A zebrafish embryo screen utilizing gastrulation identifies the HTR2C inhibitor pizotifen as a suppressor of EMT-mediated metastasis.

Metastasis is responsible for approximately 90% of cancer-associated mortality but few models exist that allow for rapid and effective screening of anti-metastasis drugs. Current mouse models of metastasis are too expensive and time consuming to use for rapid and high-throughput screening. Therefore, we created a unique screening concept utilizing conserved mechanisms between zebrafish gastrulation and cancer metastasis for identification of potential anti-metastatic drugs. We hypothesized that small chemicals that interrupt zebrafish gastrulation might also suppress metastatic progression of cancer cells and developed a phenotype-based chemical screen to test the hypothesis. The screen used epiboly, the first morphogenetic movement in gastrulation, as a marker and enabled 100 chemicals to be tested in 5 hr. The screen tested 1280 FDA-approved drugs and identified pizotifen, an antagonist for serotonin receptor 2C (HTR2C) as an epiboly-interrupting drug. Pharmacological and genetic inhibition of HTR2C suppressed metastatic progression in a mouse model. Blocking HTR2C with pizotifen restored epithelial properties to metastatic cells through inhibition of Wnt signaling. In contrast, HTR2C induced epithelial-to-mesenchymal transition through activation of Wnt signaling and promoted metastatic dissemination of human cancer cells in a zebrafish xenotransplantation model. Taken together, our concept offers a novel platform for discovery of anti-metastasis drugs.

Introduction

Metastasis, a leading contributor to the morbidity of cancer patients, occurs through multiple steps: invasion, intravasation, extravasation, colonization, and metastatic tumor formation (Nguyen et al., 2009; Welch and Hurst, 2019; Chaffer and Weinberg, 2011). The physical translocation of cancer cells is an initial step of metastasis and molecular mechanisms of it involve cell motility, the breakdown of local basement membrane, loss of cell polarity, acquisition of stem cell-like properties, and epithelial-to-mesenchymal transition (EMT) (Tsai and Yang, 2013; Lu and Kang, 2019). These cell-biological phenomena are also observed during vertebrate gastrulation in that evolutionarily conserved morphogenetic movements of epiboly, internalization, convergence, and extension progress (Solnica-Krezel, 2005). In zebrafish, the first morphogenetic movement, epiboly, is initiated at approximately 4 hr post fertilization (hpf) to move cells from the animal pole to eventually engulf the entire yolk cell by 10 hpf (Latimer and Jessen, 2010; Solnica-Krezel, 2006). The embryonic cell movements are governed by the molecular mechanisms that are partially shared in metastatic cell dissemination.

At least 50 common genes were shown to be involved in both metastasis and gastrulation progression: Knockdown of these genes in Xenopus or zebrafish induced gastrulation defects; conversely, overexpression of these genes conferred metastatic potential on cancer cells and knockdown of these genes suppressed metastasis (Yang and Weinberg, 2008; Dongre and Weinberg, 2019; Thiery et al., 2009; Nieto et al., 2016; Table 1). This evidence led us to hypothesize that small molecules that interrupt zebrafish gastrulation may suppress metastatic progression of human cancer cells.

Table 1.

A list of the genes that are involved between gastrulation and metastasis progression.

A list of the 50 genes that play essential role in governing both metastasis and gastrulation progression. The gastrulation defects in Xenopus or zebrafish that are induced by knockdown of each of these genes were indicated. The molecular mechanism in metastasis that is inhibited by knockdown of each of the same genes was indicated.

Genes	Gastrulation defects	Ref	Effects in metastasis	Ref
BMP	Convergence and extension	Kondo, 2007	EMT	Katsuno et al., 2008
WNT	Convergence and extension	Tada and Smith, 2000	Migration and invasion	Vincan and Barker, 2008
FGF	Convergence and extension	Yang et al., 2002	Invision	Nomura et al., 2008
EGF	Epiboly	Song et al., 2013	Migration	Lu et al., 2001
PDGF	Convergence and extension	Damm and Winklbauer, 2011	EMT	Jechlinger et al., 2006
CXCL12	Migration of endodermal cells	Mizoguchi et al., 2008	Migration and invasion	Shen et al., 2013
CXCR4	Migration of endodermal cells	Mizoguchi et al., 2008	Migration and invasion	Shen et al., 2013
PIK3CA	Convergence and extension	Montero et al., 2003	Migration and invasion	Wander et al., 2013
YES	Epiboly	Tsai et al., 2005	Migration	Barraclough et al., 2007
FYN	Epiboly	Sharma et al., 2005	Migration and invasion	Yadav and Denning, 2011
MAPK1	Epiboly	Krens et al., 2008	Migration	Radtke et al., 2013
SHP2	Convergence and extension	Jopling et al., 2007	Migration	Wang et al., 2005
SNAI1	Convergence and extension	Ip and Gridley, 2002	EMT	Batlle et al., 2000
SNAI2	Mesoderm and neural crest formation	Shi et al., 2011	EMT	Medici et al., 2008
TWIST1	Mesoderm formation	Castanon and Baylies, 2002	EMT	Yang et al., 2004
TBXT	Convergence and extension	Tada and Smith, 2000	EMT	Fernando et al., 2010
ZEB1	Epiboly	Vannier et al., 2013	EMT	Spaderna et al., 2008
GSC	Mesodermal patterning	Sander et al., 2007	EMT	Hartwell et al., 2006
FOXC2	Unclear, defects in gastrulation	Wilm et al., 2004	EMT	Mani et al., 2007
STAT3	Convergence and extension	Miyagi et al., 2004	Migration	Abdulghani et al., 2008
POU5F1	Epiboly	Lachnit et al., 2008	EMT	Dai et al., 2013
EZH2	Unclear, defects in gastrulation	O’Carroll et al., 2001	Invasion	Ren et al., 2012
EHMT2	Defects in neurogenesis	Lin et al., 2005	Migration and invasion	Chen et al., 2010
BMI1	Defects in skeleton formation	van der Lugt et al., 1994	EMT	Guo et al., 2011
RHOA	Convergence and extension	Zhu et al., 2006	Migration and invasion	Yoshioka et al., 1999
CDC42	Convergence and extension	Choi and Han, 2002	Migration and invasion	Reymond et al., 2012
RAC1	Convergence and extension	Habas et al., 2003	Migration and invasion	Vega and Ridley, 2008
ROCK2	Convergence and extension	Marlow et al., 2002	Migration and invasion	Itoh et al., 1999
PAR1	Convergence and extension	Kusakabe and Nishida, 2004	Migration	Shi et al., 2004
PRKCI	Convergence and extension	Kusakabe and Nishida, 2004	EMT	Gunaratne et al., 2013
CAP1	Convergence and extension	Seifert et al., 2009	Migration	Yamazaki et al., 2009
EZR	Epiboly	Link et al., 2006	Migration	Khanna et al., 2004
EPCAM	Epiboly	Slanchev et al., 2009	Migration and invasion	Ni et al., 2012
ITGB1/ ITA5	Mesodermal migration	Skalski et al., 1998	Migration and invasion	Felding-Habermann, 2003
FN1	Convergence and extension	Marsden and DeSimone, 2003	Invasion	Malik et al., 2010
HAS2	Dorsal migration of lateral cells	Bakkers et al., 2004	Invasion	Kim et al., 2004
MMP14	Convergence and extension	Coyle et al., 2008	Invasion	Perentes et al., 2011
COX1	Epiboly	Cha et al., 2006	Invasion	Kundu and Fulton, 2002
PTGES	Convergence and extension	Speirs et al., 2010	Invasion	Wang and Dubois, 2006
SLC39A6	Anterior migration	Yamashita et al., 2004	EMT	Lue et al., 2011
GNA12 /13	Convergence and extension	Lin et al., 2005	Migration and invasion	Yagi et al., 2011
OGT	Epiboly	Webster et al., 2009	Migration and invasion	Lynch et al., 2012
CCN1	Cell movement	Latinkic et al., 2003	Migration and invasion	Lin et al., 2012
TRPM7	Convergence and extension	Liu et al., 2011	Migration	Middelbeek et al., 2012
MAPKAPK2	Epiboly	Holloway et al., 2009	Migration	Kumar et al., 2010
B4GALT1	Convergence and extension	Machingo et al., 2006	Invasion	Zhu et al., 2005
IER2	Convergence and extension	Hong et al., 2011	Migration	Neeb et al., 2012
TIP1	Convergence and extension	Besser et al., 2007	Migration and invasion	Han et al., 2012
PAK5	Convergence and extension	Faure et al., 2005	Migration	Gong et al., 2009
MARCKS	Convergence and extension	Iioka et al., 2004	Migration and invasion	Rombouts et al., 2013

Here, we report a unique screening concept based on the hypothesis. Pizotifen, an antagonist for HTR2C, was identified from the screen as a ‘hit’ that interrupted zebrafish gastrulation. A mouse model of metastasis confirmed pharmacological and genetic inhibition of HTR2C suppressed metastatic progression. Moreover, HTR2C induced EMT and promoted metastatic dissemination of non-metastatic cancer cells in a zebrafish xenotransplantation model. These results demonstrated that this concept could offer a novel high-throughput platform for discovery of anti-metastasis drugs and can be converted to a chemical genetic screening platform.

Results

Small molecules interrupting epiboly of zebrafish have a potential to suppress metastatic progression of human cancer cells

Before performing a screening assay, we validated a core of our concept through comparing the genes expressed in zebrafish gastrulation with the genes which expressed in EMT-mediated metastasis. Gene set enrichment analysis (GSEA) demonstrated that 50%-epiboly, shield, and 75%-epiboly stage of zebrafish embryos expressed the genes which promote EMT-mediated metastasis: EMT induction, TGF-β signaling, wnt/β-catenin signaling, Notch signaling (Figure 1—figure supplement 1).

Figure 1—figure supplement 1.

Gene expression profiles obtained from zebrafish embryos at either 50%-epiboly (top left), shield (top right), or 75%-epiboly stage (bottom left) were analyzed based on the hallmark gene sets derived from the Molecular Signatures Database (MSigDB) (Liberzon et al., 2015).

The zebrafish transcriptomic data was sourced from White et al., 2017 eLife (Subramanian et al., 2005). Gene sets that were significantly enriched (FDR < 0.25) were presented with the normalized enrichment score (NES) and nominal p value. Source data files for analysis of either gene expression and enriched pathways are uploaded as gene set enrichment analysis (GSEA) Source data 1 and 2, respectively.

We further conducted preliminary experiments to test the hypothesis. First, we examined whether hindering the molecular function of reported genes, whose knockdown induced gastrulation defects in zebrafish, might suppress cell motility and invasion of cancer cells. We chose protein arginine methyltransferase 1 (PRMT1) and cytochrome P450 family 11 (CYP11A1), both of whose knockdown induced gastrulation defects in zebrafish but whose involvement in metastatic progression is unclear (Tsai et al., 2011; Hsu et al., 2006). Elevated expression of PRMT1 and CYP11A1 was observed in highly metastatic human breast cancer cell lines and knockdown of these genes through RNA interference suppressed the motility and invasion of MDA-MB-231 cells without affecting their viability (Figure 1—figure supplement 2A-C).

Figure 1—figure supplement 2.

Epiboly could serve as a marker for this screening.

(A) Western blot analysis of protein arginine methyltransferase 1 (PRMT1) (upper left) and cytochrome P450 family 11 (CYP11A1) (middle left) protein levels in non-metastatic human cancer cell line (MCF7) and highly metastatic human cancer cell lines (MDA-MB-231, MDA-MB-435, MIA-PaCa2, PC9, HCCLM3, PC3, and SW620); β-actin loading control is shown (bottom left). Preliminary experiments confirmed that epiboly could serve as a marker for this screening assay. Quantification analyses of western blotting bands. The analyses were performed by ImageJ. Signal strength of bands of PRMT1 (Top right) and CYP11A1 (bottom right) was normalized by that of β-actin. (B) Knockdown of PRMT1 or CYP11A1 in MDA-MB-231 cells. MBA-MB-231 cells were transfected with a control short hairpin RNA (shRNA) targeting LacZ, and one of four independent shRNAs targeting PRMT1 (clones #1 to #4), or one of two independent shRNAs targeting CYP11A1 (clones #1 to #4). Reduced PRMT1 and CYP11A1 expression, determined by western blot, in sub-cell lines of MDA-MB-231 cells expressing PRMT1 shRNA (clones #3 and #4) or CYP11A1 (clones #2 and #4), compared with controls (parental cell line MDA-MB-231 and control shRNA cells); β-actin levels shown as a loading control. Quantification analyses of western blotting bands. The analyses were performed by ImageJ. Signal strength of bands of PRMT1 (top right) and CYP11A1 (bottom right) was normalized by that of β-actin. (C) Effect of shRNAs targeting either PRMT1 or CYP11A1 on cell motility and invasion of MBA-MB-231 cells. Parental MDA-MB-231 cells and four sub-cell lines of MDA-MB-231 cells that were transfected with either shRNA targeting either LacZ, two independent shRNAs targeting PRMT1 (clones #3 and #4) or two independent shRNAs targeting CYP11A1 (clones #2 and #4) were subjected to Boyden chamber assays. (D) Zebrafish embryos treated with either vehicle (DMSO), 10 μM niclosamide, or 50 µM vinpocetine. Approximately 20 embryos were treated with either DMSO as a vehicle control, niclosamide, or vinpocetine. The treatment was started at 4 hours post fertilization (hpf) when all of embryos reached sphere stage and ended at 9 hpf when control embryos reached 80–90% epiboly stage. Each experiment was performed at least twice. Statistical analysis was determined by Student’s t test.

Next, we conducted an inverse examination of whether chemicals which were reported to suppress metastatic dissemination of cancer cells could interrupt epiboly progression of zebrafish embryos. Niclosamide and vinpocetine are reported to suppress metastatic progression (Weinbach and Garbus, 1969; Sack et al., 2011; Huang et al., 2012; Szilágyi et al., 2005). Either niclosamide- or vinpocetine-treated zebrafish embryos showed complete arrest at very early stages or severe delay in epiboly progression, respectively (Figure 1—figure supplement 2D).

These results suggest that epiboly could serve as a marker for this screening assay and epiboly-interrupting drugs that are identified through this screening could have the potential to suppress metastatic progression of human cancer cells.

132 FDA-approved drugs induced delayed in epiboly of zebrafish embryos

We screened 1280 FDA, EMA, or other agencies-approved drugs (Prestwick, Inc) in our zebrafish assay. The screening showed that 0.9% (12/1280) of the drugs, including antimycin A and tolcapone, induced severe or complete arrest of embryonic cell movement when embryos were treated with 10 μM. 5.2% (66/1280) of the drugs, such as dicumarol, racecadotril, pizotifen, and S(-)eticlopride hydrochloride, induced either delayed epiboly or interrupted epiboly of the embryos. 93.3% (1194/1280) of drugs have no effect on epiboly progression of the embryos. 0.6% (8/1280) of drugs induced toxic lethality. Epiboly progression was affected more severely when embryos were treated with 50 μM; 1.7% (22/1280) of the drugs induced severe or complete arrest of it. 8.6% (110/1280) of the drugs induced either delayed epiboly or interrupt epiboly of the embryos. 4.3% (55/1280) of drugs induced a toxic lethality (Figure 1A and B, Table 2). Among the epiboly-interrupting drugs, several drugs have already been reported to inhibit metastasis-related molecular mechanisms: adrenosterone or zardaverine, which target HSD11β1 or PDE3 and -4, respectively, are reported to inhibit EMT (Nakayama et al., 2020; Kolosionek et al., 2009); racecadotril, which targets enkephalinase, is reported to confer metastatic potential on colon cancer cell (Sasaki et al., 2014); and disulfiram, which targets ALDH (aldehyde dehydrogenase), is reported to confer stem-like properties on metastatic cancer cells (Liu et al., 2013). This evidence suggests that epiboly-interrupting drugs have the potential for suppressing metastasis of human cancer cells.

Figure 1.

A chemical screen for identification of epiboly-interrupting drugs.

(A) Cumulative results of the chemical screen in which each drug was used at either 10 µM (left) or 50 µM (right) concentrations. 1280 FDA, EMA, or other agencies-approved drugs were subjected to this screening. Positive ‘hit’ drugs were those that interrupted epiboly progression. (B) Representative samples of the embryos that were treated with indicated drugs.

The zebrafish transcriptomic data was sourced from White et al., 2017 eLife (Subramanian et al., 2005). Gene sets that were significantly enriched (FDR < 0.25) were presented with the normalized enrichment score (NES) and nominal p value. Source data files for analysis of either gene expression and enriched pathways are uploaded as gene set enrichment analysis (GSEA) Source data 1 and 2, respectively.

(A) Western blot analysis of protein arginine methyltransferase 1 (PRMT1) (upper left) and cytochrome P450 family 11 (CYP11A1) (middle left) protein levels in non-metastatic human cancer cell line (MCF7) and highly metastatic human cancer cell lines (MDA-MB-231, MDA-MB-435, MIA-PaCa2, PC9, HCCLM3, PC3, and SW620); β-actin loading control is shown (bottom left). Preliminary experiments confirmed that epiboly could serve as a marker for this screening assay. Quantification analyses of western blotting bands. The analyses were performed by ImageJ. Signal strength of bands of PRMT1 (Top right) and CYP11A1 (bottom right) was normalized by that of β-actin. (B) Knockdown of PRMT1 or CYP11A1 in MDA-MB-231 cells. MBA-MB-231 cells were transfected with a control short hairpin RNA (shRNA) targeting LacZ, and one of four independent shRNAs targeting PRMT1 (clones #1 to #4), or one of two independent shRNAs targeting CYP11A1 (clones #1 to #4). Reduced PRMT1 and CYP11A1 expression, determined by western blot, in sub-cell lines of MDA-MB-231 cells expressing PRMT1 shRNA (clones #3 and #4) or CYP11A1 (clones #2 and #4), compared with controls (parental cell line MDA-MB-231 and control shRNA cells); β-actin levels shown as a loading control. Quantification analyses of western blotting bands. The analyses were performed by ImageJ. Signal strength of bands of PRMT1 (top right) and CYP11A1 (bottom right) was normalized by that of β-actin. (C) Effect of shRNAs targeting either PRMT1 or CYP11A1 on cell motility and invasion of MBA-MB-231 cells. Parental MDA-MB-231 cells and four sub-cell lines of MDA-MB-231 cells that were transfected with either shRNA targeting either LacZ, two independent shRNAs targeting PRMT1 (clones #3 and #4) or two independent shRNAs targeting CYP11A1 (clones #2 and #4) were subjected to Boyden chamber assays. (D) Zebrafish embryos treated with either vehicle (DMSO), 10 μM niclosamide, or 50 µM vinpocetine. Approximately 20 embryos were treated with either DMSO as a vehicle control, niclosamide, or vinpocetine. The treatment was started at 4 hours post fertilization (hpf) when all of embryos reached sphere stage and ended at 9 hpf when control embryos reached 80–90% epiboly stage. Each experiment was performed at least twice. Statistical analysis was determined by Student’s t test.

Table 2.

A list of the drugs that interfere with epiboly progression in zebrafish.

Related to Figure 1. A list of positive ‘hit’ drugs that interfered with epiboly progression. Gastrulation defects or status of each of the zebrafish embryos that were treated with either 10 or 50 μM concentrations are indicated.

Chemical name	Chemical formula	Effect of 10 µM	Effect of 50 µM
Acitretin	C21H26O3	Delayed	Delayed
Adrenosterone	C19H24O3	Delayed	Delayed
Albendazole	C12H15N3O2S	Severe delayed	Severe delayed
Alfadolone acetate	C23H34O5	Delayed	Delayed
Alfaxalone	C21H32O3	Delayed	Delayed
Alprostadil	C20H34O5	Delayed	Delayed
Altrenogest	C21H26O2	Slightly delayed	Delayed
Ampiroxicam	C20H21N3O7S	Non-effect	Delayed
Anethole-trithione	C10H8OS3	Delayed	Delayed
Antimycin A	C28H40N2O9	Delayed	Delayed
Avobenzone	C20H22O3	Delayed	Delayed
Benzoxiquine	C16H11NO2	Non-effect	Delayed
Bosentan	C27H29N5O6S	Delayed	Delayed
Butoconazole nitrate	C19H18Cl3N3O3S	Delayed	Toxic lethal
Camptothecine (S,+)	C20H16N2O4	Severe delayed	Severe delayed
Carbenoxolone disodium salt	C34H48Na2O7	Delayed	Toxic lethal
Carmofur	C11H16FN3O3	Slightly delayed	Delayed
Carprofen	C15H12ClNO2	Severe delayed	Toxic lethal
Cefdinir	C14H13N5O5S2	Delayed	Delayed
Celecoxib	C17H14F3N3O2S	Delayed	Delayed
Chlorambucil	C14H19Cl2NO2	Slightly delayed	Delayed
Chlorhexidine	C22H30Cl2N10	Non-effect	Toxic lethal
Ciclopirox ethanolamine	C14H24N2O3	Delayed	Severe delayed
Cinoxacin	C12H10N2O5	Delayed	Severe delayed
Clofibrate	C12H15ClO3	Non-effect	Severe delayed
Clopidogrel	C16H16ClNO2S	Non-effect	Delayed
Clorgyline hydrochloride	C13H16Cl3NO	Delayed	Delayed
Colchicine	C22H25NO6	Non-effect	Delayed
Deptropine citrate	C29H35NO8	Delayed	Delayed
Desipramine hydrochloride	C18H23ClN2	Delayed	Delayed
Diclofenac sodium	C14H10Cl2NNaO2	Delayed	Severe delayed
Dicumarol	C19H12O6	Delayed	Severe delayed
Diethylstilbestrol	C18H20O2	Delayed	Toxic lethal
Dimaprit dihydrochloride	C6H17Cl2N3S	Slightly delayed	Delayed
Disulfiram	C10H20N2S4	Delayed	Delayed
Dopamine hydrochloride	C8H12ClNO2	Delayed	Delayed
Eburnamonine (-)	C19H22N2O	Delayed	Delayed
Ethaverine hydrochloride	C24H30ClNO4	Delayed	Delayed
Ethinylestradiol	C20H24O2	Delayed	Severe delayed
Ethopropazine hydrochloride	C19H25ClN2S	Delayed	Delayed
Ethoxyquin	C14H19NO	Non-effect	Delayed
Exemestane	C20H24O2	Slightly delayed	Delayed
Ezetimibe	C24H21F2NO3	Slightly delayed	Delayed
Fenbendazole	C15H13N3O2S	Non-effect	Delayed
Fenoprofen calcium salt dihydrate	C30H30CaO8	Slightly delayed	Delayed
Fentiazac	C17H12ClNO2S	Toxic lethal	Toxic lethal
Floxuridine	C9H11FN2O5	Delayed	Toxic lethal
Flunixin meglumine	C21H28F3N3O7	Delayed	Toxic lethal
Flutamide	C11H11F3N2O3	Delayed	Toxic lethal
Fluticasone propionate	C25H31F3O5S	Non-effect	Delayed
Furosemide	C12H11ClN2O5S	Delayed	Delayed
Gatifloxacin	C19H22FN3O4	Delayed	Delayed
Gemcitabine	C9H11F2N3O4	Delayed	Delayed
Gemfibrozil	C15H22O3	Delayed	Toxic lethal
Gestrinone	C21H24O2	Delayed	Delayed
Haloprogin	C9H4Cl3IO	Delayed	Toxic lethal
Hexachlorophene	C13H6Cl6O2	Delayed	Severe delayed
Hexestrol	C18H22O2	Slightly delayed	Delayed
Ibudilast	C14H18N2O	Non-effect	Delayed
Idazoxan hydrochloride	C11H13ClN2O2	Slightly delayed	Delayed
Idazoxan hydrochloride	C11H13ClN2O2	Non-effect	Delayed
Idebenone	C19H30O5	Severe delayed	Toxic lethal
Indomethacin	C19H16ClNO4	Non-effect	Delayed
Ipriflavone	C18H16O3	Delayed	Severe delayed
Isotretinoin	C20H28O2	Non-effect	Severe delayed
Isradipine	C19H21N3O5	Non-effect	Delayed
Lansoprazole	C16H14F3N3O2S	Slightly delayed	Delayed
Latanoprost	C26H40O5	Non-effect	Delayed
Leflunomide	C12H9F3N2O2	Delayed	Severe delayed
Letrozole	C17H11N5	Non-effect	Delayed
Lithocholic acid	C24H40O3	Non-effect	Delayed
Lodoxamide	C11H6ClN3O6	Non-effect	Delayed
Lofepramine	C26H27ClN2O	Non-effect	Delayed
Loratadine	C22H23ClN2O2	Delayed	Delayed
Loxapine succinate	C22H24ClN3O5	Delayed	Delayed
Mebendazole	C16H13N3O3	Severe delayed	Severe delayed
Mebendazole	C22H26N2O2	Non-effect	Delayed
Meloxicam	C14H13N3O4S2	Delayed	Toxic lethal
Methiazole	C12H15N3O2S	Delayed	Delayed
Mevastatin	C23H34O5	Non-effect	Delayed
MK 801 hydrogen maleate	C20H19NO4	Slightly delayed	Delayed
Nabumetone	C15H16O2	Non-effect	Severe delayed
Naftopidil dihydrochloride	C24H30Cl2N2O3	Slightly delayed	Delayed
Nandrolone	C18H26O2	Delayed	Delayed
Naproxen sodium salt	C14H13NaO3	Delayed	Delayed
Niclosamide	C13H8Cl2N2O4	Delayed	Delayed
Nifekalant	C19H27N5O5	Delayed	Delayed
Niflumic acid	C13H9F3N2O2	Delayed	Delayed
Nimesulide	C13H12N2O5S	Non-effect	Delayed
Nisoldipine	C20H24N2O6	Delayed	Toxic lethal
Nitazoxanide	C12H9N3O5S	Severe delayed	Severe delayed
Norethindrone	C20H26O2	Non-effect	Delayed
Norgestimate	C23H31NO3	Slightly delayed	Delayed
Oxfendazol	C15H13N3O3S	Slightly delayed	Delayed
Oxibendazol	C12H15N3O3	Severe delayed	Severe delayed
Oxymetholone	C21H32O3	Slightly delayed	Delayed
Parbendazole	C13H17N3O2	Severe delayed	Severe delayed
Parthenolide	C15H20O3	Non-effect	Delayed
Penciclovir	C10H15N5O3	Non-effect	Delayed
Pentobarbital	C11H18N2O3	Non-effect	Delayed
Phenazopyridine hydrochloride	C11H12ClN5	Delayed	Toxic lethal
Phenothiazine	C12H9NS	Non-effect	Delayed
Phenoxybenzamine hydrochloride	C18H23Cl2NO	Non-effect	Delayed
Pizotifen malate	C23H27NO5S	Delayed	Severe delayed
Pramoxine hydrochloride	C17H28ClNO3	Slightly delayed	Delayed
Prilocaine hydrochloride	C13H21ClN2O	Non-effect	Delayed
Primidone	C12H14N2O2	Slightly delayed	Delayed
Racecadotril	C21H23NO4S	Slightly delayed	Delayed
Riluzole hydrochloride	C8H6ClF3N2OS	Non-effect	Delayed
Ritonavir	C37H48N6O5S2	Non-effect	Severe delayed
S(-)Eticlopride hydrochloride	C17H26Cl2N2O3	Delayed	Delayed
Salmeterol	C25H37NO4	Non-effect	Delayed
Streptomycin sulfate	C42H84N14O36S3	Non-effect	Delayed
Sulconazole nitrate	C18H16Cl3N3O3S	Delayed	Delayed
Tegafur	C8H9FN2O3	Delayed	Delayed
Telmisartan	C33H30N4O2	Severe delayed	Toxic lethal
Tenatoprazole	C16H18N4O3S	Non-effect	Delayed
Terbinafine	C21H25N	Non-effect	Delayed
Thimerosal	C9H9HgNaO2S	Non-effect	Delayed
Thiorphan	C12H15NO3S	Delayed	Delayed
Tolcapone	C14H11NO5	Severe delayed	Severe delayed
Topotecan	C23H23N3O5	Delayed	Delayed
Tracazolate hydrochloride	C16H25ClN4O2	Severe delayed	Delayed
Tribenoside	C29H34O6	Delayed	Delayed
Triclabendazole	C14H9Cl3N2OS	Delayed	Delayed
Triclosan	C12H7Cl3O2	Delayed	Severe delayed
Trioxsalen	C14H12O3	Delayed	Delayed
Troglitazone	C24H27NO5S	Severe delayed	Toxic lethal
Valproic acid	C8H16O2	Non-effect	Delayed
Voriconazole	C16H14F3N5O	Non-effect	Delayed
Zardaverine	C12H10F2N2O3	Slightly delayed	Delayed
Zuclopenthixol dihydrochloride	C22H27Cl3N2OS	Delayed	Delayed

Identified drugs suppressed cell motility and invasion of human cancer cells

It has been reported that zebrafish have orthologues to 86% of 1318 human drug targets (Gunnarsson et al., 2008). However, it was not known whether the epiboly-interrupting drugs could suppress metastatic dissemination of human cancer cells. To test this, we subjected the 78 epiboly-interrupting drugs that showed a suppressor effect on epiboly progression at a 10 μM concentration to in vitro experiments using a human cancer cell line. The experiments examined whether the drugs could suppress cell motility and invasion of MDA-MB-231 cells through a Boyden chamber. Before conducting the experiment, we investigated whether these drugs might affect viability of MDA-MB-231 cells using an MTT assay. Out of the 78 drugs, 16 of them strongly affected cell viability at concentrations less than 1 μM and were not used in the cell motility experiments. The remaining 62 drugs were assayed in Boyden chamber motility experiments. Out of the 62 drugs, 20 of the drugs inhibited cell motility and invasion of MDA-MB-231 cells without affecting cell viability. Among the 20 drugs, hexachlorophene and nitazoxanide were removed since the primary targets of the drugs, D-lactate dehydrogenase and pyruvate ferredoxin oxidoreductase, are not expressed in mammalian cells. With the exception of ipriflavone, whose target is still unclear, the known primary targets of the remaining 17 drugs are reported to be expressed by mammalian cells (Figure 2A and Table 3).

Figure 2.

Pizotifen, one of epiboly-interrupting drugs, suppressed metastatic dissemination of human cancer cells lines in vivo and vitro.

(A) Effect of the epiboly-interrupting drugs on cell motility and invasion of MBA-MB-231 cells. MBA-MB-231 cells were treated with vehicle or each of the epiboly-interrupting drugs and then subjected to Boyden chamber assays. Fetal bovine serum (1% v/v) was used as the chemoattractant in both assays. Each experiment was performed at least twice. (B) Western blot analysis of HTR2C levels (top) in a non-metastatic human cancer cell line, MCF7 (breast) and highly metastatic human cancer cell lines, MDA-MB-231 (breast), MDA-MB-435 (melanoma), PC9 (lung), MIA-PaCa2 (pancreas), PC3 (prostate), and SW620 (colon); GAPDH loading control is shown (bottom). (C) Effect of pizotifen on cell motility and invasion of MBA-MB-231, MDA-MB-435, and PC9 cells. Either vehicle or pizotifen treated the cells were subjected to Boyden chamber assays. Fetal bovine serum (1% v/v) was used as the chemoattractant in both assays. Each experiment was performed at least twice. (D) and (E) Representative images of dissemination of 231R, shLacZ 231R or shHTR2C 231R cells in zebrafish xenotransplantation model. The fish larvae that were inoculated with 231R cells were treated with either vehicle (top left) or the drug (lower left) (D). The fish larvae that were inoculated with either shLacZ 231R or shHTR2C 231R cells (lower left) (E). White arrows head indicate disseminated 231R cells. The images were shown in 4× magnification. Scale bar, 100 µm. The mean frequencies of the fish showing head, trunk, or end-tail dissemination were counted (graph on right). Each value is indicated as the mean ± SEM of two independent experiments. Statistical analysis was determined by Student’s t test.

(A) Quantification analyses of western blotting bands in Figure 2B. The analyses were performed by ImageJ. Signal strength of bands of HTR2C (left) and DRD2 (right) was normalized by that of GAPDH. (B) Western blot analysis of DRD2 levels in non-metastatic human cancer cell line, MCF7 (breast) and highly metastatic human cancer cell lines, MDA-MB-231 (breast), MDA-MB-435 (melanoma), MIA-PaCa2 (pancreas), PC3 (prostate), and SW620 (colon); GAPDH loading control is shown (bottom). GAPDH control was obtained in the same experiment from Figure 2B. (C) Effect of S(-)eticlopride hydrochloride on cell motility and invasion of MBA-MB-231, MDA-MB-435, and PC9 cells. Either vehicle- or pizotifen-treated cells were subjected to Boyden chamber assays. Fetal bovine serum (1% v/v) was used as the chemoattractant in both assays. Each experiment was performed at least twice. Statistical analysis was determined by Student’s t test.

(A) Representative images of dissemination of 231R cells in zebrafish xenotransplantation model. The fish larvae that were inoculated with 231R cells were treated with either vehicle (top left, bottom left) or pizotifen (top right, bottom right). (B) Representative images of dissemination of MIA-PaCa2 cells in zebrafish xenotransplantation model. The fish were inoculated with MIA-PaCa2 cells, and treated with either vehicle (top left) or drug (lower left). White arrow heads indicate disseminated MIA-PaCa2 cells. The images were shown in 4× magnification. Scale bar, 100 μm. The mean frequencies of the fish showing head, trunk, or end-tail dissemination were tabulated (right). Each value is indicated as the mean ± SEM of two independent experiments. Statistical analysis was determined by Student’s t test.

Table 3.

Primary targets of the identified drugs.

The identified drugs	Primary targets of the identified drugs
Hexachlorophene	D-Lactate dehydrogenase (D-LDH), not expressed in mammalian cells
Troglitazone	Agonist for peroxisome proliferator-activated receptor α and γ (PPARα and -γ)
Pizotifen malate	5-Hydroxytryptamine receptor 2C (HTR2C)
Salmeterol	Adrenergic receptor beta 2 (ADRB2)
Nitazoxanide	Pyruvate ferredoxin oxidoreductase (PFOR), not expressed in mammalian cells
Valproic acid	Histone deacetylases (HDACs)
Dicumarol	NAD(P)H dehydrogenase quinone 1 (NQO1)
Loxapine succinate	Dopamine receptor D2 and D4 (DRD2 and DRD4)
Adrenosterone	Hydroxysteroid (11-beta) dehydrogenase 1 (HSD11β1)
Riluzole hydrochloride	Glutamate R andvoltage-dependent Na+ channel
Naftopidil dihydrochloride	5-Hydroxytryptamine receptor 1A (HTR1A) andα1-adrenergic receptor (AR)
S(-)Eticlopride hydrochloride	Dopamine receptor D2 (DRD2)
Racecadotril	Membrane metallo-endopeptidase (MME)
Ipriflavone	Unknown
Flurbiprofen	Cyclooxygenase 1 and 2 (Cox1 and -2)
Zardaverine	Phosphodiesterase III/IV (PDE3/4)
Leflunomide	Dihydroorotate dehydrogenase (DHODH)
Olmesartan	Angiotensin II receptor alpha
Disulfiram	Aldehyde dehydrogenase (ALDH)Dopamine β-hydroxylase (DBH)
Zuclopenthixol dihydrochloride	Dopamine receptors D1 and D2 (DRD1 and -2)

Blocking Dopamine receptor D2 with S(-) Eticlopride hydrochloride suppressed cell motility and invasion of highly metastatic human cancer cells in a dose-dependent manner.

We confirmed that highly metastatic human cancer cell lines expressed target genes through western blotting analyses. Among the genes, serotonin receptor 2C (HTR2C), which is a primary target of pizotifen, was highly expressed in only metastatic cell lines (Figure 2B and Figure 2—figure supplement 2A). Clinical data also shows that that HTR2C expression is correlated with tumor stage of breast cancer patients and is higher in metastatic and Her2/neu-overexpressing tumors (Pai et al., 2009). Pizotifen suppressed cell motility and invasion of several highly metastatic human cancer cell lines in a dose-dependent manner (Figure 2C). Similarly, dopamine receptor D2 (DRD2), which is a primary target of S(-)eticlopride hydrochloride, was highly expressed in only metastatic cell lines, and the drug suppressed cell motility and invasion of these cells in a dose-dependent manner (Figure 2—figure supplement 2A-C).

Figure 2—figure supplement 2.

Pizotifen suppressed metastatic dissemination of MDA-MB-231 and MIA-PaCa2 cells in a zebrafish xenotransplantation model.

(A) Representative images of dissemination of 231R cells in zebrafish xenotransplantation model. The fish larvae that were inoculated with 231R cells were treated with either vehicle (top left, bottom left) or pizotifen (top right, bottom right). (B) Representative images of dissemination of MIA-PaCa2 cells in zebrafish xenotransplantation model. The fish were inoculated with MIA-PaCa2 cells, and treated with either vehicle (top left) or drug (lower left). White arrow heads indicate disseminated MIA-PaCa2 cells. The images were shown in 4× magnification. Scale bar, 100 μm. The mean frequencies of the fish showing head, trunk, or end-tail dissemination were tabulated (right). Each value is indicated as the mean ± SEM of two independent experiments. Statistical analysis was determined by Student’s t test.

These results indicate that a number of the epiboly-interrupting drugs also have suppressor effects on cell motility and invasion of highly metastatic human cancer cells.

Pizotifen suppressed metastatic dissemination of human cancer cells in a zebrafish xenotransplantation model

While a number of the epiboly-interrupting drugs suppressed cell motility and invasion of human cell lines in vitro, it was still unclear whether the drugs could suppress metastatic dissemination of cancer cells in vivo. Therefore, we examined whether the identified drugs could suppress metastatic dissemination of these human cancer cells in a zebrafish xenotransplantation model. Pizotifen was selected to test since HTR2C was overexpressed only in highly metastatic cell lines supporting the hypothesis that it could be a novel target for blocking metastatic dissemination of cancer cells (Figure 2B). Red fluorescent protein (RFP)-labelled MDA-MB-231 (231R) cells were injected into the duct of Cuvier of Tg (kdrl:eGFP) zebrafish at 2 dpf and then maintained in the presence of either vehicle or pizotifen. Twenty-four hours post injection, the numbers of fish showing metastatic dissemination of 231R cells were measured via fluorescence microscopy. In this model, the dissemination patterns were generally divided into three categories: (i) head dissemination, in which disseminated 231R cells exist in the vessel of the head part; (ii) trunk dissemination, in which the cells were observed in the vessel dilating from the trunk to the tail; (iii) end-tail dissemination, in which the cells were observed in the vessel of the end-tail part (Nakayama et al., 2020).

Three independent experiments revealed that the frequencies of fish in the drug-treated group showing head, trunk, or end-tail dissemination significantly decreased to 55.3% ± 7.5%, 28.5 ± 5.0%, or 43.5% ± 19.1% when compared with those in the vehicle-treated group; 95.8% ± 5.8%, 47.1 ± 7.7%, or 82.6% ± 12.7%. Conversely, the frequency of the fish in the drug-treated group not showing any dissemination significantly increased to 45.4% ± 0.5% when compared with those in the vehicle-treated group; 2.0% ± 2.9% (Figure 2D, Figure 2—figure supplement 2 and Table 4).

Table 4.

Effects of pharmacological inhibition of HTR2C on metastatic dissemination of MDA-MB-231 cells in zebrafish xenografted models.

Related to Figure 2D. The numbers and frequencies of the fish showing the dissemination patterns in vehicle- or pizotifen-treated group were indicated. The fish showed both patterns of dissemination were redundantly counted in this analysis.

		Experiment_#1	Experiment_#2	Experiment_#3	Average of experiments
Drug: VehicleCell: MDA-MB-231	Non-dissemination	0% n1 = 0/17	0% n2 = 0/12	6.66% n3 = 1/15	2.22% ± 3.84%
	Head	58.82% n1 = 10/17	91.66% n2 = 11/12	6.66% n3 = 1/15	72.38% ± 17.15%
	Trunk	52.94% n1 = 9/17	8.33% n2 = 1/12	20% n3 = 2/15	27.09% ± 23.13%
	End-tail	100% n1 = 17/17	100% n2 = 12/12	86.66% n3 = 13/15	95.55% ± 7.69%
Drug: PizotifenCell: MDA-MB-231	Non-dissemination	55% n1 = 11/20	31.57% n2 = 6/19	45.45 % n3 = 10/22	44.01% ± 11.77%
	Head	5% n1 = 1/20	31.57% n2 = 6/19	18.18% n3 = 4/22	18.25% ± 13.28%
	Trunk	5% n1 = 1/20	10.52% n2 = 2/19	4.45% n3 = 1/22	6.69% ± 3.32%
	End-tail	45% n1 = 9/20	57.89% n2 = 11/19	50% n3 = 11/22	50.96% ± 6.50%

Similar effects were observed in another xenograft experiments using an RFP-labelled human pancreatic cancer cell line, MIA-PaCa-2 (MP2R). In the drug-treated group, the frequencies of the fish showing head, trunk, or end-tail dissemination significantly decreased to 15.3% ± 6.7%, 6.2% ± 1.3%, or 41.1% ± 1.5%; conversely, the frequency of the fish not showing any dissemination significantly increased to 46.3% ± 8.9% when compared with those in the vehicle-treated group; 74.5% ± 11.1%, 18.9% ± 14.9%, 77.0% ± 9.0%, or 17.2% ± 0.7% (Figure 2—figure supplement 2A and Table 5).

Table 5.

Effects of pharmacological inhibition of HTR2C on metastatic dissemination of Mia-PaCa2 cells in zebrafish xenografted models.

Related to Figure 4. The numbers and frequencies of the fish showing the dissemination patterns in vehicle- or pizotifen-treated group were indicated. The fish showed both patterns of dissemination were redundantly counted in this analysis.

		Experiment_#1	Experiment_#2	Average of experiments
Drug: VehicleCell: MIA-PaCa2	Non-dissemination	17.64% n1 = 3/17	16.66% n2 = 2/12	17.15% + 0.69%
	Head	82.35% n1 = 14/17	66.66% n2 = 8/12	74.50% + 11.09%
	Trunk	29.41% n1 = 5/17	8.33% n2 = 1/12	18.87% + 14.90%
	End-tail	70.58% n1 = 12/17	83.33% n2 = 10/17	76.96% + 9.01
Drug: PizotifenCell: MIA-PaCa2	Non-dissemination	40% n1 = 4/10	52.63% n2 = 10/19	46.31% + 8.93%
	Head	20% n1 = 2/10	10.52% n2 = 2/19	15.26% + 6.69%
	Trunk	10% n1 = 1/10	5.26% n2 = 1/19	7.63% + 3.34%
	End-tail	40% n1 = 4/10	42.05% n2 = 8/19	41.4% + 1.48%

Figure 4.

HTR2C induced epithelial-to-mesenchymal transition (EMT)-mediated metastatic dissemination of human cancer cells.

(A) The morphologies of the MCF7 and HaCaT cells expressing either the control vector or HTR2C were revealed by phase contrast microscopy. (B) Immunofluorescence staining of E-cadherin, EpCAM, vimentin, and N-cadherin expressions in the MCF7 cells from A. (C) Expression of E-cadherin, EpCAM, vimentin, N-cadherin, and HTR2C was examined by western blotting in the MCF7 and HaCaT cells; GAPDH loading control is shown (bottom). (D) Effect of HTR2C on cell motility and invasion of MCF7 cells. MCF7 cells were subjected to Boyden chamber assays. Fetal bovine serum (1% v/v) was used as the chemoattractant in both assays. Each experiment was performed at least twice. (E) Representative images of dissemination patterns of MCF7 cells expressing either the control vector (top left) or HTR2C (lower left) in a zebrafish xenotransplantation model. White arrow heads indicate disseminated MCF7 cells. The mean frequencies of the fish showing head, trunk, or end-tail dissemination tabulated (right). Each value is indicated as the mean ± SEM of two independent experiments. Statistical analysis was determined by Student’s t test.

(A) Quantification analyses of western blotting bands in Figure 4C. The analyses were performed by ImageJ. Signal strength of bands of E-cadherin, EpCAM, N-cadherin, vimentin, ZEB1, and HTR2C was normalized by that of GAPDH. (B) Representative images of dissemination patterns of MCF7 cells expressing either the control vector (top left, middle left, bottom left) or HTR2C (top right, middle right, bottom right) in a zebrafish xenotransplantation model.

To eliminate the possibility that the metastasis suppressing effects of pizotifen might result from off-target effects of the drug, we conducted validation experiments to determine whether knockdown of HTR2C would show the same effects. Sub-clones of 231R cells that expressed short hairpin RNA (shRNA) targeting either LacZ or HTR2C were injected into the fish at 2 dpf and the fish were maintained in the absence of drug. In the fish that were inoculated with shHTR2C 231R cells, the frequencies of the fish showing head, trunk, and end-tail dissemination significantly decreased to 6.7% ± 4.9%, 6.7% ± 0.7%, or 20.0% ± 16.5%; conversely, the frequency of the fish not showing any dissemination significantly increased to 80.0% ± 4.4% when compared with those that were inoculated with shLacZ 231R cells; 80.0% ± 27.1%, 20.0% ± 4.5%, 90.0% ± 7.7%, or 0% (Figure 2E and Table 6).

Table 6.

Effects of genetic inhibition of HTR2C on metastatic dissemination of MDA-MB-231 cells in zebrafish xenografted models.

Related to Figure 2E. The numbers and frequencies of the fish showing the dissemination patterns in the zebrafish that were inoculated with either shLacZ or shHTR2C MDA-MB-231 cells were indicated. The fish showed both patterns of dissemination were redundantly counted in this analysis.

		Experiment_#1	Experiment_#2	Average of experiments
shLacZ	Non-dissemination	0% n1 = 0/10	0% n2 = 0/10	0%
	Head	60% n1 = 6/10	100% n2 = 10/10	80%± 28.28%
	Trunk	30% n1 = 3/10	10% n2 = 1/10	20% ± 14.14%
	End-tail	80% n1 = 8/10	100% n2 = 10/10	90% ± 14.14
shHTR2C	Non-dissemination	80% n1 = 12/15	76.84% n2 = 14/19	76.84 ± 4.46%
	Head	6.66% n1 = 1/15	15.78% n2 = 3/19	11.22% ± 6.45%
	Trunk	6.66% n1 = 1/15	5.26% n2 = 1/19	5.96% ± 0.99%
	End-tail	20% n1 = 3/15	26.31% n2 = 5/19	23.15% ± 4.46%

These results indicate that pharmacological and genetic inhibition of HTR2C suppressed metastatic dissemination of human cancer cells in vivo.

Pizotifen suppressed metastasis progression of a mouse model of metastasis

We examined the metastasis-suppressor effect of pizotifen in a mouse model of metastasis (Tao et al., 2008). Luciferase-expressing 4T1 murine mammary carcinoma cells were inoculated into the mammary fat pads (MFP) of female BALB/c mice. On day 2 post inoculation, the mice were randomly assigned to two groups and one group received once daily intraperitoneal injections of 10 mg/kg pizotifen while the other group received a vehicle injection. Bioluminescence imaging and tumor measurement revealed that the sizes of the primary tumors in pizotifen-treated mice were equal to those in the vehicle-treated mice on day 10 post inoculation. The primary tumors were resected after the analyses. Immunofluorescence (IF) staining also demonstrated that the percentage of Ki67-positive cells in the resected primary tumors of pizotifen-treated mice were the same as those of vehicle-treated mice (Figure 3A–C), additionally, both groups showed less than 1% cleaved caspase 3 positive cells (Figure 3—figure supplement 1). Therefore, no anti-tumor effect of pizotifen was observed on the primary tumor. After 70 days from inoculation, bioluminescence imaging detected light emitted in the lungs, livers, and lymph nodes of vehicle-treated mice but not those of pizotifen-treated mice (Figure 3C). Vehicle-treated mice formed 5–50 metastatic nodules per lung in all 10 mice analyzed; conversely, pizotifen-treated mice (n = 10) formed 0–5 nodules per lung in all 10 mice analyzed (Figure 3D). Histological analyses confirmed that metastatic lesions in the lungs were detected in all vehicle-treated mice; conversely, they were detected in only 2 of 10 pizotifen-treated mice and the rest of the mice showed metastatic colony formations around the bronchiole of the lung. In addition, 4 of 10 vehicle-treated mice exhibited metastasis in the liver and the rest showed metastatic colony formation around the portal tract of the liver. In contrast, none of 10 pizotifen-treated mice showed liver metastases and only half of the 10 mice showed metastatic colony formation around the portal tract (Figure 3E). These results indicate that pizotifen can suppress metastasis progression without affecting primary tumor growth.

Figure 3.

Pizotifen suppressed metastatic progression in a mouse model of metastasis.

(A) Mean volumes (n = 10 per group) of 4T1 primary tumors formed in the mammary fat pad of either vehicle- or pizotifen-treated mice at day 10 post injection. (B) Ki67 expression level in 4T1 primary tumors formed in the mammary fat pad of either vehicle- or pizotifen-treated mice at day 10 post injection. The mean expression levels of Ki67 (n = 10 mice per group) were determined and were calculated as the mean ration of Ki67-positive cells to 4’,6-diamidino-2-phenylindole (DAPI) area. (C) Representative images of primary tumors on day 10 post injection (top panels) and metastatic burden on day 70 post injection (bottom panels) taken using an IVIS Imaging System. (D) Representative images of the lungs from either vehicle- (top) or pizotifen-treated mice (bottom) at 70 days post tumor inoculation. Number of metastatic nodules in the lung of either vehicle- or pizotifen-treated mice (right). (E) Representative hematoxylin and eosin (H&E) staining of the lung (top) and liver (bottom) from either vehicle- or pizotifen-treated mice. Black arrow heads indicate metastatic 4T1 cells. (F) The mean number of metastatic lesions in step sections of the lungs from the mice that were inoculated with 4T1-12B cells expressing short hairpin RNA (shRNA) targeting for either LacZ or HTR2C. (G) Representative H&E staining of the lung and liver from the mice that were inoculated with 4T1-12B cells expressing shRNA targeting for either LacZ or HTR2C. Black arrow heads indicate metastatic 4T1 cells. Each value is indicated as the mean ± SEM. Statistical analysis was determined by Student’s t test.

Figure 3—figure supplement 1.

Cleaved caspase 3 expression level in 4T1 primary tumors formed in the mammary fat pad of either vehicle- or pizotifen-treated mice at day 10 post injection.

To eliminate the possibility that the metastasis suppressing effects of pizotifen might result from off-target effects, we conducted validation experiments to determine whether knockdown of HTR2C would show the same effects. The basic experimental process followed the experimental design described above except that sub-clones of 4T1 cells that expressed shRNA targeting either LacZ or HTR2C were injected into the MFP of female BALB/c mice and the mice were maintained without drug. Histological analyses revealed that all of the mice (n = 5) that were inoculated with 4T1 cells expressing shRNA targeting LacZ showed metastases in the lungs. The mean number of metastatic lesions in a lung was 26.4 ± 7.8. In contrast, only one of the mice (n = 5) were inoculated with 4T1 cells expressing shRNA targeting HTR2C showed metastases in the lungs and the rest of the mice showed metastatic colony formation around the bronchiole of the lung. The mean number of metastatic lesions in the lung significantly decreased to 10% of those of mice that were inoculated with 4T1 cells expressing shRNA targeting LacZ (Figure 3F–H).

Taken together, pharmacological and genetic inhibition of HTR2C showed an anti-metastatic effect in the 4T1 model system.

HTR2C promoted EMT-mediated metastatic dissemination of human cancer cells

Although pharmacological and genetic inhibition of HTR2C inhibited metastasis progression, a role for HTR2C on metastatic progression has not been reported. Therefore, we examined whether HTR2C could confer metastatic properties on poorly metastatic cells.

First, we established a stable sub-clone of MCF7 human breast cancer cells expressing either vector control or HTR2C. Vector control expressing MCF7 cells maintained highly organized cell-cell adhesion and cell polarity; however, HTR2C-expressing MCF7 cells led to loss of cell-cell contact and cell scattering. The cobblestone-like appearance of these cells was replaced by a spindle-like, fibroblastic morphology. Western blotting and IF analyses revealed that HTR2C-expressing MCF7 cells showed loss of E-cadherin and EpCAM, and elevated expressions of N-cadherin, vimentin, and an EMT-inducible transcriptional factor ZEB1. Similar effects were validated through another experiment using an immortal keratinocyte cell line, HaCaT cells, in that HTR2C-expressing HaCaT cells also showed loss of cell-cell contact and cell scattering with loss of epithelial markers and gain of mesenchymal markers (Figure 4A–C and Figure 4—figure supplement 1A). Therefore, both the morphological and molecular changes in the HTR2C-expressing MCF7 and HaCaT cells demonstrated that these cells had undergone an EMT.

Figure 4—figure supplement 1.

HTR2C promoted EMT-mediated metastatic dissemination of poorly metastatic human cancer cells in a zebrafish xenotransplantation model.

(A) Quantification analyses of western blotting bands in Figure 4C. The analyses were performed by ImageJ. Signal strength of bands of E-cadherin, EpCAM, N-cadherin, vimentin, ZEB1, and HTR2C was normalized by that of GAPDH. (B) Representative images of dissemination patterns of MCF7 cells expressing either the control vector (top left, middle left, bottom left) or HTR2C (top right, middle right, bottom right) in a zebrafish xenotransplantation model.

Next, we examined whether HTR2C-driven EMT could promote metastatic dissemination of human cancer cells. Boyden chamber assay revealed that HTR2C expressing MCF7 cells showed an increased cell motility and invasion compared with vector control-expressing MCF7 cells in vitro (Figure 4D). Moreover, we conducted in vivo examination of whether HTR2C expression could promote metastatic dissemination of human cancer cells in a zebrafish xenotransplantation model. RFP-labelled MCF7 cells expressing either vector control or HTR2C were injected into the duct of Cuvier of Tg (kdrl:eGFP) zebrafish at 2 dpf. Twenty-four hours post injection, the frequencies of the fish showing metastatic dissemination of the inoculated cells were measured using fluorescence microscopy. In the fish that were inoculated with HTR2C expressing MCF7 cells, the frequencies of the fish showing head, trunk, and end-tail dissemination significantly increased to 96.7% ± 4.7%, 68.8% ± 6.4%, or 89.5% ± 3.4%; conversely, the frequency of the fish not showing any dissemination decreased to 0% when compared with those in the fish that were inoculated with vector control expressing MCF7 cells; 33.1% ± 18.5%, 0%, 56.9% ± 4.4%, or 43% (Figure 4E, Figure 4—figure supplement 1B and Table 7).

Table 7.

Effects of HTR2C overexpression on metastatic dissemination of MCF7 cells in zebrafish xenografted models.

Related to Figure 4E. The numbers and frequencies of the fish showing the dissemination patterns in the zebrafish that were inoculated with MCF7 cells expressing either vector control (VC) or HTR2C were indicated. The fish showed both patterns of dissemination were redundantly counted in this analysis.

		Experiment_#1	Experiment_#2	Average of experiments
VC	Non-dissemination	46.15% n1 = 6/13	40% n2 = 4/10	43.07% ± 4.35%
	Head	46.15% n1 = 6/13	20% n2 = 2/10	33.07% ± 18.49%
	Trunk	0% n1 = 0/13	0% n2 = 0/10	0%
	End-tail	53.84% n1 = 7/13	60% n2 = 6/10	56.92% ± 4.35%
HTR2C	Non-dissemination	0% n1 = 0/14	0% n2 = 0/15	0%
	Head	100% n1 = 14/14	93.33% n2 = 14/15	96.66% ± 4.71%
	Trunk	64.28% n1 = 9/14	73.33% n2 = 11/15	68.80% ± 6.39%
	End-tail	85.71% n1 = 12/14	93.33% n2 = 14/15	89.52% ± 5.38%

These results indicated that HTR2C promoted metastatic dissemination of cancer cells through induction of EMT, and suggest that the screen can easily be converted to a chemical genetic screening platform.

Pizotifen induced mesenchymal-to-epithelial transition through inhibition of Wnt signaling

Finally, we elucidated the mechanism of action of how pizotifen suppressed metastasis, especially metastatic dissemination of cancer cells. Our results showed that HTR2C induced EMT and that pharmacological and genetic inhibition of HTR2C suppressed metastatic dissemination of MDA-MB-231 cells that had already transitioned to mesenchymal-like traits via EMT. Therefore, we speculated that blocking HTR2C with pizotifen might inhibit the molecular mechanisms which follow EMT induction. We first investigated the expressions of epithelial and mesenchymal markers in pizotifen-treated MDA-MB-231 cells since the activation of an EMT program needs to be transient and reversible, and transition from a fully mesenchymal phenotype to a epithelial-mesenchymal hybrid state or a fully epithelial phenotype is associated with malignant phenotypes (Kröger et al., 2019). IF and FACS analyses revealed 20% of pizotifen-treated MDA-MB-231 cells restored E-cadherin expression. Also, western blotting analysis demonstrated that 4T1 primary tumors from pizotifen-treated mice has elevated E-cadherin expression compared with tumors from vehicle-treated mice (Figure 5A–C and Figure 5—figure supplement 1). However, mesenchymal markers did not change between vehicle and pizotifen-treated MDA-MB-231 cells (data not shown). We further analyzed E-cadherin-positive (E-cad+) cells in pizotifen-treated MDA-MB-231 cells. The E-cad+ cells showed elevated expressions of epithelial markers KRT14 and KRT19; and decreased expression of mesenchymal makers vimentin, MMP1, MMP3, and S100A4. Recent research reports that an EMT program needs to be transient and reversible and that a mesenchymal phenotype in cancer cells is achieved by constitutive ectopic expression of ZEB1. In accordance with the research, the E-cad+ cells and 4T1 primary tumors from pizotifen-treated mice had decreased ZEB1 expression compared with vehicle-treated cells and tumors from vehicle-treated mice (Figure 5D and Figure 5—figure supplement 2). In contrast, HTR2C-expressing MCF7 and HuMEC cells expressed ZEB1 but not vehicle control MCF7 and HuMEC cells (Figure 4C and Figure 5—figure supplement 3). HTR2C-expressing MCF7 cells expressed not only ZEB1 but also Twist1 and Snail. In contrast, pizotifen-treated MDA-MB-231 cells showed decreased expression of ZEB1 and Twist1 compared with that in vehicle-treated cells. Furthermore, in the primary tumors of pizotifen-treated mice, only ZEB1 expression was decreased compared with those of vehicle-treated mice. These results indicate that HTR2C-mediated signaling induced EMT through up-regulation of ZEB1 and blocking HTR2C with pizotifen induced mesenchymal-to-epithelial transition through down-regulation of ZEB1 (Figure 5—figure supplement 4).

Figure 5.

Pizotifen restored mesenchymal-like traits of MDA-MB-231 cells into epithelial traits through blocking nuclear accumulation of β-catenin.

(A) Immunofluorescence (IF) staining of E-cadherin in either vehicle- or pizotifen-treated MDA-MB-231 cells. (B) Surface expression of E-cadherin in either vehicle (black)- or pizotifen (red)-treated MDA-MB-231 cells by FACS analysis. Non-stained controls are shown in gray. (C) Protein expressions levels of E-cadherin, ZEB1, and β-catenin in the cytoplasm and nucleus of 4T1 primary tumors from either vehicle- or pizotifen-treated mice are shown; Luciferase, histone H3, and β-tubulin are used as loading control for whole cell, nuclear, or cytoplasmic lysate, respectively. (D) Protein expression levels of epithelial and mesenchymal markers and ZEB1 in either vehicle- or pizotifen-treated MDA-MB-231 cells or E-cadherin-positive (E-cad+) cells in pizotifen-treated MDA-MB-231 cells are shown. (E) IF staining of β-catenin in the MCF7 cells expressing either vector control (top left, bottom left) or HTR2C (top right, bottom right). (F) Expressions of β-catenin in the cytoplasm and nucleus of MCF7 cells. (G) IF staining of β-catenin in either vehicle (top left, bottom left) or pizotifen-treated MDA-MB-231 cells (top right, bottom right). (H) Expressions of β-catenin in the cytoplasm and nucleus of MDA-MB-231 cells and the E-cad+ cells.

The analyses were performed by ImageJ. Signal strength of bands of E-cadherin, EpCAM, p-GSK-3β, GSK-3β were normalized by that of luciferase. Signal strength of bands of nuclear and cytoplasmic β-catenin was normalized by that of histone H3 and β-tubulin, respectively.

The analyses were performed by ImageJ. Signal strength of bands of E-cadherin, EpCAM, KRT18, KRT19, MMP1, MMP3, vimentin, and S100A4 was normalized by that of GAPDH.

The analyses were performed by ImageJ. Signal strength of bands of nuclear and cytoplasmic β-catenin was normalized by that of histone H3 and β-tubulin, respectively. Signal strength of bands of p-GSK-3β and GSK-3β was normalized by that of GAPDH.

Protein expression levels of Twist1 in either vehicle- or pizotifen-treated MDA-MB-231 cells are shown (middle): GAPDH loading control is shown (bottom). Protein expression levels of Snail and Twist1 of 4T1 primary tumors from either vehicle- or pizotifen-treated mice are shown (right); luciferase is used as loading control for whole cell. Luciferase control was obtained in the same experiment from Figure 5C.

The analyses were performed by ImageJ. Signal strength of bands of nuclear and cytoplasmic β-catenin was normalized by that of histone H3 and β-tubulin, respectively. Signal strength of bands of p-GSK-3β and GSK-3β was normalized by that of GAPDH.

Figure 5—figure supplement 1.

Quantification analyses of western blotting bands in Figure 5C.

The analyses were performed by ImageJ. Signal strength of bands of E-cadherin, EpCAM, p-GSK-3β, GSK-3β were normalized by that of luciferase. Signal strength of bands of nuclear and cytoplasmic β-catenin was normalized by that of histone H3 and β-tubulin, respectively.

Figure 5—figure supplement 2.

Quantification analyses of western blotting bands in Figure 5D.

The analyses were performed by ImageJ. Signal strength of bands of E-cadherin, EpCAM, KRT18, KRT19, MMP1, MMP3, vimentin, and S100A4 was normalized by that of GAPDH.

Figure 5—figure supplement 3.

Quantification analyses of western blotting bands in Figure 5F.

The analyses were performed by ImageJ. Signal strength of bands of nuclear and cytoplasmic β-catenin was normalized by that of histone H3 and β-tubulin, respectively. Signal strength of bands of p-GSK-3β and GSK-3β was normalized by that of GAPDH.

Figure 5—figure supplement 4.

Expression of Snail and Twist1 was examined by western blotting in the MCF7 cells (left); GAPDH loading control is shown (bottom).

Protein expression levels of Twist1 in either vehicle- or pizotifen-treated MDA-MB-231 cells are shown (middle): GAPDH loading control is shown (bottom). Protein expression levels of Snail and Twist1 of 4T1 primary tumors from either vehicle- or pizotifen-treated mice are shown (right); luciferase is used as loading control for whole cell. Luciferase control was obtained in the same experiment from Figure 5C.

We further investigated the mechanism of action of how blocking HTR2C with pizotifen induced down-regulation of ZEB1. In embryogenesis, serotonin-mediated signaling is required for Wnt-dependent specification of the superficial mesoderm during gastrulation (Beyer et al., 2012). Wnt signaling plays critical role in inducing EMT. In cancer cells, overexpression of HTR1D is associated with Wnt signaling (Sui et al., 2015; Zhan et al., 2017). This evidence led to a hypothesis that HTR2C-mediated signaling might turn on transcriptional activity of β-catenin and that might induce up-regulation of EMT-TFs. IF analyses revealed β-catenin was accumulated in the nucleus of HTR2C-expressing MCF7 cells but it was located in the cytoplasm of vector control-expressing cells (Figure 5E). Nuclear accumulation of β-catenin in HTR2C-expressing MCF7 cells was confirmed by western blot (Figure 5F and Figure 5—figure supplement 2). In contrast, pizotifen-treated MDA-MB-231 cells showed β-catenin located in the cytoplasm of the cells. Vehicle-treated cells showed that β-catenin accumulated in the nucleus of the cells. (Figure 5G), and western blotting analysis confirmed that it was located in the cytoplasm of pizotifen-treated MDA-MB-231 cells (Figure 5H and Figure 5—figure supplement 5). Furthermore, immunohistochemistry and western blotting analyses showed that β-catenin accumulated in the nucleus, and phospho-GSKβ and ZEB1 expression were decreased in 4T1 primary tumors from pizotifen-treated mice compared with vehicle-treated mice (Figure 5C and Figure 5—figure supplement 1). These results indicated that HTR2C would regulate transcriptional activity of β-catenin and pizotifen could inhibit it.

Figure 5—figure supplement 5.

Quantification analyses of western blotting bands in Figure 5H.

The analyses were performed by ImageJ. Signal strength of bands of nuclear and cytoplasmic β-catenin was normalized by that of histone H3 and β-tubulin, respectively. Signal strength of bands of p-GSK-3β and GSK-3β was normalized by that of GAPDH.

Taken together, we conclude that blocking HTR2C with pizotifen restored epithelial properties to metastatic cells (MDA-MB-231 and 4T1 cells) through a decrease of transcriptional activity of β-catenin and that suppressed metastatic progression of the cells.

Discussion

Reducing or eliminating mortality associated with metastatic disease is a key goal of medical oncology, but few models exist that allow for rapid, effective screening of novel compounds that target the metastatic dissemination of cancer cells. Based on accumulated evidence that at least 50 genes play an essential role in governing both metastasis and gastrulation progression (Table 1), we hypothesized that small molecule inhibitors that interrupt gastrulation of zebrafish embryos might suppress metastatic progression of human cancer cells. We created a unique screening concept utilizing gastrulation of zebrafish embryos to test the hypothesis. Our results clearly confirmed our hypothesis: 25.6% (20/76 drugs) of epiboly-interrupting drugs could also suppress cell motility and invasion of highly metastatic human cell lines in vitro. In particular, pizotifen, which is an antagonist for serotonin receptor 2C and one of the epiboly-interrupting drugs, could suppress metastasis in a mouse model (Figure 3A–E). Thus, this screen could offer a novel platform for discovery of anti-metastasis drugs.

Among the 20 drugs which suppressed both epiboly progression and cell motility and invasion of MDA-MB-231 cells, hexachlorophene and troglitazone showed the strongest effect on suppressing cell motility and invasion of MDA-MB-231 cells. However, the drug could not suppress cell motility and invasion of other highly metastatic human cancer cell lines: MDA-MB-435 and PC3. With the exception of pizotifen and S(-)eticlopride hydrochloride, the remaining 18 drugs could not show the suppressor effect on more than three highly metastatic human cancer cell lines. These results indicate that the strength of interrupting effect of a drug on epiboly progression is not proportional to the strength of suppressing effect of the drug on metastasis.

We have provided the first evidence that HTR2C, which is a primary target of pizotifen, induced EMT and promoted metastatic dissemination of cancer cells (Figure 4A–E). Clinical data shows that HTR2C expression is correlated with tumor stage of breast cancer patients and is higher in metastatic and Her2/neu-overexpressing tumors (Pai et al., 2009). That would support our finding.

Pharmacological inhibition of DRD2 with S(-)eticlopride hydrochloride suppressed cell invasion and migration of multiple human cancer cell lines in vitro. However, overexpression of DRD2 could not induce EMT on MCF7 cells. Therefore, we stopped focusing on DRD2 and S(-)eticlopride hydrochloride.

There are at least two advantages to the screen described herein. One is that the screen can easily be converted to a chemical genetic screening platform. Indeed, our screen succeeded to identify HTR2C as an EMT inducer (Figure 4A–E). In this research, 1280 FDA approval drugs were screened, this is less than a few percent of all of druggable targets (approximately 100 targets) in the human proteome in the body. If chemical genetic screening using specific inhibitor libraries were conducted, more genes that contribute to metastasis and gastrulation could be identified. The second advantage is that the screen enables one researcher to test 100 drugs in 5 hr with using optical microscopy, drugs, and zebrafish embryos. That indicates this screen is not only highly efficient, low-cost, and low-labor but also enables researchers who do not have high-throughput screening instruments to conduct drug screening for anti-metastasis drugs.

Materials and methods

Zebrafish embryo screening

Zebrafish embryos at two-cell stage were collected at 20 min after their fertilization. Each drug was added to a well of a 24-well plate containing approximately 20 zebrafish embryos per well in either 10 or 50 μM final concentration when the embryos reached the sphere stage. Chemical treatment was initiated at 4 hpf and approximately 20 embryos were treated with two different concentrations for each compound tested. The treatment was ended at 9 hpf when vehicle- (DMSO) treated embryos as control reach 80–90% completion of the epiboly stage. The compounds which induced delay (<50% epiboly) in epiboly were selected as hit compounds for in vitro testing using highly metastatic human cancer cell lines. The study protocol was approved by the Institutional Animal Care and Use Committee of the National University of Singapore (protocol number: R16-1068).

Reagents

FDA, EMA, and other agencies-approved chemical libraries were purchased from Prestwick Chemical (Illkirch, France). Pizotifen (sc-201143) and S(-)eticlopride hydrochloride (E101) were purchased from Santa Cruz (Dallas, TX) and Sigma-Aldrich (St Louis, MO).

Cell culture and cell viability assay

MCF7, MDA-MB-231, MDA-MB-435, MIA-PaCa2, PC3, SW620, PC9, and HaCaT cells were obtained from American Type Culture Collection (ATCC, Manassas, VA). Luciferase-expressing 4T1 (4T1-12B) cells were provided from Dr Gary Sahagian (Tufts University, Boston, MA). All culture methods followed the supplier’s instruction. Cell viability assay was performed as previously described (Nakayama et al., 2020). PCR-based mycoplasma testing on these cells was performed once in 4 months.

Plasmid

A DNA fragment coding for HTR2C was amplified by PCR with primers containing restriction enzyme recognition sequences. The HTR2C coding fragment was amplified from hsp70l:mCherry-T2A-CreERT2 plasmid (Huang et al., 2012).

Immunoblotting

Western blotting was performed as described previously (Nakayama et al., 2020). Raw data of images of western blotting analyses are uploaded as source data for western. Anti-PRMT1 (A33), anti-CYP11A1 (D8F4F), anti-E-cadherin (4A2), anti-EpCAM (VU1D9), anti-vimentin (D21H3), anti-N-cadherin (D4R1H), anti-ZEB1 (D80D3), anti-histone H3 (D1H2), anti-β-tubulin (9F3), and anti-GAPDH (14C10) antibodies were purchased from Cell Signaling Technology (Danvers, MA). Anti-HTR2C (ab133570) and anti-DRD2 (ab85367) antibodies were purchased form Abcam (Cambridge, UK). Anti-phospho-GSK3β (Ser9) (F-2), anti-GSK3β (1F7), anti-KRT18 (DC-10), anti-KRT19 (A53-B/A2), anti-MMP1 (3B6), anti-MMP2 (8B4), anti-S100A4 (A-7), anti-luciferase (C-12), anti-ki67 (ki-67), and anti-β-catenin (E-5) antibodies were purchased from Santa Cruz Biotechnology (Dallas, TX).

Flow cytometry

Cells were stained with FITC-conjugated E-cadherin antibody (67A4, Biolegend, San Diego, CA). Flow cytometry was performed as described (Nakayama et al., 2009) and analyzed with FlowJo software (TreeStar, Ashland, OR).

shRNA-mediated gene knockdown

The shRNA-expressing lentivirus vectors were constructed using pLVX-shRNA1 vector (632177, TAKARA Bio, Shiga, Japan). PRMT1-shRNA_#3-targeting sequence is GTGTTCCAGTATCTCTGATTA; PRMT1-shRNA_#4-targeting sequence is TTGACTCCTACGCACACTTTG. CYP11A1-shRNA_#4-targeting sequence is GCGATTCATTGATGCCATCTA; CYP11A1-shRNA_#4-targeting sequence is GAAATCCAACACCTCAGCGAT. Human HTR2C-shRNA-targeting sequence is TCATGCACCTCTGCGCTATAT. Mouse HTR2C-shRNA-targeting sequence is CTTCATACCGCTGACGATTAT. LacZ-shRNA-targeting sequence is CTACACAAATCAGCGATT.

Immunofluorescence

IF microscopy assay was performed as previously described (Nakayama et al., 2020). Goat anti-mouse and goat anti-rabbit immunoglobulin G (IgG) antibodies conjugated to Alexa Fluor 488 (A-11029 and A-11034, Life Technologies, Carlsbad, CA) and diluted at 1:100 were used. Nuclei were visualized by the addition of 2 μg/ml of 4’,6-diamidino-2-phenylindole (DAPI) (62248, Thermo Fisher, Waltham, MA) and photographed at 100× magnification by a fluorescent microscope BZ-X700 (KEYENCE, Osaka, Japan).

Boyden chamber cell motility and invasion assay

These assays were performed as previously described (Nakayama et al., 2020). In Boyden chamber assay, either 3 × 105 MDA-MB-231, 1 × 106 MDA-MB-435 or 5 × 105 PC9 cells were applied to each well in the upper chamber.

Zebrafish xenotransplantation model

Tg(kdrl:eGFP) zebrafish was provided by Dr Stainier (Max Planck Institute for Heart and Lung Research). Embryos that were derived from the line were maintained in E3 medium containing 200 μM 1-phenyl-2-thiourea (P7629, Sigma-Aldrich, St Louis, MO). Approximately 100–400 RFP-labelled MBA-MB-231 or MIA-PaCa2 cells were injected into the duct of Cuvier of the zebrafish at 2 dpf. The fish were randomly assigned to two groups. One group was maintained in the presence of pizotifen-containing E3 medium and the other group was maintained in vehicle-containing E3 medium.

Spontaneous metastasis mouse model

4T1-12B cells (2 × 104) were injected into the #4 MFP while the mice were anesthetized. To monitor tumor growth and metastases, mice were imaged biweekly by IVIS Imaging System (ParkinElmer, Waltham, MA). The primary tumor was resected 10 days after inoculation. D-Luciferin Potassium Salt (LUCK-100) was purchased from GoldBio (St Louis, MO). The study protocol (protocol number: BRC IACUC #110612) was approved by A*STAR (Agency for Science, Technology and Research, Singapore).

Gene set enrichment analysis

Gene expression profiles obtained from zebrafish embryos at either 50%-epiboly, shield, or 75%-epiboly stage were analyzed based on the hallmark gene sets derived from the Molecular Signatures Database (MSigDB) (Subramanian et al., 2005; Liberzon et al., 2015). The zebrafish transcriptomic data was sourced from White et al., 2017. Gene sets that were significantly enriched (FDR < 0.25) were presented with the normalized enrichment score (NES) and nominal p value. Source data files for analysis of either gene expression and enriched pathways are uploaded as GSEA Source data 1 and 2, respectively.

Histological analysis

All OCT-embedded primary tumors, lungs, and livers of mice from the spontaneous metastasis 4T1 model were sectioned on a cryostat. Eight micron sections were taken at 500 µm intervals through the entirety of the livers and lungs. Sections were subsequently stained with hematoxylin and eosin. Metastatic lesions were counted under a microscope in each section for both lungs and livers.

Statistics

Data were analyzed by Student’s t test; p < 0.05 was considered significant.

We are so impressed with this new and ambitious concept for chemical screening using zebrafish embryos to find a novel anti-metastasis drug, Pizotifen. We hope many researchers will use this screening system for anti-cancer drug discovery.

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "A chemical screen based on an interruption of zebrafish gastrulation identifies the HTR2C inhibitor Pizotifen as a suppressor of EMT-mediated metastasis" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Didier Stainier as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Three reviewers basically agree that your paper should be published in eLife. And I also think it is a very unique and interesting paper suitable for eLife. However, two reviewers think that additional data are necessary to strengthen your conclusions. I would be very grateful if you could perform these experiments in order to publish your work with more impact.

Comprehensive analysis using sequencing data is needed to support the concept that human cancer metastasis mimic/recapitulate zebrafish gastrulation. In addition, Pizotifen's functional part needs additional experiments and more supportive data. Details are in the recommendations to the authors.

Reviewer #1 (Recommendations for the authors):

The authors need to provide a detailed catalog numbers of the reagents used in this study.

Reviewer #2 (Recommendations for the authors):

1. It is not clear how the authors narrowed down the list from 66 hits to 22 compounds and further why Pizotifen was selected for further study. Can the authors provide additional information or data to justify exclusion of hits. Based on the data the compounds that effect dopamine based signaling would have been a more obvious pathway to study (4/22 hits).

2. The selection of human cancer cell lines to validate the findings are not clear. Is HTR2C relevant only to the subset of cancer cell lines studied. Analyses of TCGA or other data sets to show relevance of HTR2C gene expression and survival and or metastasis would greatly strengthen the rationale for picking these cell lines. Is HTR2C relevant for metastasis in all cancers or only in breast cancer. Also, it appears in figure 2B that MCF non metastatic breast cancer cells express higher levels of HTR2C than metastatic skin (MDA-MB-435) and lung (PC9) cells.

3. Some of the data provided does not support the conclusions drawn. For example, In figure 3 C. The luciferase images clearly show that Pizotifen treated mice have smaller tumors while the data in 3 A and B do not support this data. The measurements of tumor volume appear to be taken too early to actually be reproducible, also the data for proliferation and apoptosis was not shown. The authors should provide more data to justify the conclusions drawn.

4. Figure 5 A, 5B, 5G. The effects shown in the texts are not clearly visible. Can the authors provide higher magnification (40X) images to show differences noted in the text.

5. In figure 5D or 5H the effect on Pizotifen on ZEB1 or β-catenin reduction is not observed. Does this suggest this compound is effecting EMT and metastasis by other mechanisms?

6. It is not clear if HTR2C selectively only inhibits ZEB1 mediated metastases or are its effects more general i.e. its expression also effects other EMT regulators for example SNAIL, SLUG and or TWIST expression. Showing the effect of Pizotifen/ HTR2C on other EMT regulators especially the ones above is recommended.

Reviewer #3 (Recommendations for the authors):

1. Line 142: Mentioned DRD2 but did not elaborate on it and explain why it wasn't chosen as a target. In addition, please provide the rationale, why authors forgo other drugs, including the two with the strongest effect, Hexachloroghene and Troglitazone, to focus on Pizotifen.

2. Line 307, please indicate: in which cancer cells is overexpression of HTR1D associated with Wnt-signaling that enables induction of EMT?

3. Figure 5E and 5G: the figures do not show convincing nuclear accumulation of ß-catenin. Please show zoom in insert figures.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "A chemical screen based on an interruption of zebrafish gastrulation identifies the HTR2C inhibitor Pizotifen as a suppressor of EMT-mediated metastasis" for further consideration by eLife. Your revised article has been reviewed by 3 peer reviewers and the evaluation has been overseen by Didier Stainier as the Senior Editor, and a Reviewing Editor.

We are impressed with Dr. Nakayama's manuscript with a new and ambitious concept for the chemical screening for cancer metastasis using zebrafish embryos. However, two reviewers have requested that the authors clarify the mechanisms related to EMT. I also agree with their opinions because EMT is one of the most important factors for cancer metastasis and is well-studied already. It is very much expected by other researchers and would strengthen your story.

Reviewer #1 (Recommendations for the authors):

EMT is just one of the numerous steps of both gastrulation and tumor metastasis, as they mentioned that the transcriptomic data for zebrafish embryo development at the epiboly/gastrulation stage are based on the whole embryos which include all other activities and are not specific to EMT. As far as they observe phenotypic changes of whole embryo and focus on the similarity between gastrulation and metastasis in the 1st screening, what they should analyze in RNA-seq data is "global similarity" but not EMT. While they mentioned that this point is not really an objective of this study, I still feel this is one of the most important point that can be resolved, although they eventually find a drug that affects EMT.

Insufficiency of mechanistic parts has not been addressed.

Overall, I didn't feel in the revised manuscript that the authors have addressed many of my concerns, and recommend to resubmit to the other journal.

Reviewer #2 (Recommendations for the authors):

The authors have addressed most of the major comments except for Major comment 6. The authors provide no data (only text below) showing expression of SNAIL, SLUG or TWIST in MCF7 cells with and without Pizotifen, or any data from the mouse xenograft experiments. Since the authors emphasize the role of ZEB1 showing that other EMT factors are not affected or not consistently affected would help. At a minimum the authors should add their statement below to the text.

6. It is not clear if HTR2C selectively only inhibits ZEB1 mediated metastases are its effects more general i.e. its expression also effects other EMT regulators for example SNAIL, SLUG and or TWIST expression. Showing the effect of Pizotifen/ HTR2C on other EMT regulators especially the ones above is recommended.

Author response: HTR2C-expressing MCF7 cells expressed not only ZEB1 but also Twist1 and SNAIL. In contrast, Pizotifen-treated MDA-MB-231 cells showed decreased expression of ZEB1 and Twist1 compared with that in vehicle-treated the cells. In the primary tumors of Pizotifen-treated mice, only ZEB1 expression was decreased compared with those of vehicle-treated mice."

Reviewer #1 (Recommendations for the authors):

The authors need to provide a detailed catalog numbers of the reagents used in this study.

I added catalog numbers of the reagents used in this study.

Reviewer #2 (Recommendations for the authors):

1. It is not clear how the authors narrowed down the list from 66 hits to 22 compounds and further why Pizotifen was selected for further study. Can the authors provide additional information or data to justify exclusion of hits. Based on the data the compounds that effect dopamine based signaling would have been a more obvious pathway to study (4/22 hits).

Zebrafish embryo screen identified 78 drugs as epiboly-interrupting drugs. Sixteen of the 78 drugs strongly affected cell viability at concentrations less than 1µM and were not used in the cell motility experiments. The remaining 62 drugs were assayed in Boyden chamber motility experiments. Twenty of the 62 drugs enabled to inhibit cell motility and invasion of MDAMB-231 cells without effecting cell viability and proceeded to next examination. The 42 drugs were passed over.

Inhibiting Dopamine receptor D2 (DRD2) suppressed metastatic dissemination of human cancer cells in a zebrafish xenografted model. However, overexpression of DRD2 could not induce EMT on MCF7 cells. Therefore, we have stop focusing DRD2. We incorporated this point into discussion part, line 777 to 781.

2. The selection of human cancer cell lines to validate the findings are not clear. Is HTR2C relevant only to the subset of cancer cell lines studied. Analyses of TCGA or other data sets to show relevance of HTR2C gene expression and survival and or metastasis would greatly strengthen the rationale for picking these cell lines.

According to cancer cell line encyclopedia, HTR2C mRNA expression is observed in cell lines which are established from embryo carcinoma, ovarian cancer, brain cancer, neuroblastoma, and myeloma. Ectopic expression of HTR2C is observed in not only the human cancer cell lines used in this study but also other cancer cell lines.

The reason why we select the human cancer cell lines used in this study, is that they are commonly used in an experimental study of metastasis and show metastasis in high frequency when they are xenografted into mice.

Is HTR2C relevant for metastasis in all cancers or only in breast cancer.

Pai et al., Breast Cancer Research 2009 shows that HTR2C expression is correlated with tumor stage of breast cancer patients and is higher in metastatic and Her2/ neu-overexpressing tumors (Vaibhav P Pai et al., 2009). We incorporate this point to main text in line 270-272.

Also, it appears in figure 2B that MCF non metastatic breast cancer cells express higher levels of HTR2C than metastatic skin (MDA-MB-435) and lung (PC9) cells.

Serotonin binds to HTR2C and activates HTR2C-mediated signaling. Biochemically, tryptophan hydroxylase 1 (TPH1) catalyzes amino acid tryptophan and yields serotonin. Serotonin in in vitro condition is provided from either an autocrine production of serotonin in cancer cells or culture media. Figure 5C clearly showed nuclear accumulation of β-catenin and p-GSK-3β were not observed in MCF7 cells. The result suggest serotonin would not exist in culture media and an autocrine production of serotonin would activate HTR2C-mediated signaling. Pai et al., revealed that MDA-MB-231 cells express TPH1 but MCF7 cells not. According to cancer cell line encyclopedia, TPH1 mRNA expressions in MDAMB-435 and PC9 cells is approximately five and two-fold rather than that in MCF7 cells, respectively. Taken together, both of HTR2C expression and an autocrine production of serotonin by TPH1 would need to activate HTR2C-mediated signaling.

3. Some of the data provided does not support the conclusions drawn. For example, In figure 3 C. The luciferase images clearly show that Pizotifen treated mice have smaller tumors while the data in 3 A and B do not support this data. The measurements of tumor volume appear to be taken too early to actually be reproducible, also the data for proliferation and apoptosis was not shown. The authors should provide more data to justify the conclusions drawn.

4T1 tumors grow aggressively in the primary sites and the tumor sizes often exceed the limits that are allowed in most animal protocols. Therefore, in 4T1 metastasis model, the primary tumors are usually resected at day 10 post inoculation. Following the conventional protocol on the model, we resected primary tumors at day 10 post inoculation. The images of top left and top right of Figure 3C were taken before the resection. In the image of bottom left of Figure 3C, bioluminescence emissions are detected around the mammary fat pad of vehicle-treated mice but the emissions might arise from either the lymph node or unresected primary tumor.

We have added data of apoptosis status in primary tumor of either vehicle or Pizotifen-treated mice as Figure S4. Proliferation status in primary tumor of either vehicle or Pizotifen-treated mice are indicated in Figure 3B.

4. Figure 5 A, 5B, 5G. The effects shown in the texts are not clearly visible. Can the authors provide higher magnification (40X) images to show differences noted in the text.

We inserted higher magnification images in Figure 5E and 5G. By using three different methods: immunofluorescence images, FACS and western blotting, we demonstrate Pizotifen recovers E-cadherin expression on a part of MDA-MB-231 cells. Immunofluorescence images in Figure 5A point that a part of Pizotifen-treated MDA-MB-231 cells show Ecadherin expression. Higher magnification image in Figure 5A would have missed the point.

5. In figure 5D or 5H the effect on Pizotifen on ZEB1 or β-catenin reduction is not observed. Does this suggest this compound is effecting EMT and metastasis by other mechanisms?

Quantification analyses of western blotting bands in Figure 5H reveals that b-catenin accumulation in the nucleus of Pizotifen-treated MDA-MB-231 cells decreased compared with those of vehicle-treated ones. Similar analyses in Figure 5D shows ZEB1 expression of Pizotifen-treated MDA-MB-231 cells slightly decreased compared with those of vehicle-treated ones. ZEB1 expression was significantly decreased in E-cadherin positive fraction of Pizotifen-treated MDA-MB-231 cells compared with ZEB1 expression in vehicle-treated ones. Therefore, we conclude Pizotifen affects ZEB1 expression.

6. It is not clear if HTR2C selectively only inhibits ZEB1 mediated metastases or are its effects more general i.e. its expression also effects other EMT regulators for example SNAIL, SLUG and or TWIST expression. Showing the effect of Pizotifen/ HTR2C on other EMT regulators especially the ones above is recommended.

HTR2C-expressing MCF7 cells expressed not only ZEB1 but also Twist1 and SNAIL. In contrast, Pizotifen-treated MDA-MB-231 cells showed decreased expression of ZEB1 and Twist1 compared with that in vehicle-treated the cells. In the primary tumors of Pizotifen-treated mice, only ZEB1 expression was decreased compared with those of vehicle-treated mice.

Reviewer #3 (Recommendations for the authors):

1. Line 142: Mentioned DRD2 but did not elaborate on it and explain why it wasn't chosen as a target.

We incorporated this point to discussion part, line 777-781.

Inhibiting Dopamine receptor D2 (DRD2) suppressed cell invasion and migration of multiple human cancer cell lines in vitro However, overexpression of DRD2 could not induce EMT on MCF7 cells. In contrast, pharmacologic and genetic inhibition of HTR2C suppressed metastatic progression; conversely, overexpression of HTR2C induced EMT. Therefore, we focus on Pizotifen.

In addition, please provide the rationale, why authors forgo other drugs, including the two with the strongest effect, Hexachloroghene and Troglitazone, to focus on Pizotifen.

We incorporated this point to discussion part, line 763-771. We have two reasons why we did not focus on Hexachloroghene. One is that Hexachloroghene strongly inhibited cell invasion and migration of MDA-MB-231 cells but did not affect cell invasion and migration of other cells: MDA-MB-436 and PC9 cells.

Second is that commercial products including Hexachloroghene killed 15 babies in the

United States and 39 babies in France in 1972. Even if we could demonstrate

Hexachloroghene has suppressor effect on metastasis, an issue of the toxic effects might abolish the new discovery. Therefore, we stopped focusing Hexachloroghene.

We have one reason why we did not focus on Troglitazone. Troglitazone has potential high liver toxicity and leads to drug-induced hepatitis. Once troglitazone was approved by FDA, it was withdrawn from the British market in December 1997, from the United States market in 2000, and from the Japanese market soon afterwards. Even if we could demonstrate Troglitazone has suppressor effect on metastasis, an issue of the toxic effects might abolish the new discovery. Therefore, we stopped focusing Troglitazone.

2. Line 307, please indicate: in which cancer cells is overexpression of HTR1D associated with Wnt-signaling that enables induction of EMT?

This paragraph is changed into "Wnt-signaling plays critical role in inducing EMT. In cancer cells, overexpression of HTR1D is associated with Wnt-signaling".

3. Figure 5E and 5G: the figures do not show convincing nuclear accumulation of ß-catenin. Please show zoom in insert figures.

We inserted higher magnification images in Figure 5E and 5G.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

EMT is just one of the numerous steps of both gastrulation and tumor metastasis, as they mentioned that the transcriptomic data for zebrafish embryo development at the epiboly/gastrulation stage are based on the whole embryos which include all other activities and are not specific to EMT. As far as they observe phenotypic changes of whole embryo and focus on the similarity between gastrulation and metastasis in the 1st screening, what they should analyze in RNA-seq data is "global similarity" but not EMT. While they mentioned that this point is not really an objective of this study, I still feel this is one of the most important point that can be resolved, although they eventually find a drug that affects EMT.

We compared the genes expressed in zebrafish gastrulation with the gene related with EMT through performing gene set enrichment analysis (GSEA). The analysis revealed that 50% epiboly, shield and 75%-epiboly stage of zebrafish embryos expressed the genes which promote EMT-mediated metastasis: EMT induction, TGF-β signaling, wnt/β-catenin signaling and Notch signaling (Figure 1—figure supplemental 1). This data demonstrates that a part of genes are commonly expressed between zebrafish gastrulation and EMT-mediated metastasis of human cancer cells.

Reviewer #2 (Recommendations for the authors):

The authors have addressed most of the major comments except for Major comment 6. The authors provide no data (only text below) showing expression of SNAIL, SLUG or TWIST in MCF7 cells with and without Pizotifen, or any data from the mouse xenograft experiments. Since the authors emphasize the role of ZEB1 showing that other EMT factors are not affected or not consistently affected would help. At a minimum the authors should add their statement below to the text.

6. It is not clear if HTR2C selectively only inhibits ZEB1 mediated metastases are its effects more general i.e. its expression also effects other EMT regulators for example SNAIL, SLUG and or TWIST expression. Showing the effect of Pizotifen/ HTR2C on other EMT regulators especially the ones above is recommended.

Author response: HTR2C-expressing MCF7 cells expressed not only ZEB1 but also Twist1 and SNAIL. In contrast, Pizotifen-treated MDA-MB-231 cells showed decreased expression of ZEB1 and Twist1 compared with that in vehicle-treated the cells. In the primary tumors of Pizotifen-treated mice, only ZEB1 expression was decreased compared with those of vehicle-treated mice."

We added following statement at line 321-328 and graphic data as Figure 5—figure supplement 4. “HTR2C-expressing MCF7 cells expressed not only ZEB1 but also Twist1 and Snail. […] These results indicate that HTR2C-mediated signaling induced EMT through up-regulation of ZEB1 and blocking HTR2C with Pizotifen induced mesenchymal to epithelial transition through downregulation of ZEB1 (Figure 5—figure supplement 4).”

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Strain, strain background (Zebrafish)	AB line	National University of Singapore
Strain, strain background (Zebrafish)	Tg (kdrl:eGFP) zebrafish	Provided by Dr Stainier
Strain, strain background (Mus musculus)	BALB/c	Charles River Laboratories
Cell line (Homo sapiens)	MDA-MB-231	ATCC	HTB-26
Cell line (Homo sapiens)	MCF7	ATCC	HTB-22
Cell line (Homo sapiens)	MDA-MB-435	ATCC	HTB-129
Cell line (Homo sapiens)	MIA-PaCa2	ATCC	CRM-CRL-1420
Cell line (Homo sapiens)	PC3	ATCC	CRL-3471
Cell line (Homo sapiens)	SW620	ATCC	CCL-227
Cell line (Homo sapiens)	PC9	RIKEN BRC	RCB0446
Cell line (Homo sapiens)	HaCaT	CLI	300493
Cell line (BALB/c Mus)	4T1-12B	Provided from Dr Gary Sahagian
Antibody	PRMT1 (A33)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_2449	WB (1:1000)
Antibody	CYP11A1 (D8F4F)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_14217	WB (1:1000)
Antibody	E-cadherin (4A2)(Mouse monoclonal)	Cell Signaling Technology	Cat#_14472	WB (1:1000)IF (1:100)
Antibody	EpCAM (VU1D9)(Mouse monoclonal)	Cell Signaling Technology	Cat#_2929	WB (1:1000)IF (1:100)
Antibody	Vimentin (D21H3)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_5741	WB (1:1000)IF (1:100)
Antibody	N-cadherin (D4R1H)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_13116	WB (1:1000)IF (1:100)
Antibody	ZEB1 (D80D3)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_3396	WB (1:1000)
Antibody	Histone H3 (D1H2)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_4499	WB (1:1000)
Antibody	β-Tubulin (9F3)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_2128	WB (1:1000)
Antibody	GAPDH (14C10)(Rabbit polyclonal)	Cell Signaling Technology	Cat#_2118	WB (1:1000)
Antibody	HTR2C (ab133570)(Rabbit polyclonal)	Abcam	Cat#_ab133570	WB (1:1000)
Antibody	DRD2 (ab85367)(Rabbit polyclonal)	Abcam	Cat#_ab85367	WB (1:1000)
Antibody	Phospho-GSK3β (Ser9) (F-2)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-373800	WB (1:1000)
Antibody	GSK3β (1F7)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-53931	WB (1:1000)
Antibody	KRT18 (DC-10)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-6259	WB (1:1000)
Antibody	KRT19 (A53-B/A2)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-6278	WB (1:1000)
Antibody	MMP1 (3B6)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-21731	WB (1:1000)
Antibody	MMP2 (8B4)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-13595	WB (1:1000)
Antibody	S100A4 (A-7)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-377059	WB (1:1000)
Antibody	Luciferase (C-12)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-74548	WB (1:1000)
Antibody	ki67 (ki-67)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-23900	WB (1:1000)
Antibody	β-Catenin (E-5)(Mouse monoclonal)	Santa Cruz Biotechnology	Cat#_sc-7963	WB (1:1000)IF (1:100)
Antibody	FITC-conjugated E-cadherin antibody (67A4)	Biolegend	Cat#_324104	FACS (1:100)
Antibody	Anti-mouse anti-rabbit immunoglobulin G (IgG) antibodies conjugated to Alexa Fluor 488	Life Technologies	A-11029	IF (1:100)
Antibody	Anti-goat anti-rabbit immunoglobulin G (IgG) antibodies conjugated to Alexa Fluor 488	Life Technologies	A-11034	IF (1:100)
Recombinant DNA reagent	pLVX-shRNA1	Clontech	Cat#_ 632,177
Recombinant DNA reagent	pCDH-CMV-MCS-EF1α-Hygro	System Biosciences	Cat#_CD515B-1	Gene expression vector
Recombinant DNA reagent	pMDLg/pRRE	Addgene	Addgene Plasmid #12251 RRID:Addgene_12251	Lentivirus packaging vector
Recombinant DNA reagent	pRSV-rev	Addgene	Addgene Plasmid #12253 RRID:Addgene_12253	Lentivirus packaging vector
Recombinant DNA reagent	pMD2.G	Addgene	Addgene Plasmid #12259 RRID:Addgene_12259	Lentivirus packaging vector
Recombinant DNA reagent	Providing pCMV-h5TH2C-VSV	Provided from Dr Herrick
Chemical compound, drug	FDA-approved chemical libraries	Prestwick Chemical
Chemical compound, drug	Pizotifen	Santa Cruz Biotechnology	Cat#_sc-201143
Chemical compound, drug	S(-)Eticlopride hydrochloride	Santa Cruz Biotechnology	Cat#_E101
Software, algorithm	GraphPad Prism7	GraphPad Software Inc	RRID:SCR_002798	Data analysis
Software, algorithm	FlowJo	BD Biosciences	RRID:SCR_008520	FACS data analysis