under an atmosphere of 5% CO2 at 37 C. monitored every 4 days with a caliper and tumor volume was calcu- lated as volume ¼ length width2 0.5. The humane endpoint for Genomic deletion by CRISPR/Cas9 system survival analysis of tumor bearing mice was setup by the professionals CARMIL3-deficient tumor cells were generated by CRISPR/Cas9 in the animal center of SJTU (Shanghai, China). system. In detail, 4T1 cells were transiently transfected by lipo2000 For experimental lung metastasis model, 4.5 105 4T1 cells (Invitrogen) with a pair of single guide RNA (sgRNA), GCTGCC- suspended in 200 mL PBS were intravenously injected to BALB/c mice ATCCGCCGAGAGTG, and CCAGCAACATCGTGTGAAAC, through tail vein. Two weeks later, mice were sacrificed and the lungs which were cloned to lentiCRISPR v2 plasmid (Addgene #52961) were subjected to tumor cell metastasis with 6-thioguanine assay. For respectively. The pair of sgRNA sequences were from Zhang and lung metastasis model of TRAMP C-1 cells, 1.5 105 TRAMP C-1 colleagues (27), resulting in a deletion of the genomic DNA suspended in 100 mL PBS were intravenously injected to C57/B6 mice fragment within 4T1 cells, which was verified by genomic DNA through tail vein. Eighteen days later, the lung tissue was fixed with amplification with a pair of primer (forward: CCCATAACAAAGT- Bouin solution and the metastatic nodules were counted under dis- CACCAA; reverse: CGTTCCAGGAGACCAGAT) flanking the two secting microscope. targeting sites (Fig. 2A), followed by sequencing. The TRAMP C-1 cells were infected by lentivirus containing the control or the RNA-seq and signaling pathway enrichment analysis targeting sgRNA CCAGCAACATCGTGTGAAAC, followed by Total RNA of 4T1 WT and Carmil3 KO cells (three samples in each isolation of single-cell–derived clones and sequencing of the target- group) were prepared with TRIzol (Invitrogen). The RNA sequencing ing genomic locus. The single-cell–derived KO clone was sequenced was performed by Annoroad company based on Illumina platform.

Figure 2. Carmil3 KO in tumor cells reduces tumor progression and prolongs survival. A, Carmil3 KO 4T1 tumor cell line was generated by CRISPR/Cas9 technology with a pair of sgRNA deleting a DNA fragment in the genomic locus of Carmil3. B, DNA gel electrophoresis image of genotyping result, PCR with the pair of primers showed in A. C–E, WT or Carmil3 KO 4T1 tumor cells were injected to the mammary fat pad of BALB/c mice. (WT group n ¼ 8, KO group n ¼ 6). Tumor growth curve (C) and tumor weight at day 28 after inoculation (D) were shown. Percent survival was also determined (E). The data was plotted as mean SEM. F, Result of qRT-PCR determining mRNA level of CARMIL3 in normal and cancerous human prostate cell lines. G and H, Control or shCARMIL3 PC-3M cells were injected into the bone marrow space of tibial metaphysis in NOD-SCID mice (n ¼ 5 per group). The tibial bones were harvested for histologic and IHC analysis (G) and the metastatic tumor growth was monitored (H). Statistical method: C, D, F,andH, Student t test; E, log-rank test. , P < 0.05; , P < 0.01; , P < 0.001. See also Supplementary Fig. S2.

AACRJournals.org Mol Cancer Res; 2019 OF3

Wang et al.

The signaling pathway enrichment analysis was performed through normal counterparts (Fig. 1A), higher in stages 1 to 3 than normal Gene Set Enrichment Analysis (GSEA) and Database for Annotation, tissues (Fig. 1B), higher in triple-negative breast cancer (TNBC) Visualization and Integrated Discovery (DAVID). The raw data of the comparing with normal tissues (Supplementary Fig. S1C). Analysis RNA-seq is available in GEO (GSE122378). of a publicly available dataset (GSE61723) of TNBC also revealed that CARMIL3 mRNA expression was significantly higher in lymph node Western blot analysis metastatic tumor samples compared with normal and primary tumor The cultured cells were washed with cold PBS, then scraped down samples (Fig. 1C and D). In addition, we also observed that CARMIL3 and lysed with RIPA buffer (#P0013B, Beyotime) plus protease expression was significantly increased in grade 3 tumor tissues and phosphatase inhibitor cocktail (Roche) for 30 minutes on ice, (Fig. 1E) and was upregulated in liver metastases compared with followed by centrifugation at 12,000 rpm for 15 minutes at 4C. Then, primary tumors (Fig. 1F). An analysis of the pathologic tissue data in the supernatant was harvested, of which the concentration was the Human Protein Atlas (HPA) database via IHC staining of CAR- determined with BCA kit (#23227, Thermo Fisher Scientific). Totally, MIL3 expression in breast cancer patient sample (ID: 2565) shows that 30 mg protein was loaded for every sample. The antibodies used are there was CARMIL3 present in the cortical membrane, cytoplasm, as listed below: E-cadherin (#610181), Paxillin (#610051), and b-catenin well as the nucleus of breast cancer cells (Supplementary Fig. S1D). (#610153), CD44 (#553131) were from BD Biosciences; SNAIL Next, we analyzed the association of breast cancer patients’ survival (#ab53519) was from Abcam; ZEB2 (#14026-1-AP), His-Tag antibody with the expression of CARMIL3. The data show that CARMIL3 was (#66005-1-Ig), GAPDH (#60004-1-Ig), and TWIST1 (#25465-1-AP) upregulated in PAM50 luminal A subtype tumor compared with were from ProteinTech; FLAG antibody (#F1804) was from Sigma- normal samples (Fig. 1G) and high expression of CARMIL3 was Aldrich; HA-tag antibody (#RLM3003) was from Ruiying Biological; positively correlated with poor survival of patients with breast cancer Capping protein subunit b (CPb, #AB6017) was from Millipore; in various cohorts (Fig. 1H–J). Taken together, these observations a-Tubulin (#sc8035) from Santa Cruz Biotechnology; Axin1 suggest that high CARMIL3 expression is involved in breast cancer (#AF3287) was from R&D Systems; phospho-b-catenin (Ser33/37/ metastasis and correlates with poor survival of patients with breast Thr41, #9561) was from Cell Signaling Technology. Secondary anti- cancer. Moreover, we performed multivariate Cox regression analysis bodies against mouse, rabbit, and goat were from LI-COR Odyssey. with breast cancer patient datasets (TCGA-BRCA PAM50 luminal A dataset, N ¼ 303; GSE42568 dataset, N ¼ 104). The result showed that Plasmids and reagents CARMIL3 expression represented an independent predictor for breast pTriEx-hCARMIL3-(His tagged), pCDH-mCARMIL3 (His tagged, cancer patient survival (Supplementary Table S1A and S1B). Collec- N terminal), pCDH-hCARMIL3 (His tagged, N terminal), MSCV- tively, these results suggest that CARMIL3 can potentially serve as a FLAG-mCARMIL3-Puro, pEGFP-C2-mCARMIL3 (GFP tagged), biomarker of breast cancer disease progression and prognosis. Similar pCDNA3.1-mIGF2BP1 (His tagged), and pCMV-3xFlag-hb-catenin associations of CARMIL3 with prostate cancer were also observed were constructed in this study. Flag-tagged Axin1 plasmid was a gift (Supplementary Fig. S1E–S1J). from Prof. Lin Li of Shanghai Institutes for Biological Sciences (Shanghai, China). pCMV-HA-Ub was purchased from the Miaoling- Loss of Carmil3 impairs tumor growth and prolongs animal bio Company. survival in vivo FLAG-tagged PSMD3 and H2AFY plasmids were purchased from To study CARMIL3 functionally, we generated Carmil3 knocking OriGene. Protein A/G PLUS-Agrose (#sc2003) from Santa Cruz out (KO) 4T1 murine mammary tumor cell line via CRISPR/Cas9 Biotechnology; entinostat (MS-275, #S1053), IWR-1 (#S7086), and (Fig. 2A). The KO efficiency of single-cell–derived clone was verified at SB216763 (#S1075) were from Selleckchem; MG132 (#C2211), Tri- the genomic DNA level by PCR (Fig. 2B) and sequencing (Supple- chostatin A (#T8552), and 5-Aza-20-deoxycytidine (#A3656) were mentary Fig. S2A), as well as at the mRNA level by quantitative RT- from Sigma-Aldrich. PCR (Supplementary Fig. S2B). Subsequently, wild-type (WT) or Carmil3 KO 4T1 cells were inoculated to the mammary fat pad of Statistical analysis syngeneic BALB/c mice, tumor volume (Fig. 2C) and weight on day Unpaired Student t test was used for performing statistical analysis 28 (Fig. 2D) were monitored. CARMIL3 deficiency in the tumor with GraphPad Prism5 or R (version 3.4.4) unless stated. For survival moderately reduced tumor growth, with a significant extension in analysis of patients with cancer as well as tumor-bearing mice, log-rank survival (Fig. 2E). The proliferative ability of Carmil3 KO cells, test was used. Differences were considered to be significant when P < measured by MTS (Supplementary Fig. S2C) and BrdU incorporation 0.05. assays (Supplementary Fig. S2D), was deficient compared with WT More information of the Materials and Methods is available in the cells. Moreover, cell-cycle analysis showed that there were less cells Supplementary Data. distributed in the S-phase (Supplementary Fig. S2E). To determine whether CARMIL3 deficiency affects the tumor microenvironment (TME), we examined immune cell infiltration into Results 4T1 tumors and tumor draining lymph nodes by flow cytometry. The þ High CARMIL3 expression correlates with poor outcome of result shows that NK1.1 cell infiltration of TME was significantly patients with cancer increased (Supplementary Fig. S2F), indicating that the loss of CAR- We analyzed the genomic alteration frequency of copy number MIL3 may induce innate immune responses, which could contribute to variance (CNV) of CARMIL3 in TCGA and found that gain of the reduced growth of Carmil3 KO tumors in vivo. CARMIL3 widely occurred in different types of cancer (Supplemen- In addition, we used a mouse xenograft model to explore the tary Fig. S1A). Comparing mRNA levels in normal versus cancerous functional role of CARMIL3 in prostate cancer. First, we determined tissues showed that CARMIL3 expression was upregulated across that CARMIL3 mRNA levels in prostate cancer cell lines were higher different types of cancerous tissues (Supplementary Fig. S1B). CAR- than those in normal prostate cell lines WPMY-1 and BPH-1 (Fig. 2F). MIL3 mRNA level was higher in primary breast carcinoma than their Moreover, higher expression of CARMIL3 was observed in C4-2B and

OF4 Mol Cancer Res; 2019 MOLECULAR CANCER RESEARCH

CARMIL3 Regulates Tumor Metastasis

PC-3M (prostate cancer cell lines with bone metastasis) than those in invasive abilities in vitro, whereas CARMIL3 overexpression can their nonmetastatic counterparts, C4-2 and PC-3. We then silenced enhance tumor cell–migratory capacity. CARMIL3 expression in PC-3M cells using shRNA, which resulted in Carmil3 KO tumor-bearing mice exhibited strongly reduced num- lower tumor incidences and slower tumor growth in NOD-SCID mice ber of spontaneously metastatic tumor cells in the lungs (Supplemen- (Fig. 2G and H). Thus, CARMIL3 plays a crucial role in the growth of tary Fig. S3C; Fig. 3E). Hematoxylin and eosin (HE) staining of lung human prostate cancer cells. tissue sections of tumor-bearing Carmil3 KO mice also showed much fewer metastatic nodules (Fig. 3F). These observations suggest that CARMIL3 deficiency reduces tumor metastasis CARMIL3 is required for spontaneous tumor metastasis. The effect of CARMIL3 deficiency on tumor cell proliferation was We then sought to determine the impact of CARMIL3 on tumor moderate, whereas its impact on animal survival was more prominent cell colonization in the lung by directly injecting intravenous 4T1 (extended by 20 days). We thus further explored CARMIL30s role in andTRAMPC-1tumorcellstomiceandcomparedtheirlung cell mobility, migration, and invasion. The “scratch assay” showed that colonization efficiencies two weeks later. We observed that Carmil3 the mobility of Carmil3 KO cells was reduced compared with WT cells KO 4T1 tumor cells hardly formed nodules in the lung, compared (Supplementary Fig. S3A). By the transwell assay, Carmil3 KO cells with WT tumor cells (Fig. 3G), and the number of colonizing tumor exhibited impaired migration (Supplementary Fig. S3B) and reduced cells was significantly reduced (Fig. 3H). Similarly, CARMIL3- invasive ability (Fig. 3A). Overexpression of mCARMIL3 in 4T1 by the deficent TRAMP C-1 also reduced tumor cell colonization to the Tet-On inducible system showed that mCARMIL3 significantly lung (Fig. 3I and J). Next, we extended the observation to human increased the cells’ migratory ability (Fig. 3B). Consistent with this prostate cancer cells PC-3M. Likewise, in the experimental bone observation, forced expression of mCARMIL3 in the poorly metastatic metastasis model of prostate cancer via injecting PC-3M cells into 4TO7 cells also promoted migration (Fig. 3C). Furthermore, over- the left cardiac ventricle of NOD/SCID mice, CARMIL3-silenced expression of CARMIL3 in the human breast cancer cell line MDA- tumor cells presented strongly reduced bone metastasis (Fig. 3K MB-231 enhanced the migratory capacity of the cancer cells (Fig. 3D). and L). Taken together, these results demonstrate that CARMIL3 is Taken together, these observations demonstrate that CARMIL3 defi- necessary for efficient lung colonization during tumor metastasis of ciency strongly reduces tumor cell mobility and their migratory and mammary and prostate tumor cells.

Figure 3. CARMIL3 deficiency reduces tumor metastasis. A, Transwell invasion assay of WT and Carmil3 KO 4T1 cells and the cell numbers were quantified. B–D, Results of transwell migration assay with cancer cells overexpressing CARMIL3, 4T1 (B), 4TO7 (C), and MDA-MB-231 (D). E and F, 4T1 tumor cells were injected to the mammary fat pad of BALB/c mice (WT n ¼ 8, KO n ¼ 6). The lung metastasis of tumor cells was determined by 6-thioguanine assay (E) and HE staining (F). G and H, 4T1 cells were intravenously injected to BALB/c mice (n ¼ 5) through tail vein. Representative images of the whole lung tissues were shown (G) and the tumor cells colonized to the lung were quantitatively analyzed by 6-thioguanine assay (H). I and J, TRAMP C-1 cells were intravenously injected to C57/B6 mice through tail vein (n ¼ 4or5). Representative images of lung tissues were shown (I) and the metastatic nodules (indicated by arrow) were quantified (J). K and L, Control or shCARMIL3 PC-3M cells were injected into the left cardiac ventricle of NOD-SCID mice (n ¼ 5). The bone metastasis formation (K) was monitored and the metastatic lesions were harvested for histologic and IHC analysis (L). The data was plotted as mean SEM. Statistical method: Student t test. , P < 0.05; , P < 0.001. See also Supplementary Fig. S3.

AACRJournals.org Mol Cancer Res; 2019 OF5

Wang et al.

The interacting proteins of CARMIL3 IGF2BP1 (also known as CRD-BP), one of the RNA-binding CARMIL3 shares some of the conserved domains with CARMIL proteins (28), was verified to interact with CARMIL3 (Fig. 4D). family proteins (14). Thus, we hypothesized that CARMIL3 might Similarly, in the case of core histone macro-H2A.1, which is encoded have interacting protein partners that are instrumental for its functions by the H2AFY gene, was also confirmed to interact with CARMIL3 in tumor metastasis. FLAG-tagged CARMIL3 was overexpressed in (Fig. 4E), and so was the protein of 26S proteasome non-ATPase HEK293T cells, followed by immunoprecipitation (IP) with FLAG M2 regulatory subunit 3 (encoded by PSMD3; Fig. 4F). These results affinity gel (Fig. 4A; Supplementary Fig. S4A). Mass spectrometry indicate that CARMIL3 is involved in multiple protein–protein inter- (MS) analysis identified many potential CARMIL3-interacting pro- actions, which may have multiple molecular functions in tumor cells. teins (Supplementary Table S2) including actin cytoskeleton proteins, These observations prompted us to determine the localization of such as capping proteins (CP). Protein network mapping and gene CARMIL3 protein in mammalian cells. CARMIL3 was engineered to ontology analysis showed that these proteins are closely related with fuse with enhanced GFP (eGFP), followed by transfecting into 4T1 and actin cytoskeleton (Fig. 4B; Supplementary Fig. S4B). The interaction HEK293T cells. We then tracked the location of EGFP-CARMIL3 between CPb and CARMIL3 was further confirmed by IP in HEK293T within the cells. We observed that the fusion protein not only in the and TRAMP C-1 cells (Fig. 4C). It is not surprising that CARMIL3 was cortical plasma membrane and the cytoplasm, but also in the nucleus found to bind CP protein because CARMIL3 shares similar CP- (Supplementary Fig. S4C and S4D; Fig. 4G), which was consistent with binding motifs with CARMIL family proteins (14). the very first observation of its kind (23). This result indicates that there

Figure 4. CARMIL3 is required for actin cytoskeleton remodeling. A, HEK293T cells were transfected with MSCV-FLAG or MSCV-FLAG-CARMIL3, followed by IP with FLAG antibody. Images of SDS-PAGE gel after silver staining was shown. B, The protein interacting network of CARMIL3 in mass spectrometry analysis. C, Western blot result of IP following forced expression of MSCV-FLAG or MSCV-FLAG-CARMIL3 in HEK293T and TRAMP C-1 cells, CPb: capping proteins subunit beta. D–F, Co-IP result of Western blot in HEK293T cells, transfected with indicated plasmids. G, Confocal images showing the localization of GFP-tagged CARMIL3 in HEK293T and 4T1 cells. H, Phalloidin staining of 4T1 and TRAMP C-1 cells. F-actin level indicated the mean size of stress fibers. I, Representative microscopic images of 4T1 cells and TRAMP C-1 cells under bright field. J, Representative immunofluorescent images of 4T1 cells. K, Quantification of 4T1 cells with different morphology. Top, DIC image of representative cells. L, The spreading areas of 4T1 and TRAMP C-1 cells. Scale bar, 50 mm. M, Representative immunofluorescent images of 4T1 cells stained with Paxillin antibody; right, the quantification of the focal adhesions. The data was plotted as mean SEM. Statistical method: Student t test. , P < 0.001. See also Supplementary Fig. S4 and Supplementary Table S2.

OF6 Mol Cancer Res; 2019 MOLECULAR CANCER RESEARCH

CARMIL3 Regulates Tumor Metastasis

are multiple localizations of CARMIL3 protein in mammalian cells, encodes E-cadherin, was among the core gene set in the cell which explains its numerous interacting proteins and further suggests adhesion pathway and its expression is strongly diminished by that CARMIL3 is a versatile molecule that likely exerts multiple CARMIL3 deficiency (Fig. 5C). functions in mammalian cells. E-cadherin is mostly expressed by epithelial cells and is a crucial component of AJ, which is anchored to the actin cytoskeleton via its Defective actin cytoskeletal organization in CARMIL3-deficient cytoplasmic domain in conjunction with a and b-catenins. We cells sought to determine CARMIL3's impact on AJ by comparing the Because we found that CARMIL3 was able to bind CP, which cell morphology between WT and Carmil3 KO cells cultured in high increases the rate of actin-based motility by promoting filament densities. CARMIL3-deficient 4T1 cells displayed reduced cell–cell nucleation via the Arp2/3 complex (29), we hypothesized that CAR- adhesion (Fig. 5F). By Western blot, we observed that the protein MIL3 might regulate actin filaments dynamics during tumor cell level of core AJ molecules E-cadherin and b-catenin was reduced in movement. Phalloidin staining of actin filaments of 4T1 and TRAMP Carmil3 KO cells (Fig. 5G), which could be rescued by the recon- C-1 tumor cells showed that the cell morphology and actin organi- stitution of the expression of CARMIL3 via the inducible Tet-On zation changed dramatically in Carmil3 KO cells (Fig. 4H). Prominent 3G system in Carmil3 KO 4T1 cells (Supplementary Fig. S5C). stress fibers were significantly reduced. We then compared the mor- Immunofluorescence assay corroboratedwiththeWesternblotdata phology of WT and Carmil3 KO 4T1 cells as well as TRAMPC C-1 showing the two strongly reduced AJ molecules in CARMIL3- cells. We observed that membrane protrusions of Carmil3 KO cells deficient 4T1 and TRAMP-C1 cells (Fig. 5H). Furthermore, over- both were reduced (Fig. 4I). Given that CP binds the barbed end of expression of mCARMIL3 in 4TO7 cells increased both E-cadherin actin filaments and is important for the assembly of cortical actin and and b-catenin (Fig. 5I). These observations suggest that CARMIL3 for cases of actin-based motility, such as the formation of membrane is required for AJ formation. protrusions at the leading edge of migrating cells (14), we measured CP A more extended analysis of the effect of CARMIL3 on cell adhesion distribution and observed that Carmil3 KO 4T1 cells displayed less CP molecules revealed that the mRNA expression of Esrp1 and Cd44v, distributions in the leading edge (Fig. 4J), which explains the reduced two markers of EMT as well as important regulators of metastatic lung membrane protrusions in CARMIL3-deficient cells. colonization (31), was totally lost in Carmil3 KO cells (Fig. 5J). The Given that migrating and invading tumor cells require actin cyto- expression of the cell adhesion molecule, CD44, was also reduced in skeletal reorganization, which is concomitant with cell polarization, we Carmil3 KO cells (Fig. 5K). Taken together, these results suggest that further investigated the effects of CARMIL3 on polarization by CARMIL3 defi ciency leads to loss of cell adhesion, which explains, at measuring the shape and spreading area of tumor cells. The various least partially, why lack of CARMIL3 within tumor cells reduces their morphologies of 4T1 tumor cells were quantified, which revealed that metastasis and organ colonization. CARMIL3 deficiency attenuated polarization (Fig. 4K). The spreading We performed GSEA analysis with TCGA-breast cancer specimens area was also determined and the result showed that CARMIL3 based on its gene expression profile of mRNA (RNA-seq). Totally, depletion decreased tumor cell spreading-out (Fig. 4L). Collectively, 1,100 samples from 1,093 patients with breast cancer were analyzed, of the observations demonstrate that CARMIL3 is required for tumor cell which there were 67 patient samples whose mRNA expression was polarization. defined as high expression with the Z-score >1. The GSEA revealed We then tested whether loss of CARMIL3 had any effect on focal that low CARMIL3 expression was a prominent feature in EMT adhesions (FA), which are closely related to cancer metastasis with (Fig. 5L). Part of the mRNA and protein expression (RPPA) profiles complex plasma membrane-associated macromolecular assemblies of the core gene set was shown in Fig. 5M, of which quantitative that serve to physically connect the actin cytoskeleton to integrins analysis exhibited that high CARMIL3 expression correlated with high engaging with the surrounding extracellular matrix (30). Measuring expression of epithelial cell marker genes such as CDH1 and CLDN7, Paxillin by immunofluorescence revealed that loss of CARMIL3 while low CARMIL3 expression correlated with high expression of significantly decreased the number of FAs formed in 4T1 tumor cells mesenchymal cell marker genes like SNAI2, ZEB1, ZEB2, VIM, MMP2, (Fig. 4M). and TGFB1 (Fig. 5N; Supplementary Fig. S5D). These results indicate that low CARMIL3 expression in cancer cells shifts transcriptional Altered transcriptional profile and loss of cell adhesion in profiles from epithelial cells to mesenchymal cells. CARMIL3-deficient tumor To more broadly understand the molecular function of CAR- CARMIL30s role in regulating Cdh1 transcription downstream of MIL3 in tumor cells, we compared the transcriptome between WT HDACs and Carmil3 KO 4T1 cells by RNA-seq analysis. Totally, more than Given the key role E-cadherin plays in tumor metastasis in the one thousand genes were differentially expressed (Fig. 5A). These context of EMT, we wished to further understand the molecular include downregulated genes like Gpc4, Cdh1, Esrp1, Lamc2,aswell mechanism by which CARMIL3 regulates the expression of Cdh1. as upregulated genes like Lcn2, Cxcl5, Nos2,andSaa3 (Fig. 5B). The We first examined the mRNA expression of the common EMT most differentially expressed genes were further verified by quan- markers and the main transcriptional repressors of Cdh1 including titative real-time PCR (Fig. 5C). KEGG pathway enrichment anal- Vimentin, Cdh2, Twist1, Snail1, Zeb2, and Tcf3. None of them was ysis via GSEA revealed that CARMIL3 was positively correlated altered by CARMIL3 deficiency (Fig. 6A; Supplementary Fig. S6A). It with signaling of cell adhesion, focal adhesion, and actin cytoskel- has been reported that Cdh1 transcription is epigenetically regulated eton (Fig. 5D; Supplementary Fig. S5A). Biological process analysis through histone deacetylation and DNA methylation (32). To test this through DAVID with genes whose expression were altered over possibility in the regulation of Cdh1 transcription by CARMIL3, 4-fold exhibited that CARMIL3 was mostly involved in regulating trichostatin A (TSA), a histone deacetylase (HDAC) inhibitor, and cell adhesion (Supplementary Fig. S5B). The cell adhesion molecule 5-aza-20-deoxycytidine (50-AZA-dC), a DNA methylation inhibitor, (CAM) enrichment plot and the expression profiles of the core gene were used to treat WT and Carmil3 KO 4T1 cells. Treatment with TSA set in this pathway are shown in Fig. 5E.Ofnote,Cdh1,which but not with 50-AZA-dC partially restored Cdh1 mRNA expression

AACRJournals.org Mol Cancer Res; 2019 OF7

Wang et al.

Figure 5. CARMIL3 deficiency leads to loss of cell adhesion. A, Heatmap of differentially expressed genes (DEG) between WT and Carmil3 KO 4T1 cells determined by RNA-seq. B, Volcano map of DEG. Top list genes were labeled. C, qRT-PCR result of WT and Carmil3 KO 4T1 cells to verify the result of RNA-seq. D and E, KEGG pathway enrichment of CARMIL3 with gene expression profiles of WT and Carmil3 KO 4T1 cells via GSEA, the enrichment plot of cell adhesion molecules (CAM) was shown (D) and the expression profile of core gene sets enriched in the CAMs pathway was shown (E). FC, fold change; K, KO; W, WT. F, Representative images of cultured high- density 4T1 WT and Carmil3 KO cells. G, Western blot result of WT and Carmil3 KO cells with indicated antibodies. H, Immunofluorescent images of WT and Carmil3 KO cells with indicated antibodies. I, Western blot result of control (Ctr) and CARMIL3-overexpressing (OE) 4TO7 cells. J, qRT-PCR result of WT and Carmil3 KO 4T1 cells. K, Immunofluorescent images of WT and Carmil3 KO 4T1 cells stained with CD44 antibody. L, GSEA pathway analysis with the whole transcriptomic gene expression profile (RNA-seq) of TCGA-BRCA primary tumor samples n ¼ 1,093, CARMIL3 high group was defined by Z-score >1. M, Heatmap of EMT-related genes enriched in L; data from cBioPortal. N, Protein expression (RPPA) of adhesion molecules, CDH1 and CLDN7, in TCGA-BRCA primary tumor samples grouped by CARMIL3 mRNA expression. Student t test was used for statistical analysis. See also Supplementary Fig. S5.

(Fig. 6B), which was also corroborated at the level of protein expres- upregulated in Carmil3 KO cells. The results of TWIST1 and Snail1 sion (Fig. 6C and D). The restoration of Cdh1 mRNA level by TSA was varied in the three experiments, while the result of ZEB2 was consistent dose dependent (Supplementary Fig. S6B). Collectively, these observa- throughout. Thus, we focused on ZEB2. To determine whether the tions indicate that CARMIL3 regulates Cdh1 transcription in an downregulation of Cdh1 in Carmil3 KO cells was due to the increased HDAC-dependent manner. ZEB2 protein, we used short hairpin RNA (shRNA) to knockdown Because Cdh1 is a target gene of CARMIL3, we then tested whether (KD) Zeb2 mRNA expression (Fig. 6H), which resulted in resto- CARMIL3 was regulated by HDACs. Analyzing GEO dataset GSE6770 ration of Cdh1 transcription in Carmil3 KO cells (Fig. 6I–K), revealed that Hdac2 KO in mouse embryonic heart tissue increased the so did the epithelial cell morphology (Fig. 6L). Next, we compared mRNA expression of Carmil3 (Fig. 6E). In accordance with this E-cadherin and ZEB2 protein expression between parental and the observation, TSA treatment of 4T1 cells also upregulated Carmil3 lung metastatic 4T1 WT and Carmil3 KO cells (Fig. 6M). The result expression signiﬁcantly (Fig. 6F). Taken together, these results suggest showed that E-cadherin protein level was enhanced in lung met- that CARMIL3 expression is regulated by HDAC. astatic cells and ZEB2 expression was higher in Carmil3 KO cells than WT cells, both in parental and lung metastatic cells, suggesting Regulation of Cdh1 transcription by CARMIL3 involving ZEB2 that deﬁciency of CARMIL3 resulted in upregulation of ZEB2 and We then analyzed in detail the protein expression of several major downregulation of E-cadherin, thereby reducing the metastatic lung Cdh1 transcriptional repressors, including TWIST1, SNAI1 and ZEB2, colonization of tumor cells. This observation is consistent with the by Western blot analysis and one represent result from three inde- notion that tumor cells regain epithelial properties to colonize pendent experiments is shown (Fig. 6G), showing that ZEB2 was distant sites during metastasis.

OF8 Mol Cancer Res; 2019 MOLECULAR CANCER RESEARCH

CARMIL3 Regulates Tumor Metastasis

Figure 6. CARMIL3 is involved in regulating E-cadherin transcription downstream of HDACs. A, qRT-PCR result of EMT related gene expression in WT and Carmil3 KO 4T1 cells. B–D, 4T1 WT and Carmil3 KO cells were treated with 150 nmol/L trichostatin A (TSA) for 24 hours or 2 mmol/L 5-aza-20-deoxycytidine (50-AZA-dC) for 48 hours, followed by qRT-PCR (B), Western blot (C), quantification of C by ImageJ (D). E, Expression of Carmil3 in WT and Hdac2 KO mouse tissues of E17.5 heart ventricles, 1436010_at (ID_REF) profile from GEO GSE6770. F, qRT-PCR determining Carmil3 mRNA level with 4T1 cells treated with DMSO or TSA. G, Western blot result of WT and Carmil3 KO 4T1 cells with indicated antibodies. H, qRT-PCR determining the shRNA-mediated knocking down (KD) efficiency against Zeb2. Scr (Scramble), Shzeb2 (shRNA targeting Zeb2). I–K, Measurement of E-cadherin expression after Zeb2 KD in WT and Carmil3 KO 4T1 cells by qRT-PCR (I) and Western blot (J). K, Quantification of J by ImageJ. L, Representative images showing the morphology of WT and Carmil3 KO 4T1 cells after Zeb2 KD. M, WT or Carmil3 KO 4T1 cells were intravenously injected to BALB/c mice (n ¼ 3). Two weeks later, the lung tissues were harvested for collagenase D digestion and the lung metastatic cells were isolated by 6-thioguanine selection, followed by Western blot analysis. N, Correlation analysis of mRNA expression with primary tumor samples of TCGA BRCA dataset, n ¼ 1,093, method: spearman. O, Kaplan–Meier analysis of BC patient survival of TCGA-BRCA PAM50 luminal A subtype, log-rank test. P, Schematic model of CARMIL3 in regulating Cdh1 transcription. Statistical method: Student t test; , P < 0.001; , P < 0.05. See also Supplementary Fig. S6.

Correlation analysis of CARMIL3 with ZEB2 and E-cadherin using addition, Kaplan–Meier analysis of patients with breast cancer also primary tumor samples from TCGA-breast cancer cohort showed that showed that CARMIL3-high and ZEB2-low expression positively CARMIL3 expression negatively correlated with ZEB2 expression correlated with poor survival outcome, while CARMIL3-low and (Fig. 6N), whereas positively associated with CDH1 expression both CDH1-low expression associated with better survival outcome on mRNA (Supplementary Fig. S6C) and protein (Fig. 5N) level. In (Fig. 6O). We also tried to understand how CARMIL3 regulated

AACRJournals.org Mol Cancer Res; 2019 OF9

Wang et al.

ZEB2 by examining whether CARMIL3 had any impact on ZEB2 4T1 cells were treated with MG132, an inhibitor of 26S proteasome, protein degradation via proteasome (Supplementary Fig. S6D). The followed by Western blot analysis. The result showed that b-catenin observation indicated that CARMIL3 could not regulate ZEB2 protein protein expression was generally enhanced following MG132 treat- degradation through ubiquitylation. Taken together, these observa- ment in both WT and Carmil3 KO cells, and there was no longer any tions suggest that CARMIL3 promotes Cdh1 transcription via inhibit- difference in b-catenin protein expression between these cells ing ZEB2 (Fig. 6P), which also associates with cancer disease (Fig. 7A). In contrast, E-cadherin protein expression was not altered progression. by MG132 treatment. These data demonstrate that b-catenin is regulated by CARMIL3 at the level of protein through the 26S CARMIL3 protects b-catenin from ubiquitin-mediated proteasome degradation system (UPS). His-tagged CARMIL3 and proteasome degradation FLAG-tagged b-catenin were cotransfected into HEK293T cells fol- Because our data indicate that CARMIL3 is required for both E- lowed by IP. The data showed that b-catenin-FLAG coimmunopre- cadherin and b-catenin expression (Fig. 5G), we were interested in cipitated with CARMIL3-His (Fig. 7B). This b-catenin–CARMIL3 understanding how b-catenin is regulated by CARMIL3. First, the interaction was corroborated by reciprocal IP (Supplementary mRNA level of Ctnnb1 (encoding b-catenin) was compared by qRT- Fig. S7B). PCR and there was no statistical difference between WT and Carmil3 Next, FLAG-tagged b-catenin and HA tagged ubiquitin with or KO 4T1 cells (Supplementary Fig. S7A), indicating that CARMIL3 without His-tagged CARMIL3 were expressed in HEK293T cells, does not regulate Ctnnb1 transcription. followed by IP with anti-FLAG antibody and IB with anti-HA anti- The cytosolic b-catenin can be degraded by the ubiquitin mediated body. The result showed that in the presence of CARMIL3, b-catenin proteasome system (33). We reasoned that CARMIL3 might protect ubiquitylation was reduced (Fig. 7C). Similarly, in the same experi- b-catenin from degradation by the proteasome. WT and Carmil3 KO mental setup, reciprocal IP with anti-HA antibody and IB with anti-

Figure 7. CARMIL3 protects b-catenin from polyubiquitinoylation and degradation. A, 4T1 WT and Carmil3 KO cells were treated with DMSO or 10 mmol/L MG132, followed by Western blot with indicated antibodies; W, WT; K, KO. B, HEK293T cells were transfected with CARMIL3-His or b-catenin-FLAG-expressing plasmid. After 48 hours, the cells were harvested for protein extraction, followed by immunoprecipitation (IP) with His antibody and Western blot analysis. C, Various combination of plasmids expressing CARMIL3-His, b-catenin-FLAG, and ubiquitin-HA were cotransfected into HEK293T cells. The total amount of plasmid used for transfection in every cell sample was equalized with control empty vector, respectively. Forty-eight hours later, the cells were harvested for protein extraction, followed by IP with FLAG M2 antibody. Western blot result was shown. Asterisk indicates heavy chain of the anti-HA antibody. D, Reciprocal IP with anti-HA antibody using samples from C, followed by Western blot with indicated antibodies. E, One microgram b-catenin-FLAG expression plasmid with 5 mg control (TriEx) or CARMIL3-His expression plasmid were cotransfected into HEK293T cells cultured in 10-cm dish. Twenty-four hours later, the cells were subcultured, then maintained for another 12 hours, followed by treatment with cycloheximide (CHX). At different time point post cycloheximide treatment, the cells were harvested for Western blot analysis. F, Quantiﬁcation of the protein levels in E with Image Studio software, , P < 0.05, Student t test. G and H, Co-IP result of WB in HEK293T cells, transfected with indicated plasmids. I, Western blot result with anti-Axin1 antibody using the same samples in C. J, Western blot result of WT and Carmil3 KO 4T1 cells with indicated antibodies. K, WT and Carmil3 KO 4T1 cells were treated with DMSO, 25 mmol/L IWR-1-endo or 10 mmol/L SB216763 for 24 hours, followed by Western blot. L, Kaplan– Meier analysis of breast cancer patient survival of TCGA-BRCA PAM50 luminal A subtype, Log-rank test. M, The schematic model of the molecular function of CARMIL3 in regulating b-catenin. See also Supplementary Figs. S7 and S8.

OF10 Mol Cancer Res; 2019 MOLECULAR CANCER RESEARCH

CARMIL3 Regulates Tumor Metastasis

FLAG antibody also showed that CARMIL3 decreased the ubiquityla- cytoskeletal organization, which is required for cell polarization and tion of b-catenin (Fig. 7D). These results demonstrate that CARMIL3 FA formation of migrating tumor cells. Moreover, global transcrip- can physically protect b-catenin from polyubiquitylation. tional profile analysis by RNA-seq suggests that CAMRIL3 deficiency As additional evidence, we performed a time-course experiment of significantly reduces cell adhesion. Biochemical and cellular charac- cycloheximide treatment to block new protein synthesis in HEK293T terizations further reveal that lack of CARMIL3 leads to loss of AJ that cells cotransfected with b-catenin-FLAG expression plasmid together involves E-cadherin and b-catenin, two crucial molecules of EMT and with control (TriEx) or CARMIL3-His expression plasmid. The result MET, as well as metastatic colonization. Furthermore, CARMIL3 has a confirmed that CARMIL3 was able to strongly prevent the degradation strong role in regulating Cdh1 transcription downstream of HDACs of b-catenin (Fig. 7E and F). These data provides further evidence that and it sustains Cdh1 transcription by inhibiting the transcriptional CARMIL3 protects b-catenin from ubiquitin-mediated proteolysis. repressor ZEB2 at the protein level, whereas it protects b-catenin from ubiquitination-mediated degradation by the 26S proteasome Interaction of CARMIL3 with the destruction complex of through interfering with the binding between b-catenin and the b-catenin destruction complex, which results in inhibition of phosphorylation Notably, like b-catenin, CARMIL3 was coimmunoprecipitated and ubiquitylation. In addition, the correlations of CARMIL3 with along with ubiquitin (Fig. 7D), which was consistent with the MS both E-cadherin and b-catenin with respect to expression and survival data showing that ubiquitin was among the top of the list of potential are verified in breast cancer patient samples. Collectively, these CARMIL3-interacting proteins (Supplementary Table S2). This observations are summarized in Supplementary Fig. S8. was also confirmed by a reverse IP (Fig. 7G). b-catenin is phosphor- CARMIL3 was initially identified by Hsu and colleagues as an ylated and ubiquitylated by the destruction complex, consisting of oncofetal protein able to increase the proliferation, anchorage- Axin1, APC, the Ser/Thr kinases GSK-3, casein kinase 1 (CK1), and the independent growth, and tumorigenesis of transformed cells in xeno- E3-ubiquitin ligase b-TrCP and subsequently degraded by the ubi- grafts (23). In the 4T1 model, CARMIL3 has a moderate effect on quitin-mediated proteasome (34). We proceeded to test whether promoting tumor cell proliferation, whereas it exhibits a strong CARMIL3 could bind with the destruction complex by examining enhancing effect on tumor metastasis, including tumor cell migration, the interaction between CARMIL3 with Axin1, the key scaffolding invasion, and metastatic colonization. In cancer metastasis, EMT is component of the destruction complex. Indeed, CARMIL3 interacted necessary for converting stationary epithelial tumor cells into motile with Axin1 (Fig. 7H). We also observed that CARMIL3 reduced the mesenchymal cells during metastasis, whereas MET is essential for binding of b-catenin with Axin1 (Fig. 7I). These results suggest that mobile mesenchymal cells to regain the ability to revert to stationary CARMIL3 interacts with the destruction complex of b-catenin via epithelial tumor cells, to proliferate and colonize at secondary tissues/ Axin1, which may explain how CARMIL3 protects b-catenin from organs, forming metastatic niches (5, 11). In this context, we show ubiquitylation. that loss of CARMIL3 results in upregulation of ZEB2 and down- We further examined whether CARMIL3 had any effect on b-cate- regulation of E-cadherin. ZEB2 is one of the core transcriptional nin phosphorylation. The Western blot result showed that the phos- repressors of E-cadherin. These molecular changes maintain cancer phorylated b-catenin level was increased in Carmil3 KO cells (Fig. 7J). cells in a mesenchymal state with the concomitant loss of their ability Moreover, treatment of WT and Carmil3 KO 4T1 cells with SB216763, to complete MET for colonization in the lungs. By the same logic, an inhibitor of GSK3b that phosphorylates b-catenin, totally reversed CARMIL3-sufficient cancer cells maintain the epithelial cell adhesion the protein level of b-catenin (Fig. 7K). Collectively, these results by inhibiting ZEB2 expression, thereby enhancing metastatic coloni- suggest that CARMIL3 inhibits b-catenin phosphorylation, hereby zation of tumor cells in secondary organs. A parallel example of this protecting it from polyubiquitylation and proteasome degradation. nature is the inactivation of Twist1, a transcription factor similar to Moreover, we analyzed the association of CARMIL3 and b-cate- ZEB2, in distant sites allows reversion of EMT for disseminated tumor nin with cancer disease. Correlation analysis of CARMIL3 mRNA cells to proliferate and form metastases (13). Similarly, knockdown of and b-catenin protein expression in primary tumor samples of a Id1 expression in metastasizing cells prevents MET and dramatically patient with breast cancershowed that CARMIL3 positively corre- reduces their lung colonization (35). In addition to the loss of E- lated with b-catenin (Supplementary Fig. S7C). Kaplan–Meier cadherin expression, CARMIL3-deficient cells have also lost b-catenin analysis of patients with breast cancer also showed CARMIL3-high expression (Fig. 5G), which is another important regulator of metas- and b-catenin–high expression associated with poor survival out- tasis independent of the ZEB2/E-cadherin pathway. come (Fig. 7L). Taken together, these observations suggest that Tumor cells undergo developmental processes to acquire migra- CARMIL3 maintains the protein level of b-catenin to facilitate tory and invasive properties that involve a dramatic reorganization cancer progression (Fig. 7M). of actin cytoskeleton and the concomitant formation of membrane protrusions required for invasive growth (8). Furthermore, actin filaments and related proteins undergo wave-like movement that is Discussion closely related with cancer cell membrane protrusion, enabling In this study, we uncover a pivotal role of CARMIL3 in tumor tumor cells to acquire metastatic plasticity and overcome tissue metastasis. This role was initially suggested by our bioinformatic barriers during metastasis. CP is closely related with actin assembly analysis of clinical cancer patient specimens revealing that high dynamics and is able to bind to the fast-growing barbed ends of CARMIL3 expression correlates with poor outcome of breast and actin filament and terminates its elongation at the leading edge of prostate cancer patients. Genetic manipulation of the Carmill3 gene via motile cells, thereby promoting filament nucleation and branching CRISPR/Cas9 in murine breast and prostate cancer cells in vitro and through the Arp2/3 complex and increases the rate of actin-based in vivo demonstrates that CARMIL3 is required for tumor cells’ mobility (29). CP knockdown decreases the size of lamellipodia mobility, migration, and invasion, as well as efficient metastatic whileincreasesthenumberoffilopodia of B16 melanoma cells (36). lung colonization. Mechanistically, IP-MS analysis identifies a group Moreover, CP is sequestered by protein V1 in an inactive complex of CAMRIL3-interacting proteins including CP involved in actin where CARMIL1 can retrieve CP from the CP:V1 complex, which

AACRJournals.org Mol Cancer Res; 2019 OF11

Wang et al.

promotes actin network assembly at protruding edge (37). Knock- In summary, this original study uncovers a novel molecule that is ing down of CARMIL1 and CARMIL2 expression decreases the vitally important for tumor metastasis with a moderate effect on tumor migratory ability of cancer cells in vitro (16, 17). Here we find that growth. It opens new avenues in which to explore not only CARMIL3's CARMIL3 also interacts with CP and loss of CARMIL3 results in unknown activities in normal physiology but also innovative thera- defective actin cytoskeletal organization, which attenuates cell peutic strategies to block cancer metastasis. polarization and formation of focal adhesions, hereby impairing the migratory and invasive ability of tumor cells. Disclosure of Potential Conflicts of Interest CARMIL family proteins have overlapping functions, for instance, No potential conflicts of interest were disclosed. both CARMIL1 and CARMIL2 are capable of binding with CP (16, 21). However, they also have distinct properties. Even though both CAR- Authors’ Contributions MIL1 and CARMIL2 regulate cell migration, one CARMIL isoform Conception and design: H. Wang, X. Ma (CARMIL1 or CARMIL2) is not able to rescue the knockdown Development of methodology: H. Wang, X.-G. Cui phenotypes of the other (17). Moreover, their expression pattern is Acquisition of data (provided animals, acquired and managed patients, provided different and share limited sequence similarities (17, 20). To note, facilities, etc.): H. Wang, C. Wang, G. Peng, D. Yu Analysis and interpretation of data (e.g., statistical analysis, biostatistics, CARMIL3 shares less than 40% of sequence similarities with CAR- computational analysis): H. Wang, C. Wang, X. Ma MIL1 and CARMIL2. Our study for the first time demonstrates that Writing, review, and/or revision of the manuscript:H. Wang, X.-G. Cui, Y.-H. Sun, CARMIL3 is critical for tumor metastasis in vivo, which is mediated X. Ma not only by CP, but also by transcriptional regulation of E-cadherin Administrative, technical, or material support (i.e., reporting or organizing data, and ubiquitin-mediated proteasomal degradation of b-catenin. constructing databases): H. Wang, X. Ma EMT program is finely tuned by several layers of regulation, Study supervision: X.-G. Cui, Y.-H. Sun, X. Ma including the transcriptional and translational machinery, expres- Acknowledgments sion of noncoding RNAs, alternative splicing and protein stabili- We thank Dr. Xuanming Yang of Shanghai Jiao Tong University for kindly Cdh1 ty(38).TheSNAI,ZEB,andTWISTfamilyfactorsarecore providing 4T1 cell line and the help with the lung metastasis assay of 4T1 tumor transcriptional repressors that associate with the NuRD complex model. We also thank Prof. Lin Li of Shanghai Institutes for Biological Sciences for including HDAC1 and HDAC2 to deacetylate the H3K9/14Ac at the kindly providing the Flag-tagged Axin1 plasmid. The work was supported by a grant promoter region (32). Treatment with TSA, the inhibitor of from Natural Science Foundation of China (#81872353; to X. Ma) and by an award HDACs, increased Cdh1 transcription. Meanwhile, Carmil3 expres- from the Shanghai Medical Guidance (Chinese and Western Medicine) Science and Carmil3 Technology Support Project (17411960200; to X. Cui); by an award from The Top- sion was also increased, suggesting that expression is level Clinical Discipline Project of Shanghai Pudong (PWYgf2018-03; to X. Cui), by fi regulated by histone acetylation. Here, we also nd that CARMIL3 awards from the National Natural Science Foundation of China (no. 81773154, inhibits ZEB2 expression at the protein level, which may explain 81772747, 81974391; to X. Cui); by an award from the Program of Shanghai that in the absence of CARMIL3, treatment with TSA could not Academic/Technology Research Leader (no. 19XD1405100; to X. Cui); by an award increase Cdh1 expression to the same level expressed by WT 4T1 from the Clinical Peak Discipline Construction Project of Pudong New Area cells. Of relevance here, our data also suggests that CARMIL3 is able Government (to X. Cui); by awards from the National Natural Science Foundation of China (81773154 and 81301861; to C. Wang); and by an award from Shanghai to locate in the nucleus and interact with the histone variant Natural Science Foundation of China (13ZR1450700; to C. Wang). MacroH2A, which is also implicated in inhibiting MET (39). fi However, the speci c mechanism by which CARMIL3 is involved The costs of publication of this article were defrayed in part by the payment of page in regulation of Cdh1 transcription by HDACs remains to be charges. This article must therefore be hereby marked advertisement in accordance elucidated. In addition, our data (Supplementary Fig. S6D) indicates with 18 U.S.C. Section 1734 solely to indicate this fact. that CARMIL3 does not regulate ZEB2's ubiquitylation. Thus, the issue how CARMIL3 regulates ZEB2 protein level remains to be Received July 13, 2019; revised August 27, 2019; accepted November 1, 2019; further elucidated. published first November 6, 2019.

References 1. Massague J, Obenauf AC. Metastatic colonization by circulating tumour cells. 11. Mittal V. Epithelial mesenchymal transition in tumor metastasis. Annu Rev Nature 2016;529:298–306. Pathol 2018;13:395–412. 2. Lambert AW, Pattabiraman DR, Weinberg RA. Emerging biological principles 12. Chaffer CL, Weinberg RA. A perspective on cancer cell metastasis. Science 2011; of metastasis. Cell 2017;168:670–91. 331:1559–64. 3. Brabletz T, Kalluri R, Nieto MA, Weinberg RA. EMT in cancer. Nat Rev Cancer 13. Tsai JH, Donaher JL, Murphy DA, Chau S, Yang J. Spatiotemporal regulation of 2018;18:128–34. epithelial-mesenchymal transition is essential for squamous cell carcinoma 4. Nieto MA, Huang RY, Jackson RA, Thiery JP. EMT: 2016. Cell 2016;166:21–45. metastasis. Cancer Cell 2012;22:725–36. 5. Alderton GK. Metastasis: epithelial to mesenchymal and back again. Nat Rev 14. Edwards M, Zwolak A, Schafer DA, Sept D, Dominguez R, Cooper JA. Capping Cancer 2013;13:3. protein regulators ﬁne-tune actin assembly dynamics. Nat Rev Mol Cell Biol 6. Gunasinghe NP, Wells A, Thompson EW, Hugo HJ. Mesenchymal-epithelial 2014;15:677–89. transition (MET) as a mechanism for metastatic colonisation in breast cancer. 15. Stark BC, Lanier MH, Cooper JA. CARMIL family proteins as Cancer Metastasis Rev 2012;31:469–78. multidomain regulators of actin-based motility. Mol Biol Cell 2017; 7. Parsons JT, Horwitz AR, Schwartz MA. Cell adhesion: integrating cytoskeletal 28:1713–23. dynamics and cellular tension. Nat Rev Mol Cell Biol 2010;11:633. 16. Yang C, Pring M, Wear MA, Huang M, Cooper JA, Svitkina TM, et al. 8. Yilmaz M, Christofori G. EMT, the cytoskeleton, and cancer cell invasion. Mammalian CARMIL inhibits actin ﬁlament capping by capping protein. Cancer Metastasis Rev 2009;28:15–33. Dev Cell 2005;9:209–21. 9. Christofori G. New signals from the invasive front. Nature 2006;441:444–50. 17. Liang Y, Niederstrasser H, Edwards M, Jackson CE, Cooper JA. Distinct 10. Harris TJC, Tepass U. Adherens junctions: from molecules to morphogenesis. roles for CARMIL isoforms in cell migration. Mol Biol Cell 2009;20: Nat Rev Mol Cell Biol 2010;11:502. 5290–305.

OF12 Mol Cancer Res; 2019 MOLECULAR CANCER RESEARCH

CARMIL3 Regulates Tumor Metastasis

18. Johnson B, McConnell P, Kozlov AG, Mekel M, Lohman TM, Gross 27. Shalem O, Sanjana NE, Hartenian E, Shi X, Scott DA, Mikkelsen TS, et al. ML, et al. Allosteric coupling of CARMIL and V-1 binding to capping Genome-scale CRISPR-Cas9 knockout screening in human cells. Science 2014; protein revealed by hydrogen-deuterium exchange. Cell Rep 2018;23: 343:84–7. 2795–804. 28. Noubissi FK, Elcheva I, Bhatia N, Shakoori A, Ougolkov A, Liu J, et al. CRD-BP 19. Sakiyama M, Matsuo H, Shimizu S, Chiba T, Nakayama A, Takada Y, et al. mediates stabilization of betaTrCP1 and c-myc mRNA in response to beta- Common variant of leucine-rich repeat-containing 16A (LRRC16A) gene is catenin signalling. Nature 2006;441:898–901. associated with gout susceptibility. Hum Cell 2014;27:1–4. 29. Akin O, Mullins RD. Capping protein increases the rate of actin-based motility 20. Liang Y, Cucchetti M, Roncagalli R, Yokosuka T, Malzac A, Bertosio E, et al. The by promoting filament nucleation by the Arp2/3 complex. Cell 2008;133:841–51. lymphoid lineage-specific actin-uncapping protein Rltpr is essential for costi- 30. Kuo JC. Mechanotransduction at focal adhesions: integrating cytoskeletal mulation via CD28 and the development of regulatory T cells. Nat Immunol mechanics in migrating cells. J Cell Mol Med 2013;17:704–12. 2013;14:858–66. 31. Yae T, Tsuchihashi K, Ishimoto T, Motohara T, Yoshikawa M, Yoshida GJ, et al. 21. Roncagalli R, Cucchetti M, Jarmuzynski N, Gregoire C, Bergot E, Audebert S, Alternative splicing of CD44 mRNA by ESRP1 enhances lung colonization of et al. The scaffolding function of the RLTPR protein explains its essential role metastatic cancer cell. Nat Commun 2012;3:883. for CD28 co-stimulation in mouse and human T cells. J Exp Med 2016;213: 32. Tam WL, Weinberg RA. The epigenetics of epithelial-mesenchymal plasticity in 2437–57. cancer. Nat Med 2013;19:1438–49. 22. Schober T, Magg T, Laschinger M, Rohlfs M, Linhares ND, Puchalka J, et al. A 33. Aberle H, Bauer A, Stappert J, Kispert A, Kemler R. b-catenin is a target for the human immunodeficiency syndrome caused by mutations in CARMIL2. ubiquitin–proteasome pathway. EMBO J 1997;16:3797–804. Nat Commun 2017;8:14209. 34. Stamos JL, Weis WI. The b-catenin destruction complex. Cold Spring Harb 23. Hsu CC, Chiang CW, Cheng HC, Chang WT, Chou CY, Tsai HW, et al. Perspect Biol 2013;5:a007898. Identifying LRRC16B as an oncofetal gene with transforming enhancing capa- 35. Stankic M, Pavlovic S, Chin Y, Brogi E, Padua D, Norton L, et al. TGF-b-Id1 bility using a combined bioinformatics and experimental approach. Oncogene signaling opposes Twist1 and promotes metastatic colonization via a mesen- 2011;30:654–67. chymal-to-epithelial transition. Cell Rep 2013;5:1228–42. 24. Spence EF, Dube S, Uezu A, Locke M, Soderblom EJ, Soderling SH. 36. Mejillano MR, Kojima S, Applewhite DA, Gertler FB, Svitkina TM, Borisy GG. In vivo proximity proteomics of nascent synapses reveals a novel Lamellipodial versus filopodial mode of the actin nanomachinery: pivotal role of regulator of cytoskeleton-mediated synaptic maturation. Nat Commun the filament barbed end. Cell 2004;118:363–73. 2019;10:386–401. 37. Fujiwara I, Remmert K, Piszczek G, Hammer JA. Capping protein regulatory 25. Charoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder D, et al. cycle driven by CARMIL and V-1 may promote actin network assembly at Pan-cancer immunogenomic analyses reveal genotype-immunophenotype rela- protruding edges. Proc Natl Acad Sci U S A 2014;111:E1970–9. tionships and predictors of response to checkpoint blockade. Cell Rep 2017;18: 38. De Craene B, Berx G. Regulatory networks defining EMT during cancer initiation 248–62. and progression. Nat Rev Cancer 2013;13:97–110. 26. Liu J, Lichtenberg T, Hoadley KA, Poisson LM, Lazar AJ, Cherniack AD, et al. An 39. Pliatska M, Kapasa M, Kokkalis A, Polyzos A, Thanos D. The histone variant integrated TCGA pan-cancer clinical data resource to drive high-quality survival macroH2A blocks cellular reprogramming by inhibiting mesenchymal-to- outcome analytics. Cell 2018;173:400–16. epithelial transition. Mol Cell Biol 2018;38:pii:e00669–17.

AACRJournals.org Mol Cancer Res; 2019 OF13

Capping Protein Regulator and Myosin 1 Linker 3 Is Required for Tumor Metastasis

Huan Wang, Chao Wang, Guang Peng, et al.

Mol Cancer Res Published OnlineFirst November 6, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/1541-7786.MCR-19-0722

Supplementary Access the most recent supplemental material at: Material http://mcr.aacrjournals.org/content/suppl/2019/11/06/1541-7786.MCR-19-0722.DC1

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

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mcr.aacrjournals.org/content/early/2019/12/02/1541-7786.MCR-19-0722. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.